A family of genus‐specific RNAs in tandem with DNA‐binding proteins control expression of the badA major virulence factor gene in Bartonella henselae

Abstract Bartonella henselae is a gram‐negative zoonotic bacterium that causes infections in humans including endocarditis and bacillary angiomatosis. B. henselae has been shown to grow as large aggregates and form biofilms in vitro. The aggregative growth and the angiogenic host response requires the trimeric autotransporter adhesin BadA. We examined the transcriptome of the Houston‐1 strain of B. henselae using RNA‐seq revealing nine novel, highly‐expressed intergenic transcripts (Bartonella regulatory transcript, Brt1‐9). The Brt family of RNAs is unique to the genus Bartonella and ranges from 194 to 203 nucleotides with high homology and stable predicted secondary structures. Immediately downstream of each of the nine RNA genes is a helix‐turn‐helix DNA‐binding protein (transcriptional regulatory protein, Trp1‐9) that is poorly transcribed under the growth conditions used for RNA‐seq. Using knockdown or overexpressing strains, we show a role of both the Brt1 and Trp1 in the regulation of badA and also in biofilm formation. Based on these data, we hypothesize that Brt1 is a trans‐acting sRNA that also serves as a cis‐acting riboswitch to control the expression of badA. This family of RNAs together with the downstream Trp DNA‐binding proteins represents a novel coordinated regulatory circuit controlling expression of virulence‐associated genes in the bartonellae.

cells found in close association with aggregates of B. henselae (Dehio, 2005). In addition, B. henselae is a common cause of fever of unknown origin and endocarditis and has been detected in vegetative masses on infected heart valves from patients with culture-negative endocarditis (Edouard, Nabet, Lepidi, Fournier, & Raoult, 2015). It has also been reported to grow as large aggregates in vitro and to form biofilms (Kyme, Dillon, & Iredell, 2003).
The induction of this unique angiogenic host response as well as the aggregative growth of this bacterium is known to involve the highmolecular weight trimeric autotransporter adhesin (TAA) BadA. BadA is a large surface-exposed filament-shaped protein of 3036 amino acids and a length of 240 nm (Muller et al., 2011). The role of BadA in B. henselae pathogenesis has been studied extensively in vitro and is critical for autoagglutination, adhesion to host cells, and extracellular matrix proteins such as collagen, inhibition of phagocytosis, and induction of angiogenesis (Riess et al., 2004). In vitro studies show that the expression of badA correlates with a proangiogenic cell response via activation of HIF-1 and NF-kB, and via the secretion of vascular endothelial growth factor (VEGF) and IL-8 (Kempf et al., 2001(Kempf et al., , 2005McCord, Burgess, Whaley, & Anderson, 2005;McCord, Resto-Ruiz, & Anderson, 2006;Riess et al., 2004). Despite the critical role of BadA in the pathogenesis of B. henselae, very little is known about how badA is regulated and why some strains express high levels of badA and have an autoaggregative phenotype, while others express little if any badA and are not autoaggregative (Riess, Raddatz, Linke, Schafer, & Kempf, 2007).
As a zoonotic bacterium, B. henselae must be able to adapt to very diverse environments such as the cat flea vector where the temperature is low and, after a blood meal, hemin reaches toxic levels. In contrast, the vertebrate host temperature is higher and hemin availability is limited. Very little is known about how B. henselae adapts its gene expression profiles to efficiently transition from the conditions of the cat flea vector to the mammalian hosts. Reports have proposed a role for several different regulatory proteins in various aspects of host adaptation in B. henselae and B. quintana. (Battisti, Sappington, Smitherman, Parrow, & Minnick, 2006;Quebatte, Dick, Kaever, Schmidt, & Dehio, 2013;Quebatte et al., 2010;Roden, Wells, Chomel, Kasten, & Koehler, 2012). Recently, a novel alternate sigma factor activation mechanism has been shown to induce a general stress response in B. quintana . This general stress response was shown to regulate genes in response to temperature that corresponds to a shift from the vector to humans and vice versa . This same general stress response has been described in B. henselae and shown to play a role in regulating badA (Tu, Lima, Bandeali, & Anderson, 2016). However, a role of RNA in gene regulation and host adaptation in Bartonella remains unexplored despite the well-described role of RNAs in rapidly modulating gene expression profiles in other bacteria.
RNAs can regulate expression in bacteria by acting in trans on genes distal to the RNA coding region or in cis by acting on adjacent genes either downstream or on the opposite strand. Small regulatory RNAs (sRNAs) function by binding target mRNAs, binding proteins, or by targeting and degrading specific mRNAs (see [Michaux, Verneuil, Hartke, & Giard, 2014] for review of sRNA functions). sRNAs may be cis-acting or trans-acting and their biological roles have been shown to include regulation of metabolism, growth, adaptation to stress, or pathogenesis (Gong et al., 2011;Koo, Alleyne, Schiano, Jafari, & Lathem, 2011;Toledo-Arana, Repoila, & Cossart, 2007;Vogel, 2009;Waters & Storz, 2009). In contrast, cis-acting RNAs including riboswitches involve structured domains found on the noncoding regions of the mRNA being regulated. Binding of ligands or metabolites is the primary mechanism resulting in the formation of functional domains in riboswitches. Thus, the presence and concentration of these ligands are the triggers that activate the riboswitch. This sensing and response to ligands typically results in changes in mRNA secondary structure, resulting in transcriptional read-through, altered stability of the mRNA, or changes in translation efficiency thereby effecting modified gene expression (see (Winkler & Breaker, 2005) for review of riboswitches).
Here, we describe a highly transcribed family of nine small RNAs (Bartonella regulatory transcripts, that are unique to Bartonella species. All nine of these RNAs are immediately upstream of a coding region for a protein that is annotated as a helix-turn-helix DNA-binding protein/transcriptional regulator (transcriptional regulatory protein, Trp1-9). We show a role for the most highly transcribed member of this RNA family, Brt1, in regulating badA and we further show that the cognate Trp1 positively controls badA expression. We hypothesize that the Brt family of RNAs are capable of functioning as a two-tiered system including both cis-acting and trans-acting roles to coordinate regulation of badA, which is responsible for autoaggregation, host cell attachment, the proangiogenic host response, and biofilm formation by B. henselae.

