Regulation of toxin production in the pathogenic clostridia

Authors


Summary

The genus Clostridium comprises a large, heterogeneous group of obligate anaerobic, Gram-positive spore forming bacilli. Members of this genus are ubiquitous in the environment and although most species are considered saprophytic, several are pathogenic to both humans and animals. These bacteria cause a variety of diseases including neuroparalysis, gas gangrene, necrotic enteritis, food poisoning, toxic shock syndrome and pseudomembraneous colitis, which in most cases arise as a consequence of the production of potent exotoxins. Treatment options are often limited, underscoring the need for new treatment strategies and novel therapeutics. Understanding the fundamental mechanisms and signals that control toxin production in the pathogenic clostridia may lead to the identification of novel therapeutic targets that can be exploited in the development of new antimicrobial agents.

Introduction

The pathogenic clostridia are an important group of bacteria that can cause disease in humans and other animals. Arguably the most important clostridial species in human disease are Clostridium botulinum, Clostridium difficile, Clostridium perfringens, Clostridium sordellii and Clostridium tetani. C. botulinum and C. tetani are neurotoxigenic and result in flaccid and spastic paralysis, respectively (Montecucco and Schiavo, 1994). C. perfringens is the causative agent of a myriad of different diseases ranging from gas gangrene and necrotic enteritis to food poisoning (Rood, 1998; Uzal and McClane, 2011), depending on the particular toxins produced by the infecting strain. C. difficile is an increasingly important nosocomial pathogen, which is considered the leading cause of antibiotic-associated diarrhoea and pseudomembranous colitis in the developed world (Carter et al., 2011a; 2012) and C. sordellii is an emerging pathogen that causes a range of diseases including myonecrosis, sepsis and shock (Carter et al., 2011a). The clostridial species that are important human pathogens also cause numerous neurotoxic, histotoxic and enterotoxic animal diseases. In addition to these species, Clostridium chauvoei, Clostridium colinum, Clostridium novyi, Clostridium septicum and Clostridium spiroforme also cause severe disease in numerous domestic and livestock animals (Songer, 1996; 1998; 2010). Although these organisms are responsible for very different diseases, in most cases the symptoms of disease are associated with the production of potent exotoxins that result in damage to the host (Stevens et al., 2012). Collectively the pathogenic clostridia produce some of the most potent toxins known, including the C. botulinum neurotoxin (BoNT), which is the most potent toxin identified (Schechter and Arnon, 2000; Katona, 2012).

Many of the clostridial toxins have been purified and subjected to extensive structural and functional analysis. However, understanding the mechanisms and signals that control their production has been limited by the difficulties encountered in genetically manipulating the pathogenic clostridia. Indeed, the construction of genetically defined mutants in many clostridial species remains a challenge and methods have yet to be developed for successful genetic manipulation of these bacteria. However, the recent development of techniques that have facilitated the genetic manipulation of some of these species (Lyristis et al., 1994; Chen et al., 2005; O'Connor et al., 2006; Heap et al., 2007; Bradshaw et al., 2010; Carter et al., 2011a) has resulted in a deeper understanding of the genes and regulatory pathways that control toxin production at the molecular level. One important outcome of this recent work may be that novel targets will be identified and targeted in the development of therapeutics for the treatment of these devastating clostridial infections, for which few effective remedies are currently available. Here, we summarize and discuss the current understanding of the regulatory mechanisms and signals used to control toxin production in the pathogenic clostridia, with a particular emphasis on C. botulinum, C. difficile, C. perfringens, C. sordellii and C. tetani.

