Development of TaqMan-Based Quantitative PCR for Sensitive and Selective Detection of Toxigenic Clostridium difficile in Human Stools

Background Clostridium difficile is the main cause of nosocomial diarrhea, but is also found in asymptomatic subjects that are potentially involved in transmission of C. difficile infection. A sensitive and accurate detection method of C. difficile, especially toxigenic strains is indispensable for the epidemiological investigation. Methods TaqMan-based quantitative-PCR (qPCR) method for targeting 16S rRNA, tcdB, and tcdA genes of C. difficile was developed. The detection limit and accuracy of qPCR were evaluated by analyzing stool samples spiked with known amounts of C. difficile. A total of 235 stool specimens collected from 82 elderly nursing home residents were examined by qPCR, and the validity was evaluated by comparing the detection result with that by C. difficile selective culture (CDSC). Results The analysis of C. difficile-spiked stools confirmed that qPCR quantified whole C. difficile (TcdA+TcdB+, TcdA−TcdB+, and TcdA−TcdB− types), TcdB-producing strains (TcdA+TcdB+ and TcdA−TcdB+ types), and TcdA-producing strains (TcdA+TcdB+ type), respectively, with a lower detection limit of 103 cells/g of stool. Of the 235 specimens examined, 12 specimens (5.1%) were C. difficile-positive by qPCR: TcdA+TcdB+ strain in six specimens and TcdA−TcdB− strain in the other six. CDSC detected C. difficile in 9 of the 12 specimens, and toxigenic types of the isolates from the 9 specimens were consistent with those identified by qPCR, supporting the validity of our qPCR method. Moreover, the qPCR examination revealed that the carriage rate of whole C. difficile and that of toxigenic strains in the 82 subjects over a 6-month period ranged from 2.4 to 6.8% and 1.2 to 3.8%, respectively. An average qPCR count of C. difficile detected was 104.5 cells/g of stool, suggesting that C. difficile constituted a very small fraction of intestinal microbiota. Conclusion Our qPCR method should be an effective tool for both clinical diagnosis and epidemiological investigation of C. difficile.


Introduction
Clostridium difficile is the most common cause of healthcareassociated infectious diarrhea [1]. This bacterium can be categorized into three types on the basis of its production of two major toxins, TcdA and TcdB, namely i) TcdA-positive, TcdBpositive (A + B + ) type; ii) TcdA-negative, TcdB-positive (A 2 B + ) type; and iii) TcdA-negative, TcdB-negative (A 2 B 2 ) type. Toxigenic strains, A + B + and A 2 B + types, are responsible for C. difficile infection (CDI). The elderly are particularly at increased risk of CDI possibly because of age-related changes in intestinal microbiota, weakened immune systems, and the presence of underlying diseases [2]. On the other hand, several surveys have reported that toxigenic C. difficile was detected in stools of asymptomatic elderly people [3][4][5]; such asymptomatic carriers may play a role in the transmission of CDI [5]. Selective detection of toxigenic C. difficile strains is therefore important, not only for clinical diagnosis but also for epidemiological investigations of C. difficile for infection control. Moreover, the quantification method of toxigenic C. difficile is useful in basic research such as studies to evaluate correlation of C. difficile count in the intestine and severity of the CDI symptom or to find an association between C. difficile colonization and the presence or absence of other intestinal bacteria.
C. difficile, including toxigenic strains in stool specimens, is traditionally detected by the culture method using selective medium. The C. difficile selective culture (CDSC) is a valuable method with advantages, such as high detection sensitivity and availability of isolates for characterization. However, the culture method is generally labor-intensive and time-consuming. In addition, since the culture method identifies predominant strains of C. difficile in stools, toxigenic strains that coexist with a larger population of non-toxigenic strains can be overlooked, which can fall into a false-negative result of toxigenic C. difficile detection. PCR-based methods have solved these shortcomings and have been widely used to detect toxigenic C. difficile in stools [6][7][8][9]. Several kinds of commercially available PCR kits are reported to be useful and reliable [10,11]. However, their quantitative application remains insufficient. A recent study indicated the possibility of quantification by showing that the threshold cycle (C T ) values of a commercial PCR targeting the tcdB gene were inversely correlated with C. difficile culture counts [12].
The purpose of this study was to develop a sensitive quantification method of toxigenic C. difficile and to evaluate its validity. We developed a new TaqMan real-time quantitative PCR (qPCR) method using sets of primers and a double-labeled fluorescent probe targeting the 16S rRNA, tcdA, and tcdB genes, and analyzed C. difficile in stools collected from elderly residents of nursing homes by means of qPCR.