| Bacterial strains and growth conditions
The Houston-1 strain of B. henselae which was isolated from an HIV-infected patient (Regnery et al., 1992) was used for all experiments and was cultured on heart infusion agar supplemented with 1% bovine hemoglobin (chocolate agar) and incubated for 3-4 days at 37°C in the presence of 5% CO 2 . For some experiments, bacteria were grown in Schneider's liquid media (Sigma Aldrich, S9895) supplemented with 10% fetal bovine serum for 3-4 days at 37°C in the presence of 5% CO 2 as previously described (Riess et al., 2008).
For experiments using strains harboring pNS2-derived plasmids, the culture medium was supplemented with 50 μg/ml kanamycin (Gillaspie et al., 2009). The construction of a nonpolar in-frame deletion mutant of badA (BH01510) has been previously described (Lima, Cha, Amin, Smith, & Anderson, 2014). The ∆badA mutant served as a control for experiments examining badA expression and biofilm formation. All B. henselae strains used in this study are described in Table 2. Growth curves indicate that there are no growth defects for the strains used in this study ( Figure S2). All manipulations of B. henselae have been approved by the USF Institutional Biosafety Committee.

| RNA isolation
B. henselae Houston-1 was cultured in Schneider's liquid media at 37°C, 5% CO 2 for 72 hr, and collected by centrifugation for RNA extraction. The bacteria were exposed to RNAProtect Cell Reagent (Qiagen, 76506) and pelleted and frozen at −80°C. The pellets were thawed at 4°C before RNA isolation. Total RNA was extracted using the RNeasy Mini Kit (Qiagen, 74104). Following extraction, RNA was treated with DNase (Thermo Fisher Scientific, AM1907). RNA quality was analyzed using the Agilent 2100 Bioanalyzer.  (Alsmark et al., 2004).

