A protease and a lipoprotein jointly modulate the conserved ExoR-ExoS-ChvI signaling pathway critical in Sinorhizobium meliloti for symbiosis with legume hosts

In a previous suppressor analysis to identify mutations that alleviated the cell envelope defects associated with the podJ1 deletion, we found two genes (SMc00067 and SMc03872) whose interruption or deletion led to consistent and significant reduction in EPS-I production [44]. SMc00067 (annotated as lppA) encodes a 148-amino acid lipoprotein, while SMc03872 (here named jspA, for podJ suppressor protease A) encodes a 497-amino acid metalloprotease that contains an M48 peptidase domain, with a conserved HEXXH active site, and a LysM domain, commonly associated with peptidoglycan binding (Fig 2) [63,78,79]. BLAST searches against representative bacterial species indicated that both genes are highly conserved within the Rhizobiales group of alpha-proteobacteria, based on shared synteny and protein sequences (Fig 2A and S1 Table) [80]. Outside of the Rhizobiales group, orthologs of LppA were rare or difficult to identify, while the sequence similarities of JspA homologs were generally lower than those found within Rhizobiales (S1 Table). Both LppA and JspA contain lipoprotein signal peptides at their N-termini, each with a stretch of hydrophobic amino acids followed by an invariant Cys within the lipobox motif (Fig 2B) [81]. In the original annotation for LppA in S. meliloti Rm1021, the protein starts seven codons upstream of the LAGC lipobox, with VVASGVA, but N-terminal extension of 12 codons adds more hydrophobic amino acids, allowing a more optimal signal sequence; thus, we have numbered the amino acid sequence accordingly. The lipoprotein signals suggest that each protein is exported out of the cytoplasm and attached to the inner or outer membrane [82].

Fig 2. Schematics of the genomic regions around lppA and jspA and of their protein products.

(A) S. meilioti lppA (SMc00067) and jspA (SMc03872) share synteny with their respective orthologs in closely related alpha-proteobacteria, such as B. abortus and A. tumefaciens. Gene and ORF names are shown as annotated, with pentagonal arrows indicating directionality. Arrows with the same colors in different species represent probable homologs, with red arrows indicating lppA or jspA orthologs; genes without annotated functions or obvious orthologs in corresponding regions are depicted with shades of grey. RR and HK signify response regulators and histidine kinases. The drawing is to scale; bar indicates 1 kb. (B) lppA encodes a 148-aa lipoprotein, while jspA encodes a 497-aa metalloprotease. Both LppA and JspA contain lipoprotein signal peptides at their N-termini; the sequences of these leader peptides are shown, with red arrows indicating cleavage sites before the invariant cysteine of the lipobox motifs, underlined. The N-terminus of LppA was originally annotated as the 13th amino acid (V₁₃) shown here, but extension of 12 amino acids provides a better signal sequence. JspA also contains M48 peptidase and LysM domains; key amino acids of the peptidase domain are displayed. Grey numbers indicate residues that border the predicted protein domains.

https://doi.org/10.1371/journal.pgen.1010776.g002

To verify that LppA and JspA contribute to EPS-I production, we performed complementation analysis by introducing plasmids carrying lppA or jspA under the control of a taurine-inducible promoter (P_tau) [83] into wild-type Rm1021 or deletion mutants. Serial dilutions of liquid cultures were spotted onto plates containing calcofluor, which fluoresces when bound to EPS-I [84]. Consistent with previously published results [44], ΔlppA or ΔjspA mutants carrying the empty vector exhibited lower levels of EPS-I production, with 60–70% of calcofluor fluorescence compared to wild type carrying the vector, in the presence or absence of the inducer (Fig 3A and 3B and S2 Table). Compared to their counterparts with the vector, wild-type and mutant strains with plasmids carrying lppA or jspA showed elevated fluorescence levels in the presence of taurine (Fig 3A and 3B); this increase in fluorescence was not observed in the absence of taurine (S2 Table). Similar results were obtained with the closely related species S. medicae strain WSM419: deletion of the lppA or jspA ortholog in that wild-type strain reduced fluorescence on plates containing calcofluor (S1A Fig), and complementation with the heterologous S. meliloti gene rescued the defect (S1B Fig, strains JOE5290 and JOE5264). At higher taurine concentrations (5 and 10 mM), induction of jspA expression in S. meliloti Rm1021 inhibited colony formation (Fig 3C), but we did not observe this effect with lppA expression, even with the highest possible concentration of taurine (100 mM) (Fig 3A). While induction of lppA or jspA expression promoted EPS-I production in the wild type or corresponding deletion mutants, lppA expression in the ΔjspA mutant (Fig 3A, bottom row) and jspA expression in the ΔlppA mutant (Fig 3B, bottom row) failed to increase calcofluor fluorescence (S2B Table). In addition, although overexpression of jspA at 10 mM taurine increased EPS-I production in the ΔlppA mutant, it did not cause growth arrest in that background (Fig 3C, bottom row). These results suggest that LppA and JspA require each other to stimulate EPS-I production and accomplish their physiological activities, albeit overexpression of JspA may be able to bypass some of the requirement for LppA.

Fig 3. Calcofluor fluorescence, indicating EPS-I production, of strains expressing lppA or jspA.

Ten-fold serial dilutions (10⁻² to 10⁻⁵) of logarithmic-phase cultures were spotted onto LB plates containing calcofluor and allowed to grow for three days prior to fluorescence imaging. Darker spots on representative images indicate brighter fluorescence. Fluorescence levels were measured relative to the S. meliloti Rm1021 wild-type (WT) strain carrying an empty vector on each plate. (A) WT strains or ΔlppA or ΔjspA mutants carrying the vector (pCM130 or pJC478) or a plasmid with lppA under the control of a taurine-inducible promoter (pJC532) were grown on plates containing 100 mM taurine. (B, C) WT strains or ΔjspA or ΔlppA mutants carrying the vector or a plasmid with jspA under the control of a taurine-inducible promoter (pJC535) were grown on plates containing (B) 2.5, (C) 5, or 10 mM taurine. Error bars represent standard deviations. Relative fluorescence intensities were not calculated for 5 and 10 mM taurine due to growth inhibition of select strains. *, p < 0.05; ns, not significant.

