Clin Microbiol Infect 2011; 17: 166–175
Clostridium difficile infection is most often induced by antibiotic treatment. Recently, morbidity and mortality resulting especially from C. difficile PCR ribotype 027 have increased significantly. In addition, more severe disease has been associated with C. difficile PCR ribotype 078 strains. Thus, reliable typing methods for epidemic control are needed. In the present study, we compared an automated repetitive extragenic palindromic sequence-based PCR (rep-PCR) method (DiversiLab; Bacterial Barcodes, Inc., Athens, GA, USA) to PCR ribotyping and pulsed-field gel electrophoresis (PFGE) typing using 205 isolates of C. difficile (including 24 previously characterized isolates). Among the 181 clinical isolates, a total of 31 different PCR ribotypes, 38 different PFGE types and subtypes and 28 different rep-PCR types were found. Six major rep-PCR groups (DL1–DL6) harboured 86% of the clinical isolates. All isolates belonging to PCR ribotypes 027 and 001 clustered in their own rep-PCR groups, enabling us to screen out the hypervirulent ribotype 027 strain. Within the PCR ribotype 001, four subgroups were found using rep-PCR. Overall, in 75% (135/181) of the isolates, the classification attributed following rep-PCR and PCR ribotyping was comparable. In conclusion, the automated rep-PCR-based typing method represents an option for first-line molecular typing in local clinical microbiology laboratories. The method was easy to use as well as rapid, requiring less hands-on time than PCR ribotyping or PFGE typing. The conventional PCR ribotyping or PFGE, however, are needed for confirmatory molecular epidemiology. In addition, more epidemiology-oriented studies are needed to examine the discriminatory power of automated rep-PCR with isolates collected from a larger geographical area and during a longer period of time.
Clostridium difficile is the main causative agent of antibiotic-associated diarrhoea. C. difficile infection (CDI) can present as severe disease, and particularly the PCR ribotype 027 has been recently associated with increased virulence [1–3]. The major risk factors for C. difficile infection include high age, hospitalization, immune-compromising conditions and exposure to certain antimicrobial agents, especially fluoroquinolones and cephalosporines. The C. difficile PCR ribotype 027 strain has been found in both hospital and community environments, as well as in soil, rivers, lakes and meat products [4,5]. It is commonly associated with outbreaks with increased morbidity and mortality . The epidemic-associated C. difficile 027 strain has a 18-bp deletion in the tcdC gene, the negative regulator of toxin production , and produces the binary toxin (CDT) [1–3,8]. Interestingly, McCannell et al.  described a one-nucleotide deletion in the tcdC regulator gene of C. difficile 027 leading to a premature stop codon; the authors suggested the production of a truncated TcdC protein as a result of this deletion. The excessive toxin production characteristic for C. difficile 027 is considered to be a result of this defect in TcdC, with the normal inhibition of toxin production before the stationary phase being missing in these strains .
There are also deletions larger than 18 bp in the tcdC amplification products. Such deletions have been described earlier by Stare et al. . These large deletions do not lead to truncations but result in amino acid deletions in the putative protein products. Curry et al. reported  that the tcdC variants with these larger deletions also contain stop codons comparable to the one detected in C. difficile 027, and thus these variants may be potential hyperpoducers of toxins A and B. Recently, Goorhuis et al.  reported that C. difficile PCR ribotype 078 can also cause more severe CDI and increased mortality. This ribotype has a 39-bp deletion as well as a stop codon leading to a mutation in its tcdC gene. Therefore, we chose to include also isolates possessing these characteristics in the present study.
Healthcare institutions require accurate and rapid diagnosis for the early detection of possible outbreaks . Strain typing can help with the infection control procedures by tracking the source and spread of C. difficile infections. Several molecular typing methods, such as PCR ribotyping, pulsed-field gel electrophoresis (PFGE) and multilocus variable-number of tandem repeat analysis are used as tools with C. difficile . PCR ribotyping is the reference standard in typing C. difficile in Europe and PFGE is the reference standard in the USA. Recently, an automated DiversiLab system (Bacterial Barcodes, Inc., Athens, GA, USA) using repetitive extragenic palindromic sequence-based PCR (rep-PCR) has become available. The rep-PCR method uses primers that target noncoding repetitive sequences interspersed throughout the bacterial genome [15,16]. The amplified DNA fragments, when separated by electrophoresis, constitute a genomic fingerprint that can be used for bacterial subspecies discrimination and strain delineation . The commercial semi-automated rep-PCR device, DiversiLab, offers better standardization and greater reproducibility than manual, gel-based rep-PCR .