Reference strains and culture conditions
All of the C. difficile strains (DSM 1296 T , ATCC 43255, ATCC 43596, ATCC 43598, ATCC 700057, CCUG 20309, CCUG 37780, CCUG 37785, NTCT 13307, NTCT 13366) and the other organisms belonging to the Clostridium genus that we used were cultured anaerobically in modified Gifu anaerobic medium broth (Nissui Pharmaceutical Co., Ltd., Tokyo, Japan) supplemented with 1% glucose (1% Glc-mGAM) at 37uC. Total bacterial cell counts of fresh cultures were determined by using the 49, 6-diamidino-2-phenylindole (DAPI) staining method in accordance with the method of Jansen et al. [13]. On the basis of the DAPI counts, each fresh culture was diluted to obtain 1 mL of bacterial suspension containing 10 9 cells. Twenty microliters of suspension containing 2610 7 cells was stored at -80uC until use for DNA extraction.

Development of primers and probes
The primers and probes listed in Table 1 were newly developed for this study. 16S rRNA gene-targeted primers (CD16SrRNA-F/ R) and a probe (CD16SrRNA-P) for the detection of whole C. difficile (both toxigenic and non-toxigenic strains) were designed as follows. The sequences of C. difficile strains and other clostridial species were obtained from Ribosomal Database Project (http:// rdp.cme.msu.edu/). Multiple alignments of these genes with the Clustal X program (http://www.clustal.org/clustal2/) were used to identify highly conserved regions as targets of the C. difficile specific primers and probe. For the detection of TcdA-producing strains (A + B + ) and TcdB-producing strains (A + B + and A 2 B + ), respective primers-probe sets targeting tcdA (tcdA-F/R/P) and tcdB (tcdB-F/R/P) were developed by using a procedure similar to that with CD16SrRNA-F/R/P. The sequences of tcdA and tcdB of C. difficile and other Large Clostridial Toxin genes as references obtained from GenBank were used for multiple alignments and identification of target sites. The accession numbers of gene sequences used for the design of the primers and probes were listed in Table S1.

Stool specimens
Stool specimens were collected from 82 elderly residents from four nursing homes in France (11 males and 71 females; ages 66 to 94 years [average 6 standard deviation, 8466.2 years]). Stools were collected once every 3 months, three times in total (S 1 , S 2 , and S 3 ) from each subject. Since no subjects had abdominal symptoms at these samplings, no diarrheal stools or CDI-suspected stools were included. The subject ID was composed of five digits: the first two digits specified the site number (from 01 to 04) and the latter three digits specified the subject number (from 001).

Ethics statement
This study was conducted in compliance with the Declaration of Helsinki, good clinical practice (ICH topic E6; CPMP/ICH/135/ 95), and applicable regulatory requirements (French Public Health Code). The Independent Ethics Committee in Caen in France, COMITE DE PROTECTION DES PERSONNES NORD OUEST III, approved this study. Written informed consent forms were obtained from the subjects or their legal representatives.

