OUP user menu

Heterogeneity of large clostridial toxins: importance of Clostridium difficile toxinotypes

Maja Rupnik
DOI: http://dx.doi.org/10.1111/j.1574-6976.2008.00110.x 541-555 First published online: 1 May 2008

Abstract

Clostridium difficile toxinotypes are groups of strains defined by changes in the PaLoc region encoding two main virulence factors: toxins TcdA and TcdB. Currently, 24 variant toxinotypes (I–XXIV) are known, in addition to toxinotype 0 strains, which contain a PaLoc identical to the reference strain VPI 10463. Variant toxinotypes can also differ from toxinotype 0 strains in their toxin production pattern. The most-studied variant strains are TcdA−, TcdB+ (A−B+) strains and binary toxin CDT-producing strains. Variations in toxin genes are also conserved on the protein level and variant toxins can differ in size, antibody reactivity, pattern of intracellular targets (small GTPases) and consequently in their effects on the cell. Toxinotypes do not correlate with particular forms of disease or patient populations, but some toxinotypes (IIIb and VIII) are currently associated with disease of increased severity and outbreaks worldwide. Variant toxinotypes are very common in animal hosts and can represent from 40% to 100% of all isolates. Among human isolates, variant toxinotypes usually represent up to 10% of strains but their prevalence is increasing.

Keywords
  • Clostridium difficile
  • bacterial toxins
  • virulence factors
  • A−B+ strains
  • animals
  • BI/NAP1/027 strains

Clostridium difficile causes toxin-mediated diseases ranging from self-limited diarrhoea to severe pseudomembraneus colitis (PMC). It is considered to be the most important cause of nosocomial intestinal infections. Typically, the disease develops in elderly patients (>65 years of age), with current or previous hospitalization and recent antibiotic therapy (Johnson & Gerding, 1998; Kuijper et al., 2006). However, the incidence of community-associated infections is increasing; disease can be found in persons of any age, and antibiotic treatment or other kinds of gut flora-disturbing therapy are not always a prerequisite (Chernak et al., 2005). Clostridium difficile is gaining the status of an animal pathogen, but it can also be found in asymptomatic animals and in food (Rodriguez-Palacios et al., 2007; Rupnik, 2007).

Clostridium difficile strains can be differentiated using several typing methods: PCR-ribotyping, pulsed field gel electrophoresis (PFGE), restriction enzyme analysis (REA) and multiple-locus variable-number tandem-repeats analysis (MLVA). As in every bacterial species, different types prevail in different geographic regions and during different time intervals. In parallel to the recent changes in epidemiology described above, changes in the C. difficile population structure can be observed. In the last few years, several types of so-called variant C. difficile strains have started to spread. The aim of this review is to provide an overview of variant toxinotypes, their potential clinical impact, distribution, evolution and laboratory detection.

Clostridium difficile variant strains

Two large toxins, toxin A (TcdA, enterotoxin) and toxin B (TcdB, cytotoxin), are recognized as the main virulence factors of C. difficile, as most of the symptoms of disease can be reproduced in animal models by purified toxins alone (Lyerly et al., 1985). Both toxins are encoded on a 19.6 kb region of the chromosome termed the PaLoc (Fig. 1). The PaLoc has conserved borders and in all strains studied to date is found inserted at the same site on the chromosome (Braun et al., 1996).

Figure 1

Schematic presentation of the PaLoc region in toxinotype 0 and some representative variant toxinotypes. The 19-kb PaLoc region is composed of two toxin-encoding genes (tcdA and tcdB) and three accessory genes (tcdR, tcdE, tcdC). 3′-repetitive sequences characteristic for both toxin genes are indicated in toxinotype 0 only. PCR fragments, shown above toxinotype 0, are used for characterization of the PaLoc region; fragments B1 and A3 are used specifically for toxinotyping. Length differences (deletions or insertions) are shown as grey areas or with a graphical symbol (toxinotype XIV, A1 region). Toxinotypes II and XII are representatives of minor types that have only a single change in their PaLoc when compared with toxinotype 0. Differences are mostly observed as deletions in A3, but can also be RFLP in B1 fragment (toxinotype XII). Toxinotype XI has a significantly changed PaLoc and an A−B− phenotype. Other toxinotypes are representatives of major A+B+ toxinotypes. Restriction enzymes (the presence or the absence and not the exact positions of restriction site are shown): A – AccI, E – EcoRI, Ec – EcoRV, H – HindIII, Hc – HincII, N – NsiI, Nc – NcoI, P – PstI, R – RsaI, S – SpeI, X – XbaI.

For a long time, it was commonly accepted that we can distinguish between toxigenic C. difficile strains as those producing both toxins (A+B+) and causing disease, and nontoxinogenic strains (A−B−) producing neither toxin and not causing disease. However, later studies showed that C. difficile is a very heterogeneous species and that two types of variant isolate can be differentiated: (1) toxin-gene-variants, defined as variant toxinotypes and characterized by changes in toxin genes for TcdA and TcdB, and (2) toxin-production-variants, characterized by producing only TcdB or by the production of the additional toxin, binary toxin CDT.

As we will describe below, there is a huge overlap between both types of variant isolate, and toxin-production-variants are almost always also toxin-gene-variants. In this review, the term ‘ordinary strain’ will describe strains that produce toxins A and B (but not binary toxin CDT) and have a PaLoc identical to the reference strain VPI 10463 in which this region was first sequenced. ‘Variant strain’ will describe any strain that differs from VPI 10463 either in the DNA sequence of the PaLoc and/or in the pattern of toxin production.

