|Home | About | Journals | Submit | Contact Us | Français|
Our laboratory has developed testing methods that use real-time PCR and pyrosequencing analysis to enable the rapid identification of potential hypervirulent Clostridium difficile strains. We describe a real-time PCR assay that detects four C. difficile genes encoding toxins A (tcdA) and B (tcdB) and the binary toxin genes (cdtA and cdtB), as well as a pyrosequencing assay that detects common deletions in the tcdC gene in less than 4 h. A subset of historical and recent C. difficile isolates (n = 31) was also analyzed by pulsed-field gel electrophoresis to determine the circulating North American pulsed-field (NAP) types that have been isolated in New York State. Thirteen different NAP types were found among the 31 isolates tested, 13 of which were NAP type 1 strains. To further assess the best approach to utilizing our conventional and molecular methods, we studied the populations of C. difficile in patient stool specimens (n = 23). Our results indicated that 13% of individual stool specimens had heterogeneous populations of C. difficile when we compared the molecular characterization results for multiple bacterial isolates (n = 10). Direct molecular analysis of stool specimens gave results that correlated well with the results obtained with cultured stool specimens; the direct molecular analysis was rapid, informative, and less costly than the testing of multiple patient stool isolates.
Clostridium difficile is one of the leading causes of infectious antibiotic-associated diarrhea and pseudomembranous colitis worldwide (2, 16). This is illustrated by the increased incidence and severity of C. difficile infection, suggesting the emergence of a new hypervirulent strain (5, 13-15, 17, 25, 32).
While TcdB, a cytotoxin, is the known established virulence factor of C. difficile, toxin A (TcdA), a cytotoxic enterotoxin, works synergistically with TcdB, causing damage to the intestinal mucosa in cases of C. difficile infection (17). The genes that encode these toxins are located on the pathogenicity locus of C. difficile (4, 10, 24). Additionally, several deletions in the tcdC gene, a putative negative regulator of the expression of the toxin A (tcdA) and the toxin B (tcdB) genes, have been identified, and these deletions result in higher levels of cytotoxin expression (11). Furthermore, research has shown that some C. difficile strains produce another toxin, known as the binary toxin (19, 22, 28). The genes that encode this toxin, cdtA and cdtB, together produce an actin-specific ADP-ribosyltransferase that induces damage to the actin skeleton, leading to cytopathic effects in cell lines (1). It has been suggested that the binary toxin genes and deletions in the tcdC gene are potential virulence factors in the recent emerging hypervirulent strain (22, 29).
The “gold standard” for the detection of C. difficile toxin production is a cytotoxin assay with stool specimens or isolates from anaerobic culture. The cytotoxin assay is difficult to perform and time-consuming, and it is often less sensitive than molecular assays (20, 23, 26). Enzyme immunoassays (EIAs) are used most often, and recent reports suggest that manufacturers have improved the performance of EIA kits since their introduction; however, the disadvantages of EIAs include the lower levels of sensitivity and specificity compared to those of the gold standard methods. More importantly, culture is not specific for the identification of toxigenic strains. The laboratory at the Wadsworth Center has developed a multiplex real-time PCR assay and a tcdC gene pyrosequencing assay that rapidly identify potential virulence factors of C. difficile strains and that can be used to directly test patient stool specimens for C. difficile.
(Part of this report was presented at the 107th American Society for Microbiology General Meeting in 2007 [Toronto, Canada].)
The bacterial strains utilized in this study included a specificity panel of 55 Clostridium difficile isolates, gastrointestinal flora, and other bacterial pathogens, which are listed with their sources in Table Table1.1. Included in this panel were 30 strains of C. difficile with known toxin gene profiles. Ten strains were positive for only the clostridial toxins (tcdA, tcdB), 10 strains were positive for all four toxin genes (tcdA, tcdB, cdtA, cdtB), and 10 were nontoxigenic. Archived frozen specimens from several hospitals were also tested. Thirty-six isolates from stool specimens received by our laboratory between 1990 and 1992 were from a Veterans Affairs facility in Albany, NY (18, 21), and 113 stool specimens from 2005 to the present came from several different hospitals in New York State. In total, 149 specimens were examined.
