But quorum sensing in V. cholerae may be unique. Unlike in many other species, in which quorum sensing induces virulence, in V. cholerae the response shuts virulence down, allowing the bacteria to escape the host and re-enter the environment. This means that the signals themselves could be used as therapeutics. The high specificity of 3-hydroxy-tridecan-4-one for V. cholerae certainly makes it an excellent candidate for drug development. In fact, in a proof-of-concept experiment, Higgins and colleagues4 show that their synthetic version of the signal does indeed terminate production of known virulence

Like people, many bacteria do things in groups that they don’t do on their own. These com-munal activities can be spectacular; the marine bacterium Vibrio fischeri, for example, produces bioluminescence in the light organs of deep-sea fish. But bacterial group behaviour can also be deadly — many bacteria become virulent only when they reach a certain local concentration. Such coordinated actions require bacteria to ‘talk’ to each other by sending chemical signals, a process known as quorum sensing. Although some of the molecules involved are known, it is likely that there are many more.

It’s been known for a few years that Vibrio cholerae, the bacterium that causes cholera in humans, is ‘bilingual’ — that is, it uses two distinct signalling molecules1 to suppress its virulence2. One of these signals is a molecule that many species of bacteria use for quorum sensing3, but the identity of the second signal has remained a mystery. In this issue, Higgins et al.4 report the structure of this second signal. Their discovery represents a new structural class of quorum-sensing signal that may be exclusive to Vibrio bacteria, making it a possible lead for drug discovery.

In general, quorum sensing is straight-forward: bacteria release signals into the surrounding environment; if the signals reach a critical concentration, they are detected by bacteria in the vicinity and this stimulates a response. In V. cholerae, quorum sensing pro-ceeds through two parallel systems1, either of which is sufficient to independently initiate a response. The first of these involves the AI-2 molecule (Fig. 1), a signal used by many spe-cies of bacteria. AI-2 is detected by the sensory proteins LuxP and LuxQ, which are associated with the bacterium’s cell membrane.

Although the identity of the signal in the second system was unknown, the enzyme responsible for producing the signal had been identified as the CqsA protein. The second signal is thought to be detected by a putative membrane-associated sensor called CqsS. Both systems in V. cholerae funnel information into the same signalling cascade (through the transducing proteins LuxU and LuxO) so that an analogous functional response is produced in each case: LuxO is inactivated, resulting in increased activity of HapR, a negative regulator that represses expression of virulence genes.

In their detective work identifying the unknown quorum-sensing signal of V. cholerae, Higgins et al.4 took their cue from the known biochemistry. They introduced the cqsA gene

into Escherichia coli, creating a recombinant strain that produces much more signal than the parent V. cholerae species. By extracting the culture fluids of the E. coli, the authors obtained a mixture of compounds that they separated into its constituent parts. They next tested the purified compounds on a strain of V. cholerae that had been engineered to emit light in response to the unknown signal. Cer-tain fractions of the mixture were 10,000 times more active than controls.

Using a combination of spectroscopic techniques, Higgins et al. then identified the active compound as 3-hydroxytridecan-4-one (Fig. 1). They confirmed this by preparing the compound chemically, and testing the synthetic version in their activity assay. The authors were finally able to obtain sufficient material from cultures of natural V. cholerae for analysis, and so to prove conclusively that 3-hydroxytridecan-4-one is the second signal for this bacterium.

So why does V. cholerae adopt a ‘belt and braces’ approach to quorum sensing, using two parallel systems when one would be sufficient? There are many potential explanations, one of which relates to the distinct chemical proper-ties of the two signals. These properties might influence the stability or rates of diffusion of the signals, perhaps making one molecule supe-rior to the other for quorum sensing in a parti-cular environment. Higgins et al.4 suggest that, because V. cholerae encounters environments that are rich in other AI-2-producing bacteria (such as the large intestine), the AI-2 system might be used for interspecies signalling. Con-versely, they propose that the specificity of the Cqs signal for the Vibrio genus makes it ideal for quorum sensing with other bacteria of the same species. To test this hypothesis, it will be necessary to show that one system dominates, depending on the environmental context.

