charge disorder, where the conductivity is tem-perature-independent at low temperatures1.

Why are the effects of charged impurities dif-ferent from those caused by bonding to hydro-gen? The unique properties of graphene prevent the slowly varying potential of a charged impu-rity from completely back scattering (reversing the direction of) electrons. An exciting possi-bility is that sharp disorder caused by hydrogen

can easily backscatter electrons. The result-ing interference of forward- and backward-moving electrons would create localized states at zero temperature, just like those found in other two-dimensional systems10. Another possi-bility is that hydrogen locally modifies the band structure of graphene, so creating a band insulator.

More work will be needed to understand

both the electronic structure of hydrogen-ated graphene and the effects of sharp and smooth disorder on the scattering of its elec-trons. Never theless, these results3 suggest that graphene could be almost the ideal material in which to study transitions of metals to insula-tors. Unlike other two-dimensional systems, in which electrons are buried at the interface between two materials, graphene’s surface is exposed. This allows not only direct imaging of defects by scanning tunnelling microscopy, but also the use of techniques that probe elec-tronic structure, thus creating new windows through which metal–insulator transformations can be viewed. ■

Michael S. Fuhrer is at the Center for

Nanophysics and Advanced Materials,

Department of Physics, and Shaffique Adam

is at the Condensed Matter Theory Center,

Department of Physics, University of Maryland,

College Park, Maryland 20742, USA.

e-mails: mfuhrer@umd.edu;

adam1@umd.edu

1. Novoselov, K. S. et al. Nature 438, 197–200 (2005).

2. Zhang, Y., Tan, Y.-W., Stormer, H. L. & Kim, P. Nature 438, 201–204 (2005).

3. Elias, D. C. et al. Science 323, 610–613 (2009).

4. Anderson, P. W. Phys. Rev. 109, 1492–1505 (1957).

5. Abrahams, E. et al. Phys. Rev. Lett. 42, 673–676 (1979).

6. Bardarson, J. H. et al. Phys. Rev. Lett. 99, 106801 (2007).

7. Chen, J.-H. et al. Nature Phys. 4, 377–381 (2008).

8. Adam, S., Hwang, E. H., Galitski, V. M. & Das Sarma, S.

Proc. Natl Acad. Sci. USA 104, 18392–18397 (2007).

9. Sofo, J. O., Chaudhari, A. S. & Barber, G. D. Phys. Rev. B 75, 153401 (2007).

10. Aleiner, I. L. & Efetov, K. B. Phys. Rev. Lett. 97, 236801 (2006).

Figure 1 | Metal–insulator transitions. a, The electronic energy states of metals form a series of bands. Electrons fill the bands up to a maximum energy defined by the Fermi level. E is energy, p is momentum. b, The Fermi level of a band insulator lies in a gap in its band structure. Electrons completely fill the lower valence band, but do not cross into the upper conduction band, and so are not free to form a current. c, Imperfections in solids scatter electrons, creating standing waves that also prevent the motion of electrons, so resulting in a localized insulator. d, The electronic energy-states of graphene form two cone-shaped bands that meet at their tips. In charge-neutral graphene, the Fermi level lies at the tip of the cones. There is no bandgap, so the material is classed as a metal. Elias et al.3 show that when graphene reacts completely with hydrogen, it forms a new material (graphane) that is expected9 to be a band insulator. Reaction of graphene with smaller amounts of hydrogen (also demonstrated by Elias et al.) might produce sharp imperfections that bring about a localized insulating state.

E E

px px pxpy py py

Fermi level

a Metal b Band insulator c Localized insulator d Graphene

E

Fermi level

DRUG DISCOVERY

Not as fab as we thoughtSoumaya Zlitni and Eric D. Brown

Ever since penicillin was isolated from mould, it has been assumed that naturally occurring antibiotics are good starting points for drug-discovery programmes. The latest study shows that this isn’t always true.

Drug-resistant bacterial infections continue to occupy the headlines, amid increasingly desperate calls for new antibiotics to treat infectious diseases. Some of the most alarming reports concern ‘Gram-positive’ pathogens1, which are a pervasive nuisance in both the clinic and the world at large. The recent dis-covery2 of the potent antibiotic platensimycin was therefore greeted with great enthusiasm. Platensimycin was isolated from soil-dwelling Streptomyces platensis microbes, and represents a new class of antibiotic that acts against Gram-positive pathogens. But on page 83 of this issue, Brinster et al.3 provide a sobering lesson in what drug discoverers call target validation. The authors show that, although compounds that have the same mechanism of action as platensimycin are effective anti bacterials in soil, they are inactive in models that simulate

environments relevant to infection.The problem with existing antibiotics is that

they attack a narrow spectrum of bacterial physiology: most interfere with the synthesis of bacterial DNA, proteins or cell walls. There is therefore great interest in exploring new biological targets for antibacterial therapies. One potential target is fatty-acid biosynthesis, and the past few years have witnessed intense efforts to identify inhibitors of this process.

