Regulation of bacterial surface attachment by a network of sensory transduction proteins

Bacteria are often attached to surfaces in natural ecosystems. A surface-associated lifestyle can have advantages, but shifts in the physiochemical state of the environment may result in conditions in which attachment has a negative fitness impact. Therefore, bacteria employ numerous mechanisms to control the transition from an unattached to a sessile state. The Caulobacter crescentus protein HfiA is a potent developmental inhibitor of the secreted polysaccharide adhesin known as the holdfast, which enables permanent attachment to surfaces. Multiple environmental cues influence expression of hfiA, but mechanisms of hfiA regulation remain largely undefined. Through a forward genetic selection, we have discovered a multi-gene network encoding a suite of two-component system (TCS) proteins and transcription factors that coordinately control hfiA transcription, holdfast development and surface adhesion. The hybrid HWE-family histidine kinase, SkaH, is central among these regulators and forms heteromeric complexes with the kinases, LovK and SpdS. The response regulator SpdR indirectly inhibits hfiA expression by activating two XRE-family transcription factors that directly bind the hfiA promoter to repress its transcription. This study provides evidence for a model in which a consortium of environmental sensors and transcriptional regulators integrate environmental cues at the hfiA promoter to control the attachment decision.


Author summary
Living on a surface within a community of cells confers a number of advantages to a bacterium. However, the transition from a free-living, planktonic state to a surface-attached lifestyle should be tightly regulated to ensure that cells avoid adhering to toxic or resource-limited niches. Many bacteria build adhesive structures on the surface of their cell envelopes that enable attachment. We sought to discover genes that control development of the Caulobacter crescentus surface adhesin known as the holdfast. Our studies uncovered a network of signal transduction proteins that coordinately control the biosynthesis of the holdfast by regulating transcription of the holdfast inhibitor, hfiA. We conclude that C. crescentus uses a multi-component regulatory system to sense and integrate

Introduction
In natural ecosystems, bacteria often live in surface attached, multicellular communities known as biofilms [1]. Biofilms enhance resistance to toxic compounds, promote sorption of nutrients, and facilitate exchange of genes and gene products [2]. In aquatic systems, metabolizable substrates accumulate at surfaces [3], and thus attachment can provide a fitness advantage [4]. However, it is important for cells to be able to disperse in cases when surfaces accumulate toxins, or become depleted of substrates or electron acceptors. It is therefore not surprising that surface attachment is a highly regulated process in bacteria. The alphaproteobacterium Caulobacter crescentus exhibits a dimorphic developmental cycle that results in two cell types after division: a flagellated swarmer cell and a sessile stalked cell [5] (Fig 1A). The stalked cell is the reproductive cell type and divides asymmetrically, giving birth to the non-reproductive swarmer cell [5]. Swarmer cells differentiate into stalked cells, whereupon they can secrete a unipolar polysaccharide adhesin known as the holdfast [5][6][7][8] (Fig 1A). Holdfast-mediated adherence to solid surfaces is permanent [9][10][11]. The holdfast enables cell attachment to a variety of chemically diverse materials [12], and also facilitates partitioning of cells to the air-liquid interface in aqueous media [13]. C. crescentus tightly regulates holdfast synthesis, which helps to ensure that cells can take advantage of resources in particular niches, while not becoming restricted to sub-optimal environments.
The small protein HfiA is a potent inhibitor of C. crescentus holdfast synthesis [14]. HfiA directly targets HfsJ, a glycolipid glycosyltransferase required to build the holdfast [14]. The promoter of hfiA can integrate a number of regulatory cues that modulate its transcription and thereby affect holdfast production ( Fig 1B). Multiple cell cycle regulators bind and regulate the hfiA promoter [14]. In addition, flagellum assembly indirectly affects hfiA transcription by an unknown mechanism [15]. Environmental signals such as nutrient quality and sensory transduction proteins including the two-component system (TCS) proteins LovK-LovR [14] and MrrA [16] influence hfiA transcription by unknown and independent mechanisms. CleA, a CheY-like response regulator that tunes flagellar rotation in response to cyclic-di-GMP binding, also regulates holdfast synthesis, though it is not known if CleA-mediated holdfast regulation involves HfiA [17]. In an effort to better understand the complex regulatory network that controls expression of the holdfast inhibitor, HfiA, we have investigated the molecular basis by which the LovK-LovR TCS controls hfiA transcription.
TCS are commonly used by bacteria to sense environmental signals and regulate adaptive changes in cell physiology [18]. Archetypal TCS consist of a sensor histidine kinase (HK) and a DNA-binding response regulator (RR), which generally form simple linear pathways that couple environmental perturbation to a change in gene expression. There is good experimental support for the paradigm that TCS form insulated signaling systems, with little cross-regulation between non-cognate kinases and responses regulators [19][20][21][22], though there are examples of branched TCS pathways in which sensor kinases signal to more than one regulator, or in which signals from multiple kinases converge on a single regulator (reviewed in [23,24]). Complex cellular processes including sporulation [25], stress responses [16,[26][27][28][29], biofilm formation [30], and cell cycle [31,32] are often controlled by multiple TCS in branched pathways. In this study, we have defined a regulatory system composed of multiple TCS proteins and transcription factors that regulates holdfast development in C. crescentus.
We have previously shown that the TCS LovK-LovR regulates holdfast development and surface adhesion by influencing the transcription of hfiA [14,33]. LovK is a soluble, cytoplasmic photosensory HK that contains a LOV (light, oxygen, voltage) sensor domain [34] Nterminal to a HWE/HisKA_2 family [35,36] kinase domain. The LOV domain binds a flavin mononucleotide (FMN) cofactor [37], which confers the ability to sense blue light and the redox state of the environment [33,37]. LovK is encoded adjacent to LovR, a single domain response regulator (SDRR). The regulation of hfiA transcription by LovK-LovR must be indirect as LovR has no DNA-binding domain. We developed a genetic selection aimed at identifying genes that function together with or downstream of LovK-LovR to control hfiA transcription. We found that the hybrid histidine kinase SkaH and the canonical TCS SpdS-SpdR function with LovK-LovR to control transcription of hfiA. LovK and SpdS form heteromeric complexes with SkaH in vitro and in vivo, providing evidence for cross-regulation between these different TCS. We further demonstrate that the response regulator SpdR indirectly controls hfiA transcription by activating expression of two direct transcriptional repressors of hfiA, RtrA and RtrB. Our data provide evidence for a complex regulatory system that has the capacity to regulate hfiA expression and surface attachment in response to multiple environmental signals.

A lovK(H180A) mutant phenocopies a lovK-lovR overexpression strain
Coordinate overexpression of lovK-lovR increases the probability that any single cell will develop a holdfast, and therefore enhances adhesion of a cell population to surfaces [14,33]. This phenotype is due to the repressive effect of lovK-lovR expression on hfiA transcription [14]. During the course of our studies, we discovered that mutating the conserved histidine phosphorylation site of LovK to alanine-lovK(H180A)-resulted in decreased hfiA transcription, and a concomitant increase in surface attachment and holdfast development (Fig 1C). In other words, a strain in which the chromosomal copy of lovK is replaced with lovK(H180A) phenocopies a lovK-lovR overexpression strain. To further investigate the role of the LovK-LovR TCS proteins, we expressed either the wild-type lovK allele or the lovK(H180A) allele from a plasmid in a ΔlovK-lovR background (S1A Fig). In the absence of lovR, expression of wild-type lovK had no effect on surface attachment. On the other hand, expression of lovK (H180A) increased surface adhesion (S1A Fig). These results indicate that expressing the unphosphorylatable lovK(H180A) allele is sufficient to enhance surface attachment, and that lovR is not required for this phenotype. We further tested whether hyperadhesion in the lovK (H180A) strain was restored to wild-type levels upon ectopic expression of the wild-type lovK allele. Ectopic expression of wild-type lovK in a strain carrying lovK(H180A) at the chromosomal locus reduced cell surface attachment to wild-type levels (S1B Fig). In the converse experiment, ectopic expression of lovK(H180A) in a wild-type background did not result in a hyperadhesive phenotype (S1B Fig). We conclude that wild-type lovK is dominant over lovK (H180A).
We chose to use the strain with the lovK(H180A) allele expressed from its native chromosomal locus to investigate the molecular connection between the LovK sensor kinase and hfiA transcription as it 1) provided a cleaner and more stable genetic background than the lovK-lovR plasmid overexpression strain, 2) expressed lovK(H180A) from the native lovK promoter, and 3) did not require the addition of metabolizable inducers.

