Volume 453 Number 7191



California against cronyism p.1

The scale of funds set to be spent by the state on stem-cell research necessitates strong governance.

doi: 10.1038/453001a


A research menu p.1

More spending on agricultural science is needed to help resolve the world's food crisis.

doi: 10.1038/453001b


A place for everything p.2

More researchers must record the latitude and longitude of their data.

doi: 10.1038/453002a




doi: 10.1038/453006a



doi: 10.1038/453008a



doi: 10.1038/453011a



doi: 10.1038/453012b



doi: 10.1038/453013a

News Features




doi: 10.1038/453018a




doi: 10.1038/453022a

News & Views


Neuroscience: Hidden female talent p.41


doi: 10.1038/453041a


Electronics: The fourth element p.42


doi: 10.1038/453042a


Climate change: Natural ups and downs p.43


doi: 10.1038/453043a


Quantum physics: The squeeze goes on p.45


doi: 10.1038/453045a


Immunology: T cells hang in the balance p.46


doi: 10.1038/453046a


Networks: Teasing out the missing links p.47


doi: 10.1038/453047a


Astrophysics: Rays from the dark p.48


doi: 10.1038/453048a

追悼:John Wheeler氏(1911-2008)

Obituary: John Wheeler (1911–2008) p.50


doi: 10.1038/453050a



Rapid leukocyte migration by integrin-independent flowing and squeezing p.51

All metazoan cells carry transmembrane receptors of the integrin family, which couple the contractile force of the actomyosin cytoskeleton to the extracellular environment. In agreement with this principle, rapidly migrating leukocytes use integrin-mediated adhesion when moving over two-dimensional surfaces. As migration on two-dimensional substrates naturally overemphasizes the role of adhesion, the contribution of integrins during three-dimensional movement of leukocytes within tissues has remained controversial. We studied the interplay between adhesive, contractile and protrusive forces during interstitial leukocyte chemotaxis in vivo and in vitro. We ablated all integrin heterodimers from murine leukocytes, and show here that functional integrins do not contribute to migration in three-dimensional environments. Instead, these cells migrate by the sole force of actin-network expansion, which promotes protrusive flowing of the leading edge. Myosin II-dependent contraction is only required on passage through narrow gaps, where a squeezing contraction of the trailing edge propels the rigid nucleus.

doi: 10.1038/nature06887


Mapping and sequencing of structural variation from eight human genomes p.56

Genetic variation among individual humans occurs on many different scales, ranging from gross alterations in the human karyotype to single nucleotide changes. Here we explore variation on an intermediate scale—particularly insertions, deletions and inversions affecting from a few thousand to a few million base pairs. We employed a clone-based method to interrogate this intermediate structural variation in eight individuals of diverse geographic ancestry. Our analysis provides a comprehensive overview of the normal pattern of structural variation present in these genomes, refining the location of 1,695 structural variants. We find that 50% were seen in more than one individual and that nearly half lay outside regions of the genome previously described as structurally variant. We discover 525 new insertion sequences that are not present in the human reference genome and show that many of these are variable in copy number between individuals. Complete sequencing of 261 structural variants reveals considerable locus complexity and provides insights into the different mutational processes that have shaped the human genome. These data provide the first high-resolution sequence map of human structural variation—a standard for genotyping platforms and a prelude to future individual genome sequencing projects.

doi: 10.1038/nature06862


Control of Treg and TH17 cell differentiation by the aryl hydrocarbon receptor p.65

Regulatory T cells (Treg) expressing the transcription factor Foxp3 control the autoreactive components of the immune system. The development of Treg cells is reciprocally related to that of pro-inflammatory T cells producing interleukin-17 (TH17). Although Treg cell dysfunction and/or TH17 cell dysregulation are thought to contribute to the development of autoimmune disorders, little is known about the physiological pathways that control the generation of these cell lineages. Here we report the identification of the ligand-activated transcription factor aryl hydrocarbon receptor (AHR) as a regulator of Treg and TH17 cell differentiation in mice. AHR activation by its ligand 2,3,7,8-tetrachlorodibenzo-p-dioxin induced functional Treg cells that suppressed experimental autoimmune encephalomyelitis. On the other hand, AHR activation by 6-formylindolo[3,2-b]carbazole interfered with Treg cell development, boosted TH17 cell differentiation and increased the severity of experimental autoimmune encephalomyelitis in mice. Thus, AHR regulates both Treg and TH17 cell differentiation in a ligand-specific fashion, constituting a unique target for therapeutic immunomodulation.

doi: 10.1038/nature06880



The sculpting of Jupiter's gossamer rings by its shadow p.72

version="1.0" encoding="utf-8"?> ]>

Nature 453 7191 20080501 72754 0028-0836 1476-4687 2008Nature Publishing Group 10.1038/nature06886 The sculpting of Jupiter’s gossamer rings by its shadow Douglas P.HamiltonD P HaraldKrügerH Astronomy Department, University of Maryland, College Park, Maryland 20742-2421, USA Max-Planck-Institut für Sonnensystemforschung, Max-Planck-Strasse 2, 37191 Katlenburg-Lindau, Germany Max-Planck-Institut für Kernphysik, Postfach 103980, 69029 Heidelberg, Germany Correspondence and requests for materials should be addressed to D.P.H. (dphamil@umd.edu). &e080501-11; &nature06886-s1;

Dust near Jupiter is produced when interplanetary impactors collide energetically with small inner moons, and is organized into a main ring, an inner halo, and two fainter and more distant gossamer rings. Most of these structures are constrained by the orbits of the moons Adrastea, Metis, Amalthea and Thebe, but a faint outward protrusion called the Thebe extension behaves differently and has eluded understanding. Here we report on dust impacts detected during the Galileo spacecraft’s traversal of the outer ring region: we find a gap in the rings interior to Thebe’s orbit, grains on highly inclined paths, and a strong excess of submicrometre-sized dust just inside Amalthea’s orbit. We present detailed modelling that shows that the passage of ring particles through Jupiter’s shadow creates the Thebe extension and fully accounts for these Galileo results. Dust grains alternately charge and discharge when traversing shadow boundaries, allowing the planet’s powerful magnetic field to excite orbital eccentricities and, when conditions are right, inclinations as well.

Jupiter’s dusty rings were discovered by the Voyager spacecraft in 1979 and have subsequently been investigated with Earth- and space-based telescopes as well as in situ spacecraft. These observations have revealed a broad but very tenuous distribution of dust straddling the orbits of four small inner moons. For the most part, these moonlets bound the radial and vertical extent of the various ring structures, as summarized in Fig. 1a. There is, however, one major discrepancy: the Thebe extension, which has a vertical profile that matches Thebe’s orbital excursions, but lies fully outside its source moonlet. Some dynamical process not considered by the standard ring models is clearly at work, but why does it single out the Thebe extension?

A unique source of data on the outer jovian rings was provided by Galileo, which passed through the tenuous Thebe and Amalthea rings in 2002 and 2003 (Fig. 1b). The spacecraft’s impact-ionization dust detector recorded ∼100 impacts of micrometre- and submicrometre-sized dust grains, providing in situ information on particle sizes, speeds, impact directions and impact rates. Although the initial analysis of these data appeared in refs 4 and 11, it has not yet been reported in a scientific journal. The dust detector found a sharp transition between the Amalthea and Thebe rings in agreement with imaging, but other results were less expected. For example, some dust grains follow orbits tilted by ∼20° from Jupiter’s equator, which is surprising because particles in the visible rings have orbits tilted by only ∼1°. The instrument also measured a reduction of particle flux interior to Thebe’s orbit that was unanticipated because of the seemingly smooth increase of brightness of the optical image (red region of Fig. 1a; note that the vertical blue stripe is probably an artifact); see also ref. 5. Finally, the smallest detected dust grains, with radii of 0.2–0.3 &mgr;m, are far less prevalent near Thebe than near Amalthea, a difference that highlights the importance of electromagnetic forces. These three results, detailed in Fig. 2a–c, nicely complement images of the Thebe extension. The combined constraints allow us to demonstrate convincingly that electromagnetic processes involving Jupiter’s shadow govern the planet’s gossamer rings.

The orbits of micrometre-sized ring particles are dominated by Jupiter’s gravity, but are also subject to electromagnetic forces and, to a far lesser extent, radiation pressure from the Sun. Dust grains acquire electric charges in Jupiter’s vicinity by sweeping up local plasma and by photoemission in which sunlight incident on a dust grain liberates electrons. Typically the charge is driven rapidly to an equilibrium value that depends on the dust-grain speed, the plasma composition and temperature, and the strength of solar radiation. These effects balance at one value in Jupiter’s shadow and at another, larger, value in full sunlight.

Without planetary shadowing, the effects of electromagnetic forces on the sizes (semimajor axes a) and shapes (eccentricities e) of ring particle orbits average to zero over satellite orbital periods (∼12 h) (ref. 14). With shadowing, however, significant changes to semimajor axes and eccentricities can accumulate in phase over the ∼6 month timescale during which the long axes of elliptical orbits slowly rotate (or precess) in space. This effect is known as a shadow resonance, and it produces a ring that is asymmetric and offset away from the Sun (Fig. 3c), similar to rings driven by radiation pressure.

As the shadow resonance is a fully electromagnetic effect and magnetic fields do no work, energy measured in the reference frame that co-rotates with Jupiter’s field is conserved. This constraint leads to a constant of the motion (the dimensionless Jacobi constant):where i is a ring particle’s orbital inclination and Rsyn ≈ 2.25RJ (here RJ is Jupiter’s radius) is the radius of the synchronous orbit where a ring particle circles Jupiter at the same rate that the planet rotates.

Particles blasted off satellites start on nearly circular orbits around Jupiter, because satellite orbital speeds (∼25 km s-1) completely dominate the small ejection speeds (∼50 m s-1). In the Thebe ring, a > Rsyn and dynamical evolution leads to e ≫ i, as the dominant v × B forces are primarily radial (here v is the velocity of the dust grain relative to Jupiter’s magnetic field, B). Thus the equation shows that the semimajor axis and eccentricity increase in tandem, causing dust to spread more easily outward from Thebe than inward, as indicated by Fig. 3b. In fact, dust launched on a circular orbit from Thebe at a/Rsyn = 1.38 has CJ = 3.074; by simultaneously taking a → ∞ and e = 1, we find that the smallest possible pericentre distance is q = a(1 - e) = 2.66RJ, about 0.4RJ inside Thebe’s orbit at 3.1RJ, while the most distant orbit point Q = a(1 + e) can grow without bound. Thus, perturbations from an arbitrary time-varying dust grain charge can bring the dust grain at most 0.4RJ inward from Thebe or about halfway to synchronous orbit. This hard limit on the inner extent of the Thebe dust ring provides a natural explanation for the existence of the gap in the dust number density measurements visible at ∼2.7RJ in Fig. 2a.

We ran a number of dynamical simulations including gravitational, electromagnetic and radiation forces to test our analytic prediction of a gap in the ring particle number density. A critical parameter in these simulations is the plasma density, which has not been measured. Accordingly, we ran the model for a number of different plasma densities (ne) and display three of them (ne = 0.7, 0.5 and 2.0 cm-3) in Fig. 4. All of the numerical simulations support the existence of an inner boundary, despite the fact that radiation pressure prevents the energy integral from being strictly obeyed. Over much longer timescales, diffusion induced by radiation pressure or dissipative Poynting-Robertson drag can cause material to leak inward from Thebe. Alternatively, ring material detected inside 2.7RJ may simply have originated from Amalthea, as suggested by Fig. 4c.

We now compare this suite of models to the other observations discussed above. The visual appearance of the Thebe extension in Fig. 1a requires grains with radii larger than ∼2 &mgr;m (ref. 5), significant radial extensions, and low orbital inclinations. The latitude plots of Fig. 4 indicate that particles of the right size do maintain low inclinations for all three plasma densities, but for the most tenuous ne = 0.2 cm-3 plasma no micrometre-sized material should reach the outer edge of the Thebe extension at 3.75RJ. Requiring that the shadow resonance be strong enough to push material out to 3.75RJ therefore puts a lower limit on the plasma number density, which in our simple model is ne > 0.5 cm-3. Only the models shown in Fig. 4c–f are viable.

There are several reasons why the Amalthea ring does not show a similar extension. First, the electromagnetic force (and hence the shadow resonance) is significantly weaker at Amalthea than at Thebe (note the less perturbed Amalthea ring in Fig. 3c), and in fact vanishes just ∼0.3RJ inward at synchronous orbit where a ring particle’s velocity relative to the magnetic field is zero. In addition, dust grains on crossing orbits survive against recollision with satellites for only ∼100 yr near Amalthea, but 25 times longer near Thebe owing to that satellite’s smaller size, larger orbital inclination, and greater distance from Jupiter. Finally, an Amalthea extension less than a few per cent of the brightness of the Amalthea ring would be masked by the background debris from Thebe.

Galileo’s dust detector observed a significant population of particles with orbits tilted by ≳20°. The models indicate that for certain plasma conditions, particle orbits do in fact attain large inclinations (Fig. 4b, d, f). The mechanism forcing the larger grains to high latitudes is a vertical shadow resonance, in which the line of nodes (the intersection of the orbit plane with Jupiter’s equatorial plane) rotates extremely slowly, allowing inclinations to be pumped systematically. Similar to the standard shadow resonance, this requires a near cancellation between gravitational and electromagnetic perturbations to the nodal precession rate, which occurs for dust grains with a specific charge-to-mass ratio. Outside synchronous orbit, positive orbit-averaged charges and therefore low plasma densities are needed. The vertical shadow resonance forces micrometre and submicrometre grains onto orbits inclined by up to 20°, nicely matching the observed dust trajectories (Fig. 2b). Positive orbit-averaged charges indicate a plasma density ne < 2.0 cm-3 in our simple model. Only the models shown in Fig. 4c, d satisfy both this and the previous constraint.

Finally, Galileo observed a strong excess of 0.2–0.3 &mgr;m dust grains at Amalthea (Fig. 2c). Figure 4 shows that, for all modelled plasma densities, electromagnetic forces cause positively-charged particles smaller than ∼0.3 &mgr;m to escape entirely from the ring region. For our favoured plasma density of ne = 0.5 cm-3 (Fig. 4c, d), the distance plot shows that a narrow size range from 0.2 to 0.3 &mgr;m is retained in the Amalthea ring but escapes from the Thebe ring. This occurs because the shadow resonance, like all electromagnetic effects on ring particles, is weaker near the synchronous orbit just interior to Amalthea. Thus Thebe’s extra loss mechanism explains the differences in the size distributions observed by Galileo.

We find that the shadow resonance at Jupiter can simultaneously account for (1) the outward extension of the Thebe ring at constant inclination, (2) the gap interior to Thebe, (3) the large number of high-inclination particles, and (4) the excess of 0.2–0.3 &mgr;m grains at Amalthea in a natural model with only a single tunable parameter: the plasma density, which we constrain to 0.5 cm-3 < ne < 2.0 cm-3. Changing other parameters of the plasma model will have similar effects to adjusting ne, as ring particle dynamics is almost entirely determined by the details of the charging history (Fig. 3a). Shadow resonances are active to some extent in all diffuse dusty rings, and the vertical shadow resonance is a very promising mechanism for driving micrometre-sized ring particles to large latitudes. The origin of particles detected by the Voyager spacecraft at latitudes up to 30° near Saturn remains an unsolved problem; we suggest that the vertical shadow resonance is a plausible solution.

Overview of Jupiter’s diffuse ring system.

a, Jupiter’s gossamer rings as seen by Galileo from a location near the planet’s equatorial plane (false colour: adapted from ref. 1). Galileo observations first showed that the ring system has at least four distinct components: (1) the main ring (in white) projecting inward from the two moonlets Metis and Adrastea, (2) the vertically extended halo (white) with thickness 0.1RJ (RJ = 71,492 km) interior to the main ring, (3) a gossamer ring associated with the satellite Amalthea (small yellow rectangle), and (4) an even fainter ring associated with Thebe (large red rectangle). Crosses mark the extremes of the radial and vertical motions of the two moons caused by their eccentric and inclined orbits. The thicknesses of the two outer rings exactly match the vertical excursions of the two source satellites, and ring material appears to extend primarily inward from all four moonlets. The Thebe extension is the faint material (in blue) located outside Thebe’s orbit. b, Galileo’s trajectories during the ring passages on 5 November 2002 (solid line) and 21 September 2003 (dashed line). Dust data were acquired from most of the inbound legs (thick lines), but a spacecraft anomaly prevented measurements inside 2.33RJ and during the outbound leg. We colour-code the ring components, and indicate the approximate locations of the moons’ orbits by vertical dashed lines.

Galileo dust data.

a, Impact rate of submicrometre-sized dust particles onto the Galileo dust detector during the 5 November 2002 ring passage. The rate increased by a factor of 1,000 as the spacecraft moved inward from 3.75RJ (right to left). The main rise at Amalthea’s orbit (inner edge of the plot) is also seen in all other instrument channels, and corresponds to the transition from the Thebe to the Amalthea ring (Fig. 1). The next-most-significant feature is the dip in the rate by a factor of ∼3 from 3.0RJ to 2.7RJ. Although the error bars (±1s.d.) are large, we believe that the feature is real, as it also appears in similar data for larger particles and in Galileo’s second ring passage. Furthermore, the dip is also consistent with a detailed analysis of the Galileo images. As indicated by the error bars, much of the fine structure at significantly lower impact rates is probably due to small number statistics. b, Distribution of dust grain trajectories. Larger symbols indicate larger particles. The rotation angle indicates the phase of the spinning spacecraft; the detector points closest to ecliptic north (south) at rotation angle 0° (180°). An analysis of the 140° field of view of the instrument combined with the 60° angle between the sensor and spin axes and the spacecraft trajectory and spin axis pointing shows that particles on circular uninclined orbits could strike the detector when the spacecraft had a rotation angle between 30° and 150°. We indicate this detection limit with a pair of solid lines, and another for orbits with 20° tilts with dashed lines. A significant number of both large and small impacts occur outside the solid lines, indicating that some ring particles have large inclinations of ∼20°. c, Radii of detected dust grains in the Thebe ring (2.54–3.10 RJ; dotted) and the Amalthea ring (2.33–2.54 RJ; solid). Error bars, ±1s.d. The distributions are similar, except for a very strong excess of 0.2 &mgr;m grains in the Amalthea ring.

Variations of the electric potential during shadow passages.

a, Electric potential for 3.2 &mgr;m dust grains from Amalthea (dotted lines) and Thebe (solid lines) as a function of time. We show charge histories for three different assumed plasma densities ne in a simple two-component plasma model as dust grains travel from sunlight (flat curves) into Jupiter’s shadow (dips to lower potentials) and out again. Dust in a plasma tends to charge negatively, and higher plasma densities lead to lower grain potentials, both in sunlight and in shadow. Sunlight raises equilibrium potentials because the photoelectric effect ejects electrons from a dust grain’s surface. In reality, plasma properties probably vary with distance both from Jupiter and the equatorial plane, but we find that these non-uniform distributions lead to changes in orbital behaviour only over timescales much longer than those considered here, validating our simplifying assumption (compare ref. 19). b, The orbital evolution of a 3.2 &mgr;m dust grain launched from Thebe into an ne = 2.0 cm-3 plasma; we plot the grain’s pericentre distance q = a(1 - e), semimajor axis a, and apocentre distance Q = a(1 + e). Orbital evolution is due to the shadow resonance discussed in the text; note that dust spreads more easily outward from Thebe (at 3.1RJ) than inward. c, A numerical simulation of the asymmetric Amalthea and Thebe rings formed by 3.2 &mgr;m dust grains moving in an ne = 2.0 cm-3 plasma and crossing through the planet’s shadow (stippled region).

Radial and vertical mobility of dust.

Here we plot numerically determined minimum and maximum distances (left plots) and latitudes (right plots) reached by different-sized ring particles within two years of launch from Thebe (solid curves) and Amalthea (dotted curves). Three different charging histories parameterized by ne and 600 separate numerical simulations are presented here. The mobility of dust grains is enhanced by two factors: (1) small particle sizes which have larger electromagnetic perturbations, and (2) slow pericentre precession rates which allow the shadow resonance to pump eccentricities to extremely high values. The total precession rate is the sum of a constant gravitational term and an electromagnetic contribution that is dependent on particle size and electric charge. The two terms have the same sign for low plasma densities and are comparable in strength for particle radii of ∼2 &mgr;m (near the peaks in the radial plots). For larger particles, electromagnetic perturbations weaken and radial mobility decreases. Smaller particles precess so rapidly (owing to electromagnetic perturbations) that changes to eccentricity can add together in phase for only short periods of time and again mobility decreases. a, b, For ne = 0.2 cm-3, dust grains larger than ∼0.5 &mgr;m cannot reach outward to the distance observed in Fig. 1 (3.75RJ) nor can they attain latitudes in excess of 10°. c, d, With a plasma density of 0.5 cm-3, micrometre-sized dust from Thebe can extend out to ∼4RJ and attain latitudes of nearly 20°, as required by the observations. e, f, For ne = 2.0 cm-3, micrometre-sized dust grains cannot reach large latitudes and dust grains smaller than 2–3 &mgr;m escape from the ring entirely. An extremely slow pericentre precession rate, caused by nearly equal and opposite gravitational and electromagnetic terms, allows the shadow resonance to facilitate escape.

We thank the Galileo project for a successful mission. This research was supported by grants from NASA Exobiology (D.P.H.) and DLR (H.K.).

Author Contributions H.K. did all of the data analysis and created Figs 1 and 2. D.P.H. did the analytic work and numerical modelling, including Figs 3 and 4.

Burns, J. A. et al. The formation of Jupiter’s faint rings. Science 284, 11461150 (1999) Burns, J. A. et al. in Jupiter: Planet, Satellites & Magnetosphere (eds Bagenal, F., McKinnon, B. & Dowling, T.) 241262 (Cambridge Univ. Press, Cambridge, UK, 2003) Ockert-Bell, M. E. et al. The structure of Jupiter’s ring system as revealed by the Galileo imaging experiment. Icarus 138, 188213 (1999) Krüger, H. Jupiter’s Dust Disc, an Astrophysical Laboratory. Thesis, Ruprecht-Karls-Universität (Shaker, Aachen, 2003) Showalter, M. R., de Pater, I., Verbanac, G., Hamilton, D. P. & Burns, J. A. Properties and dynamics of Jupiter’s gossamer rings from Galileo, Voyager, Hubble and Keck images. Icarus (in the press) Horányi, M. & Burns, J. A. Charged dust dynamics — Orbital resonance due to planetary shadows. J. Geophys. Res. 96, 1928319289 (1991) de Pater, I. et al. Keck infrared observations of Jupiter's ring system near Earth’s 1997 ring plane crossing. Icarus 138, 214223 (1999) Throop, H. B. et al. The jovian rings: New results derived from Cassini, Galileo, Voyager and Earth-based observations. Icarus 172, 5977 (2004) Showalter, M. R. et al. A new look at Jupiter’s ring system: Preliminary results from New Horizons. AAS/Div. Dyn. Astron. Meet. 38, no. 12.07 (2007) Grün, E. et al. The Galileo dust detector. Space Sci. Rev. 60, 317340 (1992) Moissl, R. Galileo’s Staubmessungen in Jupiters Gossamer-Ringen. Thesis, Ruprecht-Karls-Universität. (2005) Hamilton, D. P. & Burns, J. A. Ejection of dust from Jupiter’s gossamer ring. Nature 364, 695699 (1993) Schaffer, L. & Burns, J. A. Stochastic charging of dust grains in planetary rings: Diffusion rates and their effects on Lorentz resonances. J. Geophys. Res. 100, 213234 (1995) Hamilton, D. P. Motion of dust in a planetary magnetosphere: Orbit-averaged equations for oblateness, electromagnetic, and radiation forces with application to Saturn’s E ring. Icarus 101, 244264 (1993) Hamilton, D. P. The asymmetric time-variable rings of Mars. Icarus 119, 153172 (1996) Horányi, M., Morfill, G. E. & Grün, E. Mechanism for the acceleration and ejection of dust grains from Jupiter’s magnetosphere. Nature 363, 144146 (1993) Gurnett, D. A., Grün, E., Gallagher, D., Kurth, W. S. & Scarf, F. L. Micron-sized particles detected near Saturn by the Voyager plasma wave instrument. Icarus 53, 236254 (1983) Srama, R. et al. In situ dust measurements in the inner Saturnian system. Planet. Space Sci. 54, 967987 (2006) Horányi, M. & Cravens, T. E. The structure and dynamics of Jupiter’s ring. Nature 381, 293298 (1996)

doi: 10.1038/nature06886


Phase-slip-induced dissipation in an atomic Bose-Hubbard system p.76

version="1.0" encoding="utf-8"?> ]>

Nature 453 7191 20080501 76794 0028-0836 1476-4687 2008Nature Publishing Group 10.1038/nature06920 Phase-slip-induced dissipation in an atomic Bose–Hubbard system D.McKayD M.WhiteM M.PasienskiM B.DeMarcoB Department of Physics, University of Illinois at Urbana-Champaign, 1110 West Green Street, Urbana, Illinois 61801, USA Correspondence and requests for materials should be addressed to B.D. (bdemarco@uiuc.edu). &nature06920-s1;

Phase-slips control dissipation in many bosonic systems, determining the critical velocity of superfluid helium and the generation of resistance in thin superconducting wires. Technological interest has been largely motivated by applications involving nanoscale superconducting circuit elements, such as standards based on quantum phase-slip junctions. Although phase slips caused by thermal fluctuations at high temperatures are well understood, controversy remains over the role of phase slips in small-scale superconductors—in solids, problems such as uncontrolled noise sources and disorder complicate their study and application. Here we show that phase slips can lead to dissipation in a clean and well-characterized Bose–Hubbard system, by experimentally studying the transport of ultracold atoms trapped in an optical lattice. In contrast to previous work, we explore a low-velocity regime described by the three-dimensional Bose–Hubbard model that is unaffected by instabilities, and we measure the effect of temperature on the dissipation strength. The damping rate of atomic motion (the analogue of electrical resistance in a solid) in the confining parabolic potential is well fitted by a model that includes finite damping at zero temperature. The low-temperature behaviour is consistent with the theory of quantum tunnelling of phase slips, whereas at higher temperatures a crossover consistent with a transition to thermal activation of phase slips is evident. Motion-induced features reminiscent of vortices and vortex rings associated with phase slips are also observed in time-of-flight imaging. These results clarify the role of phase slips in superfluid systems. They may also be of relevance in understanding the source of metallic phases observed in thin films, or serve as a test bed for theories of bosonic dissipation based upon variants of the Bose–Hubbard model.

Although it is widely believed to be the simplest model that captures the relevant features of boson physics in a variety of physical systems, the Bose–Hubbard model is not integrable, and a full characterization of its features is therefore a challenging theoretical problem. The Bose–Hubbard model is described by the hamiltonianwhere bi operates to remove an atom from site i, the notation indicates pairs of sites that are nearest neighbours, ni = bibi is the number of bosons on site i, and εi is the energy cost for a boson to occupy site i. Particles in the Bose–Hubbard model move by tunnelling with energy J between adjacent lattice sites and interact pairwise on the same site with energy U. In our experiment we can directly test this model because ultracold atoms trapped in an optical lattice are a realization of the Bose–Hubbard model (for sufficient lattice potential depth). Although material and electronic parameters are not easy to configure independently in solid systems, we are able to quantitatively probe the Bose–Hubbard model by precisely determining and controlling parameters such as J and U.

We study mass transport, which is the equivalent of charge transport for neutral bosons. Charge transport is studied in solids by using an electric field to apply a uniform force to the charge carriers. In contrast, we use the spatially inhomogeneous restoring force from a parabolic confining potential to excite damped harmonic motion of the centre of mass (COM) of the gas of atoms. The confining potential is included in equation (1) through the site energies , where ri is the distance from site i to the centre of the harmonic potential and k is the spring constant of the restoring force. We measure the COM motion damping rate for small COM velocities. In this regime, which is analogous to the linear, or ohmic, regime for conductivity experiments on solids, the damping rate is independent of velocity. In analogy to experiments on solids, we measure the response of the damping rate to changes in the temperature of the gas of atoms. We also examine how the damping rate depends on the ratio J/U, which is controlled by tuning the lattice potential depth. This dependence cannot easily be studied in solids.

Previous work on the transport of Bose–Einstein condensates (BECs) in optical lattices has focused on regimes not described by the Bose–Hubbard model, on low-dimensional systems, on relatively high-velocity transport and on probing the Landau and dynamic instabilities. In contrast to much of that research, we work at low velocity and do not study phenomena associated with either mentioned instability. The maximum COM velocities in our data are controlled to be smaller than the critical velocity for the Landau and dynamic instabilities, including the effect of strong interactions. We do not observe any phenomena characteristic of these instabilities, such as a significant change in the condensate fraction, strong nonlinear damping, or excitations of the condensate similar to those observed in ref. 17. Measurements of temperature-dependent dissipation in solids have proven to be a powerful tool in understanding phenomena such as phase slips and other sources of resistance; our work is the first systematic investigation of the effect of temperature on transport in an optical lattice.

The experimental sequence is shown in Fig. 1. A 87Rb BEC is created in a Ioffe–Pritchard magnetic trap using standard cooling and trapping techniques (see ref. 22 for a description of our BEC apparatus). The BEC is adiabatically transferred into a cubic optical lattice with d = 405.9 nm between sites; three atoms or fewer are confined on each lattice site for the data used in this paper. The strength of the lattice potential is characterized by a dimensionless parameter s defined by the lattice potential depth sER along each lattice direction ( is the recoil energy and m is the atomic mass). After the BEC is transferred into the lattice, COM motion is generated by applying a rapid impulse to the BEC along the vertical direction z. The COM velocity is measured using time-of-flight imaging after the motion has been allowed to freely evolve for up to 200 ms.

The COM motion we observe is described well by damped harmonic motion; Fig. 2 shows representative data. We fit the time evolution of the BEC COM velocity assuming the equation of motion for the COM coordinate z in the impulse direction, where is the effective mass and dots denote differentiation with respect to time. The damping rate &ggr;, the oscillation frequency and the initial velocity (which ranges from ∼0.8–1.8 mm s-1 for the data used in this paper) are left as free parameters in the fit. Our model of the BEC COM motion assumes that the COM is acted on by the restoring force -kz, from the harmonic confinement, and a dissipative force . The damping rate &ggr;, which is the exponential decay rate for the amplitude of the motion, is the equivalent of electrical resistance in a solid. This can be understood by considering that ohmic resistance in a material—regardless of the source of dissipation—leads to a force on the charge carriers proportional to -&rgr;v, where &rgr; is the resistivity and v is the charge carrier velocity.

The temperature dependence of the damping rate &ggr; is shown in Fig. 3. We measure the ratio of temperature T to the temperature Tc critical for condensation in the magnetic trap by determining the fraction of atoms in the BEC before transfer into the lattice. In Fig. 3 we show the values of &ggr; measured for s = 2 and s = 6, which correspond to lattice potential depths that sample only the superfluid region of the Bose–Hubbard phase diagram at zero temperature. For s = 2, equation (1) requires tunnelling energy corrections beyond tight binding, which must be considered in making a detailed comparison between theory and the data shown in Fig. 3b.

The damping rate data are plotted against inverse temperature because we fit the data to a model, , of thermally activated damping that permits a finite zero-temperature damping rate &ggr;0 (&ggr; is the damping rate at infinite temperature). The data show excellent agreement with this model, excluding the highest temperature data point for each value of s. The fitted &ggr;0 values, namely 1.4 ± 0.4 Hz for s = 2 and 13 ± 2 Hz for s = 6, are inconsistent with zero at a confidence level greater than 95%. Consistency with a model that includes finite dissipation at zero temperature is, by definition, the equivalent of metallic behaviour in a solid—metals, superconductors and insulators are defined as materials that possess finite, zero and infinite resistance at zero temperature, respectively. We have verified that the levelling off of &ggr; at low temperature is not caused by saturation of T/Tc in the lattice, by measuring the fraction of atoms in the BEC after release from the lattice. This fraction does not change significantly after transfer into the lattice for the range of s and T/Tc values in Fig. 3.

The data shown in Fig. 3 are qualitatively consistent with a predicted crossover between quantum tunnelling and thermal activation of phase slips. Phase-slip models of dissipation were first used to describe the intrinsic critical velocity of superfluid 4He (ref. 1) and the appearance of resistance in a superconducting wire. In the context of the work reported here, phase slips permit the BEC COM velocity to relax through generation of topological phase structures such as vortices and vortex rings. The COM motion of the BEC is metastable because there is an activation barrier to a phase slip event occurring and driving the system towards the zero-velocity ground state. Phase slips occur when the system tunnels through the activation barrier or when thermal fluctuations produce the required activation energy. The dissipation rate should therefore change from exhibiting thermally activated behaviour (that is, behaviour exponentially dependent on the inverse temperature) to being temperature independent at a characteristic temperature for any system in which phase slips are the dominant dissipation mechanism. The data in Fig. 3 show excellent agreement with this crossover behaviour, except at the highest temperature.

Other dissipation mechanisms may contribute to the damping rate at high temperature. For example, mutual friction between the BEC and the thermal component may play a role at high temperature and for large values of s, where we observe relative motion between the BEC and the thermal component (see Fig. 2b). Mutual friction, first observed in second-sound experiments in superfluid 4He (ref. 23) and later in experiments with atomic BECs, arises from interactions between atoms in the BEC and the thermal component. Because we observe the thermal component typically to be immobile at high temperature and for large values of s (similar behaviour was observed in ref. 26), mutual friction will tend to bring the BEC to rest. Mutual friction, or any dissipation mechanism induced by the thermal component, cannot explain the measured value of &ggr;0, because the thermal component vanishes at zero temperature.

In the quantum tunnelling regime, the phase-slip rate is predicted to be proportional to e-S, where S is an action characterizing the process of quantum fluctuations driving the system to lower velocity. A generic scaling law for the action, , was derived for the Bose–Hubbard model in the appendix to ref. 21. In Fig. 4a we plot the measured damping rate for different values of (and thus S, up to the constant of proportionality) at two temperatures in the temperature-independent damping regime. The data fit to a line on a log–linear scale, in agreement with the relation predicted for quantum tunnelling of phase slips; the data show excellent agreement with the predicted scaling law. The systematic increase in the damping rate for the higher temperature data in Fig. 4a is consistent with a residual rate of thermally activated phase slips.

Figure 4b, c shows features consistent with a vortex ring and a vortex directly observed in BECs undergoing COM motion. The cores of vortex rings and vortices are visible in time-of-flight images if the free expansion time is long enough for the expanded core size to be optically resolved and if the core aligns properly with the imaging axis. Figure 4b, c show typical time-of-flight images in which vortex rings or vortices are observed and are recorded after 50 ms of free expansion following 2 ms of COM motion for s = 8. The features observed in these images that are consistent with a vortex ring viewed edgeways and a single vortex observed along its core are marked in red and were not detectable at the shorter expansion time used for the data in Fig. 3 and Fig. 4a. Vortex rings and single vortices nucleated by phase slips must lie in a plane perpendicular to the direction of COM motion; single vortex lines may be oriented along any direction in this plane. Vortex rings will therefore be detected edgeways by our imaging system, but single vortices will only rarely align with the imaging axis and be clearly resolved. We observe these features with approximately 20% probability for s = 8, consistent with a random generation process such as phase slips; we do not observe vortex features if COM motion is not excited. We find that vortices and vortex rings are most likely to be detected close to the edge of the BEC, where the energy gap to nucleation is smallest.

In conclusion, we observe temperature-dependent damping of COM motion in an optical lattice that is consistent with dissipation caused by phase slips. The parabolic potential used to confine the atoms gives rise to an inhomogeneous density distribution, which may enhance the effect of phase slips because the activation barrier is suppressed at the edge of the BEC. This system may therefore be comparable to thin superconducting wires and strips, in which vortices entering and leaving at the boundaries strongly influence current flow. The effects of inhomogeneous density and finite size on phase-slip dynamics in this system remain to be conclusively theoretically addressed. The technique used in this paper can be extended to probe transport properties in models that are relevant to solid materials, such as two-dimensional and disordered Bose–Hubbard models. Direct imaging of vortex rings and vortices nucleated by phase slips may be used to address the microscopic dissipation dynamics in this system.

Methods Summary

The optical lattice is created using three pairs of orthogonally polarized laser beams with wavelengths of 812 nm. These beams are weakly focused to a 120-&mgr;m waist, and slightly frequency-offset to eliminate residual cross-dimensional interference resulting from imperfect polarization. COM motion is excited by changing the strength of a confining magnetic potential for 5 ms. The signal-to-noise ratio in measurements of the COM velocity is improved by reducing the lattice potential to zero over an interval of 200 &mgr;s, thereby eliminating atoms effectively diffracting from the lattice potential.

The temperature is determined by measuring the fraction of atoms in the BEC. For our lowest temperature data (T/Tc = 0.6 and T/Tc = 0.3), accurately detecting the atoms in the thermal component requires images taken at high and low optical depths.

The oscillation data used to measure &ggr; are checked in two ways for nonlinear response. First, no significant change in the fitted &ggr; value is measured if the first period of motion is excluded from the fit. Second, fitting the data to a nonlinear damping model with , where &agr; is the coefficient of nonlinear damping and &Ggr; is the damping rate for &agr; = 0, typically increases the reduced &khgr;2 value of the fit by 10–20%. We find no clear dependence of &agr; on T/Tc or J/U, and the fitted value of &agr; averaged across all of the data is 0.68 ± 0.07. Ultimately, our sensitivity to weak velocity dependence is limited because of finite signal-to-noise ratio in measurements of the COM velocity.

Optical lattice apparatus

87Rb BECs are created in the state in a magnetic trap with a 21.7-Hz radial frequency and 14.5-Hz axial frequency. The optical lattice potentialis formed from three retro-reflected laser beams with orthogonal wavevectors and polarizations along the directions , and . The beam directions make 60° and 45° angles with the vertical axis z. Two of the optical lattice beams are respectively frequency-offset by +19.66 and -176.94 MHz from the third to prevent residual cross-dimensional interference. The magnetic trap is left on during and after transfer into the lattice, to support the atoms against gravity. The impulse used to excite COM motion is created by briefly changing the equilibrium position of the BEC in the magnetic trap. The harmonic potential is not spherically symmetric as was specified for simplicity in equation (1). The percentage difference between the weakest and the geometric mean of the oscillator frequencies along the lattice directions is 8% for s = 2, decreasing to 2% for s = 6. For s = 2, equation (1) also requires corrections for next-nearest-neighbour tunnelling.

COM velocity measurement

Standard absorption imaging is used to measure the COM position of the atoms, which is translated into COM velocity, after release from the optical lattice and the harmonic potential. Before turning off the magnetic trap, the lattice is linearly ramped down over a 200-&mgr;s interval to improve the image signal-to-noise ratio. We checked that this ‘band-mapping’ step does not affect the measured COM velocity, by comparing our data with some for which the lattice potential was turned off over 350 ns. Images are fitted to a Thomas–Fermi profile if only a BEC is visible, or the combination of a Thomas–Fermi and a gaussian profile if atoms in a thermal component are present. Although the COM of each component is left as a free parameter in the fit, we do not analyse motion of the thermal component in this paper.

Temperature measurement

We determine temperature for T/Tc > 0.6 by measuring the fraction of atoms in the BEC using a single absorption image taken after the BEC’s release from the magnetic trap. The atoms are optically pumped (or ‘repumped’) to the state before imaging. The thermal fraction at our lowest temperatures is not visible for images in which the maximum optical depth is small enough to accurately measure the number of atoms in the BEC. Therefore, the condensate fractions for T/Tc = 0.312 and T/Tc = 0.6 are determined by taking two images. One image is of partially repumped atoms, and is used to determine the total optical depth of the BEC. The second image is of fully repumped atoms, and is used to determine the total optical depth of the thermal component by fitting the low-optical-depth region of the image. We compare the optical depths determined from the two images by calibrating the fractional change in optical depth for partial repumping. The calibration is performed using a thermal gas and taking images at two free-expansion times.

We measure T/Tc using the fraction of atoms in the BEC because there are no proven techniques for determining T in an optical lattice. Interactions may change T/Tc for BECs loaded into the lattice for the highest value of s in Fig. 4a (refs 32–34); the extent to which this effect plays a role in experiments has not been resolved. The temperature of the gas was controlled by altering the evaporative cooling procedure, resulting, for fixed s, in a varying number of atoms in the BEC and a varying Tc value. The average Tc value for the data used in this paper is 0.13 &mgr;K, determined from the total number of atoms and the magnetic trap oscillator frequency; Tc spans a 0.07-&mgr;K range for the data in Fig. 3. The heating caused by the dissipation observed in our experiment cannot be detected within our experimental uncertainty in condensate fraction or absolute temperature.

Technical noise

We rule out several technical noise sources as dissipation mechanisms that could explain the damping of COM motion. Anharmonicity in the dipole potential may effectively damp COM motion for large values of s. To check for anharmonic behaviour, we measure COM motion when the retro-reflected lattice laser beams are removed, which eliminates the lattice potential and reduces the depth of the dipole potential by a factor of ∼2. Using this technique we measure a value of &ggr; consistent with zero for lattice laser intensities corresponding to s = 9 (or s = 18 if the retro-reflected beams are present), eliminating trap anharmonicity as an effective dissipation source for the data in Fig. 3. Relative motion between the lattice and the harmonic potential or fluctuations in s (caused by retro-reflecting mirror motion and lattice laser intensity fluctuations, respectively) can lead to the dephasing of dipole mode motion by transferring atoms into states with different values of in excited bands; we do not, however, observe population outside of the lowest energy band. The total spontaneous emission rate per atom is less than 0.3 Hz for s = 6, so momentum diffusion caused by scattering light from the optical lattice laser beams is insignificant. The lattice depth varies by less than 3% across the BEC, so spatial variation in the effective mass can play no role in the dissipation timescales measured in our data.

Experimental apparatus and experimental sequence.

Over a 100-ms interval a 87Rb BEC (blue sphere) is transferred from a magnetic trap (copper coloured) into an optical lattice by superimposing three pairs of counterpropagating lattice beams (red). A 7-ms impulse is applied to excite COM motion in the overall potential V experienced by the atoms, which is the combination of the lattice potential and a parabolic confining potential provided by the magnetic and optical dipole trap. After allowing up to 200 ms for the motion of the BEC to evolve, the lattice and magnetic trap are turned off and the BEC is released. The COM velocity of the BEC is measured after 24 ms of free expansion by taking an image. The inset is not to scale—the BEC diameter ranges from 30–40 &mgr;m for the data used in this paper (corresponding to 75–100 lattice sites) and the maximum amplitude of COM motion in the harmonic potential is 10 &mgr;m.

Representative data used to measure the damping rate &ggr;.

Each data point represents the result of one experimental cycle, in which the COM velocities of the BEC (filled circles) and the thermal component (open circles) are measured. a, Data for s = 2, T/Tc = 0.85 and 8 × 105 BEC atoms. b, Data for s = 6, T/Tc = 0.93 and 2.7 × 105 BEC atoms. In a, the error bar is the standard deviation in the COM velocity if no impulse is applied, and the solid line is a fit to the data and is used to determine the damping rate &ggr;. False-colour images that are used to measure velocity are shown as insets in b, and are labelled by the evolution time at which they were taken; the dashed lines bisecting the 0.47-mm field of view are there to guide the eye. Regions of high density are coloured red, and low-density regions are coloured blue. In each image, the narrow component is the BEC and the broad component is the thermal component.