Nutritional and environmental stimuli

Environmental stimuli play an important role in controlling toxin production in the pathogenic clostridia. For example, the production of α-toxin, β-toxin, β2-toxin and perfringolysin O (PFO) is activated following direct cell–cell contact with eukaryotic host cells such as Caco-2 cells (Vidal et al., 2009b). Temperature is also able to modulate toxin production in the clostridia, with the production of α-toxin in C. perfringens being optimal at 25°C (Katayama et al., 1999), whereas in C. difficile maximal toxin production occurs at 37°C, with reduced yields at 22°C and 42°C (Karlsson et al., 2003). A number of other environmental stimuli also have an impact on toxin production in C. difficile, including bicarbonate concentration (Karlsson et al., 1999), sub-inhibitory concentrations of certain antibiotics (Chilton et al., 2012; Aldape et al., 2013) and the presence of short-chain fatty acids such as butyric acid (Karlsson et al., 2000), although the molecular mechanisms controlling these responses are not well understood. One of the most important groups of environmental cues is likely to be nutritional signals. In C. botulinum, the presence of arginine and derivatives such as proline, glutamate and ammonia repress neurotoxin production (Johnson and Bradshaw, 2001), as does the aromatic amino acid tryptophan (Johnson and Bradshaw, 2001). Similarly, amino acids such as proline, cysteine, and certain branched chain amino acids (BCAAs), repress toxin production in C. difficile (Karlsson et al., 1999). In this bacterium, the global transcriptional regulator CodY appears to be responsible for mediating the observed amino acid-based repression, with increased amounts of toxin produced by a C. difficile codY mutant in the presence of BCAAs compared to the wild-type strain (Dineen et al., 2007). CodY represses toxin production predominately by repressing the expression of tcdR, a critical regulator of toxin production, which is discussed in more detail later in this review. In this instance, the presence of BCAAs results in increased CodY binding to the tcdR promoter and the subsequent repression of tcdR expression (Dineen et al., 2010), leading to reduced toxin production.

Catabolite repression

The presence of glucose, or other rapidly metabolizable carbon sources, is also an important nutritional signal that inhibits the production of several toxins in the pathogenic clostridia, including toxin A (TcdA) and toxin B (TcdB) in C. difficile (Dupuy and Sonenshein, 1998), and α-toxin and PFO in C. perfringens (Mendez et al., 2012). In both organisms, carbon catabolite repression is thought to be responsible since glucose inhibition of toxin production is not observed in ccpA mutants that no longer produce the carbon catabolite repressor protein CcpA (Antunes et al., 2011; Mendez et al., 2012). Further support for the role of CcpA in mediating glucose-dependent inhibition of toxin production in C. difficile comes from the finding that the tcdA, tcdB and toxin regulator genes tcdR and tcdC are direct targets of the CcpA protein in vitro (Antunes et al., 2012). The observation that C. difficile ccpA mutants produce a reduced amount of toxin in comparison to the wild-type strain (Antunes et al., 2011) suggests that toxin regulation is complex and that unidentified regulators must also be involved in the CcpA-regulation cascade. In C. perfringens, CcpA also regulates C. perfringens enterotoxin (CPE) production in a growth phase-dependent manner that is not linked to the presence of glucose (Varga et al., 2004), suggesting that unknown factors must play a role in CcpA-mediated regulation of CPE.

Growth phase and quorum sensing

Growth phase signals also play an important role in the regulation of toxins by the pathogenic clostridia. The production of α-toxin (Vidal et al., 2009b), PFO (Vidal et al., 2009b), NetB (Cheung et al., 2010), β-toxin (Fernandez-Miyakawa et al., 2007) and ε-toxin (Fernandez-Miyakawa et al., 2007) by C. perfringens occurs during or after late exponential growth, while CPE is produced exclusively during sporulation (Duncan, 1973). Unlike most clostridial toxins, CPE is not actively secreted; instead, it accumulates in the mother cell during sporulation before being released along with the spore when the mother cell lyses (Duncan, 1973). In C. botulinum and C. tetani, the production of BoNT and TeNT, respectively, is associated with the transition to stationary-phase growth (Mellanby and Green, 1981; Couesnon et al., 2006), as is the production of TcdA and TcdB in C. difficile (Hundsberger et al., 1997). The precise growth phase signals involved in toxin production remain mostly unknown; however, the role of quorum sensing has become increasingly clear in recent years.

