Genes related to redox and cell curvature facilitate interactions between Caulobacter strains and Arabidopsis

Bacteria play an integral role in shaping plant growth and development. However, the genetic factors that facilitate plant-bacteria interactions remain largely unknown. Here, we demonstrated the importance of two bacterial genetic factors that facilitate the interactions between plant-growth-promoting (PGP) bacteria in the genus Caulobacter and the host plant Arabidopsis. Using homologous recombination, we disrupted the cytochrome ubiquinol oxidase (cyo) operon in both C. vibrioides CB13 and C. segnis TK0059 by knocking out the expression of cyoB (critical subunit of the cyo operon) and showed that the mutant strains were unable to enhance the growth of Arabidopsis. In addition, disruption of the cyo operon, metabolomic reconstructions, and pH measurements suggested that both elevated cyoB expression and acid production by strain CB13 contribute to the previously observed inhibition of Arabidopsis seed germination. We also showed that the crescent shape of the PGP bacterial strain C. crescentus CB15 contributes to its ability to enhance plant growth. Thus, we have identified specific genetic factors that explain how select Caulobacter strains interact with Arabidopsis plants.


Introduction
Terrestrial plants and microbes have been coevolving for over 100 million years [1], and their interactions contribute to global biogeochemical cycles and agricultural fecundity [2]. Recent advances in microbial ecology have facilitated taxonomical and functional classifications of plant-associated microbes (PAMs), and core plant microbiomes (conserved microbial taxa) have begun to be identified across various plant species and diverse geographic regions [3,4]. For instance, sequence-based approaches have highlighted the abundance of Alphaproteobacteria species in (endosphere) and around (rhizosphere) the roots of many plant genera such as Arabidopsis, Glycine, Hordeum, Panicum, Sorghum, Triticum, and Zea mays across diverse geographical regions [5][6][7][8][9][10][11][12]. Pioneering work borne out of the last decade has expedited our understanding of PAMs and has highlighted the prevalence of plant-growth-promoting bacteria (PGPB) [5,8,13,14]. The seminal works of Bulgarelli et al. (2015) [5] and Lundberg et al. the ability of its microbial members to metabolize root exudates (primarily carbon), and 'hub strains' tend to play integral roles in the assembly and maintenance of plant microbiomes. However, detailed functional roles for hub strains have yet to be established, and the degree to which they function as PGPB remains elusive. Recent communications have commented on the prevalence of reductive and oxidative (redox) enzyme coding genes in the genomes of PAMs [15][16][17][18], and functional interactions between PAMs and their hosts have been further understood by implementing inoculumbased synthetic communities to explore and verify the requirement of select microbial genes for a given function (e.g., root colonization) [18,19]. Nonetheless, functional genetics approaches that seek to resolve the function of redox related activities in the context of PGPB assays have not been communicated and many reports consider only correlative data involving common PGP factors (1-aminocyclopropane deaminase (ACC deaminase), cytokinin biosynthesis, indole-3-acetic acid (IAA) production, nitrogen fixation, and phosphate solubilization) as proxies to assess the potential of a bacterial strain to enhance plant growth [20,21]. However, common PGP factors can also negatively correlate with plant fitness [22].
The genus Caulobacter, a member of the class Alphaproteobacteria, possesses many strains that have been isolated from the endosphere and rhizosphere of Arabidopsis, Citrullus, Lavandula and Zea mays [23][24][25][26], which in part implicates members of the Caulobacter genus as representative microbial hub species [27]. Moreover, select Caulobacter strains have been shown to increase plant biomass and alter root architecture relative to uninoculated conditions [22][23][24]. Functional roles that explain Caulobacter-mediated plant growth enhancement, however, have not been reported [22][23][24]. A recent report from Luo et al. (2019) [23] demonstrated that Caulobacter sp. RHG1 cells localize to regions of lateral root formation in Arabidopsis and increase root length and lateral root formation compared to the roots of uninoculated plants. Similarly, we previously identified six Caulobacter strains that could increase plant weight and root length relative to control conditions [22], and our results suggested that common PGP factors did not explain the plant growth enhancement that we observed in our system.
To identify presumptive genes that facilitate Caulobacter-mediated plant growth enhancement, we previously employed a genome-wide association study (GWAS) and observed that the genomes of PGP Caulobacter strains harbored~2-fold more genes with predicted reactive oxygen species (ROS) scavenging functions compared to the genomes of non-PGP Caulobacter strains. Specifically, we observed an extra operon (cyo) that is predicted to code for the biosynthesis of gomphrenin-I, which is a betalain-type ROS scavenging molecule that has been shown previously to exhibit high ROS scavenging activity [28]. Since ROS act as intracellular signaling molecules and facilitate plant growth and development [29][30][31][32], we hypothesized that this additional ROS scavenging-related operon may play a role in Caulobacter-mediated plant growth enhancement.
