A virulent genotype of Microsporum canis is responsible for the majority of human infections

Dermatophytes belonging to the Arthroderma otae complex comprise three closely related anamorphic species: Microsporum canis, Microsporum audouinii and Microsporum ferrugineum. M. canis is known to mate with tester strains of A. otae, at least in the laboratory. These mating experiments, however, have detected only a handful of (+) mating type strains, whilst the great majority of isolates displays the opposite (–) type. This observation has led several workers (Hasegawa & Usui, 1975; Hironaga et al., 1980; Weitzman & Padhye, 1978) to conclude that the (+) mating type may now be on the brink of extinction and that recombination no longer occurs in nature, or is at least highly reduced. Adaptation to carriage by furred mammals as primary hosts may explain the unbalanced distribution of mating types, reducing the probability of encountering a partner of the opposite sex.

All three species are able to cause human tinea capitis and tinea corporis, especially in prepubescent children. M. audouinii and M. ferrugineum are anthropophilic, generally being transmitted from human to human, and have evolved in Africa and Asia, respectively (Kaszubiak et al., 2004). M. canis has a worldwide distribution and is zoophilic, with humans predominantly only being infected after contact with mammals. The natural habitat of this species is the furred skin of cats, dogs and horses, where it generally resides asymptomatically. The occasional human-to-human infections are self-limiting after a few transmissions.

Epidemiological studies of human and animal infections by contagious strains of M. canis have remained unresolved due to a lack of polymorphic molecular markers. The detection of sources and routes of infection, and identification of contaminated spaces in hospitals, kindergartens, schools and animal nurseries would contribute to optimized therapy, prophylaxis and a hygienic regimen, and thus save financial resources. In addition, human infections due to M. canis tend to be moderately inflammatory, but cases of severe kerion-like tinea capitis or highly inflammatory ringworm infections do occur in similar patient populations (Ernst 1980; Pryce & Verbov, 1992; Stephens et al., 1989; Terragni et al., 1993). The question arises as to whether animal hosts harbour mixed genotypes of M. canis that differ in their degree of virulence to humans. In other words, do all lineages of M. canis have the same potential to infect humans, or is virulence limited to a subset of isolates within a genetically diverse population? Do these genotypes differ in predilection and pathogenicity? Do strains from particular animal host species compose monophyletic groups within M. canis with decreased gene flow? In order to address these questions, there is an urgent need for well-characterized neutral markers to analyse the population structure of M. canis.

In recent studies, several DNA markers (randomly amplified polymorphic DNA, sequencing of internal transcribed spacer and non-transcribed spacer regions of rRNA genes, intergenic spacers of nuclear DNA, and mitochondrial DNA genes) have been applied, but the degree of polymorphism was low within the species (Kaszubiak et al., 2004; Yu et al., 2004). Typing systems based on microsatellite markers have been shown to detect diversity at all levels from species down to individuals in pathogenic fungi such as Histoplasma, Coccidioides and Penicillium (Carter et al., 2001; Fisher et al., 2000, 2004). In the present study, we report on the application of microsatellite markers to a global set of M. canis strains to reveal patterns of genetic variation in this species.

Fungal strains. A total of 137 strains was analysed, most of them acquired from 1996 to 2002 (Table 1). Of these, 101 strains were identified morphologically as M. canis, 29 were M. audouinii and 7 were M. ferrugineum. Strains of the latter two species were used as outgroups. Thirty-three M. canis strains were isolated from epidemiologically unrelated animals (mainly cats and dogs) in Germany, whilst fifteen strains originated from horses. Fifty-three strains were obtained from epidemiologically unrelated humans (mostly children) at geographically distant locations (Austria, Mexico, Turkey, Korea, The Netherlands, USA, New Zealand and the Dominican Republic). Tinea capitis and tinea corporis was diagnosed in 22 and 18 cases, respectively, in addition to 7 cases of tinea faciei. No clinical data were available on the remaining six patients.

Table 1. Fungal strains analysed in this study

DNA extraction. DNA was extracted using the CTAB (N-cetyl-N,N,N-trimethylammonium bromide) method (Gräser et al., 1999) after growing the fungus on Sabouraud glucose agar (Difco Laboratories).

