d-amino acids signal a stress-dependent run-away response in Vibrio cholerae

To explore favourable niches while avoiding threats, many bacteria use a chemotaxis navigation system. Despite decades of studies on chemotaxis, most signals and sensory proteins are still unknown. Many bacterial species release d-amino acids to the environment; however, their function remains largely unrecognized. Here we reveal that d-arginine and d-lysine are chemotactic repellent signals for the cholera pathogen Vibrio cholerae. These d-amino acids are sensed by a single chemoreceptor MCPDRK co-transcribed with the racemase enzyme that synthesizes them under the control of the stress-response sigma factor RpoS. Structural characterization of this chemoreceptor bound to either d-arginine or d-lysine allowed us to pinpoint the residues defining its specificity. Interestingly, the specificity for these d-amino acids appears to be restricted to those MCPDRK orthologues transcriptionally linked to the racemase. Our results suggest that d-amino acids can shape the biodiversity and structure of complex microbial communities under adverse conditions.

To explore favourable niches while avoiding threats, many bacteria use a chemotaxis navigation system. Despite decades of studies on chemotaxis, most signals and sensory proteins are still unknown. Many bacterial species release d-amino acids to the environment; however, their function remains largely unrecognized. Here we reveal that d-arginine and d-lysine are chemotactic repellent signals for the cholera pathogen Vibrio cholerae. These d-amino acids are sensed by a single chemoreceptor MCP DRK co-transcribed with the racemase enzyme that synthesizes them under the control of the stress-response sigma factor RpoS. Structural characterization of this chemoreceptor bound to either d-arginine or d-lysine allowed us to pinpoint the residues defining its specificity. Interestingly, the specificity for these d-amino acids appears to be restricted to those MCP DRK orthologues transcriptionally linked to the racemase. Our results suggest that d-amino acids can shape the biodiversity and structure of complex microbial communities under adverse conditions. Most bacteria can sense and respond to a wide variety of signals to thrive in changing environments. One such adaptive response is chemotaxis, by which motile bacteria monitor changes in local concentrations of certain substances and initiate a signalling cascade that, over time, biases bacterial movement towards favourable conditions (for example, nutrients) or away from toxic compounds 1 .
Albeit species-specific components exist, the core of the chemotaxis signalling pathway consists of chemoreceptors, known as methyl-accepting chemotaxis proteins (MCPs), the adaptor protein CheW, the histidine kinase CheA and the response regulator CheY. In some species such as Escherichia coli, the MCPs form a stable ternary sensory complex with CheW and CheA that typically clusters at cell poles. Upon ligand binding or other environmental perturbations, the MCPs change conformation and modulate CheA kinase activity with the help of CheW. CheA in turn phosphorylates CheY, thereby allowing it to bind to the flagellar motor to induce clockwise rotation and, thus, cell tumbling 2,3 . To ensure a rapid response to chemotactic stimuli, the phosphorylation of CheY can be controlled by a specific phosphatase 4 . Although the chemotaxis signalling core components are highly conserved among bacteria and archaea, the number and type of chemoreceptors are highly variable between species and their specificities remain largely uncharacterized [5][6][7] .
Compared to E. coli, which has a single chemosensory system and 5 MCPs, the facultative pathogen Vibrio cholerae has a very sophisticated chemotaxis system organized into three sets of chemosensory pathways (Che systems I/F9, II/F6 and III/F7) and at least 45 chemoreceptors 8 . Thus far, only system II/F6 has been demonstrated to control motility 9 , while the function of the others remains unclear. The high number of MCPs has been suggested to reflect the complexity of V. cholerae's life cycle 10 ; however, except for a handful of input signals (for example, l-amino acids 11 , taurine 12 , oxygen 13 ) there is almost no information about the ligand specificity of most MCPs. Article https://doi.org/10.1038/s41564-023-01419-6 recapitulated by a catalytically inactive mutant (BsrV K95A, ref. 23), thus suggesting that this phenotype was directly associated with BsrV's activity to produce d-amino acids. To ascertain that the absence of secreted d-amino acids accounted for the motility defect that we found, we chemically complemented the ΔbsrV mutant by supplementing the soft-agar plates with innocuous concentrations of individual DAAs. Their corresponding l-enantiomers were used as controls (Extended Data Fig. 1). We found that only d-arginine (d-Arg) and d-lysine (d-Lys) restored the motility of the ΔbsrV mutant to wild-type levels (Fig. 1d), suggesting that these specific DAAs play a specific role in V. cholerae's ability to spread in soft agar.
To narrow down the role of d-Arg and d-Lys in V. cholerae's motility, we first assessed the integrity of the flagellum. However, ΔbsrV cells showed no significant differences in flagellum composition or structure compared to the wild type ( Supplementary Fig. 1). As motility is functionally related to chemotaxis, we reasoned that d-Arg and d-Lys modulate V. cholerae's chemotactic response. In this line, we tested the activity of diverse DAAs in capillary assays. We used the ΔbsrV genetic V. cholerae releases millimolar concentrations of diverse d-amino acid (DAAs) into the environment 14,15 (Fig. 1a). DAAs are produced from their l-amino acid (LAA) counterparts by the broad-spectrum racemase BsrV, which is conserved in many bacterial species 14 . These molecules regulate diverse cellular processes (for example, cell wall biogenesis 16,17 , biofilm integrity [18][19][20][21] , spore germination 22 or bacteria-bacteria interactions 15 ) but overall, the physiological role of extracellular DAAs is mainly dependent on each specific d-amino acid type and bacterial species. In V. cholerae, it is known that d-Met and d-Leu modulate cell wall biogenesis, but the function of other DAAs is unclear.