| RNA-seq
Data analysis was carried out using the CLC Genomics Workbench platform (CLC bio).

| Northern blot
A 29 base oligonucleotide probe (Brtprobe , Table S6) was designed to base-pair to a conserved sequence found in all nine Brts. For detecting antisense Brt1 RNA, a second probe which hybridized to RNA transcribed from the opposite (antisense) strand was used (asBrtprobe, Table S6). The probe was labeled with [γ-32 P]ATP using T4 polynucleotide kinase (Promega, M4101). Unincorporated [γ-32 P]ATP was then removed using a QIAquick Nucleotide Removal Kit (Qiagen, 28304). B. henselae strains were cultured in Schneider's medium or Schneider's medium with 50 μg/ml kanamycin for 72 hr and total RNA was isolated with the RNeasy Mini Kit and treated with Turbo DNase. Four microgram of DNase-treated total RNA was separated on a 10% denaturing polyacrylamide mini gel. The RNA was transferred to an Immobilon-Ny + Membrane (Millipore). Following transfer, the RNA was UV cross-linked to the membrane. Prior to hybridization, the membrane was prehybridized in ULTRAhyb-Oligo Buffer (Thermo Fisher Scientific, AM8663) at 42°C for 1 hr. The labeled probe was denatured by heating at 95°C for 5 min and added to the membrane to allow for hybridization overnight at 42°C in a hybridization oven.
After hybridization, the membrane was washed for 15 min, each at 42°C sequentially with: 2X SSC, 1X SSC, and 0.5 SSC. The membrane was then exposed to a phosphorimager screen overnight. The image was developed using a Typhoon 9410 (GE HealthCare).

| Bioinformatic analysis
brt1-9 transcription start and stop sites were determined by RNA-seq according to the start and end of the RNA transcripts. Brt nucleotide and Trp nucleotide and amino acid sequences were obtained from the B. henselae Houston-1 genome sequence from NCBI (accession # BX897699). Nucleotide sequence alignment and percent identity was carried out using the multiple sequence alignment tool (MUSCLE) from EMBL-EBI (Edgar, 2004). RNAfold was used to predict the secondary structures of individual Brt nucleotide sequences (Lorenz et al., 2011).
The minimum free energy was calculated based on predicted secondary structure to indicate the stability of the predicted structure.
Individual Brt nucleotide sequences were searched against the NCBI database to find related sequences in other bacteria species. The CLC Sequence View 6 (CLC bio) software was used to align Trp amino acid sequences and for phylogenetic analysis. Target predictions for individual Brts were carried out using RNApredator (Eggenhofer, Tafer, Stadler, & Hofacker, 2011) and IntaRNA (Busch, Richter, & Backofen, 2008;Wright et al., 2014).

| Antisense Brt1 RNA knockdown plasmid construction
PCR primers (Brt1asR, Brt1asF, Table S6) were expected to amplify the entire Brt1 gene sequence from B. henselae Houston-1 genomic DNA. The resulting amplicon was directionally ligated into the pN-S2Trc plasmid using the BamHI and XbaI restriction sites such that high levels of antisense RNA complementary to Brt1 would be transcribed from the Trc promoter in that plasmid. The plasmids were transformed into competent DH12S E. coli cells and selected for colonies harboring the plasmid with 50 μg/ml kanamycin. PCR was used to screen for successful ligation of the brt1 gene into the pNS2Trc plasmid in the expected orientation. After confirmation by sequencing, the plasmid was then electroporated into B. henselae Houston-1 as previously described (Resto-Ruiz, Sweger, Widen, Valkov, & Anderson, 2000), to produce strain Bhh72 (Table 2). qRT-PCR with asBrt1 RNAspecific primers showed high levels of asBrt1 in Bhh72 compared to the control, confirming transcription of this antisense RNA (Figure 6a).