Toxinotyping

In the genus Clostridium, toxinotypes have at least two definitions. In C. perfringens, the term ‘toxinotype’ describes a strain with a specific combination of toxins derived from a set of 12 known toxins. In C. difficile, toxinotype is defined as a group of strains with identical changes in the PaLoc when compared with the other strains (Rupnik et al., 1998).

Toxinotyping is an RFLP-PCR-based method. The entire 19 kb PaLoc region is amplified to produce ten PCR fragments (Fig. 1) that are subsequently compared with the reference strain VPI 10463 for length and restriction site polymorphisms. Each toxin gene is amplified in three fragments, approximately corresponding to the catalytic domain (PCR fragments A1 and B1), the putative translocation domain (A2 and B2) and the repetitive domain (A3 and B3) of each protein (Rupnik et al., 1997).

Amplifying all 10 fragments is too laborious for screening purposes and a modification of the method for detection of variant strains was described (Rupnik et al., 1998). The two most variable fragments are B1 covering the 5′-end of tcdB and A3 covering the 3′-end of tcdA. Therefore, for the toxinotyping of large strain collections, the B1 and A3 fragments are amplified and the HincII/AccI (B1) and EcoRI (A3) restriction patterns are determined. According to the patterns observed in B1 and A3 fragments, a strain is grouped within a given toxinotype. Currently, seven types of B1 HincII/AccI restriction patterns and 11 types of A3 fragment patterns (combinations of deletions and EcoRI restriction patterns) are known.

If previously unknown restriction types are observed in B1 and/or A3, a strain is defined as a new toxinotype and in this case the entire PaLoc is analysed with ten PCRs. A detailed description of toxinotyping is available at http://www.mf.uni-mb.si/mikro/tox.

Clostridium difficile toxinotypes and changes in the PaLoc region

The reference strain VPI 10463 represents the toxinotype 0 (zero). All strains that differ from this strain can be currently assigned to one of 24 variant toxinotypes, designated by roman numerals (I–XXIV). Their characteristics are summarized in Fig. 1 and Table 1. Changes in the PaLoc seen in these variant toxinotypes are insertions, deletions and point mutations. Interestingly, the three types of mutations seem to prevail at specific parts of PaLoc (Fig. 1). Deletions are found mostly in tcdA and to date no form of significant deletion in tcdB is known. The reason for this is probably the difference in the repetitive regions characteristic for the 5′-ends of genes tcdA and tcdB. Not only is this region larger in tcdA and consists of 30 repeats; the repeats are conserved and can be grouped into five classes at the nucleotide as well as the amino acid level. In tcdB 24 repeats can be recognized and these can be grouped into homology groups only at the amino acid level (Eichel-Streiber et al., 1992). The presence of homologous short stretches of DNA in tcdA probably enables recombination events resulting in deletions or insertions of fragments in the repetitive regions. In tcdB, the repeats are not conserved on the nucleotide level and therefore recombination does not occur.

View this table:
Table 1

Properties of variynt toxinotypes I to XXIV

ToxinotypeSerogroups ribotypes REA typesChanges in PaLocMinor/major toxinotypeCytopathic effectReferenceApproximate number of known strains
tcdBtcdAtcdC
0Several sero-, ribo- and REA typesReferenceReference1Most prevalentDBraun et al. (1996) (PaLoc)
IC 003, 012, 102 ndAs 0Deletion1Minor/A3DRupnik et al. (1998)<10
IIA14 103 ndAs 0Deletion1Minor/A3DRupnik et al. (1998)<10
III (a,b,c)A1, A5, G, nt 027, 034, 075, 080 BIRFLPRFLP2 or 3MajorDRupnik et al. (1998)>100
IVA1, A5 023, 034, 075, 080 ndRFLPRFLP4MajorDRupnik et al. (1998)<50
VA15 066, 078 ndRFLPRFLP3MajorDRupnik et al. (1998)<50
VIA15, A9, E 045, 063, 066 ndRFLPRFLP deletion3MajorDRupnik et al. (1998)<20
VIIE 063 ndRFLPRFLP deletion3MajorDRupnik et al. (1998)1
VIIIall F, some X 017, 047 CF1, 2, 3, 4, 5, 6; CG1, 3RFLPRFLP deletion1MajorSRupnik et al. (1998)>100
IXA15, A16 019 ndRFLPRFLP1MajorSRupnik et al. (1998)<50
XA 036 CY1RFLPRFLP, deletionntMajorSRupnik et al. (1998)2
XIE 033 AA1Not presentRFLP, deletion3MajorRupnik et al. (2001)<50
XIIS1 056 ndRFLPAs 01Minor/B1DRupnik et al. (2001)<50
XIIIK 070 ndAs 0Deletion1Minor/A3DRupnik et al. (2001)1
XIVH 111 ndRFLPRFLP3MajorDRupnik et al. (2001)2
XVC 122 ndRFLPRFLP2MajorDRupnik et al. (2001)2
XVIC nd ndRFLPRFLP, deletion3MajorDRupnik et al. (2003)1
XVIIX nd ndRFLPRFLP, deletionntMajorSRupnik et al. (2003)1
XVIIIH nd ndAs 0Deletion1Minor/A3DRupnik et al. (2003)1
XIXnt nd ndAs 0Deletion1Minor/A3DRupnik et al. (2003)1
XXnt nd ndRFLPRFLP1Minor/A3DRupnik et al. (2003)1
XXIC nd ndRFLPRFLP1MajorDGeric et al. (2004)<10
XXIInd nd ndRFLPRFLP1MajorDGeric et al. (2004)<10
XXIIIA9 nd ndRFLPRFLP1MajorDStare et al. (2007)<10
XXIVnd nd ndAs 0As 02Minor/tcdCndStare et al. (2007)1
  • * For minor toxinotypes the altered PaLoc region is indicated; major toxinotypes have changes in the majority of 10 PCR fragments overlaping the PaLoc.