In the present study, we assessed the usefulness of the DiversiLab semi-automated rep-PCR for typing C. difficile and compared these results with those obtained with PCR ribotyping and PFGE.
Materials and Methods
Bacterial strains and culture conditions
A total of 205 C. difficile isolates from THL (National institute of Health and Welfare, former KTL) (99 consecutive isolates collected from September 2007 to March 2008 from hospitals in different regions of Finland) and from HUSLAB (Laboratory of Helsinki University Central Hospital) (82 consecutive isolates collected from 1–18 January, 2008), and 24 isolates (collected from 5 November 2007 to 17 March 2008) with large (39 bp and 54 bp) deletions in tcdC were studied. PCR analyses detecting virulence genes and screening for tcdC gene deletions were performed at HUSLAB  and THL. Helsinki University Hospital is responsible for the secondary and tertiary care of approximately 1.5 million people. The culture samples from this area received by HUSLAB are both from the hospitals as well as from outpatients of this geographical area, the Helsinki and Uusimaa district in southern Finland. The culture samples of HUSLAB analyzed in the present study were from patients treated in 15 different hospitals, as well as from patients in 16 different outpatient clinics. The isolates with large deletions were chosen subsequent to routine screening at the laboratory .
C. difficile strains were cultured in anaerobic atmosphere on CCFA plates (cycloserine–cefoxitin–fructose–egg yolk agar) and incubated at 35°C for 42 h. Colonies with typical morphology, fluorescence and odour were identified as C. difficile.
DNA was extracted from colonies on CCFA plates using the UltraClean microbial DNA isolation kit (Mo Bio Laboratories, Solona Beach, CA, USA) and diluted to 25–50 ng/μL. The DNA was amplified using the DiversiLab Clostridium kit (catologue no DL-CD01; Bacterial Barcodes, Inc.) for DNA fingerprinting in accordance with the manufacturer’s instructions; 2 μL of genomic DNA, 18 μL of the rep-PCR master mix, 2 μL of primer mix provided in the kit, 0.5 μL of AmpliTaq polymerase and 2.5 μL of 10× PCR buffer (Applied Biosystems Roche, Branchburg, NJ, USA) were added for a total of 25 μL per reaction. PCR was run on a preheated thermal cycler (DNA Engine Tetrad 2; Peltier Thermal Cycler BioRad, Hercules, CA, USA). The thermal cycling parameters were: initial denaturation at 94°C for 2 min, followed by 35 cycles of denaturation at 95°C for 30 s, annealing at 60°C for 30 s, extension at 70°C for 1.5 min, and final extension at 70°C for 3 min. The specific positive and negative controls included in the kit were run with each reaction set for the validation of amplification. The rep-PCR products were detected and the amplicons separated using microfluidics lab-on-a-chip technology and analyzed using the DiversiLab system (Bacterial Barcodes, Inc.). Further analysis was performed with the web-based DiversiLab software (version 3.4) using the band-based modified Kullback–Leibler distance for the calculation of percent similarities. This calculation places more weight on band presence than on intensity variations. The automatically generated dendograms, similarity matrices, electropherograms, virtual gel images, scatter plots and selectable demographic fields were also used in the interpretation.
The manufacturer provides guidelines for strain-level discrimination; a similarity of more than 97% is considered as indistinguishable (no differences in fingerprints); a similarity of more than 95% is considered as similar (one or two band difference in fingerprints); and a similarity of <95% is considered as different. In the present study, 90% similarity was taken as the optimal cut-off for clustering.
tcdC gene sequencing
Whole-gene sequencing of tcdC was performed using isolates with deletions >18 bp. The entire tcdC gene was amplified for sequencing as described by Spigaglia and Mastrantonio .
PCR ribotyping was performed according to the protocol of the Anaerobe Reference Laboratory, Cardiff, UK , using the Cardiff-ECDC culture collection as a set of reference strains. After gel electrophoresis, the band patterns were analyzed using the BioNumerics software, version 5.0 (Applied Maths NV, Sint-Martens-Latem, Belgium). If the gel pattern of an isolate did not match any of the reference strain patterns, it was given an intralaboratory ribotype name (unknown 1, 2, 3, etc.).