Accessory gene regulator (agr)-based quorum sensing

The accessory gene regulator (agr) system of S. aureus, the most extensively characterized quorum sensing system of any Gram-positive pathogen, regulates the expression of many virulence factors in this bacterium, including several toxins (Novick, 2003). This system comprises the agrAC genes, encoding a two-component signalling system, and the agrBD genes, which encode an autoinducing peptide (AIP) processing and export protein and the AIP pre-peptide, respectively (Novick, 2003). Homologues of the S. aureus agrBD genes have been identified in a number of the toxigenic clostridia, including C. difficile (Sebaihia et al., 2006), C. botulinum (Cooksley et al., 2010) and C. perfringens (Ohtani et al., 2009, Vidal et al., 2009a) (Fig. 1). It is evident that agr-based quorum sensing plays an important role in controlling toxin production in both C. botulinum and C. perfringens. C. botulinum contains two homologues of the agrBD genes, named agrBD1 and agrBD2 (Cooksley et al., 2010). Inactivation of either agrD1 or agrD2 resulted in lower BoNT production compared to the wild-type strain, with the latter mutant displaying the greatest reduction (Cooksley et al., 2010). Similarly, inactivation of the agrBD genes has a profound effect on toxin production in C. perfringens. Mutants in agrB in a number of different strain backgrounds display a complete loss of PFO production (Ohtani et al., 2009; Vidal et al., 2009a), and reduced levels of α-toxin (Ohtani et al., 2009; Vidal et al., 2009a), κ-toxin (Ohtani et al., 2009), CPE (Li et al., 2011), β-toxin (Vidal et al., 2012), β2-toxin (Li et al., 2011) and ε-toxin (Chen et al., 2011a) production compared to the respective wild-type strains. Importantly, toxin production in a C. perfringens agrB mutant could be restored upon the addition of a synthetic C. perfringens AgrD peptide to an in vitro grown culture, supporting the hypothesis that agr-based quorum sensing controls toxin production in this bacterium (Vidal et al., 2012). Further evidence of the role of the C. perfringens agr system in modulating virulence was obtained using both a rabbit small intestinal loop model and a mouse challenge model, which confirmed that the agrB mutant was attenuated in virulence (Vidal et al., 2012). The situation is less clear in C. difficile since no agrBD mutants have been reported. However, a recent study showed that insertional inactivation of agrA, a response regulator located in the agr2 locus, in strain R20291 resulted in a twofold reduction in tcdA gene expression and a colonization defect in a mouse model of CDI. These results suggest that the agr2 quorum sensing system might play a role in regulating the virulence response of this organism (Martin et al., 2013). Since similar agrBD homologues appear widespread among the clostridia, it is likely that future research will show that similar quorum-sensing systems play an important role in controlling virulence factor expression in many toxigenic clostridial species, which may provide important targets for therapeutic development.

Figure 1.

Alignment of confirmed and putative AgrD amino acid sequences. The alignment shows the empirically confirmed AgrD peptides from S. aureus groups I–IV (Sau-group I–IV) (Otto, 2001), C. perfringens strain CN3685 (Cpe-CN3685 AgrD) (Vidal et al., 2012) and C. botulinum strain ATCC3502 (Cbo-3502 AgrD1/AgrD2) (Cooksley et al., 2010). Putative AgrD peptides identified by bioinformatic analysis of the genome sequences of C. difficile strain R20291 (Cdi-R20291 AgrD1/AgrD2), C. novyi strain ATCC19402 (Cno-19402 AgrD), C. sordellii strain ATCC9714 (Cso-9714 AgrD1/AgrD2/AgrD3), C. chauvoei strain JF4335 (Cch-JF4335 AgrD) and C. septicum strain BX96 (Cse-BX96 AgrD) are also shown. The peptides are split into three regions representing the N-terminal amphipathic region, the AIP encoding region and the C-terminal charged region. The amino acids predicted or confirmed to form the thiolactone/lactone ring structure of the mature AIP are highlighted in yellow and a highly conserved proline residue found in all confirmed and putative AgrD peptides is highlighted in blue.

