|Home | About | Journals | Submit | Contact Us | Français|
Toxinotyping is a PCR-restriction fragment length polymorphism (RFLP)-based method for differentiation of Clostridium difficile strains according to the changes in the pathogenicity locus (PaLoc), a region coding for toxins A and B. Toxinotypes are a heterogenous group of strains that are important in the development of molecular diagnostic tests and vaccines and are a good basis for C. difficile phylogenetic studies. Here we describe an overview of the 34 currently known toxinotypes (I to XXXIV) and some changes in nomenclature.
Clostridium difficile causes intestinal diseases in humans and animals and is currently among the most important health care-associated multidrug-resistant pathogens. Disease is toxin mediated, and three toxins are produced by C. difficile. Toxin A and toxin B belong to the large clostridial cytotoxins, and the third one, designated binary toxin (CDT), belongs to the clostridial binary toxins (1, 2). Despite the fact that CDT clearly contributes to virulence, toxins A and B are still recognized as the main virulence factors (2, 3). Hence, both these toxins are targets for laboratory diagnostic tests and for vaccine development (4). The variability in the genes coding for toxins A (tcdA) and B (tcdB) is therefore of significant practical importance. This variability can be assessed by C. difficile toxinotyping, a PCR-restriction fragment length polymorphism (RFLP)-based typing method distributing strains into toxinotypes. C. difficile toxinotypes are a group of strains that have changes in the toxin A and B coding region, known as the PaLoc (pathogenicity locus), compared to a reference strain, VPI 10463, in which the toxin genes were sequenced for the first time. Strains similar to VPI 10463 are defined as nonvariant strains of toxinotype 0. All strains with changes in genes tcdA and tcdB are defined as variant strains and are distributed into 34 toxinotypes designated with roman numerals (I to XXXIV).
The method was first described in 1998 (5) and reviewed again a decade later (6). Practical aspects of toxinotype determination have also been updated (7). Here we describe some recent changes in the toxinotyping scheme and give a summary of currently known C. difficile toxinotypes.
The PaLoc contains five genes (Fig. 1); two of them encode two large protein toxins (tcdA for toxin A and tcdB for toxin B). Three additional genes are involved in regulation (tcdR and tcdC) and putatively in transport of the toxin (tcdE). In the majority of strains, the PaLoc is inserted in the same chromosomal site, with the only exception currently described as toxinotype XXXII (8). In nontoxigenic strains, a short, 115-bp sequence is typically found in the insertion site instead of the PaLoc (9). Recently, 75-bp and 7.2-kb sequences at the PaLoc insertion site have also been described (10, 11). Horizontal gene transfer of the PaLoc has been previously shown (12), but the type of the mobile element or transfer mechanism has not been reported so far.
On the basis of two groups of unusual (A-negative [A−] B-positive [B+]) C. difficile strains (13,–15), we developed a set of 10 PCRs covering the entire PaLoc to detect variability in this region (16). These PCRs were then used to screen a large C. difficile collection from Michele Delmee's laboratory (UCL, Brussels, Belgium) containing several hundred serogrouped strains (5). During that study, further variations of the PaLoc, some with only a single representative, others with many, were found. Initially, all six toxin-covering PCR fragments were amplified and analyzed for RFLP in each strain, which was time-consuming. It soon became obvious that the B1 and A3 fragments were those that allowed detection and differentiation of all variant strains (strains with changes in toxin genes). Hence, these two regions were selected as markers for differentiation of variant C. difficile strains in a method termed toxinotyping (5).
The changes observed in the toxin genes are RFLPs, deletions, and insertions. RFLPs are typically present in the tcdB gene and deletions in tcdA (Fig. 1; see also Table S1 in the supplemental material). Deletions in tcdA are largely observed in the receptor binding repetitive regions and could be explained by events of recombination between short repetitive sequences, which are highly conserved in tcdA but not in tcdB. Restriction sites detected as RFLPs are much more variable in the tcdB gene than in the tcdA gene for almost any restriction enzyme tested, suggesting a high frequency of point mutations in tcdB but not in tcdA. Whole-genome sequencing (WGS) studies are confirming this observation (10; S. Janezic, unpublished). Point mutations are mostly located in the catalytic region in some toxinotypes and in receptor binding regions in other toxinotypes, and they are distributed throughout the tcdB gene in the third group of toxinotypes. No mechanism is known or suggested to exist that would lead to an increased frequency of point mutations and increased variability in only one of the two very homologous genes (tcdA and tcdB), which are located close to one another in the chromosome. Insertions are present mainly in the catalytic part of tcdA and are due to the presence of a large mobile element, an ISTron, which is spliced at the mRNA level and hence does not affect toxin production (17) (Fig. 1).