Preparation of stool specimens for each analysis
Stool specimens within 24 h after defecation were collected and immediately examined for C. difficile toxins by enzyme immunoassay (EIA) at study sites. A portion of each stool was concurrently collected into an empty tube, and stored at -20uC until transportation. The stool samples were transported in a frozen state from the sites to our laboratory, and stored at -20uC until use for the following pretreatment for qPCR and CDSC.
After being thawed, each stool was weighed and suspended in 9 volumes of Dulbecco's PBS (-) (Nissui Pharmaceutical) to make a 10% (w/v) stool homogenate (100 mg stool/mL). One hundred microliters of stool homogenate was used immediately for CDSC. Two milliliters of the 10% stool homogenate was centrifuged at 16,0006g for 5 min, and the supernatant was discarded. The stool pellets (200 mg) were stored at -80uC until use for DNA extraction.

DNA extraction
DNA was extracted from pure cultured bacteria and stool pellets by using a FastDNA SPIN Kit for Feces (MP Biomedicals, Illkirch, France) in accordance with the manufacturer's instructions, with the exception that the first wash step with glass beads, sodium phosphate buffer, and PLS solution provided in the kit was skipped only in the extraction from pure culture bacteria. This is because the wash resulted in reduction in yield probably due to a difficulty of precipitation of bacterial cells with only glass beads in the buffer. DNA extracted from the pure cultured bacteria (2610 7 cells) or 200 mg of stool pellets was finally dissolved in 100 mL of provided buffer.

Selectivity of qPCR using new primers-probe sets
The selectivity of qPCR using the newly designed primers-probe sets was determined. DNA fractions extracted from pure cultures of each strain shown in Table 2 at doses corresponding to 10 5 cells per reaction were applied to qPCR with 16SrRNA-F/R/P, tcdA-F/R/P, and tcdB-F/R/P. C T values within 63.3 of that obtained with the standard strain (C. difficile DSM 1296 T ) were considered positive (+), whereas C T values of more than 50 were considered negative (-) ( Table 2). To evaluate the accuracy of quantitative detection among the target C. difficile strains, each analytical curve of the 10 C. difficile strains was generated by using the respective primers-probe sets ( Table 3).
Comparison of analytical curve of C. difficile-spiked stool with that of C. difficile pure culture A stool specimen lacking any amplification by qPCR using CD16SrRNA-F/R/P was selected, as this response suggested that the specimen lacked indigenous populations of C. difficile. The C. difficile-negative stool was diluted with Dulbecco's PBS (-) to make a 10% (w/v) stool homogenate (100 mg stool/mL). The number of cells in a pure culture of C. difficile DSM 1296 T in 1% Glc-mGAM broth was counted by using DAPI staining, and serial dilutions of the pure culture ranging from 10 4 to 10 9 cells/mL were prepared. Twenty microliters of the serial dilutions (containing C. difficile cells ranging from 2610 2 to 2610 7 ) were spiked into the 2 mL of the C. difficile-negative stool homogenates (containing 200 mg of stool) to obtain stool specimens with C. difficile at final concentrations ranging from 10 3 to 10 8 cells/g of stool. DNA extracted from the stool specimens was applied to qPCR, and the obtained C T values were used to generate an analytical curve of the C. difficile-spiked stool. DNA extracted from 2610 7 cells of pure cultured C. difficile and its serial dilutions were applied to qPCR to generate a standard analytical curve of the C. difficile pure culture. These two analytical curves were compared to evaluate the lower detection limit and detection accuracy of this qPCR method.

Determination of bacterial count by qPCR
Whole C. difficile, TcdA-producing strains, and TcdB-producing strains were enumerated by qPCR with 16SrRNA-F/R/P, tcdA-F/R/P, and tcdB-F/R/P, as follows. C. difficile DSM 1296 T A + B + strain was selected as a standard strain for generating standard analytical curves for all three target C. difficile groups. Five microliters of 10-fold serial dilutions of DNA extracted from the pure culture of the C. difficile strain were applied to PCR to obtain a standard analytical curve ranging from 10 1 to 10 5 cells/5-mL reaction. Five microliters of the DNA solution extracted from 200 mg of stool and its 2-and 4-fold dilutions were applied to PCR as a template containing the corresponding DNA from 10, 5, or 2.5 mg of stool. The consequent C T values of the stool specimens were applied to the standard analytical curve, and the corresponding bacterial counts (cells/g of stool) were calculated as qPCR counts.

