The curli regulator CsgD mediates stationary phase counter-silencing of csgBA in Salmonella Typhimurium

Summary

Integration of horizontally acquired genes into transcriptional networks is essential for the regulated expression of virulence in bacterial pathogens. In Salmonella enterica, expression of such genes is repressed by the nucleoid-associated protein H-NS, which recognizes and binds to AT-rich DNA. H-NS-mediated silencing must be countered by other DNA-binding proteins to allow expression under appropriate conditions. Some genes that can be transcribed by RNA polymerase (RNAP) associated with the alternative sigma factor σ^S or the housekeeping sigma factor σ⁷⁰ in vitro appear to be preferentially transcribed by σ^S in the presence of H-NS, suggesting that σ^S may act as a counter-silencer. To determine whether σ^S directly counters H-NS-mediated silencing and whether co-regulation by H-NS accounts for the σ^S selectivity of certain promoters, we examined the csgBA operon, which is required for curli fimbriae expression, and is known to be regulated by both H-NS and σ^S. Using genetics and in vitro biochemical analyses, we found that σ^S is not directly required for csgBA transcription, but rather up-regulates csgBA via an indirect upstream mechanism. Instead, the biofilm master regulator CsgD directly counter-silences the csgBA promoter by altering the DNA-protein complex structure to disrupt H-NS-mediated silencing in addition to directing the binding of RNAP.

Newman – csgB in Salmonella Abbreviated Summary

H-NS silences transcription of the csgB promoter during log phase through the formation of a nucleoprotein filament that occludes RNA polymerase. The transcription factor CsgD, produced during stationary phase, binds to the csgB promoter and bends the DNA to relieve H-NS silencing and allow RNA polymerase coupled with either σ^S or σ⁷⁰ to transcribe the csgB gene.

Introduction

Horizontal gene transfer is a primary driver of bacterial evolution, allowing bacteria to rapidly acquire new traits (Ochman et al., 2000), and can account for up to a third of all genes in some bacterial species (Boto, 2010). However, newly acquired genes must be incorporated into established regulatory networks in order to allow coordinated expression and recipient cell benefit while avoiding negative impacts on fitness. In a process called “xenogeneic silencing” (Navarre et al., 2006) newly-acquired genes are silenced to protect a cell from potentially detrimental effects of unregulated expression. This silencing can be alleviated under appropriate conditions by other regulatory proteins, via a process termed “counter-silencing” (Navarre et al., 2007; Will et al., 2015). Due to their low DNA-sequence specificity, nucleoid-associated proteins (NAPs) are well-suited to the regulation of new horizontally-acquired genes, as they are able to interact with the wide array of promoters that might be acquired by a host cell via horizontal gene transfer. One NAP that has emerged to have an essential role in this process is H-NS.

At approximately 20,000 copies per cell, H-NS is one of the most abundant NAPs (Ali Azam et al., 1999), binding DNA with relatively low sequence specificity, and preferentially interacting with AT-rich DNA by way of an AT-hook motif (Gordon et al., 2011). Upon nucleation at high affinity sites, the protein oligomerizes outwards along a promoter region, controlling over 200 genes in Salmonella enterica serovar Typhimurium (Lucchini et al., 2006, Navarre et al., 2006). This formation of a filament inhibits gene transcription in a variety of ways (Grainger, 2016). Because horizontally-acquired DNA is typically AT-rich relative to the host genomic average, H-NS is able to preferentially silence new genes, acting as a xenogeneic silencer (Navarre et al., 2006, Lucchini et al., 2006).

S. Typhimurium has served as a useful model organism for understanding the regulatory integration of horizontally-acquired genes, as many of its virulence genes are horizontally-acquired (Mills et al., 1995; Shea et al., 1996). This orally-acquired Gram-negative pathogen is of global importance, most commonly causing enteric fever (typhoid) and enterocolitis/diarrhea, resulting in as many as 1.3 billion cases of infection and over 3 million deaths annually (Coburn et al., 2007). One horizontally-acquired locus in Salmonella is the csg region (Bäumler et al., 1997), which is important for both virulence and bacterial survival in the environment (Tükel et al., 2005).

The csg region is present in many enteric pathogens, including many Escherichia and Salmonella spp., and encodes multi-subunit protein polymers known as curli fimbriae, which promote community behavior and host colonization. In the environment, curli are important for adhesion to surfaces, cell aggregation, and biofilm formation, and are induced under conditions of low temperature and osmolarity, as well as in stationary phase (Olsén et al., 1993). Organized as two divergently transcribed operons, the csg region consists of csgBA, which encodes the major structural subunits of the fibrous polymers, and csgDEFG, which encodes the major biofilm transcriptional regulator (CsgD) and the protein components important for secretion and assembly of the curli structures (CsgEFG). Although regulation of this operon has been extensively studied in E. coli, significant differences in CsgD binding at the csgBA promoters of E. coli and Salmonella enterica (Ogasawara et al., 2011) and limited conservation of the intergenic region (Figure S1) warrant further characterization of csg regulation in Salmonella.

Like many genes expressed in stationary phase, the csg genes are regulated by the alternative sigma factor, σ^S (Olsen et al., 1993). σ^S is present at relatively low concentrations in exponentially growing cells, but directly or indirectly controls more than 500 genes during stationary phase or during periods of stress (Battesti et al., 2011). Promoter recognition sequences for σ^S and the housekeeping sigma factor σ⁷⁰ are nearly identical (Hengge-Aronis, 2002). However, σ^S is essential for the transcription of certain promoters in vivo. The csgBA operon in E. coli is σ^S-dependent but transcribed independently of σ^S in an hns mutant strain, and can be transcribed by either σ^S or σ⁷⁰ in vitro (Arnqvist et al., 1994). Studies of other σ^S-dependent virulence genes, such as hdeAB in E. coli, (Shin et al., 2005) and spv (Robbe-Saule et al., 1997) and csiD (Marschall et al., 1998) in S. Typhimurium, have revealed a similar pattern, wherein σ^S is only required in vivo if H-NS is present. Based on these observations, it has been suggested that H-NS is a determinant of sigma factor specificity, owing to a unique ability of σ^S to overcome H-NS mediated silencing (Typas et al., 2007). Several studies have investigated a possible counter-silencing role for σ^S (Colland et al., 2000, Marschall et al., 1998, Shin et al., 2005, Olsén et al., 1993), while other researchers have posited that this sigma-factor may indirectly regulate csgBA expression by acting upstream in a complex regulatory cascade (Weber et al., 2006), with relief of H-NS silencing occurring through alternative mechanisms.

