To follow chemical gradients, cells of Escherichia coli
utilize several types of transmembrane receptors, including the Tsr receptor that senses serine and the Tar receptor that senses aspartate1–3
. These sensors form mixed trimers of homodimers, even in the absence of other chemotaxis proteins4,5
, and together with a coupling protein, CheW, and the kinase, CheA, form large receptor arrays6–8
. These arrays were predicted to facilitate receptor collaborative operation9,10
. Indeed, synergetic effects between homodimers of different types have been demonstrated both in-vivo11
. Here we aimed to test directly whether coupling between homodimers can emerge due to direct physical interactions between neighboring dimers in receptor oligomers. To avoid coupling modes involving cytoplasmic proteins, we made measurements in strains in which CheW and CheA are not expressed. Note that while large clusters appear to be the dominant receptor state in the presence of CheW and CheA, it is the trimer-of-dimers structure that is the dominant state in their absence4,5,13
. We showed previously that changes in the relative orientation of adjacent homodimers in response to binding of ligand can be detected by labeling receptors with YFP and measuring fluorescence anisotropy14,15
. Changes in fluorescence anisotropy were traced to changes in energy transfer, i.e., to homo-FRET, that occur upon ligand binding. When serine (aspartate) is added to trimers of homodimers of Tsr-YFP (Tar-YFP), the YFPs move farther apart. These responses are absent in mutants that fail to form trimers and can be diluted out by expression of unlabeled receptors14,15
The rationale of the experiments described here is the following: by forming mixed oligomers composed of homodimers of fluorescently labeled Tar and homodimers of unlabeled Tsr, we can selectively stimulate either type of receptor while reading out changes in the relative orientation of the labeled receptors. Thus, we can directly address the following question: does binding of ligand to the Tsr dimer trigger conformational changes in the Tar dimers? This strategy is illustrated in . Two extremes are possible: (i) In the absence of coupling, the conformational response of Tsr will not affect Tar and the anisotropy level will not change; (ii) if, however, the trimers form a single allosteric unit (tight coupling) the binding of ligand to Tsr can only toggle the whole structure from the active to the inactive conformation, resulting in a change in anisotropy similar to that observed upon binding of ligand to Tar. The actual behavior turns out to be more complex.
Figure 1 The effect of direct and indirect stimuli on Tar-YFP. (a) The experimental rationale. A mixed trimer is shown containing one Tsr dimer (red) and two Tar-YFP dimers (green). The conformation of the former can be affected directly by binding of serine and (more ...)
We measured the anisotropy in cells without other chemotaxis proteins (flhD
cells) expressing different levels of Tsr, together with Tar-YFP, . At low levels of Tsr expression (lower, blue trace) the addition of serine had little effect, while the anisotropy increased upon addition of aspartate, as observed previously15
; the YFPs moved farther apart. At higher levels of Tsr expression (upper, red trace), the baseline anisotropy was higher and increased only marginally upon addition of aspartate, as expected from dilution of the YFP label in mixed trimers. However, serine lowered the anisotropy and increased the response to aspartate, as if binding of serine to Tsr in mixed trimers caused the Tar-YFPs to move closer together. Thus, the response of Tar-YFP homodimers to binding of serine to Tsr dimers in a mixed trimer, albeit small, was opposite of what one might expect.
The size of the downward shift in anisotropy (, red curve) increased with the concentration of serine, as shown in (inset). The corresponding dose-response curve is shown in the main plot (filled circles) together with a similar dose-response curve measured with a pure population of Tsr-YFP receptors (empty circles) and the curve measured in vitro16
for binding of serine to Tsr (line). The near identity of these curves confirms that the downward shift observed with mixed trimers (, red curve; and, , inset) is due to the response of the Tsr homodimers to binding of serine, even though it is being read out through a conformational response of the Tar-YFP homodimers. We confirmed that the effect is not unique to Tsr by reversing the roles of Tar and Tsr (data not shown). We also confirmed that the effect appears when Tsr and Tar-YFP are expressed in a strain that is deleted for all other chemotaxis components (UU1581; data not shown). We obtained similar results with constructs where the flexible tail at the C-terminal end of Tar was either present or absent (when YFP was fused to the C-terminus of wildtype Tar or to Tar1-527
, missing 27 C-terminal residues15
), confirming that the effect does not rely on a possibly large effective volume occupied by the YFPs. Thus, we conclude that binding of ligand to one homodimer, e.g., Tsr, in a mixed trimer affects the conformations of the other homodimers, e.g., Tar. However, it does so in a manner that is qualitatively different from that triggered by direct binding of ligand to Tar. Finally, we find that when both the indirect and direct stimuli are present (e.g. both serine and aspartate) the direct stimulus is dominant (, red curve).