Two-component signal transduction systems

The most common way by which bacteria detect and respond to changes in their environment is through two-component signal transduction systems. These phosphorelay systems generally consist of a sensor histidine kinase that detects a specific environmental signal, and a response regulator that converts that signal into a cellular response by controlling the expression of a particular subset of genes (Capra and Laub, 2012).

Among the pathogenic clostridia, the best-studied two-component system is found in C. perfringens. The production of several toxins, as well as many extracellular enzymes, is regulated by the VirSR two-component signal transduction system, with inactivation of either virS or virR resulting in an altered toxin production profile compared to the wild-type strain (Lyristis et al., 1994; Shimizu et al., 1994; Ohtani et al., 2003; Cheung et al., 2010; Ma et al., 2011) and in the case of virS, an attenuated in vivo virulence phenotype in a mouse myonecrosis model (Lyristis et al., 1994). This global regulatory network consists of the VirS sensor histidine kinase and its cognate response regulator, VirR (Lyristis et al., 1994; Shimizu et al., 1994). The regulatory cascade (Fig. 2) begins with the detection of a specific signal by VirS. For many years the nature of this signal was unknown, but recent work has suggested that one signal might be the AgrD quorum sensing AIP (Ohtani et al., 2009), described earlier. Following the detection of a stimulus, VirS autophosphorylates and then donates the phosphoryl group to VirR, leading to its activation. The activated VirR is then able to control the transcription of its target genes (Cheung et al., 2009).

Figure 2.

Model of the regulation of gene expression by the VirSR two-component signal transduction system. Positive regulation is indicated by the green arrow and plus symbols. Negative regulation is denoted by red lines and minus symbols. Upon detection of an external signal (potentially an AIP), the VirS sensor histidine kinase (shown in orange) autophosphorylates and then becomes the phospho-donor for its cognate response regulator, VirR (shown in blue). Phosphorylated VirR directly regulates the expression of pfoA, ccp, netB, vrr, virT, virU, CPF_1074,CPF_0461 and CPR_0761 by binding to the VirR boxes located in the promoter regions located upstream of these genes. VirR indirectly regulates the indicated toxin or virulence factor genes via the VR-RNA, VirU or VirT regulatory RNA molecules. The VR-RNA-controlled CPE1446/CPE1447 system positively regulates the expression of hyaluronidase genes but negatively regulates ccp, plc, pfoA and nanI transcription. The VirSR-independent VirX sRNA positively regulates the expression of pfoA, plc and colA, and negatively controls the transcription of cpe and the sporulation sigma factor genes sigF, sigE, sigG and sigK. Adapted from Cheung et al. (2013).

The genes comprising the VirSR regulon can be divided into two groups. The first group consists of six genes that are directly activated by VirR (Cheung et al., 2010; Ohtani et al., 2010), and includes the toxin genes pfoA (Cheung et al., 2004) and netB (Cheung et al., 2010), which encode for PFO and NetB toxin respectively. The second group in C. perfringens strain 13, includes 147 genes that VirR activates indirectly (Ohtani et al., 2010), again including several toxin genes such as plc, colA (Banu et al., 2000; Shimizu et al., 2002) and cpb2 (Ohtani et al., 2003), which encode α-toxin, κ-toxin and β2-toxin, respectively. Direct regulation of genes is mediated through binding of VirR to regions known as VirR boxes in the promoter regions of target genes (Cheung et al., 2004). VirR boxes contain two imperfect direct repeats that are essential for VirR-regulated transcription (Cheung and Rood, 2000). The maintenance of correct helical phasing and spacing between the VirR boxes and the −35 region of target gene promoters is also critical for optimal transcriptional activation (Cheung et al., 2004). Indirect regulation involves a different mechanism that utilizes an intermediate regulatory RNA molecule known as VR-RNA (Shimizu et al., 2002), discussed later in this review, the expression of which is positively controlled by VirR (Shimizu et al., 2002). It is noteworthy that virSR gene homologues have been identified in a number of the pathogenic clostridia; however, the role of these genes in regulating toxin production is yet to be determined.