  • distributed in groups <10; <50; <100; >100.

  • type according to deletions observed after NciI restriction (Geric Stare, 2007).

  • § toxinotype III can be further subgrouped into IIIa, IIIb and IIIc according to the changes in B2 and B3 PCR fragments (Geric et al., 2004); toxinotype IIIb includes the strains of ribotype 027.

  • nt – nontypeable.

  • nd – not done.

In contrast, the point mutations detected by RFLPs of amplified PCR products are significantly more common in the tcdB gene than in the tcdA gene (Fig. 1). For each part of tcdB, it is very easy to find an enzyme that would result in a large number of RFLP patterns, while in tcdA only a single or two different RFLPs would be observed with all known toxinotypes. Some variant toxin genes were completely sequenced (GenBank numbers Z23277, AF217292, AJ011301) and for some others only partial sequences are available (AJ294944, Y10689). All sequenced data confirm that tcdB is more variant and tcdA is more conserved. Additionally, specific regions in tcdB are more likely to mutate. As will be described below in the section termed Variant Toxins, most of the mutations are present in the catalytic region of both toxins.

Aside from the two toxin genes, there are three other genes within the PacLoc: tcdC, tcdR and tcdE (Braun et al., 1996). tcdC, which encodes a negative regulator of toxin expression, is highly variable. There are four types of deletions present in different toxinotypes (Stare et al., 2007; Spigaglia & Mastrantonio, 2002). Curry (2007) have sequenced tcdC in a collection of 199 strains and found 17 different tcdC genotypes. Several of these genotypes have either nonsense mutations or nucleotide deletions that, due to frameshifts, introduce a nonsense mutation. tcdC is also significantly changed in toxinotype X (strain 8864) (Soehn et al., 1998).

Insertions larger than one codon are observed upstream of tcdR (c. 150 bp in several toxinotypes), between tcdB and tcdE (only in toxinotype X; Song et al., 1999), and in tcdA of toxinotypes XIV, XVII, XXII and XXIII (Mehlig et al., 2001; Geric et al., 2004). This large insertion of about 2000 bp is actually a C. difficile-specific mobile element ISTron (Braun et al., 2000). It is spliced from mRNA, and TcdA produced by these toxinotypes is of the usual size.

Minor and major toxinotypes

Some variant toxinotypes have changes only in one part of the toxin genes and are defined as minor toxinotypes (I, II, XII, XIII, XVIII – XX, XXIV). Usually, the sole difference within the entire PaLoc is a deletion or insertion in the A3 fragment. Minor toxinotypes are mostly represented by only one known isolate. In cases where several known isolates are described per toxinotype (like toxinotype I), the strains are not similar one to another when tested by other genotyping methods. One interesting exception is the commonly found toxinotype XII. By definition, it is a minor toxinotype because the changes are limited to one part of toxin genes only. As mentioned above, all minor toxinotypes have changes in tcdA (A3). In contrast, toxinotype XII has an altered B1 fragment. Some of the toxinotype XII strains were also tested by other typing methods and all were serogroup S1 and ribotype 056. Another minor toxinotype (XXIV) differs from toxinotype 0 only in the tcdC region.

Major toxinotypes show changes in all three fragments or at least most fragments of both toxin genes (Fig. 1, Table 1). While some major toxinotypes only contain a single known representative, others can include from 10 to several hundred known strains (Table 1).

Correlation of toxinotypes with production of toxins TcdA, TcdB and binary toxin CDT

Three toxins are described in C. difficile. TcdA and TcdB belong to the group of large clostridial toxins (LCT) characterized by their large size (260–308 kDa), cytotoxicity for cultured cells and a common molecular mechanism of action (modification of small intracellular GTPases by glucosylation) (Eichel-Streiber et al., 1996). In addition, the LCTs are capable of autocatalytic proteolysis, which cleaves the catalytic domain from the holotoxin during the internalization process (Reineke et al., 2007). Binary toxin CDT is unrelated to TcdA and TcdB but is similar to clostridial binary toxins produced by C. perfringens type E and C. spiroforme. It is composed of two unlinked protein molecules, one of them responsible for host cell binding (CDTb) and the other for enzymatic toxic function (CDTa). It is encoded on a specific region of the chromosome termed CdtLoc, which includes the regulatorly gene cdtR and the two toxin genes (cdtA and cdtB) (Carter et al., 2007). Despite the identical name, binary toxin CDT should not be confused with cytolethal distending toxins produced by other bacterial species, which that have a completely different structure and function. Binary toxin CDT acts intracellularly and directly modifies actin molecules by ADP-ribosylation (Perelle et al., 1997). The role of binary toxin CDT in pathogenesis is still controversial (Geric et al., 2006).

Seven toxin production types can be differentiated according to the combination of all three toxins produced in C. difficile: A+B+CDT−, A−B+CDT−, A-B+CTD+, A+B−CDT+, A+B+CDT+, A−B−CDT+, A−B−CDT−. TcdA and TcdB phenotypes are determined by toxin gene amplification and/or toxin production. CDT is usually confirmed only by the presence of intact cdt gene(s) as the toxin itself is rarely analysed.

