The dimer interface of PhoQ

Relating a protein structure to its function is often not as straightforward as we would like. As any protein crystallographer knows, many biomolecules will only form crystals in specific, limited conditions, with particular buffers, protectants, and co-solutes. Sometimes even the water must be replaced with, say, isopropanol in order to form a crystal. History has shown that despite this, most protein crystals imply models that match with biochemical data. Sometimes, however, crystal structures of a single protein or two highly homologous proteins disagree substantially. In that case, it is up to the biochemists to determine which structure (if any) correctly represents the physiologically relevant ensemble.

That is just what happened in a study soon to be released in the Journal of Molecular Biology from the lab of Bill DeGrado. Goldberg, Soto et al. wish to resolve a question about the dimerization of the periplasmic domain of PhoQ, part of a two-component signaling system in bacteria. These systems sense an external stimulus (in this case divalent cations) and activate a set of genes. PhoQ deals in the first half of the equation, and exists as a dimer (pair of bound proteins) that stretches from the periplasmic space (where it senses the signal) across the inner membrane and into the cytoplasm of the bacterium. Because it is in general very difficult to crystallize membrane proteins in their entirety, the isolated pieces (domains) of this protein were crystallized independently. Unfortunately, the sensor domain does not dimerize strongly in solution, and the crystal structure of the E. coli PhoQ (explore this structure at the PDB):
does not match up particularly well with the structure determined for the homologous protein in S. enterica (explore this structure at the PDB):

Goldberg et al. note that the differences in these structures are unlikely to reflect actual differences between the proteins, because the amino acid sequences are 85% identical. That doesn't guarantee that the structures will be the same (see this post), but as a rule nearly-identical structures are what we would expect. Goldberg et al. set out to determine which structure represents the physiological form, and they decided to do it in vivo, with the full-length protein.

In order to do this, they engineered several mutants of PhoQ, each of which had a cysteine at a particular position in the N-terminal helix of the sensor domain (visible side chains in above figures). Cysteines have the property that they can form disulfide bonds with other cysteines when the chemical environment is right. Consequently, if the cysteine of one dimer is in close proximity to the cysteine from the other, they will stick together or cross-link. Once these proteins are expressed in E. coli and the cells are broken open, the concentration of this cross-linked species will be proportional to the frequency with which the relevant side-chains encounter one another. As you can see, the two structures imply very different predictions about the patterns of cross-linking that will be observed. If the first structure is correct, the successful disulfide bonds should be distributed evenly along the helix. By contrast, if the second structure is correct, the mutations that produce cross-links will all be located at one end.

Goldberg et al. observe the first of these patterns. The side chains that are visible in the figures above are the ones where mutating in a cysteine produced a significant amount of cross-linking. Obviously they are distributed along the entire length of the N-terminal helix. When the researchers ran a simulation that attempted to predict a structure on the basis of the cross-linking data, they also came up with a structure that looked very similar to the first crystal structure above. While not all of the cross-links observed appear to be compatible with the structure, this may just be a result of a few additional conformations in the ensemble.

In this case, the experiment was performed on the E. coli protein, so it is still possible that the structure for the S. enterica protein reflects actual differences arising from the 15% difference in homology. Obviously the next step would be to perform the similar set of experiments on the S. enterica protein; doubtless these experiments are already underway. As the authors note, it is also possible that the two structures represent the active and inactive forms of the protein. Repeating the experiment in minimal media and adjusting the levels of magnesium may address this possibility.

1. Goldberg, S., Soto, C., Waldburger, C., DeGrado, W. (2008). Determination of the physiological dimer interface of the PhoQ sensor domain. Journal of Molecular Biology DOI: 10.1016/j.jmb.2008.04.023