Motility Control Through an Anti-Activation Mechanism in Agrobacterium tumefaciens

Summary

Many bacteria can migrate from a free-living, planktonic state to an attached, biofilm existence. One factor regulating this transition in the facultative plant pathogen Agrobacterium tumefaciens is the ExoR-ChvG-ChvI system. Periplasmic ExoR regulates activity of the ChvG-ChvI two-component system in response to environmental stress, most notably low pH. ChvI impacts hundreds of genes, including those required for type VI secretion, virulence, biofilm formation, and flagellar motility. Previous studies revealed that activated ChvG-ChvI represses expression of most of class II and class III flagellar biogenesis genes, but not the master motility regulator genes visN, visR, and rem. In this study, we characterized the integration of the ExoR-ChvG-ChvI and VisNR-Rem pathways. We isolated motile suppressors of the non-motile ΔexoR mutant and thereby identified the previously unannotated mirA gene encoding a 76 amino acid protein. We report that the MirA protein interacts directly with the Rem DNA-binding domain, sequestering Rem and preventing motility gene activation. The ChvG-ChvI pathway activates mirA expression and elevated mirA is sufficient to block motility. This study reveals how the ExoR-ChvG-ChvI pathway prevents flagellar motility in A. tumefaciens. MirA is also conserved among other members of the Rhizobiales suggesting similar mechanisms of motility regulation.

Graphical Abstract

Agrobacterium tumefaciens is a plant pathogen that responds to low pH, stressful environments via a complex two-component regulatory system. This regulatory system can completely inhibit bacterial locomotion by flagella, and in this study, we define the mechanism for this inhibition and identify a novel small protein regulator that blocks flagellar gene expression.

Introduction

The bacterial flagellum is one of the most complex and sophisticated externalized structures that bacteria produce. Flagellar biosynthesis genes in bacteria are under hierarchical regulation, resulting in morphogenetic control of flagellum assembly (Chevance & Hughes, 2008, Soutourina & Bertin, 2003). The proper temporal and stoichiometric control of flagellar biogenesis dictates the inside-out assembly of flagella and provides checkpoints to ensure that the requisite flagellar components are produced in sufficient quantities (Chevance & Hughes, 2017). Based on the best-studied systems in Escherichia coli and Salmonella enterica as models, the top genes of the hierarchy (class I) encode transcription factors that activate expression of genes specifying the flagellar machinery components (class II) (Kawagishi et al., 1992). Class II genes encode proteins that make up the hook-basal body complex of the flagellum. The basal body includes the type III secretion system that exports hook and filament proteins outside of the cell for filament assembly and the torque-generating motor proteins required for filament rotation. Assembly of the basal body permits class II proteins to activate expression of class III genes (Chevance & Hughes, 2008). In the Gammaproteobacteria (such as E. coli, S. enterica, and Pseudomonas aeruginosa) and the Gram-positive bacterium Bacillus subtilis, class II genes encode the entire hook-basal body complex, and class III genes include hook-filament junction proteins and flagellin proteins (Mukherjee & Kearns, 2014).

In A. tumefaciens and other closely related Alphaproteobacteria (APB) in the order Rhizobiales, there are two tiers of class I transcription factors that initiate flagellar gene expression; these have been separated into class IA and class IB. Class IA consists of the LuxR-FixJ-type transcription factors VisN and VisR (Rotter et al., 2006, Sourjik et al., 2000, Tambalo et al., 2010, Xu et al., 2013). When either visN (ATU_RS02580) or visR (ATU_RS02585) are absent, transcription of flagellar biosynthesis and chemotaxis genes is abolished and cells are non-motile. VisN and VisR are predicted to function as heteromultimers because of the redundancy of their single mutant phenotypes (Sourjik et al., 2000). In Sinorhizobium meliloti, the motility gene expression defect of visNR mutants is directly caused by loss of expression of the class IB gene, rem (Rotter et al., 2006). Rem is an orphan OmpR-like transcription factor that is required for expression of all motility and chemotaxis genes. Rem from S. meliloti directly binds to and activates expression of several promoters that control flagellar basal body protein-encoding genes (Rotter et al., 2006). The hierarchical system of motility gene transcription in APB requires the class IB gene rem for expression of all class II and III flagellar and chemotaxis genes. Ectopic expression of rem in class IA visNR mutants is epistatic to the ΔvisNR motility defect, driving normal motility and demonstrating the regulatory hierarchy in this pathway.

Flagellar assembly and swimming motility in A. tumefaciens and related rhizobia are repressed in acidic conditions via the pH-responsive ExoR-ChvG-ChvI signaling system (Heckel et al., 2014, Yao et al., 2004, Yuan et al., 2008). In neutral conditions, the periplasmic protein ExoR binds to the sensor kinase ChvG and prevents initiation of phosphotransfer to ChvI in this two-component system (Chen et al., 2008, Wells et al., 2007). Upon cultivation in low pH growth media, ExoR is proteolytically inactivated, unfettering the ChvG-ChvI system, and impacting a large regulon (Lu et al., 2012). In A. tumefaciens, the activity of ChvG-ChvI leads to dramatic decreases in motility gene expression (Heckel et al., 2014). The expression of nearly every motility-related gene is decreased when A. tumefaciens cells encounter acidic conditions, including both flagellar biosynthetic gene clusters and chemotaxis genes found throughout the A. tumefaciens genome. This phenomenon occurs whether cells are exposed to low pH media or in an ΔexoR (ATU_RS08400) mutant (Heckel et al., 2014, Yuan et al., 2008). The mechanism by which activated ChvG-ChvI so effectively represses motility gene expression has remained unclear. We hypothesized that the low pH-activated ChvG-ChvI system functions to disrupt the flagellar gene expression hierarchy.

Another complex phenotype that is regulated by environmental pH in A. tumefaciens is attachment to surfaces. The acid-responsive ExoR-ChvG-ChvI system has a dramatic effect on attachment and biofilm formation: ΔexoR mutants are completely deficient for attachment and fail to form biofilms on both abiotic and plant surfaces (Tomlinson et al., 2010). A. tumefaciens cells grown in acidic conditions also fail to form biofilms. The attachment defect of acid-exposed cells and ΔexoR mutants is dependent on the ChvG-ChvI system, and mutation of either regulator gene in an ΔexoR mutant background restores biofilm formation capabilities (Heckel et al., 2014). Thus, a constitutively active ChvG-ChvI system serves to inhibit both swimming motility and biofilm formation. The uniformly negative impact that active ChvG-ChvI has on motility and attachment is atypical, because these two complex phenotypes are often considered as inverse processes; that is, bacteria will progress through either a motile or sessile program depending on the chemical profile of their environment (Prüß, 2017). The VisN and VisR regulators have a reciprocal impact on motility and biofilm formation (Xu et al., 2013). visN and visR mutants are non-motile but also over-produce the adhesive unipolar polysaccharide (UPP), which results in a hyper-attachment phenotype. This is despite the observation that non-motile and aflagellate A. tumefaciens mutants have a significant deficiency in surface attachment under non-flowing conditions, a phenotype likely due to inefficient engagement with surfaces (Merritt et al., 2007). The increased UPP production in visN and visR mutants is due to elevated levels of the intracellular second messenger cyclic diguanylate monophosphate (cdGMP) and is mediated through control of specific cyclic-di-GMP synthases (diguanylate cyclases, DGCs) (Xu et al., 2013). It is unclear to what extent the ExoR regulon intersects with the VisNR regulon to coordinate the complex developmental phenotypes of motility and biofilm formation.

We previously described transcriptome analyses for both the VisNR regulon (Xu et al., 2013) and the ExoR regulon (Heckel et al., 2014). The VisNR regulon was determined by comparing the transcriptomes of a ΔvisR strain and wild-type A. tumefaciens, and the ExoR regulon was determined by comparing ΔexoR to wild-type. The VisNR and ExoR regulons shared a striking congruence of down-regulated genes in the motility and chemotaxis category; 51 genes from over a dozen operons display severely reduced expression in both mutant backgrounds. This is not surprising for the VisNR regulatory network; VisNR are class IA motility regulator proteins, so we expect expression of all flagellar synthesis and chemotaxis genes to be dependent on VisN and VisR. Importantly, expression of rem (ATU_RS02820), encoding the class IB regulator was greatly reduced in the ΔvisR mutant compared to wild-type. Although almost all the motility and chemotaxis genes are also dramatically decreased in the ΔexoR mutant, conspicuously absent are the class I motility regulators visN, visR, and rem (Heckel et al., 2014). These findings suggest that the ExoR-ChvG-ChvI pathway impacts motility downstream of rem expression. In this study, we examined the interaction between these pathways and discovered a previously undefined ExoR-ChvG-ChvI-regulated target gene that directly impacts Rem-dependent gene activation.

