Investigating the role of the carbon storage regulator A (CsrA) in Leptospira spp.

Carbon Storage Regulator A (CsrA) is a well-characterized post-transcriptional global regulator that plays a critical role in response to environmental changes in many bacteria. CsrA has been reported to regulate several metabolic pathways, motility, biofilm formation, and virulence-associated genes. The role of csrA in Leptospira spp., which are able to survive in different environmental niches and infect a wide variety of reservoir hosts, has not been characterized. To investigate the role of csrA as a gene regulator in Leptospira, we generated a L. biflexa csrA deletion mutant (ΔcsrA) and csrA overexpressing Leptospira strains. The ΔcsrA L. biflexa displayed poor growth under starvation conditions. RNA sequencing revealed that in rich medium only a few genes, including the gene encoding the flagellar filament protein FlaB3, were differentially expressed in the ΔcsrA mutant. In contrast, 575 transcripts were differentially expressed when csrA was overexpressed in L. biflexa. Electrophoretic mobility shift assay (EMSA) confirmed the RNA-seq data in the ΔcsrA mutant, showing direct binding of recombinant CsrA to flaB3 mRNA. In the pathogen L. interrogans, we were not able to generate a csrA mutant. We therefore decided to overexpress csrA in L. interrogans. In contrast to the overexpressing strain of L. biflexa, the overexpressing L. interrogans strain had poor motility on soft agar. The overexpressing strain of L. interrogans also showed significant upregulation of the flagellin flaB1, flaB2, and flaB4. The interaction of L. interrogans rCsrA and flaB4 was confirmed by EMSA. Our results demonstrated that CsrA may function as a global regulator in Leptospira spp. under certain conditions that cause csrA overexpression. Interestingly, the mechanisms of action and gene targets of CsrA may be different between non-pathogenic and pathogenic Leptospira strains.

Introduction Leptospira spp. are gram-negative, spiral-shaped bacteria categorized into non-pathogenic and pathogenic strains. Non-pathogenic Leptospira spend their entire life in the natural environment. Pathogenic Leptospira can survive in the environment and cause leptospirosis in susceptible hosts. It is estimated that 1 million people suffer from severe leptospirosis each year and there are approximately 60,000 deaths, mostly in developing tropical countries [1].
The ability to survive in a wide range of environments is crucial for both pathogenic and non-pathogenic Leptospira spp. The pathogenic strains have to complete the zoonotic cycle to live in distinct habitats, including survival in aqueous or terrestrial environment [2], kidneys of their reservoir hosts [3], or target organs of their susceptible hosts [4]. Most transmission occurs when people are exposed to water and soil contaminated by urine of reservoir animals. Leptospira can then enter in the host through abraded skin or mucous membrane followed by hematogenous spread to the target organs. The mechanism underlying the long-term survival of pathogenic Leptospira under nutrient-poor conditions is not completely understood but biofilm formation may play an important role [5,6]. Furthermore, omics studies revealed the changes in gene expression profiles in Leptospira in response to different environmental conditions such as temperature shift [7], physiologic osmolarity [8], serum exposure [9], iron limitation [10], in vivo cultivation on dialysis membrane chamber [11], and in the presence of biofilm [12]. These transcriptome studies highlighted the role of global gene regulation which is a crucial process employed by the bacteria to deal with the changes in the environment. However, due to the lack of efficient genetic manipulation, knowledge of gene regulation is not well understood in Leptospira spp. Some regulators have been characterized such as the peroxidase stress regulator PerR [10], the KdpE sensor potassium transport activator [13], DNA repair LexA [14], the sigma factor RpoN [15,16] and, more recently, the pathogenspecific two-component system LvrAB [17]. Besides these regulators, all leptospiral genomes also possess a gene that encodes the putative CsrA [18][19][20].
Carbon Storage Regulator A (CsrA) (or its homolog RsmA) is one of the most studied RNA binding proteins in bacteria [21]. This protein is widely conserved in more than 1,500 bacterial species. A transposon mutant of csrA in Escherichia coli was first reported to display pleiotropic phenotypes including alteration in glycogen accumulation, adhesion ability, and cell size compared to the wild type strain [22]. Due to substantial pleiotropic effects, several omics studies have reported the effect of CsrA on global transcriptomic changes [23][24][25][26][27][28][29][30][31][32][33] (Table 1), showing that CsrA is a global regulator in both gram-positive and gram-negative bacteria. CsrA regulates gene expression at the post-transcriptional level by binding to mRNA targets, and affects mRNA stability and translation [34]. This protein could negatively or positively regulate mRNA expression. For negative regulation, CsrA binds to the Shine-Dalgarno region and prevents ribosome access to the targeted mRNA thus blocks the translation process of the bacteria [35][36][37]. In addition, CsrA may bind to mRNA targets that overlaps the start codon [38]. For positive regulation, CsrA binds to mRNA target and prevents the target from being cleaved by the RNase [39,40]. In E. coli, the consensus sequence of the CsrA binding site is 5'RUA-CARGGAUGU'3 where the GGA motif is located in a hairpin loop which is a critical binding site for CsrA [41]. The involvement of CsrA in the regulation of various bacterial processes has been reported including carbon metabolism [42], motility [40], biofilm formation [43], quorum sensing [44], stress response [45], as well as virulence-associated traits such as iron acquisition [46], invasion [47], and type III secretion [48].
Among the phylum of Spirochaetes, CsrA was extensively studied in Borrelia burgdorferi, the causative agent of Lyme disease. The csrA mutant of B. burgdorferi showed that there was a decrease in the expression of some virulent-associated proteins and attenuation in the mouse model [49], but these data were not confirmed by another study [50]. CsrA also acts as a repressor of the flagellin protein FlaB [51]. RNA sequencing (RNA-seq) showed that 13% of the genes were differentially expressed in the csrA mutant [23].
Our objective was to investigate the role of CsrA in Leptospira spp. We generated a csrA deletion mutant and csrA overexpressing strains to answer this question. In non-pathogenic L. biflexa, csrA was required for growth under starvation conditions. RNA-seq revealed that in rich-nutrient conditions, deletion of csrA had minimal impact on global gene regulation. We showed that CsrA is a repressor of flagellin transcripts but no alteration of motility phenotype in both deletion mutant and overexpressing strains was observed. In the pathogen L. interrogans, overexpression of csrA resulted in motility defect and CsrA could bind to flagellin transcripts. Our results demonstrated that the mechanisms of action and gene targets of CsrA appear to be different between pathogenic and non-pathogenic Leptospira strains.

