SpoIIID-mediated regulation of σ^K function during Clostridium difficile sporulation

Summary

The spore-forming bacterial pathogen Clostridium difficile is a leading cause of health-care-associated diarrhea worldwide. Although C. difficile spore formation is essential for disease transmission, the regulatory pathways that control this developmental process have only been partially characterized. In the well-studied spore-former Bacillus subtilis, the highly conserved σ^E, SpoIIID and σ^K regulatory proteins control gene expression in the mother cell to ensure proper spore formation. To define the precise requirement for SpoIIID and σ^K during C. difficile sporulation, we analyzed spoIIID and sigK mutants using heterologous expression systems and RNA-Seq transcriptional profiling. These analyses revealed that expression of sigK from a SpoIIID-independent promoter largely bypasses the need for SpoIIID to produce heat-resistant spores. We also observed that σ^K is active upon translation, suggesting that SpoIIID primarily functions to activate sigK. SpoIIID nevertheless plays auxiliary roles during sporulation, as it enhances levels of the exosporium morphogenetic protein CdeC in a σ^K-dependent manner.Analyses of purified spores further revealed that SpoIIID and σ^K control the adherence of the CotB coat protein to C. difficile spores, indicating that these proteins regulate multiple stages of spore formation. Collectively, these results highlight that diverse mechanisms control spore formation in the Firmicutes.

Introduction

Clostridium difficile is a leading cause of health-care-associated infections worldwide (Carroll and Bartlett, 2011; Depestel and Aronoff, 2013). While nosocomial infections by C. difficile have increased markedly in the past decade (Depestel and Aronoff, 2013; Valiente et al., 2014), it is now apparent that this enteropathogen is also a significant cause of community-acquired diarrhea (Khanna et al., 2012; Deshpande et al., 2013). In severe cases, C. difficile-associated disease can lead to pseu-domembranous colitis, toxic megacolon and death. These pathologies are primarily caused by the large glucosylat-ing toxins TcdA and TcdB, which are secreted by C. difficile and induce significant inflammation of the gut (Lyras et al., 2009; Kuehne et al., 2010; Pruitt and Lacy, 2012; Shen, 2012; Peniche et al., 2013).

While the toxins cause these disease symptoms, the ability of C. difficile to produce metabolically dormant spores is critical to its success as a pathogen, as spores are the primary infectious agent of this obligate anaerobe (Shaughnessy et al., 2011; Deakin et al., 2012; Francis et al., 2013; Paredes-Sabja et al., 2014). C. difficile infections begin when spores ingested from the environment survive passage through the gastrointestinal tract, germinate in response to specific bile salts in the small intestine and establish a toxin-secreting, vegetative cell population in the colon (Carroll and Bartlett, 2011; Francis et al., 2013). During growth in the gut, C. difficile induces a transcriptional program that leads to spore formation (Janoir et al., 2013); excretion of C. difficile spores presumably facilitates its transmission and increases disease recurrence (Maroo and Lamont, 2006; Gerding and Johnson, 2010; Dubberke, 2012; Depestel and Aronoff, 2013). As a result, spores represent an important environmental reservoir for C. difficile (Shaughnessy et al., 2011; Hoover and Rodriguez-Palacios, 2013), especially as they are highly resistant to common disinfectants, inert to antibiotics and readily transmissible (Lawley et al., 2010; Carroll and Bartlett, 2011; Paredes-Sabja et al., 2014). However, even though spores are critical for C. difficile pathogenesis, the mechanisms controlling their formation have only been partially characterized.

Bacterial spore formation is an ancient, highly regulated developmental process (de Hoon et al., 2010; Higgins and Dworkin, 2012; McKenney et al., 2012). Sporulation begins when asymmetric division creates a larger mother cell compartment and smaller forespore compartment. The mother cell engulfs the forespore and then assembles a series of protective structures around the forespore. These structures include the cortex, a modified layer of peptidoglycan that helps maintain metabolic dormancy (Popham, 2002; Setlow, 2006; Leggett et al., 2012), and the spore coat, a series of proteinaceous shells that protect the spore from enzymatic and oxidative insults (Henriques and Moran, 2007; McKenney et al., 2012). The spore coat is encased by an additional layer known as the exosporium in many Firmicutes, including C. difficile (Henriques and Moran, 2007; Paredes-Sabja et al., 2014).

Extensive studies in the model organism Bacillus subtilis have revealed that a complex regulatory network coordinates these morphological changes. Regulation at the transcriptional level is mediated by a hierarchical cascade consisting of the activation of a master transcriptional regulator, Spo0A, followed by the expression of genes encoding four sporulation-specific sigma factors, σ^F, σ^E, σ^G and σ^K (de Hoon et al., 2010; Higgins and Dworkin, 2012). The sigma factors are also subject to significant regulation at the post-translational level (Errington, 2003; Higgins and Dworkin, 2012). In B. subtilis, dephosphorylation of an anti-anti-σ factor liberates σ^F from binding to its anti-σ factor (LaBell et al., 1987; Margolis et al., 1991; Arigoni et al., 1996), regulated proteolysis activates both σ^E and σ^K by removing an inhibitory peptide (Cutting et al., 1990; Lu et al., 1990; Peters and Haldenwang, 1991), and small molecule transport by a secretion apparatus is proposed to activate σ^G through an unknown mechanism (Meisner et al., 2008; Camp and Losick, 2009; Doan et al., 2009). These activation mechanisms restrict sigma factor function to specific compartments and create a criss-cross dialogue between the forespore and mother cell (Losick and Stragier, 1992), with σ^F activating σ^E in the mother cell, σ^E activating σ^G in the forespore and σ^G activating σ^K in the mother cell (Errington, 2003; Higgins and Dworkin, 2012).

While the sporulation sigma factors are conserved across all spore-forming bacteria (de Hoon et al., 2010; Galperin et al., 2012), recent studies have shown that the order in which they function exhibits considerable diversity between Firmicutes. In C. difficile, the forespore and mother cell lines of gene expression appear to be uncoupled, with σ^E activation being only partially dependent on σ^F activity, σ^F activating σ^G in the forespore independently of σ^E, and σ^E activating σ^K in the mother cell independently of σ^G (Fimlaid et al., 2013; Pereira et al., 2013; Saujet et al., 2013). In C. acetobutylicum, σ^K functions at two developmental stages: during sporulation initiation and at later stages of forespore development (Al-Hinai et al., 2014). In C. pefringens, σ^K may also function at two developmental stages: prior to σ^F activation and downstream of σ^E activation (Harry et al., 2009), while in C. botulinum, σ^K appears to regulate sporulation at earlier stages than in C. difficile and B. subtilis (Kirk et al., 2012).

These observations highlight the diverse mechanisms regulating sporulation sigma factor function in the Clostridia relative to those defined in B. subtilis. Indeed, σ^G activation does not appear to require the σ^F- and σ^E-controlled secretion apparatus in C. difficile in contrast with B. subtilis (Meisner et al., 2008; Camp and Losick, 2009; Doan et al., 2009; Fimlaid et al., 2013; Pereira et al., 2013; Saujet et al., 2013). C. difficile σ^K also does not appear to depend on proteolytic activation (Fimlaid et al., 2013) unlike B. subtilis and C. perfringens (Harry et al., 2009; Higgins and Dworkin, 2012), as C. difficile encodes a σ^K lacking an N-terminal inhibitory propeptide and is missing the genes encoding the σ^K processing machinery (Haraldsen and Sonenshein, 2003; de Hoon et al., 2010). Nevertheless, it is unclear if additional post-translational mechanisms regulate σ^K activity in C. difficile, as unknown post-translational mechanisms controlσ^G function in C. difficile (Fimlaid et al., 2013) and B. subtilis (Camp and Losick, 2009). Interestingly, the absence of the σ^K propeptide appears to be conserved exclusively among clostridial species belonging to the Peptostreptococacceae family such as C. difficile (Fig. S1; Yutin and Galperin, 2013), suggesting that Peptoclostridium species do not use the same mechanisms for regulating σ^K function as other Firmicutes.

Nested within this larger regulatory framework of the sporulation sigma factors are a series of feed-forward loops (FFLs) designed to induce pulses of gene expression in B. subtilis (Eichenberger et al., 2004; de Hoon et al., 2010). In sporulation FFLs, a sigma factor induces the production of a transcription factor that subsequently acts in concert with the sigma factor to regulate downstream sporulation gene expression. Although the FFLs defined in B. subtilis are not as widely conserved among the Clostridia as the sporulation sigma factors (de Hoon et al., 2010), the σ^E, SpoIIID, σ^K FFL is conserved across the Firmicutes in terms of gene presence (de Hoon et al., 2010; Galperin et al., 2012). In B. subtilis, σ^E activates the transcription of spoIIID (Kunkel et al., 1989), which encodes SpoIIID, a helix-turn-helix transcription factor (Chen et al., 2014). SpoIIID, in conjunction with σ^E, activates the expression of sigK (Kroos et al., 1989); it also represses both σ^E- and σ^K-regulated genes (Halberg and Kroos, 1994; Zhang et al., 1997; Ichikawa and Kroos, 2000; Eichenberger et al., 2004). Later in sporulation, a negative feedback loop consisting of σ^K-mediated downregulation of sigE and spoIIID expression ensures optimal spore formation (Halberg and Kroos, 1992; Zhang and Kroos, 1997; Zhang et al., 1999). SpoIIID additionally regulates σ^K production by activating transcription of spoIVCA, which encodes the recombinase required for excising the skin element that disrupts the sigK gene (Kunkel et al., 1989; Stragier et al., 1989).

The conservation and importance of this mother cell regulatory network during spore formation in the Clostridia has not been fully explored. In C. perfringens, spoIIID expression does not depend on σ^E (Harry et al., 2009), while in C. difficile, σ^E activates spoIIID, and SpoIIID together with σ^E activate sigK (Saujet et al., 2013) similar to B. subtilis. However, excision of the skin element disrupting C. difficile sigK gene (Haraldsen and Sonenshein, 2003) does not appear to be under SpoIIID-or σ^E-mediated control (Saujet et al., 2013). Excision of the skin element has been proposed to regulate the timing of C. difficile σ^K function based on studies done in merodiploid strains (Haraldsen and Sonenshein, 2003), although plasmid complementation studies have yielded mixed results (Fimlaid et al., 2013; Pereira et al., 2013). A recent microarray study in C. difficile showed that SpoIIID represses many σ^E-regulated genes (Saujet et al., 2013) similar to B. subtilis (Halberg and Kroos, 1994; Zhang et al., 1997; Ichikawa and Kroos, 2000; Eichenberger et al., 2004). While only two σ^K regulon genes were identified as being dependent on SpoIIID, subsequent qRT-PCR analyses revealed that SpoIIID is required for the activation of nine additional σ^K regulon genes (Saujet et al., 2013). Thus, SpoIIID likely regulates genes beyond those previously identified. Furthermore, the specific role of SpoIIID and σ^K in regulating spore formation has yet to be fully characterized. In particular, it remains unclear if SpoIIID is required for sporulation beyond its role in activating sigK transcription; whether post-translational mechanisms regulate σ^K; and how SpoIIID and σ^K regulate functional C. difficile spore formation (Fimlaid et al., 2013; Pereira et al., 2013; Saujet et al., 2013).

In this study, we sought to define the precise requirement for SpoIIID and σ^K in regulating C. difficile sporulation. In particular, we assessed the importance of SpoIIID function beyond activating sigK expression using comparative RNA-Seq and by uncoupling sigK expression from SpoIIID activation. We also tested the intrinsic activity of σ^K using an inducible sigK expression system and analyzed the germination efficiency and coat composition of sigK and spoIIID mutant spores.

Results

Loss of SpoIIID leads to defects in spore formation

In order to test the role of SpoIIID in regulating σ^K function during C. difficile sporulation, we first constructed a spoIIID mutant using the ClosTron gene disruption system (Heap et al., 2010) using the erm-sensitive parental strain JIR8094 (Fig. S2). Similar to the spoIIID mutant constructed in the 630Δerm background by Saujet et al. (2013), the JIR8094 spoIIID mutant appeared to be stalled after forespore engulfment based on phase contrast microscopy analyses (data not shown). As the spoIIID mutant morphology resembled that of our previously constructed sigK mutant (Fimlaid et al., 2013), we compared sporulating cultures of spoIIID and sigK strains by fluorescent microscopy using the lipophilic dye FM4–64 (to stain mother cell and forespore membranes) and Hoechst 33342 (to stain cell nucleoids). The most terminal phenotype observed for the spoIIID and sigK mutants was intense staining of the engulfed forespore by FM4–64 and exclusion of the nucleoid stain (Fig. 1), similar to previous analyses of spoIIID (Saujet et al., 2013) and sigK (Fimlaid et al., 2013) mutants. Additionally, spoIIID and sigK mutants forespores failed to exclude the FM4–64 membrane stain in contrast with wild type (Fig. 1, white arrows). These results suggest that the spoIIID and sigK mutants complete engulfment but not membrane fission, as they exclude the Hoechst stain (Setlow et al., 1991) but stain intensely with FM4–64 membrane dye (Pogliano et al., 1999; Sharp and Pogliano, 1999). As phase-bright fore-spores exclude FM4–64 in our assay (Fig. 1; Fimlaid et al., 2013), the spoIIID and sigK mutants appear defective in post-engulfment events, such as coat assembly (McKenney and Eichenberger, 2012) and/or cortex maturation.

Fig. 1.

Comparison of spoIIID and sigK mutant morphologies using light microscopy. C. difficile strains were grown on sporulation media for 18 hours and evaluated by live differential interference contrast (DIC) and fluorescence microscopy; the membrane was visualized using FM4-64 (red), while the nucleoid was visualized using Hoechst (blue). The merge of these images is shown for each strain. Yellow arrowheads indicate forespore regions that stain with FM4-64 and Hoechst; green arrowheads indicate forespores that stain with FM4-64 but not Hoechst; and white arrowheads with pink outline indicate mature forespores that exclude both the FM4-64 and Hoechst stains. Scale bars represent 5 µm.

Based on these observations, we predicted that the spoIIID mutant would be defective in coat formation around the forespore, as coat layers are not visible around the forespore of a sigK mutant by transmission electron microscopy (TEM) (Fimlaid et al., 2013; Pereira et al., 2013). Accordingly, TEM analysis of the spoIIID mutant failed to visualize coat laminations around the forespore (Fig. 2A), although both mutant strains appeared to produce cortex in contrast with the analogous mutants in B. subtilis (Kroos et al., 1989; Cutting et al., 1991a; Driks et al., 1994). Taken together, these analyses indicate that the spoIIID mutant morphologically resembles the sigK mutant.

Fig. 2.

Comparison of C. difficile sigK and spoIIID mutants by transmission electron microscopy (TEM).

A. TEM analysis of the forespore regions of wild-type (WT), sigK⁻ and spoIIID⁻ strains at 24 hours of growth on sporulation media.

B. TEM analysis of purified spores from the indicated strains. Black triangles indicate regions that resemble coat layers, while white triangles indicate regions consistent with cortex. Scale bars represent 100 nm.

To determine whether the morphological defects in the spoIIID mutant corresponded to sporulation defects, we measured heat-resistant spore formation in the spoIIID mutant using a plate-based sporulation assay. Loss of SpoIIID resulted in a 50,000-fold defect in spore formation relative to wild type (Fig. 1), whereas loss of σ^K failed to produce heat-resistant spores within the detection limits of this assay consistent with our previous findings (Fig. 1; Fimlaid et al, 2013). The heat-resistance defect of the sigK mutant was at least tenfold more severe than the defect reported by Pereira et al. (2013), while the defect of the spoIIID mutant was ~ 16-fold more severe than that measured by Saujet et al. (2013). These differences may reflect differences in sporulation assay conditions (plate-based vs. broth-based) and/or strain background (JIR8094 vs. 630Δerm). Importantly, complementation of the spoIIID mutant with a pMTL83151 plasmid that expresses spoIIID restored coat formation around the forespore and heat-resistant spore formation at levels similar to wild type and a previously complemented spoIIID mutant (Saujet et al., 2013; Fig. S3).

spoIIID and sigK mutant spores exhibit germination defects

As the inability to produce heat-resistant spores could result from defects in spore assembly and/or germination, we assessed the germination efficiency of spores isolated from the sigK and spoIIID mutants. Notably, while both mutants produced spores that could survive standard spore purification methods (Sorg and Dineen, 2009; Adams et al., 2012), sigKr spores failed to germinate altogether, and spoIIID⁻ spores exhibited a ~ 2-log defect in germination (Tables 1 and Table S1). Heat treatment had no effect on the germination efficiency of spoIIID⁻ spores (Table 1), suggesting that the defects detected by the heat-resistance assay (Fig. 1) are due at least in part to germination defects for the spoIIID mutant. It should be noted that spore isolation from the sigK and spoIIID mutants was markedly less efficient relative to wild type, as might be expected given the morphological defects observed by TEM (Fig. 2).

Table 1.

Germination efficiency of spoIIID⁻ and sigK⁻ spores in response to heat.

	Untreated	Heat-activated
WT	1 ± 0.4	1 ± 0.4
spoIIID−	0.01 ±0.002	0.01 ± 0.002
sigK−	< 10−6	< 10−6

Heat-activated spores were incubated at 60°C for 30 minutes prior to plating. The germination efficiencies represent the average of spores recovered from four replicates ± standard error of the mean (SEM).

To assess the effect of the sigK and spoIIID mutations on spore assembly, the morphologies of sigK⁻ and spoIIID⁻ spores were analyzed by TEM. The sigK and spoIIID mutants produced spores containing a thick cortex layer but no detectable coat layers (Fig. 2B), similar to the phenotype observed in sporulating cells (Fig. 2A). The ability of these mutants to produce spores with cortex, in contrast with their respective mutants in B. subtilis (Kroos et al., 1989; Cutting et al., 1991a) likely permitted their purification. Regardless, these analyses failed to detect morphological differences between the sigK and spoIIID mutant spores.

Whole genome transcriptional profiling of the mother cell line of gene expression

To gain insight into the observed functional differences between the sigKand spoIIID mutants, we used RNA-Seq analyses to compare the global transcriptional profiles of the sigK and spoIIID mutants relative to each other and to wild type. To further delineate how SpoIIID and σ^K control mother cell gene expression, we also compared these mutant transcriptional profiles to that of a sigE mutant. Three biological replicates of each strain were grown on sporulation media for 18 hours, and RNAwas isolated and processed for whole transcriptome sequencing. The wild-type, sigE⁻, and sigK⁻ RNA samples were identical to those used in a previous study (Fimlaid et al., 2013). However, new cDNA libraries were prepared using an alternate method (TruSeq). This new library construction dramatically improved alignment to the C. difficile 630 reference genome by 80–90% relative to the previous library construction method (Nugen) (Table S2; Fimlaid et al., 2013). Genome coverage and sequencing counts for each strain and replicate can be found in Table S2.

The DeSeq variance analysis package (Anders and Huber, 2010) was used to identify genes that were differentially expressed between the mutant strains and wild type by ≥ 4-fold with an adjusted p-value of ≤ 0.05. These pairwise analyses revealed that 187, 50 and 50 genes, respectively, exhibited reduced expression levels in the sigE⁻, sigK⁻ and spoIIID⁻ strains (Fig. 3; Tables S3–S5). Only 11, 3 genes and 0 gene, respectively, were expressed at higher levels in the sigE, sigK and spoIIID mutants relative to wild type (Table S6). These findings stand in contrast with previous microarray analyses, which identified 84 genes as being overexpressed in a spoIIID mutant relative to wild type (Saujet et al., 2013). Indeed, only two SpoIIID-activated genes were shared between the current data set and the previous microarray study (Fig. S4A; Saujet et al., 2013), even though our RNA-Seq analyses overlapped with previous RNA-Seq and microarray analyses of σ^E- and σ^K-activated genes (~ 70% overlap; Fig. S4B and C).

Fig. 3.

Venn diagram of genes identified as being dependent on either σ^E, SpoIIID and/or σ^K by RNA-Seq. Genes were defined as being dependent on their respective sigma factor for expression if their transcript levels were decreased by ≥ 4-fold with an adjusted p-value of ≤ 0.05 in the mutant strains relative to wild type. The pink circle represents σ^E-dependent genes, the green circle represents SpoIIID-dependent genes; and the purple circle represents σ^K-dependent genes. The genes identified in these analyses are listed in Tables S3-S5.

Comparison of the transcriptional profiles of the sigK⁻ and spoIIID⁻ strains identified seven genes that were differentially expressed ≥ 4-fold with an adjusted p-value of ≤ 0.05 between the two mutants, although the expression level differences were relatively small (between four- and sevenfold; Table S7). All seven genes were overexpressed in the spoIIID mutant relative to the sigK mutant, suggesting that SpoIIID could repress their expression. Four of these genes are part of the eight-gene spoIIIA operon, which was previously shown to be expressed at higher levels in a spoIIID mutant by microarray analyses (Saujet et al., 2013). However, subsequent qRT-PCR analyses on RNA isolated from five separate biological replicates of wild-type, spo0A⁻, sigE⁻, sigK⁻ and spoIIID⁻ strains grown under identical conditions revealed no statistically significant difference in either spoIIIAA or spoII-IAG expression between the spoIIID⁻ strain and wild type (Fig. 4). Conversely, these qRT-PCR analyses revealed large, statistically significant differences in gene expression between the spoIIID⁻ train and swild type for σ^K regulon genes, sleC, cotE, CD3580 and cdeC (Fig. 4; 100-fold, 55-fold, 224-fold and 27-fold respectively; p < 0.001), consistent with results of the RNA-Seq analyses (Tables S4 and S5). While no differences in gene expression levels for sleC, cotE and CD3580 were observed between the sigK and spoIIID mutants, cdeC transcript levels were significantly under-expressed in the spoIIID mutant than in the sigK mutant (~ 3-fold difference, p < 0.05; Fig. 4). qRT-PCR analyses also confirmed that sigK requires SpoIIID for transcriptional activation, with sigK transcript levels being 83-fold lower in the spoIIID mutant relative to wild type (p < 0.0001; Fig. 4). Taken together, the RNA-Seq and qRT-PCR analyses suggest that SpoIIID primarily functions to activate σ^K-regulated gene expression rather than repress sporulation gene expression.

Fig. 4.

qRT-PCR analyses of σ^E-, σ^K- and SpoIIID-regulated genes. Transcript levels for the indicated genes as measured by qRT-PCR on five biological replicates. Samples were distinct from those used for RNA-Seq. spoIIIAA, spoIIIAG and spoIVA were previously shown to be σ^E-activated genes; sleC, cotE, CD3580 and spoIIID were previously shown to be σ^K-activated (Fimlaid et al., 2013; Saujet et al., 2013). Transcript levels were calculated relative to the spo0A⁻ strain after normalization to the housekeeping gene rpoB using the standard curve method. Error bars indicate the standard error of the mean. Statistically significant changes in transcript levels were determined relative to WT and are represented by adjusted p-values determined by a Dunnett’s one-way ANOVA. ****p < 0.0001, ***p < 0.001, *p < 0.01. ND indicates not determinable as the region amplified is interrupted by the spoIIID targetron insertion. sigK transcripts could be measured in the sigK strain because the region amplified precedes the sigK targetron insertion.

In agreement with the transcriptional analyses, Western blotting showed that both the spoIIID and sigK mutants fail to produce σ^K-regulated gene products (SleC, CotE, CD3580 and CotE) even though they produce wild-type levels of the σ^E-regulated gene products SpoIVA and CotB (Fig. S5; Fimlaid et al., 2013). As the spoIIID⁻ strain also failed to produce σ^K (Fig. S5), the results comprehensively show that a hierarchical cascade consisting of σ^E, SpoIIID and σ^K controls mother cell gene expression.

Heterologous production of σ^K largely bypasses the need for SpoIIID during sporulation

As there was almost perfect overlap between genes that were activatedin a σ^E-, σ^K- and SpoIIID-dependent manner relative to wild type (Fig. 3), and little difference in the expression levels between σ^K- and SpoIIID-dependent genes (Table S7), the transcriptional analyses suggested that SpoIIID is mainly required to activate sigK transcription. This model would predict that the need for SpoIIID during sporulation can be suppressed by uncoupling sigK expression from SpoIIID activation. To test this hypothesis, we complemented the spoIIID mutant with a construct consisting of a rearranged sigK (lacking the skin element) expressed from a promoter exclusively activated by σ^E (P_sipL, referred to herein as P_E, Saujet et al., 2013). This SpoIIID-independent promoter alters the order in which sigK is expressed in the mother cell, as it no longer requires σ^E-mediated activation of spoIIID transcription or subsequent SpoIIID-mediated activation of sigK expression. Fusion of the P_E promoter to the gene encoding a SNAP-tag reporter (Donovan and Bramkamp, 2009; Nicolle et al., 2010) confirmed that this promoter is σ^E- but not SpoIIID- or σ^K-dependent (Fig. S6). It should be noted that Pereira et al. previously validated the use of SNAP-tags for anaerobic imaging in C. difficile (Pereira et al., 2013). The resulting P_E-sigK construct was expressed from the multicopy plasmids pMTL83151 (designated P_E-sigK) and pMTL84151 (designated 4P_E-sigK), with pMTL84151 being a higher copy plasmid in C. difficile (Heap et al., 2009). Spore formation in spoIIID mutants heterologously producing σ^K was assessed using phase contrast microscopy and heat resistance assays. SpoIIID-independent expression of sigK from the pMTL83151 plasmid restored production of phase-bright, heat-resistant spores to the spoIIID mutant (spoIIID⁻/P_E-sigK) at levels equivalent to the spoIIID complementation strain (spoIIID⁻/spoIIID; Fig. 5). In contrast, heterologous expression of sigK from the higher copy pMTL84151 vector in the spoIIID mutant (spoIIID⁻/4P_E-sigK) decreased heat-resistant spore formation 30-fold relative to the spoIIID⁻/P_E-sigK and the spoIIID⁻/spoIIID complementation strains (Fig. 5). Quantitative Western blotting for σ^K in the four biological replicates used for the heat resistance assays indicated that the spoIIID⁻/P_E-sigK strain produced slightly higher σ^K levels relative to WT, whereas the spoIIID⁻/4P_E-sigK strain produced almost fourfold higher σ^K levels relative to WT and ~ 3-fold more σ^K relative to the spoIIID⁻/P_E-sigK strain (Figs 6A and Fig S7). Although these differences were not statistically significant (Fig. S7), σ^K-dependent gene product levels appeared to increase slightly in the spoIIID⁻/4P_E-sigK strain relative to the wild-type and spoIIID⁻/P_E-sigK strains (Fig. 6A, lanes 4 and 5). Regardless, P_E-sigK strains in the spoIIID⁻ background produced σ^K-dependent gene products SleC, CotE, CD3580 and CdeC without affecting the levels of the σ^E-dependent gene product SpoIVA (Fig. 6A). Interestingly, CdeC levels were decreased in the spoIIID⁻/P_E-sigK and spoIIID⁻/4P_E-sigK strains relative to the wild-type and spoIIID complementation strains. This result suggests that SpoIIID regulates CdeC levels in a σ^K-dependent manner using transcriptional and/or potentially post-transcriptional methods.

Fig. 5.

SpoIIID-independent expression of sigK restores heat-resistant spore formation to spoIIID and sigK mutants. Phase contrast microscopy of C. difficile spoIIID⁻ and sigK⁻ strains carrying either empty vector (EV) or the indicated complementation constructs. The P_EsigK construct expresses a rearranged sigK (i.e. lacks the skin element) from a σ^E-regulated promoter (P_sipL, referred to as P_E) in a SpoIIID-independent manner. The sigK complementation construct expresses a rearranged sigK gene from its native promoter; the spoIIID complementation construct expresses spoIIID from the P_E-promoter. The pMTL83151 vector was used for all strains with the exception of those carrying the 4P_EsigK constructs; these latter strains carry the P_EsigK construct on the high-copy pMTL84151 plasmid. Strains were grown on sporulation media for 24 hours. The efficiency of heat-resistant spore formation was determined for each strain relative to wild type carrying an empty vector from five biological replicates. White arrowheads indicate mature phase-bright spores or forespores, and black arrowheads indicate immature phase-dark spores. Representative clones of each strain are shown, but at least two clones for each strain were tested. Scale bars represent 10 µm.

Fig. 6.

σ^K is active in strains expressing sigK in a SpoIIID-independent manner.

A. Western blot analyses of C. difficile strains spoIIID⁻ and sigK⁻ strains carrying either empty vector (EV) or the indicated complementation constructs. The P_E-sigK construct expresses a rearranged sigK (i.e. lacks the skin element) from a σ^E-regulated promoter (P_sipL, referred to as P_E) in a SpoIIID-independent manner. The sigK complementation construct expresses a rearranged sigK gene from its native promoter; the spoIIID complementation construct expresses spoIIID from the P_E-promoter. The pMTL83151 vector was used for all strains with the exception of those carrying the 4P_EsigK constructs; these latter strains carry the P_EsigK construct on the high-copy pMTL84151 plasmid. Strains were grown on sporulation media for 24 hours. SleC, CotE, CdeC and CD3580 (conserved hypothetical protein of unknown function) are σ^K-regulated gene products (Fig. S5, Fimlaid et al., 2013; Saujet et al., 2013). SpoIVA and CotB are σ^E-regulated gene products (Fig. S5, Fimlaid et al., 2013; Saujet et al., 2013). Spo0A was used as a loading control (Putnam et al., 2013). pro-SleC denotes the pro-enzyme form of SleC (Adams et al., 2012); activated SleC was not detected in sporulating cells.

B. qRT-PCR analysis of spoIIID transcript levels in the indicated strains. spoIIID transcript levels were calculated relative to WT carrying empty vector (EV) after normalization to the housekeeping gene rpoB using the standard curve method. Data shown represents the averages of four biological replicates. Error bars indicate the standard error of the mean. Statistically significant changes in transcript levels were determined relative to WT and are represented by adjusted p-values determined by a Dunnett’s one-way ANOVA ***p < 0.001. ND indicates not determinable as the region amplified is interrupted by the spoIIID targetron insertion.

Unfortunately, we were unable to measure SpoIIID levels in these strains by Western blotting, as antibodies raised against either C. difficile or B. subtilis SpoIIID, or a peptide derived from C. difficile SpoIIID, failed to specifically detect SpoIIID in sporulating C. difficile (data not shown). Nevertheless, analysis of spoIIID transcript levels by qRT-PCR revealed that the spoIIID⁻/spoIIID complement strain overexpresses spoIIID relative to wild type (p < 0.001; Fig. 6B), although this overexpression still allows for functional spore formation at levels similar to a previously described spoIIID complementation strain (Figs 5 and Fig S3; Saujet et al., 2013). Taken together, these results indicate that expression of a rearranged sigK gene in a SpoIIID-independent manner largely bypasses the requirement for SpoIIID during sporulation. The results further suggest that, even if SpoIIID represses σ^E-dependent gene transcription (Saujet et al., 2013), it is not required for heat-resistant spore formation when σ^K is heterologously produced.

As the sequence of sporulation gene expression can be critically important during spore formation in B. subtilis (Zhang and Kroos, 1997; Zupancic et al., 2001; Costa et al., 2007; Henriques and Moran, 2007; Wang et al., 2007a), we next examined whether changing the order of sigK expression in a sigK mutant affected functional spore formation. To this end, we tested whether the P_E-sigK construct, encoded on either the pMTL83151 or pMTL84151 vector, could complement the sporulation defect of the sigK⁻ strain. Expression of the rearranged sigK gene in a SpoIIID-independent manner from the pMTL83151 plasmid in the sigK mutant (sigK⁻/P_E-sigK) restored phase-bright, heat-resistant spore formation by greater than four logs (Fig. 5), although there was a 30-fold reduction in heat-resistance compared with the same construct in the spoIIID~background (Fig. 5). Expression of the P_E-sigK construct from the high copy pMTL84151 plasmid in the sigK mutant (sigK⁻/4P_E-sigK) failed to restore heat-resistant spore formation altogether (Fig. 5). Notably Western blot analyses revealed that both the sigK⁻/4P_E-sigKand sigK⁻/P_E-sigK strains overproduced σ^K (Fig. 6A; ~ 30-fold, p < 0.0001 and ~ 8-fold, p < 0.02; Fig. S7) and the σ^K-regulated gene products SleC, CotE and CD3580 relative to wild type. However, the levels of the σ^E-regulated proteins SpoIVA and CotB were decreased in the sigK⁻/ 4P_E-sigK strain relative to wild type and the other complementation strains (compare lanes 1, 7, 8, and 9; Fig. 6A). Transcription of spoIVA and other σ^E-activated genes (spoIIIAA, spoIIIAG and spoIIID) was also significantly lower (~ 50–100-fold, p < 0.001) relative to wild type as detected by qRT-PCR (Fig. 6B and Fig. S8). Taken together, the results suggest that altering the order of sigK expression in the presence of SpoIIID still permits heat-resistant spore formation but overproducing σ^K can inhibit sporulation.

Consistent with the previous qRT-PCR analyses of cdeC transcript levels (Fig. 4), the spoIIID mutant strain carrying empty vector produced - 10-fold less transcripts than the sigK mutant strain (p < 0.01; Fig. S8). Although no significant difference in cdeC transcript levels was observed between the sigK and spoIIID mutants expressing the P_E-sigK and 4P_E-sigK complementation constructs (Fig. S8), more CdeC was detected in the sigK⁻ background relative to a spoIIID⁻ background (Fig. 6A). Given that the sigK mutant expresses spoIIID at wild-type levels (Fig. 6B), these observations are consistent with SpoIIID functioning to both activate cdeC transcription and CdeC production and/or stability in a σ^K-dependent manner. CdeC levels were reduced in the sigK⁻/sigK complement relative to wild type for unknown reasons.

To test whether SpoIIID can increase cdeC transcription (Fig. 4) in the absence of σ^K, we overexpressed spoIIID in a sigK mutant (which already expresses spoIIID at wild-type levels; Fig. 6B). Overexpression of spoIIID in the sigK mutant failed to significantly increase transcript levels of cdeC relative to the parental mutant (Fig. S9), and no CdeC was detected in Western blot analyses in the sigK mutant overexpressing spoIIID (Fig. S10). These results indicate that SpoIIID requires σ^K in order to enhance cdeC expression. Overexpression of spoIIID in the sigK mutant also failed to increase the expression of other σ^K-regulated genes (specifically sleC, cotE and CD3580; Fig. S9), or detection of their gene products (Fig. S10), relative to the mutant strain carrying empty vector. Taken together, our results are consistent with the notion that SpoIIID-mediated regulation of sporulation gene expression requires σ^K.

Effect of SpoIIID-independent sigK expression on spore coat composition

To test whether heterologously expressing a rearranged sigK (lacking the skin element) in a SpoIIID-independent manner affected spore coat protein maturation and/or incorporation of coat specific proteins (Fig. 6A), we purified spores from wild-type, spoIIID⁻ and sigK⁻ strains carrying either empty pMTL83151 vector or the various complementation constructs. Spores could be isolated with varying efficiency from all of these strains except for the sigK⁻/4P_E-sigK strain, consistent with this strain having sporulation defects that may relate to its overproduction of σ^K (Fig. 5 and Fig. S7). Importantly, the germination efficiency of the purified spores correlated with the heat resistance assays (Table 2). Western blot analyses revealed that the processing of the spore coat-extractable proteins CotE, SleC, CdeC and SpoIVA (Permpoonpattana et al., 2011; Adams et al., 2012; Barra-Carrasco et al., 2013; Putnam et al., 2013) was reduced in complementation strains with lower levels of heat-resistant spore formation relative to wild type (spoIIID⁻/4P_E-sigK and sigK⁻/P_E-sigK; Figs 5 and Fig 7, lanes 5 and 8). Quantitation of the extent of processing in the mutants relative to wild type revealed that SleC, CotE and SpoIVA are processed ~ 10- to 100-fold less efficiently in the spoIIID⁻/4P_E-sigK and sigK⁻/P_E-sigK strains (Table S10). The effect was not as pronounced for CdeC, although sigK⁻/P_E-sigK spores exhibited a fivefold decrease in CdeC processing relative to wild type. The significance of the altered levels of processing is unclear at present. Furthermore, the levels of the coat-extractable proteins CD3580 (Fig. S11) and CotB (Permpoonpattana et al., 2013) were decreased in these strains relative to wild type (Fig. 7, lanes 5 and 8) despite being produced at higher levels during sporulation (Fig. 6), suggesting that these mutants have defects in coat protein incorporation during spore assembly.

Table 2.

Germination efficiency of spoIIID⁻ and sigK⁻ complementation strain spores.

	Untreated	Heat- activated
WT/EV	1 ± 0.3	1
spoIIID−/EV	0.002 ± 0.001	0.002
spoIIID−/spoIIID	0.5 ± 0.3	1.9
spoIIID−/PE-sigK	0.3 ± 0.1	0.1
spoIIID−/4PE-sigK	0.05 ± 0.02	0.1
sigK−/EV	< 10−6	< 10−6
sigK−/sigK	0.6 ± 0.2	1.5
sigK−/PE-sigK	0.006 ± 0.002	0.01

The P_EsigK construct expresses sigK from a σ^E-regulated promoter (P_sipL) independently of SpoIIID. The pMTL83151 vector was used for all strains with the exception of the spoIIID mutant carrying the 4P_EsigK construct, which expresses the P_EsigK construct on the high-copy pMTL84151 plasmid. The germination efficiencies for untreated spores represents the average of spores recovered from four replicates ± SEM. Heat-activated spores were incubated at 60°C for 30 minutes prior to plating. The germination efficiency for heat-activated spores represents the average of two replicates.

Fig. 7.

Effect of heterologous sigK expression in sigK⁻ and spoIIID⁻ strains on spore coat composition. Western blot analyses of purified spores isolated from C. difficile sigK⁻ and spoIIID⁻ strains carrying either empty vector (EV) or the indicated complementation constructs using the indicated antibodies. The P_EsigK construct expresses a rearranged sigK (i.e. lacks the skin element) from a σ^E-regulated promoter (P_sipL, referred to as P_E) in a SpoIIID-independent manner. The sigK complementation construct expresses a rearranged sigK gene from its native promoter; the spoIIID complementation construct expresses spoIIID from the P_E-promoter. The pMTL83151 vector was used for all strains with the exception of the spoIIID⁻/4P_E-sigK strain and the sigK⁻/4P_E-sigK, which express the P_EsigK construct from the high-copy number pMTL84151 plasmid. Spore formation by the sigK strain carrying 4P_EsigK was too low to permit efficient isolation of purified spores. SpoIVA is a σ^E-regulated gene product that is predicted to localize to the basement layer of the inner spore coat (McKenney and Eichenberger, 2012) and undergoes processing during incorporation into wild-type spores (Putnam et al., 2013); both full-length and processed forms of SpoIVA are shown. CotB and CotE are σ^E- and σ^K-regulated, spore surface-localized gene products respectively (Permpoonpattana et al., 2011; 2013). CotE undergoes processing during incorporation into wild-type spores (Permpoonpattana et al., 2011); both full-length and cleaved forms are shown. SleC undergoes processing during sporulation; pro-SleC denotes the pro-enzyme form of SleC (Adams et al., 2012); activated SleC was not detected in purified spores.

Notably, CotB was largely absent from spoIIID and sigK mutant spores (Fig. 7), even though it is produced at wild-type levels in sporulating cells of these mutants (Fig. S5; Fimlaid et al., 2013). These results indicate that the spoIIID and sigK mutant spores are impaired in their ability to incorporate at least one coat protein on their surface.

σ^K production alone is sufficient to activate σ^K-regulated gene expression

While these analyses revealed that expression of sigK in a SpoIIID-independent manner during sporulation is sufficient to lead to active σ^K production and functional spore formation, they did not indicate whether σ^K activity depends upon additional factors produced during sporulation. To test whether σ^K is active upon translation, we used an inducible promoter system to heterologously express sigK under vegetative cell conditions. A construct containing a rearranged sigK (lacking the skin element) and its presumed ribosome binding site was cloned downstream of a tetracycline-inducible promoter (Fagan and Fairweather, 2011) and conjugated into both wild type and a sigE mutant. The resulting strains were grown in BHIS broth to permit vegetative cell growth and induced with increasing concentrations of anhydrotetracycline. It should be noted that sporulation is undetectable under these growth conditions (Putnam et al., 2013). Whole cell lysates were analyzed by Western blotting to assess production of σ^K and σ^K-regulated gene products. Dose-dependent production of σ^K and σ^K-dependent gene products, SleC and CotE, was observed in both wild type and the sigE mutant during vegetative cell growth (Fig. 8). These results indicate that C. difficile σ^K is active upon production and does not require post-translational activation by additional sporulation proteins, in contrast with B. subtilis (Cutting et al., 1990; Lu et al., 1990).

Fig. 8.

Inducible expression of sigK during vegetative cell growth produces active σ^K. Western blot analyses of σ^K and σ^K-dependent gene products SleC and CotE in wild-type and sigE mutant strains expressing sigK under the control of a tetracycline-inducible promoter from the pRPF185 plasmid (Fagan and Fairweather, 2011). Strains were grown in BHIS broth for 2 hours in the presence of varying concentrations of anhydrotetracycline (ATc).

Discussion

Extensive studies of σ^E-, SpoIIID- and σ^K-mediated regulation of mother cell gene expression in B. subtilis have shown that this regulatory system fine-tunes the timing of sporulation gene transcription in a manner that ensures proper spore formation (Lu and Kroos, 1994; Ichikawa and Kroos, 2000; Wang et al., 2007a,b). A FFL consisting of σ^E-mediated activation of spoIIID leads to the SpoIIID-dependent rearrangement of the sigK gene, activation of sigK expression and subsequent SpoIIID-mediated down-regulation of σ^E- and σ^K-regulated genes (Kroos et al., 1989; Kunkel et al., 1989; Halberg and Kroos, 1994; Halberg et al., 1995). Proteolytic activation of σ^K (Cutting et al., 1990; Lu et al., 1990) induces a negative feedback loop that creates a pulse in SpoIIID function (Halberg and Kroos, 1992; Zhang and Kroos, 1997; Zhang et al., 1999). In addition to its roles in stimulating the rearrangement of the sigK gene, activating its transcription and promoting the post-translational activation of σ^K, SpoIIID plays additional undefined role(s) in regulating sporulation (Lu and Kroos, 1994). Our present analysis of this system suggests a simpler regulatory architecture in C. difficile: SpoIIID is primarily required to activate sigK expression, and σ^K is active upon translation.

Our results further show that, in contrast with B. subtilis, C. difficile (i) SpoIIID and σ^K are dispensable for cortex production (Fig. 2); (ii) SpoIIID does not appear to repress σ^E regulon gene expression (Table S8); (iii) SpoIIID enhances σ^K levels through a post-transcriptional mechanism (Figs. 6) and (iv) SpoIIID increases the levels of a σ^K-regulated exosporium protein, CdeC, through tran-scriptional and post-transcriptional mechanisms (Figs 4 and 6). Furthermore, σ^K-regulated gene products are required for coat proteins to polymerize and stay adhered to purified spores (Figs 2 and 7). These results, combined with previously published work (Pereira et al., 2013), indicate that coat protein localization to the developing fore-spore is not sufficient to predict its incorporation into mature C. difficile spores, suggesting that C. difficile may use a different mechanism for retaining coat onto the spore surface relative to B. subtilis.

Our finding that the major requirement for SpoIIID during sporulation is to indirectly activate σ^K-dependent genes is based on the following observations: (i) producing σ^K in a SpoIIID-independent manner (spoIIID⁻/P_E-sigK) is sufficient to restore heat-resistant spore formation to a spoIIID mutant at levels similar to the spoIIID complementation strain (Fig. 5), and (ii) σ^K- and SpoIIID-dependent genes essentially overlap in comparative RNA-Seq studies (Fig. 3). Our findings stand in contrast with a previous microarray analysis showing that SpoIIID represses many σ^E-regulated genes (Saujet et al., 2013). As our qRT-PCR analyses failed to detect overexpression of genes identified as being SpoIIID-repressed in the microarray study (Fig. 4; Saujet et al., 2013), and a spoIIID mutant heterologously expressing sigK produces functional spores (Fig. 5), SpoIIID-mediated repression of gene expression, if it occurs, is not required for spore formation. Several factors could account for the observed differences in gene expression between these two studies, such as differences in growth conditions (solid vs. liquid media), expression cutoff values (fourfold vs. twofold) and/or sporulation time point (18 vs. 15 hours; Saujet et al., 2013). Consistent with the latter hypothesis, after 24 hours of growth in liquid sporulation media, qRT-PCR analyses by Saujet et al. detected 12 σ^K-activated genes as being overexpressed in the spoIIID mutant relative to wild type (Saujet et al., 2013). Importantly, all of those genes require SpoIIID for expression in the current study (Fig. S4A and Table S5).

The difference in expression cutoff values likely does not explain the difference in the identification of SpoIIID-regulated genes between the two studies as the number of genes identified as being repressed in a SpoIIID-dependent manner only increased by seven when the expression cutoff values were relaxed from fourfold to twofold; the number of genes identified as being repressed in a σ^E- or σ^K-dependent manner increased only slightly (27 and 8 respectively; Table S8). In contrast, the number of genes identified as being activated in a SpoIIID-, σ^E- or σ^K-dependent manner with the twofold cutoff increased by 17, 86 and 24 relative to the fourfold cutoff (Fig. S12 and Table S9). While the overlap between σ^K- and SpoIIID-activated genes remained high (62; Fig. S12), this reduced stringency increased the false discovery rate. For example, with the relaxed cutoff, 36 of the 86 genes now identified as σ^E-activated are not Spo0A–dependent (data not shown). sigG and sspB are identified as being σ^E-activated even though qRT-PCR analyses have previously shown that they are expressed independently of σ^E (Fimlaid et al., 2013), and spoIIID is identified as being σ^K-dependent even though qRT-PCR analyses indicate that it is expressed independently of σ^K (Fig. 6B).

While our current RNA-Seq analysis identified 30 genes that have not been previously identified as being induced during sporulation in a σ^E-dependent manner (Fig. S4), this increased level of detection likely results from improved fidelity in the library construction. This conclusion is based on the observation that, even though the libraries were constructed from the sameRNAsamples, 99% of the reads mapped to the C. difficile genome in the current study, whereas 10–20% mapped to the C. difficile genome in our previous study (Fimlaid et al., 2013).As a result, the RNA-Seq analyses described in the current study were likely more sensitive at detecting σ^E-dependent transcripts. Consistent with this hypothesis, the majority of genes newly identified as σ^E-dependent in the current study are expressed at low levels (27/30 have a base mean < 100); the remaining three genes were not previously annotated.

Our findings further indicate that C. difficile σ^K has intrinsic activity upon translation (Fig. 8) as inducible sigK expression in vegetative cells leads to the production of σ^K-dependent gene products SleC and CotE in both a wild-type and sigE⁻ background (Fig. 8). While this result could be predicted by the lack of an inhibitory pro-peptide on C. difficile σ^K (Fig. S1; Haraldsen and Sonenshein, 2003), we note that both σ^F and σ^G are activated through post-translational mechanisms that do not involve proteolytic cleavage in C. difficile and B. subtilis (Meisner et al., 2008; Camp and Losick, 2009; Doan et al., 2009; Fimlaid et al., 2013). Although it remains formally possible that additional post-translational mechanisms regulate the level of σ^K activation during sporulation, our results confirm that σ^K activity does not require forespore-mediated signaling as it does in B. subtilis (Cutting et al., 1990; Lu et al., 1990; Fimlaid et al., 2013; Pereira et al., 2013; Saujet et al., 2013). While most other Firmicutes are predicted to produce σ^K with a pro-peptide (de Hoon et al., 2010), Peptoclostridium spp. such as C. difficile (Yutin and Galperin, 2013) appear to lack the inhibitory N-terminal pro-peptide (Fig. S1). Based on these observations, we predict that σ^K produced by other Peptoclostridium spp. is likely to be active upon translation.

A functional consequence of this mode of regulation is that the sequence in which sigK is expressed does not appear to be as important for producing spores in C. difficile as in B. subtilis. Expression of a rearranged sigK gene in the absence of SpoIIID in B. subtilis fails to lead to heat-resistant spore formation (Kunkel et al., 1990) as the post-translational activation of σ^K is also SpoIIID-dependent (Lu and Kroos, 1994). In contrast, in C. difficile, expression of the analogous construct in a spoIIID mutant restores heat-resistant spore formation and production of σ^K-dependent gene products at levels equivalent to the spoIIID complementation strain (Figs 5–8).

Interestingly, a previous study indicated that plasmid complementation of a sigK mutant could only be achieved by expressing a sigK containing a truncated skin element (Pereira et al., 2013). We hypothesize that the lack of complementation arises from the use of the high copy number pMTL84121 plasmid as expressing a rearranged sigK from the equally high copy number pMTL84151 plasmid fails to restore sporulation to a sigK mutant (our unpublished observation), while the same construct expressed from the lower copy pMTL83151 plasmid complements sporulation close to wild-type levels (sigK⁻/sigK; Fig. 5; Fimlaid et al., 2013). Furthermore, in the current study, complementation strains that overproduce σ^K relative to wild type exhibit lower levels of functional spore formation (spoIIID⁻/4P_E-sigK, sigK⁻/ P_E-sigK and sigK⁻/4P_E-sigK; Fig. 5 and Fig. S7). As high levels of σ^K correlate with reduced σ^E-regulated gene expression and protein production (Figs 6 and 7 and Fig. S8), a possible explanation for these observations is that σ^K overproduction reduces sporulation by down-regulating σ^E function through competition for RNA polymerase. Consistent with this hypothesis, Western blot analyses indicate that σ^E levels are similar in heterologous sigK complementation strains relative to wild type (data not shown). Nevertheless, our results do not rule out the possibility that skin element excision regulates when active σ^K is produced, as previously proposed (Haraldsen and Sonenshein, 2003; Pereira et al., 2013). Identifying the recombinase responsible for excision would be interesting as the gene encoding the putative recombinase, CD1231, was not detected to be σ^E- or SpoIIID-regulated in this study or in a previous study (Saujet et al., 2013).

Although SpoIIID regulates sporulation primarily through its ability to activate sigK transcription, we present data indicating that SpoIIID plays auxiliary regulatory roles. In particular, SpoIIID enhances the levels of the σ^K-dependent gene product CdeC, a major exosporium morphogenetic protein (Barra-Carrasco et al., 2013) in a σ^K-dependent manner (Figs 6 and 7). Given that cdeC is the most highly expressed gene induced during sporulation (Fimlaid et al., 2013) and one of the most abundant proteins detected in C. difficile spores (Lawley et al., 2009), SpoIIID appears to be necessary to generate large amounts of CdeC protein during sporulation by acting in concert with σ^K to increase cdeC transcription (Fig. 4 and Fig. S8) and CdeC levels through a post-transcriptional mechanism (Figs 6 and 7). As the former observation predicts that SpoIIID directly binds to the cdeC promoter region, it will be interesting to determine whether SpoIIID directly activates the transcription of other σ^K-regulated genes and to define its binding site. As cdeC encodes an exosporium protein in contrast with the other σ^K-regulated genes that we tested, we measured transcript levels of an additional two exosporium genes, bclA1 and bclA3 (Escobar-Cortes et al., 2013; Pizarro-Guajardo et al., 2014). No difference in their transcription between the sigK and spoIIID mutants was observed (Fig. S13).

Notably, the reduced levels of CdeC in spoIIID⁻/P_E-sigK spores did not strongly affect heat-resistant spore formation (Fig. 5 and Table 2). It nevertheless remains possible that this reduction could decrease the resistance of these mutant spores to ethanol or other physical insults given that loss of CdeC in the R20291 background increases spore susceptibility to lysozyme and ethanol presumably due to loss of the exosporium layer (Barra-Carrasco et al., 2013).

We also observed that SpoIIID enhances σ^K levels in sporulating cells heterologously expressing a rearranged sigK copy from a σ^E-dependent promoter (P_E-sigK) as σ^K levels were markedly higher in the sigK⁻ relative to the spoIIID⁻ background (Fig. 6 and Fig. S7), even though no significant differences in sigK transcript levels were detected (Fig. S8). While the mechanism by which SpoIIID modulates σ^K or CdeC levels is unclear, it seems possible that SpoIIID may cooperate with σ^K to increase the expression of an unknown gene(s) whose product(s) increase the levels of σ^K, CdeC and potentially other σ^K-regulated gene products.

Our study also identifies a specific role for SpoIIID and σ^K in regulating spore assembly in C. difficile. TEM analyses failed to detect either coat polymerization or localization around the forespore in both sigK and spoIIID mutants (Fig. 2), even though previous work has shown that the σ^E-regulated gene product and surface-exposed spore protein, CotB (Permpoonpattana et al., 2011; 2013; Pereira et al., 2013), localizes around the forespore of a sigK mutant (Pereira et al., 2013). Thus, it was somewhat surprising that forespore-localized CotB failed to be incorporated into purified sigK⁻ and spoIIID⁻ spores (Figs 6 and 7). This finding suggests that additional σ^K-regulated gene product(s) control tethering of part of the coat to spores. Importantly, coat localization to the C. difficile forespore does not necessarily imply retention to the spore surface, highlighting the importance of analyzing coat protein incorporation into purified spores.

Interestingly, the morphology of the sigK⁻ and spoIIID⁻ spores (Fig. 2) resembled in part the phenotype of B. subtilis spoVID⁻ spores, which produce cortex but fail to retain polymerized coat on their surface (Beall et al., 1993). However, in contrast with a B. subtilis spoVID⁻ strain, no polymerized coat is visible in C. difficile sigK and spoIIID mutants (Fig. 2). This observation implies that SpoIIID and σ^K control the production of proteins required for coat protein polymerization (defined as the formation of multiple coat laminations as detected by TEM).

Collectively, our results support a model in which SpoIIID is needed to activate sigK transcription, which leads to the production of intrinsically active σ^K without the need for post-translational activation. Because of this simplified regulatory architecture (relative to B. subtilis), C. difficile spore formation is largely unaffected when the order of sigK expression and σ^K production is altered. Nevertheless, σ^K plays a key role in controlling spore maturation by activating the transcription of genes whose products appear to be required for polymerizing coat proteins into multiple layers, adhering a coat protein to the outer forespore membrane and processing coat proteins during spore formation. Identifying the σ^K-regulated gene products that mediate these processes will yield important insights into C. difficile spore formation and the development of spore resistance, and may identify new targets for therapeutic intervention.

Experimental procedures

Bacterial strains and growth conditions

All C. difficile strains are listed in Table S11 and were derived from the parent strain JIR8094 (Dineen et al., 2007), an erythromycin-sensitive derivative of the sequenced clinical isolate 630 (Sebaihia et al., 2006). C. difficile strains were grown on solid or broth BHIS (brain heart infusion broth supplemented with yeast extract; Sorg and Dineen, 2009). The BHIS media was supplemented as indicated with taurocholate (TA; 0.1% w/v), thiamphenicol (5–10 µg ml^–1), kanamycin (50 µg ml^–1), cefoxitin (16µgml^–1), FeSO4 (50 µM) and/or erythromycin (10 µg ml^–1). Brain heart infusion broth was supplemented with anhydrotetracycline (ATc; 5–200 ng ml^–1) for induction of the P_tet promoterin C. difficile inducible expression constructs (Fagan and Fairweather, 2011). C. difficile was grown at 37°C under anaerobic conditions, using a gas mixture containing 85% N₂, 5% CO₂ and 10% H₂.

The Escherichia coli strains HB101/pRK24 and BL21(DE3) were used for conjugations and protein expression respectively. E. coli strains were grown in Luria-Bertani broth (LB) at 37°C and shaking at 225 r.p.m. Medium was supplemented when appropriate with chloramphenicol (20 µg ml^–1), ampicillin (50–100 µg ml^–1), carbenicillin (50 µg ml^–1) or kanamycin (30 µg ml^–1) as indicated.

E. coli strain construction

All E. coli strains and plasmids are listed in Table S11, and all primers are listed in Table S12. For disruption of spoIIID, a modified plasmid containing the retargeting group II intron, pCE245 (a gift from C. Ellermeier, University of Iowa), was used as the template. Primers #656, 657, 658 and 532, the EBS Universal primer as specified by the manufacturer (Sigma Aldrich), carrying targeting regions for disrupting spoIIID at base pair 201, were used to amplify the targeting sequence from the template. The resulting retargeting sequence was digested with BsrGI and HindIII and cloned into the conjugation vector pJS107 (a gift from J. Sorg, University of Texas A&M and a derivative of pJIR750ai, Sigma Aldrich) (Francis et al., 2013). The resulting ligation was transformed into DH5a and confirmed by sequencing. The resulting plasmid was transformed into HB101/pRK24.

To construct the spoIIID complementation construct, PCR splicing by overlap extension (SOE) was used to fuse the σ^E-regulated promoter of sipL to spoIIID (P_E). Primer pair #700 and 941 was used to amplify the 5’ promoter region SOE product, and primer pair #940 and 939 was used to amplify the 3’ gene SOE product. The resulting fragments were combined, and the flanking primer pair #700 and 939 was used to amplify the P_sipL-spoIIID product including the TAA stop codon. The resulting PCR product was digested with NotI and XhoI, ligated into pMTL83151 (Heap et al ., 2009) digested with the same enzymes and transformed into DH5α. The resulting pMTL83151-P_E-spoIIID plasmid was transformed into HB101/ pRK24. This strategy was employed because a gene fragment consisting of spoIIID under the control of its native promoter appeared to be toxic when expressed from pMTL83151, which is a high copy plasmid in E. coli.

To construct a strain for producing recombinant CD3580-His₆ for antibody production, primer pair #498 and 499 was used to amplify the CD3580 gene lacking the stop codon from 630 genomic DNA. The resulting PCR product was digested with NcoI and XhoI, ligated to pET28a digested with the same enzymes and used to transform DH5α. The resulting pET28a-CD3580 expression plasmid was used to transform BL21(DE3) for protein expression.

PCR SOE was also used to construct sigK complementation constructs in which sigK lacking the skin element was expressed from the SpoIIID-independent σ^E-regulated promoter of sipL (P_E). Primer pair #700 and 894 was used to amplify the 5’ SOE product, while primer pair #893 and 737 was used to amplify the 3’ SOE product. The resulting fragments were combined, and the flanking primer pair #700 and 737 was used to amplify the P_EsigK construct including the TAAstop codon. The resulting PCR product was digested with NotI and XhoI, ligated into either pMTL83151 or pMTL84151 (which has a higher copy number than pMTL83151 in C. difficile; Heap et al., 2009), digested with NotI/XhoI and transformed into DH5α. The resulting pMTL83151-P_E-sigK and pMTL84151-P_E-sigK plasmids were individually transformed into HB101/pRK24.

To construct the inducible sigK strain, PCR was used to amplify the sigK gene with its native ribosome binding sites, using primer pair #1626 and 1620. The resulting product was digested with SacI and BamHI, ligated to pRPF185 (Fagan and Fairweather, 2011) (a gift from R. Fagan, University of Sheffield), digested with the same enzymes and used to transform DH5α. The resulting pRPF185-P_Tet-sigK plasmid was transformed into HB101/pRK24.

C. difficile strain construction

A spoIIID mutant in JIR8094 was constructed using TargeTron-based gene disruption as described previously (Fig. S2; Heap et al., 2007; Putnam et al., 2013). The pJS107/ spoIIID TargeTron construct in E. coli was conjugated into C. difficile JIR8094 using the E. coli HB101/pRK24 donor strain #798 as previously described (Putnam et al., 2013). Briefly, thiamphenicol resistant C. difficile clones were selected and regrown on BHIS plates containing thiamphenicol and FeSO₄ to induce expression of the Targetron system. Erythromycin-resistant clones were selected and then isolation streaked on BHIS plates supplemented with erythromycin (10 µg ml⁻¹). Positive clones were screened by colony PCR for a 2-kb insertion in spoIIID using primer pair #698 and 699 (Fig. S2). Three independent clones were phenotypically characterized.

C. difficile plasmid complementation strains were constructed using conjugation as previously described (Putnam et al.,2013). Briefly, HB101/pRK24 donor strains carrying the appropriate complementation constructs were mixed with C. difficile recipient strains. C. difficile strains containing the complementation plasmids were selected on BHIS agar supplemented with thiamphenicol (10 µg ml⁻¹), kanamycin (50 µg ml⁻¹) and cefoxitin (10 µg ml⁻¹). At least two independent clones from each complementation strain were phenotypically characterized.

Sporulation assay

C. difficile strains were grown from glycerol stocks on BHIS-TA or with both TA and thiamphenicol (5 µg ml⁻¹) for strains with pMTL83151-derived or pMTL84151-derived plasmids. Colonies arising on these plates were then used to inoculate 70:30 agar plates (containing 5 µg ml⁻¹ thiamphenicol as appropriate) for 18 to 24 hours as previously described (Putnam et al., 2013). Sporulation-induced lawns were harvested in PBS, washed once, resuspended in PBS and used in downstream analyses.

Spore purification

Sporulation was induced by growing C. difficile strains on 70:30 plates (with thiamphenicol when appropriate) for 7 days, and spores were harvested in ice-cold sterile water as previously described (Sorg and Dineen, 2009; Adams et al, 2012). Briefly, spore suspensions were washed multiple times in ice-cold sterile water and treated twice with DNase (New England Biolabs) at 37°C for 30 minutes. Spore suspensions were resuspended in ice-cold sterile water and incubated at 4°C overnight. Spores were enriched using a HistoDenz (Sigma Aldrich) gradient, evaluated for purity by phase contrast microscopy, and the optical density of the suspension was measured.

Heat resistance assay

Sporulation was induced by growing C. difficile strains on 70:30 plates (containing 5 µg ml⁻¹ thiamphenicol as appropriate as previously described (Fimlaid et al., 2013). After 24 hours of growth, cells were harvested into 1.0 ml PBS, and the samples were divided into two tubes. One tube was heat-treated at 60°C for 30 minutes. Both the untreated and heat-treated samples were serially diluted, and the dilutions were plated onto pre-reduced BHIS-TA. After 22 hours, colonies were counted and total cell counts were determined. For each strain tested, sporulation efficiency was determined by calculating the ratio of heat-resistant spores produced by a given strain (i.e. heat-resistant spores were normalized to total cells) relative to ratio of heat-resistant spores produced by wild type (heat-resistant wild-type spores normalized to total wild-type cells) for five biological replicates.

Spore germination assay

Approximately 2 × 10⁶ purified spores were divided into two tubes, one of which was heat-treated at 60°C for 30 minutes, as heat-activation of spores can increase germination efficiency in Bacillus sp. (Hyatt and Levinson, 1968; Setlow, 2014), although it should be noted that this has not been observed in C. difficile (Dembek et al., 2013). Both the untreated and heat-treated samples were serially diluted, and dilutions were plated onto pre-reduced BHIS-TA. After 22 hours, colonies arising from germinated spores were counted. Germination efficiency was determined by calculating the ratio of germinated spores for each strain relative to wild type.

Fluorescence and light microscopy

C. difficile strains were grown for 18 hours on 70:30 media, harvested into 1 ml of PBS containing 1 µg ml^–1 FM4–64 (Molecular Probes) and 15 µg ml⁻¹ Hoechst 33342 (Molecular Probes) and visualized on agarose pads as previously described (Fimlaid et al., 2013) except with the following modifications. DIC and fluorescence microscopy were performed using a Nikon PlanApo Vc 100 × oil immersion objective (1.4 NA) on a Nikon Eclipse Ti2000 epifluorescence microscope. Five fields for each sample were acquired with an EXi Blue Mono camera (QImaging) with 2×2 binning, hardware gain setting of 2.6, and driven by NIS-Elements software (Nikon). Images were subsequently imported into Adobe Photoshop CS6 for pseudocoloring and minimal adjustments in brightness/contrast levels. Phase-contrast microscopy for imaging the samples used for sporulation assays was performed as previously described (Putnam et al., 2013).

Electron microscopy

One hundred microliters of bacterial cell suspension samples from sporulation assays and purified spores were prepared as previously described (Putnam et al., 2013).

RNA processing

RNA for RNA-Seq was extracted from WT, sigE⁻, sigK⁻ and spoIIID⁻ C. difficile after 18 hours of growth on 70:30 sporulation media as previously described (Fimlaid et al ., 2013). Briefly, RNA was extracted using a FastRNA Pro Blue Kit (MP Biomedical) and a FastPrep-24 automated homogenizer (MP Biomedical). Contaminating genomic DNA was depleted using three successive DNase treatments, and samples were tested for genomic DNA contamination using quantitative PCR for 16S rRNA and the sleC gene. DNase-treated RNA (5 µg) was mRNA enriched using a Ribo-Zero Magnetic Kit (Epicentre), and the quality of total RNA was validated using an Agilent 2100 Bioanalyzer.

RNA isolated for qRT-PCR was processed identically except that mRNA enrichment was done using an Ambion MICROBExpress Bacterial mRNA Enrichment Kit (Invitrogen). Reverse transcription of enriched RNA was done using the SuperScript® First Strand cDNA Synthesis Kit (Invitrogen) with random hexamer primers.

RNA-Seq library construction and sequencing

Enriched mRNA (100 ng) was submitted to Illumina (Hayward, CA) for massively parallel sequencing on a MiSeq. cDNA synthesis and library preparation was carried out using the TruSeq stranded mRNA Sample Prep kit (without polyA selection and using fragmentation), according to manufacturer’s instructions. Paired end sequencing of samples was performed using a total of 10 pM of library in each flow cell lane. The samples were indexed and pooled in equal amounts to generate equal read coverage.

RNA-Seq analysis

Whole genome transcriptional profiling analyses were performed as previously described (Fimlaid et al., 2013). Genome coverage and sequencing counts for each strain and replicate are provided in Table S2. Differentially expressed genes were identified based on a minimum ≥ 4-fold-change (higher in the reference sample than the query) and maximum p-value ≤ 0.05. Tables showing genes with differential expression relative to wild type during sporulation are provided in Tables S3–S6. Genes differentially regulated between the spoIIID and sigK mutants are listed in Table S7. The RNA-Seq data set has been uploaded to GEO under accession number GSE63777.

Antibody production

The anti-CD3580 antibody used in this study was raised in rabbits by Cocalico Biologicals (Reamstown, PA). His₆-tagged CD3580 antigen was purified on Ni²⁺-affinity resin as previously described (Adams et al., 2012) from E. coli strain #547. It should be noted that the anti-CotE antibody used in the current study was based on the same expression construct described in a previous study (Fimlaid et al., 2013) with the exception that the recombinant C-terminal chitinase domain (Permpoonpattana et al., 2013) was purified using gel filtration chromatography on a Superdex 200 10/300 GL as previously described (Adams et al., 2012) and submitted for antibody production to Cocalico Biologicals (Reamstown, PA).

Western blot analyses

C. difficile cell pellets were processed as previously described (Fimlaid et al., 2013; Putnam et al., 2013). Briefly, pellets were freeze-thawed three times, diluted in EBB buffer [8 M urea, 2 Mthiourea, 4% (w/v) SDS, 2% (v/v) β-mercaptoethanol] and incubated at 95°C for 20 minutes. Samples were centrifuged for 5 minutes at 15,000 r.p.m., resuspended, and 4 × sample buffer [40% (v/v) glycerol, 1 M Tris pH 6.8, 20% (v/v) β-mercaptoethanol, 8% (w/v) SDS, and 0.04% (w/v) bromo-phenol blue] was added. Samples were then heated at 95°C for 15 minutes, pelleted, then reheated at 95°C for 5 minutes and re-pelleted. For the Westerns in Fig. S5, ~ 10 µg of protein per lane was loaded. For the Westerns in Fig. 6 and Figs. S10, ~ 20 µg of protein per lane was loaded. For the Westerns in Fig. 7 and Fig S10 and S11, ~ 2 µg of protein per lane was loaded. The protein quantification was done in triplicate using the Pierce 660 nm Protein Assay (Thermo Scientific). Protein samples were resolved by SDS-PAGE and transferred to Millipore Immobilon-FL membrane. The membranes were blocked in Odyssey Blocking Buffer and analyzed with the rabbit polyclonal antibodies, anti-σ^K, anti-σ^E, anti-CotB, anti-CdeC (Fimlaid et al., 2013), anti-SpoIVA (Putnam et al., 2013), anti-CD3580, and anti-CotE antibodies were used at a 1:2,000 dilution and anti-SleC (Adams et al., 2012) and the mouse monoclonal antibody anti-Spo0A (Fimlaid et al., 2013) at a 1:10,000 dilution. IRDye 680CW and 800CW infrared dye-conjugated secondary antibodies were used at 1:20,000 dilutions. The Odyssey LiCor CLx was used to detect secondary antibody fluorescent emissions for Western blots.

Quantitative RT-PCR

For the RNA-Seq validation, transcript levels of sleC, cotE, spoIVA, cdeC (cd1067) and rpoB (housekeeping gene) were determined from cDNA templates prepared from five biological replicates of WT, spo0A, sigE, sigK and spoIIID in technical duplicates as previously described (Fimlaid et al., 2013). Transcript levels of CD3580, sigK, spoIIIAA, spoIIIAG and spoIIID, were analyzed using gene-specific primer pairs #794 and 795; #1405 and 1406; #1511 and 1512; #1297 and 1298; and #698 and 699 respectively. Quantitative real-time PCR was performed using SYBR Green JumpStart Taq Ready Mix (Sigma), 50 nM of gene specific primers and an ABI PRISM 7900HT Sequence Detection System (Applied Biosystems). Transcript levels were normalized to the housekeeping gene rpoB using the standard curve method and calculated relative to the spo0A⁻ strain for the RNA-Seq validation studies (Fig. 4) and relative to wild type carrying empty pMTL83151 vector for all other analyses.

For transcriptional analyses of strains expressing spoIIID and sigK complementation constructs, expression levels of the same genes were examined on cDNA templates prepared from four biological replicates as described above. Transcript levels were normalized to the housekeeping gene rpoB using the standard curve method and calculated relative to the wild-type strain carrying empty pMTL83151 vector.

Inducible expression of sigK

C. difficile JIR8094 or sigE⁻ containing pRPF185-sigK were grown from glycerol stocks onto BHIS-TA plates supplemented with thiamphenicol. After 24 hours of growth the strains were inoculated into 1.5 ml pre-reduced BHIS broth supplemented with thiamphenicol (5 µg ml⁻¹). When broth cultures reached late-log phase, they were diluted 1:10 into 3 ml of pre-reduced BHIS supplemented with thiamphenicol (5 µg ml⁻¹) and grown to mid-log phase. sigK expression was induced in the cultures by adding anhydrotetracycline (diluted in 100% ethanol) to the indicated final concentrations. After 2 hours of growth, the OD600 was measured, and the cultures were pelleted, normalized for cell density, resuspended in EBB buffer and processed for Western blot analyses.