Bacterial cell shape has previously been shown to facilitate adsorption and may be a prerequisite for select cellular functions (e.g., ROS scavenging for plant host). For example, Persat et al. (2014) [40] demonstrated that the curvature of Caulobacter cells enhances colonization in flow, albeit curvature diversity may be selected for based on the environmental context. Similarly, the spiral shape of the bacterium Helicobacter pylori remains a prerequisite for effective stomach colonization and subsequent pathogenesis [33]. Recent larger scale analyses have even demonstrated that spatiotemporal distributions (i.e., proximity to plant roots as a function of time) of bacterial species may be predicated on cell shape and structure [34]. However, cell curvature has yet to be examined in the context of PGP factors. Thus, we investigated Caulobacter cell shape in the context of plant-microbe interactions and hypothesized that the crescent shape of C. crescentus cells may contribute to the Caulobacter-mediated plant growth enhancement that we previously observed [22].
To test our hypotheses, we disrupted the cytochrome ubiquinol oxidase subunit 1 (EC 1.10.3-) (hereafter cyoB) gene in two different PGP Caulobacter species (C. vibrioides CB13 and C. segnis TK0059) and compared the impact that each mutant strain had on the growth of Arabidopsis relative to that provided by their parental strains (wild-type). To determine if cell curvature facilitates PGP factors, we compared the effect of a creS mutant (required for Caulobacter crescent cell shape) on plant growth relative to its PGP parental strain (C. crescentus CB15). In addition, ROS play critical roles during seed germination [32], and we observed previously that CB13 severely inhibits seed germination rates, but it still increases plant weight relative to that of uninoculated plants [22]. Therefore, we explored whether differential gene expression patterns of cyoB across PGP Caulobacter strains occurred. As such, we reasoned that since elevated ROS levels are required for the seed-to-seedling transition in Arabidopsis [35][36][37], and previous reports have linked increased ROS scavenging activity to seed germination suppression [38,39], CB13 may exhibit elevated cyoB gene expression levels relative to other PGP strains, which would suggest that CB13 may dampen ROS levels in Arabidopsis seeds below the required oxidative window that drives seed germination [29]. Moreover, we determined that CB13 likely inhibits Arabidopsis seed germination (in part) by lowering proximal pH concentrations. Taken together, our results suggest functional roles for betalainrelated gene products and cell curvature regarding Caulobacter-mediated plant growth enhancement and demonstrate that pH reducing metabolic factors may cause CB13 to inhibit seed germination.

cyoB and creS contribute to Caulobacter-mediated plant growth enhancement
Since our previous analyses suggested that the expression of betalain synthesis related genes may contribute to the Caulobacter-mediated plant growth enhancement that we observed in our system [22], we knocked-out the expression of the cyoB gene (part of the cyoA-D operon; EC 1.10.3-) that is predicted to code for an enzyme that is involved in the biosynthesis of betalain. Using homologous recombination, we disrupted the function of the cyoB gene in two Caulobacter strains, C. vibrioides CB13 (CB13ΔcyoB) and C. segnis TK0059 (C. segnisΔcyoB) to subsequently test our hypothesis that a functional cyo operon is a PGP factor for more than one Caulobacter species. Operationally defining plant growth enhancement as increased plant weight (PW), we observed that both CB13ΔcyoB and C. segnisΔcyoB were unable to significantly enhance plant growth relative to control conditions and their parental strains (Fig 1A and S1A and S1B Table).