Temperature dependence of the damping rate &ggr;.

a, s = 6; b, s = 2. Each data point represents the value of &ggr; determined from BEC COM motion at fixed temperature and fixed s. The red data points are excluded from the fit (solid line) to a model of thermally activated damping. The fitted energy gap &Dgr;E is kB × (0.5 ± 0.1) &mgr;K for s = 2 and kB × (0.33 ± 0.08) &mgr;K for s = 6 (where kB is Boltzmann’s constant). The error bars in &ggr; represent the uncertainty in the fit to COM velocity data. The standard deviations in the data used to determine T/Tc for each data point are used for the error bars in (T/Tc)-1.

Scaling of &ggr; with .

a, The tight-binding limit is used to calculate from s. The filled circles are data for T/Tc = 0.76 and 3 × 105 BEC atoms and the open circles are data for T/Tc = 0.312 and 1.1 × 106 BEC atoms. The error bars in &ggr; represent the uncertainty in the fit to the COM velocity; there is an uncertainty of less than 3% in . The slope of -1.6 ± 0.1 determined by linear fit (solid and dashed lines) on the log–linear scale is consistent between the two temperatures. Time-of-flight images exhibiting features consistent with a vortex ring and a single vortex are shown in false colour in b and c, respectively; other excitations present in c are not clearly resolved. The images are fitted to a Thomas–Fermi profile, and the residual is shown in greyscale. The edge of the BEC determined from the fit is shown as a white ellipse in each inset.

We thank R. Barankov, E. Demler, P. Goldbart, N. Goldenfeld, D. Pekker and P. Phillips for discussions, and D. S. Jin, N. Mason and J. V. Porto for critically reading this manuscript. This work was supported by the National Science Foundation, the Office of Naval Research and the UIUC Research Board. D.M. acknowledges support from the Carver Foundation and NSERC.

Langer, J. S. & Fisher, M. E. Intrinsic critical velocity of a superfluid. Phys. Rev. Lett. 19, 560563 (1967) Langer, J. S. & Ambergaokar, V. Instrinsic resistive transition in narrow superconducting channels. Phys. Rev. 164, 498510 (1967) Mooij, J. E. & Nazarov, Y. V. Superconducting nanowires as quantum phase-slip junctions. Nature Phys. 2, 169172 (2006) Tinkham, M. Introduction to Superconductivity 2nd edn, Ch. 8 (McGraw Hill, New York, 1996) Bezryadin, A., Lau, C. N. & Tinkham, M. Quantum suppression of superconductivity in ultrathin nanowires. Nature 404, 971974 (2000) Bollinger, A. T., Rogachev, A. & Bezryadin, A. Dichotomy in short superconducting nanowires: Thermal phase slippage vs. Coulomb blockade. Europhys. Lett. 76, 505511 (2006) Goldman, A. M. Superconductor-insulator transitions in the two-dimensional limit. Physica E 18, 16 (2003) Phillips, P. & Dalidovich, D. The elusive Bose metal. Science 302, 243247 (2003) Fisher, M. P., Weichman, P. B., Grinstein, G. & Fisher, D. S. Boson localization and the superfluid-insulator transition. Phys. Rev. B 40, 546570 (1989) Greiner, M. et al. Quantum phase transition from a superfluid to a Mott insulator in a gas of ultracold atoms. Nature 415, 3944 (2002) Jaksch, D. et al. Cold bosonic atoms in optical lattices. Phys. Rev. Lett. 81, 31083111 (1998) Cataliotti, F. S. et al. Josephson junction arrays with Bose-Einstein condensates. Science 293, 843846 (2001) Henderson, K. et al. Experimental study of the role of atomic interactions on quantum transport. Phys. Rev. Lett. 96, 150401 (2006) Fertig, C. D. et al. Strongly inhibited transport of a degenerate 1D Bose gas in a lattice. Phys. Rev. Lett. 94, 120403 (2005) Cristiani, M. et al. Experimental properties of Bose-Einstein condensates in one-dimensional optical lattices: Bloch oscillations, Landau-Zener tunneling, and mean-field effects. Phys. Rev. A 65, 063612 (2002) Mun, J. et al. Phase diagram for a Bose-Einstein condensate moving in an optical lattice. Phys. Rev. Lett. 99, 150604 (2007) DeSarlo, L. et al. Unstable regimes for a Bose-Einstein condensate in an optical lattice. Phys. Rev. A 72, 013603 (2005) Fallani, L. et al. Observation of dynamical instability for a Bose-Einstein condensate in a moving 1D optical lattice. Phys. Rev. Lett. 93, 140406 (2004) Cristiani, M. et al. Instabilities of a Bose-Einstein condensate in a periodic potential: an experimental investigation. Opt. Express 12, 410 (2004) Wu, B. & Niu, Q. Superfluidity of Bose–Einstein condensate in an optical lattice: Landau–Zener tunnelling and dynamical instability. N. J. Phys. 5, 104.1104.24 (2003) Polkovnikov, A. et al. Decay of a superfluid currents in a moving system of strongly interacting bosons. Phys. Rev. A 71, 063613 (2005) White, M., Gao, H., Pasienski, M. & DeMarco, B. Bose-Einstein condensates in rf-dressed adiabatic potentials. Phys. Rev. A 74, 023616 (2006) Vinen, W. F. Mutual friction in a heat current in liquid helium II. III. Theory of the mutual friction. Proc. R. Soc. Lond. A 242, 493515 (1957) Stamper-Kurn, D. M. et al. Collisionless and hydrodynamic excitations of a Bose-Einstein condensate. Phys. Rev. Lett. 81, 500503 (1998) Haljan, P. C., Coddington, I., Engels, P. & Cornell, E. A. Driving Bose-Einstein-condensate vorticity with a rotating normal cloud. Phys. Rev. Lett. 87, 210403 (2001) Ferlaino, F. et al. Dynamics of a Bose-Einstein condensate at finite temperature in an atom-optical coherence filter. Phys. Rev. A 66, 011604 (2002) McCumber, D. E. & Halperin, B. I. Time scale of intrinsic resistive fluctuations in thin superconducting wires. Phys. Rev. B 1, 10541070 (1970) Matthews, M. R. et al. Vortices in a Bose-Einstein condensate. Phys. Rev. Lett. 83, 24982501 (1999) Anderson, B. P. et al. Watching dark solitons decay into vortex rings in a Bose-Einstein condensate. Phys. Rev. Lett. 86, 29262929 (2001) Plourde, B. L. T. et al. Influence of edge barriers on vortex dynamics in thin weak-pinning superconducting strips. Phys. Rev. B 6401, 014503 (2001) Greiner, M. et al. Exploring phase coherence in a 2D lattice of Bose-Einstein condensates. Phys. Rev. Lett. 87, 160405 (2001) Blakie, P. B. & Porto, J. V. Adiabatic loading of bosons into optical lattices. Phys. Rev. A. 69, 013603 (2004) Ho, T.-L. & Zhou, Q. Intrinsic heating and cooling in adiabatic processes for bosons in optical lattices. Phys. Rev. Lett. 99, 120404 (2007) Rey, A., Pupillo, G. & Porto, J. V. The role of interactions, tunneling, and harmonic confinement on the adiabatic loading of bosons in an optical lattice. Phys. Rev. A. 73, 023608 (2006) Denschlag, J. H. et al. A Bose-Einstein condensate in an optical lattice. J. Phys. B 35, 30953110 (2002)

doi: 10.1038/nature06920


The missing memristor found p.80

version="1.0" encoding="utf-8"?> ]>

Nature 453 7191 20080501 80834 0028-0836 1476-4687 2008Nature Publishing Group 10.1038/nature06932 The missing memristor found Dmitri B.StrukovD B Gregory S.SniderG S Duncan R.StewartD R R. StanleyWilliamsR S HP Labs, 1501 Page Mill Road, Palo Alto, California 94304, USA Correspondence and requests for materials should be addressed to R.S.W. (stan.williams@hp.com). &e080501-6; &nature06932-s1;

Anyone who ever took an electronics laboratory class will be familiar with the fundamental passive circuit elements: the resistor, the capacitor and the inductor. However, in 1971 Leon Chua reasoned from symmetry arguments that there should be a fourth fundamental element, which he called a memristor (short for memory resistor). Although he showed that such an element has many interesting and valuable circuit properties, until now no one has presented either a useful physical model or an example of a memristor. Here we show, using a simple analytical example, that memristance arises naturally in nanoscale systems in which solid-state electronic and ionic transport are coupled under an external bias voltage. These results serve as the foundation for understanding a wide range of hysteretic current–voltage behaviour observed in many nanoscale electronic devices that involve the motion of charged atomic or molecular species, in particular certain titanium dioxide cross-point switches.

More specifically, Chua noted that there are six different mathematical relations connecting pairs of the four fundamental circuit variables: electric current i, voltage v, charge q and magnetic flux ϕ. One of these relations (the charge is the time integral of the current) is determined from the definitions of two of the variables, and another (the flux is the time integral of the electromotive force, or voltage) is determined from Faraday’s law of induction. Thus, there should be four basic circuit elements described by the remaining relations between the variables (Fig. 1). The ‘missing’ element—the memristor, with memristance M—provides a functional relation between charge and flux, dϕ = Mdq.

In the case of linear elements, in which M is a constant, memristance is identical to resistance and, thus, is of no special interest. However, if M is itself a function of q, yielding a nonlinear circuit element, then the situation is more interesting. The i–v characteristic of such a nonlinear relation between q and ϕ for a sinusoidal input is generally a frequency-dependent Lissajous figure, and no combination of nonlinear resistive, capacitive and inductive components can duplicate the circuit properties of a nonlinear memristor (although including active circuit elements such as amplifiers can do so). Because most valuable circuit functions are attributable to nonlinear device characteristics, memristors compatible with integrated circuits could provide new circuit functions such as electronic resistance switching at extremely high two-terminal device densities. However, until now there has not been a material realization of a memristor.

The most basic mathematical definition of a current-controlled memristor for circuit analysis is the differential formwhere w is the state variable of the device and is a generalized resistance that depends upon the internal state of the device. In this case the state variable is just the charge, but no one has been able to propose a physical model that satisfies these simple equations. In 1976 Chua and Kang generalized the memristor concept to a much broader class of nonlinear dynamical systems they called memristive systems, described by the equationswhere w can be a set of state variables and and f can in general be explicit functions of time. Here, for simplicity, we restrict the discussion to current-controlled, time-invariant, one-port devices. Note that, unlike in a memristor, the flux in memristive systems is no longer uniquely defined by the charge. However, equation (3) does serve to distinguish a memristive system from an arbitrary dynamical device; no current flows through the memristive system when the voltage drop across it is zero. Chua and Kang showed that the i–v characteristics of some devices and systems, notably thermistors, Josephson junctions, neon bulbs and even the Hodgkin–Huxley model of the neuron, can be modelled using memristive equations. Nevertheless, there was no direct connection between the mathematics and the physical properties of any practical system, and hence, almost forty years later, the concepts have not been widely adopted.

Here we present a physical model of a two-terminal electrical device that behaves like a perfect memristor for a certain restricted range of the state variable w and as a memristive system for another, wider (but still bounded), range of w. This intuitive model produces rich hysteretic behaviour controlled by the intrinsic nonlinearity of M and the boundary conditions on the state variable w. The results provide a simplified explanation for reports of current–voltage anomalies, including switching and hysteretic conductance, multiple conductance states and apparent negative differential resistance, especially in thin-film, two-terminal nanoscale devices, that have been appearing in the literature for nearly 50 years.

Electrical switching in thin-film devices has recently attracted renewed attention, because such a technology may enable functional scaling of logic and memory circuits well beyond the limits of complementary metal–oxide–semiconductors. The microscopic nature of resistance switching and charge transport in such devices is still under debate, but one proposal is that the hysteresis requires some sort of atomic rearrangement that modulates the electronic current. On the basis of this proposition, we consider a thin semiconductor film of thickness D sandwiched between two metal contacts, as shown in Fig. 2a. The total resistance of the device is determined by two variable resistors connected in series (Fig. 2a), where the resistances are given for the full length D of the device. Specifically, the semiconductor film has a region with a high concentration of dopants (in this example assumed to be positive ions) having low resistance , and the remainder has a low (essentially zero) dopant concentration and much higher resistance .

The application of an external bias v(t) across the device will move the boundary between the two regions by causing the charged dopants to drift. For the simplest case of ohmic electronic conduction and linear ionic drift in a uniform field with average ion mobility &mgr;V, we obtainwhich yields the following formula for w(t):By inserting equation (7) into equation (5) we obtain the memristance of this system, which for simplifies to:The q-dependent term in parentheses on the right-hand side of this equation is the crucial contribution to the memristance, and it becomes larger in absolute value for higher dopant mobilities &mgr;V and smaller semiconductor film thicknesses D. For any material, this term is 1,000,000 times larger in absolute value at the nanometre scale than it is at the micrometre scale, because of the factor of 1/D2, and the memristance is correspondingly more significant. Thus, memristance becomes more important for understanding the electronic characteristics of any device as the critical dimensions shrink to the nanometre scale.

The coupled equations of motion for the charged dopants and the electrons in this system take the normal form for a current-controlled (or charge-controlled) memristor (equations (1) and (2)). The fact that the magnetic field does not play an explicit role in the mechanism of memristance is one possible reason why the phenomenon has been hidden for so long; those interested in memristive devices were searching in the wrong places. The mathematics simply require there to be a nonlinear relationship between the integrals of the current and voltage, which is realized in equations (5) and (6). Another significant issue that was not anticipated by Chua is that the state variable w, which in this case specifies the distribution of dopants in the device, is bounded between zero and D. The state variable is proportional to the charge q that passes through the device until its value approaches D; this is the condition of ‘hard’ switching (large voltage excursions or long times under bias). As long as the system remains in the memristor regime, any symmetrical alternating-current voltage bias results in double-loop i–v hysteresis that collapses to a straight line for high frequencies (Fig. 2b). Multiple continuous states will also be obtained if there is any sort of asymmetry in the applied bias (Fig. 2c).

Obviously, equation (7) is only valid for values of w in the interval [0, D]. Different hard-switching cases are defined by imposing a variety of boundary conditions, such as assuming that once the value of w reaches either of the boundaries, it remains constant until the voltage reverses polarity. In such a case, the device satisfies the normal equations for a current-controlled memristive system (equations (3) and (4)). Figure 3a, b shows two qualitatively different i–v curves that are possible for such a memristive device. In Fig. 3a, the upper boundary is reached while the derivative of the voltage is negative, producing an apparent or ‘dynamical’ negative differential resistance. Unlike a true ‘static’ negative differential resistance, which would be insensitive to time and device history, such a dynamical effect is simply a result of the charge-dependent change in the device resistance, and can be identified by a strong dependence on the frequency of a sinusoidal driving voltage. In another case, for example when the boundary is reached much faster by doubling the magnitude of the applied voltage (Fig. 3b), the switching event is a monotonic function of current. Even though in the hard-switching case there appears to be a clearly defined threshold voltage for switching from the ‘off’ (high resistance) state to the ‘on’ (low resistance) state, the effect is actually dynamical. This means that any positive voltage v+ applied to the device in the off state will eventually switch it to the on state after time . The device will remain in the on state as long as a positive voltage is applied, but even a small negative bias will switch it back to the off state; this is why a current-hysteresis loop is only observed for the positive voltage sweep in Fig. 3a, b.

In nanoscale devices, small voltages can yield enormous electric fields, which in turn can produce significant nonlinearities in ionic transport. Figure 3c illustrates such a case in which the right-hand side of equation (6) is multiplied by a window function w(1 - w)/D2, which corresponds to nonlinear drift when w is close to zero or D. In this case, the switching event requires a significantly larger amount of charge (or even a threshold voltage) in order for w to approach either boundary. Therefore, the switching is essentially binary because the on and off states can be held much longer if the voltage does not exceed a specific threshold. Nonlinearity can also be expected in the electronic transport, which can be due to, for example, tunnelling at the interfaces or high-field electron hopping. In this case, the hysteresis behaviour discussed above remains essentially the same but the i–v characteristic becomes nonlinear.

The model of equations (5) and (6) exhibits many features that have been described as bipolar switching, that is, when voltages of opposite polarity are required for switching a device to the on state and the off state. This type of behaviour has been experimentally observed in various material systems: organic films that contain charged dopants or molecules with mobile charged components; chalcogenides, where switching is attributed to ion migration rather than a phase transition; and metal oxides, notably TiO2 (refs 4, 13, 14, 21) and various perovskites. For example, multi-state and binary switching that are similar to those modelled in Figs 2c and 3c, respectively, have been observed, with some showing dynamical negative differential resistance. Typically, hysteresis such as in Fig. 3c is observed for both voltage polarities, but observations of i–v characteristics resembling Fig. 3a, b have also been reported. In our own studies of TiOx devices, i–v behaviours very similar to those in Figs 2b, 2c and 3c are regularly observed. Figure 3d illustrates an experimental i–v characteristic from a metal/oxide/metal cross-point device within which the critical 5-nm-thick oxide film initially contained one layer of insulating TiO2 and one layer of oxygen-poor TiO2-x (refs 21, 22). In this system, oxygen vacancies act as mobile +2-charged dopants, which drift in the applied electric field, shifting the dividing line between the TiO2 and TiO2-x layers. The switching characteristic observed for a particular memristive system helps classify the nature of the boundary conditions on the state variable of the device.

The rich hysteretic i–v characteristics detected in many thin-film, two-terminal devices can now be understood as memristive behaviour defined by coupled equations of motion: some for (ionized) atomic degrees of freedom that define the internal state of the device, and others for the electronic transport. This behaviour is increasingly relevant as the active region in many electronic devices continues to shrink to a width of only a few nanometres, so even a low applied voltage corresponds to a large electric field that can cause charged species to move. Such dopant or impurity motion through the active region can produce dramatic changes in the device resistance. Including memristors and memristive systems in integrated circuits has the potential to significantly extend circuit functionality as long as the dynamical nature of such devices is understood and properly used. Important applications include ultradense, semi-non-volatile memories and learning networks that require a synapse-like function.

The four fundamental two-terminal circuit elements: resistor, capacitor, inductor and memristor.

Resistors and memristors are subsets of a more general class of dynamical devices, memristive systems. Note that R, C, L and M can be functions of the independent variable in their defining equations, yielding nonlinear elements. For example, a charge-controlled memristor is defined by a single-valued function M(q).

The coupled variable-resistor model for a memristor.

a, Diagram with a simplified equivalent circuit. V, voltmeter; A, ammeter. b, c, The applied voltage (blue) and resulting current (green) as a function of time t for a typical memristor. In b the applied voltage is v0sin(&ohgr;0t) and the resistance ratio is , and in c the applied voltage is ±v0sin2(&ohgr;0t) and , where v0 is the magnitude of the applied voltage and &ohgr;0 is the frequency. The numbers 1–6 label successive waves in the applied voltage and the corresponding loops in the i–v curves. In each plot the axes are dimensionless, with voltage, current, time, flux and charge expressed in units of v0 = 1 V, , t0 ≡ 2&pgr;/&ohgr;0 ≡ D2/&mgr;Vv0 = 10 ms, v0t0 and i0t0, respectively. Here i0 denotes the maximum possible current through the device, and t0 is the shortest time required for linear drift of dopants across the full device length in a uniform field v0/D, for example with D = 10 nm and &mgr;V = 10-10 cm2 s-1 V-1. We note that, for the parameters chosen, the applied bias never forces either of the two resistive regions to collapse; for example, w/D does not approach zero or one (shown with dashed lines in the middle plots in b and c). Also, the dashed i–v plot in b demonstrates the hysteresis collapse observed with a tenfold increase in sweep frequency. The insets in the iv plots in b and c show that for these examples the charge is a single-valued function of the flux, as it must be in a memristor.

Simulations of a voltage-driven memristive device.

a, Simulation with dynamic negative differential resistance; b, simulation with no dynamic negative differential resistance; c, simulation governed by nonlinear ionic drift. In the upper plots of a, b and c we plot the voltage stimulus (blue) and the corresponding change in the normalized state variable w/D (red), versus time. In all cases, hard switching occurs when w/D closely approaches the boundaries at zero and one (dashed), and the qualitatively different i–v hysteresis shapes are due to the specific dependence of w/D on the electric field near the boundaries. d, For comparison, we present an experimental iv plot of a Pt–TiO2-x–Pt device.

This research was conducted with partial support from DARPA and DTO.

Chua, L. O. Memristor - the missing circuit element. IEEE Trans. Circuit Theory 18, 507519 (1971) Hickmott, M. T. Low-frequency negative resistance in thin anodic oxide films. J. Appl. Phys. 33, 26692682 (1962) Dearnaley, G., Stoneham, A. M. & Morgan, D. V. Electrical phenomena in amorphous oxide films. Rep. Prog. Phys. 33, 11291192 (1970) Waser, R. & Aono, M. Nanoionics-based resistive switching memories. Nature Mater. 6, 833840 (2007) Scott, J. C. & Bozano, L. D. Nonvolatile memory elements based on organic materials. Adv. Mater. 19, 14521463 (2007) Collier, C. P. et al. A [2]catenane-based solid state electronically reconfigurable switch. Science 289, 11721175 (2000) Zhitenev, N. B., Sidorenko, A., Tennant, D. M. & Cirelli, R. A. Chemical modification of the electronic conducting states in polymer nanodevices. Nature Nanotechnol. 2, 237242 (2007) Smits, J. H. A., Meskers, S. C. J., Janssen, R. A. J., Marsman, A. W. & de Leeuw, D. M. Electrically rewritable memory cells from poly(3-hexylthiophene) Schottky diodes. Adv. Mater. 17, 11691173 (2005) Lai, Q. X., Zhu, Z. H., Chen, Y., Patil, S. & Wudl, F. Organic nonvolatile memory by dopant-configurable polymer. Appl. Phys. Lett. 88, 133515 (2006) Terabe, K., Hasegawa, T., Nakayama, T. & Aono, M. Quantized conductance atomic switch. Nature 433, 4750 (2005) Kozicki, M. N., Park, M. & Mitkova, M. Nanoscale memory elements based on solid-state electrolytes. IEEE Trans. Nanotechnol. 4, 331338 (2005) Dietrich, S. et al. A nonvolatile 2-Mbit CBRAM memory core featuring advanced read and program control. IEEE J. Solid State Circuits 42, 839845 (2007) Jameson, J. R. et al. Field-programmable rectification in rutile TiO2 crystals. Appl. Phys. Lett. 91, 112101 (2007) Jeong, D. S., Schroeder, H. & Waser, R. Coexistence of bipolar and unipolar resistive switching behaviors in a Pt/TiO2/Pt stack. Electrochem. Solid State Lett. 10, G51G53 (2007) Beck, A., Bednorz, J. G., Gerber, C., Rossel, C. & Widmer, D. Reproducible switching effect in thin oxide films for memory applications. Appl. Phys. Lett. 77, 139141 (2000) Szot, K., Speier, W., Bihlmayer, G. & Waser, R. Switching the electrical resistance of individual dislocations in single-crystalline SrTiO3. Nature Mater. 5, 312320 (2006) Sawa, A., Fujii, T., Kawasaki, M. & Tokura, Y. Interface resistance switching at a few nanometer thick perovskite manganite active layers. Appl. Phys. Lett. 88, 232112 (2006) Hamaguchi, M., Aoyama, K., Asanuma, S., Uesu, Y. & Katsufuji, T. Electric-field-induced resistance switching universally observed in transition-metal-oxide thin films. Appl. Phys. Lett. 88, 142508 (2006) Oligschlaeger, R., Waser, R., Meyer, R., Karthauser, S. & Dittmann, R. Resistive switching and data reliability of epitaxial (Ba,Sr)TiO3 thin films. Appl. Phys. Lett. 88, 042901 (2006) Richter, C. A., Stewart, D. R., Ohlberg, D. A. A. & Williams, R. S. Electrical characterization of Al/AlOx/molecule/Ti/Al devices. Appl. Phys. Mater. Sci. Process. 80, 13551362 (2005) Stewart, D. R. et al. Molecule-independent electrical switching in Pt/organic monolayer/Ti devices. Nano Lett. 4, 133136 (2004) Blackstock, J. J., Stickle, W. F., Donley, C. L., Stewart, D. R. & Williams, R. S. Internal structure of a molecular junction device: chemical reduction of PtO2 by Ti evaporation onto an interceding organic monolayer. J. Phys. Chem. C 111, 1620 (2007) Chua, L. O. & Kang, S. M. Memristive devices and systems. Proc. IEEE 64, 209223 (1976) Kuekes, P. J., Snider, G. S. & Williams, R. S. Crossbar nanocomputers. Sci. Am. 293, 7278 (2005) Strukov, D. B. & Likharev, K. K. Defect-tolerant architectures for nanoelectronic crossbar memories. J. Nanosci. Nanotechnol. 7, 151167 (2007) Blanc, J. & Staebler, D. L. Electrocoloration in SrTiO - vacancy drift and oxidation-reduction of transition metals. Phys. Rev. B 4, 35483557 (1971)

doi: 10.1038/nature06932


Advancing decadal-scale climate prediction in the North Atlantic sector p.84

version="1.0" encoding="utf-8"?> ]>

Nature 453 7191 20080501 84885 0028-0836 1476-4687 2008Nature Publishing Group Supplementary Information

The file contains Supplementary Discussion on global correlation skill and root mean square error maps; surface temperature maps for 1994-2004 and 2005-2015 decades; and Supplementary Figures 1-4 with Legends.

CatherineCassidy When this paper was published on 1 May 2008, the e-mail address given for the corresponding author was incorrect. This was corrected online-only on 8 May 2008. The correct address is nkeenlyside@ifm-geomar.de. 10.1038/nature06921
Advancing decadal-scale climate prediction in the North Atlantic sector N. S.KeenlysideN S M.LatifM J.JungclausJ L.KornbluehL E.RoecknerE Leibniz Institute of Marine Sciences, Düsternbrooker Weg 20, D-24105 Kiel, Germany Max Planck Institute for Meteorology, Bundesstraße 53, 20146 Hamburg, Germany Correspondence and requests for materials should be addressed to N.S.K. (nkeenlyside@ifm-geomar.de.). &e080501-12; &nature06921-s1;

The climate of the North Atlantic region exhibits fluctuations on decadal timescales that have large societal consequences. Prominent examples include hurricane activity in the Atlantic, and surface-temperature and rainfall variations over North America, Europe and northern Africa. Although these multidecadal variations are potentially predictable if the current state of the ocean is known, the lack of subsurface ocean observations that constrain this state has been a limiting factor for realizing the full skill potential of such predictions. Here we apply a simple approach—that uses only sea surface temperature (SST) observations—to partly overcome this difficulty and perform retrospective decadal predictions with a climate model. Skill is improved significantly relative to predictions made with incomplete knowledge of the ocean state, particularly in the North Atlantic and tropical Pacific oceans. Thus these results point towards the possibility of routine decadal climate predictions. Using this method, and by considering both internal natural climate variations and projected future anthropogenic forcing, we make the following forecast: over the next decade, the current Atlantic meridional overturning circulation will weaken to its long-term mean; moreover, North Atlantic SST and European and North American surface temperatures will cool slightly, whereas tropical Pacific SST will remain almost unchanged. Our results suggest that global surface temperature may not increase over the next decade, as natural climate variations in the North Atlantic and tropical Pacific temporarily offset the projected anthropogenic warming.

The North Atlantic is a region with large natural multidecadal variability. Observations suggest that Atlantic multidecadal variability may be oscillatory, with a period of 70–80 years. Palaeo-evidence and coupled general circulation models support the oscillatory nature of Atlantic multidecadal variability, and hence the existence of a low order mode of natural variability. Although major uncertainties exist in the mechanisms of Atlantic multidecadal variability, it is widely accepted that the meridional overturning circulation (MOC) plays an important role in driving multidecadal SST variations, as shown by ocean and coupled model simulations.

Studies with coupled general circulation models assuming perfect ocean initial conditions indicate that accurate initialization of MOC may allow Atlantic multidecadal variability to be predicted a decade or more in advance. However, past MOC fluctuations have been poorly observed and large uncertainties exist among ocean model simulations and ocean analyses. Here, this difficulty is partly overcome with a simple initialization scheme, previously applied in seasonal forecasting, that consists of relaxing SST anomalies of the coupled general circulation models to observations. (These simulations, called SST-restored, are described in Methods.) Additionally, decadal predictions require accurate projections of external radiative forcing (that is, computed from concentrations of greenhouse gases and sulphate aerosols, solar cycle variations, and volcanic activity).

Using this simple initialization technique and a state-of-the-art coupled general circulation model, a set of ten-year-long, three-member ensemble hindcasts/forecasts were performed every five years from 1955 till 2005 (except for 1995, which instead is started in 1994). External radiative forcing in the hindcasts is treated following ref. 9. Three additional integrations with radiative forcing computed following observations (called twentieth century-RF) are used to estimate predictability due to external radiative forcing. (See Methods for experimental details.)

Skilful predictions of surface temperature averaged over years one to ten of the hindcasts are obtained over large parts of the North Atlantic, Europe, North America and northern Africa (Fig. 1a). In these regions, skill is well above that of persistence (Fig. 1b). Persistence, the prediction that assumes no change from the initial conditions, is a commonly used benchmark for skill. The twentieth century-RF simulations show skill over parts of northern Africa, North America and Europe, but no skill over the North Atlantic Ocean, except in the vicinity of the Equator (Fig. 1c). Over the northern North Atlantic, the climate model hindcast is significantly more skilful than the twentieth century-RF simulations (Fig. 1a). It follows that in this region a significant fraction of the skill arises from initialization of the North Atlantic Ocean, as opposed to external radiative forcing. Over western Europe and large parts of North America, initialization also leads to a significant enhancement in skill. Skill enhancements outside the North Atlantic sector are found in the central eastern tropical Pacific (Fig. 1a and below) and the equatorial Indian Ocean (Supplementary Information).

In several regions, initialization causes a significant degradation in skill compared to the twentieth century-RF simulations (Fig. 1c). Over the tropical North Atlantic and central Africa this is probably related to deficiencies, common to many coupled models, in simulating tropical Atlantic climate. Thus, large skill improvements in this region may be expected, as models improve. The SST-restored simulations (Fig. 1d) provide an estimate for the upper limit of skill of our forecast system during the period considered, as they include observed SST anomalies and external radiative forcing. Over all three land regions, the skill of the SST-restored simulations is generally higher and larger in extent than that of the hindcasts and twentieth century-RF simulations. Over North America and western Europe, the regions that show significantly enhanced skill over the twentieth century-RF simulations are larger than those of the hindcasts (Fig. 1a, d), indicating that considerably more skill may be achievable with better initialization techniques and improved models.

Skill in predicting North Atlantic SST on decadal timescales would imply, as described above, skill in initializing the Atlantic MOC. In the SST-restored simulations, the Atlantic MOC weakens from the 1950s till the mid-1960s, strengthens thereafter, peaking between the late-1980s to late-1990s, and subsequently weakens again (Fig. 2a). Although consistent with SST observations, the veracity of these results is hard to assess, as direct MOC observations are insufficient and large uncertainties exist among ocean model simulations and ocean analyses. The small ensemble spread relative to the low frequency MOC fluctuations also supports the utility of the initialization scheme.

In contrast to measurements of the MOC, regular hydrographic observations in the Labrador Sea extend back past the 1950s. As density changes in the Labrador Sea are widely accepted to force MOC variations, these observations provide an alternative method to assess MOC initialization. Simulated multidecadal fluctuations of wintertime Labrador Sea convection, weakest around 1970 and strongest in the early 1990s, broadly agree with observations of Labrador Sea Water thickness, which is closely related to convection (Fig. 2b). Furthermore, simulated MOC variations closely follow Labrador Sea convection by several years (Fig. 2). Thus, these results provide evidence that observed MOC fluctuations are, to a certain degree, skilfully initialized.

On multidecadal timescales, Labrador Sea convection is initialized directly by SST relaxation. The latter includes the history of observed atmospheric forcing, and in particular the North Atlantic Oscillation (NAO), which plays a key role in forcing multidecadal MOC variations. Consistently, observed NAO variations lead simulated MOC changes by several years, but inconsistently, they are not well related to simulated Labrador Sea convection, except perhaps on multidecadal timescales (Fig. 2). Although the simulated NAO index does not correspond well with the observations, it strengthens during the simulated period, and thus probably also contributes to forcing MOC variations (Fig. 2c).

Initialized decadal fluctuations in the Atlantic MOC are predictable a decade in advance, with ensemble-spread small compared to the signal (Fig. 3a); hindcast skill is, however, largely due to capturing the long-term trend. MOC variations in the twentieth century-RF simulations are weak, and not surprisingly, unrelated to those in the initialization simulations. Model studies indicate that multidecadal Atlantic MOC variations force inter-hemispheric dipolar SST anomalies. Observed variations of the latter are predicted a decade in advance by the hindcasts, but not by the twentieth century-RF simulations (Fig. 3b). Hindcast skill here is thus consistent with predicting Atlantic MOC variations. Consistently, North Atlantic (0–60° N) SST is better predicted by the hindcasts (r = 0.77) than the twentieth century-RF simulations (r = 0.66) (not shown). The greater skill of the hindcasts, over the twentieth century-RF simulations, in predicting decadal North Atlantic, western European (Figs 1 and 3c) and North American (Figs 1 and 3d) surface-temperature variations is thus likely to be due to skill in predicting Atlantic MOC fluctuations.

Enhanced skill over the twentieth century-RF simulations in predicting land surface temperature (Fig. 1a, c), especially over North America, may also arise from outside the Atlantic. In particular, tropical Pacific decadal SST variations are more skilfully predicted by the hindcasts, which correctly capture the 1970s climate shift, than by the twentieth century-RF simulations (Fig. 3e; Supplementary Information). Hindcast skill for global mean temperature (Fig. 4) is also high, but slightly less than the twentieth century-RF simulations. The largest difference occurs for the 1994–2004 predictions, with hindcast surface temperature cooler than observed and twentieth century-RF simulations warmer (Fig. 4). These differences stem from the predictions of internal tropical Pacific variability, with extreme warm conditions predicted by the twentieth century-RF simulations and near neutral conditions by the hindcasts; weak warm conditions were observed (Fig. 3e; Supplementary Information).

We now consider two forecasts, started in November 2000 and November 2005. The MOC is predicted to weaken almost to its 1950–2005 mean over the next decade (Fig. 3a), leading to a weakening of Atlantic SST hemispheric difference towards zero (Fig. 3b). North Atlantic (not shown), western European (Fig. 3c) and North American (Fig. 3d) surface temperatures cool towards 1994–2004 levels. In contrast, in the un-initialized (twentieth century-RF) predictions the MOC slightly weakens, the hemispheric SST difference is unchanged, and warming of surface temperatures over the latter three regions continues (Fig. 3a–d). Eastern tropical Pacific SST is forecasted to remain almost unchanged, but 0.3 K cooler than the un-initialized predictions (Fig. 3e). The differences in predicted North Atlantic and tropical Pacific variability lead to a large difference in the global mean temperature prediction: the initialized prediction indicates a slight cooling relative to 1994–2004 levels, while the anthropogenic-forcing-only simulation suggests a near 0.3 K rise (Fig. 4). In the long-term both projections agree with each other, as is found by extending the 2005 prediction till 2030 (Fig. 4). Internal decadal fluctuations were also found to offset anthropogenic global warming in a previous study, but the offset was much less pronounced and associated primarily with changes in the tropical Pacific.

To investigate the sensitivity of the predictions to greenhouse gas forcing only, the two forecasts were repeated assuming that greenhouse gases were stabilized at year 2000 values. The predictions for the MOC and surface temperature remain basically unchanged. Thus, in the near future, natural decadal variability in the Atlantic and Pacific may not only override the regional effects of global warming, but temporarily weaken it. Thus, a joint initial/boundary value problem has to be considered when forecasting North Atlantic sector and global climate variability for the coming decades.

The results presented here are promising in light of existing model biases. Experience in numerical weather prediction and seasonal forecasting has shown that skill can be considerably improved by reducing model systematic error and by more accurate forecast initialization. Thus, useful decadal predictions may be in reach.

Methods Summary Simulations

Results are based on three sets of simulations with the ECHAM5/MPI-OM coupled general circulation model (IPCC version). (1) Three integrations (twentieth century-RF/A1B), from 1860–2015, started from different points of a control simulation. Before 2000, radiative forcing follows observations (greenhouse gas and sulphate aerosol concentrations, solar cycle variations, and major volcanic eruptions), and after 2000, it follows the IPCC A1B scenario. Predictability due to radiative forcing is estimated from these simulations. (2) Three coupled integrations (SST-restored), from 1950–2005, initialized from the twentieth century-RF simulations and with identical radiative forcing, but with relaxation towards SST constructed from the coupled model climatology with observed SST anomalies superimposed. The relaxation constant is strong (0.25 d-1) between 30° S and 30° N, and decreases linearly to zero between 30° and 60° S and between 30° and 60° N; poleward of 60° S and 60° N the model is fully coupled. (3) Nine hindcasts and two forecasts (three-member, ten-years long), initialized from the SST-restored simulations during the period 1955–2005. Radiative forcing is as in the twentieth century-RF simulations, except for solar cycle variations, which are repeated from the previous 11 years, and major volcanic eruptions. The latter that occurred during a hindcast are not included, and the impact of any that occurred before the hindcast is damped away with an e-folding time of one year.

Significance tests

For nine forecasts we assume seven degrees of freedom (d.f.). As only positive correlations indicate skill, the significance level for the forecast is determined using a one-sided t-test. A correlation is determined to be significantly larger or smaller than another if the Fisher-Z transformed values pass a one-sided t-test for differences in means (with 12 d.f.). The Fisher-Z distribution of correlations of nine independent and normally distributed pairs of points is close to normal. Field significance is estimated on a 15° × 15° grid to minimize effects of spatial correlation.

Correlation skill in predicting observed ten-year mean surface temperature anomalies a decade in advance relative to other more traditional approaches.

a, Skill of nine ten-year-long predictions, evenly distributed over the period 1955–2005, made with a climate model initialized using ocean (SST) observations and run with projected changes in radiative forcing. b, As in a but given by persistence. c, As in a, but not initialized using ocean observations and with radiative forcing following observations. d, As in c, but with model SST relaxed to observations between 60° S and 60° N (seen in near perfect correlations over the ocean). Correlations exceeding 0.58 are significant at the 5% level. Regions where initialization results in a significant enhancement or reduction in skill compared to radiative-forcing-only simulations are indicated by a blue cross-hatching in a and c, respectively. Land regions where restoring to observed SST anomalies provides a significant enhancement in skill relative to radiative-forcing-only simulations are indicated by blue cross-hatching in d. Correlations in a and c are field significant at close to the 0% level, while those in b pass the field significance test at the 1% level. Details of significance estimation are given in Methods. Correlations in a, c and d are computed from the ensemble mean of three simulations. SST observations are from HADISST; land surface temperature observations are from CRUTEMP3.

SST restoring forces robust multidecadal fluctuations in the Atlantic Ocean meridional overturning circulation, Labrador Sea convection and the North Atlantic Oscillation.

a, Annual mean meridional overturning circulation (MOC) at 30° N from the SST-restored simulations. Grey shading indicates ensemble spread in all panels. b, Simulated wintertime (December–February) Labrador Sea (60–50° W, 55–65° N) mixed-layer depth and observed annual mean Labrador Sea Water (LSW) thickness. The latter is defined between isopycnals &sgr;1.5 = 34.72–34.62, and is closely related to wintertime Labrador Sea convection. Simulated Labrador Sea convection precedes MOC variations at 30° N, with a maximum correlation (0.71) when the MOC lags by three years. c, Observed and simulated ensemble mean wintertime North Atlantic Oscillation (NAO) indices. Solid lines indicate the seven-year running mean. The NAO index is defined as the sea-level-pressure (SLP) difference between Lisbon and Stykkisholmur. Observed NAO variations precede MOC variations at 30° N, with a maximum correlation (0.52) when the MOC lags by 3–4 years, but are not closely related to simulated Labrador Sea convection (r < 0.4 at any lag).

Hindcast/forecast decadal means of selected time series, compared with observations and climate model projections including only radiative forcing changes but not initialized using ocean observations.

Model projections are twentieth century-RF followed by A1B scenario simulations (‘20C-RF/A1B’). a, Maximum MOC strength at 30° N; b, Atlantic SST dipole index (60–10° W, 40–60° N minus 50–0° W, 40–60° S SST area averages), which is constructed to isolate MOC forced SST fluctuations from radiatively forced variations; c, European land surface temperature (5° W–10° E, 35–60° N average); d, North American surface temperature (120–70° W, 30–50° N average); and e, eastern Tropical Pacific SST (150–90° W, 20° S–20° N average). ae, Each point represents a ten-year centred mean; vertical bars indicate ensemble spread; verification and forecast periods are indicated (dark shading begins 2008, indicating the start of the true forecast period); two additional forecasts with radiative forcing stabilized at year 2000 levels are also shown. be, Correlation of both hindcasts and climate model projections with observations are given in brackets. Owing to insufficient observations, in a the correlation with the MOC from SST-restored simulation is given. Hindcast correlation skill is significantly greater at the 5% level than that of the standard climate projection in all cases (significance test described in Methods). Scale on right is used for the SST-restored simulation and observations; note the different ranges for right and left axes in a and b. SST are from HADISST, land temperature from CRUTEMP3 (‘CRU3’).

Hindcast/forecast decadal variations in global mean temperature, as compared with observations and standard climate model projections.

Model projections are twentieth century-RF followed by A1B scenario simulations (‘20C-RF/A1B’); ‘Stabilization’ forecasts assume greenhouse gas concentrations fixed at year 2000 levels. Each point represents a ten-year centred mean; vertical bars indicate ensemble spread; verification and forecast periods are indicated (dark shading begins 2008, indicating the start of the true forecast period). Three additional decadal means (joined by a dotted line) show the evolution of the initialized and un-initialized 2005 predictions extended till 2030. Correlation of both hindcasts and climate model projections with observations are given in brackets. Correlation of the twentieth century-RF simulation with observations is greater than that of the hindcasts, but only marginally at the 5% significance level. Observed global mean temperature anomalies are from HadCRU3.

Computing resources at the European Centre for Medium Range Weather Forecasting, the Deutsches Klimarechenzentrum and the Norddeutscher Verbund für Hoch- und Höchstleistungsrechnen are acknowledged. Deutscher Wetterdienst provided significant CPU allocations. The work was supported by the European Union ENSEMBLES and German BMBF NORDATLANTIK projects. We thank M. Esch for performing the three twentieth century-RF simulations, and F. Nevoigt for help with art work.

Author Contributions N.S.K. and M.L. designed experiments. N.S.K. performed experiments and analysis. M.L. and J.J. contributed to analysis. L.K., J.J. and E.R. provided the model and support. N.S.K. and M.L. wrote the paper. All authors discussed the results and commented on the manuscript.