d-arginine and d-lysine are chemorepellent signals
While screening for DAA-specific phenotypes, we found that a V. cholerae strain defective in d-amino acid production (ΔbsrV) exhibited a 40% motility reduction compared to wild-type cells in soft-agar plates. This reduction was fully restored upon ectopic genetic complementation with the bsrV racemase (Fig. 1b,c)   background (control hereafter) to prevent BsrV-mediated racemization of the amino acids during the assays. Importantly, this strain has wild-type chemotactic and swimming ability towards previously characterized stimuli (for example, l-Arginine, α-aminoisobutyric acid; γ-aminobutyric acid, succinate) (Extended Data Fig. 2). Interestingly, we found that both d-Arg and d-Lys promote repulsive chemotactic responses in V. cholerae ΔbsrV, while the other DAAs did not affect the bacteria (Fig. 1e). To confirm that the motility defect of V. cholerae ΔbsrV is exclusively related to chemotaxis signalling and not due to a change in V. cholerae's swimming capacity, we used a permanently active phosphomimetic cheY3 D16K,Y109K mutant (hereafter referred to as CheY3**). As previously described 24,25 , CheY3** cells are non-chemotactic because they are locked in tumbling mode but maintain the ability to swim. Our results show that the spread of a ΔcheY3 strain complemented with CheY3** on soft-agar plates was unaffected by d-Arg (Extended Data Fig. 2c), thus indicating that this DAA indeed does not regulate V. cholerae swimming. Collectively, these results show that d-Arg and d-Lys sensing induces a repellent chemotactic response in V. cholerae.

VC1313 is the chemoreceptor of d-Arg and d-Lys
The V. cholerae chemotaxis machinery comprises at least 45 putative MCPs 8 , some of which are known to be functionally redundant (for example, Mlp24 and Mlp37, which sense several LAAs) 11,12 . Therefore, to screen for MCP candidates sensing d-Arg and d-Lys, we opted for two-dimensional thermal proteome profiling (2D-TPP) (refs. 26,27) over mutagenesis. TPP is based on the principle that protein thermal stability can be affected by interactions with ligands 28 . In our TPP experiment, we treated V. cholerae cells with different concentrations of d-amino acids and quantified the resulting changes in protein thermal stability using mass spectrometry (Extended Data Fig. 3a). Considering the structural similarity and comparable chemotactic responses of d-Arg and d-Lys, we focused on d-Arg for our study. Our results revealed ~223 proteins affected in their thermal stability (false discovery rate, FDR <0.01) by d-Arg (Fig. 1f, Extended Data Fig. 3b, and Supplementary Tables 1 and  2). Some 185 proteins presented a decreased thermal stability, while 38 displayed increased thermal stability, including three putative MCPs: VC2161, VC1406 and VC1313. To further investigate whether any of these chemoreceptors reacts to d-Arg, we generated individual MCP knockouts in the ΔbsrV background and examined their chemotactic response to d-Arg. Remarkably, we found that the d-Arg response of the Δvc2161 and Δvc1406 strains was comparable to that of the control, while the Δvc1313 mutant did not respond to d-Arg. Notably, the d-Arg response in this mutant was fully restored upon genetic complementation with vc1313 ( Fig. 1g). We found that ΔbsrV Δvc1313 cells were also unresponsive to d-Lys (Extended Data Fig. 4a), indicating that this MCP senses both DAAs. Additional capillary and microfluidic-based assays 29 with a broader range of amino acid concentrations narrowed VC1313's repellent response to d-Arg concentrations in the 0.01 mM to 1 mM range (Extended Data Fig. 4). Interestingly, 1 mM d-Arg showed a slightly positive chemotactic response in ΔbsrV Δvc1313 cells. This suggests that alternative lower-affinity chemoreceptors sense d-Arg as an attractant. However, our results show that vc1313 encodes the MCP that senses d-Arg and d-Lys to initiate a repulsive chemotactic response in V. cholerae; therefore, we renamed this protein as MCP DRK for MCP sensing d-Arg (R) and d-Lys (K).