To tease out differential effects on specific Arabidopsis anatomical features as a result of bacterial cell inoculation, we measured the basal rosette diameter (BRD), inflorescence height (IH), and silique quantity (SQ) and then analyzed these parameters among inoculum conditions. Consistent with our PW data, the mutant strains had little impact on BRD, IH, and SQ relative to the control plants ( Fig 1B-1D and S1B Table). The one exception was that seeds that were inoculated with CB13ΔcyoB cells were unable to increase BRD relative to control conditions, but seeds that were inoculated with C. segnisΔcyoB cells were still able to enhance BRD relative to control conditions ( Fig 1B). Although changes in SQ were observed between parental and mutant strains, none of the strains increased SQ relative to the control conditions (Fig 1D), which aligns with our previous analyses [22]. Prior to using the mutant constructs for plant bioassays (Fig 1), we ensured that neither mutant incurred obvious growth defects relative to their parental strains by measuring the growth rates of each assayed Caulobacter strain under low aeration conditions (Fig 2A) and moderate aeration conditions (growth on PYE agar plates at ambient O 2 concentrations). Since differences in growth rates (cell density in PYE broth and colony forming rates on PYE plates) were not observed between mutant strains and their corresponding parental strains, and our bacterial cell re-isolation assays suggested that the observed differences in growth stimulation were likely not related to differential bacterial cell growth dynamics in the soil (Fig 2B and S2 Table), our data demonstrate the importance of a functional cyoB gene in the context of Caulobacter-mediated plant growth enhancement in two different Caulobacter species.
Since bacterial cell shape has been linked to colonization abilities [33,40], and Caulobacter cells can colonize plant roots both in artificial environments [23] and in natural environments [41], we tested whether the cell curvature of Caulobacter cells (i.e., using CB15 as a proxy for PGP Caulobacter strains since cell curvature is a conserved feature among the Caulobacter strains that we previously tested [22]) contributes to Caulobacter-mediated plant growth enhancement by conducting our plant bioassays with CB15ΔcreS (rod shaped as opposed to Welch's t-tests were performed to derive p-values. � = p < 0.01; �� = p < 0.001; ns = p > 0.01. Pairwise significance values between control and experimental conditions are connected by dashed, gray lines, and p-value matrix tables can be found in S1B Table. https://doi.org/10.1371/journal.pone.0249227.g001 the typical crescent shape of C. crescentus cells). We observed that plants grown to maturation in the presence of CB15ΔcreS cells were significantly smaller (PW, BRD) than those grown in the presence of CB15 cells (Fig 1A-1C and S1A and S1B Table). However, both the CB15 and CB15ΔcreS strains caused a reduction in SQ ( Fig 1D and S1B Table). Thus, these results suggest that cell curvature contributes to C. crescentus-mediated plant growth enhancement, albeit other Caulobacter species that lack the creS gene (i.e., C. flavus RHGG3) have been shown to enhance plant growth [26].

Elevated cyoB gene expression and media composition explain CB13-mediated seed germination inhibition for Arabidopsis seeds
Previously, we demonstrated that CB13 inhibited Arabidopsis seed germination more than any other Caulobacter strain we assayed (PGP or non-PGP) but still significantly enhanced plant growth relative to control conditions [22]. Given that a critical oxidative window is necessary to induce Arabidopsis seed germination [29], we hypothesized that CB13 may exhibit increased cyoB (presumptive betalain biosynthesis function) gene expression relative to other PGP Caulobacter strains, which would suggest that CB13 may dampen the ROS levels below the optimal oxidative window [29]. Additionally, we hypothesized that CB13 seed germination inhibition may be media-specific and concentration dependent since bacterial end-products have been shown to affect seed germination [42].
To test our first hypothesis, we performed RT-qPCR to determine the relative expression of two genes with predicted functions involved in betalain synthesis (cyoB and cydA; EC 1.10.3-) and found that the cyoB and cydA genes of CB13 were expressed at significantly higher levels than those of C. segnis (PGP Caulobacter strain that moderately decreases Arabidopsis germination rates but enhances plant growth) ( Fig 3A). To address any species-specific differences regarding gene expression, we also quantified the relative gene expression of these genes in two additional PGP Caulobacter strains that enhanced seed germination rates (C. crescentus CB15 and C. crescentus CBR1), and we observed that CB13 also expressed the cyoB and cydA genes at higher levels than those observed in these strains ( Fig 3A and S3 Table). Next, bacterial cultures were used to inoculate sterile Arabidopsis seeds and germination rates were measured 7 days post inoculation (DPI). We reasoned that since CB13 exhibits relatively high ROS scavenging related gene expression compared to other PGP strains, the Arabidopsis seeds that were inoculated with the knockout mutant cells (CB13ΔcyoB) would have increased germination rates relative to seeds inoculated with CB13 cells. Consistent with our hypothesis, we Effects of cyoB mutation and media composition on Arabidopsis seed germination. A) Relative gene expression of cyoB and cydA demonstrating the elevated expression of these genes by CB13 relative to other Caulobacter strains. Expression levels were determined using rho as the internal standard, and Δct values are displayed. Bars denote variance between independent replicates. B) Germination rate comparisons between experimental conditions are represented. Bars denote replication variances, and p-values were derived using a Welch's t-test ( �� = p � 0.001). A total of 50 seeds per condition were used in each independent replicate (n = 3). C) Seedlings grown in the presence of either sterile tap water (control), CB13 cells, or CB13ΔcyoB cells on defined media (DM; see Materials and Methods). Photos were captured at 18 days after seed plating (14 days after transfer to the environmental chamber). D) Germination rate comparisons (media composition and cyoB gene knockout effects) for CB13 and CB13ΔcyoB experimental conditions are displayed (MS = Murashige and Skoog; DM = defined media). Each dot corresponds to an independent experiment (S4 Table).