C. difficile strains were plated on CDC anaerobe agar and grown at 35°C under anaerobic conditions for 48 h. For molecular testing, isolated colonies were suspended in 1× PCR buffer (Applied Biosystems, Foster City, CA) at a level equivalent to a 1 McFarland standard and heat treated at 95°C for 20 min. The specimens were either used directly in the PCR or stored at −20°C until use. Direct, uncultured stool specimens were subjected to nucleic acid extraction with the QIAamp DNA stool minikit by use of a QIAcube (Qiagen, Valencia, CA), according to the manufacturer's instructions.
Unique primer and probe sets targeting C. difficile enterotoxin A (tcdA), cytotoxin B (tcdB), and binary toxin genes A and B (cdtA and cdtB, respectively) were designed by using Primer Express (version 2.0) software (Applied Biosystems) on the basis of the sequences of strains Clostridium difficile VPI 10463 (NCBI accession number X92982, ATCC 43255) and Clostridium difficile CD196. The tcdA and tcdB primers and probes were selected so that they did not overlap known variable regions. The cdtA and cdtB primers and probes were designed to bind to the full 4.3-kb fragment which encodes the functional binary toxin gene, avoiding areas of homology to box 1, box 2, and box 3 found in all C. difficile strains (22). The primers and probe sequences are summarized in Table Table2.2. BLAST (the Basic Local Alignment Search Tool) searches were performed with all selected potential primer and probe oligonucleotides to test that no cross-reaction would occur with other human or microbial pathogens. The Amplify (version 3.0) program was utilized to determine any possible primer-dimer interactions among the oligonucleotides. Probes TcdA-P, CdtB-P, and CdtA-P and all primers were synthesized by Integrated DNA Technologies (Coralville, IA). Probes TcdB-P and Inhib-P were obtained from Applied Biosystems.
All isolates were tested in duplicate, and the DNA from stool specimens was tested in triplicate. The real-time PCR assay was performed on an ABI 7500 Fast real-time PCR system (Applied Biosystems) in a 96-well optical plate format with a LightCycler-FastStart DNA master hybridization probes kit (Roche Diagnostics Corp, Indianapolis, IN). For testing of the isolates, each 25-μl reaction mixture consisted of 1× LightCycler-FastStart DNA master hybridization probes, 3.0 mM MgCl2, 900 nM of all eight primers (Table (Table2),2), 250 nM of all four probes (Table (Table2),2), sterile water, and 10 μl DNA template. For the testing of stool samples, probes TcdA-P, TcdB-P, and CdtA-P were initially added to the mixture. The master mixture was then split, and probe CdtB-P was added to half of the master mixture for the testing of the DNA isolated from the stool and probe Inhib-P was added to the other half of the master mixture; the latter portion was then seeded with the internal inhibition control, for the assessment of inhibition. The thermal cycling conditions were as follows: 1 cycle of 95°C for 10 min, followed by 40 cycles (45 cycles for testing of stool specimens) at 95°C for 15 s and 60°C for 1 min. The data analysis was performed according to the manufacturer's instructions, with no passive reference dye being used.
The analytical sensitivity of the real-time PCR was determined for each target through the use of 10-fold dilutions of DNA from a hypervirulent strain of C. difficile (strain NYS BAC-05-3671) with known densities and by the generation of standard curves. The sensitivity of detection in stool specimens was determined by spiking stool specimens known to be negative with known densities of C. difficile cells, processing them according to the protocol, and testing the purified DNA by real-time PCR. The specificity of the assay was determined by testing the DNA from 55 organisms, as shown in Table Table11.