Studies looking to examine the relationship between H-NS and RNAP have suggested multiple possible mechanisms by which H-NS represses transcription. Some observations have indicated that H-NS multimers occlude the promoter and prevent RNAP access (Lucchini et al., 2006, Will et al., 2014). These studies have correlated the linear filament or “stiffening” mode of H-NS binding with transcriptional silencing (Walthers et al., 2011, Lim et al., 2012, Liu et al., 2010). However, others have correlated RNAP occupancy with H-NS binding at more than 65% of H-NS bound promoters (Oshima et al., 2006), suggesting a model in which RNAP may be trapped by H-NS (Singh & Grainger, 2013, Dame et al., 2002, Shin et al., 2005). In some instances, H-NS may direct RNAP to the −10 site via topological changes or by blocking access to surrounding AT-rich regions that resemble −10 hexamers (Singh & Grainger, 2013). These interactions alternatively implicate H-NS binding in a “bridging” mode (Kotlajich et al., 2015), which can most readily be observed at high magnesium concentrations (Liu et al., 2010). Many previous studies investigating RNAP-H-NS interactions have utilized high magnesium concentrations in vitro (Dame et al., 2002) and high osmolarity and 37°C in vivo (Oshima et al., 2006), conditions that fail to correlate with the physiological context in which the csg genes are expressed.

In this study, we used genetic and biochemical approaches to characterize the mechanism of H-NS-mediated silencing at the csgBA promoter and investigate the role of σ^S-associated RNAP in alleviating this silencing. Contrary to some previous studies, we found that counter-silencing at this promoter is σ^S-independent, as the formation of an H-NS filament occludes RNAP binding, regardless of the associated sigma-factor. The requirement for σ^S in curli expression is more likely to be at a level upstream of the csgBA operon, as previously suggested (Weber et al., 2006), and its functional role is unlikely to be as a counter-silencer. Rather, it is the curli transcriptional regulator CsgD that disrupts H-NS-mediated silencing through the formation of a sharp bend in the promoter DNA, altering the structure of the H-NS-DNA filament and allowing directed binding of RNAP. This finding highlights a novel counter-silencing protein, supports DNA bending as a fundamental mechanism of counter-silencing, and adds to a growing body of evidence suggesting that counter-silencing is a predominant mechanism of transcriptional regulation in bacteria.

Results

H-NS depletion restores csgB expression in rpoS and csgD mutant S. Typhimurium

The csgBA operon encoding curli fimbrial production is regulated by both H-NS and σ^S. To better understand the co-regulatory relationship between H-NS and σ^S, we performed an in vivo genetic analysis of the csgBA operon. Previous studies have found that S. Typhimurium hns mutants exhibit profound growth defects and rapidly acquire compensatory mutations in rpoS, resulting in reduced σ^S activity (Navarre et al., 2006, Visick & Clarke, 1997). To avoid adventitious compensatory mutations in rpoS, we used an IPTG-inducible antisense RNA construct to conditionally deplete hns mRNA in vivo (Nakashima & Tamura, 2009) and monitored the expression of rpoS-regulated genes to verify intact functionality of the protein (Figure S3). The paired-termini anti-sense hns (asHNS) transcript hybridizes to sequences flanking the ribosome-binding site and start codon of the hns promoter, occluding these regions and inhibiting translation (Figure 1), allowing inducible control of H-NS protein levels. Although σ^S protein levels also decrease during H-NS depletion, this is likely a result of H-NS acting indirectly as a positive regulator of rpoS, a phenomenon that has been observed previously (Silva et al., 2008, Wang et al., 2012). To confirm that asHNS was not also directly inhibiting rpoS translation, the upstream portion of the hns transcript recognized by asHNS was mutated to disrupt asRNA binding but maintain a functional hns coding sequence (Figure S2). H-NS and σ^S protein levels, as determined by immunoblot, were unchanged in strains encoding the mutant hns transcript, indicating that the decrease in σ^S levels observed in the wild-type strain is due to indirect regulation via H-NS and not due to a non-specific interaction between the rpoS transcript and asHNS.

Figure 1. Depletion of H-NS using an inducible asHNS construct.

H-NS was quantified from exponential and stationary phase cultures of SLN90 containing pHN1009 (asVector) and pHN1009asHNS, which encodes an asRNA complimentary to the region flanking the ribosome binding site of hns. asRNA expression was induced with IPTG for 4 hrs, cells were harvested, and H-NS levels visualized via immunoblot analysis using an anti-HA antibody. σ^S is absent in log phase but increases in abundance during stationary phase. GroEL was detected as a loading control.

The csgB transcript levels were then measured in strains containing either the asHNS construct or an asRNA vector control to examine the relationship between H-NS, σ^S, and CsgD. During exponential growth, csgB expression is low and unaffected by the mutation of either rpoS or csgD (Figure 2a), neither of which are normally expressed under these conditions (Figure 1) (Ogasawara et al., 2010). However, in stationary phase, both rpoS and csgD are up-regulated (Ogasawara et al., 2010), resulting in over a 350-fold increase in csgB expression relative to exponentially growing cells. Both σ^S and CsgD appear to be required for this growth phase-dependent expression, as mutation of either rpoS or csgD resulted in decreased csgB transcript levels similar to those observed in exponential phase (Figure 2b). However, the depletion of H-NS with asHNS resulted in a 4- to 6-fold increase of csgB transcript levels during exponential phase and a 10- to 21-fold increase during stationary phase in rpoS and csgD mutant strains, respectively, suggesting that these proteins oppose H-NS-mediated regulation of csgB. In wild-type cells during stationary phase, depletion of H-NS had no effect because both σ^S and CsgD were present and silencing was disrupted. Although a moderate decrease in csgB expression was observed under these conditions, this is likely due to decreased concentrations of σ^S resulting from H-NS depletion (Figure 1). The regulatory phenotypes of the hns, rpoS, and csgD mutant strains were confirmed by quantifying the expression of genes regulated by either H-NS alone (pagC) or σ^S alone (katE) by quantitative RT-PCR (Figure S3).