| Overexpression of Trp1 plasmid construction
PCR primers specific for the trp1 gene were designed to amplify the gene from genomic B. henselae Houston-1 DNA from the second codon of the gene to beyond the stop codon (Trp1F, Trp1R, Figure S6).
The resulting amplicon was ligated into the BamHI and XbaI restriction sites of pNS2Trc to create an overexpressing Trp His-tag fusion protein. The plasmids were transformed into competent DH12S E. coli cells and selected for colonies harboring the plasmid with 50 μg/ml kanamycin. PCR was used to screen for successful ligation of the trp1 gene into the pNS2Trc plasmid in the intended orientation that would | 3 of 14 allow overexpression of the gene. The plasmids were then electroporated into B. henselae Houston-1 as previously described (Resto-Ruiz et al., 2000). It should be noted that the B. henselae genome does not contain the lactose repressor gene, and hence, the P trc promoter of pNS2P Trc is constitutive and not inducible in B. henselae (Gillaspie et al., 2009). Expression of 6XHis-tagged Trp1 was confirmed in the resulting strain (Bhh73, Table 2) by western blot with mouse anti-His tag antibody (Figure 5a).

| qRT-PCR of B. henselae genes
The antisense Brt1 strain and the overexpressing trp1 strain of B. henselae were used to analyze the transcription of the badA gene.
Bacteria were grown in Schneider's medium for 3 days and cells were collected by centrifugation and total RNA was extracted, treated with Turbo DNase (Thermo Fisher Scientific, AM1907), and converted into cDNA using the iScript cDNA synthesis kit (BioRad, 1708891). qRT-PCR was performed in 25 μl reaction volumes that include 12.5 μl of the 2X Maxima SYBR Green/Fluorescein qPCR kit (Thermo Fisher Scientific, K0241), 300 nmol of each gene-specific primer, and 2 μl cDNA. All reactions were performed in triplicate and 50S ribosomal protein L4 (rplD) was used as the reference gene for normalization.
The qRT-PCR cycling parameters included 95°C for 3 min, 40 cycles of 95°C for 10 s, and 60°C for 30 s, followed by 95°C for 45 s and 55°C for 1 min. Melting curve analysis was performed to confirm that no primer dimers were amplified. Results were analyzed using the comparative C T method (Schmittgen & Livak, 2008). The primer pairs used for qRT-PCR of the Brts, badA, and trp1 are shown in Table S6.

| Dot-blot assay for BadA
To further determine if qRT-PCR for transcription of the badA gene could be correlated with BadA protein production, we performed dot-blots using rabbit polyvalent antibody to BadA (a generous gift from Dr. Volkhard Kempf). Since BadA is a surface trimeric autotransporter adhesin, it exists in B. henselae as a high-molecular weight (>1 million daltons) fibrillar complexes (Riess et al., 2004). The individual subunits of BadA are approximately 330,000 daltons, making gel resolution and protein transfer to membranes inefficient and western blotting impractical. For this reason, dot-blots of intact B. henselae cells grown for 3 days were performed. Two microliters containing approximately 1 × 10 6 bacteria (standardized to 2.0 μg protein) in PBS was spotted onto a nitrocellulose filter. The membrane was then blocked with PBS +0.05% Tween 20 containing 5% skim milk.
Antibody diluted 1:200 in PBS +5% skim milk was added and the filters were incubated for 2 hr at room temperature and washed four times in PBS +0.05% Tween 20. The resulting filters were incubated in goat anti-rabbit conjugated with horseradish peroxidase (KPL, Gaithersburg, MD) diluted 1:10,000 in PBS +5% skim milk for 1 hr at room temperature. After three washes with PBS +0.05% Tween 20, the filters were incubated with ECL luminescent substrate (Thermo Scientific) for 1 min and detected using a Bio-Rad Chemidoc XRS Imaging system.