https://doi.org/10.1371/journal.pgen.1010776.g003

Since EPS-I is critical for infection thread formation during host colonization, we asked if deletion of lppA or jspA leads to a symbiosis defect. M. sativa seedlings were inoculated with wild-type Rm1021 or ΔlppA or ΔjspA derivatives, and M. truncatula seedlings were inoculated with wild-type WSM419 or its ΔlppA or ΔjspA derivatives because WSM419 forms more efficient symbiosis with M. truncatula than S. meliloti strains [85–87]. For each host species, we did not observe obvious differences in plant growth and the development of root nodules on nitrogen-free medium over the course of four weeks: the average numbers of pink, nitrogen-fixing nodules per plant were similar 21 and 28 days after inoculation with wild-type or mutant strains (S3 Table), suggesting that the two genes are not required for symbiosis. To examine more closely if LppA and JspA contribute to efficient host colonization, we conducted competitive infection assays in which seedlings were inoculated with mixtures containing equal numbers of two strains, and bacteria were recovered from root nodules 28 days post-inoculation to determine occupancy rates (Materials and methods). One or both strains were marked with distinct antibiotic resistance to facilitate identification via plating after extraction from nodules. M. sativa plants were inoculated with mixtures of Rm1021 derivatives (Fig 4A), while M. truncatula plants were inoculated with mixtures of WSM419 derivatives (Fig 4B). Consistent with previous reports [44,88], we found that a small percentage of the nodules contained mixed populations of bacteria, while the majority of nodules were dominated by one of two strains. Discounting those nodules containing mixed populations, roughly equal numbers of nodules (45–55%) were occupied by each strain when the inoculum contained two wild-type strains (marked or unmarked). In contrast, lppA or jspA mutants were recovered from significantly fewer nodules (6–26% of the nodules) when competed against wild-type strains (Fig 4 and S4 Table). These results align with previous demonstration that JspA is important for protection against the NCR247 antimicrobial peptide and for competitiveness during symbiosis between S. meliloti and M. sativa [63]. Furthermore, our results show that both JspA and LppA contribute to competitiveness, in two different model symbiotic interactions.

Fig 4. Proportions of root nodules colonized by each bacterial strain after competitive infection.

The predominant strain colonizing each nodule was determined after seedlings were inoculated with equal mixtures of two strains. (A) S. meliloti Rm1021 and its derivatives were used to infect M. sativa, while (B) S. medicae WSM419 and its derivatives were used to infect M. truncatula. Rm1021 (Ω) is a derivative of Rm1021 marked with resistance to spectinomycin, while WSM419^R are derivatives of WSM419 marked with resistance to spectinomycin or neomycin. Mutations in jspA or lppA in Rm1021 were deletions or transposon insertions, while those in WSM419 were all deletions. Percentages (± standard deviations) below each competition indicate the mean proportions of nodules containing the jspA or lppA mutant, wild type (WT), or a mixture of the two, while the graphs depict relative abundance when mixed nodules are excluded. Dark grey circles indicate the percentage of nodules occupied by WT for individual competition trials. Error bars represent standard deviations. *, p < 0.05; **, p < 0.01. Bottom row of each table [Trials (nodules)] indicates the number of trials (and total number of nodules assessed) per competition. Detailed results from the competitive symbiosis assays are available in S4 Table.

https://doi.org/10.1371/journal.pgen.1010776.g004

Next, we investigated how JspA and LppA may influence EPS-I production by determining if they affect gene expression in S. meliloti. First, we used a transcriptional fusion to the β-glucuronidase (GUS) reporter gene [36] to measure expression of exoY, encoding a galactosyltransferase required for EPS-I biosynthesis [49]. Expression levels were examined in both PYE and LB rich media because our past experiences indicated that differences between genotypes could be more apparent in one particular medium [44]. Consistent with EPS-I levels monitored via calcofluor fluorescence (Fig 3 and S2 Table), deletion of jspA or lppA reduced exoY expression to 55–72% of wild-type levels, in both LB and PYE media (Fig 5 and S5A Table). Second, we examined reporter fusions to flaC and mcpU [45], respectively encoding a flagellin and a chemoreceptor [89,90], as expression of genes involved in flagellar motility and chemotaxis often change in opposition to those involved in EPS-I production [44,91,92]. Deletion of jspA or lppA increased expression of these two genes significantly in LB medium (to 134–159% of wild type) (Fig 5A), but less clearly in PYE medium (to 114–129% of wild type) (S5A Table). Finally, expression levels of exoY, flaC, and mcpU in the ΔjspA ΔlppA double mutant were similar to those in the single mutants (Fig 5A and S5A Table), again suggesting that JspA and LppA function in the same genetic pathway.

Fig 5. Expression levels of select promoters in different genetic backgrounds.

Expression was measured using transcriptional fusions to uidA (encoding the GUS reporter). (A) Expression levels from exoY, flaC, and mcpU promoters in ΔjspA, ΔlppA, and ΔjspA ΔlppA mutants were compared against those in wild type (WT) grown in LB. (B, C) Expression from the exoY promoter was measured in different strains overexpressing (B) jspA or (C) lppA when grown under conditions listed below the graphs. (D) Expression levels from chvI, SMc01580, and mcpU promoters were assessed when JspA or JspA_E148A was induced in PYE with 10 mM taurine for 4.5 hours. Error bars represent standard deviations. Red asterisks within bars indicate statistically significant differences when compared against the wild type or vector-bearing strain (leftmost strain in each plot), while black asterisks above bars represent significant differences between two strains under comparison: *, p < 0.05; **, p < 0.01; ns, not significant. Data for GUS reporter expression are available in S5 Table.

https://doi.org/10.1371/journal.pgen.1010776.g005

We also used the transcriptional fusions to exoY and flaC to assess the effects of jspA or lppA overexpression from a plasmid-borne, taurine-inducible promoter (with the same P_tau-regulated constructs as those in Fig 3). Induction of jspA expression with 10 mM taurine in PYE for three hours was sufficient to significantly alter expression of both exoY and flaC in the wild-type background (Fig 5B and S5B Table). Such induction also complemented the drop in exoY expression seen in the ΔjspA mutant. In contrast, induction of lppA expression did not affect flaC or exoY significantly under various conditions tested (S5B Table). We only found a modest increase in exoY transcription in the wild-type background when lppA expression was induced for six hours with 100 mM taurine in LB medium (Fig 5C and S5B Table). Such induction also sufficed to complement the drop in exoY expression in the ΔlppA mutant. Notably, overexpression of jspA could not reverse the decrease in exoY expression in the ΔlppA mutant (Fig 5B), and overexpression of lppA could not reverse the same effect in the ΔjspA mutant (Fig 5C). Overall, expression analysis with transcriptional reporters reflected the results obtained with calcofluor fluorescence (Fig 3): lppA appears to require higher levels of induction compared to jspA to cause detectable physiological changes, and both genes need each other to function efficiently.

To determine if membrane anchoring of LppA and proteolytic activity of JspA are critical for function, we mutated the lipobox motif of LppA and the peptidase domain of JspA (Fig 1B) and assessed the mutant derivatives’ effects on EPS-I synthesis and gene expression. For LppA, we mutated Cys₂₃ of the lipobox motif to Ser and tagged both the wild-type and mutant versions at the C-terminus with an HA epitope. While constructing the lppA-HA allele, we serendipitously obtained alleles with conversion of Gly₉₆ to Trp and Ala₇₈ to Ser and decided to analyze each of the corresponding two mutants as well. Due to the relatively minor changes in reporter gene expression when lppA was overexpressed in the wild-type background (Fig 5C and S5B Table), we mainly examined the functionality of various lppA alleles in the ΔlppA background. Overexpression of these various derivatives (LppA_C23S, LppA-HA, LppA_C23S-HA, LppA_G96W-HA, LppA_A78S-HA) in the ΔlppA mutant from the plasmid-borne P_tau promoter did not increase the fluorescence levels of colonies on calcofluor plates compared to the vector-only control, whereas overexpression of LppA did (Figs 6A and S2A). Induction of LppA or LppA-HA significantly elevated expression of the exoY fusion reporter to similar levels (180–195% of that in the ΔlppA mutant with the vector), while LppA_C23S and LppA_C23S-HA did not (Fig 6B and S5C Table). The other two variants, LppA_G96W-HA and LppA_A78S-HA, increased exoY expression modestly but significantly (123–130%) (Fig 6B and S5C Table). Immunoblot analysis using antibodies against the HA epitope indicated that the steady-state levels of all HA-tagged derivatives of LppA, when constitutively expressed from plasmids, are relatively similar in wild-type, ΔlppA, and ΔjspA backgrounds, except for LppA_C23S-HA, which appears to be expressed at very low levels in all backgrounds (Fig 6E). These results are consistent with acylation at the Cys residue of the predicted lipobox allowing LppA to anchor to the membrane and mature into a stable lipoprotein. Addition of an HA epitope at the C-terminus appears to reduce the functionality of LppA, while the G96W and A78S mutations further diminish protein activity without affecting its stability detectably. According to the protein structure predicted by AlphaFold [93], Gly₉₆ is located in a hairpin loop between two beta strands, while Ala₇₈ is located within an alpha helix (S2B Fig). Mutations of these residues likely alter the tertiary structure and may be helpful in the future for defining the specific molecular activity of LppA.