Figure 2. CsgD and σ^S up-regulate csgB in stationary phase.

csgB transcript levels were quantified by qRT-PCR during A) exponential or B) stationary phase in wild-type, rpoS, csgD, and rpoS csgD S. Typhimurium strains containing pHN1009 or pHN1009asHNS, with IPTG added to deplete H-NS. Data are presented as the mean +/− SD of three replicates normalized to gyrB; * indicates p<0.05, ** indicates p<0.01, *** indicates p<0.001, and **** indicates p<0.0001 by ratio-paired t-test.

The comparable levels of csgB expression in rpoS and csgD mutants and the restoration of csgB transcription to similar levels by asHNS suggests epistatic regulation of csgB by σ^S and CsgD. Although σ^S regulates csgD (Figure 3a)(Ogasawara et al., 2010), it is unknown whether σ^S-mediated regulation of csgB occurs directly at the csgB promoter or indirectly via csgD. To experimentally uncouple csgD and rpoS, and to determine the regulatory role of CsgD independent of rpoS, the plasmid pCsgD, expressing csgD from a constitutive promoter, was transformed into wild-type and rpoS mutant strains. Wild-type cultures containing pCsgD exhibited a 30-fold increase in csgB transcription during stationary phase (Figure 3b) as a result of increased csgD expression from the multicopy plasmid (Figure 3a)(Berggren et al., 1995). In an rpoS mutant, csgB transcription decreased but was restored to wild-type levels in strains containing pCsgD, suggesting that σ^S is required for CsgD synthesis but not directly required for csgB transcription. H-NS depletion in an rpoS mutant did not result in any further increase in csgB expression in strains containing pCsgD (Figure 3b), suggesting that CsgD is able to completely alleviate H-NS-mediated silencing.

Figure 3. Constitutive csgD expression in trans stimulates σ^S-independent csgB transcription.

A) csgD transcript levels were quantified using qRT-PCR in wild-type and rpoS S. Typhimurium containing pCsgD, which contains csgD downstream of a constitutive promoter. B) csgB transcript levels were quantified in response to csgD constitutive expression in strains containing both pCsgD and pHN1009asHNS. Data are presented as the mean +/− SD of three replicates normalized to gyrB expression; ** indicates p<0.01, and *** indicates p<0.001 by ratio-paired t-test.

H-NS depletion restores RDAR phenotypes in an rpoS mutant

Expression of curli, along with cellulose and other polysaccharides, results in the formation of RDAR (red, dry, and rough) colonies in many Salmonella strains under conditions of low temperature (30°C) and osmolarity, as part of a multicellular stress response (Römling, 2005). The RDAR phenotype is dependent on the expression of a functional csg region. Wild-type cultures are typically unable to form RDAR colonies at 37°C, but cultures expressing the asHNS construct at 37°C exhibit a partial RDAR phenotype (Figure 4a), suggesting that the temperature-sensitivity is a product of H-NS-mediated regulation. Although the colony structure of the asHNS strain is distinct, this may reflect a decrease in overall fitness due to the prolonged absence of H-NS, altering multicellular behavior, rather than a specific function of H-NS (Navarre et al., 2006). rpoS mutant cells are unable to form RDAR colonies (Figure 4a). However, depletion of H-NS resulted in a rough phenotype in an rpoS mutant after three days at 30°C and a partial restoration of the RDAR phenotype at 37°C.

Figure 4. Depletion of H-NS restores biofilm formation in σ^S-deficient S. Typhimurium.

A) Wild-type and rpoS strains of S. Typhimurium were spotted on Congo Red plates and incubated at 30 or 37°C to determine their RDAR phenotypes. B) rpoS strains containing pCsgD exhibited enhanced RDAR morphology in an rpoS mutant when compared to an rpoS strain containing the control vector.

Cultures expressing csgD constitutively from pCsgD were not temperature-sensitive, forming RDAR colonies at both 30°C and 37°C, as previously described (Kader et al., 2006), and pCsgD partially rescued the RDAR phenotype in rpoS strains, even in the presence of H-NS (Figure 4b), suggesting that CsgD is a critical determinant of RDAR colony formation. However, none of the strains tested were capable of fully complementing an rpoS mutation and forming wild-type RDAR colonies, indicating that additional σ^S-dependent factors contribute to RDAR morphology.

CsgD alone can counter-silence csgB in vitro

To understand the direct roles of σ^S, H-NS, and CsgD in the regulation of csgB, the csgB transcriptional circuit was reconstituted in vitro from individual purified components on large supercoiled targets, which are likely to be more physiologically and structurally indicative of csgB regulation in vivo than the small linear targets typically used in such studies. In vitro transcription assays with both Eσ⁷⁰ and Eσ^S indicated that the csgB promoter does not exhibit any basal σ-factor selectivity, as both holoenzymes transcribed csgB at similar levels (Figure 5b), an observation that has been reported for other σ^S-dependent promoters (Olvera et al., 2009, Šilar et al., 2016). To determine whether either sigma factor was able to overcome H-NS-mediated silencing at the csgB promoter, IVT assays were performed with either Eσ⁷⁰ or Eσ^S on csgB template incubated with increasing concentrations of H-NS. The two σ-factors exhibited nearly identical susceptibility to H-NS silencing at this promoter (Figure 5a), suggesting that σ^S does not function as a counter-silencer at csgB. However, when purified CsgD was incubated with the H-NS silenced template, transcription increased 4-fold with Eσ^S and 8-fold with Eσ⁷⁰ (Figure 5b), suggesting that counter-silencing at csgB is dependent on CsgD rather than σ^S.

Figure 5. CsgD counter-silences the csgB promoter in vitro.