The DNA isolated from stool specimens was also tested for the presence of PCR inhibitors. An internal control plasmid was constructed from a PCR product that contained a portion of heterologous DNA flanked by the primer binding sites of the cdtA gene. This PCR product was cloned into a plasmid vector (pCR2.1-TOPO) with a TOPO TA cloning kit (Invitrogen Life Technologies, Carlsbad, CA). The plasmid was diluted to yield a threshold cycle (CT) value of approximately 30. The control plasmid was added to the PCR mixtures containing DNA from the stool specimen and was detected by the use of a different fluorescent probe (Table (Table2).2). The CT value obtained from this reaction was compared with that obtained from a second reaction containing only the control plasmid. The reaction was considered to be inhibited if the CT value of the reaction was 3 CT units or more greater than that for the reaction containing only the control plasmid or if no amplification was found.
To determine whether deletions were present or absent, we performed a conventional PCR and then sequenced the amplified products. A primer pair was adapted to target the tcdC gene (24); it is shown in Table Table2.2. Each 100-μl amplification reaction mixture consisted of 1× PCR buffer; 2.5 mM MgCl2; 10 mM each of dATP, dCTP, dTTP, and dGTP; sterile water; 50 μM of primers C1 and C2; and 2.5 U of AmpliTaq Gold (Applied Biosystems). The PCR products were analyzed on a 2% E-gel (Invitrogen Life Technologies) with a 1-kb DNA marker (Invitrogen Life Technologies) and were visualized under UV light by using a Bio-Rad Laboratories (Hercules, CA) Gel Doc XR apparatus. The PCR products were purified with ExoSAP-IT (USB, Cleveland, OH) and were sequenced by the Applied Genomic Technologies Core facility at the Wadsworth Center. The resultant sequences were compared to the sequence of the tcdC gene (NCBI accession number X92982 for C. difficile strain VPI 10463) to determine the presence or absence of tcdC deletions.
A pyrosequencing protocol was developed to determine whether the 1-bp, 18-bp, and/or 39-bp deletions within a portion of the tcdC gene were present or absent. Amplification and sequencing primers were designed with Assay Design software (Qiagen; previously Biotage). Briefly, a 378-bp PCR product was amplified with forward primer tcdC-F, which was labeled with biotin for immobilization onto streptavidin-coated Sepharose beads. Two separate sequencing primers (primers TcdC-S1 and TcdC-S2) were used to assess the regions where the 1-bp and 18-bp deletions may be present. If the 18-bp pyrosequencing reaction failed, a third sequencing primer (primer TcdC-S3) was used to assess the sample for the 39-bp deletion, since the binding region of the 18-bp sequencing primer falls within the region of the 39-bp deletion. Primers TcdC-F and TcdC-R were synthesized by Integrated DNA Technologies and are shown in Table Table2.2. The locations of the sequences of the PCR primers and the sequencing primers within the tcdC gene are shown in Fig. Fig.1.1. The amplification reaction mixtures consisted of 1× PCR buffer; 2.0 mM MgCl2; 10 mM each of dATP, dCTP, dTTP, and dGTP; sterile water; 200 nM of primers TcdC-F and TcdC-R; 2.5 U of AmpliTaq Gold (Applied Biosystems); and 10 μl of template DNA in a final volume of 100 μl. The thermocycling conditions in a Bio-Rad Laboratories iCycler apparatus were as follows: 1 cycle of 5 min at 95°C, followed by 50 cycles of 30 s at 95°C, 30 s at 50°C, and 30 s at 72°C and a final extension for 5 min at 72°C. Pyrosequencing was performed with Pyro Gold reagents on a PyroMark vacuum prep workstation and a PyroMark ID instrument according to the manufacturer's instructions (Qiagen). The specific dispensation order for the 1-bp, 18-bp, and 39-bp deletions was GCTGAATAT, GTTAGCTCTCAGCTAGCT, and GCTCTCTCTCTCT, respectively, with single nucleotide polymorphism analysis of the nucleotides.