Confusion on toxin production by variant toxinotypes can be due to the fact that the best-known variant strains are of phenotype A−B+, and often ‘variant strain’ is a synonym for an A−B+ strain. However, we define variant C. difficile strains by changes in PaLoc (toxinotypes I–XXIV) and such strains can also produce other combinations of TcdA, TcdB and CDT (Table 2). Moreover, only one type of A−B+ strain is spread worldwide (toxinotype VIII), again adding to the common ignorance of the fact that several different A−B+ types are known. They will be described in the next section.

View this table:
Table 2

Correlation of toxinotypes with toxin production

Toxin production typeToxinotypeMolecular background
A+B+CDT−0 minor types I, II, XII, XIII, XVIII, XIX, XX major type XXICDT−: absence of entire or large part of CDT locus
A+B+CDT+major types III, IV, V, VI, VII, IX, XIV, XV, XXII, XXIII minor type XXIVCDT+: presence of full length CDT locus
A−B+CDT−VIII some 0-like strainsA−: nonsense mutation at aa position 47 in tcdA gene A−: mechanism unknown
A−B+CDT+X XVI, XVII, some V-like strainsA−: rearrangement in PaLoc and large deletion causing probably changes in regulation and low or no transctiption of truncated tcdA A−: mechanism unknown
A+B−CDT+IX-likeB−: mechanism unknown
A−B−CDT+XIa, XIb some strains without PaLocA−, B−: only small nonfunctional part of PaLoc present A−B−: no PaLoc
A−B−CDT−PaLoc and CDT locus negative strainsComplete absence of tcd and cdt genes

All strains in a given toxinotype produce the same pattern of toxins. A few exceptions to this correlation are also known. These are mainly strains with a toxin production pattern different from to that expected for their toxinotype. Representatives of such rare exceptions are: (1) a strain of toxinotype 0 with an A−B+CDT+ phenotype (Johnson et al., 2003), (2) a strain of toxinotype V with an A−B+ phenotype (Geric et al., 2003) and (3) strains of toxinotype IX with the toxinotype A+B−CDT+ (will be described below).

The phenotype A+B− has been mentioned occasionally in the early literature (Cohen et al., 1998; Wilkins & Lyerly, 2003). In some reports, strains were called A+B− because tcdB could not be amplified by the primers and conditions used, but strains were not tested for the TcdB production (Cohen et al., 1998). Therefore, it is very likely that these strain(s) had variant forms of tcdB and actually belong to one of the known A+B+ variant toxinotypes. Recently, strains containing both toxin genes (as tested by PCR) but deficient in functional TcdB production were reported in community-associated cases in Canada (MacCannell et al., 2006a). The strains are positive on a commercial test detecting only TcdA and negative on cell culture cytotoxicity test. By toxinotyping, the strains were grouped in type IX, which usually produce TcdA, TcdB and CDT. In contrast, this subtype of IX is noncytotoxic. Further studies are needed to clarify the presence or the absence of (non)functional TcdB in this group of strains.

Ordinary C. difficile strains (toxinotype 0) have the phenotype A+B+CDT−. An identical phenotype is found in almost all minor and in some major variant toxinotypes (Table 2). Most of the major variant toxinotypes will, in addition to TcdA and TcdB, also produce the binary toxin CDT (A+B+CDT+).

Although a considerable proportion of C. difficile strains have a truncated form of CdtLoc, the locus encoding binary toxin CDT (Carter et al., 2007; Stare et al., 2007), it is important to note that the presence of an intact CdtLoc and production of binary toxin CDT correlates almost perfectly with variant toxinotypes. Thus, CDT is produced almost exclusively by strains with a significantly altered PaLoc, whereas toxinotype 0 strains or minor toxinotypes with slight changes in the PaLoc are very unlikely to produce binary toxin CDT (Stubbs et al., 2000) (Table 2). There are, however, exceptions to this rule. Two strains were described that contain tcdA and tcdB genes identical to toxinotype 0 but were CDT positive (Spigaglia & Mastrantonio, 2002; Stare et al., 2007). The only change in the PaLoc is in tcdC, and such strains are grouped in toxinotype XXIV (Stare et al., 2007).

Strains with another phenotype, A−B−CDT+, can be further differentiated into two subgroups. One is characterized by complete absence of PaLoc but has the CDT genes (Geric et al., 2003; Geric et al., 2006). The other group (toxinotype XI) has intact CDT genes, but possesses a truncated and therefore nonfunctional PaLoc. While PaLoc-negative, CDT-positive strains are very rare, toxinotype XI is quite regularly found.

A−B+ toxinotypes

A−B+ strains were the first well-characterized variant strains of C. difficile (Borriello et al., 1992; Lyerly et al., 1992; Depitre et al., 1993). There are currently four known toxinotypes exhibiting the A−B+ phenotype: VIII, X, XVI and XVII (Table 2). Additionally, a single strain from toxinotype 0 (Johnson et al., 2003) and another strain toxinotyped as V (Geric et al., 2003) are known to have an A−B+ toxin expression pattern.

As mentioned above, the term ‘variant strain’ is still used by some groups as a synonym for A−B+ strains. Moreover, it is used only for strains of toxinotype VIII. In a way this is understandable, because toxinotype VIII is a large group with significantly more than 100 isolates described worldwide. In contrast, only one or two strains are known for the other three A−B+ toxinotypes (Table 1).

It is beyond the scope of this article to provide an exhaustive overview of the properties and clinical importance of A−B+ strains and the reader is referred to a recent review by Drudy (2007a). Here, we will discuss only molecular reasons for the absence of TcdA production. However, they are known only for toxinotypes VIII and X.