RpoS sigma factor regulates MCP DRK
The mcp DRK locus is localized in the vicinity of vc1312, which encodes BsrV-the broad-spectrum racemase that produces d-Arg and d-Lys (Extended Data Fig. 5a). Analysis of the expression profiles across different publicly available RNAseq datasets for V. cholerae strongly suggests that these two genes are co-regulated and could form an operon (Extended Data Fig. 5b). Using promoter-probe plasmids, we found a promoter region upstream of vc1313 (Pvc1313) but not of vc1312 (Extended Data Fig. 5c). Moreover, Pvc1313 activity increases in stationary phase, consistent with a previous study showing that BsrV is induced under the same conditions, in an RpoS-dependent fashion 30 (Extended Data Fig. 5d).
To confirm this result, we decided to monitor the expression and localization of MCP DRK labelled with super-folding green fluorescence protein (sfGFP-MCP DRK ) throughout V. cholerae growth. As predicted, cultures inoculated with stationary phase cells displayed polar MCP DRK localization in nearly 90% of the cells. This percentage rapidly dropped as the cells started to grow until no foci were detected at the onset of exponential growth (Fig. 2a). As cultures plateaued, polar foci reappeared, reaching the starting levels. We found a direct correlation between sfGFP-MCP DRK foci and protein levels at different optical densities by western blot (Fig. 2b). This demonstrates that expression of MCP DRK in V. cholerae is indeed linked to the stationary phase. As RpoS is highly induced during glucose starvation, we cultured V. cholerae on a mixture with glucose and succinate and monitored expression and localization of MCP DRK . Similar to other bacteria, V. cholerae cells follow a diauxic growth caused by catabolite repression; cells consume d-glucose first, followed by a short pause, then resuming growth by metabolizing succinate until it is spent 30 (Fig. 2c). We again observed that the stationary phase cells from the inoculum exhibited polar sfGFP-MCP DRK foci in basically all cells, but these vanished when growth started. Interestingly, foci rapidly increased to maximum levels when cells entered the first growth arrest caused by glucose depletion. When cells resumed growth on succinate, foci decreased to ~60% and rose up again when succinate was consumed. These results indicate that the induction and polar localization of MCP DRK respond to growth arrest rather than high population densities.
To confirm that MCP DRK 's expression depended on RpoS, we cloned the sfGFP-MCP DRK reporter into a ΔrpoS mutant background. In contrast to wild-type stationary phase cultures where most cells displayed foci, the ΔrpoS cells showed no polar localization of the MCP (Fig. 2d). Consistently, ΔrpoS phenocopied the lack of DAA production and taxis response to d-Arg of the ΔbsrV and Δmcp DRK mutants (Fig. 2e,f), respectively. These results demonstrate that V. cholerae displays a synchronic RpoS-dependent DAA production and MCP DRK -chemotactic response during nutrient deprivation stresses that lead to growth arrest.

MCP DRK forms part of the chemotaxis system II/F6
Similar to V. cholerae chemosensory cluster III/F7 proteins 30 , MCP DRK also forms polar foci during the stationary phase and its expression is RpoS-dependent. However, based on domain length and sequence conservation, MCP DRK belongs to the 40H (heptad) class of chemoreceptors 31 . These are membrane-bound, polarly localized MCPs predicted to signal through the chemotaxis system II/F6-the only one so far related to motility in this bacterium 9,32 . To experimentally determine the chemotaxis system that transduces the MCP DRK signalling, we constructed mutants with complete deletions of the I/F9, II/F6 and III/F7 chemotaxis systems and analysed their response to d-Arg. Neither deletion of I/F9 nor of III/F7 reduced V. cholerae's negative chemotactic response to d-Arg, strongly suggesting MCP DRK association to the II/F6 chemosensory system (Extended Data Fig. 6a). Further evidence was obtained by investigating the localization of the MCP DRK chemoreceptor in these mutant backgrounds. MCP DRK foci overlap with those of CheY3 of the chemotaxis cluster II (Extended Data Fig. 6b) and are lost in the II/ F6 but not in the III/F7 or I/F9 mutants, confirming that MCP DRK belongs to the chemotaxis system II/F6 (Extended Data Fig. 6c).

Structure of the sensory module of MCP DRK
MCP DRK is a 59.1 kDa protein (Uniprot: Q9KSE4) that is predicted to have the typical transmembrane chemoreceptor domain architecture. MCP DRK contains a ligand binding domain (LBD) predicted to belong to the 4HB_MCP protein family (Pfam PF12729) as a sensory module, a transmembrane HAMP domain (Pfam PF00672) and a cytoplasmic MCP-signalling domain (Pfam PF00015) as output module Article https://doi.org/10.1038/s41564-023-01419-6 ( Fig. 3a). To elucidate the molecular mechanisms of ligand recognition by MCP DRK , we purified the LBD domain of MCP DRK (20 kDa) with a Glutathione S-transferase (GST) tag that was cleaved off before the final size-exclusion chromatography (SEC) step. This purified LBD was monomeric in solution, both in the absence and presence of its d-Arg ligand, as determined by multi-angle laser light scattering (SEC-MALS, see Supplementary Fig. 2). We solved two X-ray crystal structures of MCP DRK 's LBD in complex with either d-Arg or d-Lys at 1.8 Å resolution (PDB 8BSA and 8BSB, Supplementary Table 3). The crystals belonged to space group P1 and contained two LBD molecules, organized head-to-tail, in the asymmetric unit. This organization in the crystal is highly unlikely to be a physiologically relevant assembly since the membrane-proximal subdomain of one subunit packs against the membrane-distal subdomain of the other. As the monomers in the asymmetric unit were virtually identical and both were ligand bound, we here only show and discuss the A-chain of each structure. The final structure models correspond to residues 32-187 or 32-189 for the d-Arg and d-Lys-bound structures, respectively.
Overall, MCP DRK -LBD forms an antiparallel four-α-helix bundle structure (4HB) (Fig. 3b). This fold is distinct from the dCACHE LBD-type fold that is found in other chemoreceptors from V. cholerae that respond to l-amino acids (Mlp24, Mlp37) (refs. 11,12), but has been found in various LBDs from other Gram-negative bacteria. A structure-based homology search via the DALI server 33 showed that MCP DRK -LBD was structurally most similar to the citrate-binding chemoreceptor MCP2201 from Comamonas testosterone (PDB ID: 5XUB, RMSD of 1.7 Å) (ref. 34) and the more promiscuous PcaY_PP chemoreceptor from Pseudomonas putida (PDB ID: 6S38, RMSD of 2.1 Å) (ref. 35), but also showed similarity to the well-characterized aspartate-binding Tar receptor from E. coli (PDB ID: 4Z9I, RMSD of 2.5 Å) (ref. 36). While these LBDs share a similar overall fold, their ligand binding sites are at entirely different locations (Fig. 3c). For instance, the model receptors Tar and Tsr bind ligands at the dimer interface with negative cooperativity [37][38][39] , while MCP DRK binds d-Arg and d-Lys centrally and remains as monomer instead ( Fig. 3c and Supplementary Fig. 3). As for MCP DRK , the chemoreceptor MCP2201 from C. testosterone also binds the ligand to a single protein chain, but at a different location (Fig. 3c) 34 .
In the MCP DRK -LBD structures, both d-Arg and d-Lys bind at the same site, at the bottom of the second and the third helix ( replicates. Error bars represent mean ± s.d. of 3 biologically independent replicates. Significant differences (unpaired t-test) are indicated by ***P < 0.001. f, DAA production by the different strains. Significant differences (unpaired t-test) are indicated by **P < 0.01 or ***P < 0.001.