A forward genetic selection identifies genes that function with lovK-lovR to repress hfiA expression
The SDRR LovR lacks a DNA-binding output domain, and thus transcriptional regulation of hfiA by LovK-LovR is likely indirect. Moreover, LovR is not required to regulate hfiA transcription when LovK lacks its histidine phosphorylation site. We hypothesized that additional regulator(s) function either downstream or in concert with LovK-LovR to repress hfiA transcription. To identify such regulator(s), we designed a forward genetic selection to isolate mutants in which the repressive regulatory connection between LovK and hfiA was disrupted (Fig 2A). Specifically, we fused the hfiA promoter (P hfiA ) to the chloramphenicol acetyltransferase (cat) gene and transformed the lovK(H180A) strain with this reporter plasmid (P hfiA -CAT). We then used growth of this strain in the presence of chloramphenicol as a proxy for P hfiA activity. By serially passaging this reporter strain, we selected for spontaneous mutants with enhanced growth in the presence of chloramphenicol (Fig 2A). We predicted that this approach would enable us to identify mutants in which the genetic connection between lovK and hfiA transcription was disrupted.
After several passages, we isolated dozens of mutants that grew faster than the parent strain. In these isolates, we first sequenced the P hfiA -CAT promoter fragment to identify clones with cis mutations in the hfiA promoter itself that increase promoter activity. In mutants with an intact P hfiA -CAT promoter fragment, we assessed P hfiA activity in two independent secondary screens: we measured activity of a second hfiA promoter fusion (P hfiA -lacZ), and we assessed activity of the native hfiA promoter by measuring the surface adhesion phenotype of these mutants. Two classes of mutants emerged from this selection: 1) strains that grew faster in chloramphenicol as a result of P hfiA -CAT promoter mutations, but that displayed unaltered surface adhesion and no change in P hfiA -lacZ activity compared to the parental lovK(H180A) strain (S2 Fig), and 2) mutants with an intact P hfiA -CAT reporter plasmid that displayed fast growth in chloramphenicol, reduced surface adhesion, and increased β-galactosidase activity from the P hfiA -lacZ plasmid (S1 Table). We mapped polymorphisms in nine spontaneous mutants from this latter class, derived from three independent selections, by whole genome sequencing.
Two mutants derived from the same flask shared a~10kb deletion that included gene CC_3560 (CCNA_03675), annotated as a hybrid histidine kinase (Fig 2B and S1 Table). A single mutant had a 19bp deletion that resulted in a frameshift in the transmembrane histidine kinase gene spdS (CC_0248) (Fig 2B and S1 Table). The remaining six mutants had lesions in either tipR or both acrB2 and tipR (Fig 2B and S1 Table). TipR is a repressor of the acrAB2--nodT operon, which encodes the components of the AcrAB2-NodT antibiotic efflux pump [38]. As several of our mutants harbored lesions in tipR, we hypothesized that enhanced growth in chloramphenicol in these strains was due to de-repressed expression of the efflux pump. Deletion of tipR in either wild-type or the lovK(H180A) background increased growth rate in the presence of chloramphenicol (S3 Fig). We concluded that disruption of tipR is a general genetic solution to combat chloramphenicol toxicity, and that these mutants were Genetic selection identifies regulators that function with lovK(H180A) to repress hfiA. A) A strategy to select for spontaneous mutants that disrupt lovK(H180A) repression of the hfiA promoter. The P hfiA -CAT transcriptional fusion allows for selection of mutants with increased hfiA promoter activity. Specifically, cultivation in M2X medium supplemented with chloramphenicol allows for selection of enhanced expression of chloramphenicol acetyltransferase (CAT) from P hfiA -CAT. B) Mutations in three genomic regions identified in the selection with specific lesions marked below the genetic loci. Lesions at the acr locus confer increased chloramphenicol resistance irrespective of lovK (S3 Fig); the mutant alleles identified at this locus are not indicated in the diagram but are listed in S1 Table. C) Predicted domain organization of the TCS proteins SkaH, SpdS and SpdR. PAS-sensory domain, HWE confounding the efficiency of our selection. We therefore modified our selection strategy and ectopically expressed a second copy of tipR from its native promoter in an effort to minimize enrichment of tipR mutants. Using this strategy, we identified nine mutants from three independent selections with a point mutation resulting in a CC_3560(H82R) allele. Altogether, our results provided evidence that the sensor kinases CC_3560 and SpdS play a role in hfiA repression by LovK-LovR. Though we likely did not saturate our selection, we decided to move forward with our initial hits of the TCS proteins CC_3560 and SpdS-SpdR.

CC_3560 (skaH), spdR and spdS regulate hfiA transcription, holdfast development, and adhesion downstream of lovK
Our selection identified mutations in two sensor kinase genes: CC_3560 and spdS. CC_3560 encodes a cytoplasmic hybrid histidine kinase ( Fig 2C) that, like LovK, belongs to the noncanonical HWE/HisKA_2 family of histidine kinases [36]. We have named this kinase skaH (sensor kinase associated hybrid) for reasons we describe in a later section. spdS is the transmembrane sensor kinase component of the SpdS-SpdR TCS, which has been reported to regulate genes involved in stationary phase adaptation in C. crescentus [39,40] (Fig 2C). Homologs of SpdS-SpdR in other species are more commonly known as RegB-RegA, and have been reported to detect changes in cellular redox state, and regulate a variety of energy generating and energy consuming processes (Reviewed in [41]).
To validate the results of our genetic selection, we tested whether the TCS genes skaH and spdS are involved in regulation of hfiA transcription, holdfast development and surface adhesion. Though mutations in spdR were not identified in our selection, we also tested its role in hfiA regulation and adhesion since it is the cognate RR of SpdS. We first generated in-frame deletions of skaH, spdS or spdR in the lovK(H180A) genetic background to determine if any of these genes were required for LovK(H180A)-dependent control of hfiA transcription, holdfast development and surface adhesion. Repression of hfiA transcription by lovK(H180A) required skaH, spdR and spdS ( Fig 3A). Similarly, skaH, spdR and spdS are required for the increase in holdfast development and surface adhesion by lovK(H180A) (Fig 3C and Fig 3D). We note that the surface adhesion phenotype of the lovK(H180A)ΔspdS strain varied depending on the growth phase of the culture. At early time points, surface adhesion of lovK(H180A)ΔspdS cultures is lower than lovK(H180A) cultures. When cultures begin to enter stationary phase, we observed a sharp increase in surface adhesion in the lovK(H180A)ΔspdS cultures that eventually matched the levels we observed in lovK(H180A) (Fig 3D). Importantly, the skaH, spdR and spdS deletion phenotypes we observed in the lovK(H180A) background were complemented by ectopic expression of these deleted genes from a xylose-inducible promoter (S4 Fig). Together, these data provide evidence that skaH and spdR are required for lovK(H180A)dependent regulation of hfiA transcription, holdfast development and surface adhesion. lovK (H180A)-dependent regulation of hfiA transcription and adhesion requires spdS at low culture density.
We next asked whether this set of TCS genes affect hfiA transcription, holdfast development and surface adhesion in wild-type cells. Activity from the P hfiA -lacZ transcriptional reporter in ΔskaH, ΔspdR and ΔspdS was indistinguishable from wild type (Fig 3B), though transcription of hfiA is high under these conditions and additional increases are difficult to measure. Deletion of spdR (but not skaH or spdS) resulted in decreased holdfast development compared to and HisKA-histidine kinase domains, REC-receiver domain, TM-transmembrane, HTH-helix-turn-helix DNAbinding domain. https://doi.org/10.1371/journal.pgen.1008022.g002 Environmental regulators of Caulobacter holdfast development wild type (Fig 3C). We further measured a reproducible and statistically significant decrease in adhesion in ΔskaH and ΔspdR cultures (S5 Fig), though we note that baseline surface adhesion in wild-type cultures is low under these assay conditions and additional decreases are difficult to measure (Fig 3D). Deletion of spdS in a wild-type background resulted in a similar adhesion phenotype to the lovK(H180A)ΔspdS strain, where we observed a marked increase in surface attachment relative to wild type as the culture entered stationary phase (Fig 3D). To better understand this ΔspdS hyperadhesive phenotype during stationary phase, we first tested whether the phenotype requires holdfast production. ΔspdS hyperadhesion requires the glycosyltransferase gene hfsJ, which is necessary for holdfast synthesis (S6A Fig). We next tested if