Bacterial strains and growth conditions
Leptospira spp. were grown in liquid Ellinghausen-McCullough-Johnson-Harris (EMJH) medium (Difco) at 30˚C or 1% agar of solid EMJH at 30˚C. The saprophyte Leptospira biflexa serovar Patoc strain Patoc1 and the pathogen Leptospira interrogans serovar Manilae strain L495 were obtained from the French National Reference Center (NRC) for Leptospirosis (Institut Pasteur, Paris, France). Escherichia coli strains were grown in Luria-Bertani (LB) medium at 37˚C. When needed, an appropriate antibiotic was added to the culture medium. Bacterial strains are listed in Table 2.
For growth curves, the bacteria were grown in EMJH medium until the culture reached exponential phase (OD 420~0 .1 to 0.2 or 2.5×10 8 cells/mL). Then, 2×10 6 bacteria were added into 10 mL of EMJH medium. The cultures were incubated at 30˚C, at 30˚C with 100 rpm shaking or at 37˚C with 200 rpm shaking. One mL of each culture was taken for OD 420 measurement every 24 h. In order to perform a growth curve in diluted EMJH, Leptospira cells were prepared as described above before inoculation into 1/5 EMJH medium diluted in sterile water.

Allelic exchange mutagenesis of leptospiral csrA
A L. biflexa csrA deletion mutant was generated by allelic exchange. Briefly, a plasmid containing a kanamycin resistance cassette was used to replace the coding sequence of csrA, LEPBIa3210, and 0.8 kb sequences that flanked the target gene was synthesized by GeneArt (Life Technologies, Grand Island, NY, USA), pretreated by UV, and used to transform L. biflexa as previously described [52]. A similar strategy was performed for the csrA homolog, LIMLP_17575, in L. interrogans serovar Manilae. The map of each suicide plasmid is shown in S1 Fig.
To check for a double crossing-over event among the kanamycin-resistant colonies of L. biflexa, genomic DNA was isolated from exponential phase cultures using a Maxwell 16 cell DNA purification kit and a Maxwell instrument (Promega, Madison, WI), and PCR was performed on DNA extracts with the following primer pairs: 1) Flk_L and Flk_R, and 2) ORF_L and ORF_R.

Construction of the plasmids and E. coli β2163 conjugation with Leptospira spp.
To construct the plasmids for complementation and overexpression, the L. biflexa and L. interrogans csrA genes were cloned into 2 different vectors. We first cloned csrA in pMaGro [53] in front of a strong promoter groES. We also synthesized a transcription fusion of csrA with a promoter of operon flgN-flgK-flgL-fliW-csrA by GeneArt (Life Technologies, Grand Island, NY, USA). This fusion was cloned into the SacI and XbaI sites of pMaORI [54]. All pMaORI constructs are shown in S2 Fig. Conjugation was performed as previously described [55]. Briefly, E. coli β2163 containing plasmid of interest was incubated with log-phase Leptospira on a membrane filter and placed on EMJH plate supplemented with 0.3 mM diaminopimelic acid and incubated for 16-20 h at 30˚C. The bacteria were then resuspended in EMJH and spread onto EMJH solid agar plates supplemented with 50 μg/mL spectinomycin. The plates were incubated at 30˚C until leptospiral colonies were observed, approximately 1 week for L. biflexa and 2 weeks for L. interrogans.

