project inwards to contact membrane-associ-ated adaptor proteins, which in turn recruit the cargo proteins.

COPII is a set of four proteins organized as two heterodimers: the Sec31/Sec13 pair gener-ates the lattice of a coat; and the Sec23/Sec24 dimer is the cargo adaptor. Sec13 is a WD40 -propeller. Sec31 also has an amino-terminal -propeller, followed by an -solenoid, rather like clathrin. Unlike clathrin, however, the -propellers of these proteins form the vertices. Previous work6,7 has shown how the assembly unit of COPII, composed mainly of Sec31 plus Sec13, can generate a set of symmetrical lat-tices in which four edges converge on a two-fold symmetrical vertex (Fig. 1c,d). Variability in the curvature of the -solenoid elements can allow for variable diameters of the COPII coat, to accommodate cargoes of various sizes. The carboxy-terminal half of Sec31, which recruits Sec23/Sec24, projects flexibly inward, towards the membrane.

COPI is a complex of seven proteins: -, -, -, -, -, - and -COP. Of these, three -, - and -COP make up a stable, shell-form-ing heterotrimer, and the remaining four form a complex similar to heterotetrameric clathrin adaptors. Lee and Goldberg1 have determined crystal structures of an -subcomplex; it consists of most of the -chain (two tandem WD40 -propellers followed by an -solenoid) and a central fragment of the -chain (a fur-ther -solenoidal segment). This basic unit makes a triskelion-like trimer, with the dual -propellers of -COP at its centre and the overlapping, antiparallel -solenoids of the - and -chains radiating outwards (Fig. 1e). The amino-terminal part of -COP, which is missing from the structure, includes a further predicted -propeller; a continuation of the legs, therefore, will place such a domain at their tips, just as in clathrin.

The authors (and another team8) also report the structure of the carboxy-terminal part of -COP in complex with -COP: the two polypeptide chains form an intertwined, -solenoid handshake. This subcomplex is linked to the -subcomplex by a flexible -chain segment, and a plausible suggestion is that it projects inwards from the -lattice to recruit the -adaptor complex. COPI vesicles seen by electron microscopy of thin-sectioned cells seem to be relatively uniform in diameter9, so the range of lattice diameters to which the COPI -triskelions must adapt may be relatively small. Lee and Goldberg pro-pose a model for such a lattice, with an - triskelion at each vertex and with postulated two-fold contacts between such triskelions at the middle of each edge (Fig. 1f). Whether this model is correct can be verified by analysis of isolated or reconstituted COPI coats by elec-tron cryomicroscopy or tomography.

Lee and Goldbergs results show that clath-rin, COPI and COPII have certain architec-tural principles in common; these include the structural motifs that contribute to the shell

and the characteristics of the linkage between the coat and the membrane. The conservation of -solenoid and -propeller substructures between these three coat proteins is prob-ably also a sign of their common evolution-ary origin. The combination seems unlikely to be coincidental, even though the roles of the -propeller components vary among the different coat proteins. A reason for the evo-lutionary persistence of the -solenoid could be its suitability for creating lattices of variable diameter. A clearer view of potential missing links may be necessary to establish a common origin with confidence.

The same -solenoid and -propeller sub-structures (including Sec13 itself) make up a large part of another membrane-curving assembly (the scaffold of the nuclear pore10), which may also have been derived from the common vesicle-coat precursor. The nuclear pore has a fixed geometry, however, and the -solenoids of nuclear porin proteins curve back on themselves to form a more rigid structure than the completely extended legs of a clathrin triskelion.

The flexible tethers that link the COP coats and clathrin to adaptor components make minimal demands on the organization of the vesicle and thus allow for a wide selection of protein and lipid cargo. The relatively open lattices of all three types of coat allow access to the underlying membrane and accommodate variable projections from its cytoplasmic sur-face. The clathrin lattice is so open that it does not constrict the neck of the budding vesicle tightly enough to drive fission; it recruits a large protein called dynamin, which has GTPase-enzyme activity, to finish the job11.

There is one further, noteworthy, common principle of vesicle-coat biochemistry. A revers-ible assemblydisassembly cycle that allows reuse of the coat components requires the input

of free energy. COPI, COPII and clathrin coats meet this requirement by harnessing hydro-lysis of the nucleotides GTP and ATP. Assem-bly and release of COPI and COPII are linked to the GTP-hydrolysis cycle of a small GTPase enzyme Arf1 in the case of COPI12 and Sar1 in the case of COPII13. COPI is itself a crucial component of the Arf-GTPase activating com-plex14. Clathrin recruits the Hsc70 ATPase, which dissociates the coat after dynamin has pinched off the internal vesicle15. The uni-fied picture of carrier-vesicle properties that derives both from these parallels and from the recent structural studies of COPI components1 enhances the likelihood that analysis of one transport system will inform characterization of the others. Stephen C. Harrison is in the Department of Biological Chemistry and Molecular Pharmacology and the Howard Hughes Medical Institute, and Tomas Kirchhausen is at the Immune Disease Institute and the Department of Cell Biology, Harvard Medical School, Boston, Massachusetts 02115, USA. e-mail: harrison@crystal.harvard.edu