Fig 6. Expression of mutant lppA and jspA alleles in S. meliloti.

(A) Calcofluor fluorescence was used to assess EPS-I production in ΔlppA mutants expressing different alleles of lppA from a taurine-inducible promoter. (B) Overexpressing different alleles of lppA in the ΔlppA mutant affected expression from the exoY promoter to varying degrees, as measured with transcriptional fusion to GUS. (C, D) Wild-type Rm1021 expressing different jspA alleles from a taurine-inducible promoter exhibit varying levels of (C) fluorescence on calcofluor plates and (D) expression from the exoY promoter. (E) Immunoblots show steady-state levels of different versions of HA epitope-tagged LppA in wild-type (WT), ΔlppA, and ΔjspA backgrounds. Samples were harvested from cultures grown in LB with 100 mM taurine for 6 hours. (F) Immunoblot shows steady-state levels of of JspA-HA and JspA_E148A-HA in different genetic backgrounds. Samples were harvested from wild type or ΔlppA or ΔjspA mutants, carrying different plasmids, grown in PYE with 10 mM taurine for 4.5 hours. Presence or absence of chromosomal jspA and lppA (+ or Δ) are indicated above each lane. Numbers to the right of immunoblots (E, F) indicate approximate molecular mass standards, in kDa. Plasmids pJC532, pJC605, pJC606, pJC607, pJC608, and pJC609 were used for expressing lppA, lppA_C23S, lppA-HA, lppA_C23S-HA, lppA_G96W-HA, and lppA_A78S-HA, while pJC535, pJC555, pJC556, pJC558, pJC559, pJC560, and pJC561 were used for jspA, jspA_E148A, jspA_E148D, jspA-HA, jspA_E148A-HA, jspA_E148D-HA, and jspA_H147A-HA, respectively. Vectors used were pCM130 (A, B, D, F) or pJC478 (C). LppA and JspA variants are named according to their altered residues: for example, C23S-HA indicates LppA_C23S-HA. Error bars represent standard deviations. Red asterisks within bars indicate statistical significance when compared against the vector-carrying strain (leftmost strain in each plot): *, p < 0.05; **, p < 0.01. Data for GUS reporter expression are available in S5 Table.

https://doi.org/10.1371/journal.pgen.1010776.g006

For JspA, we mutated residues within the HEMAH active site, His147 to Ala, or Glu148 to Ala or Asp [94], and tagged wild-type and mutant versions at the C-terminus with the HA epitope. When overexpressed from the plasmid-borne P_tau promoter in the wild-type background with 5 mM taurine, both JspA and JspA-HA elevated the fluorescence of colonies on calcofluor plates, but only JspA inhibited colony formation (Fig 6C). None of the three active site mutants (H147A, E148A, E148D), untagged or tagged, enhanced fluorescence or affected growth (Figs 6C and S2C). In addition, we expressed S. meliloti JspA and its derivatives in three related alpha-proteobacteria—S. medicae WSM419, S. fredii NGR234, and C. crescentus NA1000—to assess if JspA activity is conserved (S1B and S3 Figs). In WSM419, expression of JspA complemented the EPS-I production defect of the ΔjspA mutant, whereas expression of JspA_E148A did not (S1B Fig). In both WSM419 and NGR234, overexpression of JspA inhibited colony formation, while overexpression of JspA_E148A-HA did not (S3 Fig). Overexpression of JspA_E148A and JspA-HA inhibited growth to different extents in these two species (S3 Fig), suggesting that JspA variants have different activities in distinct genetic backgrounds: overproduction of a proteolytically inactive JspA seems more deleterious in NGR234 than in S. meliloti Rm1021 or S. medicae WSM419. In contrast, expression of JspA and its derivatives in the more distantly related C. crescentus NA1000 did not inhibit growth until higher levels of induction (10 mM taurine), possibly due to the general stress of protein overproduction because similar levels of inhibition were observed for all alleles. These results suggest that the proteolytic function of JspA is conserved in Rhizobiales but not in other alpha-proteobacteria.

Results similar to those seen with calcofluor plates were obtained when analyzing JspA derivatives with transcriptional fusion of the GUS reporter to exoY in S. meliloti: overexpression of JspA-HA significantly increased exoY transcription, but to a lesser extent compared to untagged JspA (Fig 6D and S5C Table). The E148A and E148D variants, untagged or tagged, both failed to elevate exoY expression to levels achieved by JspA and JspA-HA, indicating that the active site of the protease is necessary for function (Fig 6D and S5C Table). As with JspA, JspA-HA also increased exoY expression in the ΔjspA mutant but not in the ΔlppA mutant, again indicating that JspA needs LppA for activity (S5C Table). Immunoblotting indicated that HA-tagged versions of JspA appear to reach similar steady-state levels in wild-type, ΔlppA, and ΔjspA backgrounds (Figs 6F and S2D), demonstrating that variations in activity are not likely due to differences in protein stability in different genetic backgrounds. Together, these results suggest that the active site of JspA is critical for its proteolytic activity, but other domains contribute to function because expression of proteolytically inactive or tagged versions of the protein can interfere with cellular processes to varying extents when overexpressed in distinct species (S3 Fig); this interpretation is consistent with previous demonstration by Arnold et al. [63] that the peptidase active site, the lipobox motif, and the LysM domain of JspA are all necessary for protection against the antimicrobial activity of NCR247.