Results

Convergence of the VisN-VisR-Rem and ExoR-ChvG-ChvI Pathways

We previously demonstrated that deletion of the chvI gene (ATU_RS00165) relieves the inhibition of motility and chemotaxis genes in the ΔexoR mutant (Heckel et al., 2014). To further investigate the role of chvI, we examined the impact of mutating its site of phosphorylation. As with many two-component system response regulators, ChvI is phosphorylated at a conserved aspartate residue (D52) in its N-terminal receiver domain to become active, in this case via the activity of the ChvG sensor kinase (Fig. S1) (Cheng & Walker, 1998). We mutated the chvI codon for this aspartate to glutamate (D52E) and to asparagine (D52N), to mimic and prevent phosphorylation by ChvG, respectively. Allelic replacement mutants of the native chvI gene with these alleles in the wild type and the exoR mutant were then tested for control of motility and biofilm formation. Similar to exoR, the chvID52E mutant was non-motile and had a severe biofilm defect; the phenotypes of this mutant were not impacted by deletion of exoR (Fig. 1A). In contrast, the chvID52N mutation had no effect on motility in an otherwise wild-type background but was completely epistatic to the non-motile and non-biofilm exoR mutant phenotypes. The chvID52E mutant also phenocopies the ΔexoR deletion strain for elevated succinoglycan production (Heckel et al., 2014, Tomlinson et al., 2010), evidenced by dramatic binding of Calcofluor, a polysaccharide-specific dye, and diffuse fluorescence around the colony (Fig. 1B). In contrast, when the native copy of chvI was replaced with the chvID52N allele, colonies show a slight decrease in Calcofluor binding relative to wild type (Fig 1B). The exoR exoA (ATU_RS18935) double mutant does not produce succinoglycan, and very weakly fluoresces with Calcofluor.

Figure 1. Phenotypes from chvID52E and chvID52N mutations.

(A) The swimming and biofilm deficiencies of the ΔexoR mutant are recapitulated by the chvID52E allele. Motility was assessed as swim ring diameter in 0.3% Bacto Agar after three days of incubation at RT (room temperature) and reported values are relative to wild-type C58. Biofilms were developed on PVC coverslips in ATGN media supplemented with 22 μM FeSO₄ · 7H₂O for 48 h. Biofilm biomass was determined by solubilizing 0.1% crystal violet-stained biofilm cells in 33% acetic acid and measuring the absorbance at 600 nM (A₆₀₀). Crystal violet absorbance was normalized per cell count by dividing by the planktonic culture OD₆₀₀ (A₆₀₀/OD₆₀₀). The data are presented as biofilm biomass relative to wild type. Error bars represent standard deviation for at least three replicates for each strain. (B) Calcofluor binding by chvI mutants. Fluorescence of solid culture bacterial growth from 5 ml of cell suspension (OD₆₀₀ 0.2) after 48 h incubation on ATGN agar supplemented with 200 μg/ml Calcofluor, and visualized by UV light illumination. Calcofluor binds to the exopolysaccharide succinoglycan, the production of which is elevated in ΔexoR (C) β-galactosidase assays of A. tumefaciens derivatives harboring plasmid-borne lacZ fusions. Activity of the five indicated promoters was assessed by translational fusions to lacZ and measurement of β-galactosidase activity. Cultures were inoculated from colonies and grown to exponential phase (OD₆₀₀ 0.3 – 0.8) for promoter activity assessment. Error bars represent standard deviation for at least three replicates.

Deletion of exoR does not alter expression of a plasmid-borne rem-lacZ translational fusion (Fig. 1C) nor does it impact a similar lacZ translational fusion to visN (Fig. S2). These results are consistent with transcriptomic analysis of the ΔexoR mutant, as there is no substantial reduction of visN, visR, or rem transcript abundance in this genetic background (Heckel et al., 2014). In fact, rem-lacZ expression remains unchanged in the ΔchvI, chvID52E, and chvID52N mutants (Fig. 1C). In contrast, similarly constructed lacZ fusions to the motility genes motA (ATU_RS02760) and flgE (ATU_RS02825), encoding a motor protein and the flagellar hook protein respectively, were greatly diminished in the ΔexoR and chvID52E mutants, but equivalent to wild type in the ΔchvI null and chvID52N mutants. Conversely, activity of lacZ fusions to chvI itself and the predicted outer membrane protein aopB (ATU_RS05585) were strongly elevated in ΔexoR and chvID52E mutants (Fig. 1C); these results correlate well with prior transcriptomic data suggesting that the transcription of chvI and aopB are elevated in the ΔexoR mutant (Heckel et al., 2014). The aopB fusion was decreased relative to wild type in ΔchvI and chvID52N mutants, whereas the chvI-lacZ was equivalent to wild type. Thus, although these data suggest that phospho-ChvI activates expression of target genes such as aopB, it inhibits expression of motility genes, and does so independent of changes to rem or visNR expression. Therefore, we explored other models of how ChvI might be regulating motility gene expression.

ChvI and Rem independently bind their own target promoters but do not cross-recognize binding sites and do not interact with each other in vitro.

An alternate model for ChvI-mediated inhibition of flagellar motility gene expression was that ChvI directly competes with Rem for binding to Rem-dependent motility promoters, thereby preventing their transcriptional activation. ChvI is a PhoB/OmpR class response regulator, with a winged helix-turn-helix motif, and it regulates many genes in A. tumefaciens (Heckel et al., 2014, Okamura et al., 2000). Prior studies demonstrated binding of purified ChvI^D52E protein in vitro to the promoter region of the type VI secretion system (T6SS) genes, specifically the intergenic region between the divergent operons that initiate with impA (ATU_RS20340) and clpV (ATU_RS20345) (Wu et al., 2012). We tested the ability of purified preparations of N-terminal, hexahistidinyl-tagged derivatives of ChvI^D52E and wild type ChvI to bind to a PCR amplicon of this promoter region (defined as P_hcp) in vitro as determined by electrophoretic mobility shift assay (EMSA) (Fig. 2A–B). Consistent with the prior study, His₆-ChvI^D52E was able to robustly shift the P_hcp DNA fragment (Fig. 2A). In contrast to the prior study, however, we observed similar binding of P_hcp by the wild-type His₆-ChvI allele, presumably the non-phosphorylated form of the protein (Fig. 2B). However, addition of either protein does not shift the mobility of DNA fragments from any of the upstream regions of motility genes (data not shown). To determine more specifically where ChvI binds the hcp promoter, we created smaller fragments of different portions of P_hcp (Fig. S3). We found that at least 140 bp upstream of the start codon of clpV were required in the amplicon to observe a gel mobility shift following incubation with His₆-ChvI^D52E, well beyond the transcription start site we mapped by 5’ RACE (Rapid Amplification of cDNA Ends), to be 63 bp upstream of the start codon of clpV. We also used 5’ RACE to map the transcription start sites of chvI and aopB and found them to start 84 and 142 bp upstream of their start codons, respectively. However, we did not observe His₆-ChvI^D52E binding to these promoters by EMSA (data not shown).

Figure 2. Electrophoretic mobility shift assay of DNA binding by ChvI and Rem.

(A) Electrophoretic mobility shift assays were performed to evaluate binding of His₆-ChvI^D52E to the intergenic region between the divergent P_tss and P_hcp operons. Promoter DNA (10 nM) was incubated with increasing concentrations of purified His₆-ChvI^D52E. Binding was assessed by DNA migration of reactions separated on a 6% polyacrylamide native gel. DNA was detected by staining with SYBR Safe dye. One discrete shift (indicated by asterisk) is observed when DNA is incubated with purified His₆-ChvI^D52E. The aopB promoter was used as a negative control for binding. (B) DNA Binding of His₆-ChvI^D52E and His-ChvI (wild-type) alleles. Promoter DNA of hcp, aopB, and chvI (10 nM) was incubated with and without 600 nM of purified His-tagged alleles of ChvI and assessed for binding. (C) Binding of purified Rem to 10 nM P_motA. (D) His₆-ChvI^D52E does not compete for Rem-P_motA binding. His₆-ChvI^D52E was incubated with Rem for 10 min prior to the addition of 10 nM DNA to reactions and 20 min incubation prior to electrophoresis on a 6% acrylamide gel and staining with SYBR Safe dye.

Rem is a two-component-type response regulator, with no known cognate sensor kinase, and it initially seemed plausible that ChvG might act to phosphorylate Rem as well as ChvI (although that model would not readily explain the impact of the chvID52E mutant on motility gene expression). Alignment of Rem with well-characterized response regulators indicates that, at the position for the conserved aspartic acid that is the phosphorylation site of canonical response regulators, Rem instead has a glutamate residue at this position (E50; Fig. S1, red text). However, there is an aspartate five residues more N-terminal to this site (D45, Fig. S1, yellow text). To test whether these residues are important for Rem function, we expressed rem with the glutamate (RemE50N) or with the upstream aspartate residue (RemD45N), individually mutated to asparagines. Expression of the remE50N allele complemented a Δrem mutant for motility (Fig S4A) and biofilm formation (Fig S4B) similarly to the wild-type rem allele and had no impact on the non-motile exoR phenotype. Expression of the remD45N allele partially restored motility and biofilm formation to the Δrem mutant. Together, these results suggest that the native residues D45 and E50 are not required for the ability of Rem to bind the class II motility promoters, although the D45N mutation may partially compromise Rem activity.