Determination of toxigenic types of C. difficile predominating in individual stool specimens
By comparing the three qPCR counts targeting the 16S rRNA gene for whole C. difficile (A + B + , A 2 B + , and A 2 B 2 types), tcdB for TcdB-producing strains (A + B + and A 2 B + types), and tcdA for TcdA-producing strains (A + B + type), the toxigenic types of C. difficile predominating in each stool specimen were determined as follows. When the differences among the three qPCR counts were within 0.3 log 10 cells/g of stool (a 2-fold difference in real values), the toxigenic type of the dominant C. difficile strain was identified as A + B + . When the difference between the qPCR counts of the 16S rRNA gene and tcdB was within 0.3 log 10 cells/g of stool and these counts were higher than that of tcdA by at least 0.3 log 10 cells/g of stool, the toxigenic type was identified as A 2 B + . When the qPCR count of the 16S rRNA gene was higher than those of tcdA and tcdB by at least 0.3 log 10 cells/g of stool, the toxigenic type was identified as A 2 B 2 .

Enzyme immunoassay (EIA)
C. difficile toxins in stools were detected with a commercial EIA kit, the Xpect Clostridium Difficile Toxin A/B Test (Remel Inc., Lenexa, KS), in accordance with the manufacturer's instructions.

C. difficile selective culture (CDSC)
In accordance with the method described by Wren [14], C. difficile was isolated by stool culture with cefoxitin cycloserine egg yolk (CCEY) agar. CCEY agar (BioConnections Ltd., Knypersley United Kingdom) supplemented with 40 mL of egg yolk emulsion (BioConnections), two vials of cefoxitin-cycloserin (BioConnections), and 10 mL of lysed horse blood per liter was prepared inhouse and stored at 4uC for a maximum of 1 week before use. The 10% (w/v) stool homogenate was mixed with an equal volume of absolute ethanol and left at room temperature for 30 to 60 min. In an anaerobic glovebox (Coy Laboratory Products Inc., Grass Lake, MI), 100 mL of the alcohol-treated stool or its 10-fold dilution was inoculated onto the CCEY plates and cultured under anaerobic conditions at 37uC for 48 to 72 h. Suspected colonies of C. difficile were selected on the basis of their morphological characteristics and subjected to rapid identification by real-time PCR with the CD16SrRNA-F/R/P set. Obtained C. difficile isolates were examined by PCR for identification of toxigenic types and by cell cytotoxicity neutralization assay (CCNA) for the detection of TcdB.
Identification of toxigenic type of C. difficile isolates by PCR Toxigenic types of C. difficile isolates were identified by comparing the size of PCR amplicons for the respective toxin genes, in accordance with the method described by Kato et al. [15]. An NK11/NK9 primer set for tcdA and an NK104/NK105 primer set for tcdB were used for amplification.
TcdB detection of C. difficile isolates by CCNA The cytopathic effect of the isolates was examined with a C. difficile Toxin/Antitoxin Kit (Techlab, Blacksburg, VA). C. difficile isolates were subcultured in BHI broth anaerobically at 37uC for 5 days. The suspension was then centrifuged at 16,0006g for 5 min and the supernatant filtered with a 0.45-mm membrane filter. The filtrate was added to precultured Vero-B4 cells with or without C. difficile antitoxin. After incubation of the culture at 37uC for 24 or 48 h, the presence or absence of TcdB was determined by judging the cytopathic effects. The following criteria were employed: when at least 90% of the cells were rounded in the absence of antitoxin and the cytopathic effect was neutralized by the antitoxin, the culture was regarded as TcdB positive; when no cytopathic effect was observed in the cells both with and without antitoxin, the culture was regarded as TcdB negative.