Conserved SkaH and SpdR phosphorylation sites are required for hfiA transcriptional regulation
We next addressed the role of the conserved skaH, spdS, and spdR phosphorylation sites in regulation of hfiA transcription by lovK(H180A). The hybrid histidine kinase SkaH has a histidine phosphorylation site in its kinase domain and an aspartyl phosphorylation site in its receiver (REC) domain. We individually mutated these residues to alanine (resulting in H285A and D550A alleles, respectively) in a lovK(H180A) background. The skaH(H285A) mutation had no effect on lovK(H180A) repression of hfiA transcription ( Fig 3A). The skaH(D550A) mutation ablated hfiA repression by lovK(H180A) ( Fig 3A) and ablated its hyperadhesive phenotype (S7 Fig The transcription and adhesion phenotypes of spdS(H248A) were highly variable from day to day for reasons we do not yet understand, and thus we cannot presently draw any conclusions about this strain. Overall, we conclude that intact phosphorylation sites on the skaH and spdR REC domains are required for lovK(H180A) dependent regulation of hfiA transcription and surface adhesion.

SkaH directly interacts with LovK and SpdS in vitro and in vivo
To further explore the molecular basis of the genetic interactions we observed between TCS genes, we tested whether the corresponding TCS proteins physically interact. We first performed a bacterial two-hybrid (B2H) assay in a heterologous E. coli system [42], fusing the LovK, SkaH or SpdS kinases to the C-terminus of the T25 and T18c split domain fragments of adenylate cyclase. The transmembrane domain was not included in the SpdS construct in these experiments, which contains only the cytoplasmic kinase domain (hereafter named SpdS � ). Sensor histidine kinases generally form stable homodimers [43][44][45]. Co-expression of the same kinase fused to the T25 and T18c fragments resulted in blue colonies on X-gal medium (Fig 4A), indicative of a homomeric histidine kinase interaction that reconstitutes the split adenylate cyclase. Co-expression of the T25-SkaH fusion with either a T18c-LovK or T18c-SpdS � fusion also resulted in blue colonies ( Fig 4A) indicative of heteromeric kinase interactions. B2H reporter strains co-expressing T25-LovK and T18c-SpdS � constructs yielded white colonies (Fig 4A). In addition, we tested whether the conserved phosphorylation sites of SkaH, LovK and SpdS were required for the heteromeric kinase interactions we observed in the B2H assay (S8 Fig). Specifically, we assayed a number of combinations where we mutated the conserved phosphorylation site to an alanine in either one or both of the kinase fusions. All combinations of SkaH alleles co-expressed with either allele of LovK or SpdS � resulted in a positive interaction (S8 Fig). These data provide evidence that SkaH can directly interact with both LovK and SpdS to form a heteromeric kinase complex, and that these interactions do not require the conserved protein phosphorylation sites. We further conclude that LovK and SpdS do not interact in a B2H assay.
SkaH is a hybrid histidine kinase, so we next tested whether the heteromeric B2H interactions we observed required the REC domain of SkaH. We fused a short version of SkaH that lacked the REC domain (SkaH ΔREC ) to the T25 and T18c fragments of adenylate cyclase. Coexpression of SkaH ΔREC fusions to both T25 and T18c resulted in blue colonies when grown on agar plates containing X-gal ( Fig 4A); co-expression of T25-SkaH ΔREC and T18c-SkaH also yielded blue colonies ( Fig 4A) providing evidence that the homomeric interaction between SkaH monomers does not require the REC domain. A B2H reporter strain co-expressing T25-SkaH ΔREC and T18c-LovK yielded white colonies ( Fig 4A). Co-expressing T25-SkaH ΔREC and T18c-SpdS � yielded light blue colonies, indicating the interaction was qualitatively weaker than the observed B2H interaction between full-length SkaH and SpdS � (Fig 4A). These results provide evidence that the REC domain of SkaH is necessary for the heteromeric interaction with LovK observed in a B2H assay. The SkaH REC domain is not strictly required, but apparently contributes to the B2H interaction we observe between SkaH and SpdS � .
To test if the SkaH-LovK and SkaH-SpdS interactions observed in the heterologous B2H system occur in vivo, we tested whether LovK and/or SpdS co-purify with SkaH when affinity purified from C. crescentus lysate. Specifically, we ectopically expressed different combinations of histidine kinase fusions in C. crescentus: a maltose binding protein (MBP) N-terminally fused to SkaH (MBP-SkaH), a HA epitope tag N-terminally fused to LovK (HA-LovK), and a 3xFLAG tag C-terminally fused to full length SpdS (SpdS-3xFLAG). After co-expression in C. crescentus, cells were lysed and MBP-SkaH was immobilized on an amylose affinity resin in an effort to pull down potential heteromeric kinase complexes. Co-purifying proteins were detected by Western blot. When MBP-SkaH was co-expressed with either HA-LovK or SpdS-3xFLAG, both fusions co-eluted with MBP-SkaH after a stringent wash and elution with maltose ( Fig 4B and 4C). As a negative control, we co-expressed MBP alone with either HA-LovK or SpdS-3xFLAG. Neither HA-LovK nor SpdS-3xFLAG co-eluted with MBP ( Fig 4B and 4C), indicating that the interactions observed were not due to protein interaction with MBP or non-specific interactions with the resin. From these data, we conclude that SkaH physically interacts with both LovK and SpdS in C. crescentus.

spdR regulates hfiA transcription indirectly
A major goal of our genetic selection was to identify direct regulators of hfiA transcription downstream of LovK. The DNA-binding response regulator, SpdR, emerged as a clear candidate from our data. We utilized genetic, molecular and genomic approaches to test the role of SpdR as a regulator of hfiA transcription. As described above, alanine substitution of the conserved phosphorylation site in the SpdR REC domain (D64A), phenocopied the ΔspdR deletion mutation in a lovK(H180A) background ( Fig 3A). Moreover, a classic phosphomimetic substitution of aspartate to glutamate (D64E) in a wild-type background decreased activity from the P hfiA -lacZ transcriptional reporter, and increased surface adhesion and holdfast