The transcriptional start sites of Rem-regulated class II flagellar genes flgB (ATU_RS02735) and motA (ATU_RS02760) were determined by 5’ RACE mapping. The start sites were found to be downstream of an imperfect direct repeat sequence that is conserved among multiple A. tumefaciens predicted class II motility gene promoters (CG-WCAAGWCTCRCG-CAAGNYYNNAC; Fig. S5A) and similar to repeats bound by Rem upstream of motility genes in S. meliloti (Rotter, 2006 #1465). The divergent motility gene operons that initiate with ATU_RS02755 and motA have two of these direct repeat sequences in their intergenic region. Rem was purified using an Intein tag system (in which the tag self-cleaves to release a nearly native protein). When complexed with PCR-amplified P_flgB which contains a single putative Rem binding site, a single Rem-DNA complex is observed (Fig. S5B), which suggests that Rem binds one site at this promoter. In contrast Rem forms two shifted EMSA complexes with P_motA, (Fig. 2C), indicative of two Rem binding sites. We did not observe binding of Rem to P_rem (Fig. S5C). To test whether ChvI competes with Rem for DNA binding, purified His₆-ChvI^D52E was incubated with Rem and the P_motA amplicon. We assessed binding by mixing all three components (Rem, His₆-ChvI^D52E, and P_motA) simultaneously and also with addition of His₆-ChvI^D52E to pre-formed Rem-DNA complexes. We did not observe binding of His₆-ChvI^D52E to P_motA, nor did His₆-ChvI^D52E disrupt Rem-P_motA complexes (Fig. 2D, S6A).

Given that ChvI and Rem are both response regulators, it seemed plausible that ChvI might directly interact with Rem. A farwestern assay was used to assess interaction between Rem and ChvI. Rem was immobilized on nitrocellulose membranes and probed with His₆-ChvI^D52E. Upon development with monoclonal anti-His₆ antibody, we did not observe stable association of ChvI and Rem (Fig. S6B). Overall, the genetic interaction we have observed between the ExoR-ChvG-ChvI and the VisNR-Rem pathways could not be explained by physical interaction between ChvI and Rem or ChvI and the target promoters of Rem. We therefore hypothesized that a previously-unidentified ChvI-regulated factor inhibits Rem-dependent motility gene activation upon growth in acidic medium or in upon deletion of exoR.

Isolation of motile suppressors of ΔexoR

To identify a cellular component that mediates the motility inhibition by ChvI, a transposon-mediated motility suppressor screen was performed to identify mutants that restore motility through low density agar to a ΔexoR ΔexoA double mutant (exoA is mutated to prevent the mucoidy associated with succinoglycan overproduction in the ΔexoR mutant) (Heckel et al., 2014). For simplicity, we will denote the ΔexoR ΔexoA double mutant strain as ΔexoR unless otherwise specified. As observed before, and therefore expected, multiple transposon insertional mutations were identified in the chvI-chvG locus. In addition to reversing motility, these null mutations also rescue the biofilm deficiency of ΔexoR (Heckel et al., 2014). Expression of the chvID52N allele fails to correct the biofilm formation of ΔexoR (Fig. 1A), indicating that the biofilm deficiency of ΔexoR is dependent on the phosphorylation of ChvI. Although flagellar motility is correlated with efficient biofilm formation in A. tumefaciens (Merritt et al., 2007), the biofilm deficiency of the ΔexoR mutant is more severe than that for the non-motile mutants such as Δrem, indicating that there are additional functions besides motility under ExoR control that impact biofilm formation (Fig. S7). We screened for ΔexoR suppressor mutants that rescued the motility defect but retained the biofilm deficiency of ΔexoR, and therefore retained native chvG-chvI. We obtained three such motile, biofilm-deficient transposon mutants BCH132, BCH133, and BCH134 (Fig. 3A, Fig. S8A). To confirm that the suppressor mutations do not alter the activity of ChvG-ChvI, we assessed expression of the ChvI-regulated genes chvI and aopB through activity of lacZ translational fusions in the BCH133 mutant (Fig. S8B). We found that BCH133 retained the elevated P_chvI and P_aopB activity indicative of active ChvG-ChvI. As the other suppressor mutants shared the BCH133 motility and biofilm phenotypes, these were not tested for P_chvI and P_aopB activity, but this phenotype is expected to also be shared by these mutants. Since the ΔexoR strain has a nonmotile phenotype but does not alter rem expression (Fig. 1A, S2), we predicted that these suppressors of ΔexoR would restore motility without changing rem expression. As expected, we observed no change in P_rem activity in BCH133 relative to its ΔexoRΔexoA parent strain (Fig. S8B). Sequencing at the transposon junctions revealed BCH132 had a transposon insertion within ATU_RS03690, a bioY homolog likely involved in biotin biosynthesis. The transposon in BCH133 mapped to ATU_RS10935, which is predicted to code for an ABC transporter subunit. BCH134 carried an insertion within ATU_RS09055, which is homologous to bolA from E. coli, which negatively regulates motility and positively regulates biofilm formation (Dressaire et al., 2015). Since all three transposon mutants had similar phenotypes with unrelated transposon insertion sites, we sequenced the genome of all three mutant strains. The single mutations from all three isolates clustered to the same general genomic location, diagrammed in Fig. 3A, and were first mapped to the initial A. tumefaciens C58 reference genome (BioProject Access No. PRJNA283). The independent mutations included a single base substitution (BCH134), a base deletion in the 3’ end of the annotated hypothetical gene Atu1638 (BCH133), and a single base substitution in the intergenic region between Atu1638 and the adjacent gene Atu8018 (ATU_RS08045) (BCH132). We recreated the frameshift suppressor mutation of BCH133 in a naïve ΔexoR background (ΔexoR BCH133_FS); the mutation restored motility to ΔexoR but retained the biofilm defect of the parent strain (Fig. S8C). The point mutation had no effect on swimming motility in the wild-type background (BCH133_FS), suggesting that motility may be saturated at wild type levels, or that the effects of this mutation are specific to an ΔexoR background. Given the proximity of the mutations and the fact that they all shared the same phenotype, the locus was investigated further. Introduction of an in-frame stop codon after amino acid 19 (serine) within the Atu1638 coding sequence did not restore motility to ΔexoR (Fig. S9A), suggesting that Atu1638 may not be required for the motility defect of ΔexoR. A survey for over-annotated genes performed by Yu et al. predicted that Atu1638 was erroneously annotated as a protein-coding gene (Yu et al., 2015), and Atu1638 is not annotated in a more recent C58 genome annotation (BioProject Access. No. PRJNA224116).

Figure 3. Identification of the motility inhibitor gene mirA.

(A) Top arrows: gene diagram with Atu gene numbers (labeled above gene arrow) from the C58 reference genome (Genbank BioProject Access No. PRJNA283) (Goodner, 2001 #604). Bottom arrows: gene diagram with ATU_RS numbers (labeled above arrows) from the C58 representative genome annotation from 03-10-2020 (Genbank BioProject Access number PRJNA224116) . The direction of arrows indicates the direction of transcription. Lolipops indicate mutation sites, mutations are labeled above lollipops. The table on the right notes the genomic location (nucleotide number on C58 circular chromosome) and the resulting amino acid change for ATU_RS0850 is listed as AA change for each mutation. (B) Motility assay of mirA mutation in wild-type and ΔexoR (ΔexoA) strains as well as wild type expressing p P_N25-mirA. Colonies were inoculated into the center of 0.3% Bacto Agar and incubated for 24 h. Error bars represent standard deviation for at least three replicates. (C) Activity of the P_mirA-lacZ fusion following growth in neutral conditions (ATGN pH 7) or acidic medium (AT-MES pH 5.5) in wild-type and ΔexoR (ΔexoA) strains. Strains were grown to exponential phase prior to freezing at −20 °C and thawing prior to β-galactosidase assays. Error bars represent standard deviation for at least three replicates. (D) Western blot analysis of mirA-FLAG₃ protein levels. Strains with the native mirA locus replaced with mirA-FLAG₃ were grown to exponential phase prior to separation of whole cell lysates by SDS-PAGE (12.5% polyacrylamide gel) and transfer to a nitrocellulose membrane and probing with anti-M2 antibody. Images are from a single western blot of a single gel, with identical image treatment. Bands from total protein stained with 2,2,2 trichloroethanol (TCE) are shown as a loading control.