Considering that LppA and JspA are likely extracytoplasmic, we wondered how they influence transcription of EPS-I and flagellar genes. To uncover their mechanism of action, we performed whole-genome expression analysis using Affymetrix GeneChips [92]. Having generated mutant alleles and determined growth conditions and strain backgrounds with which significant changes in gene expression could be observed, we decided to examine the transcriptomes of Rm1021 carrying an empty vector or overexpressing wild-type JspA or loss-of-function JspA_E148A from a plasmid. As expected, overexpression of JspA caused changes in a substantial set of genes, whereas JspA_E148A did not: pairwise comparisons for changes greater than 1.5-fold revealed 198 genes with significantly different expression between strains overexpressing JspA and carrying the vector, 155 genes between JspA and JspA_E148A, and only 5 genes between JspA_E148A and the vector (Fig 7A and S6 Table). 141 gene expression changes were shared between the JspA versus vector and JspA versus JspA_E148A comparisons, and consequently these genes were deemed strong candidates for the JspA transcriptome: 80 increased expression and 61 decreased expression during JspA overexpression. Consistent with measurements of exoY and flaC reporter fusions (Fig 5 and S5B Table), a sizable portion of up-regulated genes are associated with exopolysaccharide biosynthesis, while the majority of down-regulated genes are associated with flagellar motility and chemotaxis. One up-regulated target that stood out is chvI, encoding a conserved response regulator critical for viability and symbiosis [13,30]. To verify the results of the transcriptome analysis, we constructed transcriptional fusions of the GUS reporter to chvI and select candidates of the JspA transcriptome. Measurements of GUS activity showed expected increase (for chvI, SMc01580, and pckA) or decrease (for mcpU) when jspA expression is induced with taurine from the plasmid-borne P_tau promoter, compared to the same wild type carrying the vector; in contrast, overexpression of the mutant jspA_E148A allele did not elicit significant changes (Fig 5D and S5D Table). We also evaluated expression of the chvI reporter when JspA and variants were expressed from a plasmid-borne, IPTG-inducible promoter, P_lac [95] to ensure that the observed changes could be replicated with another inducer. Overexpression of JspA and JspA-HA both increased chvI transcription, whereas JspA_E148A, untagged or tagged, did not cause similar effects (S5D Table). These reporter activities support the validity of the transcriptome analysis.

Fig 7. Venn diagrams depicting overlaps of gene sets.

(A) Circles of the Venn diagram represent the numbers of genes whose expression changed >1.5-fold in three pairwise comparisons: between strains that overexpress wild-type JspA or mutant JspA_E148A (JspA vs. JspA_E148A), between strains that overexpress JspA or carry the vector pCM130 (JspA vs. vector), or between strains that overexpress mutant JpsA or carry the vector (JspA_E148A vs. vector). The 141 genes that appeared in both the JspA vs. vector and JspA vs. JspA_E148A comparisons were grouped according to their functions, as listed on the right. ChvI belongs to the group of regulators whose gene expression increased when JspA was overexpressed. (B, C) The bottom Venn diagrams represent the overlaps of (B) genes that belong to the JspA or ExoR/ExoS transcriptome or ChvI regulon and (C) those that belong to the RpoH1 or JspA transcriptome. Gene sets and analyses of their overlaps are provided in S6 and S7 Tables. See Materials and methods for details about assignment of genes to the ChvI regulon and ExoR/ExoS transcriptome.

https://doi.org/10.1371/journal.pgen.1010776.g007

Knowing that the ExoR-ExoS-ChvI signaling system can control its own expression [30,96], we asked if JspA participates in that regulatory pathway. We compared the JspA transcriptome against the published transcriptomes of ExoR/ExoS and ChvI [21,30,36] (see Materials and methods for details) and saw substantial overlap among the three sets of genes (Fig 7B). In contrast, the JspA transcriptome had minimal overlap with the published transcriptome of RpoH1, a heat shock sigma factor (Fig 7C) [97], chosen for comparison because it represented a stress response distinct from that of the ExoR-ExoS-ChvI system. Hypergeometric probability tests [98] indicated that overlap among the JspA, ExoR/ExoS, and ChvI transcriptomes are highly significant, whereas each of the three sets overlapped poorly with the RpoH1 transcriptome (S7 Table). As noted previously [30], there are significant overlaps between the ChvI regulon and the groups of genes affected by the podJ1 mutation [44] or by NCR247-treatment [40]. To be expected for one sharing the same genetic pathway as ChvI, JspA’s transcriptome also intersects significantly with the podJ1 and NCR247 sets (S8 Table). These similarities suggest that JspA contributes to a regulatory pathway, likely the ExoR-ExoS-ChvI system, for responding to specific stress conditions, such as those caused by the podJ1 mutation or exposure to NCR247.

To determine if JspA and LppA influence the ExoR-ExoS-ChvI signaling pathway, we conducted epistasis analysis, first using Tn5 insertions in exoR and exoS that lead to overproduction of EPS-I [27]. Mutants carrying the exoS96::Tn5 insertion produce an N-terminally truncated ExoS that behaves like a constitutively active kinase [13], while mutants carrying the exoR95::Tn5 insertion produce a C-terminally altered ExoR that has lost function [23]. Loss of lppA or jspA in these backgrounds did not reduce EPS-I synthesis, suggesting that exoR and exoS are epistatic to lppA and jspA (Fig 8A and S2C Table). Considering that, like ExoR, mature JspA and LppA are predicted to reside in the periplasm, and ExoR inhibits ExoS-ChvI signaling, we hypothesized that JspA and LppA together negatively regulate ExoR activity (Fig 1). This model is consistent with the loss of JspA or LppA being unable to reduce EPS-I production if ExoR is inoperative or if ExoS is constitutively active.

Fig 8. Epistatic interaction between jspA and lppA and the exoR-exoS-chvI pathway.

(A) Calcofluor fluorescence of wild-type and mutant strains were assessed by spotting ten-fold serial dilutions of cultures onto LB plates. Strain genotypes are shown to the left of the fluorescence images. Representative images are shown, and at least two replicates were included for each comparison. Measurements of relative fluorescence are available in S2C Table. (B) Expression of the exoY-uidA reporter was monitored in strains replete with or depleted of ChvI, while jspA or jspA_E184A was ectopically expressed. Relevant alleles on the chromosome and on plasmids are indicated below the plot: first row indicates the presence or absence of chromosomal chvI (+ or Δ), second row indicates presence of empty vector or a plasmid that expresses chvI (vector or +), and the third row indicates presence of vector or a plasmid that expresses wild-type or mutant jspA (vector, +, or E148A, respectively). Strains with the ΔchvI allele (rightmost three strains) carried a complementing plasmid (pAD101) with chvI under the control of the P_lac promoter (P_lac–chvI): growth in the presence or absence of IPTG resulted in expression or depletion of ChvI. For comparison, chvI⁺ strains carried the P_lac–chvI plasmid or the corresponding parent vector (pSRKKm). All the strains also bore a compatible vector (pCM130) or its derivatives that enable taurine-regulated expression of jspA or jspA_E184A (pJC535 or pJC555). Strains were grown in LB with 10 mM taurine for 6 hours, while expressing or depleting ChvI, prior to measurement of GUS activities. Averages and standard deviations (error bars) were calculated with measurements from at least four different days (S5E Table). **, p < 0.01 between specified measurements.