Two-component signal transduction systems are also important in controlling the production of BoNT in C. botulinum. Antisense RNA targeted against three two-component systems, CLC_1093/CLC_1094, CLC_1914/CLC_1913 and CLC_0661/CLC_0663, resulted in the transcriptional activation of genes encoding both BoNT and the associated non-toxic protein (ANTP) (Connan et al., 2012), suggesting that these systems positively regulate neurotoxin production. Whether this regulation is direct or indirect, or if there is cross-talk between the three systems, remains to be determined. By contrast, the CBO0786/CBO0787 two-component system of C. botulinum strain ATCC3502 was found to negatively regulate BoNT production (Zhang et al., 2013), with mutants in either the gene encoding the sensor histidine kinase (cbo0787) or the response regulator (cbo0786) found to produce significantly higher levels of BoNT than the wild-type strain (Zhang et al., 2013). Experiments performed in vitro demonstrated that the CBO0786 response regulator bound to the BoNT and ANTP promoter regions, suggesting a direct mechanism of toxin inhibition (Zhang et al., 2013). The CBO0786 DNA binding sites identified by DNase I footprinting do not, however, appear to contain any repeat sequences, a characteristic feature of most other response regulator binding sites, and appear to show similarity only in the −10 box of the target gene promoters. This unexpected finding suggests that CBO0786 must use an as yet unknown mechanism to distinguish these binding sites from the promoters of other genes with similar pribnow boxes.

In C. difficile, toxin production is also influenced by several two-component signalling systems. For example, CDT-binary toxin is positively regulated by an orphan response regulator known as CdtR, which is located immediately upstream of the cdtA gene and within the CdtLoc, a chromosomal region encoding the structural binary toxin genes, cdtA and cdtB, as well as cdtR (Carter et al., 2007). In addition, Spo0A, the master regulator of sporulation in both Bacillus and Clostridium species, may play a role in regulating TcdA and TcdB production. Note that in C. perfringens Spo0A has previously been shown to regulate the production of CPE (Huang et al., 2004) and TpeL (Paredes-Sabja et al., 2011), providing evidence that this regulator plays a role in both clostridial toxin production and sporulation in this bacterium. At this time, however, the exact role of Spo0A in C. difficile is unclear, with a number of studies presenting conflicting data. In one of these studies, Spo0A positively regulated toxin production in strain 630Δerm (Underwood et al., 2009), while in another study Spo0A appeared to play no role in toxin production in the same strain (Rosenbusch et al., 2012). A third study using strain R20291, an epidemic BI/NAP1/027 isolate, showed that Spo0A negatively regulated toxin production (Deakin et al., 2012). The apparent discrepancy between these studies was clarified by a recent publication that suggests that Spo0A exerts differential regulatory effects on toxin production in different C. difficile strain backgrounds. In this work, the construction and complementation of spo0A mutants in two ribotype 027 isolates, M7404 and R20291, demonstrates that Spo0A acts as a negative regulator of TcdA and TcdB production in this strain background while equivalent mutations in 630Δerm and a ribotype 078 isolate, JGS6133, did not alter toxin production (Mackin et al., 2013). These results suggest that Spo0A regulation of toxin production in C. difficile may be an adaptation specific to ribotype 027 isolates.

Alternative sigma factors

In addition to two-component signalling systems, multiple alternative sigma factors regulate toxin production in the pathogenic clostridia. One such example is Sigma H (SigH), which is involved in the transition to stationary-phase growth and sporulation in both Bacillus and Clostridium species (Sauer et al., 1994). SigH negatively regulates toxin production in C. difficile, with a sigH mutant found to produce increased amounts of TcdA and TcdB in comparison to the wild-type strain (Saujet et al., 2011). By contrast, CPE production in C. perfringens is positively regulated by alternative sigma factors, with significantly reduced or abolished CPE expression observed in sigE, sigF and sigK null mutants that no longer produce the sporulation sigma factors Sigma E (SigE), Sigma F (SigF) and Sigma K (SigK) (Harry et al., 2009; Li and McClane, 2010). Three promoters, P1, P2 and P3, are responsible for coupling CPE production with sporulation (Zhao and Melville, 1998). Promoter P1 is similar to consensus SigK promoters while promoters P2 and P3 show similarity to other SigE-dependent promoters. The regulatory cascade controlling CPE production appears to be hierarchical since SigF regulates the expression of sigE and sigK, which then directly control cpe transcription from the SigE and SigK-dependent cpe promoters (Li and McClane, 2010).