https://doi.org/10.1371/journal.pone.0249227.g003 observed that seeds inoculated with CB13ΔcyoB cells germinated at a rate of~5-fold greater than did those inoculated with CB13 cells (7 DPI), and lateral root formation was increased relative to those in the CB13 inoculum condition at 18 DPI (Fig 3B and 3C). In contrast, differences in germination rates between the C. segnis and C. segnisΔcyoB inoculum conditions were not observed (S4 Table), which is consistent with the elevated expression of cyoB that we observed in CB13 cells. Given that CB13 and CB13ΔcyoB cells appeared to grow similarly on MS plates with Arabidopsis seeds (S1 Fig), the increased cyoB gene expression that we observed in CB13 (relative to other conditions) may play a role in dampening the oxidative window below the optimal concentrations that drive Arabidopsis seed-to-seedling transitions.
To test whether the growth medium impacted CB13-mediated seed germination, we first plated CB13 inoculated Arabidopsis seeds on standard MS plates (pH adjusted to 7.5) and defined media (DM) plates (0.5 mM MgSO 4 + 1 mM CaCl 2 + 1.5% Bacto agar) and calculated relative germination rates at 7 DPI. Our results suggested that the ability of CB13 cells to inhibit seed germination is media-specific since germination rates were increased when seeds were plated on DM compared to MS plates (Fig 3C and 3D, S4 Table, and S2 Fig). Importantly, the media composition (MS vs. DM) did not affect the germination rates of the uninoculated seeds (Figs 3D, S1 and S2). In addition, we reasoned that CB13-mediated seed germination inhibition (on MS plates) would be contingent on bacterial cell concentration. To address this idea, we inoculated Arabidopsis seeds with discrete concentrations of CB13 cells (OD600 nm = 1.0, 0.1, 0.01) and observed that a decrease in CB13 cell concentration led to an increase in Arabidopsis seed germination rates on MS plates. In contrast, differing CB13 cell concentrations did not appear to alter Arabidopsis seed germination rates when they were grown on DM plates (S4 Table). Moreover, seeds that were inoculated with CB13ΔcyoB cells showed increased germination rates and enhanced root growth on each media type (MS and DM) compared to seeds that were inoculated with CB13 cells (Fig 3C and 3D and S4 Table). To determine the degree to which these two variables (media composition and cyoB function) contribute to the CB13-mediated seed germination inhibition, we analyzed this dataset using a two-way ANOVA. Our results suggested that media-composition addressed~80.0% of the germination inhibition, while 15.0% of the variation was explained by the impact of the knockout mutation and the remaining~2.0% (~3.0% uncertainty) was explained by interactions between the two variables (S5 Table). Thus, both elevated cyoB gene expression and the seed plating media composition contribute to the CB13-mediated seed germination inhibition that we previously observed [22].