Article
https://doi.org/10.1038/s41564-023-01419-6 a salt bridge with Glu111. The ligands are further stabilized by hydrogen bonds with residues from all helices, namely Asn43, Thr48, Thr105, Glu111 and Asn176 (Fig. 3d,e). Trp107, from the short loop between helix 2 and 3, forms a lid that closes off the binding site and has a stacking interaction with the ligands.
To further characterize the precise mechanism of ligand specificity, we generated seven individual mutants in which each amino acid residue that interacts with the ligands according to the crystal structures was replaced with alanine (N43A, D47A, T48A, T105A, W107A, E111A and N176A). Complementation assays were performed by introducing these mutant alleles under the control of mcp DRK native promoter interrupting the lacZ locus in the Δmcp DRK strain. Chemotaxis assays demonstrated the essentiality of 5 of the 7 MCP DRK -LBD residues tested (N43, T48, W107, E111 and N176) to promote a repulsive response to d-Arg in V. cholerae (Fig. 3f). d-Arg-mediated taxis was not affected by the N43A and T105A mutations, suggesting a less relevant role for these residues in stabilizing the d-Arg and d-Lys. These in vivo assays thus confirmed the location of the ligand binding site that was found in the crystal structure.

Conservation of MCP DRK
To analyse the conservation of MCP DRK, we built a phylogenetic tree on the basis of sequence similarity (Fig. 4a). We found that this chemoreceptor is widely conserved in the Gamma-proteobacteria orders Aeromonadales, Alteromonadales and Vibrionales, which are known to be polar flagellated bacteria that are commonly found in aquatic environments. MCP DRK -like proteins cluster into three groups on the basis of their percentage of homology and genomic association with the BsrV racemase (Extended Data Fig. 7). MCP DRK orthologues with the highest protein identity (>50%) cluster together with a BsrV-like racemase (for example, many Vibrio species and Moritella sp.), suggesting coevolution of d-amino acid production and chemotactic recognition. In these orthologues, the d-Arg/d-Lys binding site residues are either fully conserved or have only subtle changes (for example, V. metschnikovii, V. cincinnatiensis and V. anguillarum; see Extended Data Fig. 7). Species with MCP DRK orthologues of slightly lower identity (48-51%) (for example, Aeromonas and Aliivibrio) form a single clade and typically show a substantial change in two important binding site residues: D > S in position 47, resulting in a loss of the salt bridge with the ligand, and W > G in position 107, which removes the stacking interaction with the ligand. Remarkably, in these species, the BsrV-like racemase does not form an operon with the chemoreceptor but is encoded elsewhere in the genome. Finally, the species encoding MCP DRK homologues of lower identity (<45%) (for example, V. owensii) contain neither the d-Arg/d-Lys binding residues nor a BsrV-like locus. To assess the potential functional divergency of the MCP DRK chemoreceptor across species, we selected three orthologues from the pool of representative genomes that we used to build the phylogenetic tree. These were V. anguillarum VAA_1440 and V. parahaemolyticus VPA_1000, which cluster with a BsrV-like racemase and show a conserved binding site. We also selected A. fischeri VF_A1069, which does not form an operon with the racemase and has a partially conserved binding site (Fig. 4b). LBD model analysis on models generated by alphafold [40][41][42] predicted defective d-Arg stabilization caused by changes in the VF_A1069 binding site as compared to the canonical one (that is, V. cholerae) (Supplementary Fig. 3). To study this, we monitored by capillary assays the response to d-Arg of the Δmcp DRK mutant strain complemented with the selected orthologues expressed under the mcp DRK native promoter. As expected, we found that expression of VAA_1440 (68.6% identity, 100% conserved binding site) and VPA_1000 (65.7% identity, 100% conserved binding site) restored the ability of the Δmcp DRK mutant to sense d-Arg. The expression of VF_A1069 (48.1% identity, 71% conserved binding site) did not complement the chemotactic response to d-Arg (Fig. 4c), probably because this orthologue does not conserve the D47 and W107 residues ( Supplementary Fig. 3) that were found to be essential for d-Arg binding (Fig. 3). Together, these results suggest that MCP DRK -like proteins specific for d-Arg and d-Lys are those encoded nearby a bsr locus.

