Denying bacteria a quorum
May 12, 2011 § 2 Comments
Many species of bacteria make decisions about which genes to express based, in part, on how dense their populations are. The most famous example of this phenomenon, called “quorum sensing”, is seen in the bioluminescent bacterium Vibrio fischeri, which turns on its light-producing genes only when its population reaches a certain level. This bacterium has evolved a rather unlikely-sounding symbiotic relationship with a squid, in which the light produced by bacteria living in the squid allows the squid to avoid casting a shadow on moonlit nights, which makes it harder for predators to see. More sinisterly, V. fischeri’s relative Vibrio cholerae uses quorum sensing to decide when to produce virulence factors and develop hard-to-treat biofims. Many other pathogenic bacteria do the same, leading to considerable interest in the question of how to block quorum sensing. A recent paper (Chen et al. 2011. A strategy for antagonizing quorum sensing. Molecular Cell PMID: 21504831) discusses one way to prevent bacteria from finding out whether they are present in sufficient numbers to have a desired effect.
There are two main known strategies for quorum sensing: one uses transmembrane receptors that sense secreted molecules in the environment, the other uses cytoplasmic, soluble receptors that act as transcription factors when the quorum signal is present. If the quorum signal isn’t present, the receptors fail to fold properly and get degraded. The transcription factor/sensor in V. fischeri is named LuxR, and the family of proteins that does the same job in other bacteria are therefore dubbed the LuxR-type transcriptional regulators. This paper focuses on a LuxR-type protein in Chromobacterium violaceum, a (rare) human pathogen that uses quorum sensing to control several apparently disconnected processes: biofilm formation, cyanide production, and synthesis of a purple pigment that may also be an antibiotic.
The natural ligand for this LuxR-type protein, called CviR, is an acyl homoserine lactone with a 6-carbon tail (C6-HSL). Any modification that extends the 6-carbon tail — say, to 8 or 10 carbons — reduces the activity of the ligand, and the authors previously showed that adding a bulky chlorophenyl group at the end of the tail produces a signaling antagonist (CL). In this paper, they set out to understand why this happens.
Leaping lightly over a lot of structural biology and biochemistry, I’ll just tell you the authors’ conclusion. The CviR protein is a dimer, and in its active form the ligand-binding domains (oval, in my sketch) and DNA-binding domains (cylinders) from each monomer are stacked neatly on top of each other. This puts the business ends of the two DNA-binding domains 30 Å apart, just right for binding to two successive turns of the DNA double helix. In the inactive form, the DNA-binding domains stack under the ligand-binding domain from the opposite monomer. In this conformation, the DNA-binding domains are 60Å apart and can’t effectively bind to their target DNA. The active form is stabilized by binding to the natural ligand (C6-HSL), while the inactive form is stabilized by the bulkier antagonist (CL). C10-HSL is inbetween the two, acting in a confused way as both an agonist and an antagonist.
By crystallizing the ligand-binding domain of CviR with various versions of the ligand, Chen et al. were able to see that the conformation of a specific residue in the ligand-binding site, Met89, is different in the inhibited form. In the active form, Met89 protrudes into the ligand-binding pocket; because the natural ligand is small, there is no problem with this positioning. But in the inhibited form, when larger ligands such as C10-HSL or CL are bound, Met89 flips away, changing the shape of the surface that’s buried in the interface between the DNA-binding domain and the ligand-binding domain and stabilizing the crossed-domain form of the protein. When the authors mutated Met89 to an amino acid with a smaller side chain (Ser or Ala), ligands that previously had an antagonistic effect turn into agonists.
Destroying an activity is one way to test whether your theory is consistent with reality. Another — perhaps stronger — way is to start with something that doesn’t have the activity in the first place, and see if you can create the activity. Fortuitously, there is another strain of C. violaceum starts off with a serine residue in the position of Met89 in its CviR protein (let’s call it CviR’ to distinguish it from the CviR above, to which it is only about 85% identical). This bacterium can use both C10-HSL and CL as agonists, in addition to C6-HSL. Chen et al. wanted to know what it would take to alter this promiscuous behavior. They cheated just a little: they assumed that Met89 would be required, so they mutated this residue in the CviR’ protein first. Using the Met-containing CviR’ as the parent for mutagenesis and selection, they found that they only needed one additional amino acid change — Asn77 to Tyr — to get a protein that is selective for agonists with small side chains: it’s antagonized by CL, but C6-HSL still acts as an agonist. Like residue 89, residue 77 is at the interface between the ligand-binding domain and the DNA-binding domain in the crossed conformation.
So the picture that emerges is that the LuxR-type proteins are in equilibrium between the closed, crossed conformation and the open, DNA-binding conformation. Ligands that bind better to the closed conformation (e.g. CL for CviR and for CviR’ with the Met89 and Tyr77 mutations) shift the equilibrium, stabilizing the closed form.
Of course, conformational changes in signaling proteins are not exactly news. Think of the switchblade-like behavior of Src kinase, for example (in which the phosphorylated tail of the protein folds back on itself and binds to the SH2 domain, blocking the activity of the kinase domain — a closed conformation that can be released by dephosphorylating the tail, or competition from another SH2 ligand). But this is a fairly dramatic difference in domain arrangement, with one arrangement stabilized by one small molecule (C6-HSL) and the other stabilized by a not-very-different small molecule (CL). The authors point out that this opens up two possibilities for drug development. One can, of course, look for molecules like CL that work in the strain you’re interested in. More unusually, one could look for molecules that bind to the protein-agonist complex (which is in the open state) and push it into the closed state. Effectively this would be equivalent to finding a small molecule that substitutes for the chlorophenyl ring in CL.
The slight worry I have about these strategies is that it’s obviously not impossible for the bacteria to change the interface between their ligand-binding domain and the DNA-binding domain without losing activity — as in CviR’ — so strategies like this might be easy for the bacteria to evolve around. Still, with biology you never know until you look.
Chen G, Swem LR, Swem DL, Stauff DL, O’Loughlin CT, Jeffrey PD, Bassler BL, & Hughson FM (2011). A strategy for antagonizing quorum sensing. Molecular cell, 42 (2), 199-209 PMID: 21504831
A general problem with potential antibiotics tackling cool things like quorum sensing or chemotaxis or cell mobility is that their targets are not conserved enough: different bugs have different proteins.
Therefore these antimicrobials end up having a very narrow spectrum, and since usually it is not immediately obvious what sort of bug is causing the desease, usability of these otherwise great compounds is – as yet! – limited. Fast and reliable diagnostics will turn the tables one day and “narrow spectrum” will become “specificity”.
Here is a great review about this: http://informahealthcare.com/doi/abs/10.1517/13543776.2010.511176
yes indeed — even with the two strains of C. violaceum the authors study, the compound that is an antagonist for one is an agonist for the other. But I think it’s still a very interesting strategy to explore.