The csgB transcriptional circuit was reconstituted in vitro using purified, supercoiled DNA template and purified protein components. A) Purified supercoiled plasmid containing a 3kb-region of DNA surrounding the csgB promoter was incubated in the presence of increasing concentrations of H-NS and either Eσ^S or Eσ⁷⁰ to assay transcriptional output. Transcript levels are normalized to the RNAP-only control. B) Either Eσ^S or Eσ⁷⁰ can initiate transcription of the csgB promoter in vitro. CsgD-mediated counter-silencing was assayed by incubating the csgB template in the presence of 50 nM CsgD before adding 130 nM H-NS. Transcript levels are expressed as relative copy number. Data are presented as the mean +/− SD of three replicates; * indicates p<0.05 by ratio-paired t-test.

Biochemical analysis of CsgD-mediated counter-silencing

To understand the structure of the nucleoprotein complexes involved in silencing and counter-silencing of csgB, Differential DNA Footprint Analysis (DDFA), a highly sensitive quantitative method for analyzing DNA-protein interactions and DNA structure in footprinting experiments (Will et al., 2014), was performed. This approach utilizes large targets and signal normalization to facilitate the analysis of DNA interactions with proteins such as H-NS that bind DNA in a relatively non-specific manner, which is not possible with conventional footprinting methods. Supercoiled csgB target was incubated with H-NS, as in the IVT studies described above, and the resulting complexes were analysed by DNase I footprinting. Analysis of the anti-sense strand revealed that although H-NS binds extensively along the csgB promoter region, two regions of concentrated binding are evident. The first is located upstream of the promoter, ranging from position −93 to −76 relative to the transcriptional start site (TSS), while a second region is detectable near the RNAP binding site from −37 to −5 (Figure 6a, Figure S4b). These locations may represent sites of H-NS nucleation, and indeed, exhibit AT-rich sequences with some similarity to a previously defined high-affinity binding site (Bouffartigues et al., 2007). The absence of hypersensitive sites on either strand (Figure 6a, Figure S6a), indicated by large upward peaks on DDFA plots and representative of DNA bends or distortions, suggests that H-NS binds the csgB promoter in a “stiffening” mode to form a nucleoprotein filament, corroborating evidence that H-NS forms filaments at low magnesium concentrations (Lim et al., 2012).

Figure 6. DNaseI Differential DNA Footprint Analysis (DDFA) of the csgB promoter region.

In vitro DNase I footprinting of the anti-sense strand at the csgB promoter was performed with 130 nM H-NS and 50 nM CsgD, as indicated. Peaks are regions of hypersensitivity, which are suggestive of distorted or bent DNA, whereas valleys indicate sites of protection. Base positions are indicated relative to the TSS. A) H-NS, B) CsgD, or E) CsgD and H-NS were added to the csgB promoter and the fluorescent peak height following DNase I cleavage determined relative to the protein-free control in relative fluorescent units (RFU). The differences compared to the H-NS control (C) and the CsgD control (D) were also analyzed. Data are presented as the mean +/− SD of three replicates.

Observed protected sites at −20 to −25 and −86 on the anti-sense strand (Figure 6b) and at −40 and −60 on the sense strand (Figure S6b) suggest the locations of the CsgD footprint, which align more closely to previous findings in E. coli (Ogasawara et al., 2011) than what has been predicted in Salmonella (Zakikhany et al., 2010). Binding closer to the promoter region introduced a hypersensitive site at position −33 (Figure 6b, Figure S6c). Hypersensitive sites observed in DNase I footprinting are suggestive of flexible DNA regions subject to looping or bending (Suck & Oefner, 1986; Bouffartigues et al., 2007; Walthers et al., 2011; Baraquet et al., 2012; Bhat et al., 2014). The induction of a bend was confirmed by atomic force microscopy (AFM) using linear DNA fragments containing the csgB promoter (Figure S5a). The presence of CsgD significantly decreases the mean distance between fragment ends in comparison to DNA alone (Figure S5c), which is indicative of DNA bending (Wiggins et al., 2006). This is significant, as recent studies of SlyA-PhoP-mediated counter-silencing suggested that DNA-bending disrupts the H-NS-DNA filament to alleviate silencing (Will et al., 2014).

CsgD and H-NS are able to concomitantly bind the csgB promoter, which can be discerned from the fact that the combined footprint of an H-NS + CsgD complex is distinct from that of each individual protein (Figure 6). Binding of CsgD to H-NS-bound DNA produces the same hypersensitive site as is observed with CsgD alone (Figure 6e). CsgD binding between −20 and −25, and near −86 (Figure 6b), overlaps both potential H-NS nucleation sites, suggesting that CsgD may disrupt filament formation through direct competition for binding at H-NS nucleation sites. Although H-NS binding was also perturbed at other positions, such as −75 and −7, these positions exhibited significant variability, suggesting that the H-NS-CsgD-DNA complexes at these locations are unstable, perhaps as a result of a disrupted nucleation site. As we have observed at other promoters (Will et al., 2014), most H-NS was not displaced despite profound changes in the architecture of the promoter introduced by CsgD, particularly upstream of the transcriptional start site at −5 and adjacent to the RNA polymerase binding site, at −25 (Figure 6e), indicating that the structure of the H-NS-DNA complex, rather than simple binding, is a critical determinant of H-NS activity. This is not entirely unexpected, as other studies have recognized that H-NS may remain bound to DNA during counter-silencing conditions, and the force imparted by RNAP elongation is sufficient to overcome bridging interactions (Dame et al., 2006, Kotlajich et al., 2015, Chandraprakash & Seshasayee, 2014).

Differential DNA Footprinting Analysis of the sense strand demonstrated CsgD binding at previously described recognition motifs near −60 (Figures S6b and S7c), again in a pattern more similar to that observed in E. coli (Ogasawara et al., 2011) than to what has been predicted in Salmonella (Zakikhany et al., 2010). H-NS is able to protect this strand at the −10 and −35 regions (Figures S6a and S7d) but is displaced in the presence of CsgD, as reflected by the positive change at these sites when CsgD and H-NS are both present as compared to H-NS alone (Figure S6c). However, the most striking distortion still appears to take place on the anti-sense strand, as noted above.