A−B+ strains are commonly defined by nonreactivity in commercial toxin A-only specific diagnostic tests. There are two possible explanations for this. One is that no TcdA is produced by A−B+ strains and the other is that a variant TcdA is produced that would not be recognized by the monoclonal antibodies used in these diagnostic tests.

Toxinotype X has a significantly changed PaLoc, with a deletion in tcdA as well as a stop codon within the remaining portion of tcdA. In addition, the negative regulator tcdC and some regulatory noncoding regions are changed or missing. This probably results in a complete absence of transcription from the tcdA promoter, as not even the truncated version of TcdA could be detected (Soehn et al., 1998).

Toxinotype VIII strains have almost the entire tcdA gene but 1.7 kb of the repetitive region is deleted. The deletion was characterized by several groups (Eichel-Streiber et al., 1999; Kato et al., 1999; Sambol et al., 2000). This large deletion within the immunodominant repetitive region has prompted speculation that TcdA is produced but not recognized by the commercial tests. However, when expressed as a recombinant protein, the repetitive regions of toxinotype VIII do react with anti-TcdA antibodies, indicating that if a variant TcdA of toxinotype VIII were produced by toxinotype VIII strains, it would also be recognized by commercial tests. Several attempts to purify TcdA from culture supernatants was unsuccessful, suggesting that toxinotype VIII is not a producer of TcdA. The reason for the absence of the toxin is a nonsense mutation introducing a stop codon and terminating TcdA translation at amino acid position 47 (Eichel-Streiber et al., 1999) (Fig. 2).

Figure 2

Characteristics of four A−B+ toxinotypes. (a) All four types have changes in tcdB and tcdA. Toxinotype XVI is similar to types V, VI and VII but has a larger deletion in tcdA. Toxinotypes X and XVII are also similar to each other and differ in the proportion of tcdA deleted and in the presence or the absence of large insertion between tcdE and tcdA. (b) Mechanism of non-TcdA production in toxinotype VIII – the most numerous and widespread A−B+ toxinotype. Toxinotype VIII strains produce a variant form of TcdB, but do not produce TcdA because of the stop codon at position 47 (indicated by *). If the repetitive domain of toxinotype VIII TcdA is expressed in a recombinant form, it does react with toxin A-specific antibodies. Restriction site designations are as shown in Fig. 1.

Recognition of toxinotypes in a routine laboratory

In a routine laboratory, culture and/or commercial toxin tests are usually used for diagnosis of C. difficile in faecal samples. A few laboratories also use cell cytotoxicity tests to detect the presence of C. difficile toxin B in stool. Most of the variant C. difficile strains will be detected, but they cannot be differentiated from ordinary strains by cultural methods. Only the A−B+ subgroup of variant strains will not react with commercial tests specific for toxin A and would therefore not be detected at all. However, most of the commercial producers offer combined assays specific for both toxins A and B. Again, such tests will detect variant strains but will not recognize them as variants.

In a routine laboratory, several tests can be used to distinguish variant C. difficile strains from ordinary ones (Rupnik et al., 2001). Thus, variant strains can be identified as: (1) strains that produce only TcdB (A−B+), (2) strains that cause a ‘sordellii-like’ cytopathic effect on cultured cells (see section Variant Toxins and Fig. 3) and (3) strains that produce a binary toxin or have the binary toxin gene most likely belonging to variant toxinotypes. No commercial tests for detection of binary toxin CDT are available but cdt genes can easily be detected by PCR (Stubbs, 2000;http://www.mf.uni-mb.si/mikro/tox).

Figure 3

Two types of cytopathic effect (CPE) caused by Clostridium difficile toxins. (a) Morphology of untreated cells; (b) ‘Difficile’ type of CPE is characterized by the remaining long protrusions and is caused by ordinary toxinotype 0 strains and most of the variant toxinotypes as well; (c) ‘Sordellii’ type of CPE is characterized by complete cell rounding. This is typical for TcsL produced by C. sordellii but is also observed in some of the variant C. difficile toxinotypes.

The most efficient, but a time-consuming method, for recognition of variant strains is amplification and analysis of B1-, A3- and tcdC PCR fragments, which is beyond the need for a routine laboratory and suitable only for reference and research settings.

Correlation of toxinotypes with other typing methods

Toxinotyping was first described using the UCL Brussels collection in which C. difficile strains are grouped using rabbit serum agglutination into more than 30 serogroups (Delme et al., 1985). Variant toxinotypes were at that time found in nine serogroups, but currently 15 serogroups are known to include variant strains. However, only serogroups F and S1 correlate perfectly with toxinotypes. Strains from serogroup F always belong to toxinotype VIII and strains of serogroup S1 always to toxinotype XII. All other serogroups that include variant toxinotypes include, in addition, toxinotype 0 or A−B− strains (Rupnik et al., 1998) (Table 1). A much better correlation was observed between toxinotyping and molecular typing methods such as PCR ribotyping (Rupnik et al., 2001) or REA typing (Geric et al., 2003; Johnson et al., 2003). In the case of the major toxinotypes, all strains of a given ribotype belong to a certain toxinotype. Because ribotyping is more discriminative than toxinotyping, a single toxinotype can include several ribotypes (Table 1). For example, toxinotype III can include ribotypes 027, 034, 075 and 080. Toxinotype III can also be further differentiated according to variation in B2 and B3 fragments (Geric et al., 2004) into IIIa, b and c. Ribotype 027 (also designated BI/NAP1/027 in several international studies) correlates completely with toxinotype IIIb.