Discussion
Bacterial chemotaxis has been the subject of decades of research. Although the signal transduction components have been extensively investigated, less is known about the identity and specificity of the chemoreceptors that bacteria use to sense their surroundings 43 . Given the intimate association between MCPs and a bacterium's lifestyle, these chemoreceptors are generally numerous and specific, suggesting that many MCP-ligand pairs remain uncharacterized 6 .
We previously reported that many bacteria release high concentrations of a variety of d-amino acids to the environment. d-amino acids display different biological functions depending on their chemical properties. For instance, stationary-phase V. cholerae cells produce d-Met to adapt their cell wall biogenesis to growth arrest mode 17 while they use d-Arg to inhibit the growth of nearby competitors 15 . While d-Arg is lethal to many species, V. cholerae's physiology appears to be unaffected by this d-amino acid 15 . Using a mutant defective in d-amino acid production, we report here that d-Arg is sensed by a previously unrecognized MCP to move the V. cholerae population away from stressful environments.
It is interesting that among all the d-amino acids produced by V. cholerae, only the toxic d-Arg (and d-Lys) induce negative chemotaxis. This means that d-Arg plays multifaceted roles in shaping the biodiversity and structure of microbial communities when nutrients   are scarce by clearing the environment of potential competitors while also migrating the community towards more favourable niches. For this 'fight and flight' strategy, V. cholerae has evolved to arrange the genes encoding the d-Arg-sensing MCP (MCP DRK ) and the broad-spectrum racemase (BsrV) together in one operon whose expression is controlled by the stress sigma factor RpoS. By synchronizing the production of d-Arg and its MCP, V. cholerae builds an efficient stress-dependent response while simultaneously preventing its futile activation under favourable conditions (Fig. 5). In fact, when the MCP DRK is not expressed, d-Arg stimulates a mild chemotactic attraction that suggests the existence of other MCPs with lower affinity for this ligand. Indeed, our TPP experiments identified potential MCP candidates whose stability is affected by d-Arg, such as VC2161 (Mlp24) and VC1406 (Fig. 1f).

S a li n iv ib ri o p ro te o ly ti c u s A lii vi br io fis ch er i Aer om ona s dha ken sis
Chemotaxis towards d-Arg may have a nutritional purpose. This idea is supported by results in E. coli showing that certain compounds with low nutritional value can be chemoattractants 44 and by the presence of a catabolic d-amino acid dehydrogenase enzyme 45 . The crystal structures of the LBD of MCP DRK showed that it has a typical antiparallel 4HB structure, which is conserved in various receptors 46 . Interestingly, this relatively simple fold has evolved to recognize a variety of substrates by using different binding sites. The binding site of Tar, Tsr and the PcaY_PP receptors is located at the dimer interface, while MCP DRK and MCP2201 bind their ligands centrally within the 4HB (Fig. 3c). MCP DRK 's d-Arg-binding pocket is located at the bottom of the second and the third helix, where only five residues are essential for recognizing this d-amino acid and similar compounds such as d-Lys. Contrary to the 4HB-containing model receptors, that is, Tar and Tsr, our purified MCP DRK -LBD construct does not oligomerize in the presence of the ligand. This is also in contrast with MCP2201, where ligand binding inside the monomer induces a structural rearrangement that prevents dimerization of the LBD, instead stimulating trimerization 34 .
Interestingly, although MCP DRK is broadly conserved among species from three Gamma-proteobacteria orders, the specificity for d-Arg/Lys seems to be evolutionarily restricted to receptors that are transcriptionally linked to broad-spectrum racemases such as BsrV. As some of these species are both tolerant to d-Arg toxicity 15 and inhabit aquatic habitats, d-Arg/d-Lys produced by one of these species could simultaneously coordinate the chemotactic response for many inhabitants of the same niche. Conversely, species that lack a Bsr homologue or encode it somewhere else in the genome display changes in the ligand binding domain that are likely to disable the chemotactic response to d-Arg/d-Lys (Fig. 4 and Extended Data Fig. 7). While some of these changes imply critical structural variations in the binding site (Supplementary Fig. 3), more subtle ones could instead recognize alternative signals that are chemically related to d-Arg. Indeed, multiple arginine analogues are known to be naturally produced, such as l-homoarginine in humans and animals 47 and l-canavanine and l-indospicine produced by certain plants 48 . Interestingly, canavanine can be racemized by microbial Bsr enzymes, suggesting that it should be possible to find both l-and d-enantiomers of arginine analogues in the natural environment 49 . Future studies should investigate how interspecies and interkingdom competing signals for the same MCP may impact chemotaxis regulation and hence bacterial capacity to respond to stress in complex polymicrobial environments.
The role of d-amino acids as chemotactic repellent signals adds to previously reported roles for these molecules in stationary-phase adaptation. An obvious ecological scenario where the d-Arg run-away response could be beneficial is abandoning spent environments. It is known that V. cholerae forms biofilms during the aquatic and intestinal phases of its life cycle 50 ; when the biofilm ages, RpoS controls biofilm detachment 51 . In a host environment, RpoS-dependent activation of chemotaxis and motility is linked to mucosal escape. Given that this response coincides with a reduction in the expression of virulence genes, it has been proposed as a preparation of V. cholerae for the next stage in its life cycle 52 . While several degradative proteins contributing to biofilm disassembly have been reported in recent years (for example, in Vibrio 25 ), the signal(s) promoting biofilm dispersal remain enigmatic. Therefore, an exciting question for the future will be whether d-Arg's dual function as a broad-spectrum toxin and chemotactic warning signal could play a role in biofilm destablization and dispersal. In this line, a study recently reported that the MCP DRK -BsrV operon (vc1313-1312) is part of the in-biofilm repressome and that constitutive expression of some of these genes interferes with biofilm formation 53 . However, the loss of MCP DRK was reported to have no effect on V. cholerae's fitness in infection or dissemination 54 . While the effect of d-Arg chemotaxis on virulence and transmission remains to be studied, the widespread presence of this chemoreceptor in other marine non-pathogenic bacteria suggests that this response could be an important and overlooked aspect of bacteria's ability to assess niche habitability.