CB13 may inhibit Arabidopsis seed germination by lowering local pH concentrations
Since our seed plating assay results indicated that CB13-mediated germination inhibition is significantly linked to the media-specific component, we leveraged the PATRIC 3.6.7 database to construct a flux balance analysis (FBA) metabolome model (ModelSEED) that predicts the relative H + ions exchanged (byproducts of nutrient cycling) in the environment (MS media) for each of the experimentally tested Caulobacter strains (Fig 4A). Our results suggested that CB13 harnesses the potential to yield more H + ions than any of the other Caulobacter strains that we analyzed (AP07, CB1, CB2, CB4, CB15, C. segnis TK0059, K31), and the increase of H + ion flux would likely not be buffered since phosphate fluxes were predicted to remain relatively constant (Fig 4A and S6 Table). In contrast, when we reconstructed the metabolomic potential for CB13 using DM + glucose as the substrate, the H + ion flux substantially decreased (S6 Table). Given that our FBA factored in substrate availability (MS and DM media) and reaction stoichiometry, it is likely that the results gained from our metabolic reconstruction analyses reflect those of our seed plating assays (Fig 3D). In addition, compared to the genomes of other Caulobacter strains, the CB13 genome harbors an additional predicted gene product that codes for an aldehyde dehydrogenase enzyme (2,5-dioxovalerate dehydrogenase; EC 1.2.1.26) (Fig 4B), which renders H+ ions as a result of its catalytic activity (carbohydrate metabolism). To test the results derived from our computational analyses, we measured the pH of bacterial cultures grown in MS media and DM media (+1% glucose to adjust for the carbon source that germinating seeds provide) at 11 DPI. Consistent with our FBA analyses, CB13 decreased the pH in the MS media below that of the other Caulobacter strains. In contrast, significant pH reductions in the DM media were not observed for any strain (Fig 4C and S7A Table). Further, when we tested the pH directly surrounding the developing seedlings on MS media (11 DPI), we observed that seedlings inoculated with CB13 and CB13ΔcyoB cells were surrounded by a pH of~6, whereas all other conditions maintained a pH of~7-8 (S7B Table). Moreover, when we artificially increased the local pH concentrations surrounding the developing seedlings (pH 7.5 to pH 10), we observed that both CB13 and CB13/CB13ΔcyoB inoculated seed conditions decreased the local pH concentrations (from pH 10 to 9), whereas the other conditions maintained a pH of 10 (S7B Table). As a result, CB13 and CB13ΔcyoB strains enhanced germination rates relative to neutral pH conditions and control conditions (S3 Fig). Since low pH has been linked to reduced Arabidopsis seed germination rates [43][44][45], it is plausible that the additional 2,5-dioxovalerate dehydrogenase encoding gene in the CB13 genome may (in part) contribute to CB13-mediated seed germination that we observed under neutral pH conditions.

Discussion
The advent of large-scale omics projects has catapulted our understanding of which bacterial genera tend to associate with plants, and recent studies have begun to hone our knowledgebase regarding the functional prerequisites of these plant-bacteria interactions [22,46]. However, many outstanding questions remain concerning the functional factors that many plantgrowth-promoting bacteria (PGPB) provide to their host(s). Here, we elucidate two underlying genetic factors (cyoB and creS) that contribute to Caulobacter-mediated plant growth enhancement (increased biomass), and provide computationally-derived factors that may explain the seed germination inhibition that we previously observed in our plant growth system [22].
Although the key molecular mechanisms that drive the interactions between PGP Caulobacter strains and Arabidopsis remain outstanding, our study demonstrates that a functional cyo operon is required for select PGP Caulobacter strains to enhance the growth of Arabidopsis plants. Moreover, given the predicted function(s) of the cyo operon our data suggest that ROS scavenging activities might impact positive interactions between PGP Caulobacter strains and Arabidopsis. However, the detailed mechanisms that govern the crosstalk between select PGP Caulobacter strains and Arabidopsis in the context of ROS scavenging abilities remain unknown. Therefore, future investigations will be aimed at understanding if and to what degree select PGP Caulobacter strains can regulate ROS levels in Arabidopsis plants to ultimately enhance plant growth. Nevertheless, it is well-established that in plants (as in other organisms) ROS develop as a result of aerobic metabolism, and they can cause irreversible DNA damage leading to cell death or alternatively drive important signal cascades that subsequently regulate normal plant growth and development [47,48]. Thus, ROS molecules must be kept in balance to maintain plant biochemical and physiological states. Given that plants and microbes have coevolved for millions of years [1], orchestrated processes (between plant and microbe) that maintain the balance of ROS have likely undergone functional selection.
In a previous paper, we proposed that ROS scavenging might be a PGP factor that select Caulobacter strains employ to enhance plant growth since they contain an extra cytochrome ubiquinol oxidase operon and the proteins produced from both the cyo and cyd operons can contribute to ROS scavenging [22]. Previous studies linked gomphrenin-I-a type of betalain -to high ROS scavenging activity [28] and suggested that even under optimal plant growth conditions additional ROS scavenging activity supplied by the local microbiome could modulate plant growth through development stages [47,49,50]. Given that PGP Caulobacter strains harbor the genomic architecture (i.e., cyo and cyd operons) to potentially biosynthesize multiple betalain types (Fig 5) and do not depend on the functionality of the cyo operon for survival (Fig 2), the cyo operon may indeed confer PGP Caulobacter strains with fitness benefits that could be deemed advantageous in plant-microbe contexts. Consistent with these predictions, when we disrupted the cytochrome ubiquinol oxidase cyoB gene, the resultant strain had lost its ability to enhance the growth of Arabidopsis (Fig 1). We also predicted that disruption of the cyo operon would not impair the function of the electron transport chain since some Caulobacter strains contain only the cyd operon [22]. Our bacterial cell growth assays (Fig 2A) and re-isolation data (Fig 2B and S2 Table) support this hypothesis since no differences were observed when the growth rates of the cyo knockouts were compared to those of their parent strains. However, we acknowledge that differences (e.g., plant root colonization ability) between parental and mutant strains could have persisted in vivo as a function of plant development, which our bacterial cell growth assays and re-isolation experiments would not have captured.