The TcdR family of alternative sigma factors

The most thoroughly studied alternative sigma factors in the clostridia are the TcdR family of proteins. These proteins are members of the extracytoplasmic function (ECF) family of alternative sigma factors, which belong to group 5 of the σ70 family (Mani and Dupuy, 2001). TcdR is the prototype member of this family and is critical for the initiation of toxin production in C. difficile (Mani and Dupuy, 2001; Mani et al., 2002). TcdR homologues have recently been identified in species other than C. difficile (Fig. 3) and in some but not all cases are involved in the regulation of toxin production (Dupuy et al., 2006). BotR and TetR, for example, control the expression of BoNT and TeNT by C. botulinum and C. tetani, respectively (Raffestin et al., 2005), while TcsR controls the production of TcsL and TscH in C. sordellii (Sirigi Reddy et al., 2013). These proteins are similar enough to substitute for one another in DNA binding assays and run-off transcription experiments in vitro (Dupuy et al., 2006). However, only BotR and TetR (Dupuy et al., 2006) or TcdR and TcsR (Sirigi Reddy et al., 2013), respectively, were functionally interchangeable in vivo. This is in keeping with the phylogenetic relatedness of C. difficile and C. sordellii, which were recently shown to be members of the Peptostreptococcaceae family and not the Clostridiaceae family to which C. botulinum and C. tetani belong, although both families belong to the order Clostridiales (Yutin and Galperin, 2013). The variations in functional interchangeability possibly reflect the divergence within regions 2.4 and 4.2 of these proteins, which mediate binding to the −10 and −35 regions of their target toxin gene promoters, respectively (Dupuy et al., 2006). Note that while the −35 region of the target genes is highly conserved, the −10 region is more variable (Dupuy et al., 2006) which may consequently restrict promoter recognition by TcdR family proteins in more distantly related species. Nevertheless, given the overall level of similarity between these proteins, it seems likely that they arose from a common ancestor and subsequently co-evolved to develop a degree of specificity with the toxin genes that they regulate.

Figure 3.

A. Phylogenetic tree showing the genetic relatedness of toxin-associated TcdR family proteins at the amino acid level. Distance values represent the divergence of each protein as the number of amino acid substitutions in proportion to the total length of the protein alignment. Multiple sequence alignment was performed using the clustalw algorithm available at: http://www.ebi.ac.uk/tools/msa/clustalw2. Phylogenetic analysis of the TcdR multiple sequence alignment was performed using the neighbour-joining method available at: http://www.ebi.ac.uk/tools/phylogeny/clustalw2_phylogeny.

B. Schematic representation of toxin loci and genes regulated by TcdR family proteins in the pathogenic clostridia. The PaLoc region of C. difficile (Braun et al., 1996) is shown, which harbours the toxin structural genes tcdA and tcdB, and the accessory genes tcdR,tcdE and tcdC. The recently identified PaLoc-like region of C. sordellii (Sirigi Reddy et al., 2013) encoding the toxin genes tcsL and tcsH and the accessory genes tcsR and tcsE is also shown. The C. tetani locus (Dupuy et al., 2006) contains the TeNT toxin gene tetX and the accessory gene tetR while the C. botulinum type A toxin locus (Dupuy et al., 2006) contains the BoNT toxin gene botA and the accessory gene botR. In addition to these genes, the C. botulinum locus encodes the non-toxic-non-haemagglutinin (ntnh) gene and the haemagglutinin genes (ha70,ha17 and ha30), the products of which form part of the BoNT complex. Similarly coloured genes encode for functionally related proteins: blue, TcdR family alternative sigma factors; yellow, large clostridial toxins; green, neurotoxins; red, putative holin-like proteins; orange, anti-sigma factor; pink, non-toxic-non-haemagglutinin protein and grey, haemagglutinin proteins.