Bacterial growth conditions
Strains are summarized in Supplementary Table 4. All V. cholerae strains are derivatives of the sequenced N16961, an El Tor clinical isolate 55 .
For growth curves, stationary-phase cultures were normalized to an OD of 0.04 and 20 μl were used for inoculating 96-well plates containing 180 μl of fresh medium. At least 3 replicates per strain and condition were inoculated in 2 independent experiments. Optical density was monitored using Biotek Gen5 (v.08) in an Eon Biotek plate reader (Biotek) at 5 min intervals at 30 °C.

Construction of mutants and plasmids
All mutants were generated by in vivo allelic exchange as previously described 56 . In brief, 1 Kb flanking regions were cloned into XbaI-digested pCVD442 to generate recombinant suicide vectors. For conjugations into V. cholerae, stationary-phase recipients and Sm10 λpir donor strains were washed of antibiotics, mixed in equal ratios and spotted onto LB without selection. After 4-6 h incubation at 37 °C, cells were streaked onto Cb/Sm-supplemented LB plates to select for transconjugants. Finally, cells were cured through a counterselection round on salt-free 10% sucrose plates. The chromosomal fluorescent sfGFP-vc1313 reporter was constructed in a similar way, directly inserting sfGFP upstream of vc1313 in-frame.
Plasmids and oligonucleotides used are summarized in Supplementary Tables 5 and 6, respectively. Plasmids were constructed by standard DNA cloning techniques. bsrV was PCR amplified using the appropriate primers including a strong ribosomal binding site (RBS) and cloned into restriction-digested pXB300 by isothermal assembly, yielding plasmid pbsrV. Similarly, for the generation of the phosphomimetic non-chemotactic mutant, cheY3 was PCR amplified with primers carrying the desired substitutions (CheY3 D16K, Y109K ), and the fragment was isothermally assembled into restriction-digested pXB300 to generate pcheY3**. Plasmids pbsrV and pcheY3** were introduced into ΔbsrV and ΔcheY3, respectively, via electroporation.
The DNA fragment encoding the entire periplasmic domain (residues Ser32-Ser191) of vc1313 (herein referred to as vc1313-LBD) was PCR amplified and subcloned into BamHI/EcoRI-digested expression vector pGEX-6P-2 by isothermal assembly so that the N-terminal GST-encoding sequence was fused in-frame to yield pGEX-vc1313-LBD. The stop codon located downstream of the multicloning region was used, therefore the resulting protein contained an additional C-terminal 12-amino acid arginine-rich tail.
Δvc1313 complementation was achieved by expressing a copy of vc1313 under the control of its native promoter introduced in neutral locus lacZ. vc1313 fragment was PCR amplified (including 500 bp upstream of the promoter region) and isothermal assembled into restriction-digested pJL1, a suicide plasmid that permits allele insertion into V. cholerae lacZ 57 . Successful integration was first verified by blue-white screening on plates containing 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-Gal, 40 μg ml −1 , Sigma, B4252) and later checked by PCR. Site-directed mutagenesis of the vc1313 gene was carried out following Q5 site-directed mutagenesis protocol (NEB) using pJL1::vc1313 plasmid as template.
The fidelity of all generated mutants and plasmids was confirmed by DNA sequencing.

Soft-agar motility assay
Single colonies of V. cholerae strains were grown overnight in LB broth at 30 °C with shaking, subcultured 1:100 in fresh broth and grown until exponential phase. A 2 μl volume of the normalized cultures were then spotted into 0.3% soft-agar plates and drops were allowed to dry at room temperature. When indicated, plates were supplemented with 5 mM of amino acids. After inoculation, plates were incubated for 8 h at 30 °C and then imaged in a LAS-3000 imaging system (Fuji). Image analysis and processing were performed using Fiji/ImageJ 58 and results were plotted using GraphPad Prism (v.8.0).

Capillary assays
Chemotactic ability was examined using previously described capillary assays 60 , with some modifications. An overnight culture of V. cholerae was diluted 1:200 into fresh M9 minimal medium and incubated until late exponential phase at 30 °C with shaking. Cells were collected by low-speed centrifugation, washed twice with motility buffer (MB) (50 mM HEPES (pH 7.4), 300 mM NaCl, 10 mM glucose, 5 mM MgCl 2 ) and finally resuspended to a final OD of 0.1 in MB. Tween20 was added to a final concentration of 0.01% to avoid cell adhesion. After preincubation of cells at r.t. for 1 h, pre-filled 1 μl microcapillary tubes (Sigma, P1421) containing the amino acids of interest (500 μM, unless specified otherwise) were placed in contact with the cell suspension and the suspension was incubated at r.t. for an additional 2 h. Finally, the number of bacteria (colony-forming units per ml) in each capillary was counted by plating serial dilutions on LB agar, and chemotaxis ratio was calculated relative to a control capillary.