The cyo operon predicted protein sequences (cyoA-D) in the genomes of both CB13 and C. segnis TK0059 share significant amino acid homology (>60%) with those of various bacterial genera, and a few of the strains within these genera have been isolated from plant microbiomes (S8 Table). The cyo operon also includes three additional genes, one annotated as a SURF1 family gene that would assist in cytochrome oxidase complex assembly and two genes that code for a sensor histidine kinase and its corresponding receptor. This gene arrangement is a conserved feature of the cyo operons found in PGP Caulobacter strains whose genomes represent all three branches of the Caulobacter phylogenetic tree (Fig 5). Since the sensor histidine kinase and receptor genes are distal to the cyoB gene, the disruption of the operon in our constructs may have eliminated the expression of these downstream genes. Therefore, the loss of sensor histidine kinase expression in the cyoB mutants could contribute to the inability to enhance plant growth. Moreover, we did not investigate the functional consequence(s) of direct mutations to the cyoA,C,D gene(s), nor did we employ mutant phenotype rescue experiments (i.e., complementation); therefore, further investigations should be targeted toward understanding the functional role(s) of each gene in the cyo operon in the context Caulobacter-mediated plant growth enhancement. Nevertheless, these experiments indicate that a functional cyoB gene is required for both C. vibrioides CB13 and C. segnis TK0059 to enhance the growth of Arabidopsis.
Our previous observations [22] suggested that the interactions between developing Arabidopsis seedlings and CB13 cells were complex since CB13 cells significantly decreased seed germination but subsequently enhanced plant biomass (data collected roughly six weeks post germination). And, given the high degree of genomic synteny among the PGP Caulobacter strains we analyzed [22], we reasoned that variations in redox related gene expression among the strains may provide insight regarding these complex interactions since ROS are critical during seedling development [29,[35][36][37][38][39]. To test our hypothesis that expression of the cyo operon might explain the severe decrease in seed germination that we observed for CB13 inoculated seeds, we plated Arabidopsis seeds with either CB13 cells or CB13ΔcyoB cells and calculated germination rates. Consistent with our hypothesis, the cyoB loss-of-function mutation facilitated an increased germination rate for Arabidopsis seeds (Figs 3B and S1 and S2), and the resultant seedlings developed slightly longer roots and more root hairs relative to those inoculated with CB13 cells (Fig 3C), which is in agreement with previous reports that showed that increased ROS concentrations can increase root length and root hair formation [32,51].
Although a functional cyoB gene partially explained the CB13-mediated seed germination inhibition that we observed, germination rates still appeared diminished compared to those in control conditions and other PGP Caulobacter strain conditions (S4 Table). To establish a theoretical framework for CB13-mediated inhibition of seed germination, we performed a metabolomic reconstruction analysis of the CB13 genome and determined that growth of CB13 might lower the pH of the surrounding microenvironment. When we measured the pH of cultures and the proximal zones surrounding developing seedlings (11 DPI), we found that, as predicted, CB13 produced more acid than any other strain, which lowered the pH in the surrounding environment (Fig 4C and S7A Table). After artificially increasing the pH surrounding the developing seedlings (from 7.5 to 10), we also observed that both the CB13 and CB13ΔcyoB inoculated seeds germinated at faster rates than they did under neutral pH conditions (S3 Fig). The pH concentrations surrounding the seedlings also dropped to~9 in both the CB13 and CB13ΔcyoB inoculated conditions, whereas each of the other conditions remained at a pH 10 (S7B Table), which suggests that the expression of the cyoB gene does not impact acid production, and CB13 inhibits Arabidopsis seed germination (in part) by lowering the surrounding pH. Next, we plated Arabidopsis seeds on defined media (DM) plates where only limited growth could occur (S1 and S2 Figs) and observed that seed germination in the presence of CB13 was greatly improved (Fig 3D). However, abundant bacterial growth alone likely does not explain germination rate inhibition by CB13 since seeds inoculated with varying concentrations of CB13 cells did not appear to impact seed germination on DM plates (S4 Table), and seeds inoculated with C. segnis cells germinated efficiently despite developing in the presence of abundant bacterial growth (S1 Fig). Taken together, our observations are consistent with several reports that link low pH to decreased germination rates [43][44][45]. However, other reports [52,53] have linked low external pH to faster germination rates, and external pH changes have also been shown to modulate IAA production, pectinase activity, and iron uptake gene expression [52][53][54][55]. Therefore, the interplay between pH and several signaling pathways probably impacts seed germination in variable and complex ways.