SkaH physically interacts with LovK and SpdS in vitro and in vivo. A)
Bacterial two-hybrid (BTH) experiments to assess interactions between histidine kinase fusions to a split adenylate cyclase. Fusions with spdS lack the transmembrane domain, notated spdS � . skaH ΔREC is a truncated version of SkaH that lacks the REC domain. Two biological replicates are shown for each co-expression combination. Protein-protein interaction between fusions reconstitutes the split adenylate cyclase encoded on pKT25 and pUT18c, and results in a blue color on agar plates containing x-gal. Strains expressing fusions that do not interact appear white. Zip = positive control. Histidine kinase protein-protein interaction models are schematized below each BTH assay as 1:1 interactions. However, the actual oligomeric state of detected interactions is not known. B and C) Co-purification experiments in which MBP is affinity purified with amylose resin and co-purifying kinases are detected by Western blot. The fusion proteins co-expressed in each experiment are schematized in the cells above each blot. Representative immunoblot against the HA and FLAG epitope tags are shown in B and C respectively. Input = clarified lysate, FT = Flow through. https://doi.org/10.1371/journal.pgen.1008022.g004 Environmental regulators of Caulobacter holdfast development development ( Fig 5A). These data provide genetic evidence that SpdR can function to repress hfiA transcription and modulate holdfast production and cell surface adhesion in a phosphorylation dependent manner.
To test whether SpdR is a direct regulator of hfiA, we performed an electrophoretic mobility shift assay (EMSA) using a fluorescently labeled hfiA promoter probe and His 6 -SpdR(D64E) recombinant protein. Increasing concentrations of His 6 -SpdR(D64E) did not retard gel mobility of the hfiA probe ( Fig 5B). As a positive control, we performed an EMSA using a fluorescently labeled probe corresponding to the promoter of cspD, a gene reported to be directly regulated by SpdR [39]. Migration of the labeled cspD promoter probe was retarded by Environmental regulators of Caulobacter holdfast development increasing concentrations of His 6 -SpdR(D64E) and this probe displayed the same double shift as previously published (Fig 5B) [39]. These results provide evidence that SpdR does not bind the hfiA promoter, and therefore support a model in which SpdR indirectly regulates hfiA transcription.

RNA sequencing analysis defines the SpdR regulon
Given the evidence that SpdR regulates hfiA transcription indirectly, we experimentally defined the transcriptional regulon of SpdR with the goal of uncovering direct hfiA regulators that are expressed downstream of SpdR. We measured steady-state transcript levels in a strain missing spdR (ΔspdR) and a strain expressing the active phosphomimetic spdR(D64E) allele by RNA sequencing (RNA-seq). The majority of genes with differential transcriptional profiles are more highly expressed in cells with the constitutively active spdR(D64E) allele (Fig 5C and  S2 Table). Importantly, hfiA is among the small set of transcripts that are downregulated in the spdR(D64E) strain relative to the ΔspdR strain (5.1-fold, p < e -16 ) (Fig 5C and S2 Table). A consensus binding motif for SpdR has been defined in C. crescentus [39,40] and for SpdR homologs in other species [46][47][48]. To predict direct targets of SpdR, we utilized the MEME motif discovery platform [49] to identify shared sequence motifs in the promoter regions (-150 to +50) of all the genes with at least 2-fold expression differences between spdR(D64E) and ΔspdR strains. This approach revealed a predicted SpdR binding site (Fig 5D) in the promoter regions of 82 genes (S2 Table), suggesting that these genes are directly regulated by SpdR. A scan of the region from -500 bp to +100 bp around the start hfiA codon failed to identify a SpdR binding sequence. These data provide additional evidence that SpdR is not a direct repressor of hfiA transcription.

An approach to identify direct hfiA transcriptional regulators downstream of SpdR
We hypothesized that transcription factor(s) regulated downstream of SpdR were responsible for repression of hfiA transcription. We took a candidate approach and examined all of the differentially expressed genes (>2-fold, P < 0.01) in our RNA-Seq dataset that contained a predicted DNA-binding domain. The six predicted transcription factors in this set were activated by SpdR (S2 Table). Thus, we reasoned that the repressive effect of SpdR on hfiA transcription was the result of activation of one or more transcription factors that directly repress hfiA ( Fig  6A). We individually overexpressed all candidate transcription factors and measured hfiA promoter activity and surface attachment. Two Xenobiotic Response Element (XRE)-family transcription factors, CC_3164 (CCNA_03267) and CC_2330 (CCNA_02415), displayed the overexpression phenotypes expected for an hfiA repressor: a decrease in activity of the hfiA promoter (Fig 6B), increase in holdfast development ( Fig 6C) and an increase in surface adhesion (S9A Fig), compared to wild-type. Importantly, the promoter regions of rtrA and rtrB each contain a predicted SpdR binding site (S2 Table), providing evidence that transcription of these genes is directly regulated by SpdR. Since CC_3164 and CC_2330 are in the SpdR regulon and regulate hfiA promoter activity, we named these genes rtrA and rtrB, for RegBA transcriptional regulator A and B, respectively.
Overexpression of each transcription factor alone provides evidence that rtrB is a less potent hfiA repressor than rtrA. This is congruent with our observation that rtrB overexpression does not influence holdfast development to the same extent as rtrA overexpression (Fig 6B and Fig  6C). Coordinate overexpression of rtrA and rtrB did not decrease hfiA expression or increase holdfast development more than lovK(H180A) or a strain overexpressing rtrA alone (Fig 6B  and Fig 6C). Strains overexpressing either one or both rtrA and rtrB had the same bulk surface adhesion phenotype as lovK(H180A) (S9A Fig). These data indicate that rtrA and rtrB regulate hfiA transcription, holdfast development, and surface adhesion.
Finally, we measured the effect of deleting rtrA and rtrB on surface adhesion in wild-type and in spdR(D64E) backgrounds (S9B Fig). In wild-type cells, deletion of rtrB but not rtrA resulted in a modest decrease in surface adhesion; the ΔrtrAΔrtrB adhesion phenotype was similar to ΔrtrB (S9B Fig). We observed similar results in a spdR(D64E) background. We conclude that rtrB has a larger effect on surface adhesion than rtrA under these assay conditions.

RtrA and RtrB are direct regulators of hfiA transcription
Having demonstrated that rtrA and rtrB repress hfiA expression and enhance holdfast development, we next tested whether RtrA and RtrB proteins directly bind the hfiA promoter. Increasing concentrations of either protein retarded mobility of a labeled hfiA promoter probe relative to the probe alone in an EMSA assay (Fig 6D & 6E). To assess the specificity of this interaction, we added 10-fold excess of unlabeled specific probe or unlabeled non-specific probe. The unlabeled specific probe competed with the labeled probe for binding to RtrA and Environmental regulators of Caulobacter holdfast development RtrB, while unlabeled non-specific probe had little effect on binding of RtrA or RtrB to the hfiA promoter region (Fig 6D & 6E). These data support a model in which RtrA and RtrB directly repress hfiA transcription.

Discussion
Adherence of a C. crescentus cell to a surface via its holdfast is effectively permanent. As such, it is not surprising that holdfast synthesis is a highly regulated process. hfiA is a potent inhibitor of holdfast synthesis, and its transcription is controlled by a suite of cell cycle and environmental regulatory proteins [14][15][16]. In this study, we sought to understand the molecular mechanism by which the TCS, LovK-LovR, controls hfiA transcription. A forward genetic selection followed by candidate gene studies led to the discovery of a network of TCS proteins and transcription factors that coordinately function with LovK to regulate hfiA transcription. Specifically, we have shown that the hybrid histidine kinase, SkaH, forms heteromeric complexes with the LovK kinase and the SpdS kinase in vivo. This suite of kinases works through the DNA-binding response regulator, SpdR, to activate expression of two transcription factors, RtrA and RtrB, which directly bind the hfiA promoter and repress its transcription. Our data provide evidence for a multi-component sensory system (Fig 7) that has the capacity to detect, and perhaps integrate, multiple signals to regulate holdfast development.