| Biofilm assay
A standard microplate adherence assay was used to measure B. henselae biofilm formation. This method is based on the standard crystal violet microplate assay for gram-negative bacteria (O'Toole, 2011) that has been adapted for B. henselae (Kyme et al., 2003). Briefly, 5 × 10 6 bacteria in a volume of 100 μl was added to each well of a 96-well polystyrene plate (Costar, Corning, NY) and the plates were incubated for 3 days at 37°C with 5% CO 2 . After incubation, the plates were washed with water and subsequently stained with 125 μl of a 0.1% crystal violet solution for 15 min. After washing, the plates were allowed to dry for 3 hr and the stain was extracted with 125 μl 30% acetic acid for 15 min. The stain solution was transferred to a new 96-well plate and the OD 550 measured using a plate reader (BioTek, Winooski, VT).

| Statistical analysis
The means ± standard errors were presented in the data. SigmaPlot software (Systat Software, San Jose, CA) was used for statistical analysis using the t-test, with a p-value of <.05 being considered statistically significant.

| RNA-seq reveals nine unannotated highly transcribed short RNAs
The transcriptome of B. henselae, grown under standard growth conditions in Schneider's liquid medium at 37°C with 5% CO 2 for 72 hr as previously described, (Riess et al., 2008) was examined using RNA-Seq. Based on growth curves for the strains used in these studies, this correlates with late log-phase growth. While the depth of coverage was low due to incomplete rRNA removal, it was not difficult to discern highly transcribed genes. Surprisingly, several small RNAs that did not map to know genes, were shown to be highly transcribed.
Specifically, a group of nine short RNAs, that are not annotated in the published genome (Alsmark et al., 2004) or subsequently analyzed genome (Tatusova, Ciufo, Fedorov, O'Neill, & Tolstoy, 2014), were noted to be highly transcribed ( Figure 1a). These RNAs varied in size from 194 to 203 nucleotides and were transcribed from noncoding regions of the B. henselae genome that did not map to or overlap any known genes. The RNAs were designated Brt1 through Brt9 (for Bartonella regulatory transcript) to identify each noncoding RNA. The genome location as well as the start and stop of transcription for all nine Brts can be found in Table S1. Specific details of the RNA-seq results are found in File S1.

| The Brt RNA loci cluster in a small region of the B. henselae genome
The nine Brt RNA loci were interspersed throughout a relatively small region (277 kb) of the B. henselae genome (Figure 1b, red oval). This region has been described as highly plastic and very likely contains several horizontally acquired mobile genetic elements (Lindroos et al., 2006). This region includes prophage genes as well as the pathogenicity island encoding the VirB type IV secretion system (Alsmark et al., 2004), which has been shown to play an important role in interaction with host cells and to play a role in the pathogenesis of B. henselae (Schulein et al., 2005). In a more recent study, the authors describe this region of the genome as a plastic "novel region" enriched in repeats with highly expressed genes for which proteins could not be detected. The authors of that study further suggested that this region contains highly transcribed genes that do not represent a bona fide protein-coding ORF, (Omasits et al., 2013), an observation consistent with noncoding regulatory RNAs. Since the genome size of B. henselae is just over 1.9 Mbp, the region harboring the Brts represents only about 15% of the total genome. This clustering of the Brts taken together with their multiple copy number is strongly suggestive of a "hot spot" area of the genome undergoing active gene duplication or horizontal gene acquisition.

| The nucleotide sequence of the nine Brt RNAs is highly conserved and are predicted to form complex secondary structures
Nucleotide sequence alignment of Brt1-Brt9 reveals a high level of conservation across the length of Brt1-9 ranging from 74% to   (Table 1). Thus, these novel RNAs appear to be genus-specific, but are not found in all Bartonella species.