Closer observation of this locus revealed that a small open reading frame (231 bp) overlaps all three mutations; this reading frame was not annotated in the A. tumefaciens C58 reference genome but is annotated by the MicroScope database as AERS4k1_1608 (Vallenet et al., 2009), and in the recent re-annotation of the A. tumefaciens C58 genome (referenced above) as ATU_RS08050 (Fig. 3A). We hypothesized that the 76 aa protein product of ATU_RS08050 may be required for the gain of motility in our isolated suppressors of ΔexoR. As detailed below, we have now designated this gene as mirA. To test whether mirA is necessary for the non-motile phenotype of ΔexoR mutants, we constructed a deletion mutant of the mirA coding sequence in both the wild type and ΔexoR backgrounds. We observed wild-type motility in both the ΔmirA and ΔexoR ΔmirA strains (Fig. 3B), suggesting that this gene is necessary for motility inhibition by the ExoR-ChvG-ChvI system. Given our prior observations that the ExoR-ChvG-ChvI pathway inhibits expression of class II flagellar genes (Heckel et al., 2014) that are activated by Rem (Rotter et al., 2006), we tentatively designated the gene as mirA, for motility inhibitor of Rem. To test whether translation of mirA is required for the non-motile phenotyoe of ΔexoR, we replaced the start codon ATG with a nonfunctional ATC; this mutation suppressed the ΔexoR motility defect (Fig. S8C). Additionally, introduction of the amber stop codon TAG into the mirA reading frame (following glutamic acid 20, outside of the overlap with the initially annotated Atu1638 coding sequence) also suppressed the ΔexoR motility defect (Fig. S9B), indicating that translation of the mirA gene product is specifically required for motility inhibition in ΔexoR mutants. Proteins in the same size range and homologous to MirA are conserved in multiple members of the Rhizobiales order. Alignment of putative MirA homologs reveals a number of invariant residues and overall sequence similarity (Fig. S10). The MirA amino acid sequence reveals no conserved domains but is predicted to be comprised of two alpha helical regions flanked by disordered segments at the termini (Hanson et al., 2017, Jones, 1999).

Expression of mirA is sufficient to inhibit motility and is under ExoR-ChvG-ChvI control

Multiple deep sequencing studies have identified an RNA adjacent to, and overlapping with the mirA gene (1,625,555-1,625,785 bp on the C58 circular chromosome); these RNAs detected by separate studies are RNA353 (140 bp, from position1,625,422-1,625,562), an unnamed sRNA of 214 bp (from position 1,625,425-1,625,639, and C1_1625426F (469 bp, from position 1,625,426-1,625,895), the last of which fully overlaps the mirA gene (Dequivre et al., 2015, Lee et al., 2013, Wilms et al., 2012). In the study by Lee, et al., C1_1625426F was predicted to encode a small open reading frame, which, in this study, we are naming mirA. To test whether expression of mirA is sufficient to achieve motility inhibition, we ectopically expressed a plasmid-borne copy of the mirA coding sequence from the constitutive coliphage T5 promoter P_N25 (Wang et al., 2000). We found that ectopic expression of the mirA coding sequence strongly inhibited motility to an equal extent as observed for the ΔexoR mutation (Fig. 3B), indicating that mirA is sufficient for motility inhibition. Plasmid-borne expression of mirA from its native promoter (including the RNA353 sequence) (pP_mirA::mirA) also showed a similar strong motility inhibition in a wild-type background (Fig. S11). Expression from this plasmid also complemented the ΔexoR ΔmirA strain, bringing motility back to the level of the ΔexoR strain.

To determine whether the expression mirA is dependent on the ExoR-ChvG-ChvI pathway, we fused the promoter and start codon of mirA to lacZ and performed a β-galactosidase assay at pH 7 and 5.5. We found that activity of P_mirA-lacZ was elevated in the ΔexoR mutant background and to a lesser extent in cells grown in pH 5.5 medium relative to pH 7 medium (Fig. 3C). Although we did not observe a change in mirA expression in our previous transcriptomic analysis of ΔexoR compared with wild type (Heckel et al., 2014), this is unsurprising as mirA was not previously annotated as a gene and was not included in the gene array that was used in that study. To test for regulation of MirA protein levels, we replaced the native copy of mirA with a C-terminal, in-frame FLAG₃ tag fusion. We found that levels of MirA-FLAG₃ were elevated in ΔexoR (Fig. 3D), further evidence that MirA is regulated by the ExoR-ChvG-ChvI system.

MirA regulates motility gene expression

Combining our data that mirA was necessary for motility inhibition in ΔexoR and that mirA gene expression is regulated by the ExoR-ChvG-ChvI system, we hypothesized that cells expressing mirA would be decreased for class II motility gene expression relative to wild type. To test our hypothesis that MirA regulates class II motility gene expression and to uncover genes regulated by MirA, we performed RNASeq on total RNA extracted from wild type cells harboring either an empty vector or the same plasmid with mirA expressed from the P_mirA native promoter. A total of 131 genes were differentially expressed (log₂ fold change <−0.50 or >0.50 and p<0.05) (Table S1, Fig. 4). The majority of these genes had decreased expression in cells ectopically expressing mirA (112 genes), and we observed elevated expression of a much smaller subset of the genes (19 genes). In strains harboring the mirA plasmid, expression of genes of the flagellar gene clusters and chemotactic genes were strongly down-regulated (Table S1; Fig. 4, black circles). Since ΔexoR mutants are non-motile but retain wild-type expression of flagellar gene activators visNR and rem (Fig. 1A, S2), we hypothesized that strains with elevated mirA expression would not lead to changes in expression of these genes. As predicted, we observed that expression of visN, visR, and rem were not significantly altered by expression of mirA (Table S1). Comparison of the transcriptomes from ΔexoR vs wild type with the derivative ectopically expressing mirA revealed that most genes in the MirA regulon are a subset of genes regulated by ExoR and VisR, with 96 genes differentially regulated when comparing ΔexoR and/or ΔvisR to wild type (log₂ fold change <−0.50 or >0.50 and p<0.05)(Table S2)(Heckel et al., 2014, Xu et al., 2013). A gene cluster that was uniquely down-regulated in the mirA dataset was a predicted prophage, spanning from ATU_RS05840-ATU_RS05900(Atu1183 – Atu1194). However, expression of these genes was extremely low in cells harboring both pmirA and the empty vector, so the importance and relevance of this cluster is unknown. Expression of the T6SS genes icmF and impI (ATU_RS020285 and ATU_RS020300, respectively) was slightly elevated in the strain harboring the mirA plasmid, and expression of these genes is also elevated in the ΔexoR mutant (Table S2)(Heckel et al., 2014). As we have demonstrated that His₆-ChvI^D52E binds at the promoter adjacent to this locus (P_hcp)(Fig. 2B), it is surprising that MirA may also regulate expression of these genes. Mutation of mirA does not affect chvI expression (Fig. S8B), so the increase in T6SS gene expression is unlikely to be due to elevated chvI expression upon mirA induction. Similarly, the increase in expression of impI (ATU_RS20300) cannot simply be explained as a decrease in Rem-dependent gene activation because expression of ATU_RS20300 is decreased in a ΔvisR mutant, where rem is not expressed (Table S2)(Heckel et al., 2014).

Figure 4. Ectopic mirA expression leads to the downregulation of motility and chemotaxis genes.

Log₂(fold change) and −log₁₀(p value) values for all A. tumefaciens genes from RNAseq of wild type C58 harboring an empty vector (pSRKGm) or pmirA(pMAT14, pP_lac::P_mirA::mirA), with three replicates per plasmid. Black shaded dots label genes belonging to flagellar or chemotaxis gene clusters; all other genes are shaded in gray.

MirA interacts with Rem in vitro

A potential mechanism of motility gene expression inhibition by MirA is through sequestration of the motility activator protein Rem. To test whether MirA interacts directly with Rem, a pull-down assay was performed with purified untagged Rem and MirA fused to a C-terminal hexahistidinyl tag (MirA-His₆) with cobalt resin, which interacts strongly with hexahistidinyl tags (Lichty et al., 2005). Rem did not bind the cobalt resin alone, but co-eluted with MirA-His₆ following co-incubation of the proteins (Fig. 5A). We also performed a farwestern experiment to separately test MirA-Rem interaction. Dilutions of purified Rem were separated by SDS-PAGE and transferred to nitrocellulose membranes, and following blocking, these were incubated with blocking agent either with or without purified MirA-His₆. Bound MirA-His₆ was detected using anti-His₆ monoclonal antibodies. The strength of anti-His₆ chemiluminescent signal correlated with the concentration of Rem protein and depended on the presence of MirA-His₆ (Fig. 5B).

Figure 5. MirA directly interacts with the Rem DNA-binding domain.

(A) Pull down assay reveals interaction between Rem and MirA-His₆. Rem was incubated alone or with MirA-His₆ for 30 min prior to addition of the samples to 10 μL of TALON cobalt resin. Following a 30 min incubation and three washes to remove weakly bound protein, samples were eluted in SDS-PAGE loading buffer and aliquots were electrophoresed on a Tris-Tricine gel labeled with 2,2,2 trichloroethanol (TCE) protein label. (B) Farwestern analysis of Rem and MirA-His₆. Purified Rem protein was separated by SDS-PAGE (12.5% acrylamide) and immobilized on a nitrocellulose membrane and probed with 0.4 μM MirA-His₆ (or no protein as a control) and incubated with monoclonal anti-His₆ antibody. Images of SDS-PAGE gels stained with 2,2,2 trichloroethanol (TCE) protein label are shown as loading controls. (C) Bacterial Adenylate Cyclase Two-Hybrid (BACTH) assay of MirA and Rem. MirA and Rem as well as the Rem receiver and DNA-binding domains were fused in frame to the T25 and T18 domains of E. coli adenylate cyclase for BACTH and assayed for lacZ activity when co-expressed in E. coli by plating 2 μL on LB agar with X-gal (160 μg/mL) and IPTG (500 μM) following a 24 h incubation.