Goldenberg, S. B., Landsea, C. W., Mestas-Nuñez, A. M. & Gray, W. M. The recent increase in Atlantic hurricane activity: Causes and implications. Science 293, 474479 (2001) Enfield, D. B., Mestas-Nuñez, A. M. & Trimble, P. J. The Atlantic Multidecadal Oscillation and its relation to rainfall and river flows in the continental U. S. Geophys. Res. Lett. 28, 20772080 (2001) Sutton, R. T. & Hodson, D. L. R. Atlantic Ocean forcing of North American and European summer climate. Science 309, 115118 (2005) Folland, C. K., Palmer, T. N. & Parker, D. E. Sahel rainfall and worldwide sea temperatures, 1901–85. Nature 320, 602607 (1986) Griffies, S. M. & Bryan, K. Predictability of North Atlantic multidecadal climate variability. Science 275, 181184 (1997) Boer, G. A study of atmosphere-ocean predictability on long time scales. Clim. Dyn. 16, 469472 (2000) Collins, M. et al. Interannual to decadal climate predictability in the North Atlantic: A multimodel-ensemble study. J. Clim. 19, 11951203 (2006) Cunningham, S. A. et al. Temporal variability of the Atlantic meridional overturning circulation at 26.5°N. Science 317, 935938 (2007) Smith, D. M. et al. Improved surface temperature prediction for the coming decade from a global climate model. Science 317, 796799 (2007) Solomon, S. et al. Climate Change 2007: The Physical Science Basis (Cambridge Univ. Press, Cambridge, UK, 2007) Bjerknes, J. Atlantic air-sea interaction. Adv. Geophys. 10, 182 (1964) Kushnir, Y. Interdecadal variations in North Atlantic sea surface temperature and associated atmospheric conditions. J. Clim. 7, 141157 (1994) Schlesinger, M. E. & Ramankutty, N. An oscillation in the global climate system of period 65–70 years. Nature 367, 723726 (1994) Knight, J. R., Allan, R. J., Folland, C. K., Vellinga, M. & Mann, M. E. A signature of persistent natural thermohaline circulation cycles in observed climate. Geophys. Res. Lett. 32, 10.1029/2005GL024233 (2005) Mann, M. E., Bradley, R. S. & Hughes, M. K. Global-scale temperature patterns and climate forcing over the past six centuries. Nature 392, 779787 (1998) Delworth, T., Manabe, S. & Stouffer, R. J. Interdecadal variations of the thermohaline circulation in a coupled ocean-atmosphere model. J. Clim. 6, 19932011 (1993) Vellinga, M. & Wu, P. Low-latitude freshwater influence on centennial variability of the Atlantic thermohaline circulation. J. Clim. 17, 44984511 (2004) Jungclaus, J. H., Haak, H., Latif, M. & Mikolajewicz, U. Arctic North Atlantic interactions and multidecadal variability of the meridional overturning circulation. J. Clim. 18, 40134031 (2005) Latif, M., Collins, M., Pohlmann, H. & Keenlyside, N. A review of predictability studies of the Atlantic sector climate on decadal time scales. J. Clim. 19, 59715987 (2006) Eden, C. & Jung, T. North Atlantic interdecadal variability: Oceanic response to the North Atlantic oscillation (1865–1997). J. Clim. 14, 676691 (2001) de Coëtlogon, G. et al. Gulf Stream variability in five oceanic general circulation models. J. Phys. Oceanogr. 36, 21192135 (2006) Keenlyside, N., Latif, M., Botzet, M., Jungclaus, J. & Schulzweida, U. A coupled method for initializing El Niño Southern Oscillation forecasts using sea surface temperature. Tellus A 57, 340356 (2005) Lee, T. C. K., Zwiers, F., Zhang, X. & Tsao, M. Evidence of decadal climate prediction skill resulting from changes in anthropogenic forcing. J. Clim. 19, 53055318 (2006) Jungclaus, J. H. et al. Ocean circulation and tropical variability in the coupled model ECHAM5/MPI-OM. J. Clim. 19, 39523972 (2006) Latif, M. et al. Is the thermohaline circulation changing? J. Clim. 19, 46314637 (2006) Curry, R. G., McCartney, M. S. & Joyce, T. M. Oceanic transport of subpolar climate signals to mid-depth subtropical waters. Nature 391, 575577 (1998) Rayner, N. A. et al. Global analyses of sea surface temperature, sea ice, and night marine air temperature since the late nineteenth century. J. Geophys. Res. 108 4407 10.1029/2002JD002670 (2003) Brohan, P., Kennedy, J. J., Harris, I., Tett, S. F. B. & Jones, P. D. Uncertainty estimates in regional and global observed temperature changes: a new dataset from 1850. J. Geophys. Res. 111 D12106 10.1029/2005JD006548 (2006) Allan, R. & Ansell, T. A new globally-complete monthly historical gridded mean sea level pressure data set (HadSLP2): 1850–2004. J. Clim. 19, 58165842 (2006)

doi: 10.1038/nature06921


Origin of a ‘Southern Hemisphere' geochemical signature in the Arctic upper mantle p.89

version="1.0" encoding="utf-8"?> ]>

Nature 453 7191 20080501 89935 0028-0836 1476-4687 2008Nature Publishing Group Supplementary Information

The file contains Supplementary Tables S1-S3, Supplementary Discussion, Supplementary Figures S1-S4 with legends and additional references.

Origin of a ‘Southern Hemisphere’ geochemical signature in the Arctic upper mantle Steven L.GoldsteinS L GadSofferG Charles H.LangmuirC H Kerstin A.LehnertK A David W.GrahamD W Peter J.MichaelP J Lamont-Doherty Earth Observatory of Columbia University, Department of Earth and Environmental Sciences, Columbia University, 61 Route 9W, Palisades, New York 10964, USA Department of Earth and Planetary Sciences, Harvard University, 20 Oxford Street, Cambridge, Massachusetts 02138, USA College of Ocean and Atmospheric Sciences, Oregon State University, Corvallis, Oregon 97331, USA Department of Geosciences, University of Tulsa, 600 South College Avenue Tulsa, Oklahoma 74104, USA Present address: Harvard Business School, Soldiers Field, Boston, Massachusetts 02163, USA. Correspondence and requests for materials should be addressed to S.L.G. (steveg@ldeo.columbia.edu). &nature06919-s1;

The Gakkel ridge, which extends under the Arctic ice cap for ∼1,800 km, is the slowest spreading ocean ridge on Earth. Its spreading created the Eurasian basin, which is isolated from the rest of the oceanic mantle by North America, Eurasia and the Lomonosov ridge. The Gakkel ridge thus provides unique opportunities to investigate the composition of the sub-Arctic mantle and mantle heterogeneity and melting at the lower limits of seafloor spreading. The first results of the 2001 Arctic Mid-Ocean Ridge Expedition (ref. 1) divided the Gakkel ridge into three tectonic segments, composed of robust western and eastern volcanic zones separated by a ‘sparsely magmatic zone’. On the basis of Sr–Nd–Pb isotope ratios and trace elements in basalts from the spreading axis, we show that the sparsely magmatic zone contains an abrupt mantle compositional boundary. Basalts to the west of the boundary display affinities to the Southern Hemisphere ‘Dupal’ isotopic province, whereas those to the east—closest to the Eurasian continent and where the spreading rate is slowest—display affinities to ‘Northern Hemisphere’ ridges. The western zone is the only known spreading ridge outside the Southern Hemisphere that samples a significant upper-mantle region with Dupal-like characteristics. Although the cause of Dupal mantle has been long debated, we show that the source of this signature beneath the western Gakkel ridge was subcontinental lithospheric mantle that delaminated and became integrated into the convecting Arctic asthenosphere. This occurred as North Atlantic mantle propagated north into the Arctic during the separation of Svalbard and Greenland.

The Gakkel ridge is the global ultraslow endmember among ocean ridges, with spreading rates decreasing from ∼15 mm yr-1 (full rate) at the Lena trough near Greenland to ∼6 mm yr-1 at the Siberian margin. It contains no significant transform offsets, which on other ridges often form mantle flow and mantle compositional boundaries. It thus provides a natural laboratory in which to study source heterogeneity and melting dynamics at ultraslow spreading rates, and where potential complications from tectonic offsets are absent. Here we report isotope and trace element data on basaltic glasses that document the Arctic mantle composition and processes contributing to its enrichment.

Prior to the 2001 Arctic Mid-Ocean Ridge Expedition, data on Gakkel lavas were restricted to three samples of basaltic cobbles and glass shards recovered on two cruises. The Arctic Mid-Ocean Ridge Expedition collected samples from the western ∼850 km of the ridge between longitudes 7° W and 86° E at over 200 stations. Michael et al. divided the studied segments into three ‘magmato-tectonic zones’: the ∼200-km-long, shallow, magmatically robust western volcanic zone (WVZ), characterized by normal ocean-ridge-rift magmatism; the ∼300-km-long, deep, magma-starved sparsely magmatic zone (SMZ) with peridotite outcropping along the ridge axis; and the ∼400-km-long, magmatically robust eastern volcanic zone (EVZ), where axial magmatism is mainly associated with central volcanoes. On the basis of cruise chemical data, Michael et al. inferred compositionally distinct mantle beneath the WVZ and the EVZ, with marked barium enrichments in WVZ lavas. Dick et al. compared the broad-scale tectonic features of the Gakkel and southwest Indian ridges and posited the existence of a new class of ultraslow-spreading ocean ridge, where ‘amagmatic’ segments like the SMZ are a consequence of spreading rates less than ∼12 mm yr-1, and are thus independent of mantle composition.

Our data confirm fundamental differences in the upper-mantle sources of WVZ and EMZ lavas (locations and data tables can be found in the Supplementary Information). The fields formed by Nd–Sr isotopes of EVZ and WVZ lavas show no overlap (Fig. 1a), and samples from the western and eastern parts of the SMZ show affinities to the adjacent EVZ and WVZ. EVZ lavas have higher 143Nd/144Nd values than do WVZ lavas and lower 87Sr/86Sr values than does any mid-ocean-ridge basalt (MORB) north of Iceland.

The contrasting WVZ and EVZ mantle compositions are shown in Fig. 1b in terms of &Dgr;8/4 (ref. 2), which reflects deviations from typical compositions of Pacific and Atlantic MORB on 208Pb/204Pb–206Pb/204Pb diagrams. The disparities also show in along-axis plots of the other radiogenic isotopes (see, for example, Supplementary Fig. 1a). West of the boundary virtually all Gakkel lavas have &Dgr;8/4 > 30; this was shown by Hart to be a primary feature of the Southern Hemisphere ‘Dupal isotope anomaly’ in Indian and some South Atlantic MORB. The Dupal-like character of the WVZ is further illustrated by comparison with global MORB: 208Pb/204Pb–206Pb/204Pb data for WVZ basalts lie in the field of Indian MORB (Fig 2a; Supplementary Fig. 2a). WVZ lavas also show a near-vertical trend in a 87Sr/86Sr–206Pb/204Pb diagram, reflecting a large range of 87Sr/86Sr values, typical of Indian MORB, whereas EVZ lavas show a shallow trend similar to that of Atlantic–Pacific basalts (Fig. 2b). Additional evidence of the WVZ’s Dupal-like character is given in Supplementary Information (Supplementary Fig. 3).

Gakkel WVZ lavas, however, differ from Southern Hemisphere Dupal basalts, in that they have low &Dgr;7/4 values; their data lie near Atlantic–Pacific basalts on this diagram (Supplementary Fig. 2b). The 207Pb/204Pb–206Pb/204Pb relationships are temporal rather than a result of the formation process, because they depend primarily on the timing of major U–Pb fractionation, whereas &Dgr;8/4 reflects the chemical composition (Th/U) and the formation age. Oceanic basalts with enriched isotopic signatures (high Sr and low Nd and Hf isotope ratios, and high &Dgr;8/4 values) are associated with both high and low &Dgr;7/4 values (ref. 5), reflecting old and young formation ages, respectively. The Gakkel data thus indicate that the formation of the Arctic Dupal-like source was more recent than that of the Southern Hemisphere Dupal basalts. Although the formation ages differ, chemical characteristics are similar: like Indian MORB, the high &Dgr;8/4 and 87Sr/86Sr values reflect high time-integrated Th/U and Rb/Sr values for the Pb and Sr in the lavas, indicating the distinctive chemical compositions of reservoirs responsible for Dupal features.

A striking feature of the Gakkel ridge is the abruptness of the isotopic boundary and its location in the middle of the SMZ (Fig. 1b; Supplementary Fig. 1a). The discontinuity is reminiscent of the Australian–Antarctic discordance of the southeast Indian ridge, which demarcates the boundary between the Indian Dupal and Pacific mantle provinces. Common features include great bathymetric depths and low magma production, as reflected by the basaltic crust being thin in comparison with that of adjoining segments. These indicate close relationships between seafloor structure, bathymetry and mantle convective processes. However, the Gakkel SMZ, unlike the Australian–Antarctic discordance, is not associated with any significant ridge offset. Along a continuously spreading ridge, lavas with isotopic affinities to the WVZ or EVZ were collected within 40 km of each other at longitudes 12.7° E and 16.0° E. The Gakkel SMZ data thus indicate a sharp geochemical boundary in the absence of any tectonic boundary. The Knipovich ridge to the southwest of the Gakkel ridge is also isotopically distinct from the WVZ (see, for example, Supplementary Figs 1b, 2a and 3a), but these ridges are separated by a 600 km offset. The marked change in the magmato-tectonic character at the Gakkel SMZ is therefore not simply a tectonic reflection of ultraslow seafloor spreading, as proposed by Dick et al., but reflects a mantle compositional and convective boundary.

Comparison with global MORB (Fig. 2) shows that the Pb–Sr–Nd isotope variations in this small Arctic region mimic the global upper-mantle variability. The Gakkel ridge is far from the Indian Ocean. Although enriched MORBs are found along North Atlantic ridges even far from hotspots, the WVZ is unique among North Atlantic MORB, with Dupal characteristics dominating along ∼350 km of the spreading axis. Their occurrence in Arctic MORB shows that Dupal characteristics are not simply an attribute of the Indian–South Atlantic oceanic mantle, but reflect processes that are more general.

The small geographical scale and well-understood tectonic history of the Eurasian basin affords opportunities to explore how the Dupal signature is created. Quaternary alkaline basalts ranging from hawaiites to nephelinitic basanites in northwestern Spitsbergen have been attributed to low-degree melting of subcontinental lithospheric mantle (SCLM), and form an enriched endmember for WVZ lavas in isotope plots (Figs 1a and 2). This relationship also holds for trace elements that are highly incompatible for melting of both anhydrous and metasomatized mantle mineral assemblages, where abundance ratios in lavas are almost the same as in the mantle source (shown for 87Sr/86Sr versus Th/La and U/La in Fig. 3a and Supplementary Fig. 4a, respectively). The trace elements and isotopes together strongly indicate that the component causing the Dupal-like characteristics is also the source of Spitsbergen alkaline basalts.

Insight into the mineral assemblage of the Spitsbergen magma source can be gleaned from the comparison of elements more compatible in amphibole or phlogopite (such as Rb, Ba and K) with other elements unaffected by amphibole or phlogopite (such as Th, U and La). In plots of 87Sr/86Sr versus Rb/La (Fig. 3b) and Ba/Th (Supplementary Fig. 4b), Spitsbergen lavas deviate from the WVZ trend at low values of Rb/La and Ba/Th , respectively; however, they are not offset in plots involving Th, U and La. Such chemical behaviour indicates the presence of amphibole or phlogopite among the residual minerals. Spitsbergen lavas include olivine, titanaugite and plagioclase, but no amphibole or phlogopite. Because these phases are unstable at asthenospheric temperatures, the data require an SCLM source for Spitsbergen lavas. Subsequent asthenospheric melting beneath the Gakkel ridge would then release the Rb and Ba, leading to the relative enrichments of these elements in the lavas of the WVZ.

The data offer insight into the long-term history of the Gakkel asthenosphere. A Rb–Sr ‘pseudo-isochron’ plot (Fig. 3c) shows that the WVZ mantle least contaminated by SCLM is distinct from the EVZ mantle, because their respective 87Sr/86Sr values differ when Rb/Sr = 0. This WVZ–EVZ offset is further confirmed by comparing isotopes and trace elements (Fig. 3a, b and Supplementary Fig. 4). The ‘least contaminated’ WVZ lavas are similar to Atlantic MORB north of Iceland, in terms of Nd–Sr isotopes and combined isotopes and trace elements (Figs 1 and 3 and Supplementary Fig. 4), whereas the EVZ basalts are distinct from both groups. These observations suggest that WVZ isotopic relationships can be explained by the addition of Spitsbergen SCLM to the North Atlantic asthenosphere, and that the WVZ and EVZ are distinct mantle provinces. The Gakkel SMZ thus bounds distinct upper-mantle regions that have experienced long-term convective isolation.

The introduction of the Svalbard SCLM to the WVZ asthenosphere, the leakage of North Atlantic mantle into the Arctic, and the formation of the boundary can be understood in the context of Arctic tectonic history (Fig. 4). The Eurasian and Atlantic oceans were separated by continent when seafloor spreading began in the Arctic and North Atlantic 55–60 million years (Myr) ago. During the early Oligocene epoch ∼34 Myr ago, Greenland and Norway–Svalbard began to separate, accompanied by northwards propagation of North Atlantic ridges, with linkage of the Arctic and North Atlantic occurring ∼10 Myr ago. Our data suggest that North Atlantic mantle flowed into the Arctic, accompanied by delamination and dispersal of the Svalbard SCLM. Whether this was simply a consequence of rifting or was facilitated by thermal erosion caused by an Arctic hotspot is beyond the scope of this work (a hotspot has been postulated to be responsible for the Yermak plateau (Fig. 4c), but its existence is disputed). In any case, the Gakkel SMZ appears to be the surface manifestation of a physical boundary within the mantle where the linkage occurred.

The origin of the Southern Hemisphere Dupal signature is an ongoing debate in mantle dynamics, and a recent summary highlights inadequacies of postulated models. Explanations include upwelling of a thermal anomaly from the core–mantle boundary; contamination of sub-Indian mantle by the Kerguelen plume; mantle modification by subduction surrounding Gondwana; ancient subduction of altered oceanic crust and sediment; recycled lower-continental or arc crust; and detachment of SCLM during continental break-up. Hybrid scenarios invoke both subduction and lithosphere delamination. It has been suggested that SCLM delamination would impart ‘too low’ an osmium isotope signature to MORB, but this depends on the age of the recycled SCLM. The combined Gakkel and Spitsbergen data and the well-constrained Arctic tectonic history provide relationships where the geochemistry and tectonics are in remarkable accord.

In summary, the Gakkel ridge is an ultraslow-spreading ridge surrounded by continental lithosphere, with clear evidence of regional SCLM contributing a Dupal geochemical signature to the asthenospheric mantle. Unlike in other regions on Earth, we have the strong evidence of Spitsbergen volcanics, with the requisite isotopic compositions to be the Gakkel endmember, and a source mineralogy definitively indicating SCLM melting. The process documented in the Arctic is chemically and geologically feasible and probably plays an important role in generating chemical heterogeneity along the mid-ocean ridge system. Dupal signatures globally may have multiple sources; nevertheless, the migrating continents that have traversed the Indian Ocean basin, where Dupal characteristics occur on a large scale, would generate substantial opportunities for analogous SCLM contributions to the sub-Indian Ocean mantle.

Methods Summary

All analyses were made on hand-picked basaltic glasses. Major elements were measured on a Cameca electron microprobe at the American Museum of Natural History, operated as a joint facility with the Lamont-Doherty Earth Observatory of Columbia University (LDEO). Trace elements were measured using a Plasma Quad 2 inductively coupled plasma mass spectrometer at LDEO or a Thermo Electron X-series inductively coupled plasma mass spectrometer at Harvard. Pb, Nd and Sr isotope ratios were measured using a VG Sector 54 multicollector thermal ionization mass spectrometer at LDEO.

Major elements

Major elements were measured at the American Museum of Natural History on the AMNH-LDEO Cameca electron microprobe. Each analysis represents an average of five individual measurements on a single glass chip. A glass chip of the Lamont glass probe standard, JDF-D2, was repeatedly analysed in each probe session to monitor long-term drift. No long-term drift was evident in any run. The errors are based on the reproducibility of JDF analyses over multiple probe sessions and are ≤1% for SiO2 and MgO.

Trace elements

Hand-picked glass separates with minimal surface alteration were analysed for trace element compositions and measured on the Plasma Quad 2 inductively coupled plasma mass spectrometer at LDEO or a Thermo Electron X-series inductively coupled plasma mass spectrometer at Harvard . Solutions were spiked with Ge, In, Tm and Bi to correct for in-run drift. Concentrations were determined using a calibration curve based on US Geological Survey and in-house standards measured during each run. Errors are 3–5%.

Isotope ratios

Pb, Nd and Sr isotope ratios were measured on separate glass chips from the trace element analyses. These were leached cold for 12 minutes with 8N HNO3. Pb was separated using AG1-X8 anion resin, Sr was separated using Eichrom Sr resin and Nd was separated in a two-column procedure using Eichrom TRU-spec resin to separate the rare-earth elements, followed by &agr;-hydroxy isobutyric acid. Sr, Nd and Pb isotopes were measured on a VG Sector 54 multicollector thermal ionization mass spectrometer at LDEO. 87Sr/86Sr ratios were normalized to 86Sr/88Sr = 0.1194. Sr isotopes were measured in multidynamic mode. Instrumental reproducibility was monitored by repeated measurements of the US National Institute of Standards and Technology SRM 987 standard, which yielded 87Sr/86Sr = 0.710246 ± 0.000016 (2-s.d. external reproducibility, n > 40). All samples were further corrected to a value of 0.71024 for SRM 987. All Nd analyses were made in multidynamic mode as NdO+. 143Nd/144Nd ratios were normalized to 146Nd/144Nd = 0.7129. Repeated measurements of the La Jolla Nd standard yielded 143Nd/144Nd = 0.511838 ± 0.000012 (2-s.d. external reproducibility, n > 30). All samples were further corrected to a value of 0.511860 for the La Jolla standard. Pb isotope analyses used a 207Pb–204Pb double spike. All measurements were performed in static mode. Measurements of unspiked–double-spiked pairs of the NBS 981 standard were replicated to 183, 284 and 300 p.p.m. (2-s.d. external reproducibility, n = 20) for 206Pb/204Pb, 207Pb/204Pb and 208Pb/204Pb ratios, respectively. These measured Pb isotope ratios were corrected to the Todt et al. values of 16.9356, 15.4891, and 36.7006, respectively, for NBS 981.

Distinct western and eastern geochemical provinces in Gakkel ridge basalts.

a, 143Nd/144Nd versus 87Sr/86Sr. Data fields for EVZ and WVZ lavas do not overlap. WVZ lava data points overlap with those for Atlantic MORBs north of Iceland and display a trend towards those for Spitsbergen Quaternary-period alkaline basalts. SMZ lavas show affinities to those in the adjacent WVZ and EVZ. W-SMZ and E-SMZ refer to lavas in the SMZ having isotopic affinities to the WVZ and EVZ, respectively. b, &Dgr;8/4 (representing deviations from the Northern Hemisphere reference line) versus distance from the western end of the Gakkel ridge near Greenland. (The key is the same as that in a.) This shows the distinct WVZ and EVZ compositions and the position and abruptness of the isotopic boundary. Three samples, labelled WVZ*, from near the WVZ–SMZ boundary have unusual chemistry (low SiO2 and incompatible element abundances, and positive Eu and Sr anomalies on trace element diagrams) and follow the WVZ isotope trend (a) but not the WVZ chemical trends (Fig. 3).

Gakkel basalts as a microcosm of global MORBs.

a, 208Pb/204Pb versus 206Pb/204Pb. WVZ and Spitsbergen lavas show the characteristic Dupal MORB offset to high 208Pb/204Pb. Global MORB data are from the Petrological Database of the Ocean Floor (http://www.petdb.org). b, 87Sr/86Sr versus 206Pb/204Pb. (The key is the same as that in a.) EVZ lavas display a trend similar to that of Atlantic-Pacific MORBs, whereas WVZ MORBs display a near-vertical trend similar to Indian and South Atlantic MORBS showing the Dupal anomaly. The WVZ data points display a trend towards those for Spitsbergen lavas.

Gakkel ridge and Spitsbergen basalts.

a, 87Sr/86Sr versus Th/La. (The key is the same as that in b.) WVZ and EVZ lavas form distinct trends. WVZ data points display a trend towards those for Spitsbergen, indicating that the Spitsbergen magma source is the enriched WVZ endmember. b, 87Sr/86Sr versus Rb/La. WVZ and EVZ lavas display distinct trends. Spitsbergen lava data points lie away from the WVZ trend, with low Rb/La values (and Ba/La values, not shown), which is consistent with partial melting in the presence of residual amphibole and/or phlogopite, and indicates a SCLM source. c, 87Sr/86Sr versus Rb/Sr. (The key is the same as that in b.) The Sr isotope ratios are distinct for the WVZ and EVZ even when Rb/Sr = 0. In a and b, for low Th/La and Rb/La values corresponding to depleted mantle composition estimates (labelled ‘upper mantle’), the WVZ and EVZ Sr isotope ratios also are different. In all three panels, ‘depleted’ WVZ compositions are consistent with Atlantic MORBs from north of Iceland, but EVZ compositions are distinct. MORB data are taken from the Petrological Database of the Ocean Floor.

Cenozoic-era evolution of the Arctic upper mantle.

a, Approximately 53 Myr ago, the nascent Eurasian basin, underlain by ‘Arctic’ upper mantle (labelled ‘A’) as sampled today in the EVZ, was separated from the North Atlantic ridges (labelled ‘NA’) by a continent. b, During the Cenozoic, the Eurasian basin continued to widen, and the North Atlantic ridges propagated north, with Greenland and Svalbard separating ∼10 Myr ago. c, Present-day situation based on the Gakkel data, with the boundary between the North Atlantic and Arctic upper-mantle provinces well within the Eurasian basin, in the middle of the SMZ. ‘YP’ labels the Yermak plateau, near the WVZ. Images modified from Blythe and Kleinspehn.

We thank A. le Roex for comments that helped to improve this paper. This work was supported by the US National Science Foundation. G.S. was supported partly by a Paul and Daisy Soros Fellowship for New Americans.

The geochemical data reported here are available in the Petrological Database of the Ocean Floor (http://www.petdb.org).

Michael, P. J. et al. Magmatic and amagmatic seafloor generation at the ultraslow-spreading Gakkel ridge, Arctic Ocean. Nature 423, 956961 (2003) Hart, S. R. A large-scale isotopic anomaly in the Southern Hemisphere mantle. Nature 309, 753757 (1984) Mühe, R., Bohrmann, H., Garbe-Schönberg, D. & Kassens, H. E-MORB glasses from the Gakkel Ridge (Arctic Ocean) at 87°N: evidence for the Earth's most northerly volcanic activity. Earth Planet. Sci. Lett. 152, 19 (1997) Dick, H. J. B., Lin, J. & Schouten, H. An ultraslow-spreading class of ocean ridge. Nature 426, 405412 (2003) Klein, E. M., Langmuir, C. H., Zindler, A., Staudigal, H. & Hamelin, B. Isotope evidence of a mantle convection boundary at the Australian-Antarctic Discordance. Nature 333, 623629 (1988) Dosso, L. et al. The age and distribution of mantle heterogeneity along the Mid-Atlantic Ridge (31–41°N). Earth Planet. Sci. Lett. 170, 269286 (1999) Shirey, S. B., Bender, J. F. & Langmuir, C. H. Three-component isotopic heterogeneity near the Oceanographer transform, Mid-Atlantic Ridge. Nature 325, 217223 (1987) Skjelkvale, B. L., Amundsen, H. E. F., Oreilly, S. Y., Griffin, W. L. & Gjelsvik, T. A primitive alkali basaltic stratovolcano and associated eruptive centers, Northwestern Spitsbergen: Volcanology and tectonic significance. J. Volcanol. Geotherm. Res. 37, 119 (1989) Ionov, D. A., Bodinier, J. L., Mukasa, S. B. & Zanetti, A. Mechanisms and sources of mantle metasomatism: Major and trace element compositions of peridotite xenoliths from Spitsbergen in the context of numerical modeling. J. Petrol. 43, 22192259 (2002) Ionov, D. A., Mukasa, S. B. & Bodinier, J. L. Sr-Nd-Pb isotopic compositions of peridotite xenoliths from Spitsbergen: Numerical modelling indicates Sr-Nd decoupling in the mantle by melt percolation metasomatism. J. Petrol. 43, 22612278 (2002) Clague, D. A. & Frey, F. A. Petrology and trace element chemistry of the Honolulu volcanics, Oahu: Implication for the oceanic mantle below Hawaii. J. Petrol. 23, 447504 (1982) Class, C. & Goldstein, S. L. Plume-lithosphere interactions in the ocean basins: constraints from the source mineralogy. Earth Planet. Sci. Lett. 150, 245260 (1997) Niida, K. & Green, D. H. Stability and chemical composition of pargasitic amphibole in MORB pyrolite under upper mantle conditions. Contrib. Mineral. Petrol. 135, 1840 (1999) Blythe, A. E. & Kleinspehn, K. L. Tectonically versus climatically driven exhumation of the Eurasian plate margin, Svalbard: Fission track analyses. Tectonics 17, 621639 (1998) Ritzmann, O. & Jokat, W. Crustal structure of northwestern Svalbard and the adjacent Yermak Plateau: evidence for Oligocene detachment tectonics and non-volcanic breakup. Geophys. J. Int. 152, 139159 (2003) Zhang, S. Q. et al. Evidence for a widespread Tethyan upper mantle with Indian-Ocean-type isotopic characteristics. J. Petrol. 46, 829858 (2005) Dupré, B. & Allègre, C. J. Pb–Sr isotope variation in Indian Ocean basalts and mixing phenomena. Nature 303, 142146 (1983) Hawkesworth, C. J., Mantovani, M. S. M., Taylor, P. N. & Palacz, Z. Evidence from the Paraña of south Brazil for a continental contribution to Dupal basalts. Nature 322, 356359 (1986) Castillo, P. The Dupal anomaly as a trace of the upwelling lower mantle. Nature 336, 667670 (1988) Arndt, N. T. & Goldstein, S. L. An open boundary between lower continental crust and mantle: its role in crust formation and crustal recycling. Tectonophysics 161, 201212 (1989) le Roex, A. P., Dick, H. J. B. & Fisher, R. L. Petrology and geochemistry of MORB from 25°E to 46°E along the Southwest Indian Ridge: Evidence for contrasting styles of mantle enrichment. J. Petrol. 30, 947986 (1989) Mahoney, J. J. et al. Isotopic and geochemical provinces of the Western Indian Ocean spreading centers. J. Geophys. Res. 94, 40334052 (1989) Barling, J., Goldstein, S. L. & Nicholls, I. A. Geochemistry of Heard Island (southern Indian Ocean): Characterization of an enriched mantle component and implications for the enrichment of the sub-Indian Ocean mantle. J. Petrol. 35, 10171053 (1994) Rehkamper, M. & Hofmann, A. W. Recycled ocean crust and sediment in Indian Ocean MORB. Earth Planet. Sci. Lett. 147, 93106 (1997) Kempton, P. D. et al. Sr-Nd-Pb-Hf isotope results from ODP Leg 187: Evidence for mantle dynamics of the Australian-Antarctic Discordance and origin of the Indian MORB source. Geochem. Geophys. Geosyst. 3 10.1029/2002GC000320 (2002) Hanan, B. B., Blichert-Toft, J., Pyle, D. G. & Christie, D. M. Contrasting origins of the upper mantle revealed by hafnium and lead isotopes from the Southeast Indian Ridge. Nature 432, 9194 (2004) Escrig, S., Capmas, F., Dupré, B. & Allègre, C. J. Osmium isotopic constraints on the nature of the DUPAL anomaly from Indian mid-ocean-ridge basalts. Nature 431, 5963 (2004) Meyzen, C. M. et al. New insights into the origin and distribution of the DUPAL isotope anomaly in the Indian Ocean mantle from MORB of the Southwest Indian Ridge. Geochem. Geophys. Geosyst. 6 10.1029/2005GC000979 (2005) Geldmacher, J., Hoernle, K., Klugel, A., van den Bogaard, P. & Bindemann, I. Geochemistry of a new enriched mantle type locality in the northern hemisphere: Implications for the origin of the EM-I source. Earth Planet. Sci. Lett. 265, 167182 (2008) le Roux, P. J. et al. Mantle heterogeneity beneath the southern Mid-Atlantic Ridge: trace element evidence for contamination of ambient asthenospheric mantle. Earth Planet. Sci. Lett. 203, 479498 (2002) Janney, P. E., le Roex, A. P. & Carlson, R. W. Hafnium isotope and trace element constraints on the nature of mantle heterogeneity beneath the central Southwest Indian Ridge (13°E to 47°E). J. Petrol. 46, 24272464 (2005) Salters, V. J. M. & Stracke, A. Composition of the depleted mantle. Geochem. Geophys. Geosyst. 5 10.1029/2003GC000597 (2004) Workman, R. K. & Hart, S. R. Major and trace element composition of the depleted MORB mantle (DMM). Earth Planet. Sci. Lett. 231, 5372 (2005) Todt, W., Cliff, R. A., Hanser, A. & Hofmann, A. W. in Earth Processes: Reading the Isotopic Code Vol. 95 (eds Basu, A. & Hart, S. R.) 429437 (American Geophysical Union, Washington DC, 1996)

doi: 10.1038/nature06919


Hydatellaceae are water lilies with gymnospermous tendencies p.94

version="1.0" encoding="utf-8"?> ]>

Nature 453 7191 20080501 94970 0028-0836 1476-4687 2008Nature Publishing Group Supplementary Figure

The file contains Supplementary Figure with Legend showing longitudinal section of recently fertilized ovule of Hydatella inconspicua. A pollen tube that entered the ovule through the micropyle formed by the two integuments can be clearly seen. The zygote contains a prominent vacuole and the primary endosperm nucleus is situated at the base of the former female gametophyte. Upper grey box contains digital superposition of pollen tube from adjacent histological section. Lower grey box contains digital superposition of primary endosperm from adjacent histological section. Bar = 10 μm. pen = primary endosperm nucleus; ps = perisperm; pt = pollen tube; z = zygote.

Hydatellaceae are water lilies with gymnospermous tendencies William E.FriedmanW E Department of Ecology and Evolutionary Biology, University of Colorado, Boulder, Colorado 80309, USA Correspondence and requests for materials should be addressed to W.E.F. (e-mail: ned@colorado.edu). &e080501-7; &nature06733-s1;

The flowering plant family Hydatellaceae was recently discovered to be allied to the ancient angiosperm lineage Nymphaeales (water lilies). Because of its critical phylogenetic position, members of the Hydatellaceae have the potential to provide insights into the origin and early diversification of angiosperms. Here I report that Hydatella expresses several rare embryological features that, in combination, are found only in members of the Nymphaeales. At maturity, the female gametophyte is four-celled, four-nucleate and will produce a diploid endosperm, as is characteristic of most early divergent angiosperm lineages. As with all members of the Nymphaeales, endosperm in Hydatella is minimally developed and perisperm is the major embryo-nourishing tissue within the seed. Remarkably, Hydatella exhibits a maternal seed-provisioning strategy that is unique among flowering plants, but common to all gymnosperms: pre-fertilization allocation of nutrients to the embryo-nourishing tissue. This exceptional case of pre-fertilization maternal provisioning of a seed in Hydatella may well be an apomorphic feature of Hydatellaceae alone but, given the newly discovered phylogenetic position of this family, potentially represents a plesiomorphic and transitional condition associated with the origin of flowering plants from gymnospermous ancestors.

For over a century, the flowering plant family Hydatellaceae was thought to belong to the Poales, a highly derived monocot order that includes the grasses. New evidence derived from DNA sequencing and phylogenetic analysis shows that this obscure group of minute aquatic plants is closely related to water lilies (Nymphaeales), which themselves were only recently discovered to be one of the most ancient extant lineages of angiosperms. As a result of these surprising phylogenetic insights, our understanding of the earliest phases of the radiation of angiosperms continues to be very much in flux.

Although the data are, at best, ambiguous, the original embryological studies of Hydatellaceae suggested that the female gametophyte might contain eight nuclei and seven cells, with antipodals that degenerate early (or are absent), and possibly two polar nuclei; as such, Hydatellaceae was thought to produce a triploid genetically biparental endosperm. Recently, however, members of the Nymphaeales and Austrobaileyales (another ancient lineage of flowering plants) have been shown to produce an unusual four-celled, four-nucleate female gametophyte (Nuphar/Schisandra-type) that lacks antipodal cells and a second polar nucleus (Fig. 1). Importantly, the target of the second fertilization event, the central cell of the female gametophyte, is haploid in Nymphaeales and Austrobaileyales, and endosperm in these ancient lineages is diploid, not triploid as in most flowering plants. Because of the recent phylogenetic insights into what constitute the most ancient angiosperm lineages, Hydatellaceae might be predicted to share some of the rare and potentially conservative embryological features of water lilies and other ancient lineages of flowering plants.

Working with field-collected material of Hydatella inconspicua, I examined female gametophyte development and ontogenetic features of the formation of embryo-nourishing reserves in the ovule/seed. The mature female gametophyte of H. inconspicua contains four uninucleate cells: the egg cell, two synergids and a uninucleate central cell (Fig. 2). No antipodal cells are evident, nor is a second polar nucleus present within the central cell. However, in many flowering plants with seven-celled, eight-nucleate female gametophytes (Polygonum-type), the three antipodal cells degenerate and the two haploid nuclei of the central cell fuse before the fertilization process. The result is a female gametophyte that is structurally indistinguishable, at maturity, from a ‘true’ four-celled female gametophyte (Nuphar/Schisandra-type) that never forms antipodal cells or a second polar nucleus. To distinguish between these two types of female gametophyte, early developmental stages must be examined (Fig. 1).

Critically, at the two-nucleate syncytial stage in H. inconspicua, both nuclei are located at the micropylar pole of the female gametophyte (Fig. 2a). This is precisely the pattern exhibited by other four-nucleate, four-celled gametophytes found in Nymphaeales and Austrobaileyales, and differs from the comparable ontogenetic stage of Polygonum-type (and Amborella-type) gametophytes where the nuclei at the two-nucleate syncytial stage are displaced to opposite poles of the female gametophyte (Fig. 1). Thus, Hydatella produces a four-celled, four-nucleate female gametophyte that precisely matches (developmentally and at maturity) the unique and potentially plesiomorphic pattern found in two of the most ancient extant angiosperm lineages (Fig. 2c). The finding that Hydatella produces a female gametophyte identical to those found in the water lilies is clearly concordant with both the phylogenetic placement of Hydatellaceae as sister to the previously recognized Nymphaeales and its general position among ancient extant angiosperm clades.

Although endosperm is initiated in H. inconspicua (Fig. 3), perisperm, a diploid tissue derived from the maternal sporophyte, is the major embryo-nourishing constituent within the seed, as is characteristic of all members of the Nymphaeales. The sexually formed endosperm ultimately occupies a very small portion of the maturing seed (Fig. 3) and does not play a significant role in the nourishment of the embryo.

Remarkably, the embryo-nourishing tissue within the ovule/seed in Hydatella begins to acquire significant carbon resources (starch) from the maternal plant before fertilization. At the two-nucleate stage of pre-fertilization female gametophyte development, the perisperm contains reserves of starch (Fig. 3). At the four-nucleate, four-celled stage of female gametophyte development, just before fertilization, cells of the perisperm are densely packed with starch grains. After fertilization (Supplementary Data) the seed continues to develop and the volume of the perisperm increases (Fig. 3). Overall seed size increases from approximately 450 µm in length by 235 µm in width at the time of fertilization to 620 µm in length by 420 µm in width at maturity. Thus in Hydatella a significant portion of the maternal commitment of embryo-nourishing carbon reserves to ovules/seeds occurs before fertilization.

Angiosperms (with the exception of Hydatellaceae) differ from extant gymnosperms (conifers, Ginkgo, cycads, Gnetales) in that maternal commitment of embryo-nourishing resources to seeds only occurs after fertilization. This ontogenetic delay in seed provisioning (compared with ancestral non-flowering seed plants) has long been viewed as an evolved and adaptive mechanism to allow the maternal plant to allocate limited resources efficiently only to those seeds that have been successfully fertilized. As such, post-fertilization maternal resource allocation has been hypothesized to be a key innovation associated with the origin and subsequent radiation of flowering plants.

The exceptional case of pre-fertilization maternal provisioning of a seed in Hydatella may well be an apomorphic feature of Hydatellaceae alone. If so, Hydatellaceae stands as an essentially unique angiosperm clade that has reverted to allocating resources to an ovule/seed before the initiation of an embryo (albeit to the perisperm and not to the female gametophyte, as in gymnosperms). Because most taxa of flowering plants use a sexually formed endosperm to nourish an embryo, pre-fertilization allocation of embryo-nourishing reserves to the seed is ontogenetically precluded. However, members of a large number of disparate clades of angiosperms, including many basal lineages (Nymphaeaceae, Cabombaceae, Trimeniaceae, Acoraceae, Ceratophyllaceae, Saururaceae, Piperaceae, Hydnoraceae), form an embryo-nourishing perisperm. None of these diverse perisperm-forming lineages (with the possible exception of Acorus) allocates significant carbon resources to the ovule/seed before fertilization. Importantly, if pre-fertilization provisioning of embryo-nourishing reserves should prove to be an apomorphy of Hydatellaceae, it is more strong proof that the earliest phases of flowering plant evolution were marked by a tremendous diversification of reproductive features.

Alternatively, pre-fertilization maternal resource allocation to ovules/seeds, and specifically to a maternally derived perisperm, in Hydatella could represent a plesiomorphic and transitional condition associated with the origin of flowering plants. If so, the ‘underdeveloped’ endosperm of Hydatellaceae and Nymphaeales is not reduced, but rather represents an intermediate condition (between gymnosperms and other angiosperms) in which the endosperm has not yet achieved its fully fledged role as the primary source of nutrients for the developing embryo. It is important to note that Amborella and most members of the Austrobaileyales contain a well-developed endosperm and do not form a perisperm (Trimenia being the reported exception). Nevertheless, the prospect that early angiosperms might have used both a perisperm and an endosperm to nourish the embryo within a seed (as previously hypothesized), and that the maternal plant allocated reserves to this perisperm before fertilization, is thoroughly congruent with the data derived from Hydatella and its new-found phylogenetic position.

Charles Darwin was among the first to recognize the immense chasm between gymnosperms and angiosperms with respect to their biological characteristics (letters to Oswald Heer in 1875 and to Joseph Hooker in 1879 and 1882). Since then, inferring the vegetative, floral and reproductive features that defined the first angiosperms, as well as their evolutionary (transformational) links to a gymnospermous ancestor, has proved to be fraught with difficulties. Indeed, the past five years have witnessed the near-global collapse of a century-old set of paradigms concerning the embryological features of the earliest angiosperms. With the present finding that maternal plants of Hydatella provision ovules/seeds with embryo-nourishing reserves before fertilization, yet another long-standing hypothesis for a presumed angiosperm-defining biological feature appears poised to be overturned.

Methods Summary

Hydatella inconspicua was collected on 6 December 2006 at Kai Iwi Lake, Northland, New Zealand, by P. Champion, New Zealand National Institute of Water and Atmospheric Research.

Plants were chemically fixed in 4% glutaraldehyde, washed in phosphate buffer and stored in water. Specimens were dehydrated through an ethanol series, then infiltrated and embedded in glycol methacrylate (JB-4 embedding kit, Electron Microscopy Sciences). Embedded flowers were serially sectioned into 4-&mgr;m thick ribbons. Sectioned flowers were stained with 0.1% toluidine blue and examined under brightfield and cross-polarization conditions. Digital imaging was on a Zeiss Axiocam digital camera using brightfield and cross polarization optics. Images were processed with Adobe Photoshop CS2. Image manipulations were restricted to operations applied to the entire image, except as noted in specific figure legends.

Schematic of four-celled, four-nucleate (Nuphar/Schisandra-type) female gametophyte development in Nymphaeales and Austrobaileyales and of seven-celled, eight-nucleate (Polygonum-type) development in most other angiosperms.

The micropylar pole is towards the top of the figure. At the two-nucleate syncytial stage (blue boxes), water-lily female gametophytes have both nuclei at the micropylar pole; in Polygonum-type female gametophytes, a nuclear migration event leads to placement of a single nucleus at each pole. Red, synergids; yellow, egg cell; brown, antipodals. cc, central cell. pn, polar nucleus.