Results
Selectivity of qPCR using primers-probe sets targeting the 16S rRNA, tcdA, and tcdB genes For selective detection of whole C. difficile (A + B + , A 2 B + , and A 2 B 2 types), TcdB-producing strains (A + B + and A 2 B + types), and TcdA-producing strains (A + B + type), respective primers-probe sets targeting 16S rRNA, tcdB, and tcdA genes were newly designed ( Table 1). The selectivity of qPCR using the three designed primers-probe sets was assessed by using DNA extracted from each pure culture (Table 2). qPCR using each primers-probe set detected the respective target strains selectively, without any amplification of non-targeted strains. To evaluate the accuracy of quantitative detection among the target C. difficile strains, each analytical curve was compared ( Table 3). The analytical curves of the 10 C. difficile strains (five A + B + strains, two A 2 B + strains, and three A 2 B 2 strains) with 16SrRNA-F/R/P had almost equal slopes. Similarly, for the seven TcdB-producing strains (five A + B + and two A 2 B + strains) with tcdB-F/R/P and the five TcdA- producing strains (five A + B + strains) with tcdA-F/R/P, the slopes of the analytical curves were equivalent. The differences in the intercepts of the analytical curves with 16SrRNA-F/R/P between the type strain and the others ranged from -0.9 to +2.1, suggesting that each target strain could be enumerated within approximately a 4-fold difference. The difference in the intercepts ranged from -1.0 to +0.8 in the case of analytical curves with tcdB-F/R/P and from -0.3 to +1.5 in the case of those with tcdA-F/R/P, thus providing further evidence of the accurate quantification of toxigenic strains. These results indicated that the newly developed TaqMan-based qPCR method was capable of detecting the target C. difficile strains selectively and with high accuracy.

Lower detection limit and detection accuracy of qPCR
The lower detection limit and detection accuracy of qPCR for C. difficile in stools were evaluated by analyzing stool samples spiked with C. difficile vegetative cells at a final concentration of 10 3 to 10 8 cells/g of stool. The analytical curve of the C. difficilespiked stool was compared with the standard analytical curve of the C. difficile pure culture ( Figure 1A-C). In the case of all the three primers-probe sets, the obtained analytical curves were nearly equivalent over the range of 10 3 to 10 8 cells, confirming that qPCR with the standard analytical curve of the pure culture allowed accurate detection of C. difficile in stools. These results also indicated that our qPCR method enabled quantitative detection of C. difficile in stools with a lower detection limit of 10 3 cells/g of stool.

Comparison of results of toxigenic C. difficile detection by qPCR and CDSC
To verify the validity of the new TaqMan-based qPCR method, we compared the results of C. difficile detection in stool specimens by qPCR with those by CDSC as a reference method. Stool specimens were collected from 82 elderly people in four different nursing homes once every 3 months for 6 months, i.e. a total of three times (S 1 , S 2 , and S 3 sampling). A total of 235 specimens were examined for C. difficile by qPCR and CDSC (Table 4). C. difficile was detected in 12 of the 235 specimens by qPCR, whereas this organism was isolated by CDSC from 9 specimens; there were thus three discrepancies between the methods. C. difficile toxins were not detected in any of the 235 specimens by EIA (data not shown).
Details of the test results from the 12 C. difficile-positive specimens from eight subjects are shown in Table 5. Whole C. difficile counts (16S rRNA target) ranged from 10 3 to 10 5 cells level/g of stool, with the exception of one specimen, #02010-S 3 , which had a much higher count. The mean qPCR counts of whole C. difficile, TcdB-producing strains, and TcdA-producing strains were 4.561.3, 4.660.4, and 4.560.6 log 10 cells/g of stool, respectively. On the basis of the qPCR counts for the three genes within each specimen, the toxigenic type of the predominating C. difficile was identified as A + B + in six specimens (#02007-S 2 , #03008-S 1 , #03008-S 2 , #03008-S 3 , #03024-S 2 , #04003-S 3 ) and as A 2 B 2 type in the other six specimens (#02010-S 3 , #02011-S 3 , #04011-S 2 , #04026-S 1 , #04026-S 2 , #04026-S 3 ). The PCR analysis on CDSC revealed that the isolates from six specimens were of A + B + type and those from the remaining three specimens were A 2 B 2 type. The A + B + isolates from the six specimens were confirmed by cell cytotoxicity assay to be capable of producing TcdB, although the toxin was not detected in any of the stool specimens by EIA. The toxin-production profiles of these isolates, as determined by CDSC analysis, were consistent with those determined by qPCR. In terms of the results of toxigenic C. difficile detection, both qPCR and CDSC gave the same six positive specimens, indicating that qPCR was as efficient in detecting toxigenic C. difficile in stools as CDSC.