https://doi.org/10.1371/journal.pgen.1010776.g008

To test this idea further, we constructed a ChvI depletion strain, in which the only copy of chvI is under the control of P_lac on a pBBR1-derived plasmid [95]. As ChvI is essential for growth on rich medium [17,18], the ChvI depletion strain grew normally in the presence of the IPTG inducer, similar to a chvI⁺ strain carrying the same plasmid grown with or without IPTG, and poorly in the absence of the inducer (S4 Fig and S9A Table). We then monitored expression of the exoY reporter when ChvI was replete or depleted and when wild-type or mutant JspA was overexpressed from the P_tau promoter on a compatible RK2-derived plasmid [83] (Fig 8B and S5E Table). In a chvI⁺ strain carrying the P_lac vector or the P_lac–chvI plasmid, constitutive expression of JspA from P_tau increased exoY expression compared to the same strain carrying the P_tau vector (Fig 8B, first four strains on the left), consistent with previous measurements (Fig 5B and S5B Table). In the ChvI depletion strain carrying the P_tau vector, shutting off chvI expression by removing IPTG for six hours reduced exoY expression (Fig 8B, fifth strain from left). When JspA was overexpressed in the ChvI depletion strain, exoY expression was high when ChvI was replete in the presence of IPTG (Fig 8B, sixth strain from left, + IPTG), comparable to that seen in chvI⁺ strains when JspA was overexpressed. However, depletion of ChvI in the absence of IPTG prevented exoY expression from becoming elevated by JspA (Fig 8B, sixth strain from left,—IPTG). A ChvI depletion strain overexpressing JspA_E148A (Fig 8B, rightmost strain) yielded similar exoY expression patterns as the depletion strain carrying the P_tau vector (Fig 8B, fifth strain from left). These results support the model that JspA functions upstream of ChvI: increasing JspA levels relieves the inhibitory activity of ExoR, in turn activating the ExoS sensor kinase and ChvI response regulator and promoting expression of EPS-I genes and thus EPS-I production. In the absence of ChvI, JspA is unable to stimulate expression of EPS-I genes, such as exoY (Fig 1).

The ExoR-ExoS-ChvI system effects changes in gene expression in response to environmental conditions, including acid stress [99–101]. Since JspA and LppA appear to act upstream of the system, we investigated if they mediate transmission of environmental signals. We compared the JspA transcriptome and ChvI regulon against sets of genes that changed expression upon acid stress, identified in three different studies [37–39]. Hypergeometric probability tests indicate that the overlaps between the JspA transcriptome or ChvI regulon and each of the three sets of acid response genes are significant but not as strong as that between JspA and ChvI: the most significant overlaps with the JspA transcriptome and ChvI regulon belong to the set identified by Hellweg, Pühler, and Weidner [38] (S10 Table). However, the overlaps between the JspA transcriptome or ChvI regulon with each of the three sets of acid response genes are comparable to, if not better than, overlaps among the three. Thus, we chose four representative genes (SMb21188, SMc01580, exoY, and chvI) from the JspA transcriptome and ChvI regulon that also appeared in one or more of the acid responses and examined their expression via reporter fusions in acidic, neutral, or basic pH. All four reporter fusions increased expression when wild-type Rm1021 was grown at pH 6 compared to pH 7; only exoY showed significant increase at pH 8.5 as well (Fig 9 and S5F Table). Deletion of jspA or lppA appeared to curtail this increase in response to acid stress, more obviously for SMb21188 and SMc01580 (Fig 9A and 9B, pink bars). For exoY and chvI, the deletions reduced reporter expression compared to wild type at neutral pH (Fig 9C and 9D, yellow bars), and the fold-change between pH 6 and pH 7 in the deletion mutants approximated that seen in wild type (Fig 9C and 9D, red percentages). Nevertheless, in the deletion mutants, the increase in exoY expression due to growth at pH 6 was impacted more severely than the increase due to growth at pH 8.5 (Fig 9C). Our results suggest that, while factors other than ChvI may help regulate exoY and chvI expression upon acid stress, JspA and LppA facilitate ChvI’s response to acid stress.

Fig 9. Effects of jspA and lppA on transcriptional responses to pH shift.

Changes in gene expression were determined using GUS fusions to (A) SMb21188, (B) SMc01580, (C) SMb20946 (exoY), and (D) SMc02560 (chvI) at their native loci, generated in such a way as to preserve the function of the gene being examined. GUS activities in wild-type (Rm1021), ΔjspA, and ΔlppA backgrounds were measured 4.5 hours after cultures were shifted from pH 7 to pH 6 (pink bars), 7 (yellow bars), or 8.5 (blue bars) in LB medium. Activity at pH 6 relative to pH 7 for each genotype is shown as the red percentage above each pink bar. Maroon * or ** within a bar for one of the mutants represents significant difference (p < 0.05 or p < 0.01, respectively) when compared to the same condition in wild type. Analogously, black * or ** above the bars indicates significant difference when activity at pH 6 or 8.5 is compared to that at pH 7 for the same genotype, while ns indicates no significant difference. Averages and standard deviations (error bars) were calculated from three to six independent measurements (S5F Table).

https://doi.org/10.1371/journal.pgen.1010776.g009

Because JspA is predicted to be a periplasmic protease, and JspA and LppA appear to promote ExoS/ChvI activity, in opposition to ExoR, which can be regulated via proteolysis [23,25], we assessed whether JspA reduces ExoR levels. We introduced a plasmid expressing jspA or jspA_E148A from P_lac into a derivative of Rm1021 with exoR-V5 instead of exoR at the native locus. Induction of jspA expression with 1 mM IPTG in rich media inhibited growth, starting 3–4 hours after induction (Figs 10A and S5 and S9 Table). Overexpression of jspA_E148A did not retard growth compared to a strain carrying the P_lac vector when cultures were grown in flasks (Figs S5B and S5C and S9 Table) but did slow growth in 48-well plates (Figs 10A and S5A), such that the doubling time of the strain overexpressing jspA_E148A was seven hours, compared to six hours for a strain carrying the vector (S9B Table). This difference in growth between culture formats may be due to variations in aeration, trace contaminants in the flasks, or other unknown factors. For cultures grown in flasks, we monitored steady-state levels of ExoR-V5 by immunoblotting with antibodies against the V5 epitope, at zero, three, and six hours after induction. ExoR-V5 level was detectably lower in the strain overexpressing JspA compared to strains carrying the vector or expressing mutant JspA_E148A, six hours after induction (Fig 10B). Our transcriptomic analysis indicated that exoR expression is slightly elevated (1.3x) (S6 Table) when jspA is constitutively induced. This increase in exoR expression was verified using transcriptional fusion to the GUS reporter (Fig 10C and S5G Table). Thus, the decrease in ExoR level is not attributable to a drop in exoR expression. Instead, JspA appears to negatively regulate ExoR at the protein level. The reduction in ExoR likely stimulates the ExoS-ChvI pathway, resulting in feedback that elevates exoR transcription, as previously described [96].

Fig 10. Effects of JspA on ExoR levels.

