The ChvG–ChvI and NtrY–NtrX two-component systems coordinately regulate growth of Caulobacter crescentus

Two-component signaling systems (TCSs) are comprised of a sensory histidine kinase and a response regulator protein. In response to environmental changes, sensor kinases directly phosphorylate their cognate response regulator to affect gene expression. Bacteria typically express multiple TCSs that are insulated from one another and regulate distinct physiological processes. There are certainly examples of cross-regulation between TCSs, but this phenomenon remains relatively unexplored. We have identified regulatory links between the ChvG–ChvI (ChvGI) and NtrY–NtrX (NtrYX) TCSs, which control important and often overlapping processes in α-proteobacteria, including maintenance of the cell envelope. Deletion of chvG and chvI in Caulobacter crescentus limited growth in defined medium and a selection for genetic suppressors of this growth phenotype uncovered interactions among chvGI, ntrYX, and ntrZ, which encodes a previously uncharacterized periplasmic protein. Significant overlap in the experimentally-defined ChvI and NtrX transcriptional regulons provided support for the observed genetic connections between ntrYX and chvGI. Moreover, we present evidence that the growth defect of strains lacking chvGI is influenced by the phosphorylation state of NtrX and, to some extent, by levels of the TonB-dependent receptor ChvT. Measurements of NtrX phosphorylation in vivo indicated that NtrZ is an upstream regulator of NtrY, and that NtrY primarily functions as an NtrX phosphatase. We propose a model in which NtrZ functions in the periplasm to inhibit NtrY phosphatase activity; regulation of phosphorylated NtrX levels by NtrZ and NtrY provides a mechanism to modulate and balance expression of the NtrX and ChvI regulons under different growth conditions. Importance Two-component signaling systems (TCSs) enable bacteria to regulate gene expression in response to physiochemical changes in their environment. The ChvGI and NtrYX TCSs regulate diverse pathways associated with pathogenesis, growth, and cell envelope function in many α-proteobacteria. We used Caulobacter crescentus as a model to investigate regulatory connections between ChvGI and NtrYX. Our work defined the ChvI transcriptional regulon in C. crescentus and revealed a genetic interaction between ChvGI and NtrYX, whereby modulation of NtrYX signaling affects the survival of cells lacking ChvGI. In addition, we identified NtrZ as a periplasmic inhibitor of NtrY phosphatase activity in vivo. Our work establishes C. crescentus as an excellent model to investigate multi-level regulatory connections between ChvGI and NtrYX in α-proteobacteria.


Introduction
Bacteria employ two-component signaling systems (TCSs) to respond to environmental cues and maintain cellular homeostasis (1). TCS sensory modules consist of two core components: a sensory histidine kinase (HK) and a response regulator (RR). In response to signal(s), the HK undergoes autophosphorylation on a conserved histidine residue, and then passes the phosphoryl group to a conserved aspartate on the receiver (REC) domain of the RR (1). HKs may also act as phosphatases, dephosphorylating phosphorylated RRs (1). RR phosphorylation generally alters the activity of effector domains that change gene expression. TCSs are modular, and the output of a particular RR may vary between different organisms (2)(3)(4)(5). TCSs that regulate host interactions in pathogens and symbionts are often conserved in related free-living organisms and enable responses to similar physiochemical cues present in the environment (2,(6)(7)(8).
C. crescentus, a free-living a-proteobacterium found in freshwater and soil environments, encodes both the ChvGI and NtrYX systems (45,46). C. crescentus ChvGI activates transcription of the small regulatory RNA chvR, which post-transcriptionally represses the TonB-dependent receptor gene chvT (6). Examination of reporters of chvR transcription indicated that ChvGI is activated by growth in defined medium, acidic pH, DNA damage, growth at stationary phase, and cell envelope stress (6,47). However, aside from chvR, genes regulated by ChvGI have not been defined. C. crescentus NtrYX is less well-characterized than ChvGI, but a recent study established that NtrX is phosphorylated in stationary phase in defined medium, as a result of acidification (48). In addition, NtrX appears to play a core role in regulating C. crescentus physiology, as ntrX is essential for growth in complex medium and ∆ntrX cells grow more slowly than wild-type (WT) cells in defined medium (49).
In this study, we initially took a reverse genetic approach to characterize the role of ChvGI in regulating C. crescentus physiology. Deletion of chvG and chvI caused a distinctive growth defect in defined medium. By exploiting this defect, we identified striking genetic interactions between chvGI and ntrY, ntrX, and ntrZ (a previously uncharacterized gene). Epistasis analysis provided evidence that unphosphorylated NtrX is detrimental to cells lacking chvG or chvI. We defined the ChvI transcriptional regulon and discovered that it overlaps significantly with genes regulated by NtrX. In addition, we found that NtrZ promotes NtrX phosphorylation in vivo, likely by inhibiting NtrY phosphatase activity. We conclude that ChvGI and NtrYX interact at multiple transcriptional levels, working both in concert and in opposition to regulate C. crescentus growth in defined medium.

