Application of direct ribotyping during an outbreak with C. difficile
During the optimization phase of our direct ribotyping technique, a C. difficile outbreak was suspected in the ICU. In our institution, the standard typing technique for C. difficile is Amplified Fragment Length Polymorphism (AFLP) on cultured strains. The suspected outbreak cluster involved six patients with the same C. difficile AFLP-type. During this outbreak, samples of five other patients became positive for C. difficile by qPCR or toxin enzyme immune assay (EIA). We performed direct ribotyping on total fecal DNA of all eleven patients. The six patients with the same C. difficile AFLP-type had identical ribotype peak profiles (figure 1). In the five other patients that became positive for C. difficile during the outbreak, direct ribotyping enabled us to allocate 1 of the 5 patients to the outbreak-cluster and 4/5 patients outside the outbreak-cluster (figure 1). Importantly, results of direct fecal ribotyping were obtained before strains were cultured and conventionally typed by AFLP.
C. difficile PCR ribotyping and toxin gene detection
After our first experience with direct ribotyping during the outbreak, we aimed to validate our technique with a larger sample set of 130 fecal samples (including the 11 samples collected during the outbreak). With the ribotype primers we amplified DNA of a total of 65 fecal samples that were previously proven to contain C. difficile by qPCR for C. difficile toxin A and/or B genes (the standard diagnostic test for C. difficile detection in our laboratory). Mean Cp value (crossing point at which the amplification curve crosses the vertical threshold line; this is inversely associated with the C. difficile load) of C. difficile toxin gene qPCR was 33 (range 27-40 cycles, Supplementary Table 2). DNA fragment peak profiles were obtained from all 65 fecal samples (3 after 1:5 dilution because of inhibition) and from all 65 cultured strains. Hence, the sensitivity of the new primers set for toxigenic C. difficile detection was 100% (n=65, 95% Confidence Interval (CI) 94.5-100%).
We observed DNA fragment peaks ranging in size from approximately 200 to 590 nucleotides, consistent with published studies when corrected for differences in primer binding sites (11, 12, 14, 15). Also the number of DNA fragments was in agreement with previously described ribotype profiles, and varied between 5 and 10 (11, 12, 14, 15)(10, 11).
To examine the specificity of our primers for C. difficile detection, we applied the primers to total DNA obtained from C. bifermentans and C. sordellii strains and 65 fecal samples with negative qPCR for C. difficile toxin genes. Of these samples, fourteen were positive by diagnostic PCR’s for other bacterial species and viruses that are well-known causes of diarrhea such as Campylobacter spp., Salmonella spp. and norovirus. No DNA fragment peak profiles were detected in these samples, indicating a diagnostic specificity of 100% (n=65, 95% CI 94.5-100%).
To assess reproducibility, DNA isolation and direct ribotyping was performed in duplicate on a subset of 40 fecal samples with a positive qPCR for C. difficile toxin A and/or B genes. DNA fragment peak profiles were observed in 40/40 paired fecal samples. Profiles of 36/40 paired fecal samples were 100% identical (90%). All discrepancies were found in larger DNA fragments (>400) in low load samples (C. difficile toxin A and/or B genes qPCR Cp values 35-39).
To examine possible technical issues of ribotyping directly on feces – for example decreased intensity of DNA fragment peaks due to PCR inhibition or appearance of nonspecific peaks due to an excess of fecal DNA – the peak profile of each fecal sample was compared with that of its corresponding cultured strain, see figure 2 for example. Peak profiles of 61/65 paired fecal samples and strains were completely identical (94%). In 3/65 samples we observed 1 peak difference. These samples had a low bacterial load in qPCR (Cp values 35-39); and it was one of the larger DNA fragment peaks (>400 nucleotides) that was missing. In 1/65 samples we observed that the three largest DNA fragments in the strain profile were missing in the profile of the fecal sample (Cp value 29).
For detection of toxin A (tcdA), toxin B (tcdB) and binary toxin (cdtA, cdtB) genes directly on total fecal DNA we used primers designed by Persson et al. and added these in our study set (figure 3) (17). All C. difficile positive fecal samples showed at least one toxin gene peak, whereas no peaks were observed in the C. difficile negative fecal samples. The presence of toxin genes specific for different ribotypes was consistent with literature (11, 19-21). In one sample with RT190, toxin A, B and binary toxin B genes were detected but not binary toxin A gene. This could be due to non-specificity of our assay; however, C. difficile strains with presence of binary toxin B but not binary toxin A gene have been described (22, 23). Also, we detected both toxin A and B gene peaks in RT017 samples, while this ribotype is known to produce only toxin B (24-26). This was observed and clarified previously by Persson et al.: “The primers used to amplify toxin A gene are located upstream of the repetitive region in the 3′-end which, in some strains, contains various deletions that render the gene product non-detectable by EIA methods. Therefore, strains that are TcdA-negative due to 3′-end deletions are still tcdA-positive according to the present multiplex PCR.” (17).
Reference ribotypes obtained by conventional ribotyping of strains
Conventional ribotyping of all 65 C. difficile strains that were cultured from the 65 fecal samples was performed by the Dutch National Reference Laboratory. These ribotyping results served as reference. The Reference Laboratory could not determine the ribotype of 2/65 strains due to unknown or absent band patterns. A ‘probable ribotype’ was determined in 5/65 strains since the band patterns of these strains were highly similar to patterns of reference strains except for a 1 or 2 bands difference. Overall, 63/65 strains of our study set were assigned to 27 different reference ribotypes.
Clustering of fecal samples based on peak profile similarity
We assessed if direct ribotyping on fecal samples was feasible as first screening tool for detection of a clonally related C. difficile cluster by performing cluster analysis based on ribosomal DNA fragment profile similarity. A heat map and dendrogram were created based on peak profiles of all 65 fecal samples with positive qPCR for C. difficile toxin A/B genes (figure 2). The resulting clusters consisted of fecal samples containing the same C. difficile ribotypes as determined by the Reference Laboratory (for example, one cluster consisted of four fecal samples that all contained RT002), except for two samples: one with RT002 and one with RT050. The ribotyping patterns in both samples lacked the larger DNA fragment peaks when compared to profiles of samples with the same reference ribotype.
In conventional ribotyping, a pattern with a single band difference is usually considered as a different ribotype. Using this definition, we assessed the performance of direct ribotyping on feces for ribotype assignment by comparing peak profiles of samples with the same reference ribotype. We observed identical peak profiles in 42/48 (90%) fecal samples containing identical ribotypes (RT001: 4 out of 5 profiles were identical, RT002:
n=3/4, RT012: n=3/3, RT014: n=4/4, RT015: 2/2, RT017: n=10/10, RT026: n=5/5, RT050 n=0/2, RT078: n=5/6, RT190: n=2/2, RT258: n=3/3, RT626: n=2/2).