(A) Plot depicts growth of exoR-V5 strains carrying the pSRKGm vector or derivatives (pJC652 or pJC653) with jspA or jspA_E184A (noted as E148A) under control of the P_lac promoter. Strains JOE5242, JOE5244, and JOE5246 were grown in 48-well plates, with 0.4 mL PYE plus 1 mM IPTG per well. Absorbance at 600 nm (A600) was measured every 30 minutes. Average readings for three different days are depicted, with surrounding shadings indicating standard deviations. In the absence of IPTG, all strains exhibited growth patterns similar to that of the vector-carrying strain in the presence of IPTG (see S5A Fig). (B) Immunoblot shows steady-state levels of ExoR-V5 and the beta subunit of ATP synthase at 0, 3, and 6 hours after inducing expression of jspA or jspA_E184A, compared against levels in the vector-carrying strain. Approximate molecular mass, in kDa, are shown to the left of the blot, while lane numbers are shown below. Growth conditions were similar to that in (A), except that strains were cultured in flasks. (C) Expression of the uidA reporter fusion to exoR from its native locus was assessed in strains carrying the vector (pCM130) or overproducing JspA or JspA_E148A (with pJC535 or pJC555). Cultures were grown with 10 mM taurine for 4.5 hours prior to measurement of GUS activities. Averages and standard deviations were calculated from at least four measurements (S5G Table). *, p < 0.05 when compared against the vector-carrying strain.

https://doi.org/10.1371/journal.pgen.1010776.g010

To eliminate changes in ExoR levels due to transcriptional regulation, we placed a FLAG-tagged version of exoR under the control of the P_lac promoter on a plasmid and induced expression with 0.5 mM IPTG. Immunoblotting with anti-FLAG antibodies revealed a band for the mature ExoR-FLAG protein at ~29 kDa, as well as a band with slightly larger molecular mass, indicative of the pre-processed form (computed to be 32 kDa with the signal peptide) (Fig 11A, lane 2). We also observed additional bands with smaller molecular masses, likely representing degradation products, the most prominent one (labeled as “deg” in Fig 11) being approximately 20-kDa in size and possibly the same as a C-terminal fragment of ExoR (ExoR_C20) detected in a previous study [23]. Co-expression of JspA variants altered the steady-state levels of mature ExoR-FLAG to different extents, whereas levels of the ExoR-FLAG precursor remained relatively uniform, consistent with ExoR being degraded in the periplasm. In the jspA⁺ background, overexpression of JspA-HA reduced the steady-state level of mature ExoR-FLAG, compared to when the strain carried the vector or expressed mutant JspA_E148A-HA (Fig 11A, lanes 2–4). In the ΔjspA background, ExoR-FLAG levels were elevated compared to wild type, and overexpression of JspA-HA reduced that elevation, while JspA_E148A-HA did not (Fig 11A, lanes 6–8). Probing with antibodies against the HA epitope indicated that steady-state levels of JspA-HA and JspA_E148A-HA were comparable. Expression of untagged versions of JspA and JspAE148A in both the ΔjspA and jspA⁺ backgrounds (Fig 11A, lanes 10–13) led to similar effects as the corresponding HA-tagged variants on ExoR-FLAG, indicating that the tagged and untagged versions of JspA behaved similarly in this assay. Next, we examined steady-state levels of ExoR-FLAG when LppA is present or absent. Again, overexpression of JspA-HA reduced ExoR-FLAG levels in both jspA⁺ and ΔjspA backgrounds (Fig 11B, lanes 1 and 3), compared to the same strains expressing mutant JspA_E148A-HA (Fig 11B, lanes 2 and 4). In contrast, in the ΔlppA background, overexpression of JspA-HA did not reduce ExoR-FLAG levels compared to overexpression of JspA_E148A-HA (Fig 11B, lanes 5 and 6). These results reinforce that JspA and LppA concertedly regulate the ExoR-ExoS-ChvI pathway by reducing ExoR protein levels.

Fig 11. Steady-state levels of ExoR-FLAG when jspA or lppA differentially expressed.

(A) Levels of ExoR-FLAG in the presence of different versions of JspA were assessed by immunoblotting with anti-FLAG antibodies (bottom blot), while expression of JspA-HA was verified with anti-HA antibodies (top blot). ExoR-FLAG expression is indicated above the blots: + signifies that expression of ExoR-FLAG from pMB859 was induced with 0.5 mM IPTG in TY medium for 4.5 hours, while—signifies that the strain carried the empty vector pSRKKm under the same conditions. Presence (+) or deletion (Δ) of the native jspA in the chromosome is also indicated. Different versions of JspA were induced with 10 mM taurine as follows: wild-type JspA from pJC614 (black +), mutant JspA_E148A from pJC615 (black *), wild-type JspA-HA from pJC616 (red #), mutant JspA_E148A-HA from pJC617 (red ^), and no expression from the vector pJC473 (-). (B) Immunoblots show steady-state levels of ExoR-FLAG and JspA-HA in the presence or absence of chromosomal lppA. The presence (+) or deletion (Δ) of native jspA or lppA on the chromosome is shown above the blots. Expression of JspA-HA, wild-type (#) or the JspA_E148A mutant (^), is indicated above the blots as in (A) (JspA-HA row). All strains in (B) expressed ExoR-FLAG from pMB859, induced with 0.5 mM IPTG. (C) Levels of ExoR-FLAG were assessed in C. crescentus NA1000 when JspA variants and LppA were co-expressed. Expression of ExoR-FLAG, LppA, and JspA is indicated above the blot for each lane, with—denoting no expression, and + denoting expression; *, #, and ^ denote expression of JspA_E148A, JspA-HA, and JspA_E148A-HA, respectively. ExoR-FLAG was induced from pMB859 with 0.1 mM IPTG in PYE medium for 4 hours, while a strain carrying the empty vector pSRKKm was used for no ExoR-FLAG expression. Expression of LppA and different versions of JspA were induced with 10 mM taurine using the following plasmids: lanes 1 and 2, pJC473 vector; lane 3, pJC614 (JspA); lane 4, pJC615 (JspA_E148A); lane 5, pJC702 (LppA and JspA); lane 6, pJC706 (LppA and JspA_E148A); lane 7, pJC707 (LppA and JspA-HA); and lane 8, pJC708 (LppA and JspA_E148A-HA). Approximate molecular mass, in kDa, are shown to the left of the blots, while lane numbers are shown below. Positions of bands representing JspA-HA and ExoR-FLAG are indicated to the right of the blots: “pre” indicates the precursor form of ExoR-FLAG, “ExoR-FLAG” the mature form, and “deg” a major degradation product. All blots were first probed with anti-FLAG antibodies and then anti-HA antibodies (see Materials and methods). For blots in (A) and (B), the anti-FLAG images were captured first, while the anti-HA images were acquired from the same respective blots after the second probing. For the blot in (C), the image was captured after the second probing.

https://doi.org/10.1371/journal.pgen.1010776.g011

To demonstrate further that JspA and LppA participate in ExoR proteolysis, we expressed the three proteins in two heterologous systems, C. crescentus NA1000 and E. coli DH10B, both of which lack clear lppA and exoR orthologs and contain weak jspA orthologs (S1 Table). In C. crescentus NA1000, ExoR-FLAG was expressed from an IPTG-inducible P_lac promoter on a pBBR1-based plasmid, while different combinations of JspA variants and LppA were co-transcriptionally expressed from a P_tau promoter on a compatible RK2-derived plasmid. When ExoR-FLAG was expressed in the absence of JspA or LppA, we detected both the mature and pre-processed forms of the protein, as well as various degradation products, by immunoblotting with anti-FLAG antibodies (Fig 11C, lane 2). Expression of JspA, but not mutant JspA_E148A, reduced the steady-state level of mature ExoR-FLAG but did not affect the pre-processed form (Fig 11C, lanes 3 and 4). Levels of mature ExoR-FLAG dropped further when LppA was expressed along with JspA or JspA-HA (Fig 11C, lanes 5 and 7); this reduction did not happen when LppA was expressed with untagged or tagged versions of JspA_E148A (Fig 11C, lanes 6 and 8). Intriguingly, one of the ExoR-FLAG degradation products, approximately 20-kDa in size and likely the ExoR_C20 fragment previously noted [23], became more prominent when only JspA, untagged or HA-tagged, was expressed (Fig 11C, lane 3; S6 Fig, lanes 5 and 9), but lost this prominence when LppA was co-expressed with JspA (Fig 11C, lanes 5 and 7; S6 Fig, lanes 7 and 11), suggesting that LppA facilitates complete degradation of ExoR.