PFGE was performed as described previously , with slight modification. Separation was performed in 1.1% Seakem Gold agarose in 0.5×TBE (0.045 M Tris-borate, 0.001 M EDTA) with thiourea (200 μM) in a CHEF-DRIII system (Bio-Rad Laboratories, Richmond, CA, USA). Running conditions consisted of one block with switching times of 5 and 55 s, and a running time of 21 h. The lambda ladder was used as DNA size standard and the Cardiff-ECDC culture collection type strains of C. difficile PCR ribotype 027 and 001 as controls. PFGE patterns were analyzed with BioNumerics (version 5.0) using the Dice coefficient to analyze the similarity of the banding patterns, and the unweighted pair group method using arithmetic averages (UPGMA) for cluster analysis. We used a ≥ 80% relatedness to define the lineages .
Rep-PCR fingerprints were generated and PCR ribotyping was performed for 181 clinical isolates (Fig. 1). The most prevalent PCR ribotypes were 027 (n = 64, 31%) and 001 (n = 54, 26%) (Table 1). The rest of the isolates represented a total of 29 different PCR ribotypes. With rep-PCR analysis, a total of 28 different profile groups were found when >90% similarity was used as threshold (Fig. 1, Table 1). Six major rep-PCR groups were detected (Fig. 1). In 75% (135/181) of the isolates, the classification obtained by rep-PCR and PCR ribotyping (PCR ribotypes 027, 001, 023, 078 and unknowns 3, 9, 19, 22, 27, 30, 37, 53, 64, 66) were comparable (Fig. 1, Table 1).
|PCR ribotype||Number of isolates||Rep-PCRa (DLb)||PFGE|
|027||64||DL 1||NAP1, sub. 1 and 2|
|001||54||DL 2 sub.1,2,3,4||Unique 3 and unique 3 sub. 1|
|023||4||DL 5||Unique 14 sub. 1 and 2|
|078||2||DL 9||Unique 11 sub. 1 and unique 13 sub. 2|
|005||2||DL 18, DL 28||Unique 5 sub. 2|
|Unknown 3||1||DL 15||Unique 22 sub. 1|
|020||8||DL 3, DL 6, DL 11||Unique 1, unique 1 sub. 4 and 5|
|002||6||DL 3, DL 4, DL 6, DL 25||Unique 21, unique 21 sub.1, 2 and 4|
|012||2||DL 7, DL 16||Unique 1 sub. 3 and ND|
|Unknown 10||1||DL 3||ND|
|Unknown 11||1||DL 3||ND|
|014||11||DL 3, DL 6, DL 11, DL 12, DL 21, DL 22||Unique 1, unique 1 sub. 2, unique 15, unique 15 sub. 3 and ND|
|056||2||DL 4, DL 8||Unique 15 sub. 4 and ND|
|003||1||DL 4||Unique 7|
|018||3||DL 3, DL 8||Unique 6 and unique 6 sub. 1|
|Unknown 19||1||DL 19||ND|
|010||1||DL 4||Unique 3|
|Unknown 22||2||DL 10||Unique 5 sub. 1 and 3|
|029||1||DL 26||Unique 5 sub. 3|
|Unknown 30||1||DL 20||Unique 29|
|070||2||DL 4, DL 13||Unique 7|
|011||1||DL 23||Unique 1 sub. 1|
|Unknown 44||1||DL 1||NAP1 sub. 1|
|Unknown 52||2||DL 4, DL 7||Unique1 sub. 4 and unique-28|
|Unknown 53||1||DL 27||Unique 3|
|Unknown 55||1||DL 5||Unique 14 sub. 1|
|Unknown 63||1||DL 3||Unique 13 sub. 3|
|Unknown 64||1||DL 14||Unique 1|
|Unknown 66||1||DL 24||Unique 1 sub. 4|
|Unknown 68||1||DL 7||Unique 18 sub. 1|
In PCR ribotype 027, a similarity of >90% was detected by rep-PCR (data not shown). Within the PCR ribotype 001, four rep-PCR subgroups and one outlier strain were detected. Most of the PCR ribotype 001 strains (63%) belonged to rep-PCR DL 2 sub. 1 group. The rep-PCR DL 2 sub. 2 group harboured 20%, whereas DL 2 sub. 3 and sub. 4 groups harboured both 7% of the isolates (Fig. 2). Both of the main PCR ribotypes, 027 and 001, clustered in their own rep-PCR groups when all the 181 clinical isolates were compared with each other (data not shown). In addition, PCR ribotype 027 and 001 both had a typical well identifiable electropherogram (Fig. 1).