RNA purification and RT-qPCR
RNA isolation was performed as previously described [56]. Briefly, Leptospira spp. were grown until the growth reached exponential phase, OD 420~0 .1 to 0.2 or~2.5 × 10 8 cells/mL. The cells were harvested and RNA was extracted using TRIZOL reagent (Thermo Fisher Scientific, Vantaa, Finland) as previously described [56]. RNA pellets were resuspended in UltraPure Dnase/Rnase Free Distilled Water (Thermo Fisher Scientific). Genomic DNA was removed by DNase treatment using the RNase-free Turbo DNA-free turbo kit (Thermo Fisher Scientific) following the manufacturer's instructions. The 500 ng of RNA were used for cDNA synthesis using iScript™ Advanced cDNA Synthesis Kit for RT-qPCR (Bio-Rad Laboratories, Hercules, CA). Quantitative reverse transcription-PCR (RT-qPCR) was performed using SYBR 1 Green Master Mix (Bio-Rad). The results were expressed as the normalized difference of the threshold cycle (ΔΔCT), using cysK and lipL32 as a reference gene for L. biflexa and L. interrogans, respectively. All primers are listed in S1 Table.

RNA-sequencing
As previously described [57], RNA integrity was examined using the RNA 6000 Nano kit with the Agilent 2100 bioanalyzer (Agilent Technologies, Wilmington, DE) and all samples used for constructing the library had RNA Integrity Number (RIN) scores >8.
The QIAseq FastSelect -5S/16S/23S kit (QIAGEN) was used to deplete ribosomal RNA according to the manufacturer's instructions. The libraries were built using the TruSeq Stranded mRNA library Preparation Kit (Illumina, USA) following the QIAseq Fastselect -5S/ 16S/23S protocol recommendations. Quality control of the libraries was made on the Fragment Analyzer. The sequencing of the libraries was performed on the Illumina NextSeq 500 platform using single-end 150bp format. The RNA-seq analysis was performed with Sequana (version 0.9.6) [58]. In particular, we used the RNA-seq pipeline (version 0.9.20, https:// github.com/sequana/sequana_rnaseq). The differential expression analysis testing included normalization conducted with DESeq2 [59,60]. For each comparison, a p-value adjustment (padj) was performed to take into account multiple testing indicating the significance (Benjamini-Hochberg adjusted p-values [61], FDR < 0.05) and the effect size (fold-change) for each comparison. Genes with an adjusted p-value (padj) lower than 0.05 and a log2FC higher or lower than 0.5 were considered differentially expressed. These datasets were deposited into the ArrayExpress database at EMBL-EBI (www.ebi.ac.uk/arrayexpress) under the accession number E-MTAB-10396.

Measurement of motility, cell length, and velocity
The motility was checked on 0.6% semisolid EMJH medium. Exponential-phase Leptospira were diluted in EMJH to obtain OD 420 = 0.1 as a starter culture. A small divot was gouged into the agar surface into which 2μL or 5μL of the inoculum was pipetted. The plates were incubated for 1 week for L. biflexa and 2 weeks for L. interrogans. The diameter of the zone for each colony was measured to the nearest millimeter.
For cell length and velocity measurement, late exponential-phase cultures (OD 420~0 .5) were diluted in EMJH broth to obtain an appropriate number of cells per field for visualization under a dark-field microscope. For cell length, approximately 100 cells per strain were measured in randomly selected fields by using cellSens software (Olympus, Hamburg, Germany). Velocity measurement was performed by video microscopy as described previously [6].
Approximately 70 cells per strain were recorded over 60 s. Trajectory analysis and speed displacement were calculated using Olympus CellSens software. Statistical analysis of motility, cell length, and velocity was performed using an Unpaired T-Test (Prism 5.03, GraphPad Software). A p-value < 0.05 was defined as statistically significant.