Similar results were observed in E. coli DH10B, in which ExoR-FLAG was again expressed from the same P_lac promoter on a pBBR1-based plasmid, while expression of wild-type or mutant JspA-HA and varying levels of LppA-HA was achieved by placing different constructs under the control of a weakened P_trc promoter on a pBR322-based plasmid [102] (see Materials and methods) (S7 Fig and S1 File). In DH10B, expression of JspA-HA alone did not reduce the steady-state level of mature ExoR-FLAG (S7 Fig, lane 2), but co-expression of JspA-HA and LppA-HA did (S7 Fig, lanes 4 and 6). Furthermore, the extent of reduction in ExoR-FLAG levels depended on the level of LppA-HA expression, such that mature ExoR-FLAG became undetectable when LppA-HA was highly overexpressed (S7 Fig, lane 8). This decrease in ExoR-FLAG levels failed to occur in the presence of mutant JspA_E148A-HA (for example, lane 7 in S7 Fig). These degradation patterns in C. crescentus and E. coli suggest that LppA assists JspA to proteolyze ExoR.

All Sinorhizobium meliloti strains used in this study are derived from Rm1021 [144], and all S. medicae strains are derived from WSM419 [145]; they are listed in S11 Table. Other alpha-proteobacterial strains used were S. fredii NGR234 [146,147] and C. crescentus NA1000 [148]. E. coli strains DH5α and DH10B (both from Invitrogen) were used for molecular cloning, gene expression, and maintenance of plasmids, which are listed in S12 Table. Sinorhizobium strains were cultured at 30°C in LB, TY, or PYE media; C. crescentus was cultured at 30°C in PYE; and E. coli strains were cultured at 30 or 37°C in LB [44]. When appropriate, antibiotics, agar, sucrose, and/or calcofluor were added at previously published concentrations [44,149]. IPTG and taurine were added as inducers [83] at concentrations described in the text. For pH shifts, LB medium was buffered with 20 mM 2-(N-morpholino)ethanesulfonic acid (MES), 3-(N-morpholino)propanesulfonic acid (MOPS), or Tris, and adjusted to pH 6, 7, or 8.5, respectively, with HCl or NaOH. Growth of cultures was monitored by measuring absorbance at 600 nm (A₆₀₀), with aliquots from tubes or flasks or with a BioTek Synergy H1m plate reader if grown in 48-well plates. N3-mediated generalized transduction, mobilization of plasmids from E. coli to Sinorhizobium or C. crescentus strains via triparental mating, and two-step allelic replacement by homologous recombination were all performed as previously described [44,150–154]. Standard techniques were used for manipulation and analysis of DNA, including PCR amplification, restriction digests, agarose gel electrophoresis, ligation, and transformation [155,156]. Plasmids and DNA fragments were purified using commercial kits (Qiagen). Elim Biopharmaceuticals synthesized custom oligonucleotides and provided Sanger DNA sequencing services.

ExoR-FLAG was expressed in E. coli from an IPTG-inducible P_lac promoter on pMB859, derived from the pSRKKm vector [95]. To co-express JspA-HA, JspA_E148A-HA, and LppA-HA, we constructed plasmids derived from pDSW204, which is compatible with pSRKKm and also allows IPTG induction with a weakened P_trc promoter [102]: pJC720, pJC730, pJC731, pJC733, pJC734, pJC735, pJC736, and pJC737 (S7 Fig and S1 File). Plasmids pJC720 and pJC733 carry wild-type jspA-HA and mutant jspA_E148A-HA, respectively, including 18 nucleotides upstream of jspA’s annotated start codon. We added lppA-HA to these plasmids in three different configurations and assessed their expression empirically. For pJC730 and pJC735, the ribosome binding site (RBS) of E. coli araB [157] was appended upstream of lppA-HA and inserted after jspA-HA or jspA_E148A-HA. For pJC731 and pJC736, lppA-HA, along with 54 nucleotides upstream of its originally annotated start codon (18 nucleotides upstream of the new start codon suggested in this report), was inserted after wild-type or mutant jspA-HA. For pJC734 and pJC737, lppA-HA and its upstream sequence were inserted in front of wild-type or mutant jspA-HA. We intended the RBS of araB to enhance expression of lppA-HA in E. coli, but that configuration (pJC730 and pJC735) yielded the lowest levels of expression (S7 Fig). Because lppA-HA is in-frame with wild-type or mutant jspA-HA on pJC731, pJC734, pJC736, and pJC737, read-through translation [158] appeared to produce low levels of fusion proteins that were sometimes detectable on Western blots (S7 Fig).

Orthologs in representative genomes and their sequence similarities to the query were determined via BLAST [80]. Genomic contexts are presented as annotated in the National Center for Biotechnology Information (NCBI) database [159]. Protein domains were predicted using InterPro [160], Pfam [161], and LipoP [81].

EPS-I production was assessed as previously described [44], with LB plates containing 0.02% calcofluor white M2R (MP Biomedicals). Liquid cultures were calibrated to the same optical density (A₆₀₀ of 0.2–0.5) and serially diluted ten-fold in water, and four or five μL of the 10⁻² to 10⁻⁶ dilutions were each spotted onto calcofluor plates containing appropriate additives, such as taurine for induction and oxytetracycline for plasmid selection. Dilutions were at times spotted onto PYE plates as well for comparison. Plates were examined and photographed after 3–4 days of incubation with a Kodak 4000MM Pro Image Station, with its associated Carestream MI software and filters (430 nm excitation and 535 nm emission). The fluorescence intensity of each spot was standardized relative to the corresponding wild-type control on the same plate, and the average values of the 10⁻² to 10⁻⁴ spots from at least three independent plates were compared.

Transcriptional fusions for β-glucuronidase (GUS) assays were constructed, and GUS activities in different strains under various growth conditions were determined, as previously described [44,162]. Fusions to uidA were introduced into the genome using nonreplicating plasmids, and the wild-type function of the corresponding gene was preserved (S12 Table, S1 File). Cells were lysed after measuring the optical density of the culture (A₆₀₀), and PNPG (p-nitrophenyl-β-D-glucuronide) was incubated with the lysed cells until the mixture turned light yellow, when absorbance at 415 nm (A₄₁₅) was measured. GUS activity was derived according to the formula: A₄₁₅ x 1000 / [(incubation time in minutes) x (culture volume in mL) x A₆₀₀]. p values and statistical significance were determined using t-test (two-tailed, unequal variance).