In 25% of the isolates, there were inconsistencies between the PCR ribotyping and rep-PCR results (Figs 1 and 3, Table 1). Instead, we found at least three cases where rep-PCR and PCR ribotyping did not correlate, but PFGE supported the rep-PCR clustering.
First, an isolate representing PCR ribotype ‘unknown 44’ (one band difference vs. PCR ribotype 027), PFGE NAP 1 subtype 1 and with an 18 bp tcdC deletion clustered into the same group with rep-PCR as isolates with PCR ribotype 027. Second, an isolate with PCR ribotype ‘unknown 53’ (one band difference vs. PCR ribotype 001) and PFGE type unique 3 clustered according to rep-PCR close to the group harbouring all other isolates of PCR ribotype 001 and PFGE type unique 3. Third, an isolate with PCR ribotype ‘unknown 55’ and PFGE type unique 14 sub. 2 clustered into the same rep-PCR group with PCR ribotype 023 and this isolate had a 54-bp deletion in tcdC as did the PCR ribotype 023 isolates. As for the rest of the isolates, definite conclusions cannot be made; only few representatives of each PCR ribotype, PFGE type or DL type were found.
None of the three methods was coherently more discriminative than the other two. Among the PCR ribotypes 002, 005, 012, 014, 018, 020, 056, 070 and the ‘unknown 52’, the rep-PCR was more discriminative than PCR ribotyping. As for the PCR ribotypes 014 and the ‘unknown 52’, PFGE was also more discriminative than PCR ribotyping. This is shown in Fig. 3 for ribotypes 002, 014 and 020. In addition, the largest rep-PCR group (DL 3), after the two harbouring PCR ribotypes 001 and 027, included isolates from several PCR ribotypes (002, 014, 018, 020, ‘unknown 10’, ‘unknown 11’ and ‘unknown 63’) and PFGE types (Fig. 1). Over half (52%) of the PCR ribotype 002, 014 and 020 isolates were grouped into the rep-PCR group DL 3. The similarity percentages determined with rep-PCR varied between 7.7% and 99.2% among the isolates belonging to PCR ribotypes 002, 014 and 020.
tcdC-A and tcdC-A variants
In addition, isolates with large tcdC deletions (n = 24) were studied (sequencing results not shown). The tcdC-A gene (classification as in Curry et al. ) has a 39-bp deletion (nucleotides 341–367) and the tcdC-A variant (tcdC-Avar) had a 54-bp deletion in the same location as tcdC-A (nucleotides 313–367) . Of the isolates possessing tcdC-A, three different PCR ribotypes, 045, 078 and 126, were detected. These ribotypes, however, fell into two different rep-PCR groups with four additional outliers (Fig. 4). With these isolates, PFGE supported neither PCR ribotyping, nor rep-PCR but led to a third delineation of isolates. All the isolates having a tcdC-Avar allele were of PCR ribotype 023 with one exception (PCR ribotype ‘unknown 55’). Using rep-PCR, this particular isolate was indistinguishable from the main clone harbouring 11 out of 14 isolates of this population (Figs 1 and 4). Using PFGE, these tcdC-Avar isolates represented the PFGE type ‘unique 14’ and its three subtypes (Fig. 4).
Reproducibility of rep-PCR
The effect of DNA isolation and PCR conditions, and technician skills on the electropherogram for each isolate type was studied with a set of isolates including PCR ribotypes 027, 001, 078 and ‘unknown 55’ (Fig. 5).
The DNA isolation step had to be carried out optimally to obtain reproducible results. Independent of this, with some isolates, the freezing-melting cycle of the template DNA between separate PCRs appeared to result in somewhat distinct rep-PCR profiles.
We studied a set of clinical C. difficile isolates by PCR ribotyping, PFGE and rep-PCR aiming to assess the utility of the automated rep-PCR for typing. The correlation of rep-PCR, PCR ribotyping and PFGE was excellent with the two major groups of isolates, PCR ribotypes 027 and 001. Among other isolates, the grouping obtained with these three methods was less coherent.