Female gametophyte development in Hydatella inconspicua.

a, Two-nucleate syncytial female gametophyte in H. inconspicua, with both nuclei (red arrowheads) at micropylar pole. Grey box contains digital superposition of second nucleus from adjacent histological section. b, Mature four-celled female gametophyte in H. inconspicua. Grey box contains digital superposition of second synergid nucleus from adjacent histological section. cc, central cell; ec, egg cell; pn, polar nucleus; sc, synergid cell.

Pre- and post-fertilization development of perisperm in H. inconspicua.

a, Longitudinal section of perisperm in pre-fertilization ovule at two-nucleate stage of female gametophyte development. b, Cross-polarization optical image demonstrating presence of starch from boxed region in a. Starch grains, which are birefringent, appear as four white quadrants separated by a black cross. c, Longitudinal section of perisperm in pre-fertilization ovule just before fertilization. d, Cross-polarization optical image demonstrating presence of large amounts of starch from boxed region in c. Starch grains, which are birefringent, appear as four white quadrants separated by a black cross. e, Longitudinal section of perisperm in post-fertilization seed with three-celled embryo and minimally developed endosperm. f, Higher magnification from boxed region in e of embryo surrounded by a small endosperm tissue that lacks significant storage reserves. The perisperm below the endosperm contains large amounts of starch (bright circular structures). Scale bars, 10 &mgr;m. em, embryo; es, endosperm; fg, female gametophyte; ps, perisperm.

I thank: P. Champion and A. Drinnan for collecting plant materials; S. Holloway for histological work; S. Renner for translation of the embryological studies of U. Hamaan; and P. Diggle, L. Hufford, J. Williams and R. Robichaux for feedback on this manuscript. This work was supported by a National Science Foundation Research Grant.

Saarela, J. M. et al. Hydatellaceae identified as a new branch near the base of the angiosperm phylogenetic tree. Nature 446, 312315 (2007) Rudall, P. J. et al. Morphology of Hydatellaceae, an anomalous aquatic family recently recognized as an early-divergent angiosperm lineage. Am. J. Bot. 94, 10731092 (2007) Williams, J. H. & Friedman, W. E. Identification of diploid endosperm in an early angiosperm lineage. Nature 415, 522525 (2002) Tobe, H., Kimoto, Y. & Prakash, N. Development and structure of the female gametophyte in Austrobaileya scandens (Austrobaileyaceae). J. Plant Res. 120, 431436 (2007) Hamann, U. Neue Untersuchungen zur Embryologie und Systematik der Centrolepidaceae. Bot. Jahr. 96, 154191 (1975) Johri, B. M., Ambegaokar, K. B. & Srivastava, P. S. Comparative Embryology of Angiosperms. (McGraw-Hill, New York, 2002) Friedman, W. E. & Carmichael, J. S. Heterochrony and developmental innovation: evolution of female gametophyte ontogeny in Gnetum, a highly apomorphic seed plant. Evolution 52, 10161030 (1998) Mathews, S. & Donoghue, M. J. The root of angiosperm phylogeny inferred from duplicate phytochrome genes. Science 286, 947950 (1999) Parkinson, C. L., Adams, K. L. & Palmer, J. D. Multigene analyses identify the three earliest lineages of extant flowering plants. Curr. Biol. 9, 14851488 (1999) Qiu, Y.-L. et al. The earliest angiosperms: evidence from mitochondrial, plastid and nuclear genomes. Nature 402, 404407 (1999) Soltis, P. S., Soltis, D. E. & Chase, M. W. Angiosperm phylogeny inferred from multiple genes as a research tool for comparative biology. Nature 402, 402404 (1999) Graham, S. W. & Olmstead, R. G. Utility of 17 chloroplast genes for inferring the phylogeny of the basal angiosperms. Am. J. Bot. 87, 17121730 (2000) Friedman, W. E. Embryological evidence for developmental lability during early angiosperm evolution. Nature 441, 337340 (2006) Hamann, U. Hydatellaceae – a new family of Monocotyledoneae. N.Z. J. Bot. 14, 193196 (1976) Hamann, U. in Families and Genera of Vascular Plants vol. IV (ed. Kubitzki, K.) 231235 (Springer, Berlin, 1998) Friedman, W. E., Gallup, W. N. & Williams, J. H. Female gametophyte development in Kadsura: implications for Schisandraceae, Austrobaileyales, and the early evolution of flowering plants. Int. J. Plant Sci. 164 (suppl.). S294S305 (2003) Friedman, W. E., Madrid, E. N. & Williams, J. H. Origin of the fittest and survival of the fittest: relating female gametophyte development to endosperm genetics. Int. J. Plant Sci. 169, 7992 (2008) Rudall, P. J. The nucellus and chalaza in monocotyledons: structure and systematics. Bot. Rev. 63, 140184 (1997) Schneider, E. L. Morphological studies of the Nymphaeaceae. IX. The seed of Barclaya longifolia Wall. Bot. Gaz. 139, 223230 (1978) Floyd, S. K. & Friedman, W. E. Developmental evolution of endosperm in basal angiosperms: evidence from Amborella (Amborellaceae), Nuphar (Nymphaeaceae), and Illicium (Illiciaceae). Plant Syst. Evol. 228, 153169 (2001) Stebbins, G. L. Flowering Plants: Evolution above the Species Level (Harvard Univ. Press, Cambridge, Massachusetts, 1974) Tiffney, B. H. in Paleobotany, Paleoecology, and Evolution (ed. Niklas, K. J.) 193230 (Praeger, New York, 1981) Cocucci, A. E. Estudios en el género Prosopanche (Hydnoraceae). III Embriologia. Kurtziana 9, 1939 (1976) Rudall, P. J. & Furness, C. A. Systematics of Acorus: ovule and anther. Int. J. Plant Sci. 158, 640651 (1997) Floyd, S. K. & Friedman, W. E. Evolution of endosperm developmental patterns among basal flowering plants. Int. J. Plant Sci. 161 (suppl.). S57S81 (2000) Prakash, N. & Bak, H. K. Flower and fruit development in Piper nigrum L. cv. Kuching. Malays. J. Sci. 7, 1119 (1982) Prakash, N. in Plant Form and Function (eds Bhatia, B., Shukla, A. K. & Sharma, H. L.) 207216 (Angkor, New Delhi, 1998) Johnson, D. S. On the development of Saururus cernuus L. Bull. Torrey Bot. Club 27, 365372 (1900) Shamrov, I. I. in Embryology of Flowering Plants (ed. Batygina, T. B.) 169170 (Science Publishers, Enfield, New Hampshire, 2006) Doyle, J. A. in Early Evolution of Flowers (eds Endress, P. K. & Friis, E. M.) 729 (Springer, Vienna, 1994)

doi: 10.1038/nature06733


Hierarchical structure and the prediction of missing links in networks p.98

version="1.0" encoding="utf-8"?> ]>

Nature 453 7191 20080501 981014 0028-0836 1476-4687 2008Nature Publishing Group Supplementary Notes

This file contains Supplementary Notes including the technical details of our hierarchical model and the methods used to fit it to empirical data. It also contains addition results on graph resampling and the prediction of missing links, and the algorithmic specifics of our experimental studies.

Hierarchical structure and the prediction of missing links in networks AaronClausetA CristopherMooreC M. E. J.NewmanM E J Department of Computer Science, and, Department of Physics and Astronomy, University of New Mexico, Albuquerque, New Mexico 87131, USA Santa Fe Institute, 1399 Hyde Park Road, Santa Fe, New Mexico 87501, USA Department of Physics and Center for the Study of Complex Systems, University of Michigan, Ann Arbor, Michigan 48109, USA Correspondence and requests for materials should be addressed to A.C. (aaronc@santafe.edu). &e080501-13; &nature06830-s1;

Networks have in recent years emerged as an invaluable tool for describing and quantifying complex systems in many branches of science. Recent studies suggest that networks often exhibit hierarchical organization, in which vertices divide into groups that further subdivide into groups of groups, and so forth over multiple scales. In many cases the groups are found to correspond to known functional units, such as ecological niches in food webs, modules in biochemical networks (protein interaction networks, metabolic networks or genetic regulatory networks) or communities in social networks. Here we present a general technique for inferring hierarchical structure from network data and show that the existence of hierarchy can simultaneously explain and quantitatively reproduce many commonly observed topological properties of networks, such as right-skewed degree distributions, high clustering coefficients and short path lengths. We further show that knowledge of hierarchical structure can be used to predict missing connections in partly known networks with high accuracy, and for more general network structures than competing techniques. Taken together, our results suggest that hierarchy is a central organizing principle of complex networks, capable of offering insight into many network phenomena.

Much recent work has been devoted to the study of clustering and community structure in networks. Hierarchical structure goes beyond simple clustering, however, by explicitly including organization at all scales in a network simultaneously. Conventionally, hierarchical structure is represented by a tree, or dendrogram, in which closely related pairs of vertices have lowest common ancestors that are lower in the tree than those of more distantly related pairs (see Fig. 1). We expect the probability of a connection between two vertices to depend on their degree of relatedness. Structure of this type can be modelled mathematically by using a probabilistic approach in which we endow each internal node r of the dendrogram with a probability pr and then connect each pair of vertices for which r is the lowest common ancestor independently with probability pr (Fig. 1).

This model, which we call a hierarchical random graph, is similar in spirit to (although different in realization from) the tree-based models used in some studies of network search and navigation. Like most work on community structure, it assumes that communities at each level of organization are disjoint. Overlapping communities have occasionally been studied (see, for example, ref. 14) and could be represented with a more elaborate probabilistic model; however, as we discuss below, the present model already captures many of the structural features of interest.

Given a dendrogram and a set of probabilities pr, the hierarchical random graph model allows us to generate artificial networks with a specified hierarchical structure, a procedure that might be useful in certain situations. Our goal here, however, is a different one. We wish to detect and analyse the hierarchical structure, if any, of networks in the real world. We accomplish this by fitting the hierarchical model to observed network data by using the tools of statistical inference, combining a maximum-likelihood approach with a Monte Carlo sampling algorithm on the space of all possible dendrograms. This technique allows us to sample hierarchical random graphs with probability proportional to the likelihood that they generate the observed network. To obtain the results described below we combine information from a large number of such samples, each of which is a reasonably likely model of the data.

The success of this approach relies on the flexible nature of our hierarchical model, which allows us to fit a wide range of network structures. The traditional picture of communities or modules in a network, for example, corresponds to connections that are dense within groups of vertices and sparse between them—a behaviour called ‘assortativity’ in the literature. The hierarchical random graph can capture behaviour of this kind using probabilities pr that decrease as we move higher up the tree. Conversely, probabilities that increase as we move up the tree correspond to ‘disassortative’ structures in which vertices are less likely to be connected on small scales than on large ones. By letting the pr values vary arbitrarily throughout the dendrogram, the hierarchical random graph can capture both assortative and disassortative structure, as well as arbitrary mixtures of the two, at all scales and in all parts of the network.

To demonstrate our method we have used it to construct hierarchical decompositions of three example networks drawn from disparate fields: the metabolic network of the spirochaete Treponema pallidum, a network of associations between terrorists, and a food web of grassland species. To test whether these decompositions accurately capture the important structural features of the networks, we use the sampled dendrograms to generate new networks, different in detail from the originals but, by definition, having similar hierarchical structure (see Supplementary Information for more details). We find that these ‘resampled’ networks match the statistical properties of the originals closely, including their degree distributions, clustering coefficients, and distributions of shortest path lengths between pairs of vertices, despite the fact that none of these properties is explicitly represented in the hierarchical random graph (Table 1, and Supplementary Fig. 3). It therefore seems that a network’s hierarchical structure is capable of explaining a wide variety of other network features as well.

The dendrograms produced by our method are also of interest in themselves, as a graphical representation and summary of the hierarchical structure of the observed network. As discussed above, our method can generate not just a single dendrogram but a set of dendrograms, each of which is a good fit to the data. From this set we can, by using techniques from phylogeny reconstruction, create a single consensus dendrogram, which captures the topological features that appear consistently across all or a large fraction of the dendrograms and typically is a better summary of the network’s structure than any individual dendrogram. Figure 2a shows such a consensus dendrogram for the grassland species network, which clearly reveals communities and subcommunities of plants, herbivores, parasitoids and hyperparasitoids.

Another application of the hierarchical decomposition is the prediction of missing interactions in networks. In many settings, the discovery of interactions in a network requires significant experimental effort in the laboratory or the field. As a result, our current pictures of many networks are substantially incomplete. An alternative to checking exhaustively for a connection between every pair of vertices in a network is to try to predict, in advance and on the basis of the connections already observed, which vertices are most likely to be connected, so that scarce experimental resources can be focused on testing for those interactions. If our predictions are good, we can in this way substantially reduce the effort required to establish the network’s topology.

The hierarchical decomposition can be used as the basis for an effective method of predicting missing interactions as follows. Given an observed but incomplete network, we generate, as described above, a set of hierarchical random graphs—dendrograms and the associated probabilities pr—that fit that network. Then we look for pairs of vertices that have a high average probability of connection within these hierarchical random graphs but are unconnected in the observed network. These pairs we consider the most likely candidates for missing connections. (Technical details of the procedure are given in Supplementary Information.)

We demonstrate the method by using our three example networks again. For each network we remove a subset of connections chosen uniformly at random and then attempt to predict, on the basis of the remaining connections, which have been removed. A standard metric for quantifying the accuracy of prediction algorithms, commonly used in the medical and machine learning communities, is the AUC statistic, which is equivalent to the area under the receiver operating characteristic (ROC) curve. In the present context, the AUC statistic can be interpreted as the probability that a randomly chosen missing connection (a true positive) is given a higher score by our method than a randomly chosen pair of unconnected vertices (a true negative). Thus, the degree to which the AUC exceeds 0.5 indicates how much better our predictions are than chance. Figure 2 shows the AUC statistic for the three networks as a function of the fraction of the connections known to the algorithm. For all three networks our algorithm does far better than chance, indicating that hierarchy is a strong general predictor of missing structure. It is also instructive to compare the performance of our method with that of other methods for link prediction. Previously proposed methods include assuming that vertices are likely to be connected if they have many common neighbours, if there are short paths between them, or if the product of their degrees is large. These approaches work well for strongly assortative networks such as collaboration and citation networks and for the metabolic and terrorist networks studied here (Fig. 3a, b). Indeed, for the metabolic network the shortest-path heuristic performs better than our algorithm.

However, these simple methods can be misleading for networks that exhibit more general types of structure. In food webs, for instance, pairs of predators often share prey species but rarely prey on each other. In such situations a common-neighbour or shortest-path-based method would predict connections between predators where none exists. The hierarchical model, by contrast, is capable of expressing both assortative and disassortative structure and, as Fig. 3c shows, gives substantially better predictions for the grassland network. (Indeed, in Fig. 2b there are several groups of parasitoids that our algorithm has grouped together in a disassortative community, in which they prey on the same herbivore but not on each other.) The hierarchical method thus makes accurate predictions for a wider range of network structures than the previous methods.

In the applications above, we have assumed for simplicity that there are no false positives in our network data; that is, that every observed edge corresponds to a real interaction. In networks in which false positives may be present, however, they too could be predicted by using the same approach: we would simply look for pairs of vertices that have a low average probability of connection within the hierarchical random graph but are connected in the observed network.

The method described here could also be extended to incorporate domain-specific information, such as species’ morphological or behavioural traits for food webs or phylogenetic or binding-domain data for biochemical networks, by adjusting the probabilities of edges accordingly. As the results above show, however, we can obtain good predictions even in the absence of such information, indicating that topology alone can provide rich insights.

In closing, we note that our approach differs crucially from previous work on hierarchical structure in networks in that it acknowledges explicitly that most real-world networks have many plausible hierarchical representations of roughly equal likelihood. Previous work, by contrast, has typically sought a single hierarchical representation for a given network. By sampling an ensemble of dendrograms, our approach avoids over-fitting the data and allows us to explain many common topological features, to generate resampled networks with similar structure to the original, to derive a clear and concise summary of a network’s structure by means of its consensus dendrogram, and to accurately predict missing connections in a wide variety of situations.

Methods Summary

Computer code implementing many of the analysis methods described in this paper can be found online at http://www.santafe.edu/~aaronc/randomgraphs/.

A hierarchical network with structure on many scales, and the corresponding hierarchical random graph.

Each internal node r of the dendrogram is associated with a probability pr that a pair of vertices in the left and right subtrees of that node are connected. (The shades of the internal nodes in the figure represent the probabilities.)

Application of the hierarchical decomposition to the network of grassland species interactions.

a, Consensus dendrogram reconstructed from the sampled hierarchical models. b, A visualization of the network in which the upper few levels of the consensus dendrogram are shown as boxes around species (plants, herbivores, parasitoids, hyperparasitoids and hyper-hyperparasitoids are shown as circles, boxes, down triangles, up triangles and diamonds, respectively). Note that in several cases a set of parasitoids is grouped into a disassortative community by the algorithm, not because they prey on each other but because they prey on the same herbivore.

Comparison of link prediction methods.

Average AUC statistic—that is, the probability of ranking a true positive over a true negative—as a function of the fraction of connections known to the algorithm, for the link prediction method presented here and a variety of previously published methods. a, Terrorist association network; b, T. pallidum metabolic network; c, grassland species network.

Comparison of original and resampled networks Network kreal ksamp Creal Csamp dreal dsamp

Statistics are shown for the three example networks studied and for new networks generated by resampling from our hierarchical model. The generated networks closely match the average degree ⟨k⟩, clustering coefficient C and average vertex–vertex distance d in each case, suggesting that they capture much of the structure of the real networks. Parenthetical values indicate standard errors on the final digits.

T. pallidum 4.8 3.7(1) 0.0625 0.0444(2) 3.690 3.940(6) Terrorists 4.9 5.1(2) 0.361 0.352(1) 2.575 2.794(7) Grassland 3.0 2.9(1) 0.174 0.168(1) 3.29 3.69(2)

We thank J. Dunne, M. Gastner, P. Holme, M. Huss, M. Porter, C. Shalizi and C. Wiggins for their help, and the Santa Fe Institute for its support. C.M. thanks the Center for the Study of Complex Systems at the University of Michigan for hospitality while some of this work was conducted.

Wasserman, S. & Faust, K. Social Network Analysis (Cambridge Univ. Press, Cambridge, 1994) Albert, R. & Barabási, A.-L. Statistical mechanics of complex networks. Rev. Mod. Phys. 74, 4797 (2002) Newman, M. E. J. The structure and function of complex networks. SIAM Rev. 45, 167256 (2003) Ravasz, E., Somera, A. L., Mongru, D. A., Oltvai, Z. N. & Barabási, A.-L. Hierarchical organization of modularity in metabolic networks. Science 30, 15511555 (2002) Clauset, A., Newman, M. E. J. & Moore, C. Finding community structure in very large networks. Phys. Rev. E 70, 066111 (2004) Guimera, R. & Amaral, L. A. N. Functional cartography of complex metabolic networks. Nature 433, 895900 (2005) Lagomarsino, M. C., Jona, P., Bassetti, B. & Isambert, H. Hierarchy and feedback in the evolution of the Escherichia coli transcription network. Proc. Natl Acad. Sci. USA 104, 55165520 (2001) Liben-Nowell, D. & Kleinberg, J. M. The link-prediction problem for social networks. J. Am. Soc. Inform. Sci. Technol. 58, 10191031 (2007) Girvan, M. & Newman, M. E. J. Community structure in social and biological networks. Proc. Natl Acad. Sci. USA 99, 78217826 (2002) Krause, A. E., Frank, K. A., Mason, D. M., Ulanowicz, R. E. & Taylor, W. W. Compartments revealed in food-web structure. Nature 426, 282285 (2003) Radicchi, F., Castellano, C., Cecconi, F., Loreto, V. & Parisi, D. Defining and identifying communities in networks. Proc. Natl Acad. Sci. USA 101, 26582663 (2004) Watts, D. J., Dodds, P. S. & Newman, M. E. J. Identity and search in social networks. Science 296, 13021305 (2002) Kleinberg, J. in Proc. 2001 Neural Inform. Processing Systems Conf. (eds Dietterich, T. G., Becker, S. & Ghahramani, Z.) 431438 (MIT Press, Cambridge, MA, 2002) Palla, G., Derényi, I., Farkas, I. & Vicsek, T. Uncovering the overlapping community structure of complex networks in nature and society. Nature 435, 814818 (2005) Casella, G. & Berger, R. L. Statistical Inference (Duxbury, Belmont, 2001) Newman, M. E. J. & Barkema, G. T. Monte Carlo Methods in Statistical Physics (Clarendon, Oxford, 1999) Newman, M. E. J. Assortative mixing in networks. Phys. Rev. Lett. 89, 208701 (2002) Huss, M. & Holme, P. Currency and commodity metabolites: Their identification and relation to the modularity of metabolic networks. IET Syst. Biol. 1, 280285 (2007) Krebs, V. Mapping networks of terrorist cells. Connections 24, 4352 (2002) Dawah, H. A., Hawkins, B. A. & Claridge, M. F. Structure of the parasitoid communities of grass-feeding chalcid wasps. J. Anim. Ecol. 64, 708720 (1995) Bryant, D. in BioConsensus (eds Janowitz, M., Lapointe, F.-J., McMorris, F. R., Mirkin, B. & Roberts, F.) pp. 163184 (Series in Discrete Mathematics and Theoretical Computer Science, Vol. 61, American Mathematical Society-DIMACS, Providence, RI, 2003) Dunne, J. A., Williams, R. J. & Martinez, N. D. Food-web structure and network theroy: The role of connectance and size. Proc. Natl Acad. Sci. USA 99, 1291712922 (2002) Szilágyi, A., Grimm, V., Arakaki, A. K. & Skolnick, J. Prediction of physical protein–protein interactions. Phys. Biol. 2, S1S16 (2005) Sprinzak, E., Sattath, S. & Margalit, H. How reliable are experimental protein-protein interaction data? J. Mol. Biol. 327, 919923 (2003) Ito, T. et al. A comprehensive two-hybrid analysis to explore the yeast protein interactome. Proc. Natl Acad. Sci. USA 98, 45694574 (2001) Lakhina, A., Byers, J. W., Crovella, M. & Xie, P. in INFOCOM 2003: Twenty-Second Annual Joint Conf. IEEE Computer and Communications Societies (ed. Bauer, F.) Vol. 1 332341 (IEEE, Piscataway, New Jersey, 2003) Clauset, A. & Moore, C. Accuracy and scaling phenomena in Internet mapping. Phys. Rev. Lett. 94, 018701 (2005) Martinez, N. D., Hawkins, B. A., Dawah, H. A. & Feifarek, B. P. Effects of sampling effort on characterization of food-web structure. Ecology 80, 10441055 (1999) Hanely, J. A. & McNeil, B. J. The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology 143, 2936 (1982) Sales-Pardo, M., Guimerá, R., Moreira, A. A. & Amaral, L. A. N. Extracting the hierarchical organization of complex systems. Proc. Natl Acad. Sci. USA 104, 1522415229 (2007)

doi: 10.1038/nature06830


Melanopsin cells are the principal conduits for rod-cone input to non-image-forming vision p.102

version="1.0" encoding="utf-8"?> ]>

Nature 453 7191 20080501 1021050 0028-0836 1476-4687 2008Nature Publishing Group Supplementary information

The file contains Supplementary Figures 1-7 with Legends and Supplementary Tables 1-2.

Melanopsin cells are the principal conduits for rod–cone input to non-image-forming vision Ali D.GülerA D Jennifer L.EckerJ L Gurprit S.LallG S ShafiqulHaqS Cara M.AltimusC M Hsi-WenLiaoH Alun R.BarnardA R HughCahillH Tudor C.BadeaT C HaiqingZhaoH Mark W.HankinsM W David M.BersonD M Robert J.LucasR J King-WaiYauK SamerHattarS Department of Biology, Johns Hopkins University, Baltimore, Maryland 21218, USA Faculty of Life Sciences, University of Manchester, Manchester M13 9PT, UK Department of Neuroscience, and, Molecular Biology and Genetics, Johns Hopkins University School of Medicine, Baltimore, Maryland 21205, USA Visual Neuroscience, University of Oxford, Oxford OX3 7BN, UK Department of Neuroscience, Brown University, Providence, Rhode Island 02912, USA These authors contributed equally to this work. Correspondence and requests for materials should be addressed to S. Hattar (shattar@jhu.edu) or R.J.L. (robert.lucas@manchester.ac.uk). &e080501-14; &nature06829-s1;

Rod and cone photoreceptors detect light and relay this information through a multisynaptic pathway to the brain by means of retinal ganglion cells (RGCs). These retinal outputs support not only pattern vision but also non-image-forming (NIF) functions, which include circadian photoentrainment and pupillary light reflex (PLR). In mammals, NIF functions are mediated by rods, cones and the melanopsin-containing intrinsically photosensitive retinal ganglion cells (ipRGCs). Rod–cone photoreceptors and ipRGCs are complementary in signalling light intensity for NIF functions. The ipRGCs, in addition to being directly photosensitive, also receive synaptic input from rod–cone networks. To determine how the ipRGCs relay rod–cone light information for both image-forming and non-image-forming functions, we genetically ablated ipRGCs in mice. Here we show that animals lacking ipRGCs retain pattern vision but have deficits in both PLR and circadian photoentrainment that are more extensive than those observed in melanopsin knockouts. The defects in PLR and photoentrainment resemble those observed in animals that lack phototransduction in all three photoreceptor classes. These results indicate that light signals for irradiance detection are dissociated from pattern vision at the retinal ganglion cell level, and animals that cannot detect light for NIF functions are still capable of image formation.

The retinal ganglion cells that express melanopsin (rendering them intrinsically photosensitive) send monosynaptic projections to the suprachiasmatic nucleus (SCN) and the intergeniculate leaflet (IGL), responsible for circadian photoentrainment, and the olivary pretectal nucleus (OPN), responsible for PLR. It has been reported that RGCs that do not contain melanopsin innervate the same NIF brain centres. In mice genetically engineered to lack melanopsin protein, the RGCs that would normally express this opsin still project to the NIF centres, but these cells are no longer intrinsically photosensitive. In these animals, PLR and circadian light responses are reduced but not absent, indicating that rods and cones are capable of light detection for NIF functions. The rod–cone signals for NIF functions are undoubtedly relayed to the brain by RGCs, but it is unclear whether ipRGCs, conventional RGCs, or both are responsible (Fig. 1a).

To eliminate ipRGCs, we introduced a gene (aDTA) encoding attenuated diphtheria toxin A subunit (aDTA) into the mouse gene locus encoding melanopsin (Supplementary Fig. 1a, b). Using antibody staining in retinas from animals expressing aDTA (Opn4aDTA/+), we found that 3.1 ± 1.5% (mean ± s.e.m.) of melanopsin cells remained in the Opn4aDTA/+ animals (Fig. 1b). Staining with 5-bromo-4-chloro-3-indolyl-&bgr;-d-galactoside (X-gal) in animals with the aDTA gene in one melanopsin allele and tau-LacZ in the other (Opn4aDTA/tau-LacZ) revealed that 17.2 ± 1.0% of melanopsin cells remained in the retinas of these animals at six months of age (Fig. 1c). Consistent with the elimination of ipRGCs, we observed fewer fibre terminals in the SCN, IGL and OPN of Opn4aDTA/tau-LacZ animals than in those of Opn4tau-LacZ/+ animals (Fig. 2a–c). The degree of ipRGC ablation in the heterozygous animals increased with age (Supplementary Fig. 2a–c).

To ablate ipRGCs more completely, we generated animals homozygous for aDTA (Opn4aDTA/aDTA). The morphology and thickness of the retina in the Opn4aDTA/aDTA animals were not different from those in wild types (Fig. 1d). Injection of fluorescently conjugated cholera toxin into the eye, which labels all ganglion cell fibres from the retina, showed that few fibres innervated the SCN and IGL in the Opn4aDTA/aDTA animals (Fig. 2d, e). Furthermore, comparison between age-matched Opn4aDTA/+ and Opn4aDTA/aDTA mice verified that the extent of ipRGC ablation was greater in homozygous animals (Supplementary Fig. 2d). We also observed that target innervation by other RGCs was unaffected both in the dorsal lateral geniculate nucleus, which is important for image formation, and in the rostral core of the OPN (Fig. 2e, f).

To assess whether image-forming functions are affected in animals expressing aDTA, we measured electroretinograms, optokinetic nystagmus responses, visual acuity and the ability of the animals to detect a visual cue. We found that the electroretinograms and optokinetic nystagmus responses were normal in animals lacking ipRGCs (Supplementary Fig. 3a, b). On the basis of optomotor responses, the acuity of Opn4aDTA/aDTA mice was slightly decreased compared with that of the wild-type mice (Fig. 1e) and Opn4tau-LacZ/tau-LacZ mice (Supplementary Fig. 3d). This effect is most probably due to enlarged pupil diameters in the Opn4aDTA/aDTA animals (see Fig. 3c and Supplementary Fig. 3c). We also determined that these animals could use a visual cue to locate a flag-marked platform in a water maze (Fig. 1e). These results demonstrate that image formation is functional despite the elimination of ipRGCs in Opn4aDTA/aDTA animals. We could therefore determine the relative contribution of ipRGCs to PLR and circadian photoentrainment in the context of normal image formation.

Pupil constriction regulates the amount of light entering the eye, and thus pupil diameter is negatively correlated with light intensity. At high light intensity, the iris decreases the area of the pupil by 95% (full constriction) in comparison with dark-adapted conditions (fully dilated). At low light intensity in which the pupil constricts by 50% or less, rod–cone input is the main signal (Supplementary Table 1). In contrast, the intrinsic photosensitivity through the melanopsin protein in ipRGCs is necessary for full pupil constriction (Supplementary Table 1). We found that the PLR was absent in all Opn4aDTA/+ mice at a light intensity that constricts the pupil in wild-type animals to about 50% (Fig. 3a, b). This result suggests that although the melanopsin protein is not required for PLR at low light intensity, rod–cone input still requires the ipRGCs. We propose that ipRGCs are able to encode light intensities that are below the detection level of the melanopsin photopigment by integrating rod–cone signals and relaying this information to the OPN for pupil constriction.

At high light intensity, six of nine Opn4aDTA/+ mice constricted their pupils to 95%, in a similar manner to wild-type animals (Fig. 3a, b). Full pupil constriction in these six animals indicates that the intrinsic photosensitivity persists in the remaining ipRGCs in Opn4aDTA/+ mice and that only about 17% of ipRGCs are sufficient to drive full pupil constriction. The remaining three animals had pupil constriction defects at high light intensity (59 ± 14%, in contrast with 96 ± 1% in the wild type; Fig. 3b). Interestingly, when the Opn4aDTA/+ animals were placed under 24-h light/dark cycles, animals that had full pupil constriction photoentrained (Supplementary Fig. 4a), whereas animals with defective pupil constriction had photoentrainment defects (Supplementary Fig. 4b).

The aDTA homozygotes have greater ablation of ipRGCs, and consequently all 12 Opn4aDTA/aDTA animals had defective PLR (42 ± 6%) at high light intensity (Fig. 3c, d). A similar pupil constriction (about 40%) to that observed in Opn4aDTA/aDTA mice is achieved by both wild-type and melanopsin knockout animals in response to a light stimulation that is about 1,000-fold lower in intensity. This reduction in pupil sensitivity was observed in an intensity–response curve in Opn4aDTA/aDTA mice at constrictions that are 55% or higher in the wild types (Supplementary Fig. 5). Together, these data show that the rod–cone-dependent pupil constriction is reliant on signalling through ipRGCs at all light intensities (Fig. 3 and Supplementary Fig. 5). The residual pupil constriction observed in the homozygous animals may reveal a possible role for the non-melanopsin RGCs in pupil constriction. Alternatively, this residual pupil constriction could originate from rod–cone input through a few remaining ipRGCs in the Opn4aDTA/aDTA mice.

To assess the contribution of ipRGCs to circadian photoentrainment, we analysed the wheel running activity of wild-type (n = 11) and Opn4aDTA/aDTA (n = 12) animals using 24-h light/dark cycles. Under constant dark conditions (DD), both wild-type and Opn4aDTA/aDTA animals had a functional circadian oscillator with period lengths of 23.3 ± 0.1 and 23.8 ± 0.1 h, respectively (Fig. 4a). In a similar manner to math5-/- mice with severe optic nerve hypoplasia due to RGC loss, Opn4aDTA/aDTA animals had significantly longer periods than wild types (Supplementary Table 2), confirming the loss of the fibres projecting to the SCN. A light stimulus (1,500 lx) administered for 15 min at circadian time (CT)16 delayed the phase onset of the activity of the wild types by 1.66 ± 0.23 h, whereas it did not affect the phase onset of activity in any Opn4aDTA/aDTA animals (-0.06 ± 0.09 h) (Fig. 4a). When the light/dark cycle was advanced or delayed by 6 h, the activity of all wild-type animals synchronized with the shifted cycle (Fig. 4b and Supplementary Table 2). Opn4aDTA/aDTA animals segregated into two groups with distinct responses. The first group of animals (8 of 12) free-ran completely under the shifted light/dark cycle (Fig. 4b and Supplementary Fig. 6b–f) in a similar manner to animals lacking all functional photoreceptors in the retina, indicating that they were completely blind to the light shift. Animals in the second group (4 of 12) were weakly light responsive but not photoentrained, because they did not show a stable phase relation to the light/dark cycle (Supplementary Fig. 7). Given that these four animals demonstrated very weak light responsiveness in the light/dark cycle but were not phase-shifted by the 15-min light pulse, we tested whether constant light (LL) conditions would lengthen the circadian period of Opn4aDTA/aDTA mice. Under LL, 7 of 11 wild-type animals had periods longer than 24 h, whereas the remaining animals became completely arrhythmic (Supplementary Fig. 6a). In contrast, both groups of Opn4aDTA/aDTA animals had periods shorter than 24 h (Fig. 4c, Supplementary Figs 6b–f and 7, and Supplementary Table 2). It had previously been shown that animals lacking only the melanopsin protein are similar to wild-type controls in adjusting to a 24-h light/dark cycle and have minor defects in their period lengthening under constant light and phase-delaying pulses of light. The severe defects in circadian photoentrainment observed in animals lacking ipRGCs demonstrate that these cells are required not only for intrinsically signalling light information through the melanopsin protein but also for conveying rod–cone light information to the SCN. Although several reports indicated that RGCs that do not express melanopsin innervate the SCN, our results reveal that the contribution of rod–cone signalling through these RGCs for photoentrainment is negligible.

The acute effects of light on activity (also known as masking) can be studied by using an ultradian 7-h light/dark (3.5:3.5 LD) cycle that disrupts the oscillator. Under the ultradian regime, wild types confined their activity mostly to the dark portion of the ultradian cycle (84 ± 4%). In contrast, the Opn4aDTA/aDTA animals were active nearly randomly across an ultradian day, irrespective of light (64 ± 3%; Fig. 4d). This result demonstrates that melanopsin cells are essential for a direct light-driven physiological response that is independent of the circadian oscillator.

We have shown that the loss of ipRGCs does not influence image formation and therefore the involvement of this ganglion cell class in classical vision is only modulatory. In contrast, irradiance-dependent NIF functions are substantially impaired in the absence of ipRGCs. Given that the ipRGCs constitute less than 2% of the total RGC population, it is striking that light information for circadian photoentrainment and PLR is conveyed predominantly through these cells. Moreover, the ability to form pattern vision does not affect photoentrainment.

Methods Summary Animals

All experiments were conducted in accordance with National Institutes of Health guidelines and were approved by institutional animal care and use committees of the universities involved.

Behavioural analyses

We used behavioural tests measuring the integrity of the outer retina (electroretinograms), eye-tracking functions (optokinetic nystagmus), visual acuity (optomotor), object identification (water maze), pupil constriction (PLR), the period of the circadian oscillator (wheel running activity), the adjustment of the circadian clock to different light stimulations (circadian photoentrainment, phase shifting, and constant light) and direct light effects on activity (masking).

aDTA mice

Using the homologous arms that we used previously, we targeted the aDTA gene to the melanopsin locus. The targeting construct contained the diphtheria toxin A subunit and the neomycin resistance genes. The construct was flanked by 4.4 kilobases (kb) 5′ of the ATG site of the mouse melanopsin gene, and a 1.6-kb fragment containing 654 base pairs (bp) of exon 9 plus a 946-bp 3′ untranslated region. After electroporation of the linearized construct into 129.1 mouse strain embryonic stem cells and drug selection (400 &mgr;g ml-1 G418), one positive embryonic stem clone was injected into C57BL/6 blastocysts. Chimaeric animals were mated to C57BL/6 mice to produce heterozygous animals.


Whole retinas from Opn4aDTA/+ and wild-type animals, fixed in 70% ethanol, were immunostained with the C terminus melanopsin antibody (1:500 dilution). Fluorescently conjugated secondary antibody (Alexa Fluor 488 goat anti-rabbit IgG (1:1,000 dilution; Molecular Probes) was applied for 1–3 h.

Staining with X-gal

Animals anaesthetized by intraperitoneal injection of Avertin (20 ml kg-1) were perfused intracardially with 4% paraformaldehyde, and brains and eyes were isolated. Eye-cups or brain sections (50 &mgr;m) were first incubated in buffer B (100 mM phosphate buffer at pH 7.4, 2 mM MgCl2, 0.01% sodium deoxycholate, 0.02% IGEPAL) then stained for 3 days in buffer B plus 5 mM potassium ferricyanide, 5 mM potassium ferrocyanide and 1 mg ml-1 X-gal.

Cholera toxin injections in the eye

Mice were anaesthetized with Ketamine (80 mg kg-1)/Xylazine (8 mg kg-1). Eyes were injected intravitreally with 2 &mgr;l of cholera toxin B subunit conjugated with Alexa Fluor 488 or Alexa Fluor 555 (Invitrogen). Three days after injection, brains were isolated and sectioned.


Animals maintained under 12:12 light/dark cycles for three days were used to collect data within the light phase (after 50 min of dark adaptation). The electroretinogram set-up was similar to that used previously. Mydriatics (1% tropicamide and 2.5% phenylephrine) and hypromellose solution (0.3%) were used to dilate the pupil and retain corneal moisture in anaesthetized mice. A signal conditioner (Model 1902 Mark III; Cambridge Electronic Design) differentially amplified (×3,000) and filtered (bandpass filter cutoff 0.5–200 Hz) the signal before it was digitized (Model 1401 digitizer; Cambridge Electronic Design) and recorded (sampling rate 10 kHz) by means of Signal 2.15 software (Cambridge Electronic Design).

White light (10 ms in duration) was provided by a xenon arc source (Cairn Research). Neutral density filters (Edmund Optics) were used (unattenuated intensity 320 &mgr;W cm-2 at the cornea). For scotopic recordings, flashes were administered in the dark and testing began with the dimmest stimulus. Depending on the intensity, stimuli were presented at a rate of 0.5–0.2 Hz, and 6–20 repetitions were collected and averaged. Photopic electroretinograms were recorded by presentation of an unattenuated light (20 stimuli presented at a rate of 1 Hz) against blue-filtered (Grass blue filter; Astro-Med) rod-saturating background light (160 &mgr;W cm-2).

Optokinetic nystagmus responses

A mouse stabilized with a head post was placed into an acrylic holder in a 12-inch diameter drum. Computer-generated stimuli were projected down onto the drum walls. Black and white stripes 4° in width were rotated at 5° s-1. Eye movements of mice were captured with an infrared video system (ISCAN). The fast saccade components were counted in 30-s bins with an algorithm that evaluated the high-velocity eye movements that followed a low-velocity movement in the opposite direction.

Visual acuity

A virtual cylinder OptoMotry (Cerebral Mechanics) was used to determine visual acuity by measuring the image-tracking reflex of mice. A sine-wave grating was projected on the screen rotating in a virtual cylinder. The animal was assessed for a tracking response on stimulation for about 5 s. All acuity thresholds were determined by using the staircase method with 100% contrast.

Morris water maze

To assess the ability of mice to detect a visual cue, we trained the animals to find a platform marked by a 10-cm tall, high-contrast visual cue under bright light (500 lx) in an 85-cm pool. On day 1, mice were trained with four trials 15 min apart. On the following day, latency to find the island was recorded first with the cue and then without it.


All animals were dark-adapted for at least 1 h, and the eye of each animal receiving the photic stimulus was treated with 0.1% atropine before the start of recording. Measurements were restricted to the middle of the subjective day (CT4–8). One eye of each mouse was digitally captured at a frequency of one image per second for 63 s with a charge-coupled-device camera. The light stimuli (xenon arc light source) consisted of a 60-s pulse at an intensity of 3.8 mW cm-2 or 1.8 &mgr;W cm-2 of white light.

For the Opn4aDTA/aDTA experiments, the eye receiving the photic stimulus was treated with 1% atropine. While one eye received light stimulation from a 470-nm light-emitting-diode light source (E27-B24; 161 &mgr;W cm-2; Super Bright LEDs), a digital camcorder (DCRHC96; Sony) was used to record from the other eye (for 30 s) at 30 frames s-1 under a 940-nm light (LDP). The digital video recoding was deconstructed to individual frames with Blaze MediaPro software (Mystik Media). The percentage pupil constriction was calculated as the percentage of pupil area at 30 s after initiation of the stimulus (steady state) relative to the dilated pupil size.

Wheel running activity

Mice were placed in cages with a 4.5-inch running wheel, and their activity was monitored with VitalView software (Mini Mitter). The period was calculated with ClockLab (Actimetrics). For phase-shifting experiments, each animal was exposed to a light pulse (1,500 lx; CT16) for 15 min. After 41 days of constant dark, mice were re-entrained to 12:12 light/dark cycles for 19 days. Animals were then exposed to two jet-lag models: 16 days of a 6-h advance followed by 32 days of a 6-h delay. During the last two weeks of this treatment the animals were tested for PLR. Animals were then exposed to constant light for three weeks followed by ultradian 3.5:3.5 light/dark cycles.

Elimination of ipRGCs in mouse retina.

a, Model describing how rod–cone signalling through conventional RGCs or ipRGCs contributes to NIF functions. The role of ipRGCs in image formation is speculative (dotted line). b, Melanopsin antibody staining in retinas of 18-month-old wild-type (n = 6) and Opn4aDTA/+ (n = 12) mice. The white arrowhead indicates a surviving ipRGC. Scale bar, 200 &mgr;m. c, X-gal staining from Opn4tau-LacZ/+ (n = 6) and Opn4aDTA/tau-LacZ (n = 8). The surviving cells are weakly stained (black arrows). Scale bar, 500 &mgr;m. d, Cross-sections of Giemsa-stained retinas from 18-month-old Opn4aDTA/aDTA (aDTA/aDTA; n = 3) and wild-type (n = 3) mice. The morphology of retinas is indistinguishable between genotypes. GCL, ganglion cell layer; IPL, inner plexiform layer; INL, inner nuclear layer; OPL, outer plexiform layer; ONL, outer nuclear layer; OS, outer segment. Scale bar, 50 &mgr;m. e, Left, the acuity (in cycles per degree) of Opn4aDTA/aDTA mice (DTA; n = 11; red bar) was slightly decreased in comparison with wild-type mice (WT; n = 9; black bar). Right, the latency to locate a marked platform (cue) in a water maze was similar in Opn4aDTA/aDTA (DTA; n = 14; red bar) and wild-type (WT; n = 12; black bar) mice. This latency significantly differed from that for unmarked platform (no cue) tests. All statistical comparisons used Student’s t-test (asterisk, P < 0.05; two asterisks, P < 0.01); error bars indicate s.e.m.