Symbiotic association between Sinorhizobium strains and Medicago plants was assessed as previously described [44,137,163,164]. Alfalfa (M. sativa GT13Rplus) and barrel medic (M. truncatula cultivar Jemalong) were cultivated individually in 18×150-mm glass tubes on agar slants made with standard nodulation medium (as described in [165], except with 2 mM KH₂PO₄ and 0.5 mM MES, pH 6.3) and 11.5 g/L Phyto agar (PlantMedia); seeds were surface-sterilized with 70% ethanol and 50% bleach, rinsed with water, germinated in inverted 100×25-mm Petri dishes, placed on agar slants, and allowed to grow for three days at 22°C under fluorescent lamps (16-h day length) before inoculation. M. sativa was inoculated with S. meliloti Rm1021 and its derivatives, while M. truncatula was inoculated with S. medicae WSM419 and its derivatives because WSM419 is a better symbiotic partner for M. truncatula compared to S. meliloti strains, including Rm1021 [85–87]. Bacterial cells grown to mid-logarithmic phase were suspended in water to an A₆₀₀ of 0.1, and each seedling was inoculated with 0.1 mL of the suspension [approximately 10⁷ colony-forming units (CFU)]. The numbers of white and pink nodules that developed on plant roots were recorded at 14, 21, and 28 days post-inoculation (dpi). Nodules are initially white and then turn pink due to production of leghemoglobin, indicative of nitrogen fixation [7]. For competitive colonization assays, equal volumes of two cell suspensions (with A₆₀₀ of 0.1) were mixed and then diluted ten-fold, and each seedling was inoculated with 0.1 mL of the diluted mixture (approximately 10⁶ CFU). (For three of the 12 competitive assays with S. medicae strains, trials D1, D3, and E1, the inoculating mixtures were not diluted, and each seedling received 10⁷ CFU). The CFU and ratios of strains in the inoculating mixtures were determined by plating serial dilutions on PYE containing streptomycin, nalidixic acid, neomycin, or spectinomycin. Symbiosis competitiveness was assessed 28 dpi by harvesting nodules, surface-sterilizing them individually with 10% bleach, crushing each in PYE medium, and plating serial dilutions of the extracts: 10 μL each of 10⁻¹ to 10⁻⁴ dilutions were dripped to form lines on plates, and colonies were counted after three to four days of incubation at 30°C. A nodule was considered to be dominated by a particular strain if more than 80% of the CFU from the extract can be attributed to that strain. Consistent with previous reports [44,88], the majority of nodules were dominated by a single strain (Fig 4 and S4 Table). For some nodules not dominated by a single strain, colonies recovered on permissive plates (containing streptomycin for Rm1021 and its derivatives, or nalidixic acid for WSM419 and its derivatives) were re-streaked to verify the ratios of strains (S4 Table, “patch” columns). In those cases, the nodule was assigned as mixed occupancy if neither strain gave rise to at least 90% of the colonies tested. In some instances, two or more adjacent nodules were harvested together and crushed in the same tube. If such a sample was dominated by a single strain, then it counted as a single nodule for that strain. On the other hand, if the sample yielded a mixture of two strains, then it was excluded from the final tally of that particular competition trial [S4 Table, “Mixed (Multi. Nodules)”]. In Fig 4, p-values were calculated using the Mann-Whitney-Wilcoxon test (two-tailed) for the M. sativa competitions and the t-test (two-tailed, unequal variances) for the M. truncatula competitions. The Mann-Whitney-Wilcoxon test is nonparametric but has less power for smaller sample sizes (and is ineffective for a total sample size less than eight) [166]; thus, the t-test was more suitable for analyzing the M. truncatula competitions, which had fewer trials per category compared to the M. sativa competitions. S4 Table provides the t-test p-values for competitions in both host plants.

Microarray analysis of RNA transcripts using custom Affymetrix GeneChips was conducted as previously described [43,92]. RNA purification, cDNA preparation, chip hybridization, fluidics, scanning, and data analysis were performed accordingly [92]. Strains JOE3200 (carrying pCM130 vector), JOE4140 (expressing wild-type jspA), and JOE4400 (expressing mutant jspA_E148A) were grown in PYE supplemented with 0.5 μg/mL oxytetracycline and 10 mM taurine for 4.5 hours to mid-logarithmic phase, when cells were harvested for RNA extraction. Three biological replicates were used for each strain analyzed. Nine pairwise comparisons were made between two strains: a change in signal was considered significant if p ≤ 0.05. As with other gene array platforms, our DNA chip measures mRNA abundance, which is influenced by both transcription and mRNA decay; we use the term “expression” to include the sum of all factors affecting mRNA abundance. Raw microarray data have been deposited in the NCBI Gene Expression Omnibus [167] under Accession GSE155833. The significance of overlap between transcriptomes or sets of genes was determined using hypergeometric probability test, as previously described [30,98]. Genes in the ExoR/ExoS transcriptome were deduced from interrogation of exoR95::Tn5 and exoS96::Tn5 mutants that have hyperactive ChvI activity [21]; thus, many of the changes in gene expression may be indirect or independent of the ExoS-ChvI pathway. In addition, two biological replicates for each strain were used in that experiments (instead of the three used in the current analysis for JspA transcriptome), and that could have contributed to the relatively large number of genes in the ExoR/ExoS transcriptome. The ChvI regulon contains a combination of nonredundant genes from two publications: (1) those that decreased expression in the ChvI(K214T) partial-loss-of-function mutant and increased expression in the ChvI(D52E) gain-of-function mutant compared to wild type [36], and (2) group 1 and group 2 direct targets of ChvI [30].

Immunoblotting was performed using standard procedures [155,156]: 1.5 mL of culture samples were collected, resuspended in SDS sample buffer (at 150 μL for culture A₆₀₀ of 1), boiled for 5 min, resolved by SDS-PAGE, and transferred to PVDF membrane for detection by chemiluminescence (SuperSignal West Pico). Monoclonal anti-V5 (Invitrogen R960-25; AB_2556564) was used at 1:2500 dilution, anti-HA [clone 2–2.2.14] (Thermo 26183; AB_10978021) used at 1:1000 to 1:5000 dilution, and antibody to E. coli ATPase B [7E3F2] (Abcam ab110280) used at 1 ng/mL. Peroxidase-conjugated monoclonal anti-FLAG M2 antibodies (Sigma A8592) were used at 1:2000 to 1:5000 dilution. Peroxidase-conjugated donkey anti-mouse IgG antibodies (715-035-150) were from Jackson ImmunoResearch Lab and used at 1:25,000 dilution. To examine blots containing both HA and FLAG epitopes, we probed the blots first with anti-FLAG antibodies, detected bound antibodies with chemiluminescent substrates, washed the blots with Tris-buffered saline [155] containing 0.05% Tween 20 (TBST), then probed with anti-HA primary antibodies and peroxidase-conjugated secondary antibodies, and imaged again with chemiluminescence. Similarly, to detect both V5 epitope and the beta subunit of ATP synthase, we probed blots first with anti-V5 antibodies, imaged with chemiluminescence, washed the blots with TBST, then probed with antibodies against ATPase B, and treated with chemiluminescence reagents. Images captured in the second detection showed both target epitopes. Each blot image shown is representative of at least two biological replicates.