The automated rep-PCR method proved to be easy to use with good reproducibility. A set of 13 samples could be analyzed within a single day. The current European reference method, PCR ribotyping, can be performed in 2 days for a set of 17 samples ; using capillary gel electrophoresis-based PCR ribotyping, at least 90 isolates can be typed per diem . By contrast, PFGE is considered a labour-intensive, time-consuming method that yields optimal results when performed by a technician with extensive experience. In practice, the rep-PCR profiles were easily stored in the web-based library and conveniently analyzed by the software of the system. We found this especially beneficial in everyday diagnostic work; the profiles of isolates analyzed at separate timepoints can be archived in the library and compared at chosen timepoints. We found the library generated by us to be more useful than the one provided by the manufacturer, which does not allow direct PCR ribotype or PFGE type vs. rep-PCR profile comparison.
The rapid clonal spread of the hypervirulent PCR ribotype 027 strain has created the need for rapid detection of isolates belonging to this clone. The PCR ribotype 027 strains were clearly separated into one cluster using rep-PCR. Thus, the rep-PCR method can be used for screening hypervirulent PCR ribotype 027 strains for which a typical electropherogram was detected. The two other potentially hypervirulent toxin hyperproducers (1-bp deletion leading to stop codon in tcdC) (e.g. PCR ribotypes 023 and 078) were also found to cluster in their own rep-PCR groups.
Within the PCR ribotype 001, rep-PCR was more discriminatory than ribotyping, differentiating four subgroups (Fig. 2). This phenomenon has also been reported by others who used in-house rep-PCR [24,25]. Healy et al. (19th European Congress of Clinical Microbiology and Infectious Diseases, abstract P1731) have also tested the DiversiLab-system in the typing of C. difficile. In their study, in addition to PCR ribotype 001, types 002, 027, 053, 078 and 106 also were divided into several different rep-PCR profiles. The most typical rep-PCR profile of PCR ribotype 001 isolates in the present study had only four to five peaks in the electropherogram. Differences among these isolates could be found mainly as a result of differences in the intensity of the peaks.
Healy et al. (19th European Congress of Clinical Microbiology and Infectious Diseases, abstract P1731) compared rep-PCR, PCR ribotyping and PFGE in the analysis of 73 isolates from the UK and the USA. By contrast to the results obtained in the present study, they found four different rep-PCR groups among the PCR ribotype 027 strains. Moreover, their PCR ribotype 023 isolate was identical to four PCR ribotype 027 isolates in rep-PCR. However, the electropherogram of this isolate appears to correspond quite closely to the rep-PCR-profiles of our PCR ribotype 023 strains. The discrepancies between our results and the results of Healy et al. (19th European Congress of Clinical Microbiology and Infectious Diseases, abstract P1731) may be a result of true differences between the strains circulating in the three geographic areas of the UK, the USA and Finland). The emergence of PCR ribotype 027 in Finland is rather recent. The isolates of this type from Finland are most likely clonal, and variation over time has not yet occured. In addition, Healy et al. defined the rep-PCR groups manually, whereas we used the 90% cut-off value in addition to the manual comparison.
By contrast, grouping of some ribotypes into one rep-PCR group could be detected (PCR ribotypes 014, 002 and 020). These ribotypes differ from one another by only one band. In our material 14% (25/181) of the strains presented these three PCR ribotypes and most of them (52%) belonged to the third largest rep-PCR group, DL 3. Interestingly, PCR ribotypes 014 and 002 are common PCR ribotypes among clinical isolates in Europe . However, to our knowledge, these types of isolates have not been reported to cause more severe disease.
In the present study, we concentrated on comparing automated rep-PCR, PCR ribotyping and PFGE in a compact set of consecutive samples without consideration of the associated epidemiological data. The samples were collected from one European country during a limited time period. The usefulness of the rep-PCR with isolates collected from a larger geographical area, or during a longer period of time remains yet to be determined.
In conclusion, we found that DiversiLab, the rapid typing method evaluated in the present study, is a useful tool for clinical laboratories in locally monitoring the spread of C. difficile (e.g. in hospital wards). It may even be adequate for a local outbreak investigation when all the analyses are performed in one laboratory. However, the results obtained with the three methods analyzed showed discrepancies, and therefore it is advisable to use multiple methods with a carefully focused purpose for each of them.
This study was presented in part at the 19th European Congress of Clinical Microbiology and Infectious Diseases (ECCMID), Helsinki, Finland, 17 May 2009 (O147).
All authors declare that they have no conflict of interest (commercial or otherwise).