Recombinant protein production
PCR products of full sequences of csrA amplified from L. biflexa serovar Patoc or L. interrogans serovar Manilae genomic DNA were cloned into pRSET-C (Invitrogen). The recombinant plasmids were transformed into E. coli DH5α and verified by DNA sequencing (Macrogen., South Korea). Recombinant proteins with N-terminus 6× His tag was induced in E. coli BL21 (DE3) pLysS by 1 mM IPTG at 37˚C for 4 h. The pelleted bacteria were resuspended in phosphate buffered saline (PBS) pH 7.4 and disrupted using a high-pressure homogenizer (Constant System Ltd., Northants, UK). The soluble fraction was isolated by centrifugation at 15000 ×g at 4˚C for 30 min. Protein samples were purified using Ni Sepharose columns (GE Healthcare, Buckinghamshire, UK) and dialyzed with PBS pH 7.4. To check for the purity of the purified recombinant proteins, the proteins were subjected to 15% Sodium Dodecyl Sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to nitrocellulose membranes. The membranes were blocked with blocking buffer (1% BSA in PBS pH 7.4 plus 0.05% Tween 20, PBST) before the anti-6× His tag monoclonal antibody (1:5000; KPL, MD, USA) was added. The membranes were further incubated with the horseradish peroxidase (HRP)-conjugated goat anti-mouse IgG (secondary antibody). All incubations were performed at room temperature for 1 h. After incubation, washing step was performed with PBST three times for 5 min each. Amersham ECL (GE Healthcare), an HRP substrate, was added and incubated for 1 min at room temperature before the membrane was exposed to a CCD camera (Bio-Rad) for chemiluminescent signal reading.
EMSA was performed according to the manual of LightShift™ EMSA Chemiluminescent RNA Kit (Thermo Fisher Scientific). Briefly, the binding reaction was prepared. Each binding reaction consisted of 1X binding buffer (10mM HEPES pH 7.3, 20 mM KCL, 1 mM MgCl 2 , and 1 mM DTT) 1 nM of biotinylated-RNA (LEPBIa_1872 WT, LEPBIa_1872 Mut, or LIMLP_07475), 7.5% glycerol, 10 mM DTT, 0.2 μg/μL Yeast tRNA, and various concentrations of rCsrA in a total volume of 20 μL. The binding reaction was incubated at 37˚C for 30 min. After incubation, loading buffer was added into each reaction and separated on 10% native PAGE for 1 h at 100V. The reaction was transferred onto a nylon membrane, crosslinked with UV for 1 min, blocked for 15 min with a blocking buffer, and washed once with washing buffer. A 1:300 stabilized Streptavidin-HRP in a blocking buffer was added and incubated for 15 min. The membrane was washed 5 times with washing buffer and incubated for 5 min with a substrate equilibration buffer. The membrane was incubated for 5 min in HRP substrate before chemiluminescent signal reading. For competitive EMSA assay, the binding reaction was prepared as described above except rCsrA concentration was fixed at 800 nM while unlabeled RNA (LIMLP_07475) was added to the solution at the final concentrations ranged from 0.8 nM to 8 μM (10-fold serial dilution).

Genetic organization of the csrA locus in L. interrogans and L. biflexa
The csrA locus is conserved in the pathogen L. interrogans serovar Manilae and the saprophyte L. biflexa serovar Patoc; the csrA forms with the flagellar genes to develop an operon-like structure ( Fig 1A). This operon consists of 5 consecutive genes: flgN, flgK, flgL, fliW and csrA. The genes flgK and flgL encode putative flagellar hook-associated proteins, and flgN encodes a putative chaperone for FlgK and FlgL. The gene fliW encodes a putative post-transcriptional regulator of flagellin. There is a 200-bp intergenic region located upstream of flgN, the first gene of the operon, suggesting that there is a putative promoter region. The CsrA of L. biflexa and L. interrogans share >88% sequence identity, while both share~50-60% similarity compared with CsrA from other bacteria. The amino acid alignment of leptospiral CsrA shows conserved sequences (highlighted in yellow) and 2 domains (in square boxes) reported as critical for RNA binding in E. coli [62] (Fig 1B). In addition, leptospiral CsrA is slightly longer than that of other bacteria due to additional C-terminal amino acid residues.

Allelic exchange mutagenesis and complementation of csrA in L. biflexa
In this study, the suicide plasmids containing the L. biflexa csrA and L. interrogans csrA were disrupted by a kanamycin-resistance cassette (Km R ) and transformed in saprophyte L. biflexa and the pathogen L. interrogans, respectively. Transformant colonies were only obtained in L. biflexa. We were unable to get transformant colonies from L. interrogans after 5 attempts. Among the 16 randomly selected kanamycin-resistant colonies of L. biflexa, 5 (31%) produced a~2.6 kb PCR product with Flk primers which indicated that csrA was successfully replaced with kanamycin-cassette by a double crossing-over event; for the other colonies, the kanamycin-cassette was successfully replaced by a single cross-over event (Fig 2A and S3A Fig). To further confirm the deletion of csrA in the double-crossover mutants, primers ORF-R and ORF-L were also used. While the WT produced the expected size of 199-bp PCR product, approximately 1-kb PCR products were obtained from the transformants with allelic exchange (S3B Fig). These results indicated that there was a successful allelic exchange of csrA in L. biflexa which was designated as ΔcsrA.
In order to complement the ΔcsrA, our first attempt was to express the wild-type csrA under a strong promoter of Leptospira, P groES , but no transconjugant was obtained. We

PLOS ONE
Investigating the role of CsrA in Leptospira spp.
hypothesized that the excess level of CsrA may be toxic to Leptospira. Therefore, we expressed csrA under the control of its native promoter, which is the promoter of the operon containing flgN, flgK, flgL, fliW, and csrA ( Fig 1A). The resulting plasmid was used for complementation in ΔcsrA. RT-qPCR revealed that the relative fold change of csrA in the complemented strain (ΔcsrA+pMaORI_PcsrA lb ) was 3.33-fold higher compared with WT+pMaORI ( Fig 2B), indicating overexpression of csrA. In addition, RT-qPCR was unable to detect the expression of csrA in the ΔcsrA+pMaORI, confirming the successful deletion of csrA in L. biflexa.