Bacterial two-hybrid analysis of MirA-Rem interactions

The Bacterial Adenylate Cyclase Two-Hybrid (BACTH) assay was used to further assess interaction between Rem and MirA fusion proteins expressed in E. coli. This assay detects protein-protein interactions that bring the T25 and T18 fragments of adenylate cyclase into close enough proximity to become catalytically active to synthesize cyclic AMP (cAMP) synthesis (Battesti & Bouveret, 2012). cAMP synthesis activates expression of lacZ, resulting in blue colonies in the presence of X-gal. Co-expression of the MirA-T25 and Rem-T18 fusions resulted in blue colonies, further supporting a direct interaction between MirA and Rem (Fig. 5C). A similar result was obtained for the reverse combination Rem-T25 and MirA-T18, while negative controls harboring a sole MirA- or Rem- adenylate cyclase fragment yielded negative results, indicated by white colonies. Strains expressing MirA-T25 and MirA-T18 or Rem-T25 and Rem T18 also resulted in blue colonies, suggesting self-interactions of MirA and Rem, in addition to their formation of a MirA-Rem heterocomplex. As with many two-component-type response regulator transcription factors (Gao et al., 2007), the Rem protein contains a receiver domain and DNA-binding domain. To assess which domain of Rem interacts with MirA, we tested the receiver and DNA-binding domains individually for interaction with MirA via BACTH. We observed a strong interaction of MirA with the DNA-binding domain of Rem, and little to no interaction of MirA and the receiver domain (Fig. 5C).

MirA inhibits Rem-DNA binding

The BACTH data supports a model where MirA interacts with the Rem DNA-binding domain. To test whether this interaction impairs Rem DNA-binding activity, purified Rem and MirA-His₆ protein were co-incubated prior to incubation with a target promoter of Rem, P_motA. Gel shift inhibition by MirA was monitored through EMSA (Fig. 6A). Co-incubation with increasing concentrations of MirA-His₆ led to disruption of Rem-DNA binding, even at molar ratios of the two proteins at or below 1:1. To test whether MirA-His₆ could disrupt pre-formed complexes between Rem and P_motA promoter DNA, Rem was incubated with the promoter DNA to allow complexes to form, and then MirA-His₆ was added at varying concentrations. As the concentration of MirA-His₆ increased, we observed an increase in free DNA relative to the Rem-DNA complex (Fig. 6B). To test whether DNA-binding inhibition by MirA was specific to Rem, we co-incubated MirA-His₆ with His₆-ChvI^D52E and assessed DNA binding by ChvI (Fig. 6C). Even at concentrations of MirA-His₆ that inhibited Rem-DNA binding activity, His₆-ChvI^D52E was still able to bind the P_hcp promoter DNA, suggesting that MirA inhibition of DNA binding is specific to Rem and not an inherent property of MirA.

Figure 6. Gel shift inhibition of Rem by MirA.

(A) Rem and MirA-His₆ were pre-incubated for 15 min at RT followed by addition of 5 nM P_motA DNA for a 15 min incubation at room temperature. Reactions were run on an 8% acrylamide gel and the DNA was stained with ethidium bromide (1 μg/mL) for 15 min followed by imaging with UV by BioRad ChemiDoc. (B) Reactions were performed as in (A) but Rem was pre-incubated with 5 nM P_motA DNA for 15 minutes first, followed by the addition of MirA-His₆. (C) Pre-incubation with MirA does not inhibit His₆-ChvI^D52E DNA binding. His₆-ChvI^D52E was incubated with MirA 15 min at RT prior to the addition of 5 nM P_hcp DNA, followed by 15 min incubation at RT. Products were electrophoresed on an 8% acrylamide gel (80:1 acrylamide:bisacrylamide). DNA was labeled with SYBR Green dye and visualized by BioRad ChemiDoc.

Discussion

In this study, we investigated the mechanism by which the ExoR-ChvG-ChvI pathway inhibits flagellar motility in A. tumefaciens. We identified a new regulator of flagellar gene expression in A. tumefaciens, a previously unannotated gene that encodes a 76-aa protein we have designated MirA. We find that MirA strongly inhibits motility gene expression, but does not affect transcription of the motility master regulators visN, visR, and rem. Instead, MirA forms a complex with the DNA-binding domain of the response regulator Rem, and this interaction inhibits the association of Rem with DNA (Fig. 5–6). Transcriptome data indicate that elevated mirA levels decrease transcription of the motility genes that are activated by Rem. We believe the difference in methods employed in our previous studies sufficiently explain the few MirA-regulated genes that were not identified as part of the VisR regulon (and therefore also the Rem regulon) by DNA microarray (Table S2) (Heckel et al., 2014). Our RNASeq data suggest that the primary function of MirA is to inhibit expression of motility genes in A. tumefaciens and that the ExoR-ChvG-ChvI regulators can direct this inhibition through activation of mirA expression (Figure 7).

Figure 7. Model of MirA-dependent motility inhibition.

(A) At neutral pH, VisN and VisR activate rem expression (Rotter, 2006 #1465;Xu, 2013 #1883). Rem directly activates expression of the class II motility genes, and flagella are produced. ExoR inhibits activity of the ChvG-ChvI system (Wu, 2012 #1876) (B) In acidic pH, levels of ExoR protein decrease, de-repressing ChvG-ChvI protein (Wu, 2012 #1876), which leads to activation of mirA expression. MirA forms a complex with Rem, preventing it from binding to target promoters and thereby blocking motility gene expression.

A new control mechanism for motility control in A. tumefaciens.

Bacteria tightly regulate flagellar motility gene expression, only expressing genes for the components required as the flagellar nanomachine is being assembled. Flagellar gene expression is often regulated at multiple levels by a complex regulatory network of proteins and sRNAs (Osterman et al., 2015). A well-characterized inhibitory control mechanism is through the activity of anti-sigma factors, such as FlgM, that bind to and block the activity of motility-specific sigma factors required for transcription of late-stage promoters, including those that control expression of flagellin genes (Calvo & Kearns, 2015, Hughes et al., 1993). FlgM is secreted through the partially assembled flagellum through its type III secretion system, releasing the cytoplasmic sigma factor to direct late-stage flagellar gene expression, thereby coupling biogenesis of the flagellar basal body to production of the flagellar filament. Although no motility-specific sigma factor has been identified in the agrobacteria or rhizobia, the promoter elements for the class II flagellum biosynthetic genes suggest such an alternate sigma factor exists (Rotter et al., 2006). Initially, we hypothesized that MirA might function as an anti-sigma factor, but our results clearly show that MirA blocks flagellar assembly prior to class II gene expression through interaction with the Rem transcription factor.

We propose to add MirA to a small but growing list of post-transcriptional negative regulators of flagellar gene master regulators. In Salmonella enterica, YdiV is a regulatory protein that physically interacts with the FlhD flagellar activator protein and can effectively remove the FlhD₄C₂ motility master regulatory complex from its target DNA in a similar fashion to how MirA can interfere with Rem and its ability to activate target promoters (Wada et al., 2011, Takaya et al., 2012). In contrast to MirA, YdiV is a much larger protein (237 aa) with a separate EAL domain found in phosphodiesterases that typically degrade c-di-GMP, although YdiV does not bind or turnover this second messenger (Simm et al., 2009). Interestingly, YdiV acts to target the FlhD₄C₂ complex for proteolysis by ClpXP, in addition to blocking FlhD₄C₂ from binding to its target promoters. We cannot exclude the possibility that MirA functions as an adaptor for proteolysis of Rem in addition to inhibiting its DNA binding activity. Another post-transcriptional, negative regulator of the FlhD₄C₂ complex in S. enterica, FliT, interacts with FlhC, inhibiting motility by sequestering this regulatory component (Imada et al., 2010, Yamamoto & Kutsukake, 2006). Similar to MirA, FliT self-interacts (Imada et al., 2010). FliT acts as a secretion chaperone through interactions with filament cap protein FliD and with the chaperone export protein FliJ (Evans et al., 2006). We have not observed any additional roles for MirA beyond interaction with the A. tumefaciens motility regulator Rem, but additional binding partners of MirA are possible.

In the aquatic APB Caulobacter crescentus, biogenesis of late-stage components of the polar flagellum is regulated by the MadA-FliX-FlbD system (Siwach et al., 2021). FliX is a 144-aa trans-acting regulator of flagellar gene transcription, and directly interacts with the FlbD motility regulator (Xu et al., 2011). FliX also associates with a gating component of the flagellar export machinery, FlhA, blocking export of late-stage flagellar substrates including the flagellins. The 91-aa protein MadA binds to the export machinery and stimulates displacement of FliX from FlhA, releasing it to interact with FlbD (Siwach et al., 2021). Interactions with FliX are required for FlbD to function, but at elevated levels also inhibit its DNA binding activity (Xu et al., 2011). In contrast, our results indicate that MirA is not required for Rem activity, but rather acts to strictly inhibit its ability to bind DNA. Additionally, MirA blocks expression of all the motility genes downstream of VisNR and Rem, rather than specifically affecting late-stage genes.