Microfluidic chemotaxis experiments
Microfluidics chambers were fabricated as previously described 29 . Briefly, the SU8-based negative master mold of the microfluidic devices was produced by standard photolithography techniques. Microfluidics devices were produced from the mold using polydimethylsiloxane (1:10 crosslinker-to-base ratio, thermal-cured overnight at 65 °C), which was cut to shape and bound to glass slides via oxygen plasma treatment. After production, devices were stored short-term in sterile deionised water to retain their hydrophilic properties. V. cholerae strains were grown overnight in TB at 30 °C from glycerol stocks, diluted 1:100 in 10 ml M9 medium and grown for 4.5 h at 30 °C with shaking until an OD of ~0.3. The day culture was washed three times in motility buffer and diluted to a final OD of 0.1. The microfluidic device consists of two large chambers linked by a small straight channel (length = 2 mm, width = 1 mm, height = 70 μm). The suspension of cells at OD = 0.1 was introduced in one chamber, while the second was filled with MB supplemented with indicated amounts of the tested compound. The device's inputs were then sealed with microscopy grease to prevent evaporative flows. A gradient of the compound forms rapidly (~1 h) in the connecting channel. The resulting chemotactic motion of the cells was measured in the centre of the channel 1, 2 and 3 h after loading the device. To this end, the sample was observed under phase contrast microscopy at ×10 magnification (NA 0.3) and a 10,000-frames movie was recorded at a rate of 200 frames per second using a Mikrotron Eosens 4CXP CMOS camera with a 717 × 717 μm² (512 × 512 px²) field of view. A control with the same cell suspension but only MB in the second chamber was always performed simultaneously. shaking until an OD of ~0.6. The cells were then washed into fresh TB. A small volume (3 μl) of the suspension of cells was sandwiched between two coverslips maintained 100-200 μm apart by microscopy grease. Cell motion was recorded at mid-height between the coverslips using the same microscopy protocol as described above.

Swimming speed and chemotactic bias measurement
Movies were analysed offline using the differential dynamic microscopy (DDM) and phase differential microscopy (PhiDM) methods implemented as plugins in ImageJ (https://github.com/croelmiyn/FourierIm-ageAnalysis) to measure swimming speed 61 and chemotactic drift 62 , respectively. The DDM algorithm computes the time autocorrelation of the spatial Fourier components of the images, which are then fitted with a model assuming that a fraction of the cells is non-motile and diffusing, and the other is motile, with swimming speeds being Schulz distributed around a mean. Consistent fits across the different Fourier components yield the fraction of swimming cells Φ and their average swimming speed v 0 . With PhiDM, the temporal shift in the phase of the same Fourier components is fitted to measure the population-averaged drift velocity of the cells in the field of view, v d . In the absence of external flows, this drift is solely due to the chemotactic motion of the motile cells since the non-motile cells do not drift on average. The chemotactic drift velocity of the motile cells was therefore computed as v ch = v d /Φ. Since the chemotactic drift velocity is expected to scale with the swimming speed of the cells, we also computed the chemotactic bias b = v ch /v 0 = v d /Φv 0 , which must lie in the range (−1, 1) and estimates the angular motion bias of the cells up (b > 0) or down (b < 0) the gradient.
Thermal proteome profiling V. cholerae cells were grown overnight at 37 °C in LB and diluted 100-fold into 100 ml of fresh LB. Cultures were grown aerobically at 37 °C with shaking until an OD of ~2. Cells were pelleted at 4,000 × g for 5 min, washed with 10 ml PBS and resuspended in 1.2 ml of lysis buffer (final concentration: 50 μg ml −1 lysozyme, 250 U ml −1 benzonase and 1 mM MgCl 2 in PBS). Cells were lysed by 5 repeated cycles of freeze-thawing (freezing in liquid nitrogen, followed by 5 min at 25 °C with shaking until complete thawing). Then, d-Arg was added at 4 different concentrations (0, 0.05, 0.4, 3.2 mM) and 20 μl was aliquoted to a PCR plate and immediately subjected to a temperature gradient for 3 min in a PCR machine (Agilent SureCycler 8800), followed by 3 min at room temperature. NP40 was then added to all conditions to a final concentration of 0.8%. The plate was then centrifuged at 2,000 × g for 5 min to remove cell debris and the supernatant was filtered at 500 × g for 5 min through a 0.45 μm 96-well filter plate (Millipore, MSHVN4550) to remove protein aggregates. The flow-through was mixed 1:1 with 2× sample buffer (180 mM Tris (pH 6.8), 4% SDS, 20% glycerol, 0.1 g bromophenol blue) and kept at −20 °C until analysis. Mass spectrometry-based proteomics (using tandem mass tags) was then used to quantify the amount of protein in each condition, as previously described 63 . Data were analysed by a method to detect dose-dependent changes in protein thermal stability 27 .

Western blotting to determine presence of vc1313
An overnight culture of V. cholerae sfGFP-vc1313 was diluted 1:100 into fresh LB minimal medium and incubated at 37 °C with shaking. OD 600 was measured and samples were taken as a function of time. Samples were then normalized to total protein amount and analysed by SDS-PAGE. Western blotting was performed using specific antibodies against GFP-tag (1:500, Thermo Fisher, A-11122). GFP signal was detected using Amersham Imager 600 (GE) and analysed using Fiji/ ImageJ (v.1.53f51) (ref. 58).
To follow the detailed localization of sfGFP-MCP DRK during growth, cells were added into 50 ml of TB medium and growth was followed by measuring optical density. For each datapoint, samples were analysed by fluorescence microscopy and the percentage of cells with sfGFP-MCP DRK foci was counted. For the carbon catabolite repression assays, cells were grown in MSR6 medium supplemented with 0.04% w/v d-glucose and 0.4% w/v succinate, and a similar procedure was followed.