Phenotype analysis of the ΔcsrA L. biflexa
Effects of csrA on growth and motility. The growth curve of WT, ΔcsrA, and ΔcsrA +pMaORI_PcsrA lb in regular EMJH were comparable (Fig 3A), suggesting that csrA was not essential for growth in L. biflexa. However, we found that ΔcsrA displayed poor growth in 5-fold diluted EMJH compared to the WT (Fig 3B). Complementation of the ΔcsrA partially restored the wild-type phenotype under starvation conditions (Fig 3B).
We performed soft agar assays to determine the motility of ΔcsrA, but we did not find any differences between WT and ΔcsrA (Fig 3C). Consistent with soft agar results, we did not find any difference in cell length or velocity in liquid EMJH between WT and ΔcsrA (Fig 3D and

PLOS ONE
Investigating the role of CsrA in Leptospira spp.

3E
). In addition, the motility behavior of ΔcsrA observed under the dark-field microscope was similar to WT (data not shown).

RNA-sequencing.
To investigate the role of csrA as a global gene regulator in L. biflexa, RNA-seq was performed on exponential-phase cultures of WT, ΔcsrA, and ΔcsrA+-pMaORI_PcsrA lb . With log2FC > ±0.5 and padj<0.05 as the criteria for differentially expressed gene (DEG), only 3 genes, not including csrA, were differentially expressed in ΔcsrA compared with the WT strain which is accounting for less than 0.1% of total ORF (3 in 3730) (Fig 4A, Table 3 and S2 Table). Two genes were significantly upregulated in ΔcsrA; LEP-BIa_1872 (encodes a flagellin protein FlaB3), and LEPBIa_0812 (encodes putative acyltransferase) by 2.331-, and 1.423-fold, respectively, while LEPBIa0979 (encodes oligopeptidase A) was 0.668-fold downregulated (Table 3). Furthermore, the level of LEPBIa_1872 was restored to WT level of ΔcsrA+pMaORI_PcsrA lb , suggesting that LEPBIa_1872 should be a specific gene   target of L. biflexa CsrA. In contrast, the complementation of ΔcsrA could not restore wildtype expression of LEPBIa_0812 and LEPBIa0979, suggesting that these genes are not putative gene targets of CsrA (Table 3). While a few differentially expressed genes were found in ΔcsrA, 575 transcripts consisting of 569 genes (15% of total ORF), 4 ncRNA, and 2 23S rRNA were differentially expressed in the ΔcsrA complemented strain compared with WT transcriptome (Fig 4B, S4 Fig and S2  Table). The gene csrA (LEPBIa_3210) was significantly up-regulated (1.84-fold), further confirming the upregulation of csrA observed by RT-qPCR (Fig 2B and Table 3). Among the 569 genes, clpB (LEPBIa_2449) and groL (LEPBIa2344), known genes involved in general stress response, are one of the most strongly downregulated genes (Table 3 and S2 Table), indicating that overexpression of csrA may induce stress conditions in L. biflexa. Complete set of ORF expression is shown in S2 Table. RT-qPCR was performed to validate the RNA-seq results. As shown in Table 3, the significant upregulation of LEPBIa_0812 and LEPBIa_1872 was confirmed in ΔcsrA, while LEP-BIa_0979 was not differentially expressed by RT-qPCR. The restoration of LEPBIa_1872 in complemented strain was confirmed by RT-qPCR, further confirming this gene as a specific target of CsrA.
FlaB gene as a potential target of CsrA in L. Biflexa. Because L. biflexa has 4 flaB genes, the effect of csrA on the relative expression of these flaB genes was determined (Fig 5A). RT-qPCR confirmed an upregulation of flaB3 in ΔcsrA and the expression level of flaB3 was restored in the complemented strain. We also found that flaB2 (LEPBIa_2132) was significantly upregulated in ΔcsrA and its expression was restored in the complemented strain ( Fig  5A). These results indicated that flaB2 and flaB3 are potential CsrA targets. The upregulation of both genes was correlated with RNA-seq of ΔcsrA, of which only flaB3, not flaB2 was differentially expressed (S2 Table).
CsrA regulates its targets by binding to their upstream sequences overlapping the Shine Dalgarno sequence [63]. The consensus sequence of the CsrA binding site is 5'RUACARG-GAUGU'3 [41]. The upstream sequence analysis of flaB2 and flaB3 showed potential CsrA binding sites with 4 nucleotide mismatches compared with the consensus sequence ( Fig 5B  and S3 Table). The putative binding sites were similar in terms of nucleotide composition and sequence order, 5'ACACAAAGGAGT'3 for flaB2 and 5'ACACAGGAGGGT'3 for flaB3. The Shine Dalgarno sequence (5'AGGAGG'3) was present in the upstream region of flaB3, but not flaB2, suggesting that flaB3 might be more promising to be a CsrA target. Secondary