Signaling through the LovK-LovR system
C. crescentus lovK-lovR overexpression modulates cell adhesion by influencing hfiA transcription [14], but it also indirectly represses dozens of genes in the General Stress Response (GSR) regulon [50] through an independent molecular mechanism [14]. The mechanism by which the SDRR LovR functions with LovK to regulate hfiA and the GSR remains unclear, but previous genetic data have suggested that its function may be to simply dephosphorylate LovK [33,50]. This model is based on the result that overexpression of either lovK or lovR alone has no effect on cell-to-cell adhesion [33] or on repression of GSR transcription [50], while coordinate overexpression represses GSR transcription [50], enhances cell-to-cell [33] and surface adhesion [14], and represses hfiA transcription [14]. These data support a model in which LovK is active in its unphosphorylated state. Similarly, our analysis of hfiA transcription and surface attachment in a strain that solely expresses the non-phosphorylatable lovK(H180A) allele supports such a model (S1 Fig).
What, then, occurs at the molecular level that explains these phenotypes? Unphosphorylated-LovK may function as an allosteric regulator of other kinases or kinase complexes that control SpdR phosphorylation. LovK may also dephosphorylate a receiver protein or receiver domain required for SpdR phosphorylation. There is evidence that LovK can function as a dephosphorylase in the case of the SDRR MrrA, which was recently proposed to serve as a signaling hub that controls GSR transcription. MrrA also enhances hfiA expression in a manner that requires lovK, but not lovR [16]. These data suggest MrrA is an additional TCS protein that functions upstream of LovK in the network of TCS proteins that modulate hfiA transcription. We hypothesize that non-phosphorylatable LovK(H180A) leads to high levels of phosphorylated SpdR, which results in repression of hfiA transcription. Given the fact that SkaH is required for hfiA repression by lovK(H180A), we do not favor a model in which LovK directly phosphorylates or dephosphorylates SpdR.

Interplay between HisKA and HWE/HisKA_2 families of histidine kinases
SkaH and LovK belong to the HWE/HisKA_2 histidine kinase family, which is defined by unique primary structure features in the dimerization, histidine phosphotransfer (DHp), and catalytic (CA) domains [35,36]. HWE/HisKA_2 proteins are reported to function as part of complex signal transduction systems, including the GSR in Alphaproteobacteria [16,26] and cyst development in Rhodospirillum centenum [51,52], and have been reported in unusual oligomeric forms including monomer [53] and hexamer [54]. Though SpdS is a member of the more common HisKA family of HKs, its homolog in Rhodobacter capsulatus (RegB) can form tetramers mediated by an intermolecular disulfide bond that forms through a conserved cysteine [55]. It is not known if C. crescentus SpdS forms tetramers, but we predict that it may since it contains the conserved cysteine (C248) present in R. capsulatus RegB. It is therefore possible that HWE/HisKA_2 kinases (SkaH and LovK) and a HisKA kinase (SpdS) form unusual, high-order oligomers in the cell, and that changes in the oligomeric states of these proteins affect their function as regulators of hfiA and GSR transcription. Though we have presented clear evidence for interactions between SkaH, LovK, and SpdS, the structural basis of these interactions remains undefined. It is important to consider the possibility that the heteromeric HK interactions we observe under our assay conditions do not reflect the full extent of kinase interactions in the cell. Specifically, it is possible that the LovK-SkaH-SpdS complex Environmental regulators of Caulobacter holdfast development is part of an even higher order complex of regulatory proteins, as the case for the �2 MDa stressosome complex that controls the σ B -dependent general stress response (GSR) in Bacillus subtilis [56,57].
Finally, our data provide evidence for an emerging model in which HisKA and HWE/His-KA_2-family HKs can function together to regulate gene expression. A previous study showed that the SDRR, MrrA, can be phosphorylated and/or dephosphorylated by kinases from both of these families of HKs in C. crescentus [16]. Furthermore, deletion of all the HWE/HisKA_2 HKs in Sphingomonas melonis Fr1 does not completely ablate GSR transcription [26], suggesting that one or more HisKA kinases also play a role in regulating this system. This study provides additional evidence for complex signal transduction systems involving both HisKA and HWE/HisKA_2 histidine kinase families.

A consortium of sensors regulate hfiA to control cell adhesion
Crystal violet staining data reveal a modest but reproducible defect in surface adhesion when skaH or spdR are deleted in a wild-type background (Fig 3D and S5 Fig). This defect is more pronounced as culture density increases (Fig 3D). We note that small differences in surface adhesion are challenging to measure. Nevertheless, our results are consistent with a recent genome-scale analysis of adhesion to a cellulosic surface where mutants with disruptions in skaH, spdS, spdR, and rtrB exhibited reduced cell attachment to cheesecloth in complex medium (S10 Fig) [58]. This independent genetic experiment supports a model in which these gene products function together as positive regulators of surface attachment across distinct media conditions.
The specific environmental or intracellular signals that regulate activity of the sensory systems identified in our selection are not known. However, SpdS-SpdR homologs have been studied extensively and shown to sense the cellular redox state by three mechanisms: 1) binding of oxidized quinone [59,60], 2) cysteine sulfenic acid modification [61] or 3) disulfide bond formation [55] at a conserved reactive cysteine. C. crescentus SpdS transfers a phosphoryl group to SpdR in vitro [39], but the role of SpdS-SpdR phosphorylation in vivo remains unclear. In other Alphaproteobacteria, SpdS homologs have been reported to dephosphorylate their cognate regulators [62][63][64][65], and thus we cannot exclude the possibility that SpdS may primarily function to dephosphorylate SpdR in C. crescentus cells. In fact, we observed that upon entering stationary phase, a ΔspdS mutant has dramatically enhanced surface adhesion ( Fig  3D), similar to the hyper-adhesive phosphomimetic spdR(D64E) mutant. Furthermore, we present evidence that the hyper-adhesive phenotype of the ΔspdS mutant depends on spdR ( S6B Fig and S6C Fig). This stationary phase ΔspdS phenotype suggests a model in which stationary phase signal(s) activate SpdS as a SpdR~P dephosphorylase. This model remains to be tested.
C. crescentus LovK has the capacity to function as both a redox sensor [37] and a photosensor [33]. The genetic interactions we uncovered between LovK and SpdS-SpdR may reflect a fundamental role for LovK in redox sensing in vivo. C. crescentus is reported to undergo a dynamic change in the intracellular redox state during the cell cycle [66]. It would be interesting to know if hfiA transcriptional regulation by the SkaH, LovK and SpdS sensor kinases is influenced by intracellular redox. The ability of SkaH to function as a sensor has not been characterized, though a database search using the Phyre2 profile-profile alignment algorithm [67] provides evidence for two PAS-like sensor domains [68] N-terminal to the HWE kinase domain. The multi-protein regulatory system we report here could potentially sense a variety of environmental signals (e.g. redox potential, light, small molecules) to control hfiA expression.

Downstream regulators that directly control hfiA transcription
We identified two transcription factors, rtrA and rtrB, that function downstream of SpdR to directly repress hfiA transcription and control surface adhesion. rtrA was previously identified as part of the stationary phase SpdR regulon in C. crescentus [40], while rtrB was not. Our transcriptomic experiments were conducted on cells in logarithmic phase expressing a phosphomimetic SpdR allele. Studies of RegA-dependent transcription in Rhodobacter capsulatus have shown that the composition of the RegA regulon depends on the growth medium [69]. Similar conditional regulation by SpdR may exist in C. crescentus, and thus rtrA and rtrB may be activated and control hfiA transcription under distinct sets of environmental conditions. Possible conditional regulation of rtrA and rtrB in C. crescentus is reflected in our overexpression and genetic epistasis data that show distinct holdfast and surface adhesion phenotypes for the individual rtrA and rtrB overexpression and deletion strains. Efforts to globally define the regulons of rtrA and rtrB are ongoing and will provide a more complete understanding of the broader functional roles of these direct hfiA regulators in C. crescentus.

Plasmid construction
C. crescentus DNA was amplified from colonies using KOD Xtreme hot-start polymerase (EMD Biosciences/Novagen). PCR reactions were supplemented with 5% dimethyl sulfoxide (DMSO). All plasmids were cloned in E. coli Top10 (Invitrogen, Carlsbad, CA). The sequences of all cloned products were confirmed in the target plasmids. Gene deletions and point mutations were generated by spliced overlapping PCR using external primers with specific restriction enzyme sites. PCR products were digested and ligated into the pNPTS138 plasmid [71] at the necessary restriction enzyme sites. Plasmids for ectopic gene expression were generated by ligating a digested PCR product of the gene of interest into colE1-based integrating plasmids with either a vanillate (P van ) or xylose (P xyl ) inducible promoter [72]. Plasmids are listed in S3 Table.