Loss of the ChvGI system limits growth in defined medium
To investigate the physiological role of ChvGI in C. crescentus, we generated strains with inframe deletions of chvG or chvI and examined their growth. The deletion strains grew normally in complex medium (peptone-yeast extract, PYE), but they displayed a distinctive growth defect in defined medium (M2 minimal salts medium with xylose as carbon source, M2X) (  (A) Growth curves, measured by optical density (OD660), of WT, ∆chvG, and ∆chvI strains bearing empty vector (EV) or genetic rescue plasmid (++) integrated at the xylose locus. Primary M2X cultures, inoculated from PYE plates, all grow to high density (left). However, upon back-dilution to OD660 = 0.025, ∆chvG and ∆chvI EV strains saturate at a significantly lower OD (right). Points represent averages of three biological replicates ± SD. **** = p < 0.0001, one-way ANOVA followed by Dunnett's post-test comparison to WT EV at 24 h.
(B) Growth curves, measured by CFU, corresponding to the cultures in A. Points represent averages of three biological replicates ± SD. *** = p < 0.0005, one-way ANOVA followed by Dunnett's post-test comparison to WT EV at 24 h.
Phase-contrast micrographs of primary overnight cultures grown in M2X medium. On the right, ∆chvI and ∆chvG strains carry empty vector (EV) or overexpression plasmids (++). Scale bars, 5 µm.  counts (Fig. S2C). Together, our results suggest that the diminished growth capacity of ∆chvI or ∆chvG cells in M2X medium is a function of the time since resuspension from PYE plates and growth phase.

ChvI phosphorylation is critical for growth in defined medium
Given the similarity between the phenotypes displayed by ∆chvI and ∆chvG strains, we predicted that phosphorylation of ChvI might be important for growth in M2X medium. To test this hypothesis, we generated strains harboring chvI alleles encoding changes to the conserved sites of phosphorylation (D52A, D52N, and D52E). Growth of strains carrying the non-phosphorylatable alleles chvI(D52A) and chvI(D52N) was similar to ∆chvI cells in M2X ( Fig. 2A and Fig. S3A). By contrast, chvI(D52E) cultures reached higher densities than ∆chvI cultures, suggesting that ChvI(D52E) is active, albeit not to the same level as phosphorylated ChvI ( Fig. 2A and Fig. S3A).
Substitutions of the phosphorylatable aspartate with glutamate often act as phosphomimetic mutations and constitutively activate RRs (50)(51)(52)(53). Thus, this result supports a critical role for ChvI phosphorylation during growth in M2X medium.
To further examine the importance of ChvI phosphorylation, we overexpressed chvG and chvI alleles from a xylose-inducible promoter in knockout backgrounds. Although overexpression of chvG rescued growth of ∆chvG cells (Fig 1B), overexpression of chvG(H309A), a catalytichistidine mutant, did not, indicating that phosphorylation of ChvG, and by extension ChvI, is important for growth in M2X medium (Fig. 2B). Overexpression of chvI(D52E) only partially rescued growth of ∆chvI and ∆chvG strains in M2X and these strains grew similarly to the strain encoding chvI(D52E) at the native locus ( Fig. 2A, 2B and Fig. S3B). However, in contrast to native expression of chvI(D52E), overexpression of this allele resulted in cell filamentation in M2X, but not PYE, consistent with a cell division defect ( Fig. 2C and Fig. S3C). Thus, overexpression of phosphomimetic chvI(D52E) appears to interfere with cell division and, perhaps as a result, only partially rescues growth in M2X medium.

Mutations in ntrY, ntrX, and a gene of unknown function, rescue the growth defect of ∆chvI and ∆chvG strains
To better understand why ChvGI is critical for growth in defined medium, we employed a selection strategy to isolate second-site mutations that alleviate the growth defect of ∆chvG and ∆chvI cells. ∆chvG and ∆chvI M2X-overnight cultures were back-diluted and grown until they reached high density, presumably due to proliferation of suppressor strains (Fig. 3A). We then isolated single colonies with WT-like growth in M2X and used whole-genome sequencing to identify any acquired mutations (Fig. 3B).
In multiple independent strains, we identified nonsynonymous substitutions in the genes encoding the HK NtrY and its cognate RR NtrX (Fig. 3C). The ntrY mutations (L70H and A123V) are located in predicted transmembrane helices, which are involved in transmitting information from periplasmic sensing domains to the kinase domain ( Fig. 3C) (54)(55)(56). We identified only one ntrX allele, A19V, which lies in helix 1 (⍺1) of the receiver domain (Fig. 3C). ⍺1 is involved in the interaction interface between the HK dimerization and histidine phosphotransfer (DHp) domain and the RR REC domain (57,58). We also note that the ntrX(A19V) strain was isolated on PYE plates and appeared to grow normally in M2X, suggesting that this is not a complete loss-offunction allele (49).
In addition to mutations in ntrY and ntrX, we identified three mutants with nonsynonymous substitutions in a gene encoding a putative periplasmic protein of unknown function (CCNA_03863, hereafter referred to as ntrZ) (Fig. 3C). The mutations identified (Y92C and I99N) are both located outside of the predicted signal sequence (Fig. 3C). Therefore, mutations in the NtrYX TCS and NtrZ appear to affect the growth of ∆chvI and ∆chvG cells in M2X medium.