PLOS ONE
Investigating the role of CsrA in Leptospira spp. structure prediction of 120 nucleotides upstream of flaB3 revealed that the possible binding site formed a GGA motif-containing hexaloop and localized 7 nucleotides before the start codon ( Fig 6A). These findings strongly suggest that flaB3 is a specific CsrA target in L. biflexa. Therefore, only the putative CsrA binding site of flaB3 was selected for further verification.
To confirm the interaction of L. biflexa CsrA and flaB3 transcripts in vitro, N-terminal 6× His-tag recombinant CsrA protein (rCsrA) of L. biflexa, with an approximate molecular weight of 14 kDa, was produced in E. coli (S5 Fig). Electrophoretic mobility shift assay (EMSA) was performed to investigate the interaction between L. biflexa rCsrA and synthesized 22-nucleotide RNA probe upstream of flaB3. Because GGA is a known critical binding site of CsrA, the interaction between rCrsA and GGA motif-containing LEPBIa_1872 WT probe was compared to AAA-containing LEPBIa_1872 Mut probe (Fig 6A). The rCsrA could bind to the WT probe in a dose dependent manner whereas no interaction was observed between rCsrA and the Mut probe (Fig 6B), indicating that GGA motif was critical for L. biflexa CsrA binding. Therefore, CsrA regulates flaB3 by binding to its upstream sequence at the GGA motif.

Overexpression of csrA in Leptospira spp.
To generate csrA overexpressing strains of L. biflexa and L. interrogans, we first overexpressed csrA of each strain under the control of the promoter of L. interrogans groES, which previously showed to be a strong promoter [53]. Consistent with the results observed in the complementation experiment, no colony was obtained for both L. interrogans and L. biflexa. In contrast, conjugation with the empty replicative plasmid resulted in hundreds of colonies for both L. interrogans and L. biflexa. We therefore overexpressed csrA under its native promoter, the promoter of the operon flgN-flgK-flgL-fliW-csrA. Numerous spectinomycin resistant colonies were then obtained from both leptospiral strains. RT-qPCR showed that the relative fold change of csrA was 3.70-fold higher in L. interrogans (Fig 7A) and 18.77-fold higher in L. biflexa (S6A Fig) compared to its parental WT strain.
The csrA overexpressing strain of both L. interrogans (WT+pMaORI_PcsrA li ) and L. biflexa (WT+pMaORI_PcsrA lb ) had a growth rate similar to WT in regular and 5-fold diluted EMJH (Fig 7B and S6B and S6C Fig). These results suggested that overexpression of csrA did not affect the growth of Leptospira.
To investigate the effect of csrA overexpression on motility, the soft agar assay and measurement of cell length and motility were performed. We found that the motility of the WT+-pMaORI_PcsrA lb was not deficient (S6D-S6F Fig), which was consistent with the results observed in ΔcsrA. In contrast, overexpression of csrA in L. interrogans had poor motility on soft agar ( Fig 7C); however, the cell length and velocity were not significantly different from WT (Fig 7D and 7E).

Overexpression of csrA had a distinct effect on flaB expression
The relative expression levels of the 4 flaB genes in csrA overexpressing strains of both L. biflexa and L. interrogans were investigated. In WT+pMaORI_PcsrA lb , flaB2 and flaB3 were significantly downregulated by 0.433-and 0.439-fold change (Fig 8A), respectively, which is in agreement with the data in ΔcsrA showing that CsrA acts as a repressor in L. biflexa. In contrast, the overexpressing strain of L. interrogans showed poor motility on soft agar plates and had a significant upregulation of flaB1, flaB2, and flaB4 by 3.02-, 2.10-, and 1.57-fold change,

PLOS ONE
Investigating the role of CsrA in Leptospira spp. respectively (Fig 8B), suggesting that CsrA might be involved in transcriptional activation of flagellin genes in L. interrogans.
Analysis of 5' untranslated region of L. interrogans flaBs revealed putative CsrA binding sites in flaB1, flaB2, flaB3, and flaB4 ( Fig 9A). Among 3 upregulated flaB genes, flaB4 is more likely a CsrA target because of the highest match of its upstream region to the consensus sequence ( Fig 9B and S3 Table). Secondary structure prediction of 120 nucleotides upstream of flaB4 (LIMLP_07475) revealed a GGA motif-containing hexaloop locating 6 nucleotides before the start codon ( Fig 9B). Therefore, flaB4 was selected for further binding studies. Recombinant CsrA protein of L. interrogans was produced (S5 Fig) and used for EMSA. As expected, rCsrA bound to flaB4 upstream in a dose-dependent manner (Fig 9C). The specificity was further confirmed by competitive EMSA (Fig 9D). Our results not only demonstrated that CsrA of L. interrogans is an RNA-binding protein but also showed flaB4 as a specific target.