To investigate the mechanism of counter-silencing, KMnO₄ footprinting was performed, which identifies sites of open-complex formation through reaction with single-stranded DNA exposed by RNAP binding (Sasse-Dwight & Gralla, 1989). H-NS silencing of the csgB promoter via an occlusion mechanism would be expected to result in the disappearance of an open complex upon H-NS addition. However, silencing by a trapping mechanism would be expected to still show an open complex following the addition of H-NS, and silencing would result from abortive transcription (Kotlajich et al., 2015). The addition of RNAP holoenzyme alone to the csgB promoter produced a stable open complex (Figure 7a, Figure S8b), indicating that this promoter is transcriptionally competent. However, in the absense of other proteins, the open complex formed approximately 50 bp upstream of the usual transcription start site (Arnqvist et al., 1994). This open complex was abrogated by the addition of H-NS (Figure 7b, Figure S8e), suggesting that H-NS silences the csgB promoter by occlusion rather than by bridging and trapping (Dame et al., 2002). DNase I footprinting showed that RNAP is unable to bind in the presence of H-NS, particularly in the −20 region (Figure S9b), eliminating the possibility that H-NS is able to bind but unable to form an open complex. The addition of CsgD restored the ability of RNAP to form an open complex in the presence of H-NS (Figure 7b, Figure S8f) at a location corresponding to the usual transcription start site. RNAP also forms an open complex at this location with CsgD in the absence of H-NS (Figure 7a, Figure S8d), suggesting that CsgD recruits RNAP to this site. These observations support the hypothesis that CsgD alters the H-NS nucleoprotein complex to allow RNAP binding and generate a transcriptionally competent promoter.

Figure 7. KMnO₄ Differential DNA Footprint Analysis (DDFA) at the csgB promoter region.

KMnO₄ footprinting reactions, which detect regions of single-stranded DNA generated during open-complex formation, were performed on the csgB promoter region with RNAP, CsgD, and H-NS, as indicated. Peaks, in relative fluorescent units (RFUs), indicate regions of single stranded DNA present in the experimental samples that are not present in control (no RNAP) reactions. Base positions are indicated relative to the TSS. a) KMnO₄ footprinting analysis indicates that the location of RNAP open complex formation on the naked DNA promoter is dependent on the presence or absence of CsgD. RNAP alone is capable of forming a weak open complex at an upstream site, but the addition of CsgD directs RNAP to the previously identified transcriptional start site. b) RNAP is unable to form an open complex in the presence of H-NS, but the addition of CsgD restores open complex formation at the csgB promoter. Data are presented as the mean +/− SD of three replicates.

Discussion

Previous studies have shown that regulation of the E. coli csgBA curli fimbrial operon involves the interplay of several transcription factors, including H-NS, CsgD, and σ^S, with σ^S required for expression in vivo (Olsén et al., 1993). In this study, we demonstrate that CsgD acts as a counter-silencer of H-NS in S. Typhimurium, remodeling the H-NS-DNA complex at the csgB promoter to restore RNAP binding, whether associated with σ^S or σ⁷⁰, and that the specific requirement for σ^S is indirect and upstream of csgD expression. In addition, as is typical for a classical activator, CsgD directs RNAP binding to facilitate open complex formation at the csgBA promoter. These findings provide further evidence that counter-silencing is a major mechanism of transcriptional regulation in bacteria.

Given our initial in vitro findings that σ^S is unable to counter H-NS silencing at csgB directly, we investigated whether σ^S might promote counter-silencing in concert with additional regulatory proteins. As a precedent, the regulatory proteins SlyA and PhoP act in combination to relieve H-NS-mediated repression of the Salmonella pagC promoter (Will et al., 2014). SpvR (Robbe-Saule et al., 1997), cAMP-CRP (Marschall et al., 1998) and YncC (Beraud et al., 2010) are also believed to be required for σ^S-mediated counter-silencing at their respective promoters, either by initiating sigma-specific recruitment or by altering the structure of the nucleoprotein complex at the promoter. At csgBA, the master curli regulator CsgD appeared to be a suitable candidate, as it directly binds the promoter region and up-regulates transcription.

In agreement with previous studies (Arnqvist et al., 1994), we found that csgBA expression is dependent on σ^S in vivo, but becomes σ^S-independent following depletion of H-NS by asHNS (Figure 2). Similarly, csgBA transcription is reduced in a csgD mutant but is restored by depletion of H-NS. Furthermore, transcription from the csgBA promoter becomes σ^S-independent in the absence of H-NS, as well as when csgD is constitutively expressed (Figure 3). Although previous studies suggested that H-NS is unable to silence transcription in the presence of σ^S (Shin et al., 2005), our data indicate that in the case of the csgB promoter, H-NS is able to repress σ^S-coupled RNAP both in vivo and in vitro. This suggests a direct requirement for CsgD, rather than σ^S, to mediate counter-silencing and drive transcription of the csgBA promoter. σ^S activates a feed-forward regulatory loop at the csg operon, in which σ^S either directly or indirectly stimulates expression of csgD at the onset of stationary phase (Ogasawara et al., 2010), and in turn CsgD enhances expression of σ^S indirectly through transcriptional activation of iraP, which stabilizes σ^S (Gualdi et al. 2007). The apparent σ^S-dependence of csgB may be attributable to the σ^S-CsgD feed-forward loop. Correspondingly, the lack of σ^S is analogous to the absence of CsgD, abolishing transcription of csgBA. It is possible that similar mechanisms occur at other σ^S/H-NS co-regulated promoters, although further study will be necessary to confirm this hypothesis.