The ipRGC fibres in the brain decrease in aDTA mice.

ac, X-gal staining in Opn4tau-LacZ/aDTA mice (tau-LacZ/aDTA; n = 2) shows that ipRGC innervation of the SCN, IGL and OPN is decreased in comparison with Opn4tau-LacZ/+ mice (tau-LacZ/+; n = 2). df, Ocular cholera toxin injections (left eye, green; right, red) of Opn4aDTA/aDTA mice (aDTA/aDTA; n = 11) and wild-type mice (n = 6). a, d, SCN innervation is sparse in aDTA mice. b, e, The dorsal lateral geniculate nucleus (LGN) is innervated similarly both in aDTA and wild-type animals, whereas few fibres remain in the IGL of mutant mice (outlined regions). c, f, The OPN shell is innervated by ipRGCs and the core is targeted by other RGCs. c, Fibres in the shell region are eliminated in Opn4tau-LacZ/aDTA animals. f, Fibres in the OPN core are retained in Opn4aDTA/aDTA mice. Scale bars, 200 &mgr;m.

Opn4aDTA mice have deficits in PLR.

a, All nine Opn4aDTA/+ (aDTA/+) mice showed defective PLR at a light intensity (1.8 &mgr;W cm-2; 30 s white light; Low) that induced 50% constriction in wild types. The Opn4aDTA/+ mice showed a 9.0 ± 6.0% constriction. Six of nine Opn4aDTA/+ mice had full pupil constriction at high light intensity (3 mW cm-2; High). The rest of the mutant mice (three of nine) showed defective PLR. b, Quantification of PLR data of wild-type (WT; n = 9; black squares) and Opn4aDTA/+ either photoentrained (green triangles; n = 6) or non-photoentrained (orange triangles; n = 3) animals. All statistical comparisons were made by Student’s t-test (two asterisks, P < 0.01). c, All Opn4aDTA/aDTA animals (aDTA/aDTA) constrict their pupils only to a maximum of 42% after a light pulse that causes 95% constriction in wild types (161 &mgr;W cm-2, 470 nm monochromatic light; High). d, Quantification of PLR data of wild-type (WT; n = 11; black bar) and Opn4aDTA/aDTA (n = 12; red bar) mice from c. Statistical comparisons were made by Student’s t-test (three asterisks, P < 0.001); error bars indicate s.e.m.

Opn4aDTA/aDTA mice do not photoentrain or mask.

a, Opn4aDTA/aDTA mice free-run under light/dark cycles (grey and white backgrounds; dark and light (about 700 lx), respectively). Opn4aDTA/aDTA mice do not phase shift in response to a 15-min 1,500-lx pulse of white light (CT16; yellow dots indicate light pulses). b, Opn4aDTA/aDTA mice do not photoentrain to the 24-h light/dark cycle in the delay or advance phases (red dots indicate cage changes). c, Unlike wild-type animals, no Opn4aDTA/aDTA mice lengthened their period under constant light. d, Opn4aDTA/aDTA mice do not mask under a 7-h ultradian cycle.

We thank J. Mackes and G. Harrison for help in genotyping the animals; R. Kuruvilla, M. Van Doren, B. Wendland, M. Halpern, M. Caterina, C.-Y. Su, J. Bradley and laboratory members in the Biology Department at the Johns Hopkins University for scientific discussions and comments on the manuscript. This work was supported by grants from the National Institutes of Health (to S. Hattar and K.-W.Y.), the Biotechnology and Biological Sciences Research Council (to R.J.L.) and the David and Lucile Packard and Alfred P. Sloan Foundations (to S. Hattar).

Author Contributions A.D.G. and S. Hattar wrote the paper. J.L.E., R.J.L., D.M.B. and T.C.B. gave helpful comments on the manuscript. A.D.G., J.L.E. and C.M.A. in S. Hattar’s laboratory performed all the behavioural studies on the aDTA homozygous animals, as well as the X-gal staining of the Opn4aDTA/tau-LacZ and Opn4tau-LacZ/+ animals, the morphology of the retina, the cholera toxin injections, the water maze and the optomotor studies. D.M.B. helped in analysing the brains of the Opn4aDTA/tau-lacZ and the cholera-toxin-injected animals. G.S.L. and A.R.B. in R.J.L.’s laboratory conducted all the behavioural studies on the aDTA heterozygous animals, and the electroretinogram studies. T.C.B. provided the construct and suggestions for the aDTA targeting strategy. H.C. made the optokinetic nystagmus recordings. H.-W.L. in K.-W.Y.’s laboratory performed the melanopsin immunostaining on aDTA heterozygous mice. Animals were first conceived in K.-W.Y.’s laboratory and produced by S. Hattar and S. Haq to the chimeric stage. Germline transmission was obtained independently in the laboratories of S. Hattar (with help from H.Z.) and K.-W.Y. All other authors helped in the planning, technical support and discussions of experiments.

Hubel, D. H. Eye, Brain, and Vision (Freeman, New York, 1988) Berson, D. M., Dunn, F. A. & Takao, M. Phototransduction by retinal ganglion cells that set the circadian clock. Science 295, 10701073 (2002) Hattar, S., Liao, H. W., Takao, M., Berson, D. M. & Yau, K. W. Melanopsin-containing retinal ganglion cells: architecture, projections, and intrinsic photosensitivity. Science 295, 10651070 (2002) Freedman, M. S. et al. Regulation of mammalian circadian behavior by non-rod, non-cone, ocular photoreceptors. Science 284, 502504 (1999) Czeisler, C. A. et al. Suppression of melatonin secretion in some blind patients by exposure to bright light. N. Engl. J. Med. 332, 611 (1995) Hattar, S. et al. Melanopsin and rod–cone photoreceptive systems account for all major accessory visual functions in mice. Nature 424, 7581 (2003) Panda, S. et al. Melanopsin is required for non-image-forming photic responses in blind mice. Science 301, 525527 (2003) Lucas, R. J. et al. Diminished pupillary light reflex at high irradiances in melanopsin-knockout mice. Science 299, 245247 (2003) Mrosovsky, N. & Hattar, S. Impaired masking responses to light in melanopsin-knockout mice. Chronobiol. Int. 20, 989999 (2003) Panda, S. et al. Melanopsin (Opn4) requirement for normal light-induced circadian phase shifting. Science 298, 22132216 (2002) Ruby, N. F. et al. Role of melanopsin in circadian responses to light. Science 298, 22112213 (2002) Lucas, R. J., Douglas, R. H. & Foster, R. G. Characterization of an ocular photopigment capable of driving pupillary constriction in mice. Nature Neurosci. 4, 621626 (2001) Dacey, D. M. et al. Melanopsin-expressing ganglion cells in primate retina signal colour and irradiance and project to the LGN. Nature 433, 749754 (2005) Wong, K. Y., Dunn, F. A., Graham, D. M. & Berson, D. M. Synaptic influences on rat ganglion-cell photoreceptors. J. Physiol. (Lond.) 582, 279296 (2007) Hattar, S. et al. Central projections of melanopsin-expressing retinal ganglion cells in the mouse. J. Comp. Neurol. 497, 326349 (2006) Gooley, J. J., Lu, J., Fischer, D. & Saper, C. B. A broad role for melanopsin in nonvisual photoreception. J. Neurosci. 23, 70937106 (2003) Hannibal, J. & Fahrenkrug, J. Target areas innervated by PACAP-immunoreactive retinal ganglion cells. Cell Tissue Res. 316, 99113 (2004) Morin, L. P., Blanchard, J. H. & Provencio, I. Retinal ganglion cell projections to the hamster suprachiasmatic nucleus, intergeniculate leaflet, and visual midbrain: bifurcation and melanopsin immunoreactivity. J. Comp. Neurol. 465, 401416 (2003) Sollars, P. J. et al. Melanopsin and non-melanopsin expressing retinal ganglion cells innervate the hypothalamic suprachiasmatic nucleus. Vis. Neurosci. 20, 601610 (2003) Maxwell, F., Maxwell, I. H. & Glode, L. M. Cloning, sequence determination, and expression in transfected cells of the coding sequence for the tox 176 attenuated diphtheria toxin A chain. Mol. Cell. Biol. 7, 15761579 (1987) Wee, R., Castrucci, A. M., Provencio, I., Gan, L. & Van Gelder, R. N. Loss of photic entrainment and altered free-running circadian rhythms in math5-/- mice. J. Neurosci. 22, 1042710433 (2002) Gooley, J. J., Lu, J., Chou, T. C., Scammell, T. E. & Saper, C. B. Melanopsin in cells of origin of the retinohypothalamic tract. Nature Neurosci. 4, 1165 (2001) Barnard, A. R., Hattar, S., Hankins, M. W. & Lucas, R. J. Melanopsin regulates visual processing in the mouse retina. Curr. Biol. 16, 389395 (2006) Iwakabe, H., Katsuura, G., Ishibashi, C. & Nakanishi, S. Impairment of pupillary responses and optokinetic nystagmus in the mGluR6-deficient mouse. Neuropharmacology 36, 135143 (1997) Prusky, G. T., Alam, N. M., Beekman, S. & Douglas, R. M. Rapid quantification of adult and developing mouse spatial vision using a virtual optomotor system. Invest. Ophthalmol. Vis. Sci. 45, 46114616 (2004) Vorhees, C. V. & Williams, M. T. Morris water maze: procedures for assessing spatial and related forms of learning and memory. Nature Protocols 1, 848858 (2006)

doi: 10.1038/nature06829


The aryl hydrocarbon receptor links TH17-cell-mediated autoimmunity to environmental toxins p.106

version="1.0" encoding="utf-8"?> ]>

Nature 453 7191 20080501 1061090 0028-0836 1476-4687 2008Nature Publishing Group Supplementary Information

The file contains Supplementary Methods and Supplementary Figures 1-4 with Legends.

The aryl hydrocarbon receptor links TH17-cell-mediated autoimmunity to environmental toxins MarcVeldhoenM KeijiHirotaK Astrid M.WestendorfA M JanBuerJ LaureDumoutierL Jean-ChristopheRenauldJ BrigittaStockingerB Division of Molecular Immunology, MRC National Institute for Medical Research, The Ridgeway, Mill Hill, London NW71AA, UK Institute for Medical Microbiology, University Hospital Essen, D-45122, Germany Helmholtz Center for Infection Research, D-38124 Braunschweig, Germany Ludwig Institute for Cancer Research, Brussels branch, and Experimental Medicine Unit, Universite Catholique de Louvain, B-1200 Brussels, Belgium Correspondence and requests for materials should be addressed to B.S. (bstocki@nimr.mrc.ac.uk). &e080501-10; &nature06881-s1;

The aryl hydrocarbon receptor (AHR) is a ligand-dependent transcription factor best known for mediating the toxicity of dioxin. Environmental factors are believed to contribute to the increased prevalence of autoimmune diseases, many of which are due to the activity of TH17 T cells, a new helper T-cell subset characterized by the production of the cytokine IL-17. Here we show that in the CD4+ T-cell lineage of mice AHR expression is restricted to the TH17 cell subset and its ligation results in the production of the TH17 cytokine interleukin (IL)-22. AHR is also expressed in human TH17 cells. Activation of AHR by a high-affinity ligand during TH17 cell development markedly increases the proportion of TH17 T cells and their production of cytokines. CD4+ T cells from AHR-deficient mice can develop TH17 cell responses, but when confronted with AHR ligand fail to produce IL-22 and do not show enhanced TH17 cell development. AHR activation during induction of experimental autoimmune encephalomyelitis causes accelerated onset and increased pathology in wild-type mice, but not AHR-deficient mice. AHR ligands may therefore represent co-factors in the development of autoimmune diseases.

The AHR is a ligand-dependent transcription factor that mediates a range of critical cellular events in response to halogenated aromatic hydrocarbons and non-halogenated polycyclic aromatic hydrocarbons such as 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD). AHR expression is ubiquitous in vertebrate cells, suggesting important and widespread roles, but the physiological role of AHR is not yet understood. Mice with a targeted mutation of the Ahr gene provided unequivocal evidence that the AHR is crucial to TCDD-induced toxicity and suggested a function for AHR in liver development.

Gene array analysis of CD4+ effector T-cell subsets showed that the TH17 CD4+ T-cell subset, in addition to the lineage-defining transcription factor Rorc (encoding ROR-&ggr;t, an isoform of ROR-&ggr; expressed in the thymus), expresses Ahr. Quantitative polymerase chain reaction (PCR) analysis of CD4+ effector T-cell subsets from wild-type and AHR-deficient C57BL/6 (B6) mice established that the lineage-defining transcription factors T-bet, Gata3 and Rorc are expressed in a comparable manner as are the marker cytokines interferon (IFN)-&ggr;, IL-4, IL-17A and IL-17F (Fig. 1a). There is also similar Foxp3 expression in CD4+ T cells from B6 and AHR-deficient mice (Supplementary Fig. 1b). AHR was only induced under TH17-cell-inducing conditions (IL-6 and transforming growth factor (TGF)-&bgr;) with levels of expression similar to liver (Fig. 1a, d). Under our conditions we found no AHR expression in cultures with TGF-&bgr; alone (induced T-regulatory cells; iTreg) or IL-6 alone and no expression in natural Treg cells (Supplementary Fig. 1a). Notably, AHR expression is also found in human TH17 cells (Fig. 1b).

Although CD4+ T cells from AHR-deficient mice could differentiate into TH17 cells, they lacked the expression of IL-22 (Fig. 1c). IL-22, originally defined as a hepatocyte-stimulating factor, is co-expressed with IL-17 by TH17 T cells and its expression is thought to be enhanced by dendritic-cell-derived IL-23 (refs 8, 9). The biological functions of IL-22 are not fully understood: on the one hand IL-22 seems to be pro-inflammatory, inducing dermal inflammation and pro-inflammatory gene expression in the skin, on the other hand IL-22 delivery ameliorates T-cell-mediated liver injury in T-cell-mediated hepatitis.

AHR resides in the cytoplasm in complex with Hsp90 until binding of ligand triggers conformational changes resulting in an exchange of Hsp90 for the nuclear translocation component ARNT (reviewed in ref. 1). Arnt was found expressed in all CD4+ T-cell subsets (data not shown). Ligation of AHR by 6-formylindolo[3,2-b]carbazole (FICZ), a tryptophan-derived photoproduct that is thought to be an endogenous ligand with high affinity for the AHR receptor, upregulates genes encoding xenobiotic metabolizing cytochrome P450 enzymes such as Cyp1a1 (ref. 13).

To test whether exposure of T cells to FICZ influences differentiation of naive CD4+ T cells to effector cells, we added FICZ during the in vitro differentiation of CD4+ effector T-cell subsets. The addition of FICZ did not induce Ahr expression or its downstream target Cyp1a1 in TH0, TH1, TH2 and iTreg T-cell subsets (Fig. 2a) and did not alter their expression of Ifng, Il4, Il21 nor their lineage-defining transcription factors (Supplementary Fig. 1c). However, the presence of FICZ during TH17-cell-inducing conditions led to strong upregulation of Il17a, Il17f and particularly of Il22 mRNA expression (Fig. 2a and Supplementary Fig. 2a).

Comparison of TH17 differentiation in CD4+ T cells from wild-type B6 and AHR-deficient mice by intracellular staining showed that exposure of B6 CD4+ T cells to FICZ under TH17-cell-inducing conditions strongly enhanced IL-17A and IL-17F production and increased the proportion and staining intensity of cells producing IL-22 (Fig. 2b, top panels and Supplementary Fig. 2b). In contrast, IL-17A and IL-17F production was attenuated in TH17 cells from AHR-deficient mice and no IL-22 was detectable whether FICZ was present or not (Fig. 2b, bottom panels). A similar response was seen with another AHR ligand, &bgr;-naphthoflavone, which has lower affinity for AHR and consequently requires about a 10-fold higher dose compared with FICZ (Supplementary Fig. 2c). Similar to mouse, human TH17 cells reacted to AHR ligation with increased expression of IL17A, IL17F and IL22 as well as induction of CYP1A1 (Fig. 2c).

To test whether Ahr expression on its own is essential and sufficient to drive IL-22 expression, we performed retroviral transduction of sorted naive AHR-deficient CD4+ T cells with an AHR–green fluorescent protein (GFP) construct or a GFP vector control construct. Transduction under neutral, TH1, TH2 or iTreg conditions did not reconstitute IL-22 expression even in the presence of FICZ (Supplementary Fig. 3). However, under TH17-cell-inducing conditions, reconstitution of AHR expression by retroviral transduction induced expression of IL-22 (Fig. 3a, top-right panel) and increased the proportion of IL-17-producing cells (Fig. 3b, top-right panel). Exposure to FICZ resulted in a substantial increase in IL-22 (Fig. 3a, bottom-right panel) as well as the enhanced expression of IL-17A (Fig. 3b, bottom right panel). PCR with reverse transcription (RT–PCR) analysis of cultured Ahr-transduced TH17 cells confirmed the increase in Il17 expression, the enhancement of Il22 as well as the induction of the Ahr target Cyp1a1 on exposure to FICZ (Fig. 3c). Thus, Ahr expression is only functional in CD4+ T cells that have differentiated to the TH17 lineage.

Next we immunized B6 mice with myelin oligodendrocyte peptide 35–55 (MOG35–55) in complete Freund’s adjuvant (CFA), removed draining lymph nodes 7 days later and isolated CCR6+CD4+ T cells (which are enriched in TH17 cells) to test whether Ahr is also expressed in TH17 cells generated in vivo. CCR6+CD4+ T cells expressed Ahr, Il17a, Il17f, Il22 and Rorc, whereas the CCR6-CD4+ T-cell fraction lacked expression of Ahr and TH17 markers, confirming that physiological differentiation in vivo recapitulates in vitro differentiation (Fig. 4a).

TH17 T cells have a prominent role in the pathology of autoimmune diseases such as experimental autoimmune encephalomyelitis (EAE), which is induced by immunization with MOG peptide 35–55 and CFA. Analysis of spinal cord at the height of the EAE response on day 18 showed increased numbers of CD4+ T cells producing IL-17 and IL-22 in B6 mice treated with MOG35–55/CFA and FICZ, whereas AHR-deficient mice showed reduced numbers of IL-17-producing T cells and minimal numbers of IL-22 producers (Fig. 4b). There was no difference in the numbers of Foxp3-expressing regulatory T cells in spinal cord in the three groups of mice.

Whereas B6 mice developed EAE with a mean day of onset of 13.9, AHR-deficient mice developed EAE with delayed kinetics (mean day of onset 15.6) in line with the attenuated TH17 differentiation seen in vitro. Despite the delayed onset of EAE, most AHR-deficient mice succumbed to disease eventually (Fig. 4c, e). Stimulation of AHR by inclusion of FICZ in the antigen emulsion accelerated the onset (day 11.7) and increased the severity of EAE in B6 mice, but, as expected, did not influence the onset or severity of EAE in AHR-deficient mice (day 15.8) (Fig. 4c, e). To assess the influence of AHR deficiency in haematopoietic or non-haematopoietic cells, we also induced EAE in chimaeras constructed either by injection of AHR-deficient bone marrow into irradiated wild-type mice (AHR-B6) or by injection of B6 wild-type bone marrow into irradiated AHR-deficient mice (B6-AHR). AHR-B6 chimaeras showed attenuated EAE like AHR-deficient mice, whereas B6-AHR chimaeras developed EAE with kinetics and severity similar to wild-type mice (Fig. 4d).

Studies on the effect of AHR stimulation on the immune system have so far focused exclusively on TCDD as a ligand because of its toxicological relevance. Although adverse effects of TCDD on immune responses are well documented, no direct measurements of AHR expression on highly purified polarized subsets of CD4+ T cells have been reported. TCDD, which cannot be metabolized and therefore causes prolonged stimulation of AHR in many cells of the body, is known to induce profound suppression of immune responses in wild-type, but not AHR-deficient, mice, but despite decades of research the underlying mechanisms for this profound toxicity remain unclear. More recently, it has been suggested that TCDD promotes the generation of regulatory T cells while causing the premature decline of activated CD4+ T cells. In our hands, stimulation of AHR by FICZ did not influence the number of regulatory T cells during MOG35–55-induced EAE responses (see Fig. 4b). It is conceivable that the high toxicity of TCDD for effector cells causes a proportional shift in regulatory T cells—which are, in general, more resistant to many depletion regimes—due to the death of other cells, rather than an actual expansion in numbers. Although TCDD may be the classical AHR ligand used in the toxicology field to analyse the effects of AHR activation, it is clear that it is not a natural ligand and it is widely recognized that pollutants such as TCDD are unlikely to have provided the evolutionary pressure for the function of this highly conserved system. There is ongoing debate in the toxicology field about what is the most relevant physiological ligand for AHR. Nevertheless, compelling indirect evidence shows that ultraviolet photoproducts of tryptophan, such as the high-affinity ligand FICZ studied in our paper, could be synthesized in vivo as exposure of human skin to ultraviolet light induces CYP1A1 (ref. 21). Future studies focusing on additional physiological ligands of AHR that can be metabolized by the CYP1 enzymes and therefore cause only transient AHR activation may give further insights into the consequences of AHR stimulation in TH17 cells.

Our data show that AHR, in addition to promoting the expression of IL-22, enhances TH17 cell development and the expression levels of IL-17A and IL-17F, and consequently increases autoimmune pathology. Blockade of IL-17A with an auto-vaccine as well as neutralizing antibody can completely prevent the development of EAE, emphasizing the central role of IL-17A in the pathogenesis of this autoimmune disease. Although it has been suggested that IL-22 may contribute to autoimmune pathology because, like IL-17, it disrupts blood–brain barrier tight junctions, IL-22-deficient mice do not seem to have altered susceptibility to EAE induction, suggesting that this cytokine may be dispensable for the development of pathology in EAE. Thus, the enhancement of IL-17 production by AHR ligation may be more crucial than the induction of IL-22 in determining disease severity.

It is currently thought that IL-22 expression depends on induction by IL-23 (refs 8, 9). Re-stimulation of lymph node cells from mice immunized with MOG35–55/CFA or MOG35–55/CFA and FICZ in the presence or absence of IL-23 showed that IL-23 increased the proportion of IL-17-producing cells both from wild-type and AHR-deficient mice (Supplementary Fig. 4, left), but only had a small effect on the production of IL-22 (Supplementary Fig. 4b, right), suggesting that IL-23 has some effect independent of AHR ligation, although clearly stimulation of AHR seems to be the dominant pathway for IL-22 production.

How AHR interacts with the TH17 pathway is currently unknown, but there is a substantial amount of literature (reviewed in ref. 26) describing interactions of AHR with other key regulatory proteins including nuclear factor-&kgr;B, which has a role in the induction of EAE. The core nucleotide sequence to which the nuclear AHR complex binds, also termed the ‘xenobiotic responsive element’, occurs frequently in the mammalian genome and is also represented in IL-17A, IL-17F, IL-22 and ROR-&ggr;t. Although basal expression of AHR and IL-22 is detectable in TH17 T cells in the apparent absence of a ligand, there are numerous endogenous agents that can activate AHR, such as prostaglandins, bilirubin at high concentration, modified low-density lipoprotein and various modifications of tryptophan, whose ultraviolet-light-irradiated photoconversion into the high-affinity ligand FICZ is only one example (reviewed in ref. 20).

Autoimmune diseases are multifactorial, depending on intrinsic components such as genetics, hormones or age, and environmental factors, including infections, diet, drugs and chemicals. The increasing prevalence of certain autoimmune diseases in highly industrialized countries is probably connected to such environmental factors.

Our data linking a transcription factor responsive to environmental pollutants to the TH17 programme open intriguing possibilities regarding the potential of such factors to initiate or augment autoimmune conditions, and warrant closer examination of a possible role of AHR in human autoimmune diseases.

Methods Summary Mice

C57BL/6 (B6) and AHR-deficient mice on a B6 background (B6 BRA AHRKO), originally obtained from the Jackson Laboratory via A. Smith, were bred in the specified pathogen free (SPF) facility at NIMR. All animal experiments were done according to institutional guidelines and Home Office regulations.

Human T-cell culture

Human peripheral blood mononuclear cells from a healthy volunteer were isolated by Ficoll/Paque, and CD4+ T cells were purified by magnetic sorting and cultured at 1.5 × 105 cells per well in plates coated with anti-CD3 (1 &mgr;g ml-1) and anti-CD28 (1 &mgr;g ml-1) in the presence of 10 ng ml-1 IL-1 and 40 ng ml-1 IL-6 (TH17 condition) or 3 ng ml-1 IL-12, 10 ng ml-1 IL-2 and 10 &mgr;g ml-1 anti-TGF-&bgr; (TH1 condition). Quantitative PCR analysis for TH17 markers and AHR expression was performed on day 4.

In vitro T-cell differentiation and cytokine staining

A detailed description of procedures is given in Methods.

Real-time PCR

The expression of mRNA for transcription factors and cytokines in CD4+ T-cell subsets was analysed 4–5 days after T-cell activation using specific primers from Applied Biosystems and expression was normalized to the housekeeping gene Hprt. More details and a list of primers used can be found in Methods.

Retroviral transduction

AHR was cloned into vector pIRES2-EGFP (Clontech) generating a bicistronic mRNA encoding AHR and, separated by an IRES element, EGFP. Viruses were generated by simultaneous CaCl2-mediated transient transfections of 293T cells with three plasmids providing vector, Gag-Pol, and Env functions. Details of the transduction protocol are given in Methods.

EAE induction

EAE was induced and scored as described previously. Some mice received 600 ng FICZ in the antigen emulsion. Draining lymph nodes were isolated 7 days after immunization. Spinal cord was isolated on day 18 after EAE induction for determination of cell numbers.

Real-time PCR

RNA was extracted using Trizol (Invitrogen) and 1-bromo-3-chloro-propane (Sigma) and reverse transcribed with oligo d(T)16 (Applied Biosystems) according to the manufacturer’s protocol. The cDNA served as template for the amplification of genes of interest and the housekeeping gene (Hprt) by real-time PCR, using TaqMan Gene Expression Assays and Applied Biosystems 7900HT Fast Real-Time PCR System.

Primers from Applied Biosystems

Mouse: Hprt, Mm00446968_m1; Tbx21, Mm00450960_m1; Gata3, Mm00484683_m1; Foxp3, Mm00475156_m1; Rorc, Mm01261019_g1; Ahr, Mm00478930_ml; Arnt, Mm00507836_ml; Ahr exon1 forward, 5′-CGCCTCCGGGACGCAGGTGG-3′; Ahr exon2 reverse, 5′-AAAGAAGCTCTTGGCCCTCAG-3′; Ifng, Mm00801778_m1; Il4, Mm00445259_m1; Il17a, Mm00439619_m1; Il17f, Mm00521423_m1; Il22, Mm00444241_m1; Cyp1a1, Mm00487217_m1; Il23r, Mm00519942_m1. Human: HPRT, HS99999909_m1; IL17A, HS00174383_m1; IL17F, HS00369400_m1; IL22, HS00220924_m1; AHR, HS00169233_m1; CYP1A1, HS00153120_ml.

In vitro T-cell differentiation and cytokine staining

Naive CD4+ T cells were isolated by fluorescence-activated cell sorting (FACS) using a MoFlo sorter of lymph node cell suspensions for CD44loCD25-CD4+ cells. Natural Treg cells were sorted from lymph node suspensions as CD4+CD25+ cells.

Conditions for different cell subsets were: TH0 (anti-IFN-&ggr; plus anti-IL-4 plus anti-TGF-&bgr;), TH1 (4 ng ml-1 IL-12), TH2 (10 ng ml-1 IL-4 and anti-IFN-&ggr;), iTreg (10 ng ml-1 TGF-&bgr; and anti-IFN-&ggr;) or TH17 (20 ng ml-1 IL-6 plus 1 ng ml-1 TGF-&bgr; plus 10 ng ml-1 IL-1&bgr; and anti-IFN-&ggr;). Neutralizing antibodies were used at a concentration of 10 &mgr;g ml-1: clone R46A2 anti-IFN-&ggr;, clone 11B11 anti-IL-4 and clone 1D11 anti-TGF-&bgr;.

The AHR ligands FICZ (BioMol) or &bgr;-naphthoflavone (Sigma) were added in some experiments at the start of culture. For measurements of intracellular cytokines, T cells were re-stimulated with 500 ng ml-1 phorbol dibutyrate and 500 ng ml-1 ionomycin in the presence of brefeldin A for 4 h on day 5 after initiation of cultures. IL-17A antibody was obtained from eBioSciences; IL-17F antibody was from R&D. The anti-IL-22 antibody (MH22B2) was generated in IL-22 knockout BALB/c mice immunized with recombinant human IL-22 crosslinked to ovalbumin in the presence of glutaraldehyde and cross-reacts with human and murine IL-22.

Retroviral transduction

FACS-sorted naive CD4+ T cells were plated in antibody-coated wells as described above in the absence of cytokines on day 0. On day 1, fresh retrovirus supernatant was added together with cytokines for polarization and the cells were spun for 1 h at 1,200 r.p.m. Thereafter, the cells were cultured in the presence or absence of 200 nM AHR ligand FICZ for another 4 days and then assayed by RT–PCR and intracellular staining.

Generation of bone marrow chimaeras

5 × 106 bone marrow cells from B6 donors or AHR-deficient donors were injected into lethally (9.5 Gy) irradiated AHR-deficient hosts (B6-AHR chimaeras) or B6 hosts (AHR- B6 chimaeras). Bone marrow donors were differentiated from the hosts by an allotypic marker (CD45.1). Chimaeras were tested for reconstitution 8 weeks after bone marrow transfer.

AHR is selectively expressed in the T-H17 cell subset.

a, Fluorescence-activated cell sorter (FACS)-sorted naive CD4+ T cells from B6 or AHR-deficient mice stimulated under TH0, TH1, TH2, iTreg or TH17 conditions and harvested on day 5 for quantitative PCR. mRNA levels, normalized to Hprt expression ± s.d., are shown. Rorc, mRNA encoding ROR-&ggr;t. b, mRNA levels ± s.d. of AHR in human CD4+ T cells stimulated under TH1 or TH17 conditions. c, mRNA levels ± s.d. for Il22 in CD4+ T cells subsets from B6 or AHR-deficient mice. d, mRNA levels of Ahr in CD4+ T-cell subsets and TH17 cells from three mice compared with mouse liver.

AHR ligation promotes the TH17 cell programme.

a, Mean mRNA levels ± s.d. from B6 CD4+ effector subsets generated in the presence or absence of 200 nM FICZ. b, CD4+ T cells from B6 (top panels) and AHR-deficient mice (bottom panels) cultured under TH17-cell-inducing conditions in the presence or absence of 200 nM FICZ and stained for expression of IL-17A versus IL-17F (left panels) and IL-17A versus IL-22 (right panels) after re-stimulation with phorbol dibutyrate and ionomycin. A summary of all experiments is shown in Supplementary Fig. 2b. c, mRNA levels of IL17A, IL17F, IL22 and CYP1A1 in human CD4+ T cells stimulated under TH17-cell-inducing conditions in the presence or absence of 3 &mgr;M &bgr;-naphthoflavone (&bgr;-NF).

Retroviral transduction of AHR restores IL-22 expression.

a, b, FACS-sorted naive CD4+ T cells from AHR-deficient mice cultured under TH17-cell-inducing conditions and transduced with vector control (RV-GFP) or AHR-containing construct (RV-AHR-GFP) in the presence (bottom panels) or absence (top panels) of FICZ. IL-22 intracellular staining versus GFP expression (a) and IL-17A expression in gated GFP+ cells (b) were assessed. c, Quantitative PCR for Il17a, Il22 and Cyp1a1 in TH17 cells from AHR-deficient mice transduced with control retroviral vector, control vector in the presence of 200 nM FICZ, AHR-containing vector, or AHR-containing vector in the presence of FICZ.

EAE is enhanced by AHR ligation.

a, RT–PCR analysis for Il17a, Il17f, Il22, Rorc (ROR-&ggr;t) and Ahr in FACS-sorted CCR6-CD4+ T cells or CCR6+CD4+ T cells from draining lymph nodes 7 days after MOG35–55 immunization. b, Mean numbers of IL-17A+, IL-22+ and Foxp3+ cells ± s.d. in spinal cord 18 days after immunization of B6 mice (n = 4) with MOG35–55 (Ctrl) or MOG35–55 + FICZ (FICZ) and of AHR-deficient mice with MOG35–55 (AHR) (n = 4). c, Mean clinical EAE scores ± s.e.m. of B6 (n = 12) or AHR-deficient mice (n = 12) immunized with MOG35–55 in the absence or presence of 600 ng FICZ. d, Mean clinical EAE scores ± s.e.m. of MOG35–55-immunized chimaeras from AHR-deficient donors into irradiated B6 hosts (AHR-B6 n = 8), or bone marrow from B6 donors into irradiated AHR-deficient hosts (B6-AHR n = 8). e, Incidence, mean day of onset and mean maximal scores for the mice in c. P values are P = 0.0013 for mean day of onset in B6 versus B6 + FICZ and P = 0.03 for B6 versus AHR-deficient.

We thank A. Smith for providing us with the AHR-deficient mouse strain; A. Rae and G. Preece for cell sorting; and Biological Services at NIMR, especially T. Norton and H. Boyes, for animal care and EAE scoring. This work was funded by the Medical Research Council UK.

Author Contributions M.V. and K.H. performed the experiments; A.M.W. and J.B. did the microarrays and analysis; L.D. and J.-C.R. generated and provided the anti-IL-22 antibody; and B.S. directed the research and wrote the manuscript.

Schmidt, J. V. & Bradfield, C. A. Ah receptor signaling pathways. Annu. Rev. Cell Dev. Biol. 12, 5589 (1996) Harper, P. A., Riddick, D. S. & Okey, A. B. Regulating the regulator: factors that control levels and activity of the aryl hydrocarbon receptor. Biochem. Pharmacol. 72, 267279 (2006) Schmidt, J. V., Su, G. H., Reddy, J. K., Simon, M. C. & Bradfield, C. A. Characterization of a murine Ahr null allele: involvement of the Ah receptor in hepatic growth and development. Proc. Natl Acad. Sci. USA 93, 67316736 (1996) Fernandez-Salguero, P. et al. Immune system impairment and hepatic fibrosis in mice lacking the dioxin-binding Ah receptor. Science 268, 722726 (1995) Mimura, J. et al. Loss of teratogenic response to 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) in mice lacking the Ah (dioxin) receptor. Genes Cells 2, 645654 (1997) Ivanov, I. I. et al. The orphan nuclear receptor ROR&ggr;t directs the differentiation program of proinflammatory IL-17+ T helper cells. Cell 126, 11211133 (2006) Dumoutier, L., Van Roost, E., Colau, D. & Renauld, J. C. Human interleukin-10-related T cell-derived inducible factor: molecular cloning and functional characterization as an hepatocyte-stimulating factor. Proc. Natl Acad. Sci. USA 97, 1014410149 (2000) Zheng, Y. et al. Interleukin-22, a TH17 cytokine, mediates IL-23-induced dermal inflammation and acanthosis. Nature 445, 648651 (2007) Liang, S. C. et al. Interleukin (IL)-22 and IL-17 are coexpressed by Th17 cells and cooperatively enhance expression of antimicrobial peptides. J. Exp. Med. 203, 22712279 (2006) Boniface, K. et al. IL-22 inhibits epidermal differentiation and induces proinflammatory gene expression and migration of human keratinocytes. J. Immunol. 174, 36953702 (2005) Radaeva, S., Sun, R., Pan, H. N., Hong, F. & Gao, B. Interleukin 22 (IL-22) plays a protective role in T cell-mediated murine hepatitis: IL-22 is a survival factor for hepatocytes via STAT3 activation. Hepatology 39, 13321342 (2004) Zenewicz, L. A. et al. Interleukin-22 but not interleukin-17 provides protection to hepatocytes during acute liver inflammation. Immunity 27, 647659 (2007) Rannug, A. & Fritsche, E. The aryl hydrocarbon receptor and light. Biol. Chem. 387, 11491157 (2006) Hirota, K. et al. Preferential recruitment of CCR6-expressing Th17 cells to inflamed joints via CCL20 in rheumatoid arthritis and its animal model. J. Exp. Med. 204, 28032812 (2007) Acosta-Rodriguez, E. V. et al. Surface phenotype and antigenic specificity of human interleukin 17-producing T helper memory cells. Nature Immunol. 8, 639646 (2007) Kerkvliet, N. I. Recent advances in understanding the mechanisms of TCDD immunotoxicity. Int. Immunopharmacol. 2, 277291 (2002) Funatake, C. J., Marshall, N. B., Steppan, L. B., Mourich, D. V. & Kerkvliet, N. I. Cutting edge: activation of the aryl hydrocarbon receptor by 2,3,7,8-tetrachlorodibenzo-p-dioxin generates a population of CD4+CD25+ cells with characteristics of regulatory T cells. J. Immunol. 175, 41844188 (2005) Minamimura, K., Gao, W. & Maki, T. CD4+ regulatory T cells are spared from deletion by antilymphocyte serum, a polyclonal anti-T cell antibody. J. Immunol. 176, 41254132 (2006) Cox, A. L. et al. Lymphocyte homeostasis following therapeutic lymphocyte depletion in multiple sclerosis. Eur. J. Immunol. 35, 33323342 (2005) Nguyen, L. P. & Bradfield, C. A. The search for endogenous activators of the aryl hydrocarbon receptor. Chem. Res. Toxicol. 21, 102116 (2007) Katiyar, S. K., Matsui, M. S. & Mukhtar, H. Ultraviolet-B exposure of human skin induces cytochromes P450 1A1 and 1B1. J. Invest. Dermatol. 114, 328333 (2000) Uyttenhove, C. & Van Snick, J. Development of an anti-IL-17A auto-vaccine that prevents experimental auto-immune encephalomyelitis. Eur. J. Immunol. 36, 28682874 (2006) Langrish, C. L. et al. IL-23 drives a pathogenic T cell population that induces autoimmune inflammation. J. Exp. Med. 201, 233240 (2005) Kebir, H. et al. Human TH17 lymphocytes promote blood-brain barrier disruption and central nervous system inflammation. Nature Med. 13, 11731175 (2007) Kreymborg, K. et al. IL-22 is expressed by Th17 cells in an IL-23-dependent fashion, but not required for the development of autoimmune encephalomyelitis. J. Immunol. 179, 80988104 (2007) Okey, A. B. An aryl hydrocarbon receptor odyssey to the shores of toxicology: the Deichmann lecture, International Congress of Toxicology-XI. Toxicol. Sci. 98, 538 (2007) Tian, Y., Rabson, A. B. & Gallo, M. A. Ah receptor and NF-&kgr;B interactions: mechanisms and physiological implications. Chem. Biol. Interact. 141, 97115 (2002) Hilliard, B., Samoilova, E. B., Liu, T. S., Rostami, A. & Chen, Y. Experimental autoimmune encephalomyelitis in NF-&kgr;B-deficient mice: roles of NF-&kgr;B in the activation and differentiation of autoreactive T cells. J. Immunol. 163, 29372943 (1999) Denison, M. S., Fisher, J. M. & Whitlock, J. P. Inducible, receptor-dependent protein-DNA interactions at a dioxin-responsive transcriptional enhancer. Proc. Natl Acad. Sci. USA 85, 25282532 (1988) Veldhoen, M., Hocking, R. J., Flavell, R. A. & Stockinger, B. Signals mediated by transforming growth factor-&bgr; initiate autoimmune encephalomyelitis, but chronic inflammation is needed to sustain disease. Nature Immunol. 7, 11511156 (2006)

doi: 10.1038/nature06881


BCR-ABL1 lymphoblastic leukaemia is characterized by the deletion of Ikaros p.110

version="1.0" encoding="utf-8"?> ]>

Nature 453 7191 20080501 1101140 0028-0836 1476-4687 2008Nature Publishing Group Supplementary information

The file contains Supplementary Results; Supplementary Tables 1-7; Supplementary Figures 1-8 with Legends and additional references.

BCR–ABL1 lymphoblastic leukaemia is characterized by the deletion of Ikaros Charles G.MullighanC G Christopher B.MillerC B InaRadtkeI Letha A.PhillipsL A JamesDaltonJ JingMaJ DeborahWhiteD Timothy P.HughesT P Michelle M.Le BeauM M Ching-HonPuiC Mary V.RellingM V Sheila A.ShurtleffS A James R.DowningJ R Departments of Pathology, Oncology and, Pharmaceutical Sciences and, The Hartwell Center for Bioinformatics and Biotechnology, St Jude Children’s Research Hospital, Memphis, Tennessee 38105, USA Division of Haematology, The Institute for Medical and Veterinary Science, Adelaide, South Australia 5000, Australia Section of Hematology/Oncology, University of Chicago, Chicago, Illinois 60637, USA Correspondence and requests for materials should be addressed to J.R.D. (james.downing@stjude.org). &e080501-15; &nature06866-s1;

The Philadelphia chromosome, a chromosomal abnormality that encodes BCR–ABL1, is the defining lesion of chronic myelogenous leukaemia (CML) and a subset of acute lymphoblastic leukaemia (ALL). To define oncogenic lesions that cooperate with BCR–ABL1 to induce ALL, we performed a genome-wide analysis of diagnostic leukaemia samples from 304 individuals with ALL, including 43 BCR–ABL1 B-progenitor ALLs and 23 CML cases. IKZF1 (encoding the transcription factor Ikaros) was deleted in 83.7% of BCR–ABL1 ALL, but not in chronic-phase CML. Deletion of IKZF1 was also identified as an acquired lesion at the time of transformation of CML to ALL (lymphoid blast crisis). The IKZF1 deletions resulted in haploinsufficiency, expression of a dominant-negative Ikaros isoform, or the complete loss of Ikaros expression. Sequencing of IKZF1 deletion breakpoints suggested that aberrant RAG-mediated recombination is responsible for the deletions. These findings suggest that genetic lesions resulting in the loss of Ikaros function are an important event in the development of BCR–ABL1 ALL.

Acute lymphoblastic leukaemia (ALL) comprises a heterogeneous group of disorders characterized by recurring chromosomal abnormalities including translocations, trisomies and deletions. An ALL subtype with especially poor prognosis is characterized by the presence of the Philadelphia chromosome arising from the t(9;22)(q34;q11.2) translocation, which encodes the constitutively activated BCR–ABL1 tyrosine kinase. BCR–ABL1-positive ALL constitutes 5% of paediatric B-progenitor ALL and approximately 40% of adult ALL. Expression of BCR–ABL1 is also the pathological lesion underlying CML. Data from murine studies demonstrate that expression of BCR–ABL1 in haematopoietic stem cells can alone induce a CML-like myeloproliferative disease, but cooperating oncogenic lesions are required for the generation of a blastic leukaemia. Although the p210 and p190 BCR–ABL1 fusions are most commonly found in CML and paediatric BCR–ABL1 ALL, respectively, either fusion may be found in adult BCR–ABL1 ALL. Notably, a number of genetic lesions including additional cytogenetic aberrations and mutations in tumour suppressor genes have been described in CML cases progressing to blast crisis. However, the specific lesions responsible for the generation of BCR–ABL1 ALL and blastic transformation of CML remain incompletely understood. To identify cooperating oncogenic lesions in ALL, we recently performed a genome-wide analysis of paediatric ALL. This analysis identified an average of 6.8 genomic copy number alterations in 9 BCR–ABL1 ALL cases, including deletions in genes that have a regulatory role in normal B-cell development.