Overexpression or deletion of ntrX modulates the ∆chvI growth phenotype
Several studies have noted similar phenotypes of chvGI and ntrYX mutants (25,26,41); however, to our knowledge none have established a genetic link between these TCSs. To better understand the connections between ChvGI and NtrYX in C. crescentus, we first characterized the genetic relationship between chvI and ntrX. Replacement of ntrX with the ntrX(A19V) allele restored growth of ∆chvI cells in M2X (Fig. 4A), ruling out other background mutations as causal for suppression of the ∆chvI phenotype. However, overexpression of ntrX(A19V) in the presence of the native ntrX allele did not rescue growth of ∆chvI cells (Fig. 4A). This result indicates that the ntrX(A19V) allele is recessive, and suggests that the presence of WT NtrX contributes to the growth phenotype of ∆chvI cells.
To examine this notion further, we tested the effect of deleting ntrX in the ∆chvI strain. ∆ntrX cells had a slight growth defect in M2X, but were clearly distinct from ∆chvI cells (Fig. 4B). Deletion of ntrX in a ∆chvI background rescued growth, indicating that the presence of ntrX is indeed detrimental to ∆chvI cells (Fig. 4B). Suppression of the ∆chvI growth phenotype was not due to a Growth curves, measured by optical density (OD660), of WT, ∆chvI, and suppressor strains isolated from either the ∆chvI strain (∆IS#) or the ∆chvG strain (∆GS#) upon back-dilution in M2X medium. Points represent averages of three biological replicates ± SD. **** = p < 0.0001, one-way ANOVA followed by Dunnett's post-test comparison to ∆chvI at 24 h. C. Given that the presence of ntrX limits growth of ∆chvI cells in M2X, we next tested whether overexpression of ntrX affects growth of WT cells. Overexpression of ntrX only moderately slowed growth of WT cells. However, overexpression of the nonphosphorylatable ntrX(D53A) allele dramatically impaired the growth of WT cells in M2X, similar to the defect observed in ∆chvI cells (Fig. 4C). Overexpression of either ntrX or ntrX(D53A) exacerbated the growth defect of ∆chvI cells (Fig. 4C). Together these results support a model in which unphosphorylated NtrX limits growth capacity in M2X medium, especially in ∆chvI cells.