Figure 2 The dependence of the downward anisotropy shift of on the concentration of serine. The inset shows the raw data for one sequence with cells similar to those of in which the concentration of serine was toggled between 0 and 10, 30, 100, (more ...)
In principle, an alternative interpretation of these results could be that the changes in anisotropy are due to dissociation of receptor oligomers upon ligand binding. Although evidence has been presented elsewhere against such a view (see discussion in ref. 15
and references therein), we further tested this possibility here by varying the location of YFP within the receptor structure. If receptor oligomers dissociate, the efficiency of energy transfer between YFP’s on different dimers should markedly decrease, regardless of where the YFPs are placed. However, if the response is due to conformational changes within the trimers, the sign and/or magnitude of the response should depend upon the position and orientation of the YFP, and thus could be different at different locations. We inserted YFP into the cytoplasmic tip of a full-length Tsr receptor at position 390 (for details see the legend of ). This construct is termed Tsr-YFP* and is shown schematically in . When cells expressing this construct where stimulated with serine, a clear anisotropy response was observed () but its sign was opposite of that found with the C-terminal fusion to Tsr. When Tsr-YFP* was co-expressed with increasing levels of unlabeled Tar, the baseline anisotropy increased and the amplitude of the response to addition of serine decreased (), as expected from dilution of the YFP label within mixed oligomers. When cells expressing Tsr-YFP* with unlabeled Tar at an intermediate level were stimulated by addition of aspartate, an anisotropy response also was observed () but again with a sign opposite of that seen with serine (). Thus, overall, the YFP insertion at the distal cytoplasmic tip of the dimers had a behaviour qualitatively similar to that observed with YFP close to the membrane at the C-terminal end of the receptor, but with smaller amplitude and opposite sign. These findings provide strong evidence for the view that the changes in anisotropy seen in our experiments reflect conformational changes in receptor oligomers rather than dissociation of these oligomers.
Figure 3 Anisotropy traces with YFP fused to the cytoplasmic tip of a receptor homodimer. (a) The position of YFP in the construct Tsr-YFP*: tsr1-390–yfpA206K –tsr390-end cloned into pTrc99A vector to form plasmid pAV48. (b) An anisotropy trace (more ...)
A few qualitative characteristics of trimer responses are illustrated in , which shows the trimer of viewed from the periplasmic side. In the absence of any ligand the trimer is in the active conformation, with the C-terminal ends of the homodimers close to one another (left image). Ignoring possible structural differences between Tar and Tsr, this conformation would be expected to have a three-fold symmetry. When the ligand binding sites are fully occupied, the trimer switches to the inactive conformation, with the C-terminal ends of the homodimers farther apart (right image). This conformation also has a three-fold symmetry. However, when the ligand binding sites are partially occupied, for example when only serine is bound to Tsr, the symmetry of the trimer is broken, and an asymmetric conformation is possible (middle image). To account for the data presented in , the C-terminal ends of the Tar homodimers should be closer together than they are in the absence of serine. On the other hand, to account for the monotonic change in anisotropy with ligand concentration observed with homogeneous Tar-YFP trimers or with homogeneous Tsr-YFP trimers (, open symbols), the mean distance between dimers should be larger than observed in the absence of ligand. These constraints are met by the middle image of . It is clear that the actual conformational change of trimers is more complex, as indicated by the differences in sign of the anisotropy responses of the Tsr-YFP and Tsr-YFP* constructs (see also ref. 17
); however, the requirement for qualitatively different conformations in the partially or fully bound state should be general. This requirement means, in particular, that trimers on their own do not form tightly coupled ‘two-state’ allosteric units of the sort proposed18–20
to explain recent measurements of kinase activity11,15,21–23
. Yet, the conformations of neighbouring dimers are coupled, supporting the view5,24–26
that dimers act collaboratively in trimers to propagate signals to the associated kinase proteins. Additional contributions to receptor coupling could emerge in fully assembled receptor clusters, from direct trimer-trimer interactions or via their association with CheW and CheA.
Figure 4 Characteristics of trimer responses. A trimer containing one Tsr dimer (red) and two Tar dimers (green) is shown viewed from the periplasmic side. In the absence of any ligand, the trimer has three-fold symmetry (left image). Upon binding of serine to (more ...)