Toxinotypes found in animal hosts

Cattle and pigs are two emerging important animal hosts for C. difficile and generally the diversity of strains found is much lower than in humans. Different groups have found toxinotype V to prevail in the pig population in the United States (Keel et al., 2007), the Netherlands (E. Kuijper, pers. commun.) and Slovenia (T. Pirs et al., pers. commun.). Toxinotype (V) strains are also found in the majority of bovine isolates (ribotype 078; Keel et al., 2007). On the other hand, a Canadian study described eight different ribotypes in isolates from symptomatic and asymptomatic calves and only four of them are known variant types: ribotype 078 (toxinotype V), ribotype 017 (toxinotype VIII), ribotype 027 (toxinotype III) and ribotype 033 (toxinotype XI) (Rodriguez-Palacios et al., 2006). Toxinotype XI was also isolated from calves in Slovenia (Pirs et al., pers. commun.).

Braun (2000) reported a high proportion of CDT+ strains among equine isolates (23.5%) and two of the strains were toxinotype V (M. Rupnik, personal data). Our data on strains isolated from horses in USA are in congruence with these findings, as more than 35% of the strains were CDT+ and a large proportion were toxinotype V (S. Marks and M. Rupnik, unpublished data). However, another study found 10 ribotypes among 20 equine isolates. Most of them were nonvariant and only one strain was from ribotype 078, which correlates with toxinotype V (Keel et al., 2007).

For other species, there are not many publications describing either direct use of toxinotyping or some other typing method, and therefore easy recognition of the toxinotypes is not possible (see ‘Recognition of toxinotypes in a routine laboratory’). A report on toxinotype III (ribotype 027) isolated from a Canadian visitation dog (Lefebvre et al., 2006) is worth mentioning because it shows well as to how animals can act as a possible vector for intrahospital spread or for transfer between the hospital environment and the community. Toxinotype III strains were also detected in ground meat. By ribotyping these strains were similar but not identical to PCR ribotype 027 (Rodriguez-Palacios et al., 2007).

Altogether, six toxinotypes (III, IV, V, VIII, XI and XII) out of 24 known were to date found or described in animal hosts. However, a single toxinotype (V) seems to be particularly adapted to animals and is present in three different animal hosts worldwide. It also seems that some countries, like Canada, have a larger diversity in animal C. difficile isolates than others.

Clinical importance of toxinotypes

Variant strains have been isolated from asymptomatic and symptomatic humans, from animals and from the environment. However, a given toxinotype does not correlate with a particular form of disease or with a certain type of population. The five toxinotypes most likely to be isolated from patients are III, IV, V, VIII, IX and XII (Table 4). Of these, toxinotype VIII and toxinotype IIIb (known as BI/NAP1/027 strains) are currently spreading worldwide and are associated with outbreaks and/or increased disease severity (Kuijper et al., 2006).

View this table:
Table 4

Overview of toxinotypes found in different studies

ToxinotypeBrussels serotyped collectionCardiff ribotyped collectionJapanUSA single hospitalEU study
01726335123268
I21111
II1
III651325
IV45126
V12128
VI432
VII1
VIII25811222
IX21111
X1
XI12
XII214
XIII1
XIV11
XV1
XVI1
XVII1
XVIII1
XIX1
XX1
XXI1
XXII1
XXIII
XXIV1
Tox-n.d.n.d.n.d.1357
All strains studied21910156153411
  • Data from Rupnik (1998, 2001, 2003); Geric (2004); Barbut (2007).

Of the 24 known variant toxinotypes, the strains from toxinotype III are most commonly associated with PMC. In addition to IIIb strains (BI/NAP1/027 strains), a few other non-027/ toxinotype III strains from PMC patients have been reported (Rupnik et al., 1998; Geric et al., 2004).

The first toxinotype VIII strains were isolated mainly from asymptomatic neonates and children (Depitre et al., 1993). The same paper also reported the nonvirulence of these strains in a hamster model. However, soon afterwards, toxinotype VIII-associated outbreaks and severe PMC cases were reported (Alfa et al., 2000; Limaye et al., 2000; Johnson et al., 2001; Kuijper et al., 2001; Barbut et al., 2002; Komatsu et al., 2003; Toyokawa et al., 2003; Drudy et al., 2007c). The emergence of the A−B+ toxinotype VIII has consequently changed the diagnostic methods for C. difficile with replacement of toxin A only-specific tests with specific tests that detect both toxins A and B.

C. difficile diarrhoea can be associated with any of the known antibiotic classes. Similar to the nonvariant types, the spread of variant types is enhanced by their antibiotic resistances. Particularly well described is resistance to clindamycin in A−B+ strains (VIII) (Kuijper et al., 2001; Pituch et al., 2003) and fluoroquinolone resistance in A−B+ strains (VIII) and BI/NAP1/027 strains (IIIb) (McDonald et al., 2005; Drudy et al., 2006, 2007b).

Data on the variant toxinotypes causing community-acquired disease are very limited. In the USA, toxinotypes IX and XIV/XV were reported (Chernak et al., 2005) and an A−B+ variant of toxinotype IX seems to be found often in community strains in Canada (D. MacCannell, pers. commun.). Binary toxin gene positive strains, which are most probably variant toxinotypes (Table 2), were found in community-associated cases in Hungary (Terhes et al., 2004).

The evolutionary aspect of toxinotypes

How have toxinotypes evolved?