Discussion
Leptospira spp. are ubiquitous bacteria found as free-living saprophytes in environmental water and soil or as pathogens excreted in the urine of asymptomatic hosts to cause disseminated infections in both humans and animals. Global gene regulators are required for their rapid adaptation to environmental changes. However, the knowledge of gene regulation in Leptospira remains limited. We found that csrA homolog, a well-characterized post-transcriptional global regulator, is present in all available leptospiral genomes. The leptospiral csrA is located inside an operon of genes involved in the flagellum biosynthesis (Fig 1A) like other bacteria [65]. Moreover, the csrA operon of both L. biflexa and L. interrogans are in synteny with csrA operons of other spirochete bacteria including B. burgdorferi and T. pallidum [66]. In gammaproteobacteria, non-coding RNA (ncRNA), such as csrB [67] and csrC [68], modulates CsrA function. In epsilon-proteobacteria and firmicutes that have no gene encoding ncRNA antagonist, FliW was reported as the protein antagonist of CsrA [69,70]. The Leptospira genomes do not possess csrB and csrC homologs but fliW is located adjacent to csrA (Fig 1A). Thus, FliW may function as the leptospiral CsrA antagonist. A CsrA-like ncRNA, which could regulate leptospiral CsrA activity, was also identified in L. biflexa [12].
While generation of csrA deletion was feasible in L. biflexa, we were unable to delete csrA in L. interrogans. Because targeted mutation particularly in pathogenic Leptospira is difficult and inefficient, a limited number of virulence genes have been confirmed [71]. Likewise, the present study could not successfully generate a deletion mutant in L. interrogans. Alternatively, csrA may have an essential role in the viability of L. interrogans but not in L. biflexa. The csrA mutant of Salmonella Typhimurium showed severe growth defect compared to its parental strain [47]. Moreover, CsrA might be toxic to Leptospira because we were unable to obtain any transconjugant in both L. biflexa and L. interrogans when csrA was fused to a strong promoter.
Metabolism is one of the common phenotypes regulated by CsrA in many bacteria [72][73][74][75][76]. For instance, E. coli CsrA regulates the carbon starvation gene, cstA, which plays a role in peptide transport during carbon starvation [76]. Another study reported a strong activity of CsrA during iron-limited condition [46]. In regular EMJH, growth curves of ΔcsrA and WT+-pMaORI_PcsrA lb were not different from those of their parental strains (Fig 3A and S6B Fig), suggesting that csrA was not essential for growth in L. biflexa in rich medium. However, the growth of ΔcsrA was defective under starvation condition compared to its parental WT strain. There was a relatively slow lag phase before reaching a similar growth rate as that in WT in the stationary phase, indicating that csrA is required in the early phase of growth when nutrients are limited. However, the complemented strain was unable to fully restore the phenotype. This is probably due to the overexpression (3-fold increase) of csrA in trans in ΔcsrA compared to the wild-type which may result in massive gene deregulation as shown by RNA-seq which had more than 500 differentially expressed genes ( Fig 4B). Our data suggested that CsrA is required for growth of L. biflexa under starvation so that they can survive in the environment where nutrients are limited.
Several reports showed that there was an alteration of transcriptomic profile in csrA mutant strains and those strains support CsrA as a global gene regulator (Table 1). In csrA mutant of enterohemorrhagic Escherichia coli O157:H7, 641 genes were upregulated, and 703 transcripts (~15% of total genes) were downregulated compared to its parental WT strain [26]. A total of 239 genes (13.4% of total genes) showed different expression in csrA mutant of B. burgdorferi compared to its parental WT [23]. Surprisingly, only 3 genes (<1% of total genes) were