In the study that produced the first description of toxinotypes, the modular structure of the toxin genes was noticed (5). Some variant toxin genes are a combination of a catalytic domain identical to one toxinotype and a binding domain identical to other toxinotype. With several toxinotypes that were described later, this modular structure was even more obvious (see Fig. 1 in reference 18; see also reference 19) and was confirmed with sequencing data (20). Such a structure suggests recombination events requiring the presence of two copies of toxin genes in a single cell. As mentioned before, the mobility of the PaLoc, albeit only from a toxinogenic to a nontoxinogenic strain, was proven. No strain with more than a single PaLoc copy has been described so far.
The first 10 toxinotypes (I to X) were described in the original toxinotyping paper (5). As toxinotyping correlated well with serogroups, the next obvious question was about the correlation with PCR ribotypes, another typing scheme that had just been developed (21). Toxinotyping was applied to a large C. difficile strain collection from J. Brazier's laboratory (Anaerobe Reference Unit, Cardiff, United Kingdom) containing 101 PCR ribotypes, and an additional 5 toxinotypes (XI to XV) were described (19). For toxinotype XI, which was missing a large part of the PaLoc (Fig. 1), a detailed analysis of the chromosomal insertion site was done also (22). Further variant strains were found while typing strains isolated in Japan, Indonesia, and Korea (XVI to XX) (23), and two (XXI and XXII) were found in a study on the distribution of binary toxin-positive strains in a single hospital in the United States (18). From then on, new toxinotypes were added to a database (http://www.mf.uni-mb.si/tox/) but were not regularly published. Toxinotypes XXIII and XXIV were briefly described as new types (24), and types XXV to XXIX and types XXXIII and XXXIV were not included in any previous publication and are described here for the first time. Two new A− B+ toxinotypes from Australia, XXX and XXXI, have been reported (25), while another A− B+ toxinotype, XXXII, was published recently as the only known strain with a PaLoc located outside the usual chromosomal integration site (8).
A C. difficile toxinotype was defined initially as a group of strains that had identical changes in the PaLoc region and could be differentiated from nonvariant strains (toxinotype 0; named according to the type most prevalent at the time and to strain VPI 10463, in which the toxin genes were first sequenced) (5, 7).
To assign the toxinotype in each strain, PCR fragments B1 and A3 are amplified. B1 RFLP types (types 1 to 11) are determined according to the HincII/AccI restriction profile, and A3 RFLP types (types 1 to 16) are determined according to deletions/insertions and the EcoRI restriction profile. The combination of B1 and A3 then determines the toxinotype (Table 1), with each toxinotype having a unique combination of B1 and A3 (7). Toxinotypes with the same B1 and A3 combination but with differences elsewhere would be further differentiated into subtypes (designated a, b, c, etc.). However, this was not the case in the previous scheme; some toxinotypes shared the same type of B1 and A3 and should have been differentiated further with additional PCRs. In the updated toxinotyping scheme, we have largely corrected these ambiguities by renaming several toxinotypes (Table 1; see also Table S1 in the supplemental material).
The previous toxinotype, XXIV, was identical to toxinotype 0 in all six toxin-covering PCRs and differed from it only in the tcdC gene and the presence of the binary toxin gene—a property otherwise not found in strains with nonvariant tcdA and tcdB genes (24). This toxinotype was discovered during typing of strains from Kuwait but was reported as toxinotype 0 (PCR ribotype 131) (26), although it was described as toxinotype XXIV in an earlier publication (24). In the updated scheme, we have renamed toxinotype XXIV as a subtype of toxinotype 0 (0/v) to indicate its similarity to the reference toxinotype but also to highlight the minor changes in the PaLoc (in the tcdC gene) and the presence of the binary toxin gene.