| Brt1 is among the most highly transcribed regions of the Bh genome
Based on analysis of the quantitative data from the RNA-seq, the RPKM (reads/kilobase/million mapped reads) were far higher (>10fold) for Brt1 than for any of the other Brt family of RNAs indicating that more copies of this RNA were present than for Brt2-Brt9 combined ( Figure 3a). In fact, brt1 was among the most highly transcribed regions of the B. henselae genome, with the number of reads among the highest four loci (excluding rRNAs) in the Bh transcriptome (File S1). The presence of Brt RNAs has also been confirmed by northern   (Table S2). However, none of these gene copies are found adjacent to a Brt RNA like those identified with RNAseq. The number of trp genes found in various Bartonella species and whether they are linked to a brt sequence on the genome is shown in Table S3.

| The Trps belong to a family of helix-turn-helix DNA-binding proteins
The region located immediately downstream of each brt codes for a family of small DNA-binding proteins (Trps). The deduced amino acid  (Table S2). It is interesting to note that phylogenetic analysis of the deduced amino acid sequence for all 24 putative Trps (or Trp-like proteins) shows that all eight of the proteins that are encoded downstream of the Brt RNAs cluster together, branching out from a single node on the tree (Figure 4c). Locus BH02210 encodes a protein that represents the closest member that is not preceded upstream by a Brt RNA. The remaining 15 Trp-like proteins are more distally related than those encoded by trp1-7 and trp9. They contain the conserved helix-turn-helix putative DNA-binding domain but are poorly conserved throughout the remainder of the protein.

| Overexpression of trp1 results in upregulation of badA and increased biofilm formation
Our RNA-seq data showed that the Brt RNAs are highly transcribed, in contrast to the downstream Trp-encoding genes, which are uniformly poorly transcribed (Figure 1a). This observation indicates transcriptional termination immediately upstream of the trps, at least under the conditions used for this study. Furthermore, the complex secondary structure noted at the 3′ end of the Brt1 RNA (Figure 3d) bears resemblance to those seen in riboswitches raising the possibility of a cis-acting regulatory mechanism preventing transcription of

(b) (a)
the downstream transcriptional regulatory protein genes. Since the function of these Trp proteins is unknown in B. henselae and we hypothesize that they are potentially regulated by the Brt RNAs via a riboswitch mechanism, it is crucial to determine whether they would affect the expression of key virulence factors of B. henselae. To address this possibility, we cloned the downstream trp1 gene into a plasmid with a strong constitutive promoter. We chose plasmid pNS2P Trc since it contains the trc promoter which results in a high level of constitutive transcription in B. henselae since there is no lactose repressor gene (lacI) present in the genome (Gillaspie et al., 2009). For this reason, we cloned trp1 into pNS2P Trc to generate strain Bhh73 that constitutively overexpresses trp1 (description in Table 2). Bhh73 was shown to express trp1 efficiently with Trp1 easily detected by western blot using antibody to the 6XHis-tag on the amino terminus of the fusion protein ( Figure 5a). Bhh73 was also noted to have a colony phenotype that is more adherent than the pNS2Trc vector control suggesting in-