Demographic analysis of microscopy data
Demographic analysis was performed in several steps. Fluorescence intensity profiles of sfGFP-VC1313 cells were first measured using Fiji/ ImageJ (v.1.53f51) (ref. 58). The generated data were then processed using the cellProfiles package (v.3.0.1; developed by Todd Cameron) in RStudio (v.1.4.1106; www.rstudio.com) to sort cells by length and normalize the generated intensity profiles as an average of each cell's fluorescence. Finally, the ggplot2 package was used to plot the demographics (v.3.3.5) (ref. 65).
X-ray diffraction data of the different crystals were collected on the beamlines ID30B and ID23-2 at the European Synchrotron Radiation Facility in Grenoble, France 66 . The data were processed using XDS 67 . The d-lysine and d-arginine crystals both belonged to the P1 space group and contained 2 molecules in the asymmetric unit. The crystallographic phase problem of the MCP DRK -LBD bound to d-arginine was solved by Arcimboldo 68 using 8 helices of 14 residues as a search model. The structures of MCP DRK -LBD bound to d-lysine were solved by PHASER 69 using the MCP DRK -LBD:d-Arg structure as a molecular replacement model. Coot (0.9.5) (ref. 70) was used to build the models and the structures were refined using Refmac5 (v.5.8.0267) (ref. 71) and PHENIX refine (v.1.13) (ref. 72). For complete data collection and refinement statistics, see Supplementary Table 3. The final models of the MCP DRK -LBD structures were validated using MolProbity 73 . Atomic coordinates and structure factors of the MCP DRK -LBD crystal structures (in complex with d-Arg or d-Lys) have been deposited in the Protein Data Bank (PDB codes 8BSA and 8BSB, respectively).

Phylogenetic analysis
Nucleotide and amino acid sequences were obtained from the National Center for Biotechnology Information (NCBI) 74 and the Biocyc database 75 . MCP DRK homologues were identified using the BlastP algorithm 76 against a non-redundant protein database, with an E-value threshold of 10 −10 and with V. cholerae VC1313 as reference sequence. The MUSCLE algorithm 77 was used to generate the multiple sequence alignment. The resulting alignment was then filtered by removing 90% of redundancy, and unique representative species were selected for further analysis. The Jalview (v.2.11.2.0) toolkit 78 was used to visualize all the alignments and analyse the conservancy of residues of interest. The phylogenetic tree was generated using ClustalW2 (ref. 79), and Interactive Tree of Life (iTOL, v.5) (ref. 80) was then used to further customize the phylogenetic tree and add additional information. The genetic context analysis was carried out by comparing upstream/downstream regions of the selected representatives.

Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability
The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD038312. Atomic coordinates and structure factors of the MCP DRK -LBD crystal structures (in complex with d-Arg (PDB: 8BSA) and d-Lys (PDB:8BSB)) have been deposited in the Protein Data Bank. Source data are provided with this paper. Article https://doi.org/10.1038/s41564-023-01419-6 Extended Data Fig. 2 | Comparison of chemotactic ability of V. cholerae wild-type and ΔbsrV strains. a, Chemotactic ability of ΔbsrV mutant. Both chemotactic bias and swimming speed were analysed in presence of several known chemoeffectors to compare the behaviour of wild-type (wt) and ΔbsrV mutant cells. Chemotaxis was measured in microfluidic devices, monitoring cell motion from a reservoir filled with cells in motility buffer to a reservoir filled only with 100 μM of indicated compound in motility buffer. Blank, no attractant in second reservoir; L-Arg, L-arginine; AiBu, α-aminoisobutiric acid; GABA, γ-aminobutyric acid; Succ, succinate. Each data point represents the average over the 3 time points for a given experiment (n = 4) and the line represents the mean. Significant differences (unpaired t-test) are indicated by *** (p < 0.001) and was evaluated relative to the blank for the given strains. b, Swimming speed analysis of V. cholerae grown in presence of D-Arg. Cells were grown in TB (± 500 μM D-Arg) and transferred to fresh TB for swimming speed measurements. Each point represents the mean and error bars indicate the SD over 3 biological replicates. c, Motility analysis of non-chemotactic V. cholerae cells in presence and absence of D-Arg. Relative motility on soft-agar plates supplemented or not with 5 mM D-Arg of several strains compared to wild-type (wt). ΔbsrV and ΔcheY3 strains are included as controls. Error bars in c represent mean values ± SD of 3 biologically independent replicates. Significant differences (unpaired t-test, adjusted for multiple comparison using Bonferroni-Dunn method) are indicated by *** (p < 0.001). See source data.

Nature Microbiology
Article https://doi.org/10.1038/s41564-023-01419-6 Extended Data Fig. 6 | MCP DRK forms part of the chemotaxis F6 system. a, Chemotaxis response to D-Arg of chemotaxis system mutants. ΔI/F9 and ΔIII/F7 showed similar response to D-Arg as control strain. No response was observed for ΔII/F6 system mutant. ΔbsrV strain was used as background. Black diamond represents the mean of 3 independent biological replicates. Significant differences (unpaired t-test) are indicated by *** (p < 0.001), n = 3. b, Co-localization of sfGFP-MCP DRK and CheY3-dTtomato (chemotaxis system II/ F6) fluorescence reporters during stationary phase. c, Localization of sfGFP-MCP DRK in the three chemotaxis systems knockout backgrounds. Representative micrographs of all experiments are shown. Representative micrographs of 3 independent replicates are shown in b and c, at least 3 images were acquired per condition: scale bar, 2 μm.