1. Lee, C. & Goldberg, J. Cell 142, 123132 (2010).2. Goldstein, J. L., Anderson, R. G. W. & Brown, M. S. Nature

279, 679685 (1979).3. Barlowe, C. et al. Cell 77, 895907 (1994).4. Waters, M. G., Serafini, T. & Rothman, J. E. Nature 349,

248251 (1991).5. Fotin, A. et al. Nature 432, 573579 (2004).6. Fath, S., Mancias, J. D., Bi, X. & Goldberg, J. Cell 129,

13251336 (2007).7. Stagg, S. M. et al. Cell 134, 474484 (2008).8. Hsia, K.-C. & Hoelz, A. Proc. Natl Acad. Sci. USA 107,

1127111276 (2010).9. Orci, L., Glick, B. S. & Rothman, J. E. Cell 46, 171184 (1986).10. Brohawn, S. G., Leksa, N. C., Spear, E. D., Rajashankar, K. R.

& Schwartz, T. U. Science 322, 13691373 (2008).11. van der Bliek, A. M. & Meyerowitz, E. M. Nature 351,

411414 (1991).12. Serafini, T. et al. Cell 67, 239253 (1981).13. Matsuoka, K. et al. Cell 93, 263275 (1998).14. Goldberg, J. Cell 100, 671679 (2000).15. Chappell, T. G. et al. Cell 45, 313 (1986).

ASTROPHYSICS

Making black holes from scratchMarta Volonteri

The means by which supermassive black holes form and grow have remained largely unclear. Numerical simulations show that the collision of massive galaxies can naturally lead to the creation of these objects.

Black holes are objects in which the pull of gravity is so strong that nothing not even light can escape. However, they are far from being invisible: they can be detected through the effect they have on their surroundings. Stellar-mass black holes have long been known to exist and to be the evolutionary end point of massive stars. But we now have evidence of another class of black hole, termed super massive black holes (SMBHs). These have masses of millions, or even billions, of solar masses, and

inhabit the centres of galaxies. A black hole of four million solar masses is thought to sit at the centre of the Milky Way. Stars in the Galactic Centre move too fast to be explained solely by the gravitational potential generated by the Galaxys luminous matter: a massive dark object must be lurking in the depths of the Galaxy, masterminding the movement of its puppet stars. Although the existence of SMBHs can be inferred in nearby and distant galaxies, the way in which they form is still largely mysterious.

1049

NATURE|Vol 466|26 August 2010 NEWS & VIEWS

20 Macmillan Publishers Limited. All rights reserved10

50 &

100

YEA

RS A

GO

50 YEARS AGOInterference between plant viruses was first demonstrated in plants infected with tobacco mosaic virus. Although several theories have been proposed to explain interference there is little evidence on the mechanisms involved. In the experiments described below interference was measured by the ability of an inoculum that does not produce local lesions (interfering strain) to affect the local lesions produced by another strain when the two are inoculated simultaneously to susceptible leaves The present results suggest that sap from virus-infected plants contain, in addition to infective particles, another factor that inhibits infection Centrifugation showed that infectivity and interfering activity occurred in the same zone in the gradient. If interference systems occur, then these appear to be in some way associated with virus particles. A. D. ThomsonFrom Nature 27 August 1960.

100 YEARS AGOOn August 20 occurred the bicentenary of the birth of Thomas Simpson, who may be regarded as one of the last of the English school of mathematicians of the eighteenth century. Newton, Halley, the Gregories, Muston, Demoivre, Brook Taylor, Maclaurin, had all passed away before Simpson reached middle age With the sole assistance of Edmund Stones translation of lHpitals Analyse des Infiniments Petits, Simpson wrote A New Treatise on Fluxions, which was considered a notable contribution to the literature of that comparatively new subject. In 1743 Simpson obtained a post as professor of mathematics at the Royal Military Academy, Woolwich ... After holding his post at Woolwich for eight years, he was seized with illness, caused, it was thought, by overwork He journeyed to Bosworth in February, 1761, and died there on May 14.From Nature 25 August 1910.

On page 1082 of this issue, Mayer et al.1 investigate, by means of high-resolution numerical simulations, the conditions that can drive the formation of an SMBH seed during the merger of massive galaxies in the young Universe. The authors study a process that can leave behind a black hole that is very massive at birth. Once this newly born SMBH seed starts to grow, by accreting matter from the dense cloud of gas in which it is embedded, we can hope to spot it as a quasar a compact and extremely bright object as small as a star but as luminous as an entire galaxy. SMBHs have been recognized as the engines that power quasars, and the energy emitted by material plunging onto an SMBH as the source of such power. About 10% of every parcel of matter is converted into energy during the infall.

However, SMBHs are not always active as quasars. Quasars were very common in the early Universe up until the Universe was about 40% of its current age of 14 billion years.But they slowly disappeared, leaving behind quiescent SMBHs. The masses of quiescent SMBHs in nearby galaxies correlate with the properties of their host galaxies2, suggesting that a single mechanism underpins the assem-bly of both SMBHs and galaxies. However, most of the galaxy-formation models involving quasar and SMBH evolution neglect research into the physical processes that formed SMBHs, and take simplistic approaches. This is a crucial missing ingredient: knowledge about the first SMBH seeds is necessary when investigating how SMBHs grow with their host galaxies over cosmic time.

Mayer et al.1 attack this problem by focusing on the environment in which an SMBH seed might form. They find that the collision and merging of two massive young galaxies pro-duces a massive gas disk, which is born in an unstable configuration, with a spiral pattern that transports mass towards the disks centre. Within only 100,000 years, more than 100 mil-lion solar masses of gas are accumulated in the disks central region, forming a dense cloud of gas. The core of this cloud eventually col-lapses into a supermassive star (with masses of tens of thousands of solar masses or more), the core of which ultimately collapses into an SMBH seed. The timescale for the formation of the cloud is much shorter than the 108 years that are needed to convert the disks gas into stars. Gas is therefore available for SMBH for-mation rather than being consumed to form stars. Gas collapse driven by galaxy mergers thus overcomes the major difficulty of pre-vious collapse models in isolated galaxies, in which star formation had to be artificially suppressed to prevent it from consuming the gas reservoir.

The highly dynamic, out-of-equilibrium conditions of galaxy mergers, and the details of gas infall, can be studied only with numeri-cal tools such as those used by Mayer and colleagues1. The authors results tie in nicely with Begelman and colleagues proposal3,4

that a supermassive star forms when gas-infall rates exceed the large threshold value of about 1 solar mass per year. According to this model3,4, although the stars core collapses into a black hole, material piles up on the stars surface and allows the hole to be efficiently fed; this black-hole-powered star is known as a quasistar3.

Mayer et al. find that such a large gas-infall rate can indeed be produced through galaxy mergers. Whats more, the combination of their gas-infall process1 with the quasistar model proposed by Begelman et al. can explain the existence of powerful quasars at a time when the Universe was one-tenth of its current age5. These luminous quasars can be powered only by SMBHs of billions of solar masses. (There are galaxies that weigh, in their entirety, less than these SMBHs.) Explaining the existence of such SMBHs less than one billion years after the Big Bang is challenging, and requires either a very steady growth of the SMBHs or that the seeds are quite massive. Mayer and colleagues results are appealing in this context because the SMBH seeds that form during galaxy mergers are indeed very massive at birth.

However, the black-hole formation mech-anism proposed by Mayer et al. seems to be efficient only for galaxies that were already very massive at early cosmic times; these galaxies are expected to evolve into objects the size of the Milky Way or larger. Although this finding is broadly consistent with obser-vations, which indicate that most SMBHs are hosted by large galaxies, SMBHs have also been found in dwarf galaxies6 (more than a 100 times smaller than the Milky Way). Of course, it is possible that smaller SMBHs in low-mass galaxies may have formed through other mechanisms7, for instance as remnants of the first generation of stars. These stars are predicted to have had masses up to several hundred times that of the Sun, and so probably left behind almost-massive black holes at the end of their life.

The James Webb Space Telescope (the successor to the Hubble Space Telescope), future X-ray missions such as IXO and the gravitational-wave telescope LISA will all have the technical capabilities to detect qua-sars and SMBHs in the early Universe. These therefore promise to provide further insight into the process that generates SMBHs from scratch. Marta Volonteri is in the Department of Astronomy, University of Michigan, Ann Arbor, Michigan 48109, USA.e-mail: martav@umich.edu

1. Mayer, L., Kazantzidis, S., Escala, A. & Callegari, S. Nature 466, 10821084 (2010).

2. Gltekin, K. et al. Astrophys. J. 698, 198221 (2009).3. Begelman, M. C., Volonteri, M. & Rees, M. J. Mon. Not. R.

Astron. Soc. 370, 289298 (2006).4. Begelman, M. C. Mon. Not. R. Astron. Soc. 402, 673681

(2010). 5. Fan, X. et al. Astron. J. 122, 28332849 (2001). 6. Gallo, E. et al. Astrophys. J. 714, 2536 (2010). 7. Rees, M. J. IAU Symp. Proc. 77, 237242 (1978).

1050

NATURE|Vol 466|26 August 2010NEWS & VIEWS

20 Macmillan Publishers Limited. All rights reserved10