C. difficile carriage in the elderly in nursing homes
We obtained detection rates and qPCR counts of whole C. difficile, TcdB-producing strains, and TcdA-producing strains in elderly residents of different nursing homes at three samplings (S 1 :82 specimens; S 2 :79 specimens; and S 3 :74 specimens) (See Table S2). Detection rates of whole C. difficile in S 1 , S 2 , and S 3 specimens from the four nursing homes were 2/82 (2.4%), 5/79 (6.3%), and 5/74 (6.8%), respectively. The detection rates of TcdA-producing strains were 1/82 (1.2%), 3/79 (3.8%), and 2/72 (2.7%), respectively, and those of TcdB-producing strains were exactly the same. On the basis of the qPCR counts, we determined the toxigenic types of the predominating C. difficile in each specimen; the rates of carriage of the respective types by the 82 subjects are shown in Figure 2. Whereas site 01 had no C. difficile carrier throughout the 6-month period, the other sites had some carriers of either the A + B + strain or the A 2 B 2 strain, or both. The overall carriage rate of C. difficile in the 82 subjects fluctuated below 10% during the 6-month test period, except in one case (S 3 at site 02).
As a result of review of medical records, 19 (23%) of 82 subjects received at least one course of antibiotic treatment in the study period of 6 months. C. difficile was detected in only one of the 19 subjects (#02010). The number of subjects who had hospitalization during the study period was six (7.3%), only one of which, the same #02010, was C. difficile-positive. The subject underwent several antibiotic treatments and short-term hospitalizations between S 2 and S 3 sampling.