Suppressor Mutations
NtrZ is a predicted periplasmic protein that functions upstream of NtrY to regulate levels of phosphorylated NtrX We next examined the nature of the ntrY and ntrZ suppressor mutations. Overexpression of the ntrY and ntrZ alleles identified in our suppressor strains restored growth of ∆chvI cells in M2X (Fig. 5A). In addition, overexpression of WT ntrY, but not WT ntrZ, also significantly suppressed the ∆chvI growth defect (Fig. 5A). We conclude that the ntrY and ntrZ mutations are dominant, likely gain-offunction, alleles. As deletion of ntrX rescued the growth of ∆chvI cells, overexpression of ntrY or ntrZ mutant alleles likely promotes phosphorylation and/or sequestration of NtrX. We attempted, but failed, to construct ∆chvI ∆ntrY and ∆chvI ∆ntrZ double-mutants Optical density (OD660) of cultures 24 h after backdilution to OD660 = 0.025 in M2X medium (average ± SD, N = 3). Restored ∆chvI (∆chvI ∆ntrX::ntrX) was generated by knocking ntrX into the native ntrX locus in ∆chvI ∆ntrX cells. *** = p < 0.0005, **** = p < 0.0001, one-way ANOVA followed by Dunnett's post-test comparison to ∆chvI.
Phosphorylation of NtrX is likely important for stationary phase survival (48), and thus, we hypothesized that deletion of the ntrY HK, and perhaps also ntrZ, might lead to a stationary-phase defect. Indeed, we observed significantly lower CFUs in both ∆ntrY and ∆ntrZ cultures compared to WT cultures after 48 h of growth in M2X ( Fig  5C). Notably, ∆ntrY ∆ntrZ cultures behaved similarly to the single mutants, suggesting that NtrY and NtrZ indeed function in the same pathway.
To place ntrY and ntrZ relative to one another, we evaluated the ability of ntrY and ntrZ alleles to rescue stationary-phase survival in the deletion strains. Both ∆ntrY and ∆ntrZ cells were fully rescued by ectopic overexpression of each respective WT allele (Fig. 5D).
Overexpression of ntrY(L70H), but not ntrY, also fully rescued the phenotype of ∆ntrZ cells (Fig. 5D). By contrast, neither WT ntrZ nor ntrZ(I99N) rescued ∆ntrY cells (Fig.  5D). These data provide evidence that NtrZ functions upstream of NtrY, potentially as a kinase activator and/or phosphatase inhibitor. Growth curves, measured by optical density (OD660), of WT and knockout strains upon back-dilution in M2X medium. Points are averages of three biological replicates ± SD. **** = p < 0.0001, one-way ANOVA followed by Dunnett's post-test comparison to WT at 24 h. (C) Growth curves, measured by CFU, for WT and mutant strains grown in M2X medium. Points are averages of three biological replicates ± SD. *** = p = 0.0005, one-way ANOVA followed by Dunnett's post-test comparison to ∆ntrZ at 48 h.
Cell density, measured by CFU, for ∆ntrZ and ∆ntrY strains bearing empty vector (EV) or overexpression vectors (++) at 48 h growth in M2X medium. Points are averages of three biological replicates ± SD. ** = p < 0.01, one-way ANOVA followed by Dunnett's post-test comparison to ∆ntrZ EV at 48 h.
Anti-HA western blot of lysates analyzed by Phos-tag gel electrophoresis (left). Each strain encodes ntrX-HA at the native ntrX locus. Phosphorylated (NtrX~P) and unphosphorylated (NtrX) NtrX-HA are indicated. Quantification of the percent of NtrX that is phosphorylated (% NtrX~P) (Right). Points are averages of four biological replicates ± SD. * = p < 0.05, ** = p < 0.01, *** = p < 0.001, one-way ANOVA followed by Dunnett's post-test comparison to WT. We next examined the phosphorylation state of NtrX directly using Phos-tag gel electrophoresis. To detect NtrX by western blotting, we constructed strains carrying C-terminally HA-tagged ntrX (ntrX-HA) encoded at the native ntrX locus. Phos-tag analysis of WT lysates at stationary phase revealed distinct bands for phosphorylated (NtrX~P) and unphosphorylated (NtrX) NtrX-HA (Fig.  5E). Deletion of ntrZ ablated NtrX~P, consistent with a role for NtrZ in promoting NtrX phosphorylation (Fig. 5E). Surprisingly, ∆ntrY cells displayed higher levels of NtrX phosphorylation than WT cells, suggesting that NtrY primarily acts as a phosphatase in vivo and is not required for NtrX phosphorylation. NtrZ indeed acts upstream of NtrY, as deletion of ntrZ in the ∆ntrY background did not affect NtrX phosphorylation (Fig. 5E). We conclude that NtrZ promotes NtrX phosphorylation by inhibiting NtrY phosphatase activity.
Our characterization of the ntrY, ntrX, and ntrZ mutants suggested that phosphorylation of NtrX may rescue the growth of ∆chvI or ∆chvG cells. NtrX is phosphorylated under acidic conditions, such as those encountered during stationary phase in M2G or M2X medium (48). Therefore, we tested whether the pH of M2X affected the growth of ∆chvI cultures. For both WT and ∆chvI strains, primary overnight cultures diluted in M2X at pH 7.0 reached similar CFUs at 8 h as those diluted in standard M2X (pH 7.2) (Fig. 1B, S4). WT cultures were relatively unaffected by growth in M2X at pH 6.0, but had significantly fewer CFUs in M2X pH 5.5 vs. pH 7.0 (Fig. S4C). By contrast, ∆chvI cells were markedly less fit in M2X pH 6.0 than at pH 7.0, and displayed an intermediate growth yield at pH 5.5 relative to M2X at either pH 6.0 or pH 7.0 (Fig. S4C). These results are consistent both with suppression of the ∆chvI growth defect by NtrX phosphorylation and an important role for ChvGI in acid-stress responses (6).