The previous toxinotype, XVII, is very similar to toxinotype X by RFLP analysis (see Table S1 in the supplemental material). The two types group together also in comparisons of entire tcdB sequences (S. Janezic, unpublished). Toxinotype X was therefore renamed Xa and toxinotype XVII renamed Xb. Further differences are described in Table 1.
Changes have also been made in toxinotypes IX and XXIII. Toxinotype IX is now differentiated into subtypes a to d, and toxinotype XXIII was renamed IXc. Toxinotype IXb includes strains of ribotype 244 that were reported in an outbreak of severe disease in Australia (27). The authors typed the strains as toxinotype IX but reported that in silico analysis of WGS data showed differences from the reference toxinotype IX strain suggesting that a new toxinotype should be created for this group of strains.
The last change involves the toxinotypes previously named XIV and XV, now renamed XIVa and XIVb, respectively. They differ from each other in almost all toxin gene fragments except B1 and A3. In the entire tcdB sequence, XIV and XV are the closest neighbors but are not as close as, for instance, Xa and Xb (previously X and XVII). Therefore, some toxinotypes include more closely related subtypes (subtypes Xa and Xb or subtypes XIa to XIc) and the subtypes are probably less closely related in others (subtypes IXa to IXd and subtypes XIVa and XIVb).
In the updated scheme, there are two exceptions to the rule that each toxinotype should have a unique combination of B1 and A3 fragments. The first exception is represented by toxinotype XXXI, which is identical to toxinotype X (a and b) in both B1 and A3. We have decided to keep it as a separate toxinotype for the following reasons: (i) the complete absence of the tcdA gene (Fig. 1) and (ii) the ability to divide the toxinotypes into separate groups on the basis of an analysis of a tcdB dendrogram (not shown). The second exception is represented by toxinotypes XIa to XId. As they differ one from another in the A3 fragment, they should have been designated individual toxinotypes. However, as they all share a very typical PaLoc structure with a large deletion (Fig. 1), we decided to keep them as subtypes in the new scheme.
Variant C. difficile strains can produce all combinations of the three known toxins (A, B, and CDT).
The majority of variant toxinotypes produce both toxins A and B. Typically, they also produce, or at least contain functional genes for, binary toxin (24). Toxinotypes with minor changes in toxin genes (minor toxinotypes; Table 1) do not produce binary toxin. Seven toxinotypes have an A− B+ phenotype and could be binary toxin gene positive or negative. Toxinotypes XIa to XId form the only group with part of the PaLoc present but no toxin production; however, they do produce binary toxin (A− B− CDT+). A single strain that produces toxin A only and no toxin B has been described only recently (28). The A+ B− strain does not react with standard primers used for toxinotyping and hence is not included in the toxinotyping scheme.
There are several ways to detect C. difficile strains with variant toxin genes (toxinotypes I to XXXIV) without actually performing amplification of toxin gene fragments required in toxinotyping (29). However, none of them, except WGS, can detect all variant strains.
The easiest way is to test for the presence of the binary toxin gene, as this gene(s) is present only in strains with changes in genes for toxins A and B and extremely rarely in nontoxigenic strains other than toxinotype XI (6).
Another possibility for the laboratories still performing the cytotoxicity test using cell cultures is to look for a cytopathic effect morphologically different from that seen with reference strain VPI 10463 and similar to the morphological changes caused by C. sordellii toxins. The “difficile” phenotype is characterized by rounded cells that still have long protrusions, while a “sordellii” effect shows only rounded cells with no visible protrusions. This is typical of some variant toxinotypes (6).
Variant toxinotypes also correlate well with PCR ribotypes (6); therefore, detection of certain PCR ribotypes is also suggestive of strains with variant toxin genes. Some well-known PCR ribotypes with variant toxin genes are 017, 019, 023, 027, 033, 078, 126, 176, and 244.
Widely used commercial molecular tests (such as Xpert and illumigene) detect variants with toxin A and toxin B genes well but mostly do not recognize them as variant strains. An exception is the Xpert test, which would report also a positive binary toxin gene result and a tcdC deletion at position 117. One or both of these features are also the markers for variant C. difficile strains. A new, but rare, toxinotype (XXXII) is not recognized by Xpert (8) and, due to the absence of tcdA, probably also not by the illumigene assay.