Our DNase I footprinting DDFA studies (Figure 6) are consistent with H-NS nucleation at high affinity “TpA” steps and oligomerization along lower affinity regions flanking the csgBA promoter (Lim et al., 2012, Winardhi et al., 2012), beginning upstream of the RNAP binding site and extending into the coding region, likely resulting in the occlusion of RNAP binding (Becker & Hengge-Aronis, 2001). DDFA plots of CsgD binding at csgBA revealed that CsgD binds immediately upstream of the TSS, competing with H-NS nucleation sites and inducing a sharp bend in the H-NS-bound DNA at the promoter, a structural change that is likely to disrupt H-NS-DNA filament formation (Will et al., 2014; Will et al., 2015). Previous studies have suggested that CsgD functions as a classical activator due to its binding site overlapping the −35 box of the RNAP binding site (Brombacher et al., 2003), and our observation of directed open complex formation in the presence of CsgD supports that hypothesis. However our observations also suggest that CsgD-mediated counter-silencing is essential for csgB expression

There appear to be additional uncharacterized requirements for CsgD-mediated regulation at some promoters, as CsgD is reported to regulate other genes, including its own promoter (Ogasawara et al., 2010), but is unable to counter-silence or activate its own promoter in vitro (Figure S10). A likely explanation is that an additional factor acts as a co-counter-silencer with CsgD. However, the identity of potential co-counter-silencers at csgD is presently unknown.

CsgD is a member of the LuxR-family of transcriptional regulators, which have been shown to be involved in biological processes including quorum sensing, virulence, motility, plasmid transfer, and biofilm formation in a variety of bacterial species (Chen & Xie, 2011). Previous studies have suggested that LuxR-family members can serve as either direct activators or repressors of transcription (Miyamoto et al., 1994, van Kessel et al., 2013), but there is also a precedent for LuxR and its homologs to serve as counter-silencers. In Vibrio species, LuxR acts in conjunction with σ^S at the lux bioluminescence genes to relieve H-NS silencing, as neither LuxR nor σ^S are necessary for lux expression in an hns mutant (Ulitzur et al., 1997, Ulitzur, 1998). In Pseudomonas aeruginosa, the LuxR regulators RhlR and LasR promote stationary-phase expression of lecA, which is silenced by the H-NS functional homologue, MvaT (Diggle et al., 2002), and in other enteric bacteria, LuxR regulators such as BglJ (found in Salmonella, E. coli, and Shigella) (Giel et al., 1996) and EcpR (found in both enteropathogenic and enterohemorrhagic Escherichia coli) (Martínez-Santos et al., 2012) have also been found to disrupt H-NS silencing. The recognition sequences of LuxR-family proteins tend to be degenerate in nature, allowing them to recognize a variety of foreign promoters (Tsou et al., 2009, Lee et al., 2008). Additional studies have suggested that, like CsgD, LuxR bends DNA upon binding (Stevens, Dolan et al., 1994), which might allow it to disrupt the H-NS-DNA filament and function as a counter-silencer. In view of these observations, the LuxR-family transcriptional regulators may function primarily as counter-silencers rather than classical activators. As this family of transcriptional regulators can be found in many bacterial species, counter-silencing by these proteins may be a conserved mechanism to incorporate horizontally acquired genes into host genomes.

In summary, we have demonstrated for the first time that CsgD induces distinct structural changes at the csgBA promoter and promotes transcription by disrupting H-NS silencing and directing RNAP binding and open complex formation. As σ^S does not act as a counter-silencer of csgBA expression, the basis for the σ^S-dependence of the curli operon is presently unclear. A role of σ^S in csgD regulation has not been excluded. As H-NS negatively regulates csgD expression, it remains possible that σ^S counters H-NS silencing of this gene. However, given the number of additional regulators at the csgD promoter (Ogasawara et al., 2010), the explanation is likely to be more complex and awaits further investigation.

Experimental Procedures

Bacterial strains and plasmids

All Salmonella enterica serovar Typhimurium strains were constructed in a 14028s genetic background and grown in 50 mM NaCl Luria Bertani (LB) medium, which corresponds to 100 mOsm, unless otherwise noted. This medium has low osmolarity compared to typical LB (396 mOsm). All bacterial strains, plasmids, and primers are described in Table S1. Ampicillin and kanamycin were added to media as required at concentrations of 100 μg ml⁻¹ and 50 μg ml⁻¹, respectively. Deletions of csgD and rpoS were generated using the λ-Red recombinase protocol (Datsenko & Wanner, 2000), and antibiotic resistance markers were swapped out using pCP20 (Cherepanov & Wackernagel, 1995). Phage P22 HT105/1 int-201 was used for transduction of csgD and rpoS mutations into the appropriate strains (Schmieger, 1972).

Plasmid pSLN15 was constructed by cloning a PCR fragment generated from the primers csgD-OE-F and csgD-OE-R into the BamHI and XhoI sites of pET28b (Qiagen). Plasmid pSLN16 was constructed by cloning a PCR fragment generated from the primers EF154 and EF155 with pKD4 as a template into pHR103 (Frawley et al., 2013) and inserted into the PvuI site. Plasmid pSLN17 was constructed by cloning the PCR product generated from the primers SP20 and SP21 into the XbaI and EcoRI sites of pSLN16. Plasmid pQE-80L-RpoS was constructed by cloning the PCR product generated from the primers RpoS-OE-F and RpoS-OE-R into the SacI and HindIII sites of pQE-80L (Qiagen). Plasmid pQE-80L-RpoD was constructed by cloning the PCR product generated from the primers RpoD-OE-F and RpoD-OE-R into the SacI and HindIII sites of pQE-80L (Qiagen). Plasmid pHN1009 asHNS was constructed by cloning a 128 bp fragment of the H-NS promoter region starting 36 bp upstream of the translational start site and spanning 92 bp into the gene. The fragment was amplified using primers asRNA HNS F and asRNA HNS R and inserted into the NcoI and XhoI sites of pHN1009 (Nakashima & Tamura, 2009). Plasmid pSLN7 was generated by cloning a 3.0 kb fragment of the csgB region, amplified using primers csgA IVT vector R and csgA IVT vector F, into the HinDIII and XbaI sites of pRW20 (Will et al., 2014). Plasmid pSLN19 was generated by cloning a 2.9 kb fragment of the csgD promoter region, amplified using primers csgD IVT vector F and csgD IVT vector R, and moved into the PstI and XbaI sites of pRW20. Plasmid DNA for biochemical assays was purified as previously described (Will et al., 2014).