Isolation of microsatellites (enrichment methods). Microsatellite sequences were captured using biotinylated (GT)₁₂ and (GA)₁₂ probes and immobilized on avidin-coated beads. The captured DNA was subjected to washing steps, and then eluted, amplified and cloned to produce a library enriched for the target sequence. The method was modified slightly from the one used previously by Ohst et al. (2004). Briefly, genomic DNA was isolated from a clinical isolate of M. canis (H22). Approximately 10 µg DNA was digested with DpnII and cleaned by drop dialysis for 15 min. Linkers (Sau-A, 5'-GCGGTACCCGGGAAGCTTGG-3'; Sau-B, 5'-GATCCCAAGCTTCCCGGGTACCGC-3') were ligated to both ends of the fragments using T4 DNA ligase (New England Biolabs). After purification via columns, pre-hybridization PCR was performed with the Sau-A linker only (annealing temperature of 56 °C, 15 cycles). For enrichment, the PCR product was denatured and hybridized to the biotinylated (CA or GA) probe in a solution of 6x SSC (1x SSC, 0.15 M NaCl, 0.015 M sodium citrate)/0.1 % SDS. The mixture was denatured at 95 °C and cooled slowly (over 15–20 min) to room temperature. The probe was then captured with avidin beads (VECTREX Avidin D; Vector Laboratories) in TBT buffer [100 mM Tris-HCl (pH 7.5), 0.1 % Tween 20] at 50 °C for 30 min and washed three times with TBT plus 150 mM NaCl and three times with 0.2x SSC/0.1 % SDS. The DNA was then denatured from the beads in 10 mM Tris/HCl (pH 8)/0.1 mM EDTA at 95 °C for 5 min and again PCR amplified with Sau-A. The resulting PCR product was cloned and transformed using a TOPO TA cloning kit (Invitrogen). White selection colonies were picked and checked for the repeat insert using M13 primers and (AC)₁₀ or (GA)₁₀ primers. Inserts of 300–600 bp were chosen for sequencing using an M13 primer and an automated sequencing system (3130x Genetic Analyzer). Specific primers were then designed to amplify PCR fragments in the range of 100–300 bp containing more than 12 GT or GA repeats. Amplification of each primer pair was tested on a panel of strains of M. canis, including the genomic DNA of the isolate the library was generated from.

PCR amplification of microsatellite markers using specific primers. Standard PCR conditions were as follows: reactions were performed in 50 µl volumes containing 10 mM Tris-HCl (pH 8.3), 50 mM KCl, 3 mM MgCl₂ (1.5 mM for McGT₁₇), 20 pmol each primer (McGT₁₃ forward, 5'-GATCGGAGCATGCCATACAG-3'; McGT₁₃ reverse, 5'-TCTTCCCACCCTTCTCAATG-3'; McGT₁₇ forward, 5'-GCTCTGGGATAAGGTGTTTG-3'; and McGT₁₇ reverse, 5'-GTAGCAGTAAAGCCAAGAGGG-3'), 50 µM each dNTP, 2.5 U Taq polymerase (Applied Biosystems), and 50 ng template DNA. Samples were amplified through 30 cycles as follows: initial denaturation for 10 min at 95 °C, followed by denaturation for 50 s at 95 °C, annealing for 60 s at 60 °C and extension for 60 s at 72 °C. This was followed by a final extension step of 10 min at 72 °C.

Screening for length polymorphisms. PCR products (15 µl) were loaded onto 12 % polyacrylamide gels (Rotiphorese gel 29 : 1, 40 %; Carl Roth) and the microsatellites were run for 18 h at 12 W (constant power). Gels were silver stained and dried for documentation. Repeat numbers in alleles were calculated visually using a sequenced allele with known repeat number as the reference (EMPL accession nos AM295318 and AM295319).