To extend this analysis and identify lesions that distinguish CML from BCR–ABL1 ALL, we have now examined DNA from leukaemic samples from 304 paediatric and adult cases of ALL (254 B-progenitor; 50 T-lineage), including 21 paediatric and 22 adult BCR–ABL1 ALL cases, and 23 adult CML cases (Supplementary Table 1). Samples were analysed using the 250K Sty and Nsp Affymetrix single-nucleotide polymorphism (SNP) arrays (and also the 100K arrays for most cases). This identified a mean of 8.79 somatic copy number alterations per BCR–ABL1 ALL case (range 1–26), with 1.44 gains (range 0–13) and 7.33 losses (range 0–25) (Supplementary Table 4). No significant differences were noted in the frequency of copy number alterations between paediatric and adult BCR–ABL1 ALL cases. The most frequent somatic copy number alteration was deletion of IKZF1, which encodes the transcription factor Ikaros (Table 1). IKZF1 was deleted in 36 (83.7%) of 43 BCR–ABL1 ALL cases, including 76.2% of paediatric and 90.9% of adult BCR–ABL1 ALL cases. CDKN2A was deleted in 53.5% of BCR–ABL1 ALL cases, most of which (87.5%) also had deletions of IKZF1 (Table 1 and Supplementary Table 5). Conversely, of the BCR–ABL1 ALL cases with IKZF1 deletions, 41.6% lacked CDKN2A alterations. Deletion of PAX5 occurred in 51% of BCR–ABL ALL cases, again with the majority also having a deletion of IKZF1 (95%) (Table 1 and Supplementary Table 5). No other defining copy number alterations were identified in the rare BCR–ABL1 ALL cases that lacked a deletion of IKZF1.

Ikaros is a member of a family of zinc-finger nuclear proteins that is required for normal lymphoid development. Ikaros has a central DNA-binding domain consisting of four zinc fingers, and a homo- and heterodimerization domain consisting of the two carboxy-terminal zinc fingers (Fig. 1 and Supplementary Fig. 1). Alternative splicing generates multiple Ikaros isoforms, several of which lack the amino-terminal zinc fingers required for DNA binding; however, the physiological relevance of these isoforms in normal haematopoiesis remains unclear (Supplementary Fig. 1). The IKZF1 deletions identified in BCR–ABL1 ALL were predominantly mono-allelic and were limited to the gene in 25 cases, conclusively identifying IKZF1 as the genetic target (Fig. 1). In 19 cases the deletions were confined to a subset of internal IKZF1 exons, most commonly exons 3–6 (&Dgr;3–6; N = 15). Notably, the &Dgr;3–6 deletion is predicted to encode an Ikaros isoform that lacks the DNA-binding domain but retains the C-terminal zinc fingers. The IKZF1 deletions were confirmed by fluorescence in situ hybridization (FISH) and genomic quantitative polymerase chain reaction (PCR), and were in the predominant leukaemic clone (Supplementary Table 6 and Supplementary Fig. 2). Detailed analysis failed to reveal any evidence of either IKZF1 point mutations or inactivation of its promoter by CpG methylation in primary ALL samples (data not shown and Supplementary Fig. 8).

The expression of aberrant, dominant-negative Ikaros isoforms in B- and T-lineage ALL has been previously reported by several groups, although alternative splicing has been reported to be the underlying mechanism. Importantly, the &Dgr;3–6 isoform of Ikaros has been shown to function as a dominant-negative inhibitor of the transcriptional activity of Ikaros and related family members. Moreover, mice homozygous for either an Ikzf1 null mutation or a dominant-negative Ikzf1 mutation exhibit profound defects in lymphoid development, and mice heterozygous for a dominant-negative Ikzf1 mutation develop clonal T-cell expansions and lymphoproliferative diseases, demonstrating that alteration in the level of Ikzf1 expression is oncogenic.

The high frequency of focal deletions in IKZF1 in BCR–ABL1 ALL suggests that expression of alternative IKZF1 transcripts may be the result of specific genetic lesions, and not alternative splicing of an intact gene. To explore further this possibility, we performed reverse-transcriptase PCR (RT–PCR) analysis for IKZF1 transcripts in 159 cases (Fig. 2). This demonstrated that expression of the Ik6 transcript, which lacks exons 3–6, was exclusively observed in cases harbouring the IKZF1 &Dgr;3–6 deletion (Fig. 2b). Furthermore, we detected two previously unknown Ikaros isoforms exclusively in cases with larger deletions: Ik9 in a case with deletion of exons 2–6, and Ik10 in three cases with deletion of exons 1–6 (Fig. 2a, b, and Supplementary Fig. 3). For each isoform, Ik6, Ik9 and Ik10, there was concordance between the transcripts detected by RT–PCR and the extent of deletion defined by SNP array and genomic PCR analysis (Fig. 2b). Moreover, analysis of 22 IKZF1 &Dgr;3–6 and 29 non-&Dgr;3–6 cases with a quantitative RT–PCR assay specific for the Ik6 transcript confirmed that Ik6 expression was restricted to cases with the &Dgr;3–6 deletion (P = 6.41 × 10-15, Supplementary Fig. 4). Furthermore, the Ik6 protein isoform was only detectable by western blotting in cases with a &Dgr;3–6 IKZF1 deletion (Fig. 2c). We also did not observe expression of Ik6 after the enforced expression of BCR–ABL1 in Arf null or wild-type murine haematopoietic precursors (data not shown). Together, these data indicate that the expression of non-DNA-binding Ikaros isoforms is due to IKZF1 genomic abnormalities, and not aberrant post-transcriptional splicing induced by BCR–ABL1, as has been suggested.

To identify copy number alterations in CML, we performed SNP array analysis on 23 CML cases. In addition to chronic-phase CML (CP-CML), we also examined matched accelerated phase (AP-CML, N = 7) and blast crisis (BC-CML, N = 15 (12 myeloid and 3 lymphoid)) samples (Supplementary Table 2). This identified only 0.47 copy number alterations per CP-CML case (range 0–8) (Supplementary Table 7), suggesting that BCR–ABL1 is sufficient to induce CML, but alone does not result in substantial genomic instability. Notably, no recurrent lesions were identified. In contrast, there was a mean of 7.8 copy number alterations per BC-CML case (range 0–28) (Supplementary Table 7), with IKZF1 deletions in four BC samples, including two of the three cases with lymphoid blast crisis (Fig. 3a). Two of the IKZF1 deletions involved the entire gene (CML-4-BC and CML-22-BC), one &Dgr;3–6 (CML-1-BC, which was associated with Ik6 expression by RT–PCR) and one &Dgr;3–7 (CML-7-BC). CML-7-BC also had an IKZF1 nonsense mutation in the C-terminal zinc-finger domain of exon 7 in the non-deleted allele (coding nucleotide 1520C>A, amino acid Ser507X, Fig. 3b, c). One BC sample had a CDKN2A deletion, and four cases had copy number alterations involving PAX5 (two deletions, one internal amplification and one trisomy 9), with two of these also having IKZF1 deletion. Copy number alterations were identified in two AP-CML samples. These data demonstrate an increased burden of genomic aberrations during progression of CML, with IKZF1 mutation a frequent event in the transformation of CML to lymphoid blast crisis.

To explore the mechanism responsible for the identified IKZF1 deletion, we sequenced the IKZF1 &Dgr;3–6 genomic breakpoints (Supplementary Fig. 6). The deletions were restricted to highly localized sequences in introns 2 and 6 (Supplementary Fig. 7). Moreover, heptamer recombination signal sequences (RSSs) recognized by the RAG enzymes during V(D)J recombination were located immediately internal to the deletion breakpoints, and a variable number of additional nucleotides were present between the consensus intron 2 and 6 sequences, suggestive of the action of terminal deoxynucleotidyl transferase (TdT). Together, these data suggest that the IKZF1 &Dgr;3–6 deletion arises owing to aberrant RAG-mediated recombination.

We have identified a high frequency of copy number alterations in BCR–ABL1 ALL and BC-CML, but not in CP-CML. We observed a near obligate deletion of IKZF1 in BCR–ABL1 ALL, with 83.7% of paediatric and adult cases containing deletions that lead to a reduction in dose and/or the expression of an altered Ikaros isoform. By contrast, deletion of IKZF1 was not detected in CP-CML, but was identified as an acquired lesion in two of three lymphoid BC-CML samples. These data, together with the low frequency of IKZF1 deletions in other paediatric B-progenitor ALL cases, and the lack of focal IKZF1 aberrations in recently reported genomic analysis of non-haematopoietic tumours (Supplementary Results), suggest that alterations in Ikaros directly contribute to the pathogenesis of BCR–ABL1 ALL. How reduced activity of Ikaros, and possibly that of other family members through the expression of dominant negative Ikaros isoforms, collaborates with BCR–ABL1 to induce lymphoblastic leukaemia remains to be determined. Importantly, mice with attenuated Ikaros expression exhibit a partial block of B lymphoid maturation at the pro-B-cell stage, suggesting that Ikaros loss may contribute to the arrested B-lymphoid maturation in BCR–ABL1 ALL. However, the high co-occurrence of PAX5 deletions in many cases suggests that IKZF1 deletion contributes to transformation in additional ways. The frequent co-deletion of CDKN2A (encoding INK4A/ARF) with IKZF1 in BCR–ABL1 ALL is a notable finding. This suggests that attenuated Ikaros activity may either collaborate with disruption of INK4A/ARF-mediated tumour suppression, or act through alternative uncharacterized tumour suppressor pathways in ALL. Furthermore, the identification of aberrant RAG-mediated recombination as the mechanism underlying deletions of IKZF1 suggests that the cellular target of this transforming event is downstream of the haematopoietic stem cell. Dissecting the contribution of altered Ikaros activity to BCR–ABL1 leukaemogenesis should not only provide valuable mechanistic insights, but will also help to determine if the presence of this genetic lesion can be used to gain a therapeutic advantage against this aggressive leukaemia.

Methods Summary

Two-hundred and eighty-two paediatric ALL cases, 22 adult BCR–ABL1 ALL cases, 49 samples obtained from 23 adult patients with chronic myeloid leukaemia (CML) and 36 leukaemia cell lines were studied (Supplementary Tables 1 and 2). Affymetrix 250K Sty and Nsp arrays were performed on all samples. 50K Hind 240 and 50K Xba 240 arrays were performed for 252 ALL samples (Supplementary Table 1). SNP array data were analysed using dChip (http://www.dChip.org), a reference normalization algorithm, and circular binary segmentation as previously described. IKZF1 deletions were confirmed by FISH and/or genomic quantitative PCR. Expression of Ikaros transcripts was examined by qualitative and quantitative RT–PCR, and western blotting. Genomic sequencing was performed for all IKZF1 coding exons. Methylation status of the IKZF1 promoter CpG island was performed by MALDI-TOF mass spectrometry of bisulphite-treated leukaemic blast DNA.

Patients and samples

Patients and samples comprised 282 patients with acute lymphoblastic leukaemia (ALL) treated at St Jude Children’s Research Hospital, 22 adult BCR–ABL1 ALL patients treated at the University of Chicago, and 49 samples obtained from 23 adult patients with chronic myeloid leukaemia (CML) treated at the Institute of Medical and Veterinary Science, Adelaide (Supplementary Tables 1 and 2). The CML cohort included 24 chronic phase, 7 accelerated phase and 15 blast crisis samples, and three samples obtained at complete cytogenetic response. All blast crisis samples were flow sorted to at least 90% blast purity before DNA extraction using FACSVantage s.e. (with DiVa option) flow cytometers (BD Biosciences) and fluorescein-isothiocyanate-labelled CD45, allophycocyanin-labelled CD33 and phycoerythrin-labelled CD19 and CD13 antibodies (BD Biosciences). Germline tissue was obtained by also sorting the non-blast population in seven cases. Informed consent for the use of leukaemic cells for research was obtained from patients, parents or guardians in accordance with the Declaration of Helsinki, and study approval was obtained from the SJCRH institutional review board.

Cell lines examined by SNP array

Thirty-six acute myeloid and lymphoid leukaemia cell lines were genotyped using the Affymetrix Mapping 250k Sty and Nsp arrays. These were the ALL cell lines 380 (MYC–IGH and BCL2–IGH B precursor), 697 (TCF3–PBX1), AT1 (ETV6–RUNX1), BV173 (CML in lymphoid blast crisis), CCRF-CEM (TAL–SIL), Jurkat (T-ALL), Kasumi-2 (TCF3–PBX1), MHH-CALL-2 (hyperdiploid B-precursor ALL), MHH-CALL-3 (TCF3–PBX1), MOLT3 (T-ALL), MOLT4 (T-ALL), NALM-6 (B-precursor ALL), OP1 (BCR–ABL1), Reh (ETV6–RUNX1), RS4;11 (MLL–AF4), SD1 (BCR–ABL1), SUP-B15 (BCR–ABL1), TOM-1 (BCR–ABL1), U-937 (PICALM–AF10), UOCB1 (TCF3–HLF), YT (NK leukaemia), and the AML cell lines CMK (FAB M7), HL-60 (FAB M2), K-562 (CML in myeloid blast crisis), Kasumi-1 (RUNX1–RUNX1T1), KG-1 (myelocytic leukaemia), ME-1 (CBFB–MYH11), ML-2 (MLL–AF6), M-07e (FAB M7), Mono Mac 6 (MLL–AF9), MV4-11 (MLL–AF4), NB4 (PML–RARA), NOMO-1 (MLL–AF9), PL21 (FAB M3), SKNO-1 (RUNX1–RUNX1T1) and THP-1 (FAB M5). Cell lines were obtained from the Deutsche Sammlung von Mikroorganismen und Zellkulturen, Braunschweig, Germany, the American Type Culture Collection, Manassas, Virginia, from local institutional repositories, or were gifts from O. Heidenreich (SKNO-1) and D. Campana (OP1). Cells were cultured in accordance with previously published recommendations. The paediatric BCR–ABL1 B-precursor ALL cell line OP1 (ref. 31) was cultured in RPMI-1640 containing 100 units ml-1 penicillin, 100 &mgr;g ml-1 streptomycin, 2 mM glutamine and 10% fetal bovine serum. DNA was extracted from 5 × 106 cells obtained during log-phase growth after washing in PBS using the QIamp DNA blood mini kit (Qiagen).

SNP microarray analysis

Collection and processing of diagnostic and remission bone marrow and peripheral blood samples for Affymetrix SNP microarray analysis has been previously reported in detail. Affymetrix 250K Sty and Nsp arrays were performed on all samples. 50K Hind 240 and 50K Xba 240 arrays were performed for 252 ALL samples (Supplementary Table 1). SNP array .CEL and SNP call .TXT files (generated by Affymetrix GTYPE 4.0 using the DM algorithm) have been deposited in NCBIs Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih.gov/geo/) and are accessible through GEO series accession numbers GSE9109–GSE9113. These accessions contain the following data: GSE9109, Sty and Nsp files for 304 ALL samples, and Hind and Xba files for 252 of these samples; GSE9110, Sty and Nsp files for 56 CML samples; GSE9111, Sty, Nsp, Hind and Xba files for 50 remission acute leukaemia samples used as references for copy number analysis; GSE9112, Sty and Nsp files for 36 acute leukaemia cell lines; GSE9113, a superseries containing all of the above data. The data are also available at http://www.stjuderesearch.org/data/ALL-SNP2/.


FISH for IKZF1 deletion was performed using diagnostic bone marrow or peripheral blood leukaemic cells in Carnoy’s fixative as previously described. BAC clones CTD-2382L6 and CTC-791O3 (for IKZF1, Open Biosystems) were labelled with fluorescein isothiocyanate, and control 7q31 probes RP11-460K21 (Children’s Hospital Oakland Research Institute) and CTB-133K23 (Open Biosystems) were labelled with rhodamine. At least 100 interphase nuclei were scored per case.

IKZF1 PCR, cloning, quantitative PCR and genomic sequencing

RNA was extracted and reverse transcribed using random hexamer primers and Superscript III (Invitrogen) as previously described. IKZF1 transcripts were amplified from cDNA using the Advantage 2 PCR enzyme (Clontech) as previously described using primers that anneal in exon 0 and 7 of IKZF1. PCR products were purified and sequenced directly and after cloning into pGEM-T-Easy (Promega). Genomic quantitative PCR for exons 1–7 of IKZF1, and real-time PCR to quantify expression of Ik6, were performed as previously described. All primers and probes are listed in Supplementary Table 3. Genomic sequencing of IKZF1 exons 0–7 in all ALL and CML samples was performed as previously described.

Western blotting

Whole-cell lysates of 3–6 × 106 leukaemic cells were prepared and blotted as previously described using N- and C-terminus-specific rabbit polyclonal Ikaros antibodies (Santa Cruz Biotechnology).

Methylation analysis

Methylation status of the IKZF1 promoter CpG island (PCR amplicon hg17 coordinates: chromosome 7 50121508–50121714) was performed using MALDI-TOF mass spectrometry of PCR-amplified, bisulphite-modified genomic DNA extracted from leukaemic cells as previously described.

Statistical analysis

Associations between ALL subtype and IKZF1 deletion frequency were calculated using the exact likelihood ratio test. Differences in Ik6 expression between IKZF1 &Dgr;3–6 and non-&Dgr;3–6 cases were assessed using the exact Wilcoxon–Mann–Whitney test. All P-values reported are two-sided. Analyses were performed using StatXact v8.0.0 (Cytel).

IKZF1 deletions in BCR–ABL1 ALL.

a, Domain structure of IKZF1. Exons 3–5 encode four N-terminal zinc fingers (black boxes) responsible for DNA binding. The C-terminal zinc fingers encoded by exon 7 are essential for homo- and heterodimerization. b, Genomic organization of IKZF1 and location of each of the 36 deletions observed in BCR–ABL1 B-progenitor ALL. Each line depicts the deletion(s) observed in each case. In four cases, two discontiguous deletions were observed. Hemizygous deletions are solid lines and homozygous deletions dashed. Arrows indicate deletions extending beyond the limits of the figure. The exact boundaries of the deletions were defined by genomic quantitative PCR, and for IKZF1 &Dgr;3–6, by long-range genomic PCR (red arrow). c, dChip SNP raw log2 ratio copy number data depicting IKZF1 deletions for 29 BCR–ABL1 cases and 3 B-progenitor ALL cell lines.

Ikaros isoforms in ALL blasts.

a, Domain structure of the IKZF1 isoforms detected by RT–PCR, examples of which are shown in b. b, RT–PCR for IKZF1 transcripts (using exon 0- and 7-specific primers) in representative cases with various IKZF1 genomic abnormalities. Each case expressing an aberrant isoform has a corresponding IKZF1 genomic deletion. IKZF1 &Dgr;3–6 was also detected in the BCR–ABL1 ALL cell lines SUP-B15 and OP1, and &Dgr;1–6 in the ALL cell line 380. c, Western blotting for Ikaros using a C-terminus-specific polyclonal antibody. Ik6 was only detectable in cases with IKZF1 &Dgr;3–6. The &Dgr;1–6 and &Dgr;2–6 deletions do not produce a detectable protein. In three cases with multiple focal hemizygous deletions involving different regions of IKZF1 (BCR-ABL-SNP-26, BCR-ABL-SNP-29 and BCR-ABL-SNP-31), no wild-type Ikaros was detectable by RT–PCR or western blotting, indicating that the deletions involve both copies of IKZF1 in each case.

IKZF1 deletions in blast crisis CML.

a, dChip SNP log2 ratio copy number heatmaps of four CML cases showing acquisition of IKZF1 deletions at progression to blast crisis. b, c, Pherograms of IKZF1 exon 7 sequencing demonstrating acquisition of the coding nucleotide 1520C>A, amino acid Ser507X mutation at chronic-phase (b) and blast crisis (c) in case CML-7. As this case has a concomitant hemizygous IKZF1 deletion involving exon 7, the mutation appears to be homozygous.

Frequency of recurring DNA copy number abnormalities in ALL ALL subtype (N) IKZF1 CDKN2A PAX5 C20orf94 RB1 MEF2C EBF1 BTG1 DLEU FHIT ETV6

The prevalence of recurring genomic abnormalities in BCR–ABL1 B-progenitor ALL identified by SNP array analysis is shown for each ALL subtype. The exact likelihood ratio P-value for variation in the frequency of each lesion across ALL subtypes is shown. The DLEU region at 13q14 incorporates the microRNA genes MIRN16-1 and MIRN15A.

B-progenitor (254) BCR–ABL1 (43) 36 23 22 10 8 6 6 6 4 4 3  Childhood (21) 16 10 10 7 4 2 3 4 1 2 2  Adult (22) 20 13 12 3 4 4 3 2 3 2 1 Hypodiploid (10) 5 10 10 0 0 0 1 1 0 1 2 Other B ALLs (75) 15 25 22 4 1 0 2 5 1 3 10 High hyperdiploid (39) 2 8 4 1 3 0 0 0 5 0 3 MLL-rearranged (22) 1 4 4 0 2 0 0 0 3 0 2 TCF3–PBX1 (17) 0 6 7 0 2 0 0 0 2 0 0 ETV6–RUNX1 (48) 0 14 16 6 2 0 5 7 4 6 33 T-lineage (50) 2 36 5 1 6 1 3 0 3 0 4 Total (304) 61 126 90 22 24 7 17 19 22 14 57 P-value 6.6 × 10-27 7.4 × 10-10 1.4 × 10-9 7.0 × 10-8 1.1 × 10-6 0.0004 0.0247 1.5 × 10-7 2.6 × 10-6 0.0076 9.1 × 10-15

The authors thank Z. Cai for technical help, K. Rakestraw and J. Armstrong for assistance with sequencing, R. Williams and C. Sherr for the provision of Arf null hematopoietic cells and BCR–ABL1 retroviral vectors, O. Heidenreich for providing the SKNO-1 cell line, and D. Campana for providing the OP1 cell line. This study was supported by the American Lebanese Syrian Associated Charities of St Jude Children’s Research Hospital. C.G.M. was supported by grants from the National Health and Medical Research Council (Australia), the Royal Australasian College of Physicians, and the Haematology Society of Australasia.

Author Contributions C.G.M. collected and extracted clinical samples, performed laboratory assays and analysed data. C.B.M., L.A.P., J.D. and I.R. performed laboratory assays. J.M. analysed SNP array data. D.W., T.P.H., M.M.L., C.-H.P., M.V.R. and S.A.S. collected clinical samples and data. C.G.M and J.R.D designed the study and wrote the manuscript, which was reviewed by all authors.

The primary SNP microarray data have been deposited in NCBIs Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih.gov/geo/) and are accessible through GEO Series accession numbers GSE9109GSE9113.

Ribeiro, R. C. et al. Clinical and biologic hallmarks of the Philadelphia chromosome in childhood acute lymphoblastic leukemia. Blood 70, 948953 (1987) Gleissner, B. et al. Leading prognostic relevance of the BCR-ABL translocation in adult acute B-lineage lymphoblastic leukemia: a prospective study of the German Multicenter Trial Group and confirmed polymerase chain reaction analysis. Blood 99, 15361543 (2002) Goldman, J. M. & Melo, J. V. Chronic myeloid leukemia–advances in biology and new approaches to treatment. N. Engl. J. Med. 349, 14511464 (2003) Daley, G. Q., Van Etten, R. A. & Baltimore, D. Blast crisis in a murine model of chronic myelogenous leukemia. Proc. Natl Acad. Sci. USA 88, 1133511338 (1991) Williams, R. T., Roussel, M. F. & Sherr, C. J. Arf gene loss enhances oncogenicity and limits imatinib response in mouse models of Bcr-Abl-induced acute lymphoblastic leukemia. Proc. Natl Acad. Sci. USA 103, 66886693 (2006) Melo, J. V. The diversity of BCR-ABL fusion proteins and their relationship to leukemia phenotype. Blood 88, 23752384 (1996) Melo, J. V. & Barnes, D. J. Chronic myeloid leukaemia as a model of disease evolution in human cancer. Nature Rev. Cancer 7, 441453 (2007) Mullighan, C. G. et al. Genome-wide analysis of genetic alterations in acute lymphoblastic leukaemia. Nature 446, 758764 (2007) Hahm, K. et al. The lymphoid transcription factor LyF-1 is encoded by specific, alternatively spliced mRNAs derived from the Ikaros gene. Mol. Cell. Biol. 14, 71117123 (1994) Molnar, A. & Georgopoulos, K. The Ikaros gene encodes a family of functionally diverse zinc finger DNA-binding proteins. Mol. Cell. Biol. 14, 82928303 (1994) Molnar, A. et al. The Ikaros gene encodes a family of lymphocyte-restricted zinc finger DNA binding proteins, highly conserved in human and mouse. J. Immunol. 156, 585592 (1996) Rebollo, A. & Schmitt, C. Ikaros, Aiolos and Helios: transcription regulators and lymphoid malignancies. Immunol. Cell Biol. 81, 171175 (2003) Sun, L., Liu, A. & Georgopoulos, K. Zinc finger-mediated protein interactions modulate Ikaros activity, a molecular control of lymphocyte development. EMBO J. 15, 53585369 (1996) Klug, C. A. et al. Hematopoietic stem cells and lymphoid progenitors express different Ikaros isoforms, and Ikaros is localized to heterochromatin in immature lymphocytes. Proc. Natl Acad. Sci. USA 95, 657662 (1998) Sun, L. et al. Expression of dominant-negative Ikaros isoforms in T-cell acute lymphoblastic leukemia. Clin. Cancer Res. 5, 21122120 (1999) Sun, L. et al. Expression of aberrantly spliced oncogenic ikaros isoforms in childhood acute lymphoblastic leukemia. J. Clin. Oncol. 17, 37533766 (1999) Sun, L. et al. Expression of dominant-negative and mutant isoforms of the antileukemic transcription factor Ikaros in infant acute lymphoblastic leukemia. Proc. Natl Acad. Sci. USA 96, 680685 (1999) Nakase, K. et al. Dominant negative isoform of the Ikaros gene in patients with adult B-cell acute lymphoblastic leukemia. Cancer Res. 60, 40624065 (2000) Olivero, S. et al. Detection of different Ikaros isoforms in human leukaemias using real-time quantitative polymerase chain reaction. Br. J. Haematol. 110, 826830 (2000) Nishii, K. et al. Expression of B cell-associated transcription factors in B-cell precursor acute lymphoblastic leukemia cells: association with PU.1 expression, phenotype, and immunogenotype. Int. J. Hematol. 71, 372378 (2000) Takanashi, M. et al. Expression of the Ikaros gene family in childhood acute lymphoblastic leukaemia. Br. J. Haematol. 117, 525530 (2002) Tonnelle, C. et al. Overexpression of dominant-negative Ikaros 6 protein is restricted to a subset of B common adult acute lymphoblastic leukemias that express high levels of the CD34 antigen. Hematol. J. 4, 104109 (2003) Klein, F. et al. BCR-ABL1 induces aberrant splicing of IKAROS and lineage infidelity in pre-B lymphoblastic leukemia cells. Oncogene 25, 11181124 (2006) Wang, J. H. et al. Selective defects in the development of the fetal and adult lymphoid system in mice with an Ikaros null mutation. Immunity 5, 537549 (1996) Georgopoulos, K. et al. The Ikaros gene is required for the development of all lymphoid lineages. Cell 79, 143156 (1994) Winandy, S., Wu, P. & Georgopoulos, K. A dominant mutation in the Ikaros gene leads to rapid development of leukemia and lymphoma. Cell 83, 289299 (1995) Fugmann, S. D. et al. The RAG proteins and V(D)J recombination: complexes, ends, and transposition. Annu. Rev. Immunol. 18, 495527 (2000) Weir, B. A. et al. Characterizing the cancer genome in lung adenocarcinoma. Nature 450, 893898 (2007) Kirstetter, P. et al. Ikaros is critical for B cell differentiation and function. Eur. J. Immunol. 32, 720730 (2002) Drexler, H. G. The Leukemia-Lymphoma Cell Line Facts Book 1st edn (Academic Press, London, 2001) Manabe, A. et al. Interleukin-4 induces programmed cell death (apoptosis) in cases of high-risk acute lymphoblastic leukemia. Blood 83, 17311737 (1994) Ehrich, M. et al. Quantitative high-throughput analysis of DNA methylation patterns by base-specific cleavage and mass spectrometry. Proc. Natl Acad. Sci. USA 102, 1578515790 (2005)

doi: 10.1038/nature06866


Genome-wide screen reveals APC-associated RNAs enriched in cell protrusions p.115

version="1.0" encoding="utf-8"?> ]>

Nature 453 7191 20080501 1151195 0028-0836 1476-4687 2008Nature Publishing Group Supplementary Information

The file contains Supplementary Figures 1-14 with Legends. and Legends to Supplementary Movies 1-5.

Supplementary Table S1

The file contains Supplementary Table S1. This file contains a list of RNAs significantly enriched in pseudopodia in response to both LPA and fibronectin (FN) stimulation.

Supplementary Table S2

The file contains Supplementary Table S2. This file contains a list of RNAs significantly enriched in pseudopodia (Ps) or cell bodies (CB) in response to LPA.

Supplementary Table S3

The file contains Supplementary Table S3. This file contains a list of RNAs significantly enriched in pseudopodia (Ps) or cell bodies (CB) in response to Fibronectin.

Supplementary Movie 1

The file contains Supplementary Movie 1. Confocal fluorescence time lapse imaging of a cell expressing mRFP (red), MS2-GFP (green) and the &bgr;globin-24bs/pkp4 mRNA. Overlay images of the two channels are shown. Arrows point to protrusions with localized RNA granules. Note that the RNA granules remain stationary over the course of observation. Time is indicated on the upper left corner in seconds.

Supplementary Movie 2

The file contains Supplementary Movie 2. FRAP experiment of a localized RNA granule. Shown is a cell expressing MS2-GFP and the &bgr;globin-24bs/pkp4 mRNA. A localized RNA granule at the end of a protrusion is indicated by an arrow. Fluorescence at this protrusion was bleached and its recovery recorded over ca 60 seconds. Note that fluorescence recovers only minimally during the course of observation. Time is indicated on the upper left corner in seconds.

Supplementary Movie 3

The file contains Supplementary Movie 3. FRAP experiment of a localized RNA granule. Shown is a cell expressing MS2-GFP and the &bgr;globin-24bs/pkp4 mRNA and which exhibits two localized RNA granules at the ends of two protrusions. Fluorescence at one of the granules (arrow) was bleached and its recovery recorded over ca 4 minutes. Note that fluorescence recovers only minimally during the course of observation while the second granule (arrowhead) remains stationary. Time is indicated on the upper left corner in seconds.

Supplementary Movie 4

The file contains Supplementary Movie 4. FRAP experiment of a cell expressing EB1-GFP. Fluorescence at a protrusion (arrow) was bleached and its recovery recorded over ca 60 seconds. Note that EB1-GFP comets (representing +ends of dynamic MTs) rapidly move throughout the cell body and several of them enter into protrusions over the course of observation. Time is indicated on the upper left corner in seconds.

Supplementary Movie 5

The file contains Supplementary Movie 5. FRAP experiment of a cell expressing GFP-APC which exhibits APC granules at the tips of different protrusions. Fluorescence at the tip of one protrusion (arrow) was bleached and its recovery recorded over ca 60 seconds. Note that fluorescence recovers only minimally during the course of observation while another APC granule (arrowhead) remains stationary. Time is indicated on the upper left corner in seconds.

Genome-wide screen reveals APC-associated RNAs enriched in cell protrusions StavroulaMiliS KonstadinosMoissogluK Ian G.MacaraI G Department of Microbiology, Center for Cell Signaling, Cardiovascular Research Center, University of Virginia, HSC, Charlottesville, Virginia 22908-0577, USA Correspondence and requests for materials should be addressed to I.G.M. (igm9c@virginia.edu) or S.M. (sm2ju@virginia.edu). &e080501-1; &nature06888-s1;

RNA localization is important for the establishment and maintenance of polarity in multiple cell types. Localized RNAs are usually transported along microtubules or actin filaments and become anchored at their destination to some underlying subcellular structure. Retention commonly involves actin or actin-associated proteins, although cytokeratin filaments and dynein anchor certain RNAs. RNA localization is important for diverse processes ranging from cell fate determination to synaptic plasticity; however, so far there have been few comprehensive studies of localized RNAs in mammalian cells. Here we have addressed this issue, focusing on migrating fibroblasts that polarize to form a leading edge and a tail in a process that involves asymmetric distribution of RNAs. We used a fractionation scheme combined with microarrays to identify, on a genome-wide scale, RNAs that localize in protruding pseudopodia of mouse fibroblasts in response to migratory stimuli. We find that a diverse group of RNAs accumulates in such pseudopodial protrusions. Through their 3′ untranslated regions these transcripts are anchored in granules concentrated at the plus ends of detyrosinated microtubules. RNAs in the granules associate with the adenomatous polyposis coli (APC) tumour suppressor and the fragile X mental retardation protein (FMRP). APC is required for the accumulation of transcripts in protrusions. Our results suggest a new type of RNA anchoring mechanism as well as a new, unanticipated function for APC in localizing RNAs.

To identify on a genome-wide scale RNAs that are enriched at the leading edge of migrating cells, we used a fractionation method in which cells are plated on a microporous filter and then induced to polarize and extend pseudopodial protrusions in response to a migratory stimulus. Pseudopodia and cell bodies are then physically isolated and their contents compared. We isolated pseudopodia and cell body fractions from NIH/3T3 cells extending protrusions in response to a chemotactic (lysophosphatidic acid, LPA) or a haptotactic (fibronectin) stimulus (Fig. 1a, b and Supplementary Fig. 1a). The quality of the fractionation was assessed by immunoblotting for activated focal adhesion kinase (FAK, phosphorylated at Y397), which is concentrated at the leading edge during migration. Activated FAK was enriched in the pseudopodial fraction (Supplementary Fig. 1b), whereas total FAK and Ran were not. Total RNA from pseudopodia and cell body fractions was hybridized on Affymetrix GeneChip arrays, which analyse the expression level of over 39,000 transcripts. The resulting signals were processed to identify transcripts that show an asymmetric distribution.

About 50 RNAs were significantly enriched in pseudopodia in response to both migratory stimuli (Supplementary Table 1). This enrichment was verified by quantitative polymerase chain reaction with reverse transcription (RT–PCR; Fig. 1c) and real-time RT–PCR (Supplementary Fig. 1c) analyses of representative RNAs. Notably, these RNAs did not include &bgr;-actin or Arp2/3 subunit transcripts, which have been previously observed at lamellipodial regions, possibly because in some cell types these RNAs accumulate at the leading edge in only a small percentage of cells. Most transcripts enriched in pseudopodia encode proteins with various functions, ranging from membrane traffic to cytoskeletal organization, signalling, microtubule-based transport and RNA metabolism, as well as a number of uncharacterized proteins (Supplementary Table 1). However, comparison of the primary structure of these transcripts did not reveal any readily identifiable motifs shared among them, suggesting that potential localization signals are probably defined by a combination of primary structure and higher order structure, as is common for most localized RNAs.

To dissect the localization mechanisms, we focused on the Rab13 and plakophilin 4 (Pkp4) messenger RNAs, which both showed robust localization. First, we expressed in NIH/3T3 cells the Rab13 gene encompassing the whole open reading frame and 3′ untranslated region (UTR) (Fig. 2a). A Flag tag at the 5′ end was used to distinguish the exogenous mRNA from endogenous transcript. Transfected cells were induced to extend pseudopodial protrusions in response to LPA, and pseudopodia and cell body fractions were isolated. RT–PCR analysis revealed that, like the endogenous Rab13 mRNA, the reporter transcript was enriched in pseudopodia (Fig. 2b). Pseudopodial enrichment was abolished by replacement of the 3′ UTR of the Rab13 gene with the corresponding region from the non-localized RhoA gene. Moreover, the non-localized &bgr;-globin mRNA was recruited to pseudopodia when attached to the Rab13 3′ UTR (Fig. 2a, b). All exogenous RNAs were expressed at similar levels, close to the expression level of the endogenous Rab13 mRNA (Supplementary Fig. 2a, b), and in all experiments the distribution of the endogenous Rab13 mRNA was determined in parallel, to ensure the reproducibility of the fractionation (data not shown). We conclude that the Rab13 3′ UTR is necessary and sufficient to direct RNA accumulation in pseudopodia.

To visualize the localization pattern conferred by the Rab13 3′ UTR, we used the MS2 system. The MS2 bacteriophage coat protein binds with high affinity to RNA elements, multiple repeats of which are introduced into a reporter transcript. MS2 is fused to green fluorescent protein (GFP) and carries a nuclear localization signal to force accumulation in the nucleus. When co-expressed, the MS2–GFP binds to the reporter RNA and accompanies it to the cytoplasm, thus providing indirect detection of reporter RNA distribution in the cytoplasm of live cells.

Twenty-four repeats of the MS2-binding site were introduced into the &bgr;-globin gene downstream of the coding region, followed by different 3′ UTR sequences (Fig. 2c). These constructs were co-expressed with MS2–GFP and fluorescence was monitored in live cells during spreading on a fibronectin-coated surface. Cells expressing &bgr;-globin mRNA with control 3′ UTRs derived from the vector or from two non-localized mRNAs, Rac1 and RhoA, exhibited diffuse fluorescence throughout the cytoplasm. The GFP signal overlapped entirely with the fluorescence from co-transfected mRFP protein, used as a diffuse cytosolic marker (Fig. 2d, bottom panels, and e, constructs &bgr;-globin-24bs/-, &bgr;-globin-24bs/Rac1 and &bgr;-globin-24bs/RhoA). In contrast, in a large proportion of the cells that expressed a &bgr;-globin mRNA carrying the Rab13 3′ UTR, GFP was concentrated in granules at the tips of protrusions (Fig. 2d, arrows, top panels, and e, construct &bgr;-globin-24bs/Rab13). Granular accumulation did not result from differences in expression levels, as all RNAs were expressed at similar amounts (Supplementary Fig. 2c). Furthermore, the signal was dependent on the presence of the 24 MS2-binding sites in the mRNA (Fig. 2e, construct &bgr;-globin-0bs/Rab13), confirming that it reflects the distribution of the transfected RNA. Significantly, a similar localization was conferred by the 3′ UTR of Pkp4 (Fig. 2d, e, construct &bgr;-globin-24bs/Pkp4). Therefore, certain 3′ UTRs can direct RNAs to granules at the tips of protrusions, and this property is shared among mRNAs enriched in pseudopodia.

When imaged over time, the localized RNA granules are relatively stationary (Supplementary Movie 1) and appear and disappear with surprisingly slow kinetics (Supplementary Fig. 3). This stable accumulation of RNAs could represent either the steady state of a dynamic movement of RNA molecules to and from the granules, or result from sequestration into some cellular structure. To distinguish between these possibilities, we performed fluorescence recovery after photobleaching (FRAP). Because the RNA granules vary in size (see Fig. 2d and Supplementary Fig. 3), we focused, for these experiments, on the smaller granules of approximately 1–2 &mgr;m in diameter. Whereas bleaching of the fluorescence signal in internal cytoplasmic areas was followed by rapid recovery (Fig. 3a), bleaching of the localized RNA granules was followed by very slow fluorescence recovery (Fig. 3a and Supplementary Movies 2 and 3), indicating that the transcripts present in these granules are stably associated/anchored in these structures and do not exchange rapidly with free cytoplasmic RNA molecules.

To gain insight into the identity of the structures that anchor transcripts at the tips of protrusions, the localized RNAs were expressed together with fluorescently tagged markers of various cellular structures. The RNA granules did not co-localize significantly with DCP1-containing P-bodies, focal adhesions (Supplementary Fig. 4a, c, d), cortical actin or actin stress fibres (Supplementary Fig. 4b). Notably, the granules appeared to be associated with microtubules, visualized through expression of RFP–tubulin, and were specifically concentrated at their plus ends (Fig. 3b and Supplementary Fig. 5). Consistent with a microtubule association, the RNA granules largely disappeared when cells were treated with the microtubule-depolymerizing agent nocodazole (Fig. 3d and Supplementary Fig. 6b), at concentrations that do not affect the actin cytoskeleton (Supplementary Fig. 6a). In contrast, the presence of RNA granules was not affected when the actin cytoskeleton was disrupted by cytochalasin D (Fig. 3d and Supplementary Fig. 6a, b). Notably, brief treatment with nocodazole significantly reduced the pseudopodial enrichment of four different endogenous localized RNAs (Fig. 3e), without affecting the overall integrity and number of protrusions (Supplementary Fig. 6c). We conclude that association with microtubule plus ends is a property shared by multiple RNAs found in pseudopodia.

The fact that the RNA granules remain stationary over several minutes (Supplementary Movie 1) suggested that the microtubules with which they associate do not exhibit dynamic instability. Indeed, the RNA granules did not co-localize significantly with EB1–RFP or RFP–CLIP170 (Supplementary Fig. 7a, b), two plus-end tracking proteins known to associate with plus ends of dynamic microtubules. Furthermore, FRAP analysis on protrusions of EB1–GFP-expressing cells showed that, in contrast to the RNA granules, EB1 is highly dynamic with fluorescence recovering after a few seconds (Supplementary Fig. 8a and Supplementary Movie 4). Therefore, we tested whether the RNA granules associate specifically with the plus ends of stable microtubules, which are marked by post-translational modifications of tubulin. Although the RNA granules were not attached to acetylated microtubules (Supplementary Fig. 9a), they did associate with the plus ends of detyrosinated microtubules (Glu-microtubules) (Fig. 3c and Supplementary Fig. 9b). Glu-microtubules do not significantly grow or shrink over a period of minutes, in agreement with the dynamics of the RNA granules we observe.

APC is an unusual plus-end tracking protein that associates with only a minority of microtubules, and it has been observed in particular at the plus ends of Glu-microtubules. In migrating cells APC is attached to a subset of the microtubules growing into protrusions towards the leading edge. When NIH/3T3 cells were induced to spread on a fibronectin-coated surface, we found that GFP–APC was mainly concentrated in granules at the tips of protrusions, reminiscent of the distribution exhibited by the localized RNAs (Supplementary Fig. 10a, arrows). Indeed, co-expression of a localized reporter RNA with APC tagged with three copies of orange fluorescent protein (3×OFP–APC) revealed that >90% (n = 80) of the RNA granules at the tips of protrusive areas co-localize with APC (Fig. 4a, arrows and Supplementary Fig. 10b). Furthermore, FRAP analysis showed that APC present in granules is stably anchored there, exhibiting a very slow exchange rate (Supplementary Fig. 8b and Supplementary Movie 5). Taken together, these data strongly suggest that APC and the localized RNAs are part of the same complex.

To test this hypothesis, we asked if these RNAs bind specifically to endogenous APC. Indeed, APC co-immunoprecipitated with both Rab13 and Pkp4 mRNAs (Fig. 4b and Supplementary Fig. 11a). Importantly, under high stringency conditions, APC associated preferentially with RNAs enriched in pseudopodia (Supplementary Fig. 11b, c). Moreover, APC also bound the cytoplasmic poly(A)-binding protein (PABP1). This association was specific as neither Ran (Supplementary Fig. 11a) nor importin-&bgr; (data not shown) was detected. Furthermore, interaction of APC with PABP1 was disrupted by pre-treatment with RNase, indicating that it is mediated through RNA (Fig. 4b, c). We also found that APC associates with FMRP (Fig. 4b), a known translational regulator of localized RNAs in other systems. FMRP co-localized with the RNA granules at tips of protrusions and co-precipitated with endogenous localized mRNAs (Fig. 4d, e). We conclude that APC is a component of RNP complexes that contain localized RNAs, PABP1 and FMRP.

