Compatibility of the antiparallel-to-parallel switch

Structural information on the signalling mechanism of the cytosolic PAS domain is scarce. The overall destabilization signalling mechanism has been proposed for E.

coli DcuS PASc relying on ssNMR and mutational studies (Etzkorn et al., 2008;

Monzel et al., 2013). The same model was proposed also for G. thermodenitrif-icans CitA, based on ssNMR studies of the CitApc segment (Salvi et al., 2017).

However, as shown for NarQ (Gushchin et al., 2017), signalling mechanisms are not so simple and can be a combination of many motions. Thus, alongside with the overall destabilization of the cytosolic domain, eventual structural rearrange-ments may happen and be relevant to the signalling relay from the PASc to the DHp domain.

Since the crystal structures of the CitA PASc showing alternative dimer confor-mations of figure 3.5 were solved, it was speculated that the antiparallel and open arrangements may be relevant for the signalling mechanism inG. thermodenitrifi-cans CitA. Following this idea, the existence of the alternative dimer arrangements was confirmed in the isolated PASc domain. Later, the cytosolic domain of the CitApc was studied to investigate whether the alternative dimeric conformations not only can be adopted in the context of the membrane, but that the rearrange-ment is triggered by the binding of the citrate in the periplasmic domain. The results just discussed show unambiguously that the bound state of CitApc cytoso-lic PAS domain adopt the parallel dimer arrangement, as in figure 3.5 A. On the other hand, even if sPREs still have to be measured for the citrate-free CitApc, the anti-parallel dimer (figure 3.5 C) seems to resemble the conformation of the CitA cytosolic PAS domain in absence of the ligand, based on overall visibility and amide exchange properties (see paragraph 4.3.3). Thus, if supported by sPRE data on the citrate-free CitApc, the antiparallel-to-parallel (see figure 4.15) switch

4. Discussion 70 would be the relay mechanism from the PASc to the DHp domain, while the open dimer, which could not be associated with the CitApc, either is just adopted in the isolated PASc or represents a low populated, non-detectable intermediate.

Figure 4.15: The antiparallel-to-parallel switch proposed for the signalling mechanism of G. thermodenitrificans CitA PASc.

The suggested mechanism involves certainly some drastic rearrangement of the dimeric conformation and its compatibility with previous findings on HKs needs to be discussed.

The antiparallel-to-parallel switch complements the overall destabilization model proposed previously (Etzkorn et al., 2008; Salvi et al., 2017): based on overall NMR visibility and amide exchange properties, the PASc wild-type, which adopt predominantly the parallel arrangement, is less rigid than the PASc N288D, which is found mainly in the anti-parallel dimeric conformation. Still, the swap from an-tiparallel to parallel arrangement may result not easy to conceive. First, because conformations alternative to the parallel dimer have never been reported for PAS domains and second, because the inter-subunit distance at the C-termini and at

4. Discussion 71 the N-termini of the anti-parallel structure raises questions about the compatibil-ity of this arrangement with the functional dimeric state of the full-length CitA, as explained below.

In the anti-parallel structure, the N-terminal helices run almost parallel to the membrane surface, in contrast to what observed in NarQ (Gushchin et al., 2017).

Gushchin et al. (2017) reported that the TM2 is connected to the N-terminal helix of the cytosolic HAMP domain without any break in the secondary structure in both holo- and apo-state (Gushchin et al., 2017). This would be possible for the parallel dimer, but unlikely for the anti-parallel. In addition, the 10 ˚A displace-ment at the N-termini occurring in the antiparallel-to-parallel switch was never reported and it is in contrast with what reported for E. coli NarQ. It needs to be considered however that a twelve residues linker is located between the end of the TM2 and the first amino acid of the PASc domain, while only five residues are present in NarQ between the end of TM2 and the first amino acid of the HAMP domain. Thus, the longer linker of G. thermodenitrificans CitA could permit the antiparallel-to-parallel switch, allowing the PASc domains to adopt an atomic mechanism different from the one observed in NarQ for the HAMP domain.

Also the displacement at the C-termini is much larger in the antiparallel-to-parallel switch than what reported for NarQ. In the latter, a difference of 6 ˚A is observed at the C-termini between the holo and the apo structure (Gushchin et al., 2017), while the largest motion estimated for CitA using the proposed model is about 20 ˚A. In addition, it can be questioned whether the 6 ˚A motion observed at the C-termini of the NarQ crystal structure directly triggers for the kinase activity, as more than hundred residues are located between the end of the HAMP domain and the beginning of the kinase core. The displacement at the HAMP domain could then be amplified or further modified before it reaches the kinase core. On the other hand, only 21 residues are located between the end of the PASc and the kinase core in the G. thermodenitrificans CitA. The length of the linker guaran-tees that the output signal of the PASc is relayed directly to the DHp and the kinase core and, at the same time, it could allow the antiparallel-to-parallel switch.

4. Discussion 72