Strain construction
Chromosomal allele replacements were made using standard two-step recombination using sacB for counterselection. Briefly, pNPTS138-derived plasmids were transformed into C. crescentus by electroporation. Primary integrants were selected by growth on PYE Kan plates, inoculated into liquid PYE and grown overnight without selection. Cultures were plated on PYE agar supplemented with 3% sucrose to select for clones that lost the plasmid in a second recombination event. Chromosomal allele replacements were confirmed by PCR amplification of the gene of interest from kanamycin sensitive clones. pRKlac290 transcriptional reporter plasmid P hfiA -lacZ was conjugated into C. crescentus by triparental mating [70]. C. crescentus was then grown on PYE plates supplemented with tetracycline to select for cells carrying the plasmid and nalidixic acid to counterselect against E. coli. Integrating expression plasmids were transformed into C. crescentus by electroporation. Strains carrying the plasmid were selected on PYE agar supplemented with a plasmid-specific antibiotic.

Protein purification for EMSA
For heterologous expression, pET-based plasmids were transformed into Rosetta(DE3)pLysS (Novagen). Strains were inoculated into 10 ml of LB liquid medium supplemented with Kan 50 μg/ml and grown overnight shaking at 37˚C. The overnight culture was used to inoculate a flask with 250 ml of fresh LB liquid medium supplemented with Kan 50 μg/ml and grown shaking to OD 600nm of 0.6-0.8 at 37˚C. Then, a final concentration of 1 mM isopropyl β-D-1-thiogalactopyranoside (GoldBio) was used to induce protein expression for 4 hrs under the same conditions. The cultures were harvested by centrifugation at 11,000 × g for 7 min at 4˚C, resuspended in 15 ml of resuspension buffer (10 mM Imidazole, 125 mM NaCl, 25 mM Tris pH 7.6) and saved at -20˚C until needed. When samples were thawed, DNAseI and PMSF were added to final concentrations of 250 μg/ml and 800 μM, respectively. The cells were then disrupted by one passage in a microfluidizer (Microfluidics LV1) and the resulting lysate was clarified by centrifugation at 39,000 × g for 15 min at 4˚C. Purification of His 6 -RtrB, His 6 -SUMO-RtrA and His 6 -SpdR(D64E) was performed by nickel affinity chromatography (nitrilotriacetic acid resin; GE Healthcare). After binding of the clarified lysate to the column, the sample was washed with 30-50 ml of resuspension buffer followed by 10 ml of wash buffer (75 mM Imidazole, 125 mM NaCl, 25 mM Tris pH 7.6). Protein was eluted with 1-2 ml of elution buffer (500 mM Imidazole, 125 mM NaCl, 25 mM Tris pH 7.6). Sample purity was assessed by resolving the purified protein by 12% SDS-PAGE and the gels were stained with Coomassie blue.
For purification of RtrA, ULP1 protease was added to the eluted protein to cleave the His 6-SUMO tag and this sample was dialyzed overnight at 4˚C in 1 L of buffer containing 25 mM Tris pH 7.6, 125 mM NaCl 2 , 50% glycerol. The cleaved protein was further purified by nickel affinity chromatography, where the flow through containing the cleaved untagged RtrA was collected for use in this study. For His 6 -RtrB purification, the eluted protein was dialyzed overnight at 4˚C in 1 L of buffer containing 25 mM Tris pH 7.6, 200 mM KCl, 10 mM MgCl 2 , 15mM EDTA pH 8.0 and 50% glycerol. For purification of His 6 -SpdR(D64E), the eluted protein was dialyzed overnight at 4˚C in 1 L of buffer containing 25 mM Tris pH 7.6, 125 mM NaCl 2 and 50% glycerol. Aliquots of the purified proteins were saved at -80˚C.

Genetic selection
The goal of this selection was to isolate spontaneous mutants in which the lovK(H180A) allele could no longer repress the hfiA promoter (P hfiA ). We used a lovK(H180A) strain carrying an integrated plasmid in which the promoter of hfiA was fused to cat (P hfiA -cat). The cat gene encodes chloramphenicol acetyltransferase, which confers resistance to the antibiotic chloramphenicol in a dose dependent manner. We used growth rate in the presence of chloramphenicol as a proxy for P hfiA -cat activity and enriched for spontaneous mutants that grew faster than the parental lovK(H180A) strain in the presence of chloramphenicol. We identified optimal selection conditions by comparing growth rate of wild-type and lovK(H180A) strains carrying the P hfiA -cat plasmid in a range of chloramphenicol concentrations. The highest dynamic range between growth rates in the wild-type and lovK(H180A) backgrounds was at 1-2 μg/ml chloramphenicol.
For each independent selection, one ml of an overnight culture of the parental strain was used to inoculate flasks containing 500 ml of M2X medium supplemented with a final concentration of 1-2 μg/ml of chloramphenicol. Flasks were incubated shaking at 30˚C until they reached OD 660nm of 0.6-0.9. At this point, 100-500 μl of the cultures was used to inoculate 500 ml of fresh M2X medium. Cultures were allowed to grow and this passaging process was performed until the diluted cultures grew fast enough to reach OD 660nm of 0.6-0.9 in a single overnight incubation. Upon initial inoculation, cultures took 3-5 days to reach the desired OD 660nm but after 5-7 serial passages cultures only took about 18 hrs to reach OD 660nm of 0.6-0.9. This indicated that the population was more resistant to chloramphenicol. Individual mutant clones were isolated from the final (fast-growing) culture by plating serial dilutions on PYE agar supplemented with gentamicin 5 μg/ml, which selected for cells carrying the P hfiAcat plasmid. Several isolated single colonies were picked for further analysis.
Several secondary screens were performed on the isolated mutants. First, we sequenced the hfiA promoter from the P hfiA -cat plasmid in order to distinguish clones with intact selection systems from those with cis mutations in the hfiA promoter. We then confirmed faster growth rate in the presence of chloramphenicol. Then, to distinguish mutants that result in de-repression of P hfiA from those that confer chloramphenicol resistance through other mechanisms, we evaluated bulk surface attachment of the isolated mutants by crystal violet stain or ring assay (see below) as a proxy for activity of the native hfiA promoter. Finally, mutants that displayed faster growth rate and decreased surface attachment compared to the lovK(H180A) parental strain were transformed with the P hfiA -lacZ transcriptional fusion plasmid. β-galactosidase activity was measured to confirm de-repression of the hfiA promoter in the spontaneous mutants compared to the lovK(H180A) strain. Based on the results obtained during the secondary screens, several isolates were selected and the mutations in these strains were mapped by whole genome sequencing.

Whole genome sequencing
We isolated genomic DNA from the C. crescentus parent and the mutant strains arising from our chloramphenicol selection using a standard guanidinium thiocyanate extraction and isopropanol/ethanol precipitation. The DNA was randomly sheared and libraries were prepared for whole genome shotgun sequencing using an Illumina HighSeq 2500 (50-bp single end reads). Whole-genome sequence data from the parent and mutant strains were assembled to the C. crescentus NA1000 genome (genbank accession CP001340), and polymorphisms between the parental strain and suppressor strains were identified using Geneious v 11 [73], see S1 Table. The sequenced genome libraries yielded an average of 11.4 million reads, resulting in global average depth of coverage of 141x.

Holdfast quantification and imaging
Overnight cultures grown in M2X were diluted to an approximate OD 660nm of 0.00002 in 2 ml of M2X. When the cultures reached 0.05-0.1 OD 660nm , 5 μl of 1 mg/ml Wheat Germ Agglutinin, Alexa Fluor 594 Conjugate (ThermoFisher) was added to the culture. After incubation for 10 minutes, cells were collected by centrifugation for 2 minutes and resuspended in 10-25 μl of M2X. Cells were imaged with a Leica DM5000 microscope. Fluorescent WGA staining was visualized using an RFP fluorescent filter (Chroma set 41043).