Another functional insight that we gleaned from our experiments was the impact that cell curvature had on PGP ability (Fig 1). Using a mutant strain unable to form curved or 'crescent' shaped cells [40], we demonstrated that the loss of cell curvature reduced the ability of C. crescentus CB15 to enhance plant growth (Fig 1 and S1A and S1B Table). It is highly unlikely that cell curvature alone is the causal factor for Caulobacter-mediated plant growth enhancement since some PGP Caulobacter strains lack the creS gene [26]. A functional creS gene may, however, facilitate the presumed proximity-dependent requirement for PGP factors (i.e., a functional cyoB gene) if bacterial cell attachment to root structures is a prerequisite for Caulobacter-mediated plant growth enhancement [23], but these microscale interactions (e.g., endosphere vs. rhizosphere colonization dynamics) remain to be tested. Nevertheless, recent evidence suggests that cell curvature may provide a selective advantage for niche adaptation in select contexts [40]. Additional findings have also demonstrated that cell shape, cell wall composition, and motility factors may function as valuable proxies for estimating species abundance across environmental gradients [34], albeit the exact mechanistic factors governing these host-microbe interactions have been relatively unexplored. Nonetheless, the cell curvature of CB15 cells appears to facilitate their ability to enhance plant growth, but cell shape is not a sole determinant of Caulobacter-mediated plant growth enhancement since our previous analyses demonstrated that plant growth enhancement is not a conserved feature among C. crescentus strains [22].
Taken together, these results suggest that PGP bacteria have a complex relationship with their plant hosts and the elucidation of these relationships requires careful experimentation under controlled conditions.

Bacterial growth conditions
Overnight cultures were grown in peptone yeast extract (PYE) [56] and were derived from frozen stocks. Each culture was viewed with a phase-contrast microscope to check for contamination prior to experimentation. For low aeration growth curve assays, cells were cultured overnight, and cell cultures (mid-log phase) were then diluted 100-fold to a final volume of 10 mL with a surface area to volume ratio of 0.1:1.0. Subsequent cultures were placed in an orbital incubator shaker set to 100-150 rpm. Optical densities were collected using a Klett-Summerson photoelectric colorimeter. Growth curve assays were performed three times independently, and values are reported as Klett and cells per milliliter. In addition, overnight cultures were also streaked on PYE plates, and subsequent colony growth was observed at 24-and 48-hours post-incubation. To determine pH concentrations of the assayed cultures, bacterial cultures were grown in Murashige and Skoog (MS) [57] and defined media (DM) (1 mM MgSO 4 + 0.5 mM CaCl 2 ) supplemented with 1% glucose for 11 days and pH values were determined using a pH probe (S7A Table).

Plasmid construction and bacterial mutant generation
The plasmid used to generate cyoB mutants was commercially constructed (GeneScript), and it was used to generate gene knockouts via homologous recombination. Briefly,~250 bp of the cyoB flanking regions were cloned into the vector pUC57-Kan at PfoI and Ndel (left flanking region) and BsaXl and PfIlll (right flanking region) sites. Electrocompetent cells were prepared by resuspending mid-log phase cultures [56] in 30% glycerol, and the pUC57-Kan-cyoB vector was electroporated into either C. vibrioides CB13 or C. segnis TK0059 cells using a Bio-Rad Gene Pulser (2.5 kV, 25 μF, 400 O). Subsequently, 1 mL of PYE was added to each electroporated strain, and the resultant cell suspensions were grown for three hours at 30˚C with aeration. Afterwards, cell cultures were plated on PYE+ kanamycin (50 mg/L) agar plates and incubated at 30˚C for 48-72 hours. Single colonies were aseptically streaked onto PYE + kanamycin plates, and a single colony from each plate was grown in PYE broth to generate pure cultures for DNA extraction (DNeasy Blood and Tissue Kit). To confirm that the anticipated homologous recombination events occurred without a tandem insertion of exogenous DNA (i.e., the mutant strain constructs did not harbor the wildtype allele) in the Caulobacter strain genomes, mutant strain DNA was subjected to PCR using the following primer pairs: cyoBFWD (5'-TTTGAATTCCCTGTTCTTCGCCTGGAAGT-3'), cyoBREV (5'-TTTTTTCTCGAG ACCAGAGCGATGAAGCTCAA-3'), 16sFWD (5'-GGTTACCTTGTTACGACTT-3 0 ), and 16sREV (5'-GTGCCAGCMGCCGCGGTAA-3 0 ), and subsequent Sanger sequencing was employed to validate the sequences (both the 16s rDNA and the cyoB-Kan insert). The cell curvature mutant (CB15ΔcreS) was obtained from Zemer Gitai's laboratory at Princeton University.