The pathogenicity locus (PaLoc)

In C. difficile, the tcdA and tcdB toxin genes and the tcdR gene are located within a 19.6 kb chromosomal pathogenicity locus, known as PaLoc (Braun et al., 1996) (Fig. 3B). This region contains two additional genes, tcdE, which appears to encode a bacteriophage holin-like protein possibly involved in toxin export (Govind and Dupuy, 2012), and tcdC, a putative anti-sigma factor (Carter et al., 2011b). The tcsL and tcsH toxin genes of C. sordellii were also shown to reside within a PaLoc-like region (Fig. 3B) that also harbours tcsR and a gene with homology to the C. difficile tcdE gene (Sirigi Reddy et al., 2013). Note that TcsL and TcsH belong to the large clostridial toxin, or LCT, family, which also includes C. difficile TcdA (Sullivan et al., 1982) and TcdB (Sullivan et al., 1982), TcnA from C. novyi (Bette et al., 1991) and TpeL from C. perfringens (Amimoto et al., 2007). The observation that C. sordellii tcsL and tcsH are found in a PaLoc region similar to that found in C. difficile suggests that the LCT genes might all be located within similar loci. Since C. difficile PaLoc appears to be of bacteriophage origin (Braun et al., 1996) and tcnA and tpeL from C. novyi and C. perfringens are associated with a lysogenic bacteriophage and a conjugative plasmid, respectively, the movement of these toxin loci between species from a common ancestor seems highly plausible. Whether tcnA and tpeL also reside within PaLoc-like regions is currently unknown.

TcdC

The TcdC protein is thought to be a negative regulator of toxin production in C. difficile (Dupuy et al., 2008). Interest in this regulator increased when it was reported that epidemic BI/NAP1/027 strains carry a nonsense mutation within tcdC (Warny et al., 2005), leading to the hypothesis that TcdC inactivation may be responsible for the increased virulence of these strains (Warny et al., 2005). Despite numerous studies designed to address the role of TcdC in toxin production and virulence, the picture remains unclear, with conflicting findings reported (Matamouros et al., 2007; Carter et al., 2011b; Bakker et al., 2012; Cartman et al., 2012). In one study, TcdR-dependent in vitro transcription assays in the presence of purified TcdC protein, as well as assays performed in the surrogate host C. perfringens using a tcdA promoter-reporter gene fusion, showed that TcdC represses toxin gene expression by interfering with binding of TcdR-associated RNA polymerase to the PaLoc gene promoters (Matamouros et al., 2007). However, a second study showed that the insertional inactivation of the tcdC gene in C. difficile strain 630Δerm had little impact on toxin production (Bakker et al., 2012). In a third study, the introduction of a plasmid-borne copy of tcdC into the BI/NAP1/027 strain M7404 resulted in the downregulation of toxin production and an attenuated virulence phenotype in the hamster model of infection (Carter et al., 2011b). However, in a fourth study, correction of the tcdC mutation on the chromosome of the BI/NAP1/027 isolate R20291 resulted in no discernible effect on toxin production (Cartman et al., 2012). The reasons for these conflicting data are not clear, but may be related to experimental variation between the studies, such as different growth media and strains being used, or it might be that other important differences account for the disparate in vitro and in vivo results. Whatever the reasons, further work is needed to conclusively define the role of TcdC in C. difficile, particularly on virulence capacity. However, based on currently available data it is unlikely that the tcdC mutation alone is responsible for the increased virulence of BI/NAP1/027 isolates (Carter et al., 2011b).