Discussion
We designed new oligonucleotide sets targeting the 16S rRNA, tcdB, and tcdA genes for the quantification of whole C. difficile (A + B + , A 2 B + , and A 2 B 2 types), TcdB-producing strains (A + B + and A 2 B + types), and TcdA-producing strains (A + B + type), respectively. Most clinically isolated TcdA-negative, TcdB-positive strains (A 2 B + type) possess tcdA that has a 1.8-kbp deletion in 8.0 kbp of the intact full length sequence [16][17][18][19], and therefore does not produce functional TcdA. We designed tcdA-specific primers and a probe against this deleted region, and confirmed that our PCR correctly identified intact tcdA genes and did not detect tcdA genes that were rendered non-functional due to the deletion (Table 2). Moreover, these newly developed primersprobe sets were designed against the highly conserved region of each gene and consequently included no mismatches with their target sequences. Indeed, qPCR was capable of detecting all the strains in the target group with almost equal reactivity (Table 3). Up to the present, many sets of primers and probes targeting toxin genes have been reported [6,[20][21][22]. However, in silico comparison of those oligonucleotides with their target gene sequences currently available in the database revealed that most of them included at least a few mismatches against some of their targets (data not shown); this may result in inaccurate quantification due to strain-dependent differences in reactivity. It can be expected that our qPCR method overcame this potential defect.
The lower detection limit of C. difficile in stools remained to be solved. We previously reported that reverse transcription (RT)-qPCR targeting the rRNA molecule achieved sensitive detection of C. difficile with a lower detection limit of 10 2.4 cells/g of stool [23]. However, this application was limited to C. difficile species because such a low detection limit was accomplished only by targeting rRNA molecules present in high copy numbers (10 3 to 10 4 molecules per single cell) [24,25]. Here, we developed a detection system for toxigenic strains that was based on qPCR targeting the toxin genes. To lower the limit of detection by qPCR, we subjected a highly concentrated stool DNA fraction to PCR in the presence of an agent that neutralized PCR amplification inhibitors. In general, extraction of DNA from larger amounts of stool and dissolution in smaller amounts of buffer can yield highly concentrated DNA, but the product will contain increased amounts of stool-derived PCR inhibitors. In our study, a stool DNA fraction that was prepared by using a commercial DNA extraction kit (FastDNA SPIN Kit for Feces) and contained DNA from as much as 10 mg of stool was subjected to qPCR without dilution because the commercial PCR buffer (Ampdirect Plus) neutralized the effect of the PCR inhibitors. This combination successfully lowered the detection limit to 10 3 cells/g of stool ( Figure 1). We confirmed that an application of the extracted stool DNA solution to PCR without Ampdirect plus gave no amplification (data not shown). The qPCR counts of C. difficile in most subjects fell into a low range (e.g. 10 3 or 10 4 cells/g of stool; Table 5), indicating that the lower detection limit of 10 3 cells/g of stool is important in the in-depth investigation of C. difficile present in stools of asymptomatic subjects.
In our examinations, the consistent result of C. difficile detection was obtained in 232 out of 235 specimens between qPCR and CDSC (Table 4), supporting the validity of our TaqMan PCR as a sensitive method to detect C. difficile in stools. The other three specimens were qPCR positive but CDSC negative. One possible reason of this discrepancy is the difference of C. difficile forms targeted by the methods. C. difficile has typically two forms; one is the active and infectious form, vegetative cells, and the other is the inactive form, spores. In our CDSC examination, C. difficile spores were exclusively detected since alcohol treatment killed vegetative cells. In contrast, we confirmed that our qPCR method mainly detected vegetative cells because the DNA extraction efficiency from spores was approximately 1,000-times as low as that from vegetative cells (data not shown). The discrepant result between qPCR and CDSC in the three specimens may therefore reflect that C. difficile present in the stools was composed of the majority of vegetative cells and the minority of spores, and that a smaller amount of spores were undetectable by CDSC. In our study, CDSC was performed for the purpose of detecting only spores because it was thought to be difficult to meet all the requirements for the detection of vegetative cells by culture, such as collection of fresh stools with less exposure to aerobic environment, more rapid transportation of stool from study sites to our laboratory, and immediate examination. An application of a cultural procedure for the detection of vegetative cells as well as spores could have provided further information. In addition, modification of DNA extraction method to improve the extraction efficiency from spores would make it possible to compare detection results of the same target, both vegetative cells and spores, between qPCR and culture methods. Another possible reason of the discrepancy is an issue of