In the era of WGS, toxinotyping will also change. However, a large part of the variability is observed in the repetitive regions of tcdA gene. Currently, repetitive regions are still difficult to sequence. Therefore, it is likely that toxinotyping will remain an important method for detection and differentiation of the majority of variant C. difficile strains. With WGS, further characterization of variant toxin genes and genomes will be possible and toxinotypes will be recognized as more or less coherent clusters of strains.
Toxinotypes represent the natural variants of C. difficile strains with changes in genes for toxins A and B and consequently also in the functional properties of variant toxins A and B (6). The studies of toxinotypes are important in understanding the structure-function relationships of the toxins and the evolution of the PaLoc and C. difficile species, in epidemiologic studies, and in development of vaccines and molecular diagnostic approaches (6, 10, 30).
The most prevalent toxinotypes isolated from humans worldwide are III (PCR ribotype 027), IV (PCR ribotype 023), V (PCR ribotype 078/126), and VIII (PCR ribotype 017). In a large EU study performed in 2008, toxinotypes III, IV, V, and XII were detected (31). The genes for binary toxin, another indicator for C. difficile variant toxinotypes, were found in 23% of the strains in this study. Strains of ribotype/toxinotype 078/V, 027/III, 017/VIII, 126/V, and 023/IV represented 8%, 5%, 4%, 3% and 3% of all 395 isolates from 73 hospitals and 26 countries, respectively, and were among 12 most prevalent ribotypes found (31). The same four ribotypes/toxinotypes are also very frequent in Japan, China and Korea (32), while only two (027/III and 078/V) are often isolated in the United States (33). In contrast to Europe and the United States, where 027/III is predominant, type 017/VIII is the most frequent variant toxinotype in Asia. As already mentioned above, type 244/IXb is an emerging cause of community infections in Australia (27). Toxinotype XI (all subtypes) is still rare in humans, and its role in the associated disease is debated because of the lack of toxins A and B and production of CDT only (34).
In the past, variant toxinotypes were reported to be typically associated with animal hosts (see reference 6 for further references); however, with increased numbers of studied species in different countries, the proportion of variant toxinotypes in animals is lower. Toxinotypes typically associated with animals are 078/V or 126/V (associated with pigs) and 033/XI (associated with cattle). Additional toxinotypes found in animals are I, III, IV, VIII, XII, and XIX (35).
Toxinotyping is a method to classify C. difficile strains into groups according to specific changes within toxin genes. Changes in the tcdA and tcdB toxin genes can range from very minimal deletions limited to only tcdA and a few point mutations in tcdB to significant changes in tcdB and large deletions leaving only a remnant of the PaLoc. Different toxinotypes are currently endemic or associated with outbreaks in several countries. Screening for changes in toxin genes and describing new variants are of importance in the development of new molecular diagnostic tests and vaccines.
We acknowledge Warren Fawley and Mark Wilcox (Clostridium difficile Ribotyping Network for England and Northern Ireland [CDRN] Reference Laboratory [Leeds], National Infection Service, Public Health England, United Kingdom) for providing ribotyping for some of the reference strains and A. Kocuvan for work with the toxinotype reference database.
Maja Rupnik graduated as a biologist at University of Ljubljana in 1991 and obtained her Ph.D. at the same university in 1998. From 1991 to 2003, she worked at the Biotechnical Faculty, University of Ljubljana. For two years, she was an independent research fellow. Since 2005, she has been leading the Department for Microbiological Research at the Institute for Public Health Maribor, which was reorganized into the National Laboratory for Health, Environment and Food (NLZOH) in 2014. Also, since 2005 she has been a Professor for Microbiology at the Faculty of Medicine, University of Maribor. Her research interest is focused on molecular biology and epidemiology of Clostridium difficile, on molecular typing techniques, and, lately, on gut microbiota. Her laboratory has established C. difficile toxinotyping. The honors and awards that she has obtained for her work include an Alexander von Huboldt grant, an ESCMID/bioMerieux Award for Advances in Clinical Microbiology, and a national award, the Zois Certificate of Recognition, for exceptional scientific achievements in microbiology.
The Slovenian Research Agency provided funding to M.R. and S.J. under grant numbers P3-0387 and J4-6810.
Supplemental material for this article may be found at http://dx.doi.org/10.1128/JCM.02083-15.