C-terminal HA-tagged H-NS was created using the λ-Red recombinase protocol (Datsenko & Wanner, 2000) with insertion of the tetRA locus using a PCR product generated from HNS-tetRA F/R primers into strain SL4482. The tetRA element was subsequently replaced with a PCR fragment amplified by HA-HNS fill-in F/R primers, generating strain SLN66. Using SLN66 gDNA as a template, Gibson cloning fragment 1 amplified with primers “HNS Gibson Scramble Fragment 1 F/R,” Gibson cloning fragment 2 amplified with primers “HNS Gibson scramble Fragment 2 F/R,” and Gibson cloning fragment 3 amplified with primers “H-NS Gibson Scramble Fragment 3 F and R” were PCR-amplified, gel purified and cloned at equimolar ratios into the BamHI site of pRDH10 using Gibson Assembly Master Mix (NEB) and incubated for 90 min at 50°C to construct plasmid pSLN23. pSLN23 was introduced into the S. Typhimurium SLN196 recipient strain, containing pSW172, by conjugation using E. coli S17-1λpir (FLS306) as the donor strain. Strain selection was performed as previously published (Faber et al., 2016) to create SLN221 and pSW172 was cured by growth at 37°C.

RNA isolation and quantitative PCR analysis of in vivo gene expression

2 × 10⁷ CFUs of overnight cultures grown in LB broth at 37°C were subcultured into 50 mL 100 mOsm LB broth and grown at 30°C. For exponential phase experiments, cells were split into two flasks, and one was induced with 1mM isopropyl β-D-1-thiogalactoside (IPTG) for approximately four hrs before reaching OD₆₀₀~0.6. For stationary phase experiments, cells were split at OD₆₀₀ ~0.6 into two flasks, and one was induced with 1 mM IPTG. After induction, cells were grown for another four hrs at 30°C. Cells were then pelleted and re-suspended in TRIzol reagent (Life Technologies) and incubated for a minimum of 30 min. After two chloroform extractions, isolated RNA was suspended in isopropanol and stored at −20°C until further processing. RNA was precipitated and washed per manufacturer’s instructions. Five U DNase I (Thermo Fisher Scientific) were added to each sample and incubated for one hr at 37°C to remove residual DNA. RNA was extracted with acid-phenol:chloroform (1:1) and precipitated with 95% ethanol. The final RNA pellet was dissolved in RNase-free water. RNA was converted to cDNA using the QuantiTect RT-PCR kit (Qiagen) according to the manufacturer’s instructions. cDNA was diluted appropriately (1:20) to bring samples into linear range of detection and transcripts were quantified via qPCR in a CFX90 (BioRad) using SYBR green master mix (Aparicio et al., 2005). Oligonucleotide pairs are described in Table S1. The relative expression of genes was determined by normalizing expression levels to gyrB transcript.

Phenotypic Analyses

Bacterial strains were grown in 100 mOsm NaCl LB overnight at 37°C. The cells were diluted to OD₆₀₀ 1.0, and 5 μL were spot inoculated on LB plates without salt and supplemented with the appropriate antibiotic(s), Congo red (40 μg mL⁻¹), and Coomassie Blue (20 μg mL⁻¹) (Anwar et al., 2014). For asHNS knock-down experiments, plates were spread with 40 μL of 100 mM IPTG before spotting. Plates were incubated at 28°C or 37°C for three days without inversion.

Western Blotting

The asHNS construct and asVector control were transformed into an S. Typhimurium 14028s strain carrying an N-terminal HA-tagged H-NS chromosomal construct. Depletion experiments were conducted as above, but instead of harvesting cells for RNA collection, 0.25 OD₆₀₀ cells were collected at each time point by centrifugation. Pellets were boiled for 5 min at 95°C in 30μL Laemmli buffer, and total protein was separated using a 4–15% SDS gradient gel. Protein was transferred onto a PVDF membrane and, after blocking with PBS buffer containing 0.05% Tween-20 (PBS-T) and 5% Non-fat dry milk, incubated in blocking solution with 1:2000 anti-HA (Sigma), 1:1000 mouse IgG anti-sigmaS (Biolegend), or 1:80,000 anti-GroEL (Sigma). Membranes were washed in PBS-T and probed with either anti-rabbit IgG or anti-mouse IgG-HRP conjugate secondary antibody (Bio-Rad) diluted 1:10,000 in blocking solution, and visualized with Pierce ECL Western Blotting Substrate (Pierce).

Protein Purification

Sigma Factors

Expression vectors pQE-80L-RpoS and pQE-80L-RpoD were transformed into E. coli BL21 (DE3) cells for production and purification of recombinant sigma factors. Cells were grown at 30°C to OD₆₀₀ 0.5 in LB. Expression was induced with 1 mM IPTG for four hrs before cells were pelleted by centrifugation (7,000 × g, 10 min at 4°C) and stored at −80°C. Cells were re-suspended in 20 mL binding buffer (50 mM NaH₂PO₄, 300 mM NaCl pH 8.0) with 0.5 mg mL⁻¹ lysozyme and protease inhibitors (Sigma). The solution was incubated on ice for twenty min and lysed by sonication, and supernatant was clarified with three rounds of centrifugation (15,000 × g, 30 minutes each). Lysates were then applied to 5 mL HisTrap HP columns (GE Healthcare Life Sciences) equilibrated with binding buffer. Columns were washed first with 5 column volumes (CV) binding buffer followed by 4 CV binding buffer supplemented with 20 mM imidazole. Proteins were then eluted over a 20 CV gradient from 20 mM to 500 mM imidazole in binding buffer. Fractions containing sigma factors were determined by SDS-PAGE and pooled. Eluate was dialyzed overnight at 4°C at a 1:200 ratio to gel filtration buffer (50 mM Tris pH 7.9, 500 mM NaCl, 0.1 mM EDTA, 10 mM β-mercaptoethanol, 5% glycerol) and concentrated to 2.5 mL. The sample was injected into a 130 mL Superdex 75 gel filtration column (GE Healthcare Life Sciences) equilibriated with gel filtration buffer. Elution occurred over 180 mL. Pure fractions were identified by SDS-PAGE, dialyzed in storage buffer (50 mM Tris pH 7.5, 0.5 M NaCl, 1 mM EDTA, 0.5 mM DTT, 50% glycerol), and stored at −80°C.