living and dead cells. In general, while a culture method detects only living cells, conventional qPCR detects both living and dead cells, which could result in the higher detection rate of C. difficile by qPCR. Recently, selective detection of living cells by PCR in combination with cell membrane-impermeable dyes, which modify DNA of dead cells and inhibit PCR amplification, has been reported [26,27]. We consider it necessary to apply such living cell-specific detection system to our qPCR method for the detection of authentic living bacterial cells in stools. Other reasons such as false-positive results of qPCR may also lead to discrepancy. Although the primers-probe set was carefully checked to specifically detect C. difficile, it remains a possibility of nonspecific reaction with other unknown bacteria. Despite these differences in the methodological character of qPCR and CDSC, the highly comparable results of C. difficile detection were obtained (Table 4), suggesting the effectiveness of our qPCR targeting vegetative cells as the detection method of C. difficile in stools. The C. difficile carriage rate in long-term care facilities (LTCFs) is considered higher than that in community-dwelling adults but lower than that in hospital inpatients [28]. In our previous study of 83 LTCFs residents in Japan [23], we revealed that the carriage rate of C. difficile was as high as 43%. Riggs et al. [5] reported a 51% carriage rate of toxigenic C. difficile in 68 inpatients of LTCFs in Ireland, as determined by using a culture method. However, in our present study of 82 nursing home residents in France, the carriage rates of C. difficile and toxigenic C. difficile, as determined by TaqMan-based qPCR over 6 months, ranged from 2.4% to 6.8% and 1.2% to 3.8%, respectively-much lower than those reported in the studies above. Nevertheless, the C. difficile carriage rates reported in other culture-based examinations are comparable to our current results [3,4,29,30]. For example, Walker et al. [30] reported that the carriage rates of whole C. difficile and toxigenic C. difficile in 225 LTCF residents in the United States were 7.1% and 4.0%, respectively. In a later study by Arvand et al. [29], these carriage rates in 240 elderly nursing home residents in Germany were 4.6% and 4.2%, respectively. LTCFs include various facilities, such as nursing homes, rehabilitation facilities, inpatient behavioral health facilities, and long-term chronic care hospitals. Because environmental contamination with C. difficile occurs commonly in hospitals [31,32], subjects in facilities close to hospital environments are likely to have more chances to acquire the organism. Certainly, the facilities in our study, which had relatively low C. difficile carriage rates, were nursing homes, whereas those in our previous study, which had higher carriage rates, were chronic care facilities. Thus, difference in the types of LTCFs may explain these variations in C. difficile carriage rates.
Exposure to antibiotics and frequent or prolonged hospitalization are the major risk factors for acquisition or colonization of C. difficile [31,33,34]. In our study, subject #02010, who had several antibiotic treatments and hospital stays between S 2 and S 3 , acquired C. difficile at S 3 , and the qPCR count was over 10 8 cells/ g of stool-much higher than in the other subjects (Table 5). It is likely that overgrowth of C. difficile newly acquired from the hospital environment was observed by the TaqMan-based qPCR, although this was not a case of CDI because the strain was nontoxigenic. All eight of our C. difficile-positive subjects, including the four toxigenic C. difficile carriers, had no abdominal symptoms. The mean qPCR count of C. difficile in these asymptomatic carriers was 4.5 log 10 cells/g of stool (Table 5). Naaber et al. [35] examined stools from patients with antibioticassociated diarrhea by using C. difficile species-level qPCR. They reported increased numbers of C. difficile, ranging from 5.6 to 11.2 log 10 cells/g of stool. They also revealed that the mean qPCR count in C. difficile-toxin-positive stools was higher than that in toxin-negative stools (9.3 vs. 6.3 log 10 cells/g of stool). Riggs et al. [5] also reported that the mean C. difficile count in 18 patients with CDI was higher than that in 20 asymptomatic carriers (5.6 vs. 3.6 CFU/g of stool). Thus, the bacterial number in stools may be useful for predicting the status of C. difficile carriers (i.e. symptomatic or asymptomatic) and the severity of symptoms. Our TaqMan-based qPCR method would be appropriate for such assessments because it enables accurate monitoring of C. difficile counts with an appropriate lower detection limit. We believe that use of this method could provide valuable information for the control of CDI.
In conclusion, we developed a sensitive and selective detection system for C. difficile in human stools that uses TaqMan-based qPCR. Application of qPCR to the examination of stools from nursing home residents revealed in detail the prevalence of C. difficile, including toxigenic strains, indicating that this method can be an effective tool for both clinical diagnosis and epidemiological investigation.

Supporting Information
Table S1 Accession numbers of nucleotide sequences used for the design of primers and probes.