The authors also suggest that the Cqs signal could be exploited therapeutically to dampen V. cholerae virulence. Quorum-sensing systems that trigger virulence have already been targets for therapies against several species, such as Pseudomonas aeruginosa — an opportunistic pathogen that causes infections in people with compromised immunity. Usually the goal is to block quorum sensing with a small-molecule inhibitor5. This is a difficult task, because such inhibitors must be specific, stable, easily deliverable to the infection site and able to out-compete the natural quorum-sensing signal for the target receptor.

O

OH

OB

O

O

HO CH3

HO

HO OH

LuxQ

LuxU

Deactivated LuxO

Upregulated HapR

Suppression ofvirulence

CqsS

Deactivation Deactivation

AI-2

3-Hydroxytridecan-4-one

LuxP

–

Figure 1 | Inhibiting virulence in Vibrio cholerae. In V. cholerae bacteria, the HapR regulator represses the expression of virulence genes. HapR expression is usually inhibited by the transducer protein LuxO, so that the bacteria are virulent. But V. cholerae emit two types of signal molecule that inhibit virulent behaviour in nearby V. cholerae bacteria. One of these, AI-2, is recognized by the LuxP receptor on the bacterial cell membrane. LuxP activates the LuxQ protein inside the cell, which deactivates the transducer protein LuxU. This prevents activation of LuxO, so that HapR activity is increased and virulence is suppressed. Higgins et al.4 show that the second signal molecule used by V. cholerae is 3-hydroxytridecan-4-one. This signal interacts with a putative receptor on the cell membrane (CqsS) that then deactivates LuxU, triggering the same signalling cascade described for AI-2 and LuxPQ.

MICROBIOLOGY

Bilingual bacteriaMatthew R. Parsek

Many bacteria use chemical signals to coordinate group behaviour. A signal that suppresses virulence has been identified in the bacterium that causes cholera, and could be a new therapeutic target.

805

NATURE|Vol 450|6 December 2007 NEWS & VIEWS

factors in V. cholerae. But this idea raises possi-ble public-health issues: the activation of quo-rum sensing in V. cholerae also induces active movement of the bacterium, potentially mobi-lizing the pathogen and encouraging the spread of infection from one person to another.

For several years, the repertoire of bacte-rial quorum-sensing signal molecules and receptors was thought to be rather limited and restricted to a few species. But recent studies have revealed an array of different signals, suggesting that we have only just scratched the surface of possible mechanisms. As new signals are identified and their use by bacteria is assessed, the list of quorum-sensing organisms will undoubtedly grow. We may

eventually reach a point at which bacteria that do not engage in quorum sensing are regarded as the exception, rather than the norm. The challenge now is not only to identify new sys-tems, but also to make sense of why an organism would use one type of system over another. ■

Matthew R. Parsek is in the Department of Microbiology, University of Washington, Box 357242, Seattle, Washington 98195-7242, USA.e-mail: parsem@u.washington.edu

1. Miller, M. B., Skorupski, K., Lenz, D. H., Taylor, R. K. & Bassler, B. L. Cell 110, 303–314 (2002).

2. Zhu, J. et al. Proc. Natl Acad. Sci. USA 99, 3129–3134 (2002).3. Camilli, A. & Bassler, B. L. Science 311, 1113–1116 (2006).4. Higgins, D. A. et al. Nature 450, 883–886 (2007).5. Sperandio, V. Expert Rev. Anti-infect. Ther. 5, 271–276

(2007).

PALAEOCLIMATE

Slush findAlan J. Kaufman

A coupled model of palaeoclimate and carbon cycling turns up the heat on the idea that Earth once became a giant snowball. It supports instead a milder ‘slushball Earth’ history — but piquant questions remain.

Sediments laid down in the oceans during the late Neoproterozoic era, between about 850 million and 542 million years ago, tell a dramatic story. They contain wildly varying abundances of the carbon isotope 12C, which is typically incorporated into organic matter dur-ing photosynthesis. The pattern of excess 12C in carbonates immediately above and below gla-cial deposits seems to indicate that photosyn-thesis on Earth came to a halt during a series of ice ages. These observations are a foundation of the ‘snowball Earth’ hypothesis1,2: that, just before the first appearance of animals, Earth’s surface might have been repeatedly frozen over, even at tropical latitudes.