H-NS

H-NS purification was performed as previously using pRW13 (Will et al., 2014) except that elution from the His-Trap HP column was performed as follows: 10 CV of linear gradient of buffer containing 20 mM to 250 mM imidazole followed by a 5 CV wash with buffer containing 250 mM imidazole followed by a 5 CV linear gradient of buffer containing 250 to 500 mM imidazole. Purified protein was stored in aliquots at −80°C.

CsgD

pSLN15 was transformed into E. coli BL21 (DE3) for production and purification of CsgD. Cells were grown at 37°C to OD₆₀₀~0.6 in 100 mOsm LB and then induced with 1 mM IPTG for four hrs at 30°C and harvested and stored as above. Cells were re-suspended in 15 mL binding buffer supplemented with 20 mM imidazole, 1 mM PMSF, and protease inhibitors (Sigma) and lysed by sonication. The lysate was clarified by centrifugation (three times at 20,000 × g, 30 minutes each). Lysates were then loaded onto an equilibrated 5 mL His-Trap HP column (GE Healthcare Life Sciences) and washed with 50 mL binding buffer containing 20 mM imidazole. Protein was eluted over a 120 mL linear gradient of imidazole, from 20 to 250 mM, and CsgD-containing fractions were identified through SDS-PAGE. CsgD was concentrated using an Amicon Ultra centrifugal concentrator (Millipore), and the 6xHis-tag was removed using a Thrombin Cleavage Capture Kit (Novagen) with overnight incubation at 4°C, combining 1 mg protein with 1 U Thrombin. Any remaining precipitate was isolated by centrifugation. Residual thrombin was removed with streptavidin agarose per the manufacturer’s protocol. Sample was dialyzed against binding buffer overnight at 4°C and then loaded onto 0.5 mL equilibrated Ni-NTA resin (Qiagen). Purified CsgD was collected from the flow-through and verified by SDS-PAGE. Pure cleaved CsgD was dialyzed into CsgD storage buffer (10 mM Tris pH 8.0, 10 mM MgCl₂, 100 mM KCl, 1 mM DTT, 0.1 mM EDTA, 50% glycerol) and stored at −80°C. The His×6 tag was not removed for AFM studies (Figure S5). Instead, the cell lysate was prepared and applied to a HisTrap HP column (GE Healthcare Life Sciences) as described above. The protein was then purified using a linear gradient of imidazole from 20 mM to 500 mM. CsgD-containing fractions were identified via SDS-PAGE, pooled, and dialyzed overnight into CsgD storage buffer before storing at −20°C.

In vitro transcription

All steps of IVT reactions were carried out at 30°C unless otherwise specified. Reactions were assembled in IVT buffer containing 40 mM HEPES, pH 7.3, 1 mM MgCl₂, 120 mM potassium glutamate, 0.5 mM CaCl₂, 1 mM DTT, 0.05% NP-40, 0.1 mg mL⁻¹ BSA and 10% glycerol to a final volume of 20 μL. This concentration of potassium glutamate was used to promote potential sigma-factor specificity (Rajkumari et al., 1996). To confirm that the elevated potassium glutamate concentration did not impact counter-silencing, counter-silencing IVT reactions were also performed in IVT buffer containing 50 mM potassium glutamate (Figure S11a). Purified σ^S was incubated with RNAP core (New England Biolabs) at a 10:1 ratio, whereas purified σ⁷⁰ was incubated with RNAP core at a 2:1 ratio. RNA polymerase holoenzyme was reconstituted in IVT buffer for 30 min at 30°C immediately before use in IVT reactions. Template at a final concentration of 1 nM was combined with 20 U of RiboLock RNase Inhibitor (Thermo Fisher Scientific) and incubated for 10 min in order to equilibrate. For counter-silencing reactions, CsgD was added at 50 nM for 10 min before H-NS was added to a final concentration of 130 nM and incubated for 10 min, although the order of addition was not found to influence counter-silencing (compare Figure 5b, Figure S11b). Following the addition of regulatory proteins, reconstituted RNA polymerase holoenyzme was incubated with template for 10 min at a final concentration of 10 nM. rNTPs were added to a final concentration of 1 nM and the reaction incubated for an additional 30 min at 30°C. The reaction was stopped by adding 20 μL DNase I buffer containing 10 mM Tris-HCl (pH 7.5), 2.5 mM MgCl₂, 0.1 mM CaCl₂ and 4 U of DNase I (Thermo Fisher Scientific), and incubated for 30 min at 37°C. EDTA was added to final concentration of 5 mM and the reactions incubated for 10 min at 65°C to inactivate DNase I. cDNA synthesis was performed as previously (Will et al., 2014) using the sequence-specific reverse oligonucleotide probes as indicated in Table S1.

DNase I Footprinting and Fluorescent Primer Extension

Reactions, fluorescent primer extension, and Differential DNA Footprint Analysis (DDFA) were performed as in Will et al. (Will et al., 2014) using csgB FAM sense or csgB FAM anti-sense probes as indicated in Table S1.

KMnO₄ Footprinting

Reactions were assembled as described for IVT reactions, except that DTT was omitted from the IVT buffer. Proteins were sequentially added for 10 min each at 30°C as for IVT reactions, except that RNAP was allowed to incubate for 15 min, at which point KMnO₄ was added to a final concentration of 10 mM. The reactions were allowed to proceed for 2 min and stopped by the addition of 2 μL of 14 M β-mercaptoethanol, followed by 78 μL of quenching buffer containing 1 M β-mercaptoethanol, 20 mM EDTA, and 385 mM sodium acetate (pH 7.0). Reactions were extracted with phenol:CHCl3:IAA (25:24:1) followed by chloroform and precipitated with 2 μL glycogen and ethanol. Reactions were washed 3 times with 70% ethanol, dried, and re-dissolved in 10 μL H₂0. Reactions were detected via fluorescent primer extension as above using the csgB FAM sense probe as the primer.