Minor toxinotypes with deletions in tcdA are most probably spontaneous mutants, and we have already described the mechanism causing it. Major toxinotypes have large number of changes throughout the PaLoc. The possibility that several strains with large numbers of identical mutations would emerge independently is unlikely, and therefore we consider all strains within a toxinotype to be a lineage developing from a single ancestor. This has been confirmed for toxinotypes VIII and IIIb by PFGE, MLST, microarray studies and other molecular methods (Kato et al., 1998; Moncrief et al., 2000; Lemee et al., 2004; van den Berg, 2004; McDonald et al., 2005; Stabler et al., 2006). From a study on variant and nonvariant strains from serogroups containing several toxinotypes (0, III, IV, V, VI, VIII, IX) and A−B− strains analysed with arbitrarily primed PCR, it could be concluded that all variant toxinotypes do not share a common ancestor but have evolved independently on several occasions (Wozniak et al., 2000).

How old are toxinotypes?

Variant C. difficile strains have probably existed for a long time, but they do seem to have been much less common in the past. The very first description of a variant strain is a study on toxin production in defined media, where strain 8864 (toxinotype X) was found not to produce TcdA under any conditions tested (Haslam et al., 1986). The oldest variant strains in collections known to the author are isolates from 1985 (toxinotype III, IV, XXIII). Among older variant isolates is also the first binary toxin positive strain CD196 isolated in France (Popoff et al., 1988), the first known strain of type BI/NAP1/027 in Europe.

In the early 1990s, several studies using PCR as a possible rapid diagnostic method were published. Most of them showed that toxigenic strains to always contain both toxins (TcdA and TcdB) and PCR products were mostly identical in all strains. Only two unusual strains were reported, indicating that variant strains were present but rare (McMillin et al., 1991; Boondeekhun et al., 1993). Both strains had a shorter gene or protein for TcdA, but it is not possible to conclude from the published data into which toxinotype they could be grouped.

Is the prevalence of variant toxinotypes increasing?

One can compare the prevalence of variant strains in the past and present from toxinotyping studies as well as other publications in which variant strains could be defined by ribotypes or variant toxin production (describing binary-positive strains or A−B+ strains) (Table 3).

View this table:
Table 3

Comparison of some studies detecting the variant strains in humans

Definition of variant C. difficile strainYear of studyGeographic regionType of strains studiedProportion of variant strainsReference
Toxinotypes1996Europe (mostly Belgium and France)Large serotyped collection21.5% (of all studied strains; not estimated on entire collection)Rupnik et al. (1998)
Toxinotypes1998Europe (mostly England and Wels)Large ribotyped collection7.7% (estimated on all strains in the collection)Rupnik et al. (2001a)
Binary toxin positive strains1986–1999ItalyDifferent hospitals13.7%Spigaglia & Mastrantonio (2002)
Toxinotypes5/1996 – 4/2001USASingle hospital consecutive isolates11%Geric et al. (2004)
A−B+ strains and binary toxin positive strains1/2001–6/2001SpainSingle hospital consecutive isolates9.5%Alonso et al. (2005)
Toxinotyping1980–2000Japan, Korea, IndonesiaRibotyped collection from 6 different hospitals23.5%Rupnik et al. (2003)
Binary toxin producing strains REA typing4/2001–3/2002USASingle hospital65.3%McEllistrem et al. (2005)
tcdC sequence3–12/2001 4–5/2005USASingle hospital (Pittsburgh)81.9%Curry et al. (2007)
Toxinotyping2000FranceStrains from 17 hospitals in Paris6%Goncalves et al. (2004)
Binary toxin positive strainsNot givenHungaryHospital and outpatients in the region1.8%Terhes et al. (2004)
Toxinotypes4–6/200514 EU and associated countriesConsecutive hospital isolates20.9%Barbut et al. (2007)
Toxinotypes5–6/2005Netherlands17 hospitals36%Paltansing et al. (2007)

Spigaglia & Mastrantonio (2004) have compared Italian C. difficile isolates from three different time periods and found that before 1990 no binary toxin gene positive strains were present in the collection. Among the strains isolated between 1991 and 1999, there were already 24% binary toxin-positive strains, and in the years 2000–2001 this percentage increased to 45%. As strains with binary toxin genes are very likely to be variant toxinotypes, this is a well-documented increase of the prevalence of variant strains during time.

The large strain collections are probably a reasonably good reflection of the population structure for a given geographical region and time period. Analysis of a representative number of strains contained in the Anerobe Reference Unit, Cardiff, allowed an estimation of the prevalence of variant strains to be 7.7% in 2001 (Rupnik et al., 2001). Strain collections generally contain isolates from large-scale outbreaks, which can contribute large numbers of a given type in the collection and do therefore document the emergence and/or spread of a given type. A good example is new epidemic type BI/NAP1/027 (toxinotype IIIb). Until 2003, this particular type was extremely rare in various collections. Only a single ribotype 027 strain was known in the UK C. difficile collection (Rupnik et al., 2001), a few BI strains were in REA typed the USA collection (McDonald et al., 2005) and no such type was found in collections of human isolates from the Montreal region in Canada before 2001 (MacCannell et al., 2006b). After the outbreaks caused by this type in 2003, the number of isolates has increased to several hundreds in all three collections.

The expected proportion of variant C. difficile strains is best estimated at up to 11% from data on consecutive isolates from a single hospital that was not in an outbreak situation (Table 3). But in EU hospitals, the average proportion of variant strains found in 2005 in hospital isolates was 20.9% (Barbut et al., 2007). This high percentage is again mostly due to local outbreaks of IIIb and VIII strains.

Although some published data indicate that the prevalence of variant toxinotypes is increasing, this is mainly due to the increased numbers of two specific types: VIII and IIIb. Further carefully planned epidemiological studies would be needed to monitor the changes in C. difficile populations (Table 4).