Not necessarily so, say Peltier et al. on page 813 of this issue3. They apply basic ideas about the solubility of gases to a coupled model of climate and carbon cycling4 during the frigid late Neoproterozoic era. The results that emerge might explain the oscillatory carbon-isotope compositions of carbonates across the Neoproterozoic glacial cycles, without resort-ing to the hard-snowball model. Instead, they could lend support to a milder variation on the same theme — ‘slushball Earth’.

The slushball and snowball models both predict ice sheets on continents near the Equa-tor, but with markedly different extents of ice covering the oceans. In the snowball version, the frozen planet is completely blanketed, and reflects most of the Sun’s warming rays back into space. Temperatures plummet and surface processes, including life, largely cease. Escape from the snowball state probably requires the build-up of volcanic carbon dioxide in the atmosphere over many millions of years,

resulting in torrential acid rain and the intense weathering of exposed rocks during the global thaw.

The slushball model5, by contrast, predicts open glacial oceans that would have con-strained runaway refrigeration by allowing sunlight to warm the planet’s surface, driving an active hydrological cycle6 and photosyn-thesis7 in exposed seas. The end of such an ice age need not have required extreme amounts of CO2 in the atmosphere, nor have been delayed for millions of years.

Peltier and colleagues’ new dynamic model3 shows how climate and atmospheric oxygen might have combined to prevent a runaway snowball Earth. As the oceans cool during ice ages, lower temperatures allow atmospheric gases such as oxygen to diffuse more read-ily into the deep sea, forcing the oxidation of abundant dissolved organic carbon, formed initially by photosynthesis in surface waters, to CO2. Released back to the atmosphere by this oceanic ‘respiratory’ process, the excess CO2 would warm the planet and thereby end the glacial epoch.

What is particularly interesting about this model is that climate drives the carbon cycle (and so determines the stable levels of atmos-pheric CO2). In the most recent ice ages, as well as for earlier interpretations of Neoproterozoic carbon-isotope anomalies8, the assumption has instead been the other way around. The crucial difference is that the Neoproterozoic carbon cycle was conceivably buffered by a marine pool of dissolved organic carbon that was orders of magnitude larger than that in the present-day oceans4.

A pertinent criticism of Peltier and col-leagues’ mathematical model is the uncer-tainty in its input parameters, in particular the assumption that levels of atmospheric oxygen were similar to those of today (around 21%). Biological9 and geochemical10–12 evidence indi-cates that oxygen levels were low throughout most of the Neoproterozoic, with a significant rise in breathable air around 550 million years ago — about the time animals first appeared on the planet. In that case, it seems likely that pervasive oxygenation of the atmosphere and the hydrosphere, including the vast pool of dissolved organic carbon, occurred millions of years after the extensive ice sheets of the Neoproterozoic had melted away. This rise, known as the Wonoka anomaly after the local-ity in South Australia in whose rocks it was first observed, is recorded in 550-million-year-old carbonates worldwide that are spectacularly rich in 12C.

The coupled model also does not address certain hallmark geological features of the Neoproterozoic glacial episodes. These include the unexpected appearance of iron-bearing sediments in the glacial deposits, as well as the enigmatic ‘cap carbonates’ that lie immediately above them (Fig. 1). The co-occurrence of iron-oxide cements and glacial sediments implies that levels of soluble iron increased during

Figure 1 | A soluble solution? The large (5–8-cm high) carbonate crystal fans (black to dark grey), which seem to grow out of the sea floor in this polished slab of a Neoproterozoic ‘cap carbonate’ from Brazil, suggest the presence of high concentrations of dissolved inorganic carbon in sea water after the ice ages, together with the rapid accumulation of sediments. These fans are draped by grey to white, fine-grained carbonates, which near the top become red, probably because they contain the iron-oxide mineral haematite (Fe2O3). The isotopic composition of such geological deposits is a focus of Peltier and colleagues’ model interpretation3 of Neoproterozoic climate and carbon cycling.

A. J

. KA

UFM

AN

807

NATURE|Vol 450|6 December 2007 NEWS & VIEWS