Overall, we find that MirA functions as an anti-activator of Rem. Beyond motility control, there are multiple examples of negative regulators, anti-repressors and anti-activators that directly interact with DNA-binding transcription factors to modulate their activities. In A. tumefaciens, the TraR quorum sensing transcriptional regulator is inhibited through direct interactions with its anti-activator TraM (Swiderska et al., 2001). TraM forms higher-order complexes with TraR dimers, blocking their ability to bind DNA and targeting TraR for proteolysis (Chen et al., 2007, Chen et al., 2004, Costa et al., 2012). Our BACTH results provide evidence that Rem and MirA both self-multimerize, but it is not yet clear whether these putative multimers dissociate during formation of the MirA-Rem complex, or rather assemble into higher-order complexes such as observed for TraM-TraR.

Convergence of the ExoR-ChvG-ChvI and VisNR-Rem pathways.

In our efforts to define the inhibitory mechanism of ExoR-ChvG-ChvI on motility, we gained insight into Rem and ChvI-dependent gene regulation in A. tumefaciens. Rem is part of a growing family of aspartate-less receiver domain (ALR)-containing response regulators, which lack the conserved aspartate that is typically phosphorylated for canonical two component systems (Maule et al., 2015). Rem and its orthologs in other Rhizobiales have glutamate residues at the position that typically contains a phospho-accepting aspartate residue in two-component-type response regulators. It seemed plausible that the glutamate residue was acting as a stable phosphomimic, similar to what has been observed for several other two-component response regulators (Klose et al., 1993). Mutation of this glutamate residue (E50) to asparagine did not abrogate Rem activity, nor did mutation of another aspartate proximal to this site (D45, Fig. S4). Similarly, substitution of single aspartate residues proximal to the glutamate residue at the predicted phosphorylation site of Rem from S. meliloti does not abolish Rem activity (Rotter et al., 2006). Cumulatively, these results suggest that Rem activity may not be regulated by phosphorylation, that the protein may be constitutively active, and this activity does not require the glutamate residue at position 50.

Mutation of the ChvI phosphorylation site, however, did alter ChvI function in vivo. Mutation of the conserved aspartate residue at this position to asparagine (D52N) resulted in a loss-of-function allele, whereas mutation to a glutamate residue (D52E) acted as a phospho-mimic allele (Fig. 1), similar to a subset of other response regulators (Klose et al. 1993). Consistent with our expectations for a phospho-mimic ChvI allele, purified His_6-ChvID52E binds to the P_hcpA target promoter in vitro (Wu et al., 2012). Unexpectedly, we found that purified wild-type His₆-ChvI, which we assumed to be un-phosphorylated, was also able to bind P_hcp in our assays (Fig. 2B). DNA binding in vitro by non-phosphorylated response regulators is not unheard of for two-component systems. In S. meliloti, the chvID52E allele has been shown to elevate succinoglycan production as detected by Calcofluor white fluorescence (Chen et al., 2009), which we also observed for A. tumefaciens (Fig. 1B).

The ExoR-ChvG-ChvI system is a pervasive regulatory module in the Rhizobiales and has been proposed to regulate the free-living to host invasion switch (Heavner et al., 2015). MirA may represent a fourth component of this module, facilitating the motility regulation during this transition. Our results indicate that mirA expression is elevated in the ΔexoR mutant. Previous analysis of the ΔexoR transcriptome did not identify mirA as an ExoR-regulated gene (Heckel et al., 2014), but mirA was not yet annotated in the A. tumefaciens C58 reference genome (PRJNA283) at that time. Expression of mirA is elevated in acidic conditions. Although the ExoR-ChvG-ChvI pathway is recognized to enable response to acidic pH in A. tumefaciens and its relatives, there is growing evidence that this pathway may provide a more general stress response, and thus MirA-dependent control of motility may be relevant under a wider set of conditions than strictly acidic pH.

Multiple lines of evidence we present here indicate that in A. tumefaciens ChvI does not regulate visN, visR, or rem gene expression, nor does it bind directly to Rem nor the promoters of motility genes (Fig. 1C, S2, 2D, S6). Rather, ChvI inhibits motility through activating expression of mirA. In S. meliloti, ChvI has been reported to directly regulate rem expression, in contrast to our findings in A. tumefaciens (Heckel et al., 2014, Wells et al., 2007, Ratib et al., 2018). The existence of conserved MirA homologs in multiple other Rhizobiales species which also encode the Rem transcription factor (Fig. S1, S10), suggest similar mechanisms of MirA-directed motility control.

Experimental Procedures

Bacterial strains, plasmids, and oligonucleotides

Bacterial strains and plasmids are listed in Table S3, and oligonucleotides are listed in Table S4. E. coli strains were grown in LB at 37 °C and A. tumefaciens strains were grown in LB and AT minimal medium with 15 mM (NH₄)₂SO₄, buffered to pH 7 with KH₂PO₄ to a final concentration of 79 mM and with 0.5% glucose as the carbon source (ATGN). Acidic medium was prepared with a final concentration of 20 mM MES (2-(N-morpholino)ethanesulfonic acid) hydrate buffered to pH 5.5 in place of KH₂PO₄ buffer. Growth temperatures were 28-30 °C for A. tumefaciens and 37 °C for E. coli. Antibiotics were used at the following concentrations for liquid media: Kanamycin (Km) 150 μg/mL, Gentamycin (Gm) 100 μg/mL, Carbenicillim (Cb) 25 μg/mL, Spectinomycin (Sp) 150 μg/mL for A. tumefaciens; Km 25 μg/mL, Gm 25 μg/mL, Ampicillin (Ap) 100 μg/mL, Sp 25 μg/mL for E. coli.

Allelic replacement and marker-less deletion construction

Marker-less deletion strains were constructed as previously described (Morton & Fuqua, 2012). Fragments upstream and downstream of the genes of interest (0.5-1 kb in size) were amplified using primers P1 and P2 (upstream) and P3 and P4 (downstream) (Table S2). These fragments were stitched together via 18-20 base-pair complementary sequences by a SOEing reaction. The deletion or allelic replacement construct was first ligated into the pGEM T-Easy vector for amplification in E. coli strain Top 10 F.’ Mini-prepped pGEM plasmids were digested with the appropriate enzymes and the insert ligated into the suicide vector pNPTS138 (Hibbing & Fuqua, 2011). This vector was transformed into the E. coli conjugal donor strain S17-1/λpir to transfer the deletion plasmid into the appropriate A. tumefaciens derivative. Mixtures of A. tumefaciens recipient and E. coli donor strains were spotted onto cellulose acetate filter disks as described previously (Morton & Fuqua, 2012). Primary integrants were selected by plating cells collected from filter disks onto ATGN-Km plates. Cells that subsequently lost the integrated plasmid but retained the introduced deletion were isolated through passage in the absence of antibiotic by plating onto AT medium with 5% sucrose instead of glucose (ATSN) and isolating sucrose-resistant, Km-sensitive recombinants that were then confirmed by PCR across the deletion junction, and DNA sequencing.

Construction of expression plasmids.

Controlled expression plasmids were generated by introducing A. tumefaciens coding sequences amplified from genomic DNA by PCR into the LacI^Q-encoding, IPTG (isopropyl-β-D-thiogalactopyranoside)-inducible expression vector with a gentamicin resistance (Gm^R) cassette, pSRKGm (Khan et al., 2008). For expression of mirA from its native promoter and P_lac, the mirA coding sequence and upstream region were cloned into pSRKGm at the NdeI site, with a stop codon to prevent translation across the NdeI site from the upstream lacZα. For expression from the phage promoter P_N25, mirA was cloned into pYW15c (Wang et al., 2000). Coding sequences were amplified from A. tumefaciens C58 genomic DNA or plasmids carrying engineered alleles using the 5’ and 3’ primers corresponding to each gene, specified in Table S4. Amplified fragments were ligated into the cloning vector pGEM-T Easy (Promega) or built directly into the plasmid of interest using the NEBuilder HiFi Assembly kit (New England Biolabs), transformed into E. coli Top10 F’ or DH5α/λpir for amplification and collection, and the inserts were confirmed by sequencing. Coding sequences were excised from the pGEM-T Easy cloning vectors by restriction enzyme cleavage and ligated into the appropriately cleaved pSRKGm vector using T4 DNA ligase (New England Biolabs). Derived plasmids were verified by PCR amplification across the multiple cloning site and the insert prior to transformation into A. tumefaciens cells via electroporation (Mersereau et al., 1990).

Site-directed mutagenesis of ChvI and Rem

Site-directed mutagenesis was performed to create the alleles chvID52E, chvID52N, remD45N, and remE50N. Mutagenesis was performed as previously described (Hibbing & Fuqua, 2011, Mohari et al., 2018) using the protocol described in the QuickChange Site-Directed Mutagenesis Kit (Stratagene Corp). The desired nucleotide changes were designed into the complementary mutagenesis primers (Table S4). Mutant alleles were amplified from cloning constructs carrying the wild-type alleles. After amplification of the entire plasmid, template and hemi-methylated plasmid DNA were removed from the reaction via DpnI digestion. The mutagenized plasmids were transformed into E. coli Top10 F’ or DH5α/λpir, collected, sequence verified, and the alleles were fused with controlled expression plasmids or allelic replacement plasmids.