ChvI and NtrX regulate transcription of an overlapping set of genes
Although ChvGI is known to regulate the expression of chvR, the complete ChvI transcriptional regulon is not known. To examine the regulatory link between ntrX and chvI in greater detail, we performed an RNA-seq experiment to comprehensively define ChvI-dependent gene regulation.

chvI(D52E)++/EV
We next compared our ChvI regulon with a previously published NtrX regulon that was determined using DNA microarrays (49). This experiment measured relative gene expression between C. crescentus WT (strain NA1000) and ∆ntrX cells during exponential growth in M2G medium, a condition where NtrX is expected to be largely unphosphorylated (48,49). Surprisingly, a large fraction of the genes regulated by ChvI were also represented in the genes regulated by NtrX (Fig. 6A). That is, many of the genes upregulated by ChvI also appeared to be upregulated by NtrX (and likewise for downregulated genes). Using cutoffs of 1.5 fold-change for the ChvI RNA-seq data and 2.5 fold-change for the NtrX microarray data, we established that the ChvI regulon is significantly enriched for NtrX-dependent genes (6.31-fold enrichment, p-value = 8.99 × 10 -19 , hypergeometric test). To confirm this overlap, we evaluated the effect of deleting ntrX on our ChvI regulon reporters. Strains carrying the transcriptional reporters were grown to log phase (OD660 ~ 0.1-0.2) in M2X medium before assaying β-galactosidase activity. Four of the six reporters exhibited ntrX dependence, including two genes (CCNA_00889 and chvR) not evaluated in the NtrX microarray experiment (Fig. 6C). We note that transcription from the chvI promoter (PchvI) was activated by both ntrX and chvI, raising the possibility that NtrX may indirectly affect ChvI-dependent genes via upregulation of chvIG-hprK.
Given the oppositional nature of ChvI and NtrX during growth in M2X medium, we were surprised to see such a high degree of similarity in the genes they regulate. However, a small subset of genes exhibited opposite regulation by ChvI and NtrX, which might therefore account for suppression of the ∆chvI growth defect by deletion of ntrX (Fig. S6A). We overexpressed (for those genes upregulated by ChvI) or knocked out (for those genes downregulated by ChvI) each of these genes in ∆chvI cells and tested their growth capacity in M2X medium. Only one candidate, chvT, had any effect on ∆chvI cells. Specifically, deletion of chvT partially rescued growth of ∆chvI cells in M2X ( Fig. 7A and Fig. S6B). Differential regulation of chvT expression by ChvI and NtrX may therefore contribute to the growth defect observed in M2X medium.

Discussion
The importance of the C. crescentus ChvGI system for growth in defined medium To our knowledge, the only reported physiological phenotype for chvGI mutants in C. crescentus is sensitivity to the antibiotic vancomycin (47). However, previous work indicated that ChvGI might be important for growth in defined medium (6). In fact, chvGI mutants in other ⍺-proteobacteria are sensitive to nutritional conditions, although most, with the exception of Brucella abortus, exhibit particularly poor growth in complex media (4,8,27,59). Deletion of chvG or chvI in C. crescentus caused a distinctive growth defect in M2X medium. ∆chvI cultures do grow in M2X upon inoculation from PYE plates, suggesting that the physiological state of the cell in PYE agar is initially amenable to growth in defined medium (Fig. 1). However, this tolerance is limited by time, with only higher inocula able to reach high cell density ( Fig. 1 and Fig. S2). Why then, can primary overnight cultures persist at high density until back-dilution in fresh M2X? One possibility is that the low pH of M2X at stationary phase preserves ∆chvI and ∆chvG cells, potentially by triggering phosphorylation of NtrX (Fig. 7B) (48). This model is supported by the observation that ∆chvI cells reach higher CFUs when back-diluted in M2X at pH 5.5 vs. pH 6.0 (Fig. S4C). As ~20% of the NtrX pool is phosphorylated in M2G medium at pH 6.0 vs. ~40% at pH 5.5 (by Phos-tag analysis), the pH 6-5.5 transition would significantly change the level of  unphosphorylated NtrX (48). However, as ∆chvI and ∆chvG cells also appear to be sensitive to acidic pH, additional factors are likely at play (Fig. S4C). For example, several extracytoplasmic function (ECF) sigma factors are involved in resistance to stationary-phase stress and might play a protective role in ∆chvI and ∆chvG cells (60)(61)(62).
A recent study failed to identify chvG or chvI as being important for fitness in M2X medium (63). However, in this work M2X cultures were inoculated from PYE starter cultures at high enough density to reach saturation in 5 doublings. Thus, these experiments likely mimicked our primary overnight cultures, obscuring detection of fitness defects for chvGI mutants.