Data analysis. Different approaches were used to assign strains to populations. Whilst distance-based methods proceed by calculating a pairwise distance matrix whose entries give the distance between every pair of individuals, model-based methods proceed by assuming that observations from each cluster are random draws from some parametric model. In the first case, this matrix is then displayed using a graphical presentation such as a tree. In the second case, inference for the parameters corresponding to each cluster is done jointly with inference for the cluster membership of each individual, using standard statistical methods such as the Bayesian method. The disadvantage of the distance-based method is that the identified clusters may be heavily dependent on the distance measurement and graphical presentation chosen (Pritchard et al., 2000). Therefore, we calculated genetic distances between individuals based on three different measurements, Dc (Cavalli-Sforza & Edwards, 1967), Dm and Ds (Saitou & Nei, 1987) distances, implemented in the software package POPULATIONS version 1.2.28 (). Neighbour-joining trees were constructed from the distance matrices and were displayed using TREEVIEW (). As a model-based method, a Bayesian approach was used in the program STRUCTURE version 2.1 (Pritchard et al., 2000). This method allows the assessment of confidence of the inferred clusters by fine statistical analysis, whilst genetic distance methods are more suited to exploratory data analysis. Various models were used with STRUCTURE, including the no-admixture model, which can deal with clonal reproduction. One million Markov chain Monte Carlo replications and a burn-in period of 100 000 generations were used. The probability of the data, assuming one to five populations (K), was estimated in three replicate analyses. The posterior probability and other values displaying the confidence of the number of populations were recorded.

After structuring the populations, Wright's F statistics were applied to compute the variance in allele frequencies and test for free gene flow versus population differentiation between the inferred populations. Theta (Weir, 1996) was calculated across loci and populations using MULTILOCUS version 1.3 (Agapow & Burt, 2001). Here, the null hypothesis is no population differentiation; 400 000 randomizations were used.

To test for clonality versus recombination in the M. canis sample, the overall and the in population separated (based on both cluster methods; see Results) index of association (I_A) was calculated using the software MULTILOCUS. In this test, the observed data are compared against the null hypothesis of random mating (random association of alleles from different DNA loci). When the null hypothesis is rejected, a clonal population structure is suggested.

Of the 19 typable microsatellite markers developed, we used the most polymorphic loci, Mc(GT)₁₇ and Mc(GT)₁₃, revealing four and five alleles within the M. canis set of strains, respectively. The alleles varied by seven and five dinucleotide repeats within each locus. Up to 3 alleles were found with the remaining 17 markers; however, these were represented only by single strains (data not shown). In M. audouinii and M. ferrugineum, the alleles were species specific. With marker Mc(GT)₁₃, two alleles with a single dinucleotide difference among M. audouinii strains were detected. Strains of M. audouinii and M. ferrugineum were excluded from further analysis because of the low variability detected with both markers. Only one strain of each was used as outgroup for the distance tree in Fig. 1. The data for each strain and locus are presented in Table 1.

(20K): Fig. 1. Neighbour-joining tree based on Dc distance (on the left): no asterisks, humans isolates; *, cat or dog isolates; **, horse isolates. A bar plot for K=3 using STRUCTURE (on the right) shows the same three clusters as the distance tree. The three populations (I–III) are indicated by different shades of grey. The three strains marked by asterisks grouped with cluster I by STRUCTURE and with cluster III by the distance method.

Analysis of the combined dataset of both markers detected a total of 11 multilocus genotypes among the 101 M. canis strains and 3 among the 36 M. audouinii and M. ferrugineum strains. The application of several distance methods revealed identical results (Table 1, Fig. 1). An indication of clonal reproduction within the M. canis sample set was the observation of three multilocus genotypes that were shared by multiple strains from unrelated hosts. Of the 11 multilocus genotypes, 1 was shared by 50 strains, whilst 2 genotypes comprised 14 and 15 strains . This corresponds to a genotypic diversity in this dataset of 0.71. The linkage disequilibrium analysis of the overall sample and the clone-corrected sample rejected the null hypothesis of random mating (I_A=0.19, P<0.001). The identity of spatially separated strains of the same genotype demonstrated the high degree of reproducibility of the technique used.

Independent from the underlying assumptions and using several distance measurements – the Bayesian approach and neighbour joining – always revealed trees with three clusters (I–III) within M. canis. Each method generated branches with nearly identical sets of strains, except for three strains that grouped with cluster I (Bayesian approach) or cluster III (distance approach), but with a slight affinity to cluster III (mixed genotype) using the Bayesian approach (Fig. 1). Isolates from cats and dogs were distributