Human Hypervirulent Clostridium difficile Strains Exhibit Increased Sporulation as Well as Robust Toxin Production

Abstract

Toxigenic Clostridium difficile strains produce two toxins (TcdA and TcdB) during the stationary phase of growth and are the leading cause of antibiotic-associated diarrhea. C. difficile isolates of the molecular type NAP1/027/BI have been associated with severe disease and hospital outbreaks worldwide. It has been suggested that these “hypervirulent” strains produce larger amounts of toxin and that a mutation in a putative negative regulator (TcdC) allows toxin production at all growth phases. To rigorously explore this possibility, we conducted a quantitative examination of the toxin production of multiple hypervirulent and nonhypervirulent C. difficile strains. Toxin gene (tcdA and tcdB) and toxin gene regulator (tcdR and tcdC) expression was also monitored. To obtain additional correlates for the hypervirulence phenotype, sporulation kinetics and efficiency were measured. In the exponential phase, low basal levels of tcdA, tcdB, and tcdR expression were evident in both hypervirulent and nonhypervirulent strains, but contrary to previous assumptions, toxin levels were below the detectable thresholds. While hypervirulent strains displayed robust toxin production during the stationary phase of growth, the amounts were not significantly different from those of the nonhypervirulent strains tested; further, total toxin amounts were directly proportional to tcdA, tcdB, and tcdR gene expression. Interestingly, tcdC expression did not diminish in stationary phase, suggesting that TcdC may have a modulatory rather than a strictly repressive role. Comparative genomic analyses of the closely related nonhypervirulent strains VPI 10463 (the highest toxin producer) and 630 (the lowest toxin producer) revealed polymorphisms in the tcdR ribosome binding site and the tcdR-tcdB intergenic region, suggesting that a mechanistic basis for increased toxin production in VPI 10463 could be increased TcdR translation and read-through transcription of the tcdA and tcdB genes. Hypervirulent isolates produced significantly more spores, and did so earlier, than all other isolates. Increased sporulation, potentially in synergy with robust toxin production, may therefore contribute to the widespread disease now associated with hypervirulent C. difficile strains.

Clostridium difficile is a leading bacterial nosocomial pathogen. Antibiotic treatment alters and suppresses commensal microbiota, allowing ingested C. difficile spores to germinate and colonize the gut. If the infecting strain is of the toxin-producing lineage of C. difficile (toxigenic), the resulting infection (CDI) can range from mild diarrhea to potentially fatal pseudomembranous colitis. Since 2000, highly virulent variants of toxigenic C. difficile have caused epidemics of CDI characterized by greater incidence, severity, and fatality (12, 25, 29). These “hypervirulent” (HV) strains cluster into a distinct phylogenetic group (38), as assessed by several different molecular methods (21), including multilocus variable-number tandem repeat analyses (MLVA) (27) and microarray analyses (38). Other methodologies used to type C. difficile isolates have classified human HV strains as group BI (restriction endonuclease analyses [REA] [29]), type NAP1 (pulsed-field gel electrophoresis [21]), ribotype 027 (PCR ribotyping [41]), and toxinotype III (toxin gene polymorphism typing [33, 41]). BI/NAP1/027 strains have spread rapidly and widely in the past 10 years and have been responsible for CDI epidemics worldwide (5, 7, 16, 20, 23, 29).

In response to nutrient limitation in the stationary phase of growth, toxigenic C. difficile strains produce the glucosyltransferase toxins TcdA and TcdB (17). A 19.6-kb chromosomal region, called the pathogenicity locus (PaLoc), includes the toxin genes, a putative holin (TcdE), and two regulators, TcdC (negative regulator) and TcdR (activator/sigma factor), of toxin gene expression (6). While the mechanism(s) responsible for the increased virulence of HV strains are not well understood, many studies have explored the contribution of the toxins to this phenotype. The relative amounts of the two toxins produced and their potential impacts on disease severity are subjects of debate (2, 30). Akerlund et al. found that one HV strain produced 3- to 13-fold more of the toxins than a group of non-HV strains (2). On the other hand, using a human gut model, Freeman et al. reported that mean toxin titers were not significantly higher for an HV strain than for a non-HV strain, although the HV strain had an extended duration of toxin production (13). A study by Warny et al. found that the median amounts of toxins A and B in a group of HV strains were 16- and 23-fold higher, respectively, than the median toxin amounts for a group of non-HV strains and concluded that HV strains also expressed toxins during exponential growth, although no quantitative toxin measurements were presented for the 0- to 24-h time period (41). In this study, we sought to clarify if HV strains did indeed produce more TcdA/B toxins and to determine whether any other growth-related parameters, such as sporulation, were impacted in HV strains.

The absence of functional TcdC, a negative regulator of toxin gene expression, has been proposed as an explanation for the apparently dysregulated toxin production in HV strains (36). Postulated to act as a novel anti-sigma factor, TcdC has been shown to destabilize the RNA polymerase holoenzyme containing the TcdR sigma factor (28). One study also suggested that, unlike tcdA and tcdB, tcdC is expressed only during exponential growth (17). A frameshift mutation in tcdC in many (but not all) C. difficile strains of the HV clade is predicted to result in a truncated, nonfunctional molecule (8, 28, 41). While the tcdC variation has been well described in HV strains, its direct contribution to hypervirulence is unknown. Whether the lack of this negative regulator is sufficient to account for toxin production during exponential growth, as proposed by Warny et al. (41), has not been specifically tested.

The most likely form of C. difficile in the health care environment is the highly resistant bacterial spore, which is spread to susceptible patients by either environmental contact or carriage by health care personnel (3, 10). Since HV C. difficile strains may also exhibit increased sporulation, environmental contamination burdens leading to increased rates of transmission are also a concern.

To begin elucidating the pathogenesis of HV C. difficile, we performed controlled studies to determine the course of toxin production and sporulation over the growth cycle in multiple HV C. difficile clinical isolates. We compared HV strains to known high- and low-toxin-producing non-HV strains, including those that have been associated with previous outbreaks (35, 36) and whose genetic relationships to HV strains have been estimated (27, 38). We also performed comparative genomic analyses of phylogenetically related high- and low-toxin-producing strains and suggest a possible mechanistic basis for differences in toxin production. Taken together, our studies revealed that increased toxin production is not the definitive determinant of HV strains and that current PaLoc regulation models may need additional refinement.

MATERIALS AND METHODS

C. difficile human clinical isolates were obtained from the culture collection of D. N. Gerding (Table 1). Four HV C. difficile strains isolated from geographically distinct regions and four genetically distinct toxigenic but non-HV strains were chosen. The non-HV strains included strains 630 (a known low toxin producer) and VPI 10463 (a known very high-level toxin producer) (1), which are both very rarely found in clinical settings. Both strains gave been sequenced and have been found to be closely related members of the same clade in phylogenetic analyses (38). The other two non-HV strains, J9 and K14 (representative of the REA type J and K groups, respectively) have caused hospital outbreaks and are frequently isolated from hospital settings in the United States but have never been reported as a cause of increased CDI severity and are not referred to as hypervirulent (27, 35, 36, 38). For all assays, all C. difficile strains were grown to saturation in brain heart infusion (BHI) broth (BD Biosciences, Boston, MA). Culture aliquots (1 ml) were clarified by centrifugation at 2,000 × g. Bacterial pellets were washed in phosphate-buffered saline (PBS), resuspended in fresh BHI broth at a ratio of 1:50, and allowed to grow without agitation under anaerobic conditions (85% N₂, 5% H₂, and 10% CO₂) in a Coy glove box (Coy, Grasslake, MI). Identical inocula were used for all strains and all growth experiments. Optical density readings (600-nm wavelength) (OD₆₀₀) were taken at intervals of 1 h or less for the first 18 h and again at 24 and 48 h. For total toxin quantitation, culture samples were clarified and supernatant fluids were sterile filtered and frozen at −80°C prior to use. All growth experiments were performed in their entirety at least three times, and a representative example is shown (Fig. 1A). For toxin quantitation, culture supernatant fluids were collected at mid-exponential phase (OD₆₀₀ = 0.5); at early stationary phase (defined by two consecutive nonincreasing OD₆₀₀ readings; approximately 11 h); at mid-stationary phase (15 h); and at 18, 24, and 48 h of growth.

TABLE 1.

Strain designations, origins, and characterizations

Strain designationa	Isolate no.	Yr	Source	Toxigenicb	Hypervirulentc
R23 (630)	6396	1982	Switzerland	Yes	No
Z3 (VPI 10463)	5707	1980	Eastern United States	Yes	No
J9	4478	1987	IL	Yes	No
K14	5780	1994	IL	Yes	No
BI-6	6336	2003	OR	Yes	Yes
BI-8	6477	2004	ME	Yes	Yes
BI-17	6443	2004	Montreal, Canada	Yes	Yes
BI-23	6654	2007	Eastern United States	Yes	Yes

^aStrain designation by REA type. The more common strain designations for strains R23 and Z3, as reported in the literature, appear in parentheses.

^bToxigenic status as determined by in vitro cytotoxicity, as well as toxin A immunoassays of culture supernatants.

^cHypervirulence as defined by association with recent hospital outbreaks of severe clinical disease.

FIG. 1.

(A) Growth curves of the eight strains used in this study. Absorbance at 600-nm wavelength was measured over 72 h of growth. (B) Total toxin levels over time. The toxin levels represent combined TcdA and TcdB levels by ELISA (see Materials and Methods). Exponential-phase samples were taken when each individual culture reached an OD₆₀₀ of 0.5. Means and standard errors from three biological replicates are shown. The sensitivity of the ELISA was approximately 0.8 ng/ml for TcdA and 2.5 ng/ml for TcdB.

Toxin amounts were quantified using the Wampole Tox A/B II kit (TechLabs, Inc., Blacksburg, VA). Purified toxin B (provided by TechLabs, Inc., Blacksburg, VA) was used to construct a standard curve. Samples of the supernatant fluids were diluted as needed in order to obtain readings within the linear range of the standard curve. All samples were tested in triplicate, and each experiment was performed in its entirety at least three times.

For toxin gene expression assays, total RNA was isolated from exponential-phase (OD₆₀₀ = 0.5) and stationary phase (12-h) cultures. Five milliliters of culture was harvested by centrifugation at 2,800 × g for 10 min at 4°C. The bacterial pellets were processed immediately or resuspended in lysis buffer, flash frozen in liquid nitrogen, and stored at −80°C. Lysis and extraction of RNA were performed using the Ambion RiboPure Bacteria RNA kit (Ambion, Austin, TX) according to the manufacturer's instructions, followed by DNase I digestion using Ambion Turbo DNase I. DNase I digestion was repeated twice (three times in total) for all samples. The RNA obtained after each DNase I digestion was purified using the Qiagen RNeasy RNA column purification kit, according to the manufacturer's instructions (Qiagen, Germantown, MD); quantified using a NanoDrop spectrophotometer; and stored at −20^oC. The RNA quality was assessed spectrophotometrically (260/280 nm) and by visualization on denaturing formamide/formaldehyde gels (not shown). Five hundred nanograms of pure RNA from each sample was converted to cDNA using random hexamers and the Bio-Rad iScript cDNA synthesis kit (Bio-Rad, Hercules, CA).

The primers used to amplify PaLoc genes were either synthesized using previously published sequences (for tcdA, the non-HV strain tcdB, and tcdC) (26) or specifically designed for the HV strain tcdB (tcdB4BIup, 5′-AGCTGCTTCAGTCGGAGAAA-3′; tcdB4BIdn, 5′-AATCAATTGCTTCCCCCTCT-3′), as well as for tcdR (tcdR3up, 5′-ATCAAAGTAAGTCTGTTTTTGAGGAAG-3′; tcdR3dn, 5′-TGCTCTATTTTTAGCCTTATTAACAGC-3′), and rpoA (rpoA2up, 5′-TCATTACCAGGTGTAGCAGTGAA-3′; rpoA2dn, 5′-GAGCATGGTCCTTGAGCTTC-3′).

All primers were tested in genomic-DNA amplification for all test strains to confirm the specificity and amplicon size before being used in quantitative reverse transcriptase (qRT) PCRs. To test for DNA contamination, RT-PCRs were performed for all samples using RNA alone and were found to be negative (not shown). The expression levels of all PaLoc genes tested in each isolate were normalized using the C. difficile housekeeping gene rpoA as a reference (32). Transcription rates for individual PaLoc genes relative to each other among different isolates were not determined, since the efficiency of reverse transcription varies for different genes, making between-gene transcription rate comparisons inaccurate.

A gradient of C. difficile genomic DNA was used to determine the efficiency of amplification of each primer set. Efficiencies ranged from 90 to 101%, and R² values were at least 98%. For all samples tested, 25-μl reactions were performed in triplicate using 5Prime Real Master mix SYBR Rox (Fisher, Pittsburgh, PA), 1 μl of cDNA, and 250 nM primers in an Eppendorf MasterCycler equipped for qRT-PCR detection (32). The specificity of the product was determined by dissociation (melt) curve analysis. PaLoc gene expression was analyzed relative to that of the reference gene using the ΔC_T method, according to the following formula: 2^{CTRef − CTTest} (24).

To identify genetic differences in PaLoc and non-PaLoc DNAs that might contribute to variations in toxin gene expression, we compared the entire ∼19,600-bp PaLoc sequences, as well as non-PaLoc regulator gene sequences, of the very low-toxin-producing strain 630 and the very high-toxin-producing strain VPI 10463. The two strains have previously been shown to be closely related (in the same phylogenetic clade [38]). To further characterize this relatedness of the two strains, we used the Rapid Annotations using Subsystem Technology (RAST) server (4) to annotate the recently released VPI 10463 genome sequence and to generate a predicted proteome (GenBank accession no. NZ_ABKJ00000000.2). We then compared the whole genomes, as well as proteomes, of 630 and VPI 10463. All sequences were aligned using the ClustalW algorithm (39) in the BioEdit program (15). Predicted and empirically determined promoter sequences were determined based on the work of Hundsberger et al. (17). Transcriptional terminators were predicted using the algorithm of Kingsford et al. (22), and ribosomal binding sequences were annotated manually based on an Escherichia coli consensus sequence. The thermodynamic stability of RNA secondary structures was estimated using free-energy calculations from the RNAfold Web server (14).

To assess the accumulation of spores over the growth cycle, we employed a differential spore-plating and microscopy method. At 8, 20, 28, and 48 h postinoculation into BHI broth as described above, 1-ml samples of each strain were clarified by centrifugation, and the pellets were washed in PBS, heat shocked at 65°C for 15 min (to kill vegetative cells), serially diluted, and plated on 1% taurocholate (a spore germinant) fructose agar (TFA) plates to enumerate spores (34). The plates were incubated anaerobically for 48 h, after which CFU were enumerated, and the data were represented as total spores per milliliter of starting culture. For microscopy, 10-μl samples of 48-hour cultures were applied to microscope slides, oven dried at 85°C for 5 min, and Gram stained. Spores were defined as all fully formed refractile bodies, whether free or attached to mother cell material. For each strain, 10 distinct fields were photographed and enumerated at ×100 magnification under oil immersion. Spores in all fields were counted, and data were represented as the mean number of spores per field and the percentage of spores obtained per total maximum vegetative cell count at the end of exponential phase (3.0 × 10⁸ CFU/ml at an OD₆₀₀ of 1.03).

For statistical analyses, the StatView5 software package (SAS, San Francisco, CA) was employed. Analysis of variance (ANOVA) tests were used to determine significance, and the protected least significant difference test was used for posthoc comparisons.

RESULTS

HV strains showed no statistically significant growth defects or advantages over non-HV strains (Fig. 1A). Some strains exhibited a steeper decline in optical density readings in late stationary phase, but this finding was not consistent or specific to HV or non-HV strains.

In exponential phase, toxin levels were below the enzyme-linked immunosorbent assay (ELISA) detection threshold for all strains tested (Fig. 1B). The sensitivity of the ELISA was approximately 0.8 ng/ml for toxin A and 2.5 ng/ml for toxin B. The non-HV strain VPI 10463 produced the largest amounts of toxin (∼500 ng/ml total toxin at 48 h). Toxins began to accumulate in culture supernatant fluids at between 12 and 15 h in VPI 10463, as well as in three of the HV strains. For all strains tested, the majority of toxin accumulated between 24 and 48 h. When averaged, there were no statistically significant differences between the group of four HV strains and the four non-HV strains (P ≥ 0.72). At 24 and 48 h, VPI 10463 exhibited significantly higher toxin production than all other strains. At 24 h, BI8 production was significantly higher than that of strains 630, J9, and BI23. Interestingly, while strain BI23 production was significantly lower than that of BI17, BI8, and VPI 10463 at 24 h, at 48 h it was significantly higher than that of all strains, other than VPI 10463.

The ELISA used in this study measured total production of TcdA and TcdB proteins. However, toxin gene expression may be differentially regulated, and we assessed this by quantitative determination of tcdA and tcdB gene expression. Transcriptional analyses using quantitative real-time PCR were performed on all eight strains tested as described above for both tcdA and tcdB, as well as for the regulatory-factor-encoding genes tcdC and tcdR, at both exponential and early stationary growth phases. Based on the ELISA data, we expected that the expression of tcdA and tcdB in all strains would be low/undetectable in the exponential phase of growth and high in stationary phase, and this was indeed the case. The levels of rpoA reference gene expression were comparable across all strains in the same growth phase. Overall levels of rpoA transcription were higher in exponential phase than in stationary phase, as would be expected under rapid growth conditions.

For tcdA, and consistent with the expectation described above, low levels of expression were observed during exponential growth (Fig. 2A). Levels of detectable tcdA transcripts were, on average, 100-fold lower than those of rpoA (the reference gene) and were not significantly different between HV and non-HV strains as a group. In this sensitive assay, we also determined that strain 630 was significantly lower in tcdA gene expression than strains VPI 10463, K14, BI17, and BI23 (P ≤ 0.004), while as noted earlier, toxins were not detectable in exponential-phase cultures.

FIG. 2.

(A) Transcription of tcdA during exponential phase relative to transcription of the housekeeping gene rpoA. Means and standard errors of 3 replicates are shown. (B) Transcription of tcdA during stationary phase relative to transcription of the housekeeping gene rpoA. Means and standard errors of 3 biological replicates are shown.

In contrast, high relative expression of tcdA was seen in stationary-phase samples (Fig. 2B). Strain VPI 10463 showed significantly higher tcdA expression than all other strains (P ≤ 0.05), which was consistent with the ELISA results for toxin production. The average tcdA expression of the HV strains was not significantly different from that of the non-HV strains.

Similar patterns of expression were observed for tcdB in exponential phase (Fig. 3A). There was a small but statistically significant (P ≤ 0.02) 3-fold-higher expression in the HV strains versus the non-HV strains in exponential phase. During stationary phase, strain VPI 10463 again exhibited the highest tcdB expression level of all strains tested (Fig. 3B), consistent with the toxin levels determined by ELISA.

FIG. 3.

(A) Transcription of tcdB during exponential phase relative to transcription of the housekeeping gene rpoA. Means and standard errors of 3 replicates are shown. (B) Transcription of tcdB during stationary phase relative to transcription of the housekeeping gene rpoA. Means and standard errors of 3 biological replicates are shown.

TcdR is the sigma factor that directs toxin gene expression. If toxin production were indeed dysregulated in HV strains and occurred during exponential growth, then the expression of tcdR should be expected to be dysregulated as well, since this protein is a required positive activator of toxin gene expression. However, and consistent with the above results, relative expression levels of tcdR were low and not significantly different between HV and non-HV strains during exponential phase (Fig. 4A).

FIG. 4.

(A) Transcription of tcdR during exponential phase relative to transcription of the housekeeping gene rpoA. Means and standard errors of 3 replicates are shown. (B) Transcription of tcdR during stationary phase relative to transcription of the housekeeping gene rpoA. Means and standard errors of 3 biological replicates are shown.

In contrast, during stationary phase, we observed that tcdR expression was high in all strains and correlated with high toxin levels. High-toxin-producing strains tended to have higher levels of transcript of this sigma factor than low toxin producers (Fig. 4B). Expression of tcdR in the HV strains was, on average, 1.2-fold higher than in all non-HV strains.

tcdC expression was evident in exponential phase, and contrary to previous reports of a decrease for VPI 10463 (17), this expression increased in all strains, including VPI 10463, during stationary phase (Fig. 5A and B). Although there was evidence of tcdC transcription in the HV strains, it should be noted that this expression did not result in a functional protein in the strains tested (due to the presence of a frameshift mutation at bp117); thus, the relevance of tcdC expression in these strains is unclear. Expression of tcdC during exponential phase was higher in strain K14 than in the other strains (P ≤ 0.001).

FIG. 5.

(A) Transcription of tcdC during exponential phase relative to transcription of the housekeeping gene rpoA. Means and standard errors of 3 replicates are shown. (B) Transcription of tcdC during stationary phase relative to transcription of the housekeeping gene rpoA. Means and standard errors of 3 biological replicates are shown.

To better understand some of these trends in gene expression, we performed an alignment of the PaLoc DNA sequences of two highly related strains, 630 and VPI 10463, which exhibited the lowest and highest levels of transcription of toxin and toxin-regulating genes, respectively, across all growth phases. Overall, and consistent with their phylogenetic relatedness (they share a last common ancestor), the PaLoc regions shared 99% DNA sequence identity. Nevertheless, we annotated all open reading frames (based on the annotated 630 sequence) and the experimentally derived and predicted promoters (17), as well as the predicted Rho-independent transcriptional terminators (22). We found several interesting features in the sequences of both strains, all of which may have potential functional implications for toxin production (Fig. 6). First, the negative regulator of tcdC of VPI 10463 contained a mutation that changes aspartic acid 7 of TcdC to glutamic acid. Next, the second transcriptional terminator of cdu1 (the non-PaLoc gene upstream of tcdR) overlaps with one of the predicted promoters of tcdR (tcdRp). A second predicted promoter (tcdRp′) is also present just downstream of tcdRp. Interestingly, the RNA secondary structure predicted to form from the tcdRp′ transcript was less stable in the VPI strain than in the 630 strain (free energy of −9.1 kcal/mol versus −41.01 kcal/mol, respectively). Finally, the third predicted transcriptional terminator for cdu1 lies within the coding sequence of tcdR, consistent with the localization for a transcriptional attenuator. We also aligned codY sequences (predicted promoter, coding region, and predicted terminator) from both strains 630 and VPI 10463 and found no differences in the locus that encodes a global regulator of gene expression, including that of tcdA and tcdB (data not shown).

FIG. 6.

Schematic of the C. difficile PaLoc region. Coding regions and predicted regulatory elements are depicted. The cdu1 gene has three predicted Rho-independent terminators (indicated by hairpins). The formation of any given hairpin usually prevents the formation of others downstream. Further, if ter2 forms, transcription from tcdRp is blocked. ter, terminator; p, promoter; RBS, ribosome binding site. Not to scale.

Since the C. difficile spore is the etiologic, transmissible agent, any alteration in sporulation efficiency can impact the degree of environmental dissemination. We determined the accumulation of spores over the growth cycle and found that HV strains sporulated earlier and accumulated more spores per total volume of culture than non-HV strains. HV strain spore accumulation commenced at 28 h in BHI broth, prior to that of any non-HV strain (Fig. 7A). At 48 h, HV strains BI6 and BI8 had accumulated significantly more spores than all other strains. HV strains BI17 and BI23 had accumulated significantly more spores than strains 630, VPI 10463, J9, and K14. When calculated as the number of spores formed per total number of vegetative cells at the peak of growth (efficiency), HV strains also had the greatest efficiency (up to 3.55%) compared with non-HV strains (≤0.66%; not shown).

FIG. 7.

(A) Spores as heat-resistant CFU over 48 h of growth. Means and standard errors of three biological replicates are shown. (B) Mean spores counted by microscopy per field.

Sporulation efficiency, as evaluated by a second approach (microscopy), also indicated similar trends (Fig. 7B). The highest numbers of spores were seen in HV strains BI6 and BI17, which were not significantly different from each other but which accumulated significantly more spores than the other six strains (P < 0.05).

DISCUSSION

Since the identification of the BI/NAP1/027 genetic cluster as the cause of multiple CDI epidemics, there has been great interest in elucidating the molecular mechanisms contributing to the hypervirulent phenotype, particularly those involved in toxin production. Our data revealed that the four HV C. difficile clinical isolates tested did not make more toxin than all the non-HV C. difficile strains tested. The approach we took also clarified the timing of the onset of toxin production. The evaluation of sporulation determined that significant differences existed in the production of infectious particles by HV strains, the magnitude of which exceeds any difference in the production of toxins.

In this study, we showed that HV C. difficile strains do not make the toxins TcdA and TcdB during exponential growth in BHI medium. BHI medium contains glucose, a known repressor of toxin gene expression. Though it is formally possible that the kinetics of toxin production may vary in different media, it is still unlikely that this would somehow occur only for the HV strains, since multiple studies have shown that the relief of glucose repression impacts non-HV strains (11, 18). Importantly, even our small sample of non-HV strains exhibited a wide range of toxin production. This variance illustrates the importance of the choice of comparator strains when making generalizations about the HV phenotype. We suggest that previous studies (41) may have underestimated the significance of these variances in toxin production, leading to misattributions of the potential mechanism(s) of hypervirulence.

The negative regulator TcdC has been at the center of the debate about toxin production in HV strains. A previous study of the transcription of PaLoc genes in strain VPI 10463 indicated that tcdC was highly expressed during exponential phase and that this expression diminished as growth slowed in stationary phase (17). These data, along with those indicating that TcdC prevents TcdR from complexing with RNA polymerase, led to the model in which TcdC acts as a negative regulator of toxin gene expression in exponential phase (31). Most HV clade C. difficile isolates have a frameshift mutation in tcdC that would result in a truncated 65-amino-acid protein (8). It has been hypothesized that the lack of a functional TcdC results in increased transcription of the toxin genes and thus more toxin, possibly during exponential growth (41). However, our results clearly demonstrate that there is no TcdA/B protein detectable during exponential phase in HV strains and that neither tcdA nor tcdB expression is dramatically altered compared to that in non-HV strains. These data are consistent with the absolute requirement for TcdR in toxin gene expression and the strong repression of tcdR (observed by us and others) in exponential phase (9, 17, 28). Thus, the absence of TcdC is likely not sufficient to permit toxin production during exponential growth and underscores the critical requirement for TcdR in tcdA and tcdB expression.

The prevailing model of TcdC as a negative regulator of toxin gene expression describes the fact that the TcdC protein diminishes (by an unknown regulatory mechanism) in stationary phase. However, we observed relatively high transcription of tcdC in stationary phase. This finding is consistent with recently published data. In particular, while examining the role of CodY (a global regulator of gene expression in Gram-positive bacteria) in toxin gene repression, Dineen and colleagues found that in strain 630, tcdC was expressed during both exponential and stationary phases (9). Karlsson et al. also demonstrated that tcdC expression increased in stationary phase and that this expression was suppressed by nutrient supplementation, similar to what was seen for the toxin genes (19).

We thus hypothesize that TcdC, while not required for exponential-phase tcdA and tcdB repression, may exert a modulatory effect on toxin production. The lack of this modulator in HV strains may be one reason they show robust toxin production in stationary phase. However, there are likely multiple influences on the quantity of toxins produced by various strains. TcdC protein levels may be one factor; the affinity of TcdC for its targets may be another. For example, strains 630 and VPI 10463 both have functional TcdC proteins but exhibit vastly different toxin production. A single amino acid difference between the TcdC proteins of these two strains may indicate a difference in the ability to disrupt the interaction of TcdR and RNA polymerase. Obviously, to specifically address these hypotheses, isogenic mutants will need to be generated and tested.

Our studies showing different basal levels of transcription may also suggest differences in regulatory sequences/elements. This is supported by the PaLoc computational analysis of the strains VPI 10463 and 630, which exhibit vastly different toxin production capabilities. We rationalized that subtle alterations may contribute to the differences observed in toxin production, since the genome analyses predicted an otherwise high degree of PaLoc conservation between the two strains. The results suggest the following possibilities. First, the differences in the amino acid compositions of TcdC between VPI and 630 may alter the capacity to negatively regulate toxin production. Second, the use of cdu1 transcriptional terminator 2 would inhibit transcription from the tcdRp promoter, which would lead to preferential use of the tcdRp′ promoter for transcription of tcdR. Interestingly, the secondary structure of the tcdR transcript formed from the tcdRp′ promoter is much less stable in VPI than in 630 (−9.1 kcal/mol versus −41.01 kcal/mol, respectively). This suggests that the third Rho-independent terminator of cdu1 (located in the tcdR coding sequence) is less able to attenuate transcription from the tcdRp′ promoter in the VPI strain. The preferential use of select PaLoc promoters has indeed been previously described for C. difficile (11). codY sequences (predicted promoter, coding region, and predicted terminator) were identical in both strains 630 and VPI 10463, and we found no differences in this locus, which encodes a global regulator of gene expression, including that of tcdA and tcdB (data not shown).

Toxin production and sporulation are both responses to nutrient limitation, and the relationship between them is a subject of debate. One survey of toxigenic C. difficile strains indicated an inverse relationship between the toxin yield and spore counts (2), suggesting that if bacteria sporulate early in stationary phase, there is less time in which to produce toxin. However, an extended period of toxin production under nutrient-limited conditions may result in bacterial death before sporulation is completed. For example, strain VPI 10463 (the high toxin producer) sporulates very poorly (Fig. 7A and B) (2). Other studies, however, indicate that mutants of the stationary-phase regulator Spo0A and its associated sensor kinase are impaired, not only in sporulation, but also in toxin production (40). Further research is thus required to elucidate the links between the two systems, and in particular, the role of the tcdC mutation in HV strains. One might speculate that TcdC may also disrupt sporulation-associated sigma factors.

Our data clearly demonstrate that the four strains in the hypervirulent clade of C. difficile that we tested not only sporulate earlier and with greater efficiency than other strains, but also produce robust amounts of toxin. These data are consistent with another study of different hypervirulent isolates (1). HV strains also produce toxin B with different intoxicating potentials (37). While the correlation between in vitro toxin production and clinical outcomes is not consistent (2), it is possible that the altered toxin B phenotype may have some influence on disease severity. Enhanced sporulation may increase the likelihood of disseminating infectious particles into the environment, acting synergistically with toxin to give an adaptive advantage to hypervirulent C. difficile in terms of pathogenesis.

Most likely, the synergistic confluence of multiple factors has allowed the C. difficile hypervirulent phenotype to emerge. Bacterial factors may include, but not be limited to, genetic alterations resulting in fluoroquinolone resistance, high toxin production, increased sporulation, and colonization efficiencies. Human factors, such as the increased use of antimicrobials in hospital and community settings, as well as inability to control environmental contamination, may also contribute to CDI outbreaks.