Plant growth experiments
All plant growth assays were conducted as previously described by Berrios and Ely (2020) [22]. Briefly, bacterial cultures were grown overnight and were then pelleted and rinsed (3X) with sterile tap water to remove residual metabolites. Culture concentrations were adjusted to an OD600 nm = 1.0, and sterilized Arabidopsis seeds (Ler-O) were inoculated with 500 μL of the bacterial culture (depending on the condition). Control seeds were inoculated with 500 μL of sterile tap water. Seed mixtures were incubated at room temperature for 30-45 minutes and were plated on either Murashige and Skoog (MS) or defined media (DM: 1 mM MgSO 4 + 0.5 mM CaCl 2 + 1.5% Bacto agar) with pH conditions adjusted to 7.5. The plated seeds were stratified for 4 days at 4˚C and were transferred to an environmental chamber (16:8 light/dark photoperiod) under a light intensity of~150 μM/m2/s. Germination rates were calculated (total number of germinated seeds divided by the total number of plated seeds; n = 50) at 7 DPI, and the pH concentrations surrounding developing seedlings were derived using ADVANTEC 1 Whole Range pH test strips (TOYO ROSHI KAISHA, LTD.) at 11 DPI. Seedlings along with any ungerminated seeds were transferred aseptically from MS plates to sterilized soil in prewashed plastic trays (3 X 4 grid), and plastic domes were placed over each tray to increase humidity for the first week and then the domes were removed thereafter. The plants were bottom watered as needed (1-2 times per week) with sterile tap water for 5-6 weeks. Each experiment was conducted twice (24 plants per condition), which yielded a final dataset of 48 plants per condition. Fresh plant weight (PW), inflorescence height (IH), basal rosette diameter (BRD), silique quantity (SQ) data, root architecture, and bacterial cell re-isolation data were collected for each sample as previously described [22]. One-way ANOVAs and Welch's t-tests were performed to determine significant differences within and between conditions.

RNA extraction and RT-qPCR
Bacterial cultures were grown in PYE to mid-log phase (rotational incubator at 30˚C). RNA was extracted using a Qiagen RNeasy kit according to the manufacturer's protocols. The forward and reverse primers that were used to measure cyoB gene expression in both wildtype and mutant constructs were 5'-CAACTGGCTGTTCACGATGTA-3' and 5'-GATCACGAAGGT-GACCATGAA-3', respectively, and the forward and reverse primers that were used to measure cydA gene expression were 5'-TGGTCATCATGGAGAGCATCTA-3' and 5'-ACGAAGTT-GATGCCGAACAG-3', respectively. The rho gene was used as an internal control, and the corresponding forward and reverse primers used for amplification were 5'-GCACGGT-GAAGGGCGAGG-3' and 5'-GAGTCC AGCAGGATGACGA-3', respectively. Each assay was performed twice in triplicate, and relative expression (Δct) values (internal control (rho) compared to the target gene) are reported.

Comparative genomics
Metabolomic reconstruction analyses of the genomes of Caulobacter strains were conducted in PATRIC 3.6.7 and analyzed in ModelSEED [58]. Homology-based calculations were derived from BLASTn or BLASTp for nucleotide and amino acid sequence comparisons, respectively [59]. Quantitative gene binning was performed in PATRIC 3.6.7 using subsystem and pathway functions. Gene and protein sequences were deemed homologous using E-value cutoffs of 10 −5 , query coverages of >60%, and identities of >70%.
Phylogenetic analyses were performed using CLUSTAL in MEGAX (Jones-Taylor-Thornton Model or Tamura-Nei Model). Each alignment was bootstrapped (1000X), and branch lengths depict the degree of amino acid or nucleotide substitutions among sequences. A complete list of each of the strains used in these analyses and their corresponding accession numbers can be found in S6 Table. Supporting information