Crystal violet stain surface attachment assay
Colonies from PYE agar plates were inoculated into 2 ml of M2X broth and grown overnight at 30˚C shaking at 200 rpm. Overnight liquid cultures were diluted to a final OD 660nm of 0.00625 into wells of 24-well polystyrene plates containing 1 ml of M2X or M2VX. Plates were grown shaking at 155rpm at 30˚C for varying times periods as indicated in the results. Plates were removed from the incubator and surface attached cells were measured with crystal violet stain. Briefly, the cultures were discarded and the wells were thoroughly washed with water. Then, 1.5 ml of 0.01% crystal violet in water was added to each well and the plates were incubated with shaking at 155 rpm for 5 min. The crystal violet solution was discarded and the wells thoroughly washed with water again. To dissolve the stain, 1.5 ml of 100% EtOH was added to each well and the plates incubated again with shaking for 5 min. Crystal violet stain extracted in each well was quantified by measuring absorbance at 575nm.

B-galactosidase assay
Strains were inoculated in M2X medium from colonies on PYE-agar plates and grown shaking overnight at 30˚C. Overnight cultures were then diluted back into fresh M2X to an OD 660nm of 0.00025-0.00075 and grown in a shaking incubator at 30˚C until they reached OD 660nm 0.05-0.15 (15-20 hrs). At this point, β-galactosidase activity was measured as previously described [14].

Bacterial two-hybrid
We utilized a system previously described [42]. We co-transformed plasmids bearing fusions to either the T18c or T25 domains of adenylate cyclase into the adenylate cyclase null strain, BTH101, by electroporation. An aliquot of the outgrowth was plated on LB agar supplemented with Amp 100 μg/ml, Kan 50 μg/ml, X-gal 80 μg/ml and IPTG 0.5 mM and grown at 30˚C for 24 hrs. Strains were inoculated into LB liquid medium supplemented with Amp 100 μg/ml, Kan 50 μg/ml and IPTG 0.5 mM and grown shaking overnight at 30˚C. Overnight cultures were diluted to OD 600nm of 0.05 and 5 μl of each culture were spotted into LB agar plates supplemented with Amp 100 μg/ml, Kan 50 μg/ml, X-gal 80 μg/ml and IPTG 0.5 mM. The color of each spot was evaluated after 36 hours of growth at 30˚C.

Protein pulldown from cell lysate
C. crescentus colonies from PYE agar plates were inoculated into 2 ml of M2X and grown shaking overnight at 30˚C. These cultures were diluted into 40 ml of M2X and grown until they reached OD 660nm of 0.4-0.6. The cultures were then induced with 0.5 mM vanillate for 3 hrs. The cells were harvested by centrifugation at 11,000 × g for 15 min at 4˚C, resuspended in 6 ml of standard buffer (25 mM Tris pH 7.6, 125 mM NaCl) and frozen at -20˚C. When thawed, cells were disrupted by one passage in a microfluidizer (Microfluidics LV1), PMSF was added to a final concentration of 800 uM and the lysates were cleared by centrifugation at 21,000 × g for 5 min at 4˚C. The cleared lysate was loaded into a column of approximately 200 μl amylose resin (NEB) prequilibrated in buffer, and allowed to bind. The resin was washed with 30-50 ml standard buffer and eluted with 150 μl standard buffer supplemented with 40 mM maltose. All the fractions collected for analysis were mixed 1:1 with SDS loading buffer, boiled for 2 min at 95˚C and saved at -20˚C for further Western Blot analysis. 15 μl of each pulldown fraction sample was loaded onto a Mini-PROTEAN TGX Precast 4-20% Gradient Gel (Bio-Rad). Samples were resolved at 35 mA constant current in SDS running buffer (0.3% Tris, 18.8% Glycine, 0.1% SDS). Proteins in the gel were transferred to an Immobilon-P PVDF Membrane using a Mini Trans-Blot Cell after preincubation in Western transfer buffer (0.3% Tris, 18.8% Glycine, 20% methanol). Transfer was carried out at 100 volts for 1 hr at 4˚C in Western transfer buffer. The membrane was then blocked in 5% powdered milk in Tris-buffered Saline Tween (TBST: 137 mM NaCl, 2.3 mM KCl, 20 mM Tris pH 7.4, 0.1% Tween 20) shaking for 1 hr at 4˚C. Incubation with the primary antibody, FLAG monoclonal antibody (clone FG4R) or HA-Tag monoclonal antibody, was carried out shaking overnight in 5% powdered milk TBST at 4˚C. Membrane was then washed 3 times in TBST for a total of 30 min at room temperature. Incubation with Goat anti-Mouse IgG (H+L)-HRP secondary antibody was at room temperature for 1 h in TBST. Finally, the membrane was washed 3 times in TBST for a total of 30 min at room temperature. Chemiluminescence was performed using the SuperSignal West Femto Maximum Sensitivity Substrate (Pierce) and was imaged using a ChemiDoc MP imaging system (Bio-Rad). Chemiluminescence was measured using the ChemSens program with an exposure time of 30-60 sec.

Electrophoretic Mobility Shift Assays
Promoter regions of interest were amplified by PCR using a set of appropriate primers in which the reverse primer was fluorescently labeled with AlexaFluor 488 (Integrated DNA Technology, DTT). The hfiA promoter probe was 275 bp (primers F:TGGTGGTCCTGAT CCTCCTG and R: CACTGACAACATCCTGTCCG) and the cspD promoter probe was 199 bp (primers F: CTAGGGACTGCCATCTTCGG and R: TCGTAACCAGACATCCCACC). Specific cold specific competitor probe represents the same region as the corresponding labeled probe amplified with unlabeled primers with the same sequence. The cold non-specific competitor was a 220 bp PCR product amplified from the pNPTS138 plasmid using primers F: GTAAAACGACGGCCAG and R: CAGGAAACAGCTATGAC.
For reactions in which the binding of His 6 -SpdR(D64E) was tested, DNA binding reactions were performed in 20 μl reaction volumes containing binding buffer (20mM HEPES pH 7.8, 4 mM MgCl 2 , 37 mM KCl, 1 mM CaCl 2 , 0.75 mM dithiothreitol, 8% glycerol and 75 μg/ml bovine serum albumin), 8 ng of fluorescently labeled DNA probe and increasing concentrations of His 6 -SpdR(D64E) (0, 50, 125, 250, 375 nM for incubation with the hfiA promoter probe and 0, 25, 50, 125, 250 nM for incubation with the cspD promoter probe). Reactions were incubated at room temperature for 30 min in the dark. Then, 12.5 μl of each reaction was loaded on a fresh 6% native acrylamide gel pre-run for at least 30 min in 1X Tris Acetate EDTA buffer (TAE: 40 mM Tris, 20 mM acetic acid and 1 mM EDTA) in the dark at 4˚C for 1 hr at 70 volts followed by 1.5 hr at 90 volts. The gels were imaged using the BioRad Chemidoc MP imaging system with a 60-120 sec exposure and the manufacturer's settings for AlexaFluor 488 detection.
For reactions where the binding of RtrA was tested, DNA binding reactions were performed in 20 μl reaction volumes containing binding buffer (10 mM Tris pH7.6, 2 mM MgCl 2 , 50 mM KCl, 2 mM CaCl 2 , 0.2 mM dithiothreitol, 5% glycerol, 1 mM EDTA pH 8.0 and 1 mg/ ml bovine serum albumin), 8 ng of fluorescently labeled DNA probe and increasing concentrations of RtrA (0, 100, 200, 300, 400 nM). For competition experiments, a 10-fold excess of cold competitor probe was added to the reaction. Reactions were incubated at room temperature for 30 min in the dark. 12.5 μl of each reaction was loaded on a fresh 6% native acrylamide gel pre-run for at least 30 min in 1X Tris Borate EDTA (TBE) in the dark at 4˚C for 1 hr at 100 volts. The gels were imaged using the BioRad Chemidoc MP imaging system with a 30 sec exposure and the manufacturer's settings for AlexaFluor 488 detection.
For reactions where the binding of RtrB was tested, DNA binding reactions were performed in 20 μl reaction volumes containing binding buffer (20mM HEPES pH 7.8, 4 mM MgCl 2 , 37 mM KCl, 1 mM CaCl 2 , 0.75 mM dithiothreitol, 8% glycerol and 75 μg/ml bovine serum albumin), 7 ng of fluorescently labeled DNA probe and increasing concentrations of RtrB (0, 10, 25, 50, 75, 100, 250 nM). For competition experiments, a 10-fold excess of cold competitor probe was added to the reaction. Reactions were incubated at room temperature for 30 min in the dark. 12.5 μl of each reaction was loaded on a fresh 6% native acrylamide gel pre-run for at least 30 min in 1X TAE in the dark at 4˚C for 30 min at 80 volts, and subsequently 1 hr at 100 volts. The gels were imaged using the BioRad Chemidoc MP imaging system with a 30 sec exposure and the manufacturer's settings for AlexaFluor 488 detection.