To test directly whether APC mediates anchoring of the localized RNAs, we knocked down APC expression using short-hairpin RNAs (shRNAs) (Fig. 4f and Supplementary Fig. 12a, b). Knockdown of APC did not significantly affect the ability of cells to extend protrusions (see Supplementary Figs 13 and 14, and data not shown), but reduced both the enrichment of the endogenous Rab13 and Pkp4 transcripts in pseudopodia as well as the localization of the MS2 reporter RNA in protrusions (Fig. 4f, g). It is unlikely that this effect of APC on RNA localization is mediated indirectly through the effects of APC on transcription or microtubule dynamics. APC knockdown does not affect the steady-state levels of localized RNAs (data not shown) and, consistent with previous reports, it does not affect the overall microtubule organization or the presence of Glu-microtubules (Supplementary Fig. 13). APC knockdown did affect the organization of acetylated microtubules, causing more cells to exhibit short acetylated microtubules (Supplementary Figs 14 and ref. 26). However, RNA granules are not associated with acetylated microtubules (Supplementary Fig. 9a) and can form in cells with short acetylated microtubules (Supplementary Fig. 9a, middle panels). Thus, we conclude that APC is directly required for accumulation and anchoring of RNAs in protrusions.

This study provides the first genome-wide identification of asymmetrically distributed RNAs in fibroblasts. We show that, in response to migratory stimuli, >50 mRNAs accumulate in cellular protrusions of fibroblasts, revealing that even in less differentiated polarized cells, RNA localization mechanisms are widely used. Several RNAs have also been recently reported to associate in vitro with mitotic microtubules, but the mechanism is unknown. Our data suggest a novel anchoring mechanism in which specific RNAs accumulate in stable granules at the plus ends of Glu-microtubules. The tumour suppressor protein APC has an essential role in this plus-end anchoring mechanism, and associates both with RNA-binding proteins and with specific RNAs. The disruption or loss of function of APC affects both polarization and cell migration, and we speculate that at least some of the effects of this tumour suppressor are mediated through its ability to anchor mRNAs at the tips of cellular protrusions.

Methods Summary Pseudopodia and cell body isolation

To isolate pseudopodia and cell bodies in response to LPA, we followed the protocol described by ref. 13 with some modifications. Specifically, LPA was added in the bottom chamber for 1 h and the cells were fixed with 0.3% methanol-free formaldehyde (Polysciences, Inc.). Pseudopodia and cell bodies were scraped into crosslink reversal buffer (100 mM Tris pH 6.8, 5 mM EDTA, 10 mM dithiothreitol and 1% SDS), the extracts were incubated at 70 °C for 45 min and used for protein and RNA isolation. To isolate pseudopodia and cell bodies in response to fibronectin, cells were placed in serum-free media in the upper compartment of a Transwell insert equipped with a 3-&mgr;m porous polycarbonate membrane whose underside only was coated with 5 &mgr;g ml-1 fibronectin. Cells were allowed to extend pseudopodial protrusions for 1 h and were subsequently fixed and processed as described above.

RNA isolation and analysis

RNA was isolated using Trizol LS reagent (Invitrogen) and contaminating DNA was removed by treatment with RQ1 DNase (Promega) for 30 min at 37 °C. One microgram of RNA was reverse transcribed using SuperscriptII Reverse Transcriptase (Invitrogen) and random hexamer primers, according to the manufacturer’s instructions. cDNA was used for PCR reactions in a GeneAmp PCR System 9700 (Applied Biosystems).

Microarray analysis

Biotin-labelled cRNAs were generated from total RNA of pseudopodia and cell body fractions and were hybridized to GeneChip Mouse Genome 430 2.0 arrays (Affymetrix). Details about the RNA preparation and hybridization protocols can be found at http://www.healthsystem.virginia.edu/internet/biomolec/microarray.cfm. Results were analysed using the GeneChip Operating Software (GCOS) platform or the dChip software (http://www.dchip.org). Complete lists of RNAs significantly enriched in pseudopodia or cell bodies are presented in Supplementary Tables 2 and 3.

Plasmid constructs

A plasmid encoding the human &bgr;-globin mRNA was provided by J. Lykke-Andersen. The plasmid contains all the coding exons and introns of human &bgr;-globin, 54 nucleotides of the &bgr;-globin 5′ UTR, 69 nucleotides of the &bgr;-globin 3′ UTR followed by 24 repeats of the MS2-binding site cloned into the HindIII, XbaI sites of pcDNA3 (Invitrogen). To facilitate cloning, a multiple cloning site with sequence 5′-ATCGATGGTACCGCTAGCGATATCCTCGAG-3′ was introduced between the XbaI and ApaI sites, generating plasmid &bgr;-globin-24bs/-. Various 3′ UTRs were PCR amplified from mouse genomic DNA (BD Biosciences Clontech) with primers carrying appropriate restriction sites and were ligated into the NheI and XhoI sites (for the Pkp4 and Rab13 3′ UTRs) or into the XbaI and ApaI sites (for the Rac1 and RhoA 3′ UTRs) of the &bgr;-globin-24bs/- plasmid, to generate plasmids &bgr;-globin-24bs/Pkp4, &bgr;-globin-24bs/Rab13, &bgr;-globin-24bs/Rac1 and &bgr;-globin-24bs/RhoA. The exact 3′ UTRs correspond to: nucleotides 3858–4612 of NM_026361 (Pkp4 UTR), nucleotides 91295–92092 of NT_039254 (Rab13 UTR), nucleotides 776–2240 of NM_009007 (Rac1 UTR) and nucleotides 1009–2071 of NM_016802 (RhoA UTR). To generate plasmids &bgr;-globin-0bs/Pkp4 and &bgr;-globin-0bs/Rab13, plasmids &bgr;-globin-24bs/Pkp4 and &bgr;-globin-24bs/Rab13, respectively, were digested with NotI and NheI (to remove the fragment containing the 24 MS2-binding sites), the ends were blunted with T4 DNA polymerase and re-ligated. To generate constructs Rab13/Rab13 and Rab13/RhoA, the genomic sequence containing the coding exons and intervening introns of the Rab13 gene was PCR amplified from mouse genomic DNA with primers introducing an EcoRI site at the 5′ end and a KpnI site at the 3′ end. A Flag tag was ligated at the EcoRI site in frame with the Rab13 ORF. The Rab13 3′ UTR or the RhoA 3′ UTR were ligated at the KpnI site through blunt-end ligation. The two resulting fragments were ligated into pEGFP-C1 vector from which the GFP sequence had been previously removed.

The plasmid pcNMS2, containing an oligomerization-defective MS2 coat protein mutant, was provided by J. Lykke-Andersen. The GFP sequence was PCR amplified from plasmid pEGFP-N3 (Clontech) with primers introducing BamHI and XhoI sites and was ligated at the XhoI site with oligonucleotides containing the SV40 large T-Ag NLS. The GFP–NLS fragment was ligated into the BamHI and NotI sites of pcNMS2 to generate the pcNMS2–GFP–NLS plasmid expressing the MS2–GFP protein. RFP–tubulin was generated from pEGFP–tubulin (Clontech) by replacing the NheI/XhoI GFP fragment with a PCR-amplified fragment of mRFP. To generate EB1–GFP and EB1–RFP expressing plasmids, the mouse EB1 cDNA was amplified by RT–PCR from total NIH/3T3 RNA with primers introducing a KpnI site at the 5′ end and a BglII site at the 3′ end. The EB1 fragment was either ligated into KpnI/BamHI sites of pEGFP-N3 or into KpnI/XhoI sites of pcDNA3 (Invitrogen) together with a BglII/XhoI PCR-amplified fragment of mRFP. The plasmid expressing GFP–APC was provided by I. Nathke. The SacI/BamHI APC fragment was ligated with a NheI/SacI fragment containing three copies of OFP (generated through sequential PCR amplification and ligation of individual OFP fragments) into the XbaI and BamHI sites of pKH3 to generate plasmid 3×OFP–APC. To generate RFP–FMRP, a plasmid expressing GFP–FMRP was used (provided by R. Darnell) and the NheI/SacI fragment, encoding GFP, was replaced with a NheI/SacI PCR-amplified fragment of mRFP. To generate Flag–FMRP, the SacI/EcoRI fragment encoding FMRP (from the RFP-FMRP plasmid) was ligated together with a NheI/SacI fragment encoding the Flag tag into the XbaI/EcoRI sites of pKH3. To generate RFP–DCP1B, a plasmid expressing Flag–DCP1B was used (provided by J. Lykke-Andersen). The BamHI (blunted with T4 DNA polymerase)/NotI fragment encoding human DCP1B was ligated with a HindIII/NdeI (blunted with T4 DNA polymerase) fragment of mRFP into HindIII and NotI sites of pcDNA3.

For knockdown experiments, oligonucleotides were synthesized targeting different regions of the mouse Apc mRNA. Sequences of the oligonucleotides are as follows: shAPC 2 sense oligonucleotide, 5′-GATCCCCGAATCAACCAGGCATAATATTCAAGAGATATTATGCCTGGTTGATTCTTTTTGGAAA-3′ and shAPC 2 antisense oligonucleotide, 5′-AGCTTTTCCAAAAAGAATCAACCAGGCATAATATCTCTTGAATATTATGCCTGGTTGATTCGGG-3′; shAPC 4 sense oligonucleotide, 5′-GATCCCCTAAGTGATCTGACAATAGATTCAAGAGATCTATTGTCAGATCACTTATTTTTGGAAA-3′ and shAPC 4 antisense oligonucleotide, 5′- AGCTTTTCCAAAAATAAGTGATCTGACAATAGATCTCTTGAATCTATTGTCAGATCACTTAGGG-3′; shAPC 5 sense oligonucleotide, 5′- GATCCCCCAACTACAGTGAACGTTATTTCAAGAGAATAACGTTCACTGTAGTTGTTTTTGGAAA-3′ and shAPC 5 antisense oligonucleotide, 5′- AGCTTTTCCAAAAACAACTACAGTGAACGTTATTCTCTTGAAATAACGTTCACTGTAGTTGGGG-3′. Bold characters indicate the Apc mRNA targeting sequence; underlined characters indicate the 9-bp hairpin loop. Sequences of control oligonucleotides targeting luciferase were: sense oligonucleotide, 5′- GATCCCCCGTACGCGGAATACTTCGATTCAAGAGATCGAAGTATTCCGCGTACGTTTTTGGAAA-3′, antisense oligonucleotide, 5′-AGCTTTTCCAAAAACGTACGCGGAATACTTCGATCTCTTGAATCGAAGTATTCCGCGTACGGGG-3′. The sense and antisense oligonucleotides were annealed and ligated into the BglII and HindIII sites of the pSuper vector.

Cell culture and transfection

NIH/3T3 cells were grown in DMEM supplemented with 10% calf serum, sodium pyruvate, penicillin and streptomycin (Invitrogen). For live cell imaging, plasmid constructs were transfected with Effectene (Qiagen) according to the manufacturer’s instructions. Twenty-four hours after transfection cells were plated for ∼2 h on Lab-Tek coverglass chambers (Nalge nunc International) coated with 5 &mgr;g ml-1 fibronectin and fluorescence was visualized by confocal microscopy. Where indicated, cells were treated at 37 °C for 30 min with 10 &mgr;M nocodazole or 1 &mgr;M cytochalasin D. For pseudopodia/cell body fractionation or for shRNA-mediated knockdown experiments, cells were electroporated using the Gene Pulser II electroporation system (Bio-Rad) with plasmid constructs (5 &mgr;g per 6 × 106 cells) or with pSuper constructs encoding shRNAs (40 &mgr;g per 6 × 106 cells), respectively. Electroporation efficiency, based on co-transfected GFP, was generally >70%. Electroporated cells were processed after 72 h.

Pseudopodia and cell body isolation

To isolate pseudopodia and cell bodies of cells induced to migrate with LPA, we followed the protocol described by ref. 13 with some modifications. Cells were serum-starved overnight and 1.5 × 106 cells were placed in the upper compartment of a Transwell insert (24-mm diameter, Costar) equipped with a 3-&mgr;m porous polycarbonate membrane coated on both sides with 5 &mgr;g ml-1 fibronectin. Cells were allowed to spread on the upper surface of the membrane for 2 h. LPA (150 ng ml-1) was then added in the bottom chamber to induce the cells to extend pseudopodial protrusions. After 1 h the cells were briefly rinsed with PBS and fixed with 0.3% methanol-free formaldehyde (Polysciences, Inc.) in PBS for 10 min at room temperature. Glycine was added to 250 mM for 5 min at room temperature and the cells were washed twice with PBS. To isolate pseudopodia, cell bodies on the upper membrane surface were manually removed with a cotton swab and laboratory paper and pseudopodia on the underside of the membrane were scraped into crosslink reversal buffer (100 mM Tris pH 6.8, 5 mM EDTA, 10 mM dithiothreitol and 1% SDS). Cell bodies were similarly isolated except that pseudopodia on the underside of the membrane were manually removed and cell bodies were scraped into crosslink reversal buffer. Extracts were incubated at 70 °C for 45 min to reverse the formaldehyde-induced crosslinks. Proteins were directly analysed by western blot otherwise total RNA was isolated and processed as described below.

For fractionation after nocodazole treatment, cells were processed as above except that nocodazole was added to 10 &mgr;M during the last 25 min of the assay.

To isolate pseudopodia and cell bodies of cells migrating towards fibronectin, ∼1 × 106 cells were placed in serum-free media in the upper compartment of a Transwell insert equipped with a 3-&mgr;m porous polycarbonate membrane whose underside only was coated with 5 &mgr;g ml-1 fibronectin. Cells were allowed to extend pseudopodial protrusions towards the fibronectin-coated surface for 1 h and were subsequently fixed and processed as described above.

RNA isolation and analysis

Total RNA or RNA from pseudopodia and cell body fractions was isolated using Trizol LS reagent (Invitrogen) and contaminating DNA was removed by treatment with RQ1 DNase (Promega) for 30 min at 37 °C. One microgram of RNA was reverse transcribed using SuperscriptII Reverse Transcriptase (Invitrogen) and random hexamer primers according to the manufacturer’s instructions. cDNA was used for PCR reactions in a GeneAmp PCR System 9700 (Applied Biosystems) to detect different RNAs using the following primer pairs: Rab13 F 5′-GCCTACCAGTGTTGGCTCTT-3′, Rab13 R 5′-TCCACGGTAATAGGCGGTAG-3′, Rab13 UTR F 5′-AGGCTGCTAGCGAGCATTTCTTGCCTCCTAT-3′, Rab13 UTR R 5′-AATGGCTCGAGCCATTCATTTCTTCTTCC-3′; Pkp4 F 5′-AGGCTGCTAGCCAGGGAAGTGAGGAAACC-3′, Pkp4 R 5′-AATGGCTCGAGAAAACATGAAGGGCATCC-3′; Ankrd25 F 5′-TTCAAAGCCAGAAAGCCAAG-3′, Ankrd25 R 5′-AGGTGACAAAGGGTGGTGAG-3′; Inpp1* F 5′-TTTGAAGTGGAATGGGGATAAC-3′, Inpp1* R 5′-AATAGTCAGATAGTCAAACTCATGG-3′ (these primers detect a likely RNA isoform of Inpp1 with a longer 3′ UTR); Apc F 5′-CCTCTCACCGGAGTAAGCAG-3′, Apc R 5′-GTCGTCCTGGGAGGTATGAA-3′; Arpc3 F 5′-CACGGACATTGTGGATGAAG-3′, Arpc3 R 5′-CCACCACTTGCTGGCTTTAT-3′; Flag Rab13 F 5′-TAATACGACTCACTATAGGG-3′, Flag Rab13 R 5′-TCCACGGTAATAGGCGGTAG-3′; &bgr;-globin F 5′-TTGAGTCCTTTGGGGATCTG-3′, &bgr;-globin R 5′-CACTGGTGGGGTGAATTCTT-3′; &bgr;-actin F 5′-TGTTACCAACTGGGACGACA-3′, &bgr;-actin R 5′-GCTGTGGTGGTGAAGCTGTA-3′.

The identity of the amplified products was verified either by sequencing or in the case of transfected RNAs by ensuring that no product was amplified when using RNA from vector-transfected cells. PCR reactions contained 10 mM Tris-HCl (pH 8.3), 50 mM KCl, 1.5 mM MgCl2, 0.2 mM each dNTPs, 0.5 &mgr;M each primers and 2.5 U Taq polymerase. The amount of cDNA and the number of cycles were varied for each primer pair, to ensure amplification was within the linear phase. This was verified in all experiments by including decreasing amounts of selected samples.

To calculate the enrichment of RNAs in pseudopodia, equal amounts of RNA from pseudopodia and cell body fractions were analysed by RT–PCR, the signals were quantified using WCIF ImageJ software, normalized to the control Arpc3 mRNA and enrichment in pseudopodia was defined as (pseudopodia signal/cell body signal)-1, when pseudopodia signal > cell body signal or as (-1)×[(cell body signal/pseudopodia signal)-1], when pseudopodia signal < cell body signal.

Microarray analysis

Microarray analysis was performed at the Biomolecular Research Facility at the University of Virginia. Total RNA from pseudopodia and cell body fractions was analysed on an Agilent BioAnalyser and was deemed to be of good quality. Biotin-labelled cRNAs were generated and hybridized to GeneChip Mouse Genome 430 2.0 arrays (Affymetrix), which contain 45,000 probe sets analysing the expression level of over 39,000 transcripts and variants from over 34,000 well-characterized mouse genes. Details about the RNA preparation and hybridization protocols can be found at http://www.healthsystem.virginia.edu/internet/biomolec/microarray.cfm.

In all hybridization experiments, quality assessment variables, such as background, noise, GAPDH 3′/5′ ratio, were within the acceptance limits.

For experiments analysing pseudopodia and cell body fractions from cells induced with LPA, results were analysed using the GeneChip Operating Software (GCOS) platform. RNAs were considered to be significantly enriched in pseudopodia if the P-value was significant, the fold change in signal intensity was greater than 2.2 and the absolute difference in signal intensity was greater than 100.

For experiments analysing pseudopodia and cell body fractions from cells migrating towards fibronectin, results were analysed using the dChip software (http://www.dchip.org). RNAs were considered to be significantly enriched in pseudopodia if the P-value was less than 0.05, the fold change in signal intensity was greater than 1.5 and the absolute difference in signal intensity was greater than 10 times the noise level.

Western blot, immunofluorescence and immunoprecipitation

For western blot and immunofluorescence, the following antibodies were used: mouse monoclonal anti-Ran (Transduction Laboratories), rabbit anti-pY397-FAK (Biosource International), monoclonal anti-FAK (BD Transduction Labs), anti-APC (C-20) (Santa Cruz Biotechnology), rabbit anti-PABP1 (Cell Signaling Technology), anti-FMRP (clone 1C3, Chemicon), anti-acetylated tubulin (Clone 6-11B-1, Sigma), anti-&agr;-tubulin (clone DM1A, Sigma), anti-Glu-tubulin (provided by G. Gundersen).

For immunoprecipitation, anti-APC (C-20) or control antibody (rabbit anti-HA tag, Santa Cruz Biotechnology) was bound on protein-A beads, otherwise anti-Flag M2 agarose beads (Sigma) were used. In all cases, antibody-bound beads were pre-incubated with 200 &mgr;g ml-1 Escherichia coli tRNA, 200 &mgr;g ml-1 herring sperm DNA, 200 &mgr;g ml-1 RNase-free BSA and 50 &mgr;g ml-1 glycogen. Cells spreading on fibronectin-coated plates for approximately 2 h were lysed in lysis buffer (10 mM Tris pH 7.5, 100 mM NaCl, 2.5 mM MgCl2, 0.5% Triton X-100) supplemented with protease inhibitors (leupeptin, pepstatin, aprotinin) and RNase inhibitor (0.5 U &mgr;l-1). Lysates were centrifuged at 10,000g for 10 min and incubated with antibody-bound beads at 4 °C for 3 h. Beads were washed five times either with lysis buffer or, where indicated, with high stringency wash buffer (50 mM Tris pH 7.5, 500 mM NaCl, 1% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS). The immunoprecipitated material was released by incubation in lysis buffer containing 1% SDS. A fraction of the released material was analysed by western blot. The remainder was used for total RNA isolation and RT–PCR analysis. For RNase treatment, 1 mM CaCl2 was added in the lysis buffer and after the initial centrifugation, RNase A (30 &mgr;g ml-1) and micrococcal nuclease (70 U ml-1) were added, the lysate was incubated at 30 °C for 10 min and subsequently centrifuged at 10,000g for 5 min. The supernatant was used in immunoprecipitations as described above.

Confocal microscopy and photobleaching

Imaging and photobleaching were performed with a Zeiss LSM 510 Meta confocal microscope operated by LSM-FCS software (Carl Zeiss). Temperature was maintained at 37 °C. Medium pH was controlled by addition of 25 mM HEPES buffer. EGFP was excited with the 488-nm line of an argon laser at 30% power, 5% transmission. mRFP, OFP and DsRed were excited with the 543-nm line of a HeNe laser at 80% transmission. For photobleaching, regions of interest were selected and bleached with the 488-nm laser line at 100% power, 100% transmission for two iterations. Fluorescence recovery within the region of interest was monitored for ∼60 s. For each experiment, three images were recorded pre-bleach. Mean intensities in the bleached area were measured, background signal was subtracted, intensities were corrected for bleaching during imaging and were expressed as a percentage of the pre-bleach intensity.

Several RNAs are enriched in protruding pseudopodia of migrating cells.

a, Schematic diagram depicting strategies used for isolation of pseudopodia (Ps) and cell bodies (CB). NIH/3T3 cells, plated on microporous filters, were induced to extend protrusions by adding LPA in the bottom chamber. Alternatively, the underside of microporous filters was coated with fibronectin (FN) and cells plated on top extended protrusions towards the fibronectin-coated surface. Cell body and pseudopodia fractions were subsequently isolated. b, Cells extending protrusions in response to LPA or fibronectin, as described in a, were stained with fluorescein-isothiocyanate-conjugated phalloidin. Confocal images of the top and bottom side of the filter are shown. Scale bar, 15 &mgr;m. c, Total RNA from pseudopodia and cell body fractions of cells extending protrusions in response to LPA was analysed by RT–PCR to detect the mRNAs indicated on the left. Increasing amounts of the pseudopodia sample were amplified (lanes 1 and 2) to ensure linearity of the amplification. Values on the right indicate mean pseudopodia/cell body ratios, normalized to the control Arpc3 mRNA, ± s.e.m., n = 3.

The 3′ UTRs direct RNAs to accumulate in granules at tips of protrusions.

a, Schematic of transfected constructs. Black boxes, exons; black lines, introns; grey boxes, 3′ UTR; white box, Flag tag. b, Cells transfected with the constructs depicted in a were fractionated into pseudopodia and cell body fractions after induction with LPA, and RNA enrichment towards each fraction was calculated (n = 2–4). Error bars, s.e.m. c, Schematic depicting the general structure of &bgr;-globin constructs used in d and e. 24bs, 24 MS2-binding sites. d, Imaging of live cells co-transfected with plasmids encoding mRFP, MS2–GFP and &bgr;-globin constructs with various 3′ UTRs. Shown are representative examples of the localization patterns observed when &bgr;-globin mRNA carried the UTRs indicated on the left (localized in granules at tips of protrusions (arrows) or diffuse in the cytoplasm). Scale bar, 10 &mgr;m. e, Quantification of the percentage of cells exhibiting localized RNA distribution when transfected as in d, with constructs carrying the indicated 3′ UTRs (n = 3). Error bars, s.d.

Localized RNA granules are anchored at the plus ends of detyrosinated microtubules.

a, Cells were co-transfected with MS2–GFP and the &bgr;-globin-24bs/Pkp4 RNA. Fluorescence intensity in granules at protrusive areas (left panel, red circle) or within the cytoplasm (right panel, red circle) was monitored before and after photobleaching. The arrow indicates time of bleach. Curves represent average values of ten and five independent experiments, respectively. Error bars indicate s.d. b, c, Confocal fluorescence images of cells expressing MS2–GFP, the &bgr;-globin-24bs/Pkp4 RNA and either RFP–tubulin (b) or fixed and stained with anti-Glu-tubulin antibody (c). Panels show edges of protrusive areas. Scale bar, 3 &mgr;m. d, Cells expressing mRFP, MS2–GFP and the &bgr;-globin-24bs/Pkp4 RNA were treated with nocodazole (Noc.) or cytochalasin D (Cyt.D). The percentage of cells exhibiting localized RNA distribution was quantified as in Fig. 2d, e (n = 3); asterisk, P-value < 0.005 by two-tailed t-test versus untreated control (-). Error bars, ± 1 s.d. e, Enrichment of the indicated mRNAs in pseudopodia was determined in cells extending protrusions in response to LPA in the absence or presence of nocodazole. Error bars indicate s.e.m., n = 3; asterisk, P-value < 0.005 by paired, two-tailed t-test versus untreated control.

APC associates with RNP complexes containing FMRP and is required for localization of RNAs in protrusions.

a, Confocal fluorescence images of cells expressing MS2–GFP, the &bgr;-globin-24bs/Pkp4 RNA and 3×OFP–APC. Panels show edges of protrusive areas. b, NIH/3T3 lysates were immunoprecipitated (IP) with control antibody (IgG), anti-APC antibody (anti-APC) or no antibody (-) and analysed to detect the indicated proteins and mRNAs. c, Same as in b, except that before immunoprecipitation lysates were treated (+) or not (-) with RNase. d, Lysates of cells transfected with Flag–FMRP or vector were immunoprecipitated with anti-Flag antibody and analysed to detect the indicated proteins and mRNAs. e, Confocal fluorescence images of cells expressing MS2–GFP, the &bgr;-globin-24bs/Pkp4 RNA and RFP–FMRP. Panels show edges of protrusive areas. f, Cells were transfected with shRNAs against luciferase (shLuc) or Apc (shAPC), individually or in combination as indicated. Left panel, normalized Apc mRNA levels; right panel, enrichment of the indicated mRNAs in pseudopodia in response to LPA (n = 2–3); asterisk, P-value < 0.05 by paired, two-tailed t-test versus shLuc control. g, shLuc- and shAPC-expressing cells were transfected with MS2–GFP and the &bgr;-globin-24bs/Pkp4 RNA. The percentage of cells exhibiting localized RNA distribution, as described in Fig. 2d, was quantified (n = 7); asterisk, P-value < 0.0001 by two-tailed t-test. All error bars indicate s.e.m.; all scale bars are 5 &mgr;m.

We thank Y. Bao for the microarray and real-time PCR analysis, and J. Lykke-Andersen, I. Nathke, J. T. Parsons, R. Darnell and G. Gundersen for plasmids and antibodies. S.M. is a fellow of the Leukemia and Lymphoma Society. This work was supported by a grant from the NIH to I.G.M., and by the James and Rebecca Craig Foundation.

Author Contributions S.M. performed the experiments. S.M., K.M. and I.G.M. designed the experiments and analysed the data. K.M. provided reagents. S.M. and I.G.M. wrote the manuscript.

The microarray data have been deposited in NCBI's Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih.gov/geo/) and are accessible through GEO Series accession number GSE10230.

St Johnston, D. Moving messages: the intracellular localization of mRNAs. Nature Rev. Mol. Cell Biol. 6, 363375 (2005) Forrest, K. M. & Gavis, E. R. Live imaging of endogenous RNA reveals a diffusion and entrapment mechanism for nanos mRNA localization in Drosophila. Curr. Biol. 13, 11591168 (2003) Babu, K., Cai, Y., Bahri, S., Yang, X. & Chia, W. Roles of Bifocal, Homer, and F-actin in anchoring Oskar to the posterior cortex of Drosophila oocytes. Genes Dev. 18, 138143 (2004) Erdelyi, M., Michon, A. M., Guichet, A., Glotzer, J. B. & Ephrussi, A. Requirement for Drosophila cytoplasmic tropomyosin in oskar mRNA localization. Nature 377, 524527 (1995) Jankovics, F., Sinka, R., Lukacsovich, T. & Erdelyi, M. MOESIN crosslinks actin and cell membrane in Drosophila oocytes and is required for OSKAR anchoring. Curr. Biol. 12, 20602065 (2002) Beach, D. L., Salmon, E. D. & Bloom, K. Localization and anchoring of mRNA in budding yeast. Curr. Biol. 9, 569578 (1999) Liu, G. et al. Interactions of elongation factor 1&agr; with F-actin and &bgr;-actin mRNA: implications for anchoring mRNA in cell protrusions. Mol. Biol. Cell 13, 579592 (2002) Alarcon, V. B. & Elinson, R. P. RNA anchoring in the vegetal cortex of the Xenopus oocyte. J. Cell Sci. 114, 17311741 (2001) Delanoue, R. & Davis, I. Dynein anchors its mRNA cargo after apical transport in the Drosophila blastoderm embryo. Cell 122, 97106 (2005) Condeelis, J. & Singer, R. H. How and why does &bgr;-actin mRNA target? Biol. Cell 97, 97110 (2005) Leung, K. M. et al. Asymmetrical &bgr;-actin mRNA translation in growth cones mediates attractive turning to netrin-1. Nature Neurosci. 9, 12471256 (2006) Mingle, L. A. et al. Localization of all seven messenger RNAs for the actin-polymerization nucleator Arp2/3 complex in the protrusions of fibroblasts. J. Cell Sci. 118, 24252433 (2005) Cho, S. Y. & Klemke, R. L. Purification of pseudopodia from polarized cells reveals redistribution and activation of Rac through assembly of a CAS/Crk scaffold. J. Cell Biol. 156, 725736 (2002) Moissoglu, K. & Schwartz, M. A. Integrin signalling in directed cell migration. Biol. Cell 98, 547555 (2006) Vikesaa, J. et al. RNA-binding IMPs promote cell adhesion and invadopodia formation. EMBO J. 25, 14561468 (2006) Jambhekar, A. & Derisi, J. L. Cis-acting determinants of asymmetric, cytoplasmic RNA transport. RNA 13, 625642 (2007) Bertrand, E. et al. Localization of ASH1 mRNA particles in living yeast. Mol. Cell 2, 437445 (1998) Galjart, N. CLIPs and CLASPs and cellular dynamics. Nature Rev. Mol. Cell Biol. 6, 487498 (2005) Akhmanova, A. & Hoogenraad, C. C. Microtubule plus-end-tracking proteins: mechanisms and functions. Curr. Opin. Cell Biol. 17, 4754 (2005) Infante, A. S., Stein, M. S., Zhai, Y., Borisy, G. G. & Gundersen, G. G. Detyrosinated (Glu) microtubules are stabilized by an ATP-sensitive plus-end cap. J. Cell Sci. 113, 39073919 (2000) Wen, Y. et al. EB1 and APC bind to mDia to stabilize microtubules downstream of Rho and promote cell migration. Nature Cell Biol. 6, 820830 (2004) Mimori-Kiyosue, Y., Shiina, N. & Tsukita, S. Adenomatous polyposis coli (APC) protein moves along microtubules and concentrates at their growing ends in epithelial cells. J. Cell Biol. 148, 505518 (2000) Nathke, I. S., Adams, C. L., Polakis, P., Sellin, J. H. & Nelson, W. J. The adenomatous polyposis coli tumor suppressor protein localizes to plasma membrane sites involved in active cell migration. J. Cell Biol. 134, 165179 (1996) Bagni, C. & Greenough, W. T. From mRNP trafficking to spine dysmorphogenesis: the roots of fragile X syndrome. Nature Rev. Neurosci. 6, 376387 (2005) Jin, P., Alisch, R. S. & Warren, S. T. RNA and microRNAs in fragile X mental retardation. Nature Cell Biol. 6, 10481053 (2004) Kroboth, K. et al. Lack of adenomatous polyposis coli protein correlates with a decrease in cell migration and overall changes in microtubule stability. Mol. Biol. Cell 18, 910918 (2007) Kita, K., Wittmann, T., Nathke, I. S. & Waterman-Storer, C. M. Adenomatous polyposis coli on microtubule plus ends in cell extensions can promote microtubule net growth with or without EB1. Mol. Biol. Cell 17, 23312345 (2006) Blower, M. D., Feric, E., Weis, K. & Heald, R. Genome-wide analysis demonstrates conserved localization of messenger RNAs to mitotic microtubules. J. Cell Biol. 179, 13651373 (2007) Watanabe, T. et al. Interaction with IQGAP1 links APC to Rac1, Cdc42, and actin filaments during cell polarization and migration. Dev. Cell 7, 871883 (2004) Etienne-Manneville, S. & Hall, A. Cdc42 regulates GSK-3&bgr; and adenomatous polyposis coli to control cell polarity. Nature 421, 753756 (2003)

doi: 10.1038/nature06888


Life without RNase P p.120

version="1.0" encoding="utf-8"?> ]>

Nature 453 7191 20080501 1201234 0028-0836 1476-4687 2008Nature Publishing Group Supplementary Figures

The file contains Supplementary Figures S1-S2 wit Legends.

Life without RNase P LennartRandauL ImkeSchröderI DieterSöllD Department of Molecular Biophysics and Biochemistry, Department of Chemistry, Yale University, New Haven, Connecticut 06520-8114, USA Department of Microbiology, Immunology, and Molecular Genetics, University of California, Los Angeles, California 90095, USA Correspondence and requests for materials should be addressed to D.S. (dieter.soll@yale.edu). &e080501-5; &nature06833-s1;

The universality of ribonuclease P (RNase P), the ribonucleoprotein essential for transfer RNA (tRNA) 5′ maturation, is challenged in the archaeon Nanoarchaeum equitans. Neither extensive computational analysis of the genome nor biochemical tests in cell extracts revealed the existence of this enzyme. Here we show that the conserved placement of its tRNA gene promoters allows the synthesis of leaderless tRNAs, whose presence was verified by the observation of 5′ triphosphorylated mature tRNA species. Initiation of tRNA gene transcription requires a purine, which coincides with the finding that tRNAs with a cytosine in position 1 display unusually extended 5′ termini with an extra purine residue. These tRNAs were shown to be substrates for their cognate aminoacyl-tRNA synthetases. These findings demonstrate how nature can cope with the loss of the universal and supposedly ancient RNase P through genomic rearrangement at tRNA genes under the pressure of genome condensation.

The transcription of tRNA genes generates precursor tRNA molecules with extended 5′ and 3′ termini that are processed into mature tRNA. The maturation of the 5′ terminus of a pre-tRNA relies solely on the endonucleolytic activity of the ubiquitous and essential ribonuclease P (RNase P). Both the substrate-binding domain and the active site are located in the RNA moiety of RNase P. Additional protein subunits serve as cofactors required for in vivo activity whereas in vitro cleavage was shown to be catalysed by the RNase P RNA molecule alone in all three domains of life.The number of required protein cofactors varies from one in Bacteria to up to ten subunits in Eukarya; four to five homologues of the eukaryal RNase P proteins exist in Archaea.

Surprisingly, analysis of genome sequence failed to reveal genes for RNase P in N. equitans, several species of Pyrobaculum and the bacterium Aquifex aeolicus. Although previous studies indicated that the missing bacterial enzyme is possibly replaced by an alternative activity reminiscent of bacterial RNase E-like enzymes, no explanation has been found for the obvious lack of RNase P in these Archaea. Here, we focus on the analysis of the absence of RNase P in N. equitans and the genomic rearrangements necessary to cope with its loss.

First we used a computational approach to verify that no RNase P RNA is present in the genome of N. equitans. An algorithm was developed which, based on the observation that non-coding RNA genes are easily identified in AT-rich hyperthermophiles, searches for GC-rich islands within the AT-rich genome of N. equitans. This algorithm clearly identifies all tRNAs and ribosomal RNA, as well as certain small nucleolar RNAs. However, no RNase P RNA, complete or fragmented, was discovered. Second, we investigated whether RNase P activity is detectable in cell extracts or total RNA extracts of N. equitans. A pre-tRNA based on the N. equitans initiator tRNA was produced, internally labelled in the presence of [&agr;-32P]ATP and used as substrate for RNase P cleavage assays (Fig. 1a). The 15-nucleotide leader sequence was easily released by Escherichia coli RNase P RNA, but no cleavage activity was observed with the cell extract or total RNA preparation of N. equitans (Fig. 1a). Assay conditions including time, pH, salt concentration and length of leader sequence of the pre-tRNA substrate were varied without yielding any detectable activity (data not shown).

The lack of RNase P activity raised the question of how an organism could live without RNase P and led us to investigate whether tRNAs are produced without 5′ leader sequences. This hypothesis would require that transcription of all tRNA species starts at the 5′ terminus of the mature tRNA. Therefore we analysed the promoter placement upstream of all tRNA (and tRNA half) genes in the genome of N. equitans. It is striking that each tRNA gene possesses a potential promoter positioned 26 nucleotides upstream of the mature tRNA sequence, which was shown to be the conserved distance between promoter and transcription start site in Archaea (Fig. 2). The strict distance of these tRNA gene promoters to the tRNA gene start allows the analysis of its architecture by a WebLogo representation. Apart from a described hexanucleotide TTTAAA motif that displays an invariable T at position -26, conserved adenosines are found at positions -32 and -31 and thymidine residues are conserved at positions -28, -27 and -20. A similar, but slightly less conserved, promoter placement is observed for Pyrobaculum species (Supplementary Fig. 1).

This strict promoter placement probably allows production of leaderless transcripts. Transcription is predominantly initiated with a purine (G > A) in all living organisms. Thus G is the first nucleotide in nearly all tRNAs from these organisms; however, close examination of all tRNA genes revealed a few exceptions. Therefore we determined the sequences of the 5′ termini in the tRNA population of N. equitans. This was accomplished by circularizing the tRNAs by ligating the 3′ CCA end of the mature tRNAs to the 5′ end of the same molecule. Specific tRNAs were amplified by polymerase chain reaction with reverse transcription (RT–PCR) using oligonucleotides complementary to the anticodon region of the tRNA molecule, which allowed the subsequent sequencing of the ligation site. The first tRNA of interest was tRNATyr, which requires a C1–G72 base pair (bp) as the major identity element for tyrosyl-tRNA synthetase. Sequencing revealed that the nanoarchaeal tRNATyr displays one additional G residue at position -1 at its 5′ terminus (Fig. 1b). We confirmed that this tRNA is a substrate for the N. equitans TyrRS, whereas a tRNA transcript without the G-1 residue could not be tyrosylated (Fig. 1c). The phosphorylation state of the 5′ end of the G-1 containing tRNA had no significant effect on aminoacylation activity. Another exception is N. equitans tRNAiMet, which contains C1 and an extra A-1 at its 5′ end (Fig. 1b). This is of special interest as all known archaeal initiator tRNAs contain an A1–U72 bp required for recognition by initiation factor aIF&ggr;. This RNA recognition element is shifted to positions -1  ˙ 73 in the N. equitans tRNA to form an eight-bp acceptor stem (Fig. 1b). The unique elongated acceptor stem of this initiator tRNA does not affect the methionylation by the N. equitans methionyl-tRNA synthetase (Fig. 1c). It is pertinent to note that such an eight-bp acceptor stem would be impossible if a conventional RNase P was present, as E. coli RNase P catalysed cleavage between nucleotides A-1 and C1 in an N. equitans tRNAiMet transcript (Fig. 1a). Similarly, studies on tRNATyr maturation indicate that an organism with RNase P would not generate tRNAs with the observed 5′ purine extensions. Finally, the nanoarchaeal tRNAHis contains the conserved G-1 residue required for recognition by the histidyl-tRNA synthetase. The universally occurring eight-bp acceptor stem of tRNAHis demonstrates that the protein biosynthesis machinery is able to deal with an extra 5′ nucleotide. Thus three tRNAs exist with extended 5′ termini, so that transcription can initiate with a purine. This extension is in agreement with a shift of the conserved hexanucleotide promoter element from -26 to -27 nucleotides upstream of position 1 in relation to its tRNA gene (Fig. 2).

Transcription initiation at the mature 5′ termini of nanoarchaeal tRNAs would result in triphosphorylated 5′ termini (for example found in the initiator tRNA of Halobacterium volcanii) rather than in monophosphorylated 5′ termini generated by RNase P cleavage. To address whether this is the case in vivo, we took advantage of the specificity of the vaccinia virus capping enzyme activity. This guanylyltransferase transfers radioactively labelled GTP to a ppp-5′ terminus, but not to a p-5′ terminus, of an RNA molecule. No radioactivity was transferred to the total small RNA of Methanopyrus kandleri, an archaeon that contains RNase P. However, labelling was observed in N. equitans, as witnessed by a band that co-migrates with a labelled tRNA transcript (Fig. 3). Various RNA substrates of P. aerophilum, including tRNA-sized molecules, were also labelled (Fig. 3). The labelled band from N. equitans was excised from the gel, and the RNA was eluted. RT–PCR and sequencing confirmed the presence of tRNAs. Thus, 5′ triphosphorylated tRNA species are present among small RNAs extracted from N. equitans cells. The instability of these 5′ triphosphate ends probably accounts for the fraction of 5′ monophosphorylated tRNAs that could be circularized without phosphatase pretreatment, as discussed above. It has been shown that both 5′ mono- and triphosphorylated suppressor tRNAs are equally active in protein biosynthesis. This raises the question of whether both phosphorylation states of tRNA are also accepted by the N. equitans ribosome.

The presence of a catalytic RNA moiety and the universal conservation of RNase P imply that this enzyme is ancient, possibly a remnant of the RNA world, and therefore present in the last universal common ancestor. An interesting question is why N. equitans lost RNase P. Because the N. equitans ancestor probably had leader-containing tRNAs and RNase P, we have to consider two driving forces of nanoarchaeal evolution that led to their absence. First, N. equitans has the smallest genome with the highest coding density of any sequenced genome. It is likely that this genome reduction is a strategy of adaptation to the obligatory parasitic lifestyle of N. equitans. Second, tRNA molecules are hotspots for integrative elements, as witnessed by the nearly exclusive usage of tRNA genes as attachment sites for viruses. Clearly, an organism with leader-containing tRNAs but lacking RNase P is not viable. We propose an intermediate situation that led to the complete loss of RNase P (Fig. 4). First, the activity of RNase P was reduced either by the loss of certain protein subunits of the ribonucleoprotein complex during genome reduction or by mutation of the RNA molecule. The fast evolutionary tempo of N. equitans and the integration/excision events of mobile elements at the tRNA genes allowed genome rearrangement and promoter placement that made RNase P obsolete. The organism adapted to this situation by allowing a certain ambiguity of the 5′ termini of the tRNAs, as witnessed by extended and triphosphorylated tRNAs. When every tRNA gene promoter allowed the generation of leaderless tRNA transcripts, the entire RNase P molecule was lost. Consequently, this situation prevents the insertion of integrative elements in N. equitans and therefore the expansion of its genome. Pyrobaculum shares the fast evolutionary clock owing to its ‘mutator’ phenotype and might represent the described intermediate state. Although most tRNA genes in Pyrobaculum contain a promoter in the conserved distance to the transcription start site, the promoter placement suggests the presence of small 5′ leaders for a few pre-tRNAs (Supplementary Fig. 1). Furthermore, a conventional RNase P is not detectable and only one protein subunit can be identified.

Because tRNA is the typical substrate of RNase P, it is possible that tRNA and tRNA genes predate RNase P. The emergence of RNase P would then coincide with a more flexible tRNA promoter placement or tRNA gene duplications. Genome reorganization and horizontal gene transfer are sometimes associated with recombination at tRNA genes. Such genomic perturbations allow an increased evolutionary tempo. Thus RNase P may have coevolved with an increased rate of horizontal gene transfer that facilitated the evolution of the genetic code.

Methods Summary

Oligonucleotide synthesis and DNA sequencing was performed by the Keck Foundation Biotechnology Resource Laboratory at Yale University. N. equitans cells were obtained from K. O. Stetter and M. Thomm.