| Brt1 knockdown results in upregulation of badA and increased biofilm formation
Since high level expression of the Brt family of RNAs, their size, their complex secondary structure, and their location in noncoding regions of the B. henselae genome are suggestive of trans-acting sRNA, we examined this possibility using an antisense (as) approach to block Brt function. Specifically, to determine if Brt1 functions as a small regulatory RNA that plays a role in gene regulation, we constructed an antisense (as) construct/strain designed to base-pair and destabilize the Brt1 RNA. Since trans-acting RNAs typically act by base-pairing, we reasoned that asBrt1 RNA would block the function of Brt1 and possibly also destabilize the RNA and make it prone to degradation.
Using primers that would only detect vector sequences found in the antisense Brt1 RNA, transcription of high levels of antisense Brt1 ( Figure 6a) was noted in the strain of Bh harboring the asBrt1 plasmid (Bhh72, description in Table 2). A northern blot using probes specific for both antisense-Brt1 RNA and Brt1 itself shows that Bhh72 produces antisense Brt1 (Figure 6f, lane 2) but that Brt1 RNA appears to be degraded (Figure 6g, lane 2)  To examine possible mRNA targets of Brt1, the RNApredator and IntaRNA prediction programs were run using Brt1 as the query and the genome of the Houston-1 strain of B. henselae as the target (Busch et al., 2008;Eggenhofer et al., 2011). Multiple mRNA targets were predicted with the Brt1:mRNA duplex formation as stable as −11.1 kcal/mol for the bepC (BH13400) mRNA (Table 3). Additional B. henselae genes among the top predicted targets include a regulatory protein (BH11940) containing a conserved domain known to activate sigma factor 54 (Studholme & Dixon, 2003), the truncated copy of the badA gene (BH01490) as well as two copies of the filamentous hemagglutinin genes (BH06670 and BH07150) and four ABC transporter genes (BH00170, BH01900, BH12230, and BH15040). In addition, genes encoding VirB type IV secretion system components VirB6 (BH13300) and VirB9 (BH13330) were predicted targets of Brt1 (Table 3). Three histidine kinases including BatS (BH00610), an orphan histidine kinase/response regulator (BH09640) and the general stress response kinase (BH13820) were also among the predicted targets for Brt1 (Table 3). The full-length copy of the badA gene (BH01510) mRNA was predicted to interact with Brt1 but with a predicted ∆G of −1.73 kcal/ mol indicating low stability of this interaction. The top 100 predicted mRNA targets from each algorithm are shown in Tables S4 and S5.