RNA preparation, sequencing and analysis
Strains grown on PYE agar plates were inoculated into 5 ml of M2X medium in 3 biological replicates for each strain, and grown overnight on a rolling incubator at 30˚C. Cultures were diluted to OD 660nm of 0.008 in 8 ml of fresh M2X and grown in the same manner until they reached OD 660nm of 0.30-0.35. At this point, 6 ml of each replicate were collected by centrifugation at 15,000 × g for 1 min. This cell pellet was immediately resuspended in 1 ml of TRIzol and stored at -80˚C. The samples were heated for 10 min at 65˚C. Then, after addition of 200 μl of chloroform samples were vortexed and incubated at room temperature for 5 min. Phases were separated by centrifugation at 15,000 × g for 15 min at 4˚C. The upper aqueous phase was transferred to a new tube and 0.7 volume of cold 100% isopropanol was added. Samples were then stored at -80˚C overnight. The overnight precipitation was centrifuged at 15,000 × for 30 min at 4˚C and the resulting nucleic acid pellet was washed twice with cold 70% ethanol. The pellet was centrifuged again at 15,000 × for 5 min at 4˚C, ethanol was removed and the pellet was allowed to dry. The nucleic acid pellet was resuspended with nuclease-free water. The samples were treated with Turbo DNase (Ambion, Life Technologies) and clean-up was performed with RNeasy Mini Kit (Qiagen).
Stranded cDNA libraries were prepared and sequenced by the Functional Genomics Core at the University of Chicago. Briefly, RNA samples were treated with the Ribo-zero Kit (Illumina) for rRNA removal and library preparation proceeded with Illumina ScriptSeq RNA-Seq Library Preparation Kit. All the libraries were sequenced with an Illumina HiSeq4000 instrument. Analysis of RNA-seq data was performed with CLC Genomics Workbench (Qiagen). Reads were mapped to the C. crescentus NA1000 genome (Genome accession number CP001340). We used the RNA-seq Analysis Tool to analyze the differential expression between the different strains.

RNA-seq data availability
The raw RNA-seq data for each spdR(D64E) and ΔspdR sample are available in the NCBI Gene Expression Omnibus (GEO) under accession number GSE125783.

SpdR binding motif discovery
We used the MEME Suite [49] to identify potential SpdR binding motifs in promoter regions of genes with more than 2-fold differences in expression between the spdR(D64E) and ΔspdR strains. The promoter sequences (-150 to +50, based on translation start site annotations in GenBank accession CP001340) of genes with > +2-fold change (up-regulated in spdR(D64E) relative to ΔspdR) were submitted to MEME to identify motifs. A motif with high similarity to that previously reported for C. crescentus SpdR [39,40] and for SpdR homologs [46][47][48] was found in 65 of 119 promoter sequences, each with a p-value < e -4 (listed in S2 Table; see Fig 5). Using the FIMO Motif Scanning algorithm, we applied the SpdR motif matrix from MEME to scan the promoter sequences (-150 to +50) of genes with < -2-fold change (down-regulated by spdR(D64E) strain relative to ΔspdR) for potential SpdR motif sequences. Predicted SpdR binding motifs with a p-value cutoff < e -4 were present in the promoter sequence of 17 out of 48 of the down-regulated genes, and are listed in S2 Table. Supporting information S1 Fig. Genetic analysis of lovK mediated regulation of surface adhesion. A) Strains bearing a deletion of the lovK-lovR locus were complemented with a wild-type lovK or lovK(H180A) allele. Surface attachment to polystyrene plates was measured by crystal violet stain. Cultures were grown in M2X medium. Data are representative of at least three independent experiments. Bars represent mean ± s.d.; n = 7.  [14]. Predicted start codon corresponding to original hfiA annotation in green; the revised hfiA reading frame is in yellow [14]. Experimentally identified hfiA transcriptional start sites in light blue [14] [74]. The central portion of the PhfiA-cat fusion is shown with P hfiA in gray while and cat in pink. Sites of spontaneous mutations identified in the genetic selections with the respective nucleotide change in light purple. Genome coordinates are based on the NA1000 genome sequence (CP001340; NC_011916). Strains bearing in-frame deletions of skaH, spdS or spdR in a lovK(H180A) background were transformed with either pMT680 plasmids that express the corresponding disrupted gene from a xylose inducible promoter or pMT680 as an empty vector (EV) control. A) β-galactosidase activity from the P hfiA -lacZ transcriptional fusion was measured in each strain. Cultures were grown in M2X medium to OD 660nm of 0.05-0.15. Presented data are representative of at least three independent experiments. Bars represent mean ± s.d.; n = 6. B) Surface attachment of cells grown in polystyrene plates was measured by crystal violet stain. Cultures were grown in M2X medium for 16 hrs post-inoculation. Data are representative of three independent experiments. Bars represent mean ± s.d.; n = 9. A) Surface attachment of cells grown in polystyrene plates was measured by crystal violet stain. Wild-type (WT) and lovK(H180A) carry empty vectors of pMT680 and pMT585. rtrA ++ carries pMT680-rtrA and pMT585 empty vector. rtrB ++ carries pMT585-rtrB and pMT680 empty vector. rtrA ++ rtrB ++ carries pMT680-rtrA and pMT585-rtrB. Cultures were grown for 16 hrs post-inoculation in M2X medium. Data presented are representative of at least three independent experiments. Data represent mean ± s.d.; n = 6-8. B) Surface attachment of cells grown in polystyrene plates was measured by crystal violet stain in ΔrtrA, ΔrtrB, and ΔrtrAΔrtrB in a wild-type (WT) and spdR(D64E) backgrounds. Cultures were grown for 16 hrs post-inoculation in M2X medium. Data presented are representative of at least three independent experiments. Data represent mean ± s.d.; n = 9. (TIF) S10 Fig. Global Tn-seq approach provides an additional line of evidence that genes identified in this study play a role in regulating adhesion in wild-type cells. Extracted fitness profiles of mutants characterized in a genome-wide attachment screen [58]. Briefly, a C. crescentus transposon mutant library was grown and passaged in complex PYE medium in the presence of cheesecloth to provide surface area for cell attachment. Cells were periodically sampled from the broth over several days. Hyper-adhesive mutants are titrated out of the broth by the cheesecloth and, consequently, have negative fitness scores (in which the fitness score is a log 2 ratio of mutant strain abundance relative to the average strain). Conversely, hypo-adhesive mutants are enriched in the broth resulting in positive fitness scores. Strains bearing transposon insertions in TCS genes identified in our study have positive fitness scores. hfsJ mutants, which do not produce holdfast and are non-adhesive, are presented as a reference. rtrA was not captured in this analysis; the library contained insufficient Tn-insertions in this gene. (TIF) S1