The ChvI transcriptional regulon in C. crescentus
Although ExoR or ChvI transcriptional regulons have been defined in several ⍺-proteobacteria (4,5,26,64), the C. crescentus ChvI regulon was unknown prior to this work. We employed RNAseq to detect direct and indirect transcriptional targets of ChvI in C. crescentus (Fig. 6, Table S1). Our ChvI regulon contained several classes of genes noted in other ⍺-proteobacteria, including those encoding outer membrane proteins and transporters, metabolic enzymes, lipopolysaccharide biosynthesis enzymes, and stress response proteins (4,5,26,64). We note, in particular, that the C. crescentus ChvI regulon contained a large number of genes encoding proteins involved in envelope maintenance, including nearly the entire BAM complex, the envelope integrity protein EipA, members of the Tol-Pal complex, and a variety of chaperones and proteases ( Fig. 6 and Table S1) (65)(66)(67)(68). The idea that ChvGI is involved in envelope integrity is consistent with previous observations that ChvGI is activated by envelope stress and confers resistance to antibiotics targeting the cell wall (47). However, ~40% of genes regulated by ChvI are annotated generically or as hypotheticals, and thus, more work will be required to characterize the pathways downstream of the ChvGI system. In A. tumefaciens, S. meliloti, and C. crescentus, ChvGI appears to suppress cell motility and/or chemotaxis (5,15,20,26,69,70). However, unlike in A. tumefaciens and S. meliloti, C. crescentus ChvI does not regulate any obvious flagellar or chemotaxis genes, suggesting that effects on motility may be due to post-transcriptional regulation and/or alterations in cell-cycle progression (5,26,(71)(72)(73).
Overexpression of the phosphomimetic chvI(D52E) allele induced cell filamentation in M2X, implicating ChvI in regulating cell division, and cytokinesis in particular. ChvI upregulates several genes related to cell division, including zauP, members of the Tol-Pal complex, smc, and nstA (68,(74)(75)(76)(77). We note that overexpression of a proteolytically stable mutant form of nstA induces cell filamentation and both zauP and the Tol-Pal complex are involved in regulating cytokinesis (74,77). Perhaps over-induction of these regulon genes, in combination with the cellular state in M2X medium, interferes with proper cell division.

Genetic interactions between ChvGI and NtrYX
The severe growth defect of ∆chvG and ∆chvI cells in M2X medium allowed us to uncover the first known genetic interaction between chvGI and ntrYX (Fig. 3). Importantly, deletion of ntrX suppressed the growth defect of ∆chvI cells, whereas overexpression of non-phosphorylatable ntrX(D53A) was deleterious for growth in both ∆chvI and WT cells (Fig. 4). These results suggest that the activity of unphosphorylated NtrX is particularly detrimental in cells lacking chvI. Although unphosphorylated response regulators are often assumed to be inactive, multiple RRs are known to affect transcription in their unphosphorylated states (78)(79)(80)(81)(82). Thus, we propose that phosphorylated ChvI and unphosphorylated NtrX oppose each other to regulate growth in defined medium (Fig. 7B). We note that a past study examining connections between ChvI and NtrX in S. meliloti did not test whether perturbations in NtrYX signaling might affect ∆chvI phenotypes (25). We predict that a similar ChvI-NtrX relationship may be conserved in other ⍺-proteobacteria.
As gain-of-function mutations in ntrY and ntrZ suppress the growth defect of ∆chvI cells, we also propose that phosphorylation of NtrX relieves its detrimental activity, possibly by changing its transcriptional regulon. Although the global transcriptional effects of NtrX phosphorylation have yet to be characterized, in vitro phosphorylation of B. abortus NtrX does induce conformational changes and alters, but does not weaken, binding to the ntrYX promoter (83). Interestingly, we were unable to construct ∆chvI ∆ntrY and ∆chvI ∆ntrZ strains, suggesting that these gene deletion combinations are synthetically lethal in PYE. As ntrX is also essential in PYE (20), the balance between unphosphorylated and phosphorylated NtrX may be important for growth in both complex and defined media.
To identify downstream genes that mediate the oppositional relationship between ChvI and NtrX, we compared their transcriptional regulons ( Fig. 6 and Table S1). The ChvI regulon strongly overlapped with that of NtrX, as 80% of the top 30 ChvI-activated genes are also activated by NtrX (per our >1.5-fold cutoff). By contrast, only 20% of the top 30 NtrX-activated genes exhibit >1.5-fold ChvI-dependence (49). Given that chvI, chvG, and hprK are upregulated by NtrX, it may be the case that NtrX simply activates expression of the chvIG-hprK operon, thereby altering transcription of genes in the ChvI regulon (Fig. 7B, grey arrow). However, further work is required to define the mechanism by which NtrX affects the ChvI regulon, as it may also indirectly affect ChvGI signaling or directly regulate expression of regulon genes. The overlap we observed between the ChvI and NtrX regulons may be restricted to C. crescentus and close relatives, as chvG and chvI are not transcriptionally regulated by NtrYX in Rhodobacter sphaeroides (40). However, ChvGI and NtrYX may still be transcriptionally linked in more distantly-related ⍺proteobacterial species, as ntrX is part of the A. tumefaciens ExoR regulon (26).
The majority of overlapping genes in the ChvI and NtrX regulons cannot account for the detrimental effect of unphosphorylated NtrX on ∆chvI cells. Therefore, we focused on eight genes that exhibited opposing transcriptional regulation by ChvI and NtrX (Fig. S6). Given that NtrX largely reinforces the ChvGI TCS, these oppositional effects may reflect direct transcriptional regulation by NtrX or the effects of unique NtrX regulon genes. Only deletion of chvT improved the growth of ∆chvI cells, indicating that suppression of chvT RNA levels by ChvI may be important for growth in M2X medium (Fig. 7). chvT is linked to diverse phenotypes in C. crescentus, including survival in stationary phase, sensitivity to cell wall-targeting antibiotics, and sensitivity to bacteriocins (47,84,85). However, the molecule(s) transported by ChvT remain undefined. High ChvT levels may contribute to defects in transport in defined medium and/or alter membrane integrity, impacting the viability of cells lacking ChvGI (47). It is clear, though, that altered chvT expression cannot fully explain the growth deficiency of ∆chvI cells in M2X medium. This growth defect may result from the collective action of several genes, and therefore, manipulation of individual candidates may not rescue growth in M2X. Moreover, 15 genes in the ChvI regulon are absent from the NtrX microarray dataset, raising the possibility that one or more may be differentially regulated by ChvI and NtrX (Fig. 6, Table S1). ChvI and NtrX might also interact more indirectly, as each regulates unique subsets of genes that may affect growth in M2X medium.