A selection of 31 strains of C. difficile was additionally examined by pulsed-field gel electrophoresis (PFGE) to preliminarily assess the North American pulsed-field (NAP) types in our collection and for comparison of the NAP types to potential virulence characteristics. These strains were chosen from among strains submitted between 1980 and 2008 and included strains that we had determined harbored the tcdA and tcdB toxin genes or the tcdA, tcdB, cdtA, and cdtB toxin genes. DNA was prepared from C. difficile cells grown for 7 h in reduced peptone-yeast-glucose broth at 37°C, according to established protocols (12). DNA was digested with restriction endonuclease SmaI (New England Biolabs, Beverly, MA), and the fragments were separated in 1.0% agarose gels on a clamped homogeneous electric field apparatus (CHEF Mapper; Bio-Rad Laboratories, Richmond, CA). The initial pulse time of 5.0 s was increased to 40.0 s over 19.5 h. The gels were stained with ethidium bromide, destained in distilled water, and visualized with a Gel Doc 2000 gel analysis system (Bio-Rad Laboratories). The PFGE patterns were analyzed with Bionumerics software (Applied Maths, Belgium). Dendrograms were created by use of unweighted pair group similarity and arithmetic mean with Dice coefficients, and the position tolerance was set at 1.1%. NAP types were determined by comparison with known NAP types. An isolate with ≥80% similarity to a known NAP type was considered to belong to that NAP type (11).
Twenty-three stool specimens from cases of presumed community-associated C. difficile infection were examined to determine whether the patients studied had homogeneous populations of C. difficile. The definition of presumed community-associated C. difficile infection was a positive EIA result for C. difficile toxin with a specimen collected when the patient was as an outpatient or within 72 h of patient admission to a hospital and with no evidence of hospitalization in the preceding 3 months. “Hospitalization” was defined as an overnight stay in a hospital or other skilled nursing facility. The patients were surveyed for antibiotic use preceding C. difficile infection, but antibiotic use was not used as an exclusion criterion.
Specimens were collected and stored frozen until culture could be performed. Twenty microliters of thawed stool specimen was inoculated onto in-house-prepared Clostridium difficile selective agar and cycloserine-cefoxitin-fructose agar with antibiotic concentrations of 480 μg/ml for cycloserine and 15.4 μg/ml for cefoxitin (9). The media were not prereduced before they were plated. The plates were incubated for 4 days at 37°C under anaerobic conditions. Up to 10 colonies from each specimen were subcultured to CDC anaerobe agar. One colony from each specimen was identified as C. difficile by a battery of conventional biochemical tests, including Gram staining and tests for anaerobic growth requirement, motility, catalase production, indole reaction, nitrate reduction, H2S production, growth in bile, lipase production, lecithinase production, and urease production. The isolates were also assayed for their abilities to ferment carbohydrates; the substrates included glucose, maltose, mannitol, lactose, sucrose, xylose, salicin, arabinose, and glycerol. Cellular fatty acid analysis was also performed by using the Microbial Identification System (Midi Inc., Newark, DE). The remaining isolates from each specimen were identified only by Gram staining, a test for anaerobic growth requirement, and cellular fatty acid analysis. Each isolate was then assessed for the presence of potential virulence factors by the molecular methods described above. In a further prospective analysis, two isolates from each of the 89 patient samples were examined in a continuation of our assessment of C. difficile bacterial populations and virulence characteristics.
The analytical results obtained with the isolated colonies indicate that the real-time PCR assay evaluated in the present study has a limit of detection of 1 CFU per reaction mixture. The sensitivity of both the real-time PCR assay and pyrosequencing by direct testing of stool specimens was determined to be 10 CFU per reaction mixture (results not shown). In a study that compared the results of the real-time PCR assay to those of the gold standard cytotoxin assay, the real-time assay was found to be 100% sensitive (results not shown). The specificity of the real-time PCR assay was determined to be 100%, as the C. difficile strains assessed in the multiplex real-time PCR were positive for the respective genes (tcdA and tcdB or tcdA, tcdB, cdtA, and cdtB) and all non-C. difficile bacterial isolates and nontoxigenic C. difficile were negative for all four genes, as expected.
The tcdC gene from 154 isolates was amplified and sequenced by both the pyrosequencing and the Sanger sequencing methods to determine the presence or absence of the 1-bp, 18-bp, and/or 39-bp deletions. Comparison of the results obtained by the two tests revealed a 100% correlation between the results of the two methods (results not shown).