Small regulatory RNAs

In addition to protein regulators, bacteria utilize another class of regulatory molecule known as small regulatory RNAs (sRNA). These RNA regulators can vary in length from 50 to 300 nucleotides and act either in cis or in trans (Storz et al., 2011). The majority of sRNAs interact with mRNA targets through an antisense mechanism and can alter transcription, translation and/or mRNA stability of target genes (Lalaouna et al., 2013). By use of in silico sRNA prediction algorithms and deep sequencing, putative sRNAs have been identified in several clostridial genomes including C. difficile (Soutourina et al., 2013), C. botulinum (Chen et al., 2011b), C. tetani (Chen et al., 2011b), C. novyi (Chen et al., 2011b) and C. perfringens (Chen et al., 2011b). With the exception of C. perfringens, none have been shown to regulate toxin production, although it seems likely that sRNA-mediated regulation of toxin production in the pathogenic clostridia will be experimentally proven in the future, as has been found in other bacteria.

In C. perfringens four sRNA molecules, known as VR-RNA, VirU, VirT and VirX, have been shown to regulate the production of various toxins. Of these, the best characterized is VR-RNA. This regulatory RNA is encoded by the vrr gene and is the crucial link between the VirSR system and the genes targeted in an indirect manner by this regulatory system. The mechanism by which VR-RNA regulates target gene expression is not well understood but appears to be dependent on the specific target mRNA. Regulation of κ-toxin expression, for example, relies on base pairing of the 3′ region of VR-RNA with the 5′ untranslated region (UTR) of the κ-toxin colA gene mRNA. This results in the disruption of a stem loop structure within the 5′ UTR and the exposure of a sequestered ribosome binding site. This conformational change also leads to cleavage of the colA mRNA, resulting in stabilization of the transcript and increased levels of translation (Obana et al., 2010). Although the 3′ end of VR-RNA was also found to be important for α-toxin expression, there does not seem to be any complementarity with this region and the 5′ UTR of plc mRNA, suggesting that a different mechanism to that described for κ-toxin is utilized.

The second sRNA, called VirU, has been shown to increase the levels of vrr, pfoA and virT mRNA, suggesting that it acts as a positive regulator or that it increases the stability of these mRNAs (Okumura et al., 2008). By contrast, VirT was found to negatively regulate the transcription of pfoA and colA (Okumura et al., 2008). The mechanism by which these sRNAs control their targets remains unclear, but it is postulated that they might be involved in preserving balanced gene expression by fine-tuning the transcription of VirSR-regulated genes (Okumura et al., 2008).

The final sRNA, known as VirX, activates transcription of the plc, colA and pfoA genes (Ohtani et al., 2002), and suppresses expression of CPE (Ohtani et al., 2013). While the exact mechanism by which VirX regulates plc, colA and pfoA gene expression remains to be determined, the control of CPE production is thought to result from VirX-mediated downregulation of the sporulation sigma factors SigF, SigE and SigK, which are important for activation of CPE production during sporulation (Ohtani et al., 2013), as discussed earlier.

Conclusions and future perspectives

The development of new technologies that have facilitated the study of the pathogenic clostridia at the molecular level have, for the first time, allowed the regulatory cascades and environmental signals that control the production of toxins in these bacteria to be dissected. The regulatory pathways involved are complex and multi-faceted, and detailed research is needed before a comprehensive understanding is achieved. The use of in vivo experiments is of particular importance, since in many cases it is not clear whether in vitro conditions used to study toxin production are truly reflective of the in vivo conditions that induce the production of these toxins. Nevertheless, it is becoming increasingly clear that toxin production between different clostridial species involves conserved regulatory mechanisms including quorum sensing, alternative sigma factors and two-component signalling systems. Importantly, with the development of new antibiotics in decline, understanding these regulatory networks may identify important new targets for the development of novel antimicrobial or anti-virulence compounds that act against the pathogenic clostridia and that may be efficacious in treating diseases caused by this group of bacteria.

Acknowledgements

Research at Monash University was supported by NHMRC Project Grants from the Australian National Health and Medical Research Council and an ARC Discovery Grant from the Australian Research Council. D.L. was supported by an ARC Future Fellowship from the Australian Research Council. The authors have no conflict of interest to declare.

Ancillary