Construction of lacZ fusions and β-galactosidase assays

Promoter fusions to lacZ were constructed by amplification of upstream regions from genomic DNA with the specified primers (Table S4) and ligated into pGEM-T Easy (Promega) for amplification and collection. Promoter fragments were digested from cloning constructs and ligated into reporter plasmid pRA301 (Akakura & Winans, 2002) to generate reporter fusions with the promoter region, ribosome binding site, and start codon of the gene of interest fused in-frame with the second codon of the lacZ gene to generate translational fusions.

Bacterial cultures were grown to exponential phase at 30 °C and frozen in 30% glycerol at OD₆₀₀ 12. Thawed samples were inoculated at OD₆₀₀ 0.05 in 2 mL cultures in triplicate, grown to exponential phase at 30 °C, measured for OD₆₀₀ and frozen. Thawed samples were incubated in Z buffer (Miller, 1972) with 2 drops of 0.02% SDS, 4 drops of chloroform, vortexed 10 sec, and incubated with 100 μL 4 mg/mL ortho-nitrophenyl-β-D-galactopyranoside (ONPG) at RT. Timed reactions were terminated with addition of 600 μL 1M Na₂CO₃ and clarified through centrifugation. β-galactosidase cleavage of ONPG was measured through A₄₂₀ measurement in a BioTek plate reader and specific activity was measured in Miller Units (Miller, 1972).

5’RACE mapping transcriptional start sites

Total RNA was extracted from either wild-type A. tumefaciens (for mapping the TSS of motA and flgB) or ΔexoAΔexoR (for mapping the TSS of hcp, chvI, and aopB). RNA was extracted using the RNeasy Midi Kit from QIAGEN (Valencia, CA). Residual DNA was removed from the extracted RNA using the TURBODNA-free™ Kit from Thermo Fisher Scientific (formerly Life Technologies). Enzymes provided in the TAKARA 5’-Full RACE Core Set were used for 1^st strand cDNA synthesis, degradation of hybrid DNA-RNA, and circularization of single-strand cDNA, according to the kit instructions. Synthesis of cDNA was directed by 5’-phosphorylated primers (Table S4). Amplification of promoter regions was performed with Phusion polymerase (NEB) with the A1, S1, A2, and S2 primers (Table S4) according to the TAKARA 5’-Full RACE Core Set kit instructions. Amplified promoter fragments were ligated into the pGEM-T Easy (Promega) and transformed into E. coli Top10 F’ for collection and sequencing.

RNA sequencing

1.5 mL exponential phase cultures were mixed with 1.5 mL RNAProtect Reagent. RNA was purified with the Qiagen RNEasy kit, and DNA was digested with the Turbo RNAse-free DNAse kit for 1 hr at 37 °C. The integrity of the RNA was checked with an Agilent Tapestation. From each sample, 1 ug of total RNA was used for ribosomal RNA depletion with the Ribo-Zero™ Magnetic Kit (Bacteria)(Epicenter). The libraries were then prepared with the TruSeq Stranded mRNA library preparation kit (Illumina). The cleaned adapter-ligated libraries were pooled and loaded on the NextSeq 500 with a 75 high cycle sequencing module to generate paired-end reads. Mutations were identified by mapping reads against the A. tumefaciens C58 reference genome (PRJNA283; Genbank accession numbers AE008687, AE008688, AE008689, and AE008690 for the circular, linear, At, and Ti plasmids, respectively) using the breseq computational pipeline (Deatherage & Barrick, 2014). Transcriptome data is summarized in Table S1 and is available at GEO-NCBI (GSE174467).

Motility assays

To quantify the efficiency of flagellar locomotion, strains were inoculated into the center of motility agar plates with 0.3 % Bacto Agar and the diameter of each swim ring was measured over time. Strains were inoculated from single colonies from 1.5% agar (ATGN) plates with a toothpick or sterile wire. Swim plates were incubated at RT in a sealed container with an open vial of a saturated K₂SO₄ solution to maintain a relative humidity of ~97% in the chamber. Swim ring diameters were measured with a standard ruler.

Biofilm assays

Biofilm assays were performed as previously described (Tomlinson et al., 2010). Cultures of A. tumefaciens were inoculated at an initial OD₆₀₀ 0.05 in ATGN supplemented with 22 μM FeSO₄ · 7H₂O into 12-well plates with upright polyvinyl chloride (PVC) coverslips or into 96-well PVC plates for large-scale assays. Plates were incubated at RT for 24-48 h in a sealed container with an open vial of a saturated solution of K₂SO₄ to maintain a relative humidity of ~97% in the chamber. The coverslips or wells were stained with 0.1% crystal violet dye and the dye was solubilized in 33% acetic acid. Biofilm biomass was determined by measuring the A₆₀₀ of solubilized crystal violet and normalizing by the OD₆₀₀ of the cultures at the time of harvesting. The cell density and absorbance readings were measured with a Bio-TEK Synergy HT plate reader using Bio-TEK Gen5 (version 1.07) software.

Calcofluor white staining

ATGN plates were supplemented with 200 μg/mL of Calcofluor White (fluorescent white) dye from a sterile stock solution of 20 mg/mL in water. A. tumefaciens strains were inoculated into ATGN liquid medium from colonies and grown overnight at 28°C with shaking. Cultures were normalized to an OD₆₀₀ 0.2, and 5 μL of each culture was spotted onto the plate. Plates were incubated for 2 days at 28°C and then imaged with UV light excitation using a Biorad ChemiDoc system with Image Lab software. All strains were grown and imaged on the same plate.

Isolation and characterization of motile suppressors of ΔexoR

Motile suppressor mutants of ΔexoRΔexoA were isolated as described previously (Heckel et al., 2014). Briefly, the nonmotile ΔexoRΔexoA mutant strain was mutagenized with the mariner transposon (Himar1) and inoculated into 0.3% Bacto Agar dissolved in ATGN media. After extended incubation, motile suppressor mutants were harvested from the edge of the swim ring. To identify suppressors with mutations in addition to those in the chvG-chvI locus, which restore both motility and biofilm formation to the ΔexoR mutant strain, motile suppressor mutants were screened for biofilm formation focusing on those that were biofilm-proficient. We used touchdown PCR and sequencing to identify the site of transposon insertion for these mutants as well as whole genome sequencing to identify point mutations in these genomes.

Whole genome sequencing

Genomic DNA was prepared from 1 mL culture grown in ATGN medium to mid-log phase. Genomic DNA of the ΔexoAΔexoR parent and four suppressor mutants was prepared and used to create libraries for sequencing. Libraries were prepared using a Bio Scientific NEXTflex™ Rapid DNA sequencing kit according to manufacturer instructions. Sequencing was performed with an Illumina Nextseq instrument with subsequent analysis from the IU CGB. Mutations were identified by mapping reads against the A. tumefaciens C58 reference genome (Genbank accession numbers AE008687, AE008688, AE008689, and AE008690 for the circular, linear, At, and Ti plasmids, respectively) using the breseq computational pipeline (Deatherage & Barrick, 2014).

Expression and purification of proteins

The rem allele was amplified from a pre-existing construct (primer sequences in Table S4) and ligated into pTYB12 (NEB IMPACT system of protein purification), resulting in an N-terminal intein fusion to Rem. The expression plasmid was transformed into E. coli Top10 F’ for sequence verification and collection and then transformed into E. coli BL21/λDE3 for expression and purification. Expression and purification of the intein-tagged Rem protein was performed according to the protocol described in the NEB IMPACT™ Kit. One liter of cells was grown at 37°C on an orbital shaker to an OD₆₀₀ 0.5 then induced with IPTG to a final concentration of 400 μM. Cells were induced overnight on an orbital shaker at 16°C. Cells were collected by centrifugation at 5,000 x g for 15 min at 4°C, and the pellet was re-suspended in column buffer (20 mM Tris-HCl pH 8.5, 500 mM NaCl, 1 mM EDTA). Cells were lysed by passage through an M-110L Microfluidizer Processor (Microfluidics, Westwood,MA). Cell debris was removed by centrifugation at 15,000 x g for 30 min at 4°C and the clarified extract was loaded onto a chitin resin (NEB) column pre-equilibrated with column buffer for purification at RT. After washing with 20 bed volumes of column buffer, auto-cleavage of the intein tag was induced by adding 3 bed volumes of cleavage buffer (column buffer with 50 mM DTT) to the column and incubating at RT for 40 h. Un-tagged Rem protein was eluted with column buffer and pooled fractions were dialyzed into a storage buffer (10 mM Tris-HCl, pH 8.0, 250 mM NaCl, 1 mM DTT, 10%glycerol). Aliquots were flash frozen and stored at −80°C.