Figure Figure22 summarizes the results of PFGE analysis of C. difficile isolates to obtain a preliminary understanding of the NAP types circulating in New York State. Thirteen NAP type 1 (NAP1) types were identified in the group and possessed all four toxin genes and both the 1-bp and 18-bp tcdC gene deletions. Eight additional NAP types were also identified (NAP2, NAP3, NAP4, NAP5, NAP8, NAP10, NAP11, NAP12). All the isolates with NAP types 2, 3, 4, 5, and 10 and undesignated NAP types CldS16015 and CldS16006 were C. difficile isolates that contained no binary toxin genes or tcdC gene deletions. Interestingly, NAP types 8, 11, and 12, as well as two others, types CldS16002 and CldS16005, were found to have some or all of the potential virulence characteristics. The NAP8 (2008) strain was positive for all toxin genes and the tcdC gene 18-bp deletion. The NAP11 (2006) and NAP12 (1990, 1992) strains were negative for the binary toxin genes but positive for the tcdC gene 18-bp deletion. Two strains of types CldS16002 and CldS16005 (2006), respectively, differed from the strains designated NAP1 by four bands but were found to be positive for all toxin genes and for the tcdC gene 1-bp and 18-bp deletions.
Twenty-three stool specimens were examined by culture, real-time multiplex PCR, as well as Sanger sequencing and pyrosequence analysis for tcdC gene deletions. All stool specimens were initially found to be positive for C. difficile by culture. One hundred ninety-four colonies from the 23 specimens were examined. The data in Table Table33 demonstrate that of the 194 strains isolated, 3 strains (1.5%) exhibited a molecular profile different from that of the other strains isolated from the same individual's stool specimen. Analysis of the stool specimens revealed that three stool specimens (13%) did not contain homogeneous populations. In two of these stool specimens (specimens 4 and 15), the difference was the presence of the binary toxin genes in one or more of the isolates of the patient's stool specimens. For the third specimen (specimen 1), the difference was the absence of both the binary toxin genes and the tcdC gene deletion in 1 of 10 isolates from the same patient's stool specimen. Overall, among the 89 stool specimens tested in this prospective analysis (two isolates per stool specimen), no differences in their molecular profiles were identified, as illustrated in Table Table44.
The C. difficile population data led us to evaluate the potential benefits of direct molecular analysis of stool specimens. A subset of 24 stool specimens was directly tested by the multiplex real-time PCR and pyrosequencing. The preliminary data showed a correlation between the results of direct testing and those of testing following culture. For 23 (95.8%) of the stool specimens tested, identical results and molecular profiles were obtained. However, one stool specimen was negative by direct testing, while culture of that stool specimen identified two colonies of C. difficile.
Although the binary toxin genes and the tcdC gene deletions and their contribution to the virulence of C. difficile strains causing infections are still poorly understood, characterization may be important for infection control and a proper public health response. Currently, only a few real-time PCR assays are available for the detection of toxins A and B (3, 6, 23, 31). We have developed a sensitive and specific one-tube, real-time PCR assay capable of determining in less than 2 h whether toxins A and B (tcdA and tcdB) and the binary toxins (cdtA and cdtB) are present. Pyrosequencing is a novel technique that we have implemented for the detection of the deletions in the tcdC gene. Studies have previously shown that the deletions that have been described in the tcdC gene (7, 17, 23, 27) may be responsible for increases in toxin production. Two deletions, a 39-bp deletion and an 18-bp deletion, were thought to contribute to the enhancement of disease severity and high rates of morbidity; however, newer investigations suggest that a 1-bp deletion in the tcdC gene may be the cause of the overproduction of the toxins (7). Utilization of these assays may be informative and may help to provide an understanding of the roles of these genetic factors in C. difficile infections.