| DISCUSSION
The genus Bartonella consists of over 30 species of bacteria found in a wide range of animal hosts with most thought to be transmitted by an arthropod vector (Buffet, Kosoy, & Vayssier-Taussat, 2013 and incidental transmission of this bacterium to humans results in disease (Anderson & Neuman, 1997). Thus, Bartonella species must be able to quickly adapt to the drastically different conditions associated with transmission from a vertebrate animal to an arthropod vector and back into another host. Several regulatory proteins of Bartonella have been proposed to play a role in this response (Battisti et al., 2006;Quebatte et al., 2010Quebatte et al., , 2013Roden et al., 2012). The adaptive response is also thought to be controlled by an alternate sigma factor RpoE of the general stress response system found in both B. quintana  and B. henselae (Tu et al., 2016). Genes in the B. henselae general stress response regulon include the hemin-binding proteins and the badA adhesin (Tu et al., 2016). In addition, to this general stress response system, twocomponent regulatory systems have also been described in Bartonella species. The BatR/S two-component system has been shown to control the VirB type IV secretion system of B. henselae and the asso-  Table 2 with Bhh17 (ΔbadA) and Bhh13 (Houston-1 pNS2P Trc ) as controls F I G U R E 6 Expression of badA/BadA and biofilm formation in Bhh72 (antisense-ASBrt1, Bh/pNS2P Trc asBrt1). (a) qRT-PCR with primers specific for antisense Brt1 RNA showing relative copy number (RCN) compared to reference rplD mRNA, (b) Dot-blot using rabbit anti-BadA, (c). Densitometry of dot-blot in panel (b and d) qRT-PCR with badA-specific primers, (e) biofilm formation. (f and g) Northern blots with Bhh13 RNA (lane 1) and Bhh72 RNA (lane 2). The probes used are specific to antisense Brt1 (asBrtProbe, panel f) and Brt RNA (BrtProbe, panel g). Strains are as described in Table 2 with Bhh17 (ΔbadA) and Bhh13 (Houston-1 pNS2P Trc ) as controls | 9 of 14 between species; the human-specific species either harbor none (B.  (Table 1).
The copy number generally correlates well with genome size and may be indicative of gene duplication by transposition or horizontal gene transfer. This is further supported by the presence of the Brt-coding sequences in only a small part of the B. henselae genome that has been described as highly plastic and that very likely contains several horizontally acquired mobile genetic elements (Lindroos et al., 2006).
Interestingly, all nine Brt RNAs are found upstream of genes annotated as XRE transcriptional regulators that we designated as trps in B. henselae. The fact that this gene arrangement is highly conserved suggests that both the Brt RNA gene and the trp gene may have been acquired and/or duplicated together. The intergenic region between each of the nine Brts and its cognate trp is very short ranging from 10 to 18 nucleotides. This operon-like gene arrangement, wherein the first gene is an RNA with no promoter region located before the sec- that is thought to bind the major groove of DNA (Wang et al., 2014). It is interesting to note that XREs in Pseudomonas aeruginosa have been T A B L E 3 Selected predicted targets of Brt1 RNA a shown to regulate genes controlling the biogenesis of type IV pili, flagella, and biofilms (Wang et al., 2014) RNApredator and IntaRNA algorithms, several genes were found including the truncated version of badA (BH01490 , Table 3). However, the full-length badA gene (BH01510) was not among the top targets and was predicted to only weakly interact with Brt1, raising the possibility that Brt1 acts on badA by an indirect mechanism. Alternatively, it is also possible that an RNA-binding protein, such as Hfq, is involved in these interactions and allows Brt1 to bind a target mRNA in the absence of extensive base-pairing. It should also be noted that discrepancies were observed between these two sRNA target prediction algorithms and while some targets were in common (PhaC pH adap- Since B. henselae has been described to autoagglutinate in a manner that is dependent upon expression of badA (Riess et al., 2007), we elected to extend this study to include the formation of biofilms.
BadA is found in the genomes of all characterized Bartonella species (O'Rourke, Schmidgen, Kaiser, Linke, & Kempf, 2011). While it was not the specific focus of this study, this is the first report describing the requirement of BadA in biofilm formation by B. henselae. Our B. henselae Houston-1 badA deletion mutant (Bhh17) consistently demonstrated an impaired ability to form biofilms in our assay and served as a negative control for the biofilm assay. Considering the well-characterized role of BadA in autoagglutination, it is reasonable to expect that the surface location and adherent nature of BadA results in a role for this protein in biofilm formation. The requirement for BadA to form biofilms makes it tempting to speculate on the role of this protein in establishing the vegetative masses described in patients with infective endocarditis caused by B. henselae (see (Edouard et al., 2015) for review). BadA belongs to the TAA-family of proteins that are found in many other gram-negative bacteria including Escherichia coli, Yersinia enterocolitica, Haemophilus influenzae, and Neisseria meningitidis (see (Linke, Riess, Autenrieth, Lupas, & Kempf, 2006) for review).
The long observed phenomenon of autoadherence and the sticky and adherent colony phenotypes reported in fresh clinical isolates of B. henselae (and very likely other Bartonella species) is thought to be due to the variable expression of badA (Riess et al., 2007). The Brt/Trp regulatory system described in this report may explain the change in colony phenotype upon repeated culture of B. henselae in the laboratory. It would be interesting to examine fresh isolates of B. henselae with the same strain that has been repeatedly passaged to examine not only badA expression but also brt1 and trp1 transcript levels.
Recently, the presence of BadA in B. henselae strains was shown to adversely affect the expression of the VirB type IV secretion system genes and the function of the cognate effectors (Lu et al., 2013). In the presence of full-length BadA protein, VirB expression was reduced and it could be restored upon deletion of badA (Lu et al., 2013). Thus, the presence of these two pathogenicity factors appears to be coordinated. It would be interesting to examine the role of Brt RNAs and Trp proteins in coordinating the expression of badA and the genes encoding the VirB type IV secretion system. Interestingly, some of the VirB type IV secretion system components including bepC, virB6, and virB9 were predicted targets of Brt1. Of note, the B. bacilliformis genome does not contain the Brt RNA genes and also lacks the VirB type IV secretion system. Thus, Brt RNAs are absent in one of the species of Bartonella that does not have a need to coordinately regulate these two systems.