Preparation and purification of RNA transcripts

The N. equitans initiator pre-tRNA gene was cloned with the leader sequence 5′-GGUUAUAACUUACU-3′ into a pUC19 vector that allowed for in vitro T7 RNA polymerase run-off transcription after plasmid cleavage with NsiI. The tRNA was internally labelled in the presence of [&agr;-32P]ATP and purified as described. RNase P RNA was synthesized in vitro from the plasmid pJA2′ provided by C. Guerrier-Takada and S. Altman.

Vaccinia virus capping enzyme assay

Five micrograms of the total RNA of M. kandleri, N. equitans and P. aerophilum were incubated with 5 U of capping enzyme (Ambion) and 56 pmol [&agr;-32P]GTP (800 Ci mmol-1) for 1 h at 37 °C according to the manufacturer’s instruction. The RNA was extracted with phenol:chloroform, ethanol precipitated in the presence of glycogen, washed with 70% ethanol, dried and run on a 12% acrylamide gel containing 8 M urea.

Preparation of cell lysates and RNA cleavage assays

P. aerophilum and N. equitans cells were each re-suspended in buffer containing 50 mM Tris·HCl (pH 7.5), 500 mM NaCl and 3 mM DTT, broken by sonication and subsequently centrifuged for 30 min at 30,000g. RNase P cleavage assays were performed with radioactively labelled tRNA transcript for 30 min at 37 °C in buffer containing 50 mM HEPES (pH 7.5), 20 mM MgCl2 and 500 mM NH4Cl, either using E. coli RNase P RNA transcript or cell extracts. Cleavage assays with archaeal cell extracts were performed with variation of the incubation temperature from 37 °C to 80 °C, incubation time from 10 min to 3 h, and with variation of salt concentration from 10 to 100 mM MgCl2 and from 100 to 1,500 mM NH4Cl.

Preparation and purification of RNA transcripts

The genes for N. equitans tRNAiMet, tRNAiMet &Dgr;A-1, tRNATyr and tRNATyr &Dgr;G-1 (including a hammerhead ribozyme to release a tRNA with a 5′-terminal C residue) were cloned into the pUC19 vector, which allowed for in vitro T7 RNA polymerase run-off transcription after plasmid cleavage with NsiI (tRNAiMet) or BstNI (tRNATyr). The tRNA transcripts were produced and purified as described. 5′ monophosphorylated tRNA was generated by incubating 10 µg tRNA transcript with 20 units of Antarctic Phosphatase (NEB) at 37 °C for 1 h followed by incubation with 20 units of T4 polynucleotide kinase (NEB) and 10 mM ATP at 37 °C for 1 h.

Aminoacyl-tRNA synthetase aminoacylation assay

The N. equitans metS gene (NEQ457) and tyrs gene (NEQ389) were amplified by PCR from genomic DNA. The metS gene was cloned into the XhoI/BlpI site of pET15b (Invitrogen) and the tyrs gene was cloned into the NdeI/XhoI site of pET20b (Invitrogen) to facilitate expression of the proteins in the E. coli BL21-Codon Plus (DE3)-RIL strain (Stratagene). Cultures were grown at 37 °C in Luria–Bertani medium supplemented with 100 µg ml-1 ampicillin and 34 µg ml-1 chloramphenicol. Expression of the recombinant proteins was induced by autoinduction. Cells were re-suspended in buffer containing 50 mM HEPES (pH 7.5) and 500 mM NaCl, and broken by sonication. The fractions were flocculated at 80 °C for 30 min and centrifuged for 30 min at 20,000g. Aminoacylation was performed at 50 °C in a 50 µl reaction with 50 mM HEPES (pH 7.0), 50 mM KCl, 4 mM ATP, 15 mM MgCl2 and 3 mM DTT. The methionylation assays contained 4 &mgr;M tRNA transcript, 15 µM [35S]methionine (more than 1,000 Ci mmol-1, diluted 1:100 with cold methionine) and 100 nM MetRS. The tyrosylation assays contained 2 &mgr;M tRNA transcript, 20 µM [14C]tyrosine (483 mCi mmol-1) and 100 nM TyrRS. Portions of 10 µl were removed at the time intervals indicated in Fig. 3, and radioactivity measured as described.

Reverse transcription and sequencing

Purification of total small RNA from P. aerophilum and N. equitans cells, circularization, reverse transcription with Thermoscript (Invitrogen) reverse transcriptase at 70 °C, amplification and sequencing were performed as described. The following primers were used for the amplification of tRNAGly(TCC): forward 5′-GCGCCCGTAGTCTAGTGGTAGGATG-3′, reverse 5′-TGCGCCCGCCGGGATTCGAACCCGG-3′; tRNATyr: forward 5′-GCCGGGCGTAGCTCAGCGGCAGAG-3′, reverse 5′-TCCGGGCGGGGGGATTCGAACCC-3′; and tRNAThr (TGT): forward 5′-GCCCCGGTAGCTCAGCGGCAGAGCG-3′, reverse 5′-AGCCCCGGGCGGGATTCGAACCC-3′. Circularization of total tRNA of N. equitans and P. aerophilum was performed with Thermophage ssDNA ligase (Prokaria) at 65 °C according to the manufacturer’s directions. The following primers were used for the amplification of circularized tRNAiMet: forward 5′-TCATAACCCCCAGGTCCCCGGTTCAAATCCG-3′, reverse 5′-TATGAGCCCCCCGGGCACTCCAGGCTGCC-3′; tRNATyr: forward 5′-TGTAGACCGGCAGGTCGGGGGTTCGAATCCC-3′, reverse 5′-CTACAGCCGGCCGCTCTGCCGCTGAGC-3′ and tRNAHis: forward 5′-TGTGGGACCCGGAGGTCCCGGGTTCGAATCCC-3′, reverse 5′-CCACAGCCCGGCGCTCTGCCACTAAGC-3′.

RNase P cleavage assay and tRNA genes in N. equitans.

a, RNase P cleavage assay. The internally radioactively labelled pre-tRNAiMet transcript (pre-Met) is cleaved in the presence of E. coli RNase P RNA (+) but not in the presence of cell extract of N. equitans (Ne). The cleavage product (iMet*) was isolated and sequenced. b, N. equitans tRNA sequences. The RT–PCR products were separated on a 3% ethidium-bromide-stained agarose gel and compared with a PCR marker with the 100-bp fragment indicated. The sequenced termini of the tRNAs are shown with numbering according to ref. 30. c, Methionylation and tyrosylation of 5′ extended tRNA transcripts (filled squares) at 50 °C. Control reactions without tRNA (filled circles), 5′ monophosphorylated tRNATyr (grey diamonds) and tRNA transcripts lacking the base at position -1 (grey triangles) are included. Values are the average from three assays; error bars, s.d.

Strict tRNA gene promoter placement in N. equitans.

The upstream region of all 44 tRNA genes (including 5′ tRNA halves) was aligned. Position 1 indicates the first tRNA nucleotide. Position -1 of three tRNA species is boxed in red. The conserved promoter element is visualized by a WebLogo in which the size of a letter indicates the degree of conservation of a nucleotide.

Detection of triphosphorylated tRNA.

Equal amounts of total small RNA from M. kandleri (Mk), N. equitans (Ne), P. aerophilum (Pa) and an N. equitans tRNATyr in vitro transcript (IVT) were separated on a 3% ethidium-bromide-stained agarose gel (left). These RNA preparations were reacted with vaccinia virus capping enzyme, the products loaded onto a denaturing 12 % polyacrylamide gel, and the transfer of radioactive GTP onto triphosphorylated RNA monitored by autoradiography (right). Sequencing samples of the gel-eluted labelled RNA of N. equitans verified the presence of tRNA. Full-length gels are presented in Supplementary Fig. 2.

A scenario that allows the loss of RNase P.

The pre-tRNA leader sequences are indicated in red; RNase P proteins are indicated in blue.

We thank P. O’Donoghue, J. Yuan and L. Sherrer for help and encouragement. This work was supported by grants from the National Institute of General Medical Sciences and the Department of Energy (D.S.) and the National Science Foundation (I.S.).

Altman, S. et al. Catalysis by the RNA subunit of RNase P — a minireview. Gene 82, 6364 (1989) Evans, D., Marquez, S. M. & Pace, N. R. RNase P: interface of the RNA and protein worlds. Trends Biochem. Sci. 31, 333341 (2006) Vögeli, G., Grosjean, H. & Söll, D. A method for the isolation of specific tRNA precursors. Proc. Natl Acad. Sci. USA 72, 47904794 (1975) Guerrier-Takada, C., Gardiner, K., Marsh, T., Pace, N. & Altman, S. The RNA moiety of ribonuclease P is the catalytic subunit of the enzyme. Cell 35, 849857 (1983) Pannucci, J. A., Haas, E. S., Hall, T. A., Harris, J. K. & Brown, J. W. RNase P RNAs from some Archaea are catalytically active. Proc. Natl Acad. Sci. USA 96, 78037808 (1999) Kikovska, E., Svard, S. G. & Kirsebom, L. A. Eukaryotic RNase P RNA mediates cleavage in the absence of protein. Proc. Natl Acad. Sci. USA 104, 20622067 (2007) Tsai, H. Y., Pulukkunat, D. K., Woznick, W. K. & Gopalan, V. Functional reconstitution and characterization of Pyrococcus furiosus RNase P. Proc. Natl Acad. Sci. USA 103, 1614716152 (2006) Li, Y. & Altman, S. In search of RNase P RNA from microbial genomes. RNA 10, 15331540 (2004) Willkomm, D. K., Feltens, R. & Hartmann, R. K. tRNA maturation in Aquifex aeolicus. Biochimie 84, 713722 (2002) Klein, R. J., Misulovin, Z. & Eddy, S. R. Noncoding RNA genes identified in AT-rich hyperthermophiles. Proc. Natl Acad. Sci. USA 99, 75427547 (2002) Randau, L., Münch, R., Hohn, M. J., Jahn, D. & Söll, D. Nanoarchaeum equitans creates functional tRNAs from separate genes for their 5′- and 3′-halves. Nature 433, 537541 (2005) Palmer, J. R. & Daniels, C. J. In vivo definition of an archaeal promoter. J. Bacteriol. 177, 18441849 (1995) Crooks, G. E., Hon, G., Chandonia, J. M. & Brenner, S. E. WebLogo: a sequence logo generator. Genome Res. 14, 11881190 (2004) Hausner, W. & Thomm, M. in Archaea, Evolution, Physiology and Molecular Biology (eds. Garrett, R. A. & Klenk, H. P.) 185198 (Blackwell, Malden, Massachusetts, 2007) Kobayashi, T. et al. Structural basis for orthogonal tRNA specificities of tyrosyl-tRNA synthetases for genetic code expansion. Nature Struct. Biol. 10, 425432 (2003) Farruggio, D., Chaudhuri, J., Maitra, U. & RajBhandary, U. L. The A1  ˙ U72 base pair conserved in eukaryotic initiator tRNAs is important specifically for binding to the eukaryotic translation initiation factor eIF2. Mol. Cell. Biol. 16, 42484256 (1996) Marck, C. & Grosjean, H. tRNomics: analysis of tRNA genes from 50 genomes of Eukarya, Archaea, and Bacteria reveals anticodon-sparing strategies and domain-specific features. RNA 8, 11891232 (2002) Connolly, S. A., Rosen, A. E., Musier-Forsyth, K. & Francklyn, C. S. G-1:C73 recognition by an arginine cluster in the active site of Escherichia coli histidyl-tRNA synthetase. Biochemistry 43, 962969 (2004) Gupta, R. Halobacterium volcanii tRNAs. Identification of 41 tRNAs covering all amino acids, and the sequences of 33 class I tRNAs. J. Biol. Chem. 259, 94619471 (1984) Shuman, S., Surks, M., Furneaux, H. & Hurwitz, J. Purification and characterization of a GTP-pyrophosphate exchange activity from vaccinia virions. J. Biol. Chem. 255, 1158811598 (1980) Noren, C. J. et al. In vitro suppression of an amber mutation by a chemically aminoacylated transfer RNA prepared by runoff transcription. Nucleic Acids Res. 18, 8388 (1990) Walker, S. C. & Engelke, D. R. Ribonuclease P: the evolution of an ancient RNA enzyme. Crit. Rev. Biochem. Mol. Biol. 41, 77102 (2006) Waters, E. et al. The genome of Nanoarchaeum equitans: insights into early archaeal evolution and derived parasitism. Proc. Natl Acad. Sci. USA 100, 1298412988 (2003) Das, S., Paul, S., Bag, S. K. & Dutta, C. Analysis of Nanoarchaeum equitans genome and proteome composition: indications for hyperthermophilic and parasitic adaptation. BMC Genomics 7, 186 (2006) She, Q., Shen, B. & Chen, L. Archaeal integrases and mechanisms of gene capture. Biochem. Soc. Trans. 32, 222226 (2004) Altman, S., Kirsebom, L. & Talbot, S. Recent studies of ribonuclease P. FASEB J. 7, 714 (1993) Fitz-Gibbon, S. T. et al. Genome sequence of the hyperthermophilic crenarchaeon Pyrobaculum aerophilum. Proc. Natl Acad. Sci. USA 99, 984989 (2002) Hartmann, E. & Hartmann, R. K. The enigma of ribonuclease P evolution. Trends Genet. 19, 561569 (2003) Randau, L. et al. The heteromeric Nanoarchaeum equitans splicing endonuclease cleaves non-canonical bulge–helix–bulge motifs of joined tRNA halves. Proc. Natl Acad. Sci. USA 102, 1793417939 (2005) Sprinzl, M. & Vassilenko, K. S. Compilation of tRNA sequences and sequences of tRNA genes. Nucleic Acids Res. 33, D139D140 (2005) Fechter, P., Rudinger, J., Giegé, R. & Théobald-Dietrich, A. Ribozyme processed tRNA transcripts with unfriendly internal promoter for T7 RNA polymerase: production and activity. FEBS Lett. 436, 99103 (1998) Studier, F. W. Protein production by auto-induction in high density shaking cultures. Protein Expr. Purif. 41, 207234 (2005)

doi: 10.1038/nature06833


Structural analysis of the essential self-cleaving type III secretion proteins EscU and SpaS p.124

version="1.0" encoding="utf-8"?> ]>

Nature 453 7191 20080501 1241274 0028-0836 1476-4687 2008Nature Publishing Group Supplementary Information

The file contains Supplementary Tables 1-2 and Supplementary Figures 1-13 with Legends. The Supplementary Tables describe the crystallization and the structure determination statistics. The Supplementary Figures describe additional biochemical data, structural analysis and circular dichroism experiments.

Structural analysis of the essential self-cleaving type III secretion proteins EscU and SpaS RazZarivachR WanyinDengW MarijaVuckovicM Heather B.FeliseH B Hai V.NguyenH V Samuel I.MillerS I B. BrettFinlayB B Natalie C. J.StrynadkaN C J Department of Biochemistry and Molecular Biology, and the Center for Blood Research, University of British Columbia, 2350 Health Sciences Mall, Vancouver, British Columbia V6T 1Z3, Canada Michael Smith Laboratories, University of British Columbia, 2185 East Mall, Vancouver, British Columbia V6T 1Z4, Canada Department of Microbiology and Medicine, HSB K-140, Box 357710, Seattle, Washington 98195, USA Correspondence and requests for materials should be addressed to N.C.J.S. (natalie@byron.biochem.ubc.ca). &nature06832-s1;

During infection by Gram-negative pathogenic bacteria, the type III secretion system (T3SS) is assembled to allow for the direct transmission of bacterial virulence effectors into the host cell. The T3SS system is characterized by a series of prominent multi-component rings in the inner and outer bacterial membranes, as well as a translocation pore in the host cell membrane. These are all connected by a series of polymerized tubes that act as the direct conduit for the T3SS proteins to pass through to the host cell. During assembly of the T3SS, as well as the evolutionarily related flagellar apparatus, a post-translational cleavage event within the inner membrane proteins EscU/FlhB is required to promote a secretion-competent state. These proteins have long been proposed to act as a part of a molecular switch, which would regulate the appropriate chronological secretion of the various T3SS apparatus components during assembly and subsequently the transported virulence effectors. Here we show that a surface type II &bgr;-turn in the Escherichia coli protein EscU undergoes auto-cleavage by a mechanism involving cyclization of a strictly conserved asparagine residue. Structural and in vivo analysis of point and deletion mutations illustrates the subtle conformational effects of auto-cleavage in modulating the molecular features of a highly conserved surface region of EscU, a potential point of interaction with other T3SS components at the inner membrane. In addition, this work provides new structural insight into the distinct conformational requirements for a large class of self-cleaving reactions involving asparagine cyclization.

The EscU family of proteins is an essential component of the inner membrane ring of T3SS systems. It is thought to be composed of approximately half a dozen highly conserved proteins (EscJNRSTUV in enteropathogenic E. coli). Earlier primary sequence analysis of the EscU and orthologous FlhB (flagellar) family of proteins predicted two major domains: an approximate 200-amino-acid amino (N)-terminal domain (NTD) that is involved in the association to the T3SS inner membrane basal ring; and an approximate 100-amino-acid carboxy (C)-terminal domain (CTD), which is known to undergo auto-cleavage and to mediate a switch in the secretion substrate profile of the T3SS (Supplementary Fig. 1). The two domains are connected by a proposed 30-amino-acid flexible linker, which is highly conserved (more than 80%) within the EscU/SpaS and FlhB family of proteins (Supplementary Fig. 2). In this study, constructs encompassing the predicted linker and CTD switch domain from the T3SS EscU protein of enteropathogenic E. coli (EPEC), residues 215–345 termed EscU&Dgr;214, and the SpaS protein from Salmonella typhimurium, residues 211–356 termed SpaS&Dgr;210, were purified and crystallized. Several crystal forms from the two species were obtained with diffraction up to a resolution of 1.2 Å (Fig. 1a; for simplicity, EscU numbering is used throughout the text).

The crystallographic data collectively show that the CTD domains are highly conserved in structure (despite the modest approximate 30% sequence identity of the region) (Supplementary Fig. 3). The root mean square deviations between all CTD structures of EscU from seven different space groups are approximately 0.6 Å for the 93 backbone atoms. A root mean square deviation of 1.0 Å was observed for the 94 common backbone atoms of EPEC EscU and Salmonella SpaS, indicating that the CTD has a compact stable fold including highly similar conformations of the several surface loops (Fig. 1). The only significant deviation between the CTDs of EscU and SpaS is the elongated C-terminal &agr;-helix, which is extended by a 15-residue sequence insertion in SpaS. As supported by light scattering analysis and mass spectrometry (data not shown), the EscU and SpaS CTD constructs are strongly monomeric both in solution and in our several crystal lattices.

EscU contains a cleavage site conserved within all T3SS and flagellar orthologues. The site of auto-cleavage was verified structurally here and is localized to an exposed region between strands &bgr;1 and &bgr;2 containing a highly conserved sequence quartet observed in all T3SS and flagellar EscU/FlhB variants (NPTH defined by Asn262, Pro263, Thr264, His265 in EscU). We also show that purification of the CTD of EscU in the absence of the membrane anchoring NTD did not obliterate the self-cleavage process as observed directly in the high-resolution structures (Figs 1 and 3a). In our structures, self-cleavage uniformly occurred at the predicted Asn262–Pro263 peptide, liberating the termini of the newly created peptides locally at the protein surface, but clearly not interfering with the overall integrity of the protein fold (Supplementary Fig. 4).

To address the mechanism of cleavage further, we analysed six point mutations at the auto-cleavage site in EscU (N262A, N262D in two crystal forms (that is, N262D and N262Di), P263A, T264A, H265A and R313T) (Supplementary Figs 4 and 9). By overlapping the structures of our native and mutant forms of EscU, we were able to elucidate the molecular details of a self-cleavage intein-like mechanism involving asparagine cyclization (Supplementary Fig. 12). The mechanism initiates during the post-translational folding of EscU, including creation of the type II &bgr;-turn (T2&bgr;) at residues 262–265. In this initial uncleaved form, the carbonyl oxygen of the strictly conserved Asn262 (notably localized in position one of the T2&bgr; in our structures and not at position two as previously hypothesized) is hydrogen bonded to the backbone amide of His265 and Ile266. The T2&bgr; unique to the uncleaved form enforces a 90° angle between the Asn262–Pro263 amide to Pro263–Thr264 amide (Asn262, &psgr; = 112; Pro263, ϕ = -47; &psgr; = 133), creating a distance of about 2.8 Å between the carbonyl oxygen of Asn262 and the carbonyl carbon of Pro263, and an n → &pgr;* interaction which would promote electron withdrawal from the Asn262 carbonyl oxygen (Fig. 3 and Supplementary Fig. 5). Supporting the need for this strained conformation in catalysis is the observation that the cleavage-impaired mutant P263A (it promotes very low levels of cleavage only at conditions of extreme pH) adopts a type I rather than T2&bgr; turn owing to the loss of the requisite main-chain conformation imposed by the conserved proline residue (Asn262, &psgr; = 104; Pro263, ϕ = 51; &psgr; = -128). The type I conformation does not promote the favourable hydrogen bonding and n → &pgr;* interactions that we propose are essential to efficient catalysis. As the reaction initiates, the side-chain N&dgr; of Asn262 is exposed on the protein surface and enables water molecules to absorb a proton to promote the attack of the N&dgr; lone pair on the partly positively charged carbonyl carbon, creating a tetrahedral intermediate. The negative charge that develops on the Asn262 carbonyl oxygen in this intermediate form is stabilized by hydrogen bonds to the backbone amides of His265/Ile266 and by electron withdrawal by the n → &pgr;* interaction. Collapse of the tetrahedral intermediate and subsequent cleavage are enabled by absorbing back the proton from water to the leaving amino group at Pro263, creating a new N terminus at Pro263 and a C-terminal succinimide intermediate at Asn262. The succinimide intermediate (Supplementary Fig. 6; P263Asuc in Supplementary Table 1) can subsequently be hydrolysed to create the new C-terminal negatively charged Asn262 (Fig. 3b). We can deduce, based on our non-cleaved mutant structures, that the asparagine driven auto-cleavage mechanism relies on the trans orientation between the Asn carbonyl and the following peptide carbonyl, and stabilization of the Asn carbonyl by hydrogen bonding to an electropositive acceptor.

To assess the importance of the conserved residues (Supplementary Fig. 3) to the T3SS function, we have tested EscU and SpaS non-cleavable mutants for secretion of T3SS substrates (Fig. 2). In agreement with earlier observations in Yersinia, in EPEC and Salmonella we observe full arrest of secretion of the translocon and filament components (Fig. 2a). However, in contrast to the Yersinia study, the SpaS mutants did not alter the secretion level of the molecular ruler (InvJ), but altered the secretion level of the needle component (PrgI) (Fig. 2b). Thus, our data suggest that EscU/SpaS act in the recognition of the needle components and as part of a switch between the secretion of the needle and the T3SS translocon (SipB/C in Salmonella) or filament (EspA in E. coli) components.

In addition to electrostatic features (Supplementary Fig. 7), additional cleavage-induced conformational features likely govern the cleaved, T3SS competent state. The breaking of the amide bond between Asn262 to Pro263 leads to a conformational change of the newly created N- and C-terminal ends from their positions within the T2&bgr; pre-cleavage to an extended form with completely distinct side-chain orientations post-cleavage (Fig. 3a). The newly introduced N- and C-terminal charges also contribute to the altered electrostatic profile of this highly conserved surface. The dispositions of critical side chains in the loop are also significant. For example, we have shown that an H265A mutant is fully cleaved but is unable to promote secretion (Fig. 2a), implying a direct role for the immidazole side chain in the secretion process. In the cleaved form the side chain of Asn262 is hydrogen bonded to the backbone of the now flipped strand containing Thr264 to Ile266, creating a more extended &bgr;-chain and effectively re-orienting the critical His265 by about 180°, transplanting it from one face of EscU to another (Fig. 3a). Post-cleavage, the His265 immidazole side chain sits at the centre of a newly created surface formed by (in addition to the N- and C-terminal charges at positions 262 and 263) the conserved Arg313, Leu292 and positively charged Lys261 (EscU) or Arg295 (SpaS). Collectively the cleavage creates a localized but electrostatically and conformationally unique surface that we predict is essential for the interaction with other T3SS components.

EscU and orthologues in other T3SSs, as well as the flagellar system, have been implicated in acting as part of a molecular ‘switch’ that regulates the chronological secretion of apparatus components and virulence effectors (detailed in Supplementary Fig. 10). Although our structural and mutational data of the conserved cleavage products show them to be attractive interaction sites for the other T3SS components globally constituting the inner membrane ring, and our thermal denaturation data directly indicate that cleavage is a critical event for stabilization of these inner membrane proteins (Supplementary Fig. 11), it is clear that the post-translational auto-cleavage, which occurs immediately after translation and folding in the cytoplasm, although essential to secretion, is not the switching event per se.

Our structural and biochemical data point to the conserved linker region of EscU orthologues for playing a potential role in the switching events. The proposed linker region (EscU 215–245 and SpaS 210–241) connects the N-terminal membrane-anchored domain with the globular CTD. Although earlier sequence analysis predicted a helical linker, all our structures show a large part of the linker is unstructured. Overlay of the SpaS and EscU structures (Fig. 1b) indicates conformational differences and potential flexibility in the linker that centre on a hinge-like region between the end of the linker and the initial helix of the CTD (Fig. 1b). This hinge provides a potential point of conformational freedom for the CTD from the membrane anchored N-terminal domain of EscU and its associated T3SS partners. One can envisage such a hinging motion could allow for the binding and dissociation of the EscU CTD from the T3SS inner membrane base. This would effectively promote its role as a molecular switch by appropriately changing the entrance shape, size and electrostatic nature of the T3SS pore as downstream T3SS components and effectors are bound and secreted in the specific chronology that has been previously observed.

To analyse the importance of the observed linker in T3SS function, we used our structure to design three deletion mutations (EscU(-5) 232–236, EscU(-11) 234–245, EscU(-16) 230–245) and two point mutations (G229P and G235P) in the flexible hinge region of EscU. These mutants were screened in vivo for secretion of the filament component EspA, the translocon components (EspB/EspD) and for the effector Tir in &Dgr;escU and in wild-type (WT) backgrounds (Fig. 2, and Supplementary Figs 8 and 13). In all cases, alteration of the linker length abolishes the secretion of the translocon components and effector in the &Dgr;escU knockout and decreased their secretion in the WT strain (a finding also consistent with previous mutagenesis studies in the flagellar FlhB family). Furthermore, G229P and G235P show dramatic alteration of T3SS secretion (Supplementary Fig. 13). Given our observations that these EscU mutants are expressed and folded properly in vivo (owing to their ability to perform auto-cleavage), the observed in vivo loss of function of the linker deletions and point mutations supports the proposed requirement for the appropriate juxtaposition of the auto-cleavable CTD relative to the membrane-bound NTD of EscU for the binding, localization and secretion of the T3SS assembly and effector molecules.

Methods Summary

Protein purification, site-directed mutagenesis and secretion assays were performed as previously published with small alterations (see Methods). Crystallization and crystallographic statistics are in Supplementary Tables 1 and 2.

Protein expression and purification

For purification of His-tagged SpaS&Dgr;210, EscU&Dgr;214 and all mutants, the expression construct pEscU/pSpaS was generated by PCR-amplifying the escU/spaS open reading frame from EPEC E2348/69 and S. typhimurium genomic DNA, and cloning it into the NheI/BamHI and NheI/HindIII sites of the pET-28(a) expression vector (Novagen). E. coli BL21 (DE3) transformed with pEscU/pSpaS was grown to mid-exponential phase at 37 °C in Luria-Bertani broth containing kanamycin and induced with 0.5 mM isopropyl &bgr;-D-1-thiogalactopyranoside. Cells were harvested after incubation for 16 h at 20 °C, re-suspended in buffer (20 mM Tris, 500 mM NaCl, pH 8.0), lysed using a French press, and centrifuged at approximately 60,000g for 35 min. His-tagged EscU/SpaS was purified from the soluble fraction using nickel-chelating Sepharose (Amersham). Protein was eluted using the buffer (20 mM Tris, 40 mM NaCl, 300 mM immidazole, pH 8.0) and the tag cleaved at 4 °C overnight with 1:1,000 thrombin (Sigma), leaving six extra residues at the N terminus (Gly–Ser–His–Met–Ala–Ser). Cleaved product was purified by Mono-Q anion-exchange (Pharmacia) using a linear gradient of NaCl in 20 mM Tris–HCl buffer, pH 8.0.

Site-directed mutagenesis by PCR

EscU mutants were generated using QuikChange site-directed mutagenesis (coding and antisense primers containing a single mutagenic site were used for PCR amplification).

Crystallization and structure determination

Purified EscU&Dgr;214/SpaS&Dgr;210 (at 15–20 mg ml-1) crystallized into various forms in a variety of conditions. Crystals were grown at 21 °C using sitting-drop vapour diffusion or under oil microbatch by mixing 0.5 &mgr;l of protein with the same volume of reservoir solution (Supplementary Table 2). Crystal screening and data collection were done at Beamline 8.2.2 of the Advanced Light Source, 11-1 of the Stanford Synchrotron Radiation Laboratory, and Cu K&agr; at home source. Data were reduced and scaled using the HKL2000 suite. For phasing, EscU&Dgr;214 crystals were soaked into 2 M NaI 2 M Na formate for 2 min and flash freezed in liquid nitrogen. Positions of all the seven iodide sites were found and refined by ShelX-C. Phase calculation and solvent flattening were done using ShelX-DE, which resulted in an interpretable map. The model was built using Coot, and refinement was done using Refmac after excluding 5% of the data for the R-free calculation. SpaS&Dgr;210 and EscU&Dgr;214 mutants were solved by molecular replacement using Phaser. Some residues were not observed in the crystal (Supplementary Table 1). Structural figures were prepared with PyMOL and electrostatic calculations were done with the APBS plugin.

Generation of the &Dgr;escU mutant in EPEC

The sacB gene-based allelic exchange method was used to generate an escU in-frame deletion mutant in the streptomycin-resistant derivative (Smr) of EPEC O127:H6 strain E2348/69 using the suicide vector pRE112. To make a deletion mutant of escU in EPEC, PCR was used to generate two fragments (1.1 and 1.3 kilobases, respectively) using primer pairs EPescU-1 (KpnI) (5′-GGGTACCTTATGTGTGCAAACGTTCTGG-3′) and DEPescU-R (NheI) (5′-CGCTAGCTTCACTTTTTGTTACATCGCC-3′) as well as DEPescU-F (NheI) (5′-CGCTAGCTTGATTCGTATTGCGATAGAC-3′) and EPescU-2 (SacI) (5′-GGAGCTCCTTCGGCAATATCATTGCGAG-3′). The PCR products were cloned into pCR2.1-TOPO (Invitrogen) and verified by DNA sequencing. After digestion with KpnI/NheI and NheI/SacI, respectively, the two fragments were gel-purified and cloned into pRE112 digested with KpnI/SacI in a three-way ligation. The resulting plasmid pRE-&Dgr;EPescU contained 1–2 kilobases of flanking regions on both sides of escU and the escU gene with an internal in-frame deletion from nucleotides 79 to 1007 (about 89% of the coding region). An NheI site was introduced into the deletion site. Plasmid pRE-&Dgr;EPescU was transformed into E. coli SM10&lgr;pir by electroporation, and introduced into EPEC strain E2348/69 Smr by conjugation. After sucrose selection, EPEC colonies resistant to sucrose and sensitive to chloramphenicol were screened for deletion of escU by PCR using primers EPescU-1 and EPescU-2. The EPEC escU mutants were further verified by multiple PCR reactions.

Expression plasmid of EPEC escU for complementation

Primers EPescU-HAF (SacI) (5′-CGAGCTCACGGCAAATATTCATTCTGAC-3′) and EPescU-HAR (XhoI) (5′-CCTCGAGATAATCAAGGTCTATCGCAATAC-3′) were used to amplify by PCR a fragment containing the coding region as well as a 55-bp upstream region of escU. The PCR product was cloned into pCR2.1-TOPO, verified by DNA sequencing, and subcloned into the vector pTOPO-2HA digested with SacI/XhoI to generate plasmid pEPescU-2HA. This plasmid expresses EscU with a double haemagglutinin tag (2HA) at the C terminus under the control of the Plac promoter on the vector, and can complement the EPEC &Dgr;escU mutant. The expression of EscU conferred by the plasmid was followed by western blotting using the haemagglutinin monoclonal antibody (Covance).

Type III secretion assay for EPEC

EPEC strains were grown in Luria-Bertani and then subcultured in Dulbecco’s Modified Eagle Media to induce type III secretion as previously described.

Strain construction for S. typhimurium

The following genetic manipulations were performed using the &lgr;-RED system. The tetracycline resistance gene tetA and its transcriptional repressor tetR were introduced into the spaS gene in WT S. typhimurium after amplification with the following primers: HF315, 5′-GGTGAAATCTGATATTGAAAACTCACGCCTGATTGTTGCCTTAAGACCCACTTTCACA; HF316, 5′-TCGGCATCAATTCGGGTTTAAAATAAATCCCGATCGTAATCTAAGCACTTGTCTCCTG-3′. The tetracycline resistance genes replaced the coding sequence for amino acids 259–261 of spaS. Allelic exchange and selection for tetracycline sensitivity was used to replace the tetracycline resistance with the N258A and P259A site-directed mutants amplified with the following primers: HF317, 5′-GATATTGAAAACTCACGCCTGATTGTTGCCGCCCCCACGCATATTACGATCGGGATTT-3′; HF318, 5′-ATAAATCCCGATCGTAATATGCGTGGGGGCGGCAACAATCAGGCGTGAGTTTTCAATATC-3′ for N258A; HF319, 5′-GATATTGAAAACTCACGCCTGATTGTTGCCAACGCCACGCATATTACGATCGGGATTTAT-3′; HF320, 5′-ATAAATCCCGATCGTAATATGCGTGGCGTTGGCAACAATCAGGCGTGAGTTTTCAATATC-3′ for N259A.

Western blots for S. typhimurium

Secreted proteins were prepared as previously described for S. typhimurium and SDS–polyacrylamide gel electrophoresis and western blot techniques were performed as described.

Needle shearing for S. typhimurium

Overnight cultures of WT S. typhimurium and the spaS mutants were grown in Luria-Bertani medium at 37 °C. Cultures were passed six times through a hypodermic needle (23Gx1, 0.6 mm × 25 mm; B. Braun) to release surface needle proteins. The cultures were centrifuged at 1,800g for 15 min, and the supernatants were precipitated with 10% trichloroacetic acid for 20 min at 4 °C. The samples were centrifuged at 1,800g for 20 min at 4 °C, and the pellets were re-suspended in 50 mM Tris, pH 8.0, and sample buffer. Equal amounts of supernatant proteins were loaded for each strain, and SDS–polyacrylamide gel electrophoresis and western blot techniques were performed as described.

Circular dichroism spectroscopy

The circular dichroism spectra were recorded with a Jasco spectropolarimeter (Model J-810) equipped with a Pelletier device. All spectra are averages of three scans, quartz cells of optical path length 0.2 cm and concentrations of 2.5 and 5 &mgr;M. EscU&Dgr;214 and mutants in buffer (0.1 mM Tris, pH 8.0) were used. For thermal denaturation experiments, melting curves were determined by monitoring the changes in dichroic density at 222 nm as a function of temperature in the range 25-90 °C and at a heating rate of 1 °C min-1. The thermodynamic parameters associated with the temperature-induced denaturation were obtained by nonlinear, least-squares analysis of the temperature dependence of circular dichroism. A two-state denaturation process was assumed during curve-fitting analyses.

Structure of the C-terminal domains of EscU and SpaS.

a, The native cleaved CTD of EscU and SpaS with a blue arrow pointing to the auto-cleavage site. CTD is a novel &agr;/&bgr;-fold with a mixed parallel and anti-parallel five-stranded twisted &bgr;-sheet (topology &bgr;4, &bgr;1, &bgr;2, &bgr;3, &bgr;5) flanked by two helices on each side (&agr;1, &agr;2, &agr;3 and &agr;4). b, Superposition of the CTD reveals a different fold for the N-terminal linker between EscU, EscU mutants and SpaS, as well as a longer C-terminal helix for SpaS.

Type III secretion in EPEC and S. typhimurium.

a, EPEC, escU mutants and complementation of type III secretion. Secretion of effector (Tir) and translocator proteins (EspB/D) checked under complementation of escU by WT and mutants. The cleavage state of EscU in vivo is presented for all EscU components. b, spaS WT, mutants and type III secretion. Secretions of effector (SipA), molecular ruler (InvJ), needle (PrgI) and translocator proteins (SipB/C) were analysed in WT and spaS mutants. The N258A and P259A substitutions in SpaS correspond to N262A and P263A respectively in EscU. Full-length blots/gels are presented in Supplementary Figure 1.

Auto-cleaving mechanism of EscU.

a, Non-cleaved type I &bgr;-bend of P263A (left), the non-cleaved type II beta bend of N262A (middle left), the non-cleaved type II beta bend of N262Di (middle right) and the native cleaved loop with a flipped His 265 (right). Note the identical conformations of the N, C, C&agr; and C&bgr; at position 262 in all uncleaved forms. All maps are sigma-A weighted 2Fo - Fc electron density (1.5&sgr;); water molecules are represented by the red spheres. b, Detailed mechanism for the asparagine cyclization in EscU.

We thank T.S., M.C. and P.I.L. for discussions; the staff at the Advanced Light Source beamline 8.2.2 and SSRL beamline 11-1 for data collection time and assistance; and E. Galyov at the Institute for Animal Health for the SipA, SipB and SipC antibodies. This work was supported by an Izaak Walton Killam Research post-doctoral fellowship, a Michael Smith Foundation for Health Research (MSFHR) post-doctoral fellowship and a Canadian Institutes of Health Research (CIHR) post-doctoral fellowship (all to R.Z.). N.C.J.S. and B.B.F. thank the Howard Hughes International Scholar program and the CIHR for funding. N.C.J.S. also thanks the MSFHR and the Canada Foundation of Innovation for infrastructure funding support. N.C.J.S. is also an MSFHR Senior Scholar and CIHR Investigator. S.M. acknowledges grants from the National Institutes of Allergy and Infectious Diseases (NIAID), U54 AI057141 and 5RO1 AI030479.

Author Contributions R.Z. and M.V. cloned and purified EscU and SpaS. R.Z. crystallized EscU and SpaS. R.Z. solved the structures. H.B.F. isolated the chromosomal SpaS mutants. W.D., H.B.F and H.V.N. performed the biochemical experiments in Fig. 2. B.B.F. and S.I.M. provided resources for the experiments shown in Fig. 2. N.C.J.S. provided resources for all data other than those in Fig. 2. All authors discussed the results and commented on the manuscript.

The atomic coordinates of all the structures have been deposited in the Protein Data Bank with the accession codes 3BZL, 3BZO, 3BZP, 3BZR, 3BZS, 3BZT, 3C03, 3BZV, 3BZX, 3BZY, 3BZZ, 3C00 and 3C01.

Hueck, C. J. Type III protein secretion systems in bacterial pathogens of animals and plants. Microbiol. Mol. Biol. Rev. 62, 379433 (1998) Macnab, R. M. How bacteria assemble flagella. Annu. Rev. Microbiol. 57, 77100 (2003) Ferris, H. U. & Minamino, T. Flipping the switch: bringing order to flagellar assembly. Trends Microbiol. 14, 519526 (2006) Sorg, I. et al. YscU recognizes translocators as export substrates of the Yersinia injectisome. EMBO J. 26, 30153024 (2007) Lavander, M. et al. Proteolytic cleavage of the FlhB homologue YscU of Yersinia pseudotuberculosis is essential for bacterial survival but not for type III secretion. J. Bacteriol. 184, 45004509 (2002) Ding, Y. et al. Crystal structure of a mini-intein reveals a conserved catalytic module involved in side chain cyclization of asparagine during protein splicing. J. Biol. Chem. 278, 3913339142 (2003) Paulus, H. Protein splicing and related forms of protein autoprocessing. Annu. Rev. Biochem. 69, 447496 (2000) Shemella, P. et al. Mechanism for intein C-terminal cleavage: a proposal from quantum mechanical calculations. Biophys. J. 92, 847853 (2007) Geiger, T. & Clarke, S. Deamidation, isomerization, and racemization at asparaginyl and aspartyl residues in peptides. Succinimide-linked reactions that contribute to protein degradation. J. Biol. Chem. 262, 785794 (1987) Stephenson, R. C. & Clarke, S. Succinimide formation from aspartyl and asparaginyl peptides as a model for the spontaneous degradation of proteins. J. Biol. Chem. 264, 61646170 (1989) Clarke, S. Propensity for spontaneous succinimide formation from aspartyl and asparaginyl residues in cellular proteins. Int. J. Pept. Protein Res. 30, 808821 (1987) Hodges, J. A. & Raines, R. T. Energetics of an n → &pgr;* interaction that impacts protein structure. Org. Lett. 8, 46954697 (2006) Ferris, H. U. et al. FlhB regulates ordered export of flagellar components via autocleavage mechanism. J. Biol. Chem. 280, 4123641242 (2005) Fraser, G. M. et al. Substrate specificity of type III flagellar protein export in Salmonella is controlled by subdomain interactions in FlhB. Mol. Microbiol. 48, 10431057 (2003) Zarivach, R., Vuckovic, M., Deng, W., Finlay, B. B. & Strynadka, N. C. Structural analysis of a prototypical ATPase from the type III secretion system. Nature Struct. Mol. Biol. 14, 131137 (2007) Otwinowski, Z. & Minor, W. Processing of X-ray diffraction data collected in oscillation mode. Macromol. Crystallogr. A 276, 307326 (1997) Schneider, T. & Sheldrick, G. Substructure solution with SHELXD. Acta Crystallogr. D 58, 17721779 (2002) Emsley, P. & Cowtan, K. Coot: model-building tools for molecular graphics. Acta Crystallogr. D 60, 21262132 (2004) Murshudov, G. N., Vagin, A. A. & Dodson, E. J. Refinement of macromolecular structures by the maximum-likelihood method. Acta Crystallogr. D 53, 240255 (1997) McCoy, A. J., Grosse-Kunstleve, R. W., Storoni, L. C. & Read, R. J. Likelihood-enhanced fast translation functions. Acta Crystallogr. D 61, 458464 (2005) PyMOL. v.1 (DeLano Scientific, Palo Alto, California, 2002) Baker, N. A., Sept, D., Joseph, S., Holst, M. J. & McCammon, J. A. Electrostatics of nanosystems: application to microtubules and the ribosome. Proc. Natl Acad. Sci. USA 98, 1003710041 (2001) Edwards, R. A., Keller, L. H. & Schifferli, D. M. Improved allelic exchange vectors and their use to analyze 987P fimbria gene expression. Gene 207, 149157 (1998) Deng, W. et al. Dissecting virulence: systematic and functional analyses of a pathogenicity island. Proc. Natl Acad. Sci. USA 101, 35973602 (2004) Datsenko, K. A. & Wanner, B. L. One-step inactivation of chromosomal genes in Escherichia coli K-12 using PCR products. Proc. Natl Acad. Sci. USA 97, 66406645 (2000) Maloy, S. & Nunn, W. Selection for loss of tetracycline resistance by Escherichia coli. J. Bacteriol. 145, 11101112 (1981) Kimbrough, T. G. & Miller, S. I. Contribution of Salmonella typhimurium type III secretion components to needle complex formation. Proc. Natl Acad. Sci. USA 97, 1100811013 (2000) Pegues, D. A., Hantman, M. J., Behlau, I. & Miller, S. I. PhoP/PhoQ transcriptional repression of Salmonella typhimurium invasion genes: evidence for a role in protein secretion. Mol. Microbiol. 17, 169181 (1995)

doi: 10.1038/nature06832

「Journal home」に戻る