The chvI and mirA coding sequences (wild type and mutant alleles) were amplified from cloning constructs (primer sequences in Table S4) and ligated into pET-28a(+) (Novagen/ EMB Biosciences/ Millipore). Protein-expression plasmids were transformed into E. coli Top10 F’ for collection and sequence verification and then transformed into E. coli BL21/λDE3 for expression and protein purification. One liter of cells were grown in LB at 37°C on an orbital shaker to an OD₆₀₀ of 0.6, then induced with 500 μM IPTG for 4 hours at 37°C. Cells were collected by centrifugation at 4°C at 5,000 x g for 15 min. Cell pellets containing ChvI were resuspended in lysis buffer (50 mM NaH₂PO₄ pH 6.5, 300mM NaCl, 10 mM imidazole) with 1 mM PMSF (phenylmethanesulfonyl fluoride). Cells were lysed by passage through an M-110L Microfluidizer Processor (Microfluidics). Cell debris was removed by centrifugation at 15,000 x g for 30 min at 4°C and pellets were resuspended in storage buffer (50 mM NaPO₄ pH 6.5, 500 mM NaCl). Purification was performed on 1 ml HisTrap HP (GE Healthcare) column using an FPLC system (ÄKTA P-920 pump and Superdex 75 10/300 GL column, GE Healthcare) and eluted using an imidazole gradient. Protein-containing fractions were pooled for desalting through a 5 mL HiTrap Desalting column (GE Healthcare). Desalted protein was dialyzed into a storage buffer (50 mM NaH₂PO₄ pH 8.0, 500 mM NaCl, 20% glycerol). Protein concentration was calculated following absorbance measurement by Nanodrop with the molar extinction coefficient for each protein.

Cell pellets from cells expressing MirA-His₆ were thawed and suspended in 20 mM NaH₂PO₄, 150 mM NaCl, pH 7.4 buffer with 1 mM PMSF and cells were disrupted by passage through an EmulsiFlex-C3 emulsifier, and clarified by centrifugation. Clarified lysates were loaded onto a 1 mL TALON (Takara) column, washed with 20 mM NaH₂PO_{4 ,}150 mM NaCl, pH 7.4 buffer and eluted with 20 mM NaH₂PO₄, 150 mM NaCl, 5% glycerol, pH 7.4 buffer with 100–600 mM imidazole step elution fractions. Eluted fractions were pooled and dialyzed into 20 mM NaH₂PO₄, 150 mM NaCl, 5% glycerol, pH 7.4 buffer and frozen at −20 °C. Protein concentration was determined by Pierce BCA assay as molar absorptivity measurements with MirA-His₆ were clearly erroneous.

Electrophoretic mobility shift assay

Upstream DNA fragments suspected to have binding sites were amplified and purified with the E.Z.N.A. Cycle Pure kit (Omega Bio-Tek). 10 nM DNA fragments were incubated with purified protein in binding buffer (10 mM Tris-Cl pH 7.5, 1 mM EDTA, 0.1 mM DTT, 5% glycerol, and 0.05 mg/mL BSA) and incubated 20 min at RT. Reactions were run on 6% acrylamide (80:1 acrylamide: bisacrylamide) 0.5% TBE (45 mM Tris-borate, 1 mM EDTA pH 8.0) gels and electrophoresed for 2 h at 150 V. Prior to loading, gels were pre-run with 0.5% TBE at 75 V for 1 h. Gels were stained with a 10,000:1 dilution of SYBR Safe DNA gel stain (Invitrogen) 20 min in 0.5X TBE, rinsed with distilled water and imaged on a Bio-Rad ChemiDoc system using Image Lab software. For gel shift inhibition assays, purified MirA-His₆ protein was incubated with Rem protein for fixed incubation times at RT prior to the addition of DNA, followed by 15 min incubation at RT. Reactions were run on an 8% polyacrylamide gel prior to staining for 15 min with 1 μg/mL ethidium bromide and imaging on a Bio-Rad ChemiDoc.

Pull-down assay

Purified proteins were diluted to the appropriate concentration in 50 mM Tris-HCl, 100 mM NaCl, 0.1% Tween pH 7.5 buffer. Samples were incubated 30 min at RT and loaded onto a 10 μL aliquot of Talon (Takara) cobalt resin. Following a 30 min incubation, the resin was washed 3 times with 500 μL of the same buffer as the dilutions and eluted with 20 μL Laemlli loading buffer.

Western and farwestern blotting

For detection of MirA-FLAG₃ by Western blot, strains were grown to exponential phase, concentrated to an OD₆₀₀ of 10, and boiled 5 min prior to running on a 12.5% acrylamide (37.5:1 acrylamide:bisacrylamide) SDS-PAGE gel with 2,2,2 trichloroethanol (TCE). The TCE was UV-activated prior to transfer to a nitrocellulose membrane. Membrane transfer was confirmed through UV-imaging of membrane (BioRad ChemiDoc ImageLab software). The nitrocellulose membranes were incubated with 1:5,000 monoclonal ANTI-FLAG M2 antibody produced in mouse (Sigma-Aldrich) and Goat Anti-Mouse IgG (H+L)~HRP Conjugate (BioRad). Membranes were developed with SuperSignal® West Pico Luminol/Enhancer Solution (Thermo Scientific) and the HRP signal was detected using a Biorad ChemiDoc system using Image Lab software.

For farwestern analysis, purified protein was electrophoresed by SDS-PAGE in two identical gels (12.5% acrylamide, 37.5:1 acrylamide:bisacrylamide) and transferred to nitrocellulose membranes. Membranes were blocked with 5% nonfat dry milk in TBS 0.1% Tween for 1 hour. For MirA-His₆, gels contained TCE for detection of protein in the gels. The membranes were blotted either with milk containing purified 0.4 M MirA-His₆ or 20 mM NaH₂PO₄, 150 mM NaCl, pH 7.4 buffer alone. For farwestern analysis of His₆-ChvI^D52E, the proteins on the membrane were denatured by incubation in AC buffer (100 mM NaCl, 20 mM Tris, 0.5 mM EDTA, 10% glycerol, 0.1% Tween-20) with 6 M guanidine-HCl for 30 min at RT with mixing. Proteins were renatured by incubating at RT with shaking for 30 min in AC buffer with 3M guanidine-HCl, then 30 min at RT in AC buffer with 1 M guanidine-HCl, then 30 minutes at 4°C in AC buffer with 0.1 M guanidine-HCl, then overnight at 4°C in AC buffer. The membranes were blocked in blotto (5% milk solution in tris-buffered saline with 1% Tween-20) at RT 1 h with shaking. Purified His₆-ChvID^52E protein (5 μg) was used as prey and incubated with the membrane-immobilized renatured protein in protein binding buffer (AC buffer with 5% milk powder and 1 mM DTT) overnight at 4°C with shaking. Another membrane was processed in tandem in the same conditions with no His₆-ChvI^D52E. Membranes for both proteins were then incubated with a 1:2000 dilution of rabbit α-His antibody for 1 hour at RT, then with a 1:20,000 dilution of HRP-conjugated goat anti-rabbit antibody for 1 hour at RT.

Bacterial adenylate cyclase two-hybrid assays

Genes of interest were PCR-amplified and inserted into KpnI and EcoRI-digested pKNT25 or pUT18 with NEB Hifi Assembly to construct in-frame C-terminal fusions (Euromedex). Plasmids containing T18 and T25 fusions were co-transformed into BACTH test strain BTH101. Strains were grown overnight at 30 °C and 2 μL were spotted onto LB plates containing IPTG (500 μM) and 5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside (X-gal, 160 μg/mL) and incubated at 30 °C overnight. Colonies were imaged with a Nikon camera. All strains were grown and imaged on the same plate for each experiment.

Multiple Sequence Alignment

Amino acid sequences were retrieved from Kegg, Biocyc or JGI IMG and aligned using ClustalW with the default parameters (Gap open penalty = 10, Gap extension penalty = 0.5, BLOSUM weight matrix). Amino acid sequences of proteins can be found at NCBI reference numbers WP_006313038.1 (Rem/A. tumefaciens)_, CAC45250.1 (Rem/S. meliloti), SUB44985.1 (Rem/O. anthropi), WP_006698188.1 (Rem/R. pusense IRGB74), WP_065114859.1 (Rem/A. rhizogenes) WP_010970615.1 (ChvI/A. tumefaciens), WP_003531999.1 (ChvI/S. meliloti), AYM61105.1 (VirG/A. tumefaciens), NP_417864.1 (OmpR/E. coli), AKH06072.1 (PhoB/S. enterica), WP_003502959.1 (MirA/ A. tumefaciens), CAC48483.1 (MirA/ S. meliloti), WP_004442019.1 (MirA/R. pusense IRGB74), WP_065115793.1 (MirA/A. rhizogenes), and WP_040128199.1 (MirA/Ochrobactrum/Brucella anthropi).

Data availability:

High-throughput, Illumina-based RNASeq data has been deposited in the GEO database Accession No. GSE174467 ([dataset] Alakavuklar and Fuqua, 2021). All other data that support the findings of this study are available from the corresponding author upon reasonable request.