On the role of NtrZ
NtrYX is associated with a wide range of physiological responses, from nitrogen metabolism to redox sensing and cell envelope maintenance (25,35,39,40). Despite these phenotypic observations, little is known about NtrY activity in vivo and the regulation of NtrY via its periplasmic domain. Our work reveals a surprising phosphatase pathway involving the previously uncharacterized protein NtrZ. Phos-tag analysis of NtrX phosphorylation in vivo clearly demonstrates that NtrY is dispensable for NtrX phosphorylation, and suggests that NtrY primarily acts as an NtrX phosphatase (Fig. 5). In addition, NtrZ appears to inhibit NtrY phosphatase activity, as deletion of ntrZ abolishes NtrX phosphorylation only when ntrY is present (Fig. 5). Thus, we propose that NtrZ inhibits NtrY phosphatase activity, stabilizing the pool of phosphorylated NtrX (Fig. 7). In the future, we are interested in determining whether NtrZ physically interacts with the NtrY periplasmic domain or affects NtrY activity indirectly. Several known periplasmic or membrane-bound TCS regulators directly interact with HK periplasmic domains (33,(86)(87)(88). Although our model does not exclude the possibility that NtrY phosphorylates NtrX under certain conditions, there is clearly another source of NtrX phosphorylation in C. crescentus. An additional HK may phosphorylate NtrX, although the most likely candidate, NtrB, does not phosphorylate NtrX in vitro (49). Alternatively, some metabolic intermediates, such as aspartyl-phosphate or carbamoyl-phosphate, can serve as phosphodonors for RRs in vivo and in vitro (1,(89)(90)(91)(92).
Interestingly, we initially placed ntrZ upstream of ntrY by examining the similar stationary-phase survival phenotypes of ∆ntrY and ∆ntrZ cells (Fig. 5). Given our Phos-tag results, these phenotypes suggest that both phosphorylated and unphosphorylated NtrX play important roles in stationaryphase survival. This idea is consistent with the fact that NtrX phosphorylation increases upon entry into stationary-phase, but eventually decreases (48).
Fernández et al. reported that NtrX phosphorylation is triggered by acidic pH in defined medium (48). Therefore, we hypothesize that inhibition of NtrY by NtrZ may be enhanced by acidic pH. Thus, NtrZ and ExoR potentially share a regulatory theme, in which their activities towards their respective HKs are affected by low pH (16,33,34). However, it remains to be seen if, as with ExoR, proteolysis plays a role in controlling NtrZ-dependent regulation of NtrY (16). Given the results of our suppressor selection, it would appear that mutations in NtrZ or in the transmembrane helices of NtrY may bypass regulation by acidic pH. In an alignment of 100 NtrZ homologs, ranging from 41% to 100% sequence identity, the Y92 position is conserved as an aromatic residue (95% Y) while I99 is conserved as an aliphatic hydrophobic residue (51% I, common L and V substitutions), suggesting that these residues are important for NtrZ function. Further work is required to determine how mutation of Y92 or I99 may alter interaction with the NtrY periplasmic domain or affect a different aspect of NtrZ function. An alignment of the non-cytoplasmic portion of 250 NtrY sequences, ranging from 37% to 100% identity, reveals that the L70 position is largely conserved as an aliphatic hydrophobic residue (94% L) whereas A123 is conserved as a hydrophobic residue (93% A), albeit with occasional bulky-aromatic substitutions. It is unsurprising, then, that the L70H and A123V substitutions appear to affect NtrY function. In fact, several studies have identified transmembrane mutations that alter HK kinase and phosphatase activities (93)(94)(95).
Unlike the periplasmic kinase regulator ExoR, which contains Sel1-like repeats typically involved in protein-protein interactions, NtrZ does not contain any conserved domains or motifs (95). Moreover, NtrZ appears restricted to the order Caulobacterales, albeit with a few distant homologs in other ⍺-proteobacteria. Notably, neither B. abortus, S. meliloti, nor A. tumefaciens contain NtrZ homologs, and conservation of the non-cytoplasmic region of NtrY is quite low between C. crescentus and these organisms (24-27% identity). However, L70 is largely conserved (L or V) and A123 is entirely conserved between these NtrY homologs. Thus, the L70H and A123V substitutions may have similar effects on NtrY activity in these organisms, and further analyses of these mutants may shed light on a conserved NtrY activity switch. Moreover, it will be interesting to see if NtrY also primarily acts as a phosphatase and/or is regulated by periplasmic effectors in S. meliloti, A. tumefaciens, and B. abortus.