Following several outbreaks in 2003, further characterization of the hypervirulent strain of C. difficile determined that it is PCR ribotype 027 (PCR ribotyping), NAP1 (PFGE), restriction endonuclease analysis type BI, and toxinotype III (toxinotyping) (5, 13, 16). In our analysis of the NAP types circulating in New York State, we found that NAP1 strains were first isolated in 2005. The NAP1 strains that we tested were isolated in recent years, and most contained the four toxin genes in combination with the 1-bp and 18-bp tcdC deletions. Retrospectively, we were also interested in assessing the molecular profiles of strains that had caused major outbreaks at a local facility in 1990 and 1992. We found several NAP strains that contained the 18-bp tcdC deletion; the other two NAP types were found to be negative for the binary toxin genes and tcdC deletions. We also observed other NAP types that had subsets of the genetic factors present; it is unclear whether the strains identified as being unmatched to known NAP designations have been recognized previously or whether they are novel. These data illustrate that these molecular profiles have actually been present in strains for a long time, yet it is only recently that more virulent disease has been appreciated. More work directed toward obtaining an understanding of the NAP types circulating throughout the world and the morbidity and mortality rates associated with each of the strains should continue to shed light on the hypervirulent strain of C. difficile.
The assessment of the C. difficile populations within individual stool specimens demonstrated that the C. difficile isolates recovered from a stool specimen are for the most part homogeneous in a patient with presumed community-associated C. difficile infection (described above). However, the isolates in 13% of the stool specimens examined were not homogeneous by our molecular methods, and this raises the question of whether the determination of potential genetic factors from a single C. difficile colony isolated from a stool specimen is the best practice. We are not the first to show that more than one strain can be present in a given stool specimen; however, conflicting data have been reported (30). In our analysis, we have shown that for at least one case (stool specimen 1 in Table Table3),3), the possibility exists that the single colony possessing no binary toxin genes or tcdC gene deletions could have been the one isolated from the patient's stool specimen, thus resulting in a false-negative result for that patient. The further prospective analysis of 89 additional stool specimens could have shed more light on this question; however, no heterogeneity was identified in that set of specimens.
The results of our testing have led us to conclude that characterization of C. difficile directly from stool specimens is the testing route of choice at the Wadsworth Center for the rapid identification of putative genetic factors of C. difficile. The idea of testing stool specimens directly is not a novel one (8). Our preliminary data have shown that C. difficile could not be detected by the direct method in 4% (1/24) of specimens. The specimen in which C. difficile could not be detected by the direct method was found, upon culture analysis, to have only two colonies of C. difficile, which was below our limit of detection for stool specimens. We believe, on the basis of our data, that direct testing of stool specimens combined with culture at the start of testing is the most informative approach for laboratories that have the ability to perform all of these methods. If only PCR analysis of stool specimens is utilized, there is a risk that a potentially highly virulent strain may fail to be identified. If culture is routinely performed first, followed by molecular analysis of one or even a few colonies from each patient, there is still the possibility that the hypervirulent strain will be missed among the typical C. difficile strains in a given stool specimen. The latter approach can also lead to additional costs because of the need to perform more than one molecular test per specimen. An alternative approach, which we have previously utilized for other pathogens, is to first perform molecular analysis as a screen and to then perform culture only with the specimens with negative or inconclusive results.
In conclusion, the C. difficile multiplex real-time PCR assay that we have developed is sensitive, specific, and rapid. Pyrosequencing of the tcdC gene is a novel and valuable tool for the further characterization of C. difficile isolates that cause infections and that are detected in stool specimens. These molecular assays can be applied directly for the screening of patient specimens and may provide valuable information that helps provide an understanding of these genetic factors and the roles that they may play in C. difficile infection.
We acknowledge the Wadsworth Center Applied Genomic Technologies Core Facility for Sanger sequencing of the PCR products and Adriana Verschoor and Ron Limberger for critical reading of the manuscript. We thank Angela Thompson from the CDC for her assistance with determining the NAP types from our PFGE data and Qiagen (Biotage) for technical assistance with the design of the pyrosequencing assay. We also acknowledge the technical support of Yvette Khachadorian with PFGE analysis and the Wadsworth Center Media Department for making the Clostridium difficile selective agar and cycloserine-cefoxitin-fructose agar plates for stool culture.
Funding for this study was provided in part from the Emerging Infections Program, CDC, Atlanta, GA.
Published ahead of print on 29 April 2009.