PLOS ONE
differentially expressed in ΔcsrA of L. biflexa, but only 1, LEPBIa1872 (flaB3), was significantly upregulated more than 1.5-fold change and its expression level could be restored in the complemented strain indicating that csrA is a repressor of flaB3. A small number of genes were detected by RNA-seq probably because of different mechanisms employed by CsrA to regulate its gene targets [63]. For example, CsrA post-transcriptionally regulates its gene targets by affecting their mRNA stability [35,36,77] or it can regulate its targets without any change in the number of target transcripts [76,78]. As a result, such post-transcriptional control might not be detected by RNA-seq. Proteomic profiling may be necessary to investigate posttranslational effects of CsrA as well as to identify its targets. In addition, we found that ΔcsrA of L. biflexa grew slower than WT strain under starvation conditions. Thus, transcriptomic profiling of ΔcsrA under starvation may yield more information on the target genes.
In contrast to ΔcsrA, a higher number of genes were differentially expressed in the ΔcsrA+-pMaORI_PcsrA lb overexpressing CsrA in L. biflexa (Fig 4B). This finding might be a result of deregulation of other regulators as reported in some bacteria [63]. Presumably, CsrA exerts global regulation in L. biflexa when its expression reaches a substantial level. RNA-seq of WT demonstrated low csrA expression in rich medium (low total read/sample of csrA in S2 Table), therefore deletion of csrA might not result in major transcriptomic changes. In addition, other unknown factors might inhibit csrA expression at its native locus because the expression of csrA under its native promoter was significantly higher than WT (Fig 2B). Accordingly, we cannot exclude the possibility that CsrA is a global regulator in L. biflexa especially under the conditions that upregulate csrA. The impact of CsrA on expression of other genes under such conditions require further investigation.
Motility is one of the common traits regulated by CsrA. The alteration in motility affected by CsrA as well as the molecular mechanisms of CsrA that act on motility genes have been well documented in many bacteria [40,[79][80][81][82][83][84]. Flagellin genes have been reported as targets of CsrA in many bacteria. For example, CsrA bound to 5'untranslated regions of borrelial flaB at the consensus sequences overlapping the Shine Dalgarno sequence resulted in the translational block [51]. Hag protein, which shares~47% amino acid identical to leptospiral flaB3, is regulated by CsrA using the same mechanism as Borrelia [78]. Our transcriptome analysis showed that flaB3 is a potential target of CsrA. This was further confirmed by the presence of putative CsrA binding site in the promoter and gel shift assays which showed that there was a specific binding of rCsrA to the WT flaB3 5'untranslated region through the GGA conserved residues, which is consistent with a previous report [41]. In contrast, this finding is inconsistent with the results from the RNA-seq and EMSA results which showed that there were no differences in motility on soft agar, cell length, and velocity (Fig 3C-3E). While most bacteria harbor one flagellin component [85], Leptospira have 4 homologs of the flagellin FlaB in their genome [86]. The numbers of each FlaB in L. interrogans are approximately 14000, 2000, 300, and 3500 copies for FlaB1, FlaB2, FlaB3, and FlaB4, respectively [86]. Our RNA-seq results revealed that flaB4 (LEPBIa1589) was the most transcribed flaB, more than 3-fold compared to other flaB transcripts (S2 Table). Apparently, FlaB3 (LEPBIa_1872), which is regulated by CsrA, is not a major FlaB protein, which could explain the absence of change in the motility phenotype. It is possible that flaB2 is a target of L. biflexa CsrA because it was significantly upregulated in ΔcsrA and its expression was restored in the complemented strain (Fig 5A), but the interaction was not investigated in this study. The putative CsrA binding site of flaB2 shares high similarity to the flaB3 binding site and harbors GGA motif (Fig 5B and S3 Table). However, although flaB1 and flaB4 genes contain a possible CsrA binding site (Fig 5B and S3 Table), no transcriptional change was observed. Other cooperating factors might be required for gene regulation. It is noteworthy to mention that the flagellar expression and motility phenotype observed here occurred in a nutrient-rich culture medium.
Because we were unable to generate a csrA mutant strain of L. interrogans, an overexpressing csrA strain was constructed. A 4-fold increase of csrA in L. interrogans resulted in poor motility on soft agar (Fig 7C), suggesting that csrA may regulate the motility of L. interrogans. Due to no alteration in cell length or velocity, other pathways regulated by CsrA may be responsible for this phenotype. These results indicated the crucial function of csrA in the pathogenic strain because motility is known to be a virulence factor of Leptospira [87][88][89].
The upregulation of flaB in csrA overexpressing strain of L. interrogans is in contrast to the downregulation in csrA overexpressing strain of L. biflexa, suggesting the distinct mechanisms of csrA among leptospiral strains. As previously reported, the mechanism of CsrA on motility regulation can be distinct in different bacteria. For example, the csrA mutants in E. coli and S. Typhimurium were non-motile and CsrA positively regulated the master operon in flagellum biosynthesis, flhDC [40,90]. In contrast, RsmA, a CsrA homolog of Erwinia carotovora, negatively regulated flhDC, and the rsmA mutant was hypermotile [82]. In addition, CsrA in Salmonella regulated different motility genes compared to E. coli [38]. These results indicated that CsrA in different bacteria may have distinct effects on motility. For example, the same gene in different bacteria may have different mechanisms or control of different gene targets. Comparative transcriptomic and proteomic profiles of csrA-overexpressing L. biflexa and L. interrogans might give useful information on the global role of CsrA as well as different mechanisms of CsrA between these 2 species.
In conclusion, we characterized the role of CsrA in Leptospira spp. We found that csrA of the saprophyte L. biflexa is required for starvation response and repressed the expression of flaB3 (LEPBIa_1872) without any change in motility phenotype. L. biflexa CsrA may exert a global effect under certain conditions that upregulate csrA expression. In contrast, overexpression of CsrA in pathogenic L. interrogans resulted in poor motility and CsrA may be an activator of flaB1, flaB2, and flaB4 genes. This study suggested that pathways of gene regulation by CsrA may be different in bacteria belonging to the same genus, i.e., pathogenic and nonpathogenic Leptospira spp.