Strains and Plasmids
All plasmids were constructed using standard molecular biology techniques. See Table S2 for strain, plasmid, and primer information. Plasmids for generating in-frame deletions and allelereplacements were generated by cloning homologous upstream and downstream regions into pNPTS138. Transcriptional reporter plasmids were generated by cloning 400-500 bp upstream of the open reading frame into pRKlac290. For overexpression strains, ORFs were inserted into pMT585, a plasmid for xylose-inducible expression that integrates at the xylX locus. Plasmids were transformed into C. crescentus CB15 strain by electroporation or tri-parental mating. In-frame deletion and allele-replacement strains were generated by a double recombination strategy involving sacB counterselection on PYE plates supplemented with 3% sucrose (97). For ∆ntrX knockout strains, counterselection was carried out on M2X + 0.5% sucrose plates due to their growth defect on PYE. ∆ntrX and ∆chvI ∆ntrX strains were grown only on M2 medium. Construction of ∆chvI ∆ntrY and ∆chvI ∆ntrZ mutants was attempted using both PYE and M2X counterselection methods. All C. crescentus strains were grown at 30°C. For strains carrying pRKlac290 plasmids, oxytetracycline was added to 1 µg/mL in liquid and 2 µg/mL in PYE agar or 1 µg/mL in M2X agar.

Measurement of Growth in M2X Medium
Primary M2X cultures were inoculated from plates to an approximate density of OD660 0.02-0.10 and grown overnight. Overnight cultures were back-diluted to OD660 = 0.025 and OD660 was recorded at the indicated times. To enumerate colony forming units, samples were taken at the indicated time points and 10-fold serial dilutions were plated on PYE agar. For pH experiments, M2X medium was adjusted to the indicated pH using HCl.
For experiments with controlled starting densities, cells were resuspended in M2X medium directly from PYE plates. These resuspensions were then diluted to the indicated OD660 and the cultures were titered for CFUs. In washing experiments, 1 mL resuspended cells were spun down at 8k × g for 3 min and resuspended in 1 mL fresh M2X medium once (1X) or twice (2X) before dilution. Plotting and statistical analyses were carried out using Prism (Graphpad).

Microscopy
Samples of ∆chvG and ∆chvI cells were taken from overnight M2X or PYE + 0.15% xylose cultures and imaged with a DMI6000B (Leica) microscope in phase contrast using a HC PL APO 63x/1.4na oil Ph3 CS2 objective. Images were captured using an Orca-R 2 C10600 digital camera (Hamamatsu) controlled by Leica Application Suite X (Leica). Images were processed using Fiji (98,99).

b-galactosidase Assays
For assays of WT vs. ∆chvI strains, cells were grown overnight in 2 mL PYE medium. Then, 500 µL of each culture was centrifuged at 8k × g for 3 min and resuspended in 500 µL M2X medium. Resuspended cultures were used to inoculate 2 mL M2X medium at a starting OD660 of 0.075. Cultures were grown for 4 h before assaying b-galactosidase activity. For assays of WT vs. ∆ntrX strains, cells were grown overnight in 2 mL M2X medium. Overnight cultures were diluted in M2X medium such that they would reach OD660 = 0.1-0.2 after 23.5 h of growth. b-galactosidase assays were performed as previously described using 200 µL of M2X culture + 100 µL sterile PYE medium as an emulsifier (106). Plotting and statistical analyses were carried out using Prism (Graphpad).

Data Availability
RNA-seq data are available in the NCBI's Gene Expression Omnibus (GEO) database (107)