Variant toxins produced by variant strains

Variant toxins differ from wild-type toxins in size, substrate specificity and cytopathic effect, biological effects and to some extent in immunoreactivity. Only some variant toxins have been characterized in more detail. Owing to deletions at the 3′-end of tcdA, toxinotypes VI and VII produce TcdAs with a lower molecular mass. Full length and shortened versions of variant toxins A do react with TcdA-specific monoclonal antibodies (Blake et al., 2004). In contrast, for TcdB some immunological differences were observed. TcdBs produced by A−B+ toxinotypes VIII (serogroup F) and X did not react with some mAbs specific for the reference toxin TcdBVPI10463 (Lyerly et al., 1992; Depitre et al., 1993).

LCTs can result in two types of cytopathic effects (Fig. 3). While C. difficile toxins A and B cause cell rounding with remaining protrusions (difficile type), C. sordellii toxins cause complete rounding of the cells. Some variant C. difficile TcdBs do have a ‘sordellii type’ instead of a ‘difficile type’ of cytopathic effect (Table 1) (Torres et al., 1991; Chaves-Olarte et al., 1999; Rupnik et al., 2001; Blake et al., 2004). The cytopathic effect depends on the pattern of small GTPases recognized by clostridial toxins (Chaves-Olarte et al., 2003).

The majority of the mutations in both tcdA and tcdB from C. difficile are located in the region encoding the catalytic domain. When the sequences are compared at the amino acid level, variant TcdBs have 93–96% homology throughout the entire molecule. Analysis of TcdB1470 also showed that despite mutations being found throughout the entire tcdB gene, 95% of the mutations resulting in a change in the amino acid sequence were located within the first 868 N-terminal amino acids encoding the enzymatic part of the molecule (Eichel-Streiber et al., 1995). Most amino acid changes were between residues 294 and 530. Therefore, at the amino acid level, wild-type TcdB and variant TcdB1470 had 78.6% homology in the catalytic region, but 100% and 99% homology in the translocation and receptor regions, respectively.

Mutations influence the pattern of GTPases used as a substrate and such variant patterns were described for TcdB toxins of toxinotype VIII (Chaves-Olarte et al., 1999), toxinotype X (Soehn et al., 1998) and toxinotype XIV (Mehlig et al., 2001). Wild-type TcdBVPI10463 modifies Rho, Rac and Cdc42. All three variant toxinotypes also use Rac and Cdc42 as a substrate. However, they differ in Rho modification, which is seen only in toxinotype XIV strains. Additionally, all three variant toxinotypes modify Rap, Ral and R-Ras. Biological effects were studied only for TcdB8864 (toxinotype X), which has a lethal dose in the mouse model of 6 ng/g, which is much lower than wild-type TcdBVPI10463 (50 ng/g) (Lyerly et al., 1992). Although the complete sequence is not known for all variant toxinotypes, RFLPs present in regions coding for the translocation and binding domains suggest that additional differences could also be expected in the internalization and binding properties of variant toxins.

Variability in other LCT

The group of LCTs consist of toxins A (TcdA) and B (TcdB) produced by C. difficile, toxins LT (TcsL) and HT (TcsH) produced by C. sordellii, alpha toxin (TcnA) produced by C. novyi and the recently described TcpL produced by certain C. perfringens types (Amimoto et al., 2007). This new member of the LCT group, toxin TcpL, is interesting, because it shows homology with other toxins but the repetitive binding domain is completely missing.

Although most of the genes for LCT toxins have been sequenced, the precise composition and borders of the complete toxin coding region are known only for C. difficile. Additionally, C. sordellii and C. novyi are much less frequently isolated from humans and the C. perfringens LCT was just recently described. Consequently, information on variability in LCTs other than TcdA and TcdB is limited.

Clostridium sordellii strain 6018 does not produce TcsH toxin (Green et al., 1996) and in this way resembles variant C. difficile A−B+ strains. Another two C. sordellii TcsL toxins (TcsL9048 and TcsL82) differ in their GTPases substrates and have been used for studying insulin-secreting beta cells (Tannous et al., 2001). In C. difficile, differences in GTPase recognition are due to the mutations in the catalytic domain of toxin genes. Therefore, one of these C. sordellii toxins most likely represents a variant TcsL. Some other C. sordellii strains probably have an altered modification pattern, although some differences were observed in the same strain (6018) using two different toxin preparation methods (Hofmann et al., 1996). The sequencing of aditional tcsL genes will definitely prove the extent of variability in C. sodellii LCT genes and toxins.

Summary

Clostridium difficile strains show significant interspecies heterogeneity in a region termed the PaLoc, coding for their two main virulence factors, toxins TcdA and TcdB. This is the basis for their distribution into toxinotypes I–XXIV. The consequences of major changes in the PaLoc are (1) the absence of production of one or both toxins and (2) production of toxins with altered properties. Interestingly, such variant strains were more likely to be associated with animal hosts in the past but in recent years, the number of reports of such strains in the human population is increasing. Variant strains are not easily differentiated from ordinary ones in the diagnostic laboratory. Some toxinotypes have hypervirulent potential and are emerging in many countries as a cause of severe disease and outbreaks. Although the mechanisms underlying the generation of variability in PaLoc are unknown, variant strains have been, and will be in the future, helpful in studies of the evolution of LCT and their role in pathogenesis.

Acknowledgements

M.R. is supported by the national research grant J4-7119-0406 and EU research grant LSHE-CT-2006-037870.

Footnotes

  • Editor: Neil Fairweather

References

View Abstract