Fatty acids are organic molecules that contain long, unbranched hydrocarbon chains of up to 18 carbon atoms. Their biosynthetic machinery is encoded by several genes involving the fab loci of the bacterial genome. Fab proteins come together to construct fatty acids, two carbon atoms at a time, in a cyclic process. The promise of this biosynthetic pathway as an antibacter-ial target stems from the fact that it is essential

for the formation of cellular membranes in a wide range of bacterial pathogens. The pro-cess is distinct from fatty-acid biosynthesis in humans, suggesting that antibiotics that block this mechanism could be made that are selectively toxic for bacteria.

Several compounds have already been iden-tified that inhibit specific steps in the bacterial biosynthesis of fatty acids. These include syn-thetic compounds (such as the antituberculosis compound isoniazid and the general-purpose antibiotic triclosan) and naturally occurring compounds (such as cerulenin and thiolac-tomycin, both broad-spectrum antibiotics) 4. The synthetic compounds, however, have had only niche applications — isoniazid in com-bination therapies and triclosan in soaps and plastics — whereas the natural products have never proved useful in the clinic. Neverthe-less, the discovery of platensimycin as a new addition to the roster of fatty-acid biosynth-esis inhibitors generated renewed excitement about this antibiotic class.

Yet studies dating as far back as the late 1970s have shown that Gram-positive bacteria can acquire fatty acids from their surroundings and incorporate them into their cell mem-branes5. Given that human serum is a rich source of such acids, these findings seriously undermine the idea that inhibitors of bacterial

39

NATURE|Vol 458|5 March 2009 NEWS & VIEWS

5.3 n&v MH IF 395.3 n&v MH IF 39 27/2/09 17:30:3527/2/09 17:30:35

© 2009 Macmillan Publishers Limited. All rights reserved

and colleagues’ work is likely to be regarded as a landmark that

will raise the bar for target validation in drug discovery

in the future. A likely out-come is that novel biologi-cal targets for antibiotics will be scrutinized with many more experiments

than before to understand the importance of those

targets for infection (rather than just their role in bacterial

survival in vitro).This work also chal-

lenges common assump-tions associated with the discovery of natu-rally occurring antibiot-ics from organisms such as streptomycetes. It is tempting to conclude that the very existence

of such compounds validates the target of the antibiotic as being suitable for pharmaceutical intervention. This idea stems from the notion that antibiotics are the result of an evolution-ary arms race among soil-dwelling bacteria.

But one must be mindful of the environmental conditions that define the limits of this strug-gle. Antibiotics from streptomycetes have evolved to provide those organisms with an advantage in a soil environment where fatty acids, for example, are in short supply.

Antibiotic drug discovery and develop-ment is an onerous process fraught with long development times, high costs and the risk that regulatory agencies will ultimately reject the discoverer’s application to sell the fruits of their labours6. Brinster and co-workers’ work3 will resonate with those working in drug discovery, and will inspire renewed efforts to understand the basis of bacterial pathogenesis. ■

Soumaya Zlitni and Eric D. Brown are in the

Department of Biochemistry and Biomedical

Sciences, and the Michael G. DeGroote Institute

of Infectious Disease Research, McMaster

University, Hamilton, Ontario L8N 3Z5, Canada.

e-mail: ebrown@mcmaster.ca

1. Weigel, L. M. et al. Science 302, 1569–1571 (2003).

2. Wang, J. et al. Nature 441, 358–361 (2006).

3. Brinster, S. et al. Nature 458, 83–86 (2009).

4. Heath, R. J., White, S. W. & Rock, C. O. Prog. Lipid Res. 40, 467–497 (2001).

5. Altenbern, R. A. Antimicrob. Agents Chemother. 11, 574–576

(1977).

6. Fernandes, P. Nature Biotechnol. 24, 1497–1503 (2006).

fatty-acid biosynthesis could fight infections: the effects of such drugs would be overcome if the pathogens simply take up fatty acids from serum. Brinster et al.3 conducted experiments to directly address this contradic-tion, and present com-pelling evidence that calls into question the value of antibiotic therapies that tar-get fatty-acid biosynthesis.

The authors began their investigation by treating several clinical isolates of Gram-positive pathogens — including strains resistant to com-monly used antibiotics — with triclosan and cerulenin, in growth media that either con-tained or lacked fatty acids. They found that both inhibitors are quite effective at inhibiting the growth of drug-resistant strains in standard laboratory media (which lack fatty acids), but that this effect was abolished when certain fatty acids were added individually, or when human serum was added. They also observed that the strict requirement for fatty-acid biosynthesis to sustain bacterial growth is superseded when fatty acids are readily available in the growth environment, as demonstrated by the low level of expression of fab genes in bacteria that were cultured under these conditions. Moreover, they found that mutant strains of bacteria that were defective in up to five of the fab genes flourish in fatty-acid-rich serum.

In vitro experiments are all well and good, but the gold-standard studies for drug dis-covery are those performed in vivo. Most sig-nificantly, in animal experiments testing the implications of fatty-acid biosynthesis in host infection, the authors were especially careful to run their tests under conditions that closely reflect real-world infection, a criterion some-times overlooked in such studies. Together, their results strongly support the conclusion that fatty-acid biosynthesis is a poor target for drugs against Gram-positive pathogens.

This work3 offers a profound lesson in target validation. A cornerstone of modern anti bacterial drug discovery is that cellular processes are investigated in vitro (Fig. 1), because this is quicker, cheaper and simpler than investigating these processes in animals. Such in vitro models are usually fine, because the bacterial processes that are important for life in a Petri dish are also almost always essen-tial for the survival of bacteria that have infected a host — although many more bacterial func-tions that can’t be modelled in a dish will also be crucial for infections. But the current study begs the question: what if functions important for growth in the laboratory are superfluous for growth in a host? In raising this point, Brinster

Figure 1 | Sure of the cure? Antibiotics are usually tested for their efficacy in vitro. Here, the red regions around antibiotic-

impregnated wafers indicate areas of poor bacterial growth. Brinster et al.3 show that such assays can be misleading if they don’t accurately simulate environments relevant

to infection in vivo.

ASTROPHYSICS

Capturing black-hole pairsJon M. Miller

The observed growth of galaxies suggests that the black holes thought to lurk at their centres may find each other and merge. A large survey of galaxies has finally netted two black holes in a tight pairing.

The idea of black holes merging may seem like the height of scientific extremity, but such mergers may be commonplace if the prevailing theory of structure formation in the Universe is correct. Scant context is provided by the solar neighbourhood: stars within a galaxy do not merge because they are very small relative to their separations. However, the distances that separate galaxies are not as large in com-parison to their size, and observations reaching back to the early Universe strongly suggest that galaxies grow through a hierarchy of mergers1. In a galactic merger, gravitation and dynamical friction drag black holes to the bottom of the ‘gravitational well’, where they encounter each other and may eventually merge. Nowhere is this sinking of black holes more apparent than in clusters of galaxies, best exemplified by the nearby Virgo cluster and the monstrous billion-solar-mass black hole at the heart of its central galaxy.

By carefully sifting through 17,500 spectra of quasars — extremely bright galactic nuclei thought to be powered by supermassive black holes — selected from the unprecedentedly

large Sloan Digital Sky Survey2 (SDSS), Boroson and Lauer3 (page 53) have identified the first compelling example of a tightly bound black-hole binary system in a quasar known as SDSS J153636.22+044127.0. The pair of black holes is separated by a mere 0.1 parsecs, or about 1,300 times the radius of the event horizon (a boundary in space-time beyond which nothing can escape) of the more massive member. This distance is 13 times shorter than that between the Sun and its near-est neighbour, the Alpha Centauri star system. The finding3 is an important validation of the theory of hierarchical structure growth in the Universe.

Previous work has identified potential exam-ples of black holes on their way to merging, but the case presented by Boroson and Lauer is special because the pairing is tighter and the evidence much stronger. An X-ray image taken with the Chandra X-ray Observatory of the galaxy NGC 6240 reveals two bright nuclei4 powered by the accretion of gas onto black holes, but their separation is very large. The precessing radio jet seen in another galactic

40

NATURE|Vol 458|5 March 2009NEWS & VIEWS

5.3 n&v MH IF 405.3 n&v MH IF 40 27/2/09 17:30:3527/2/09 17:30:35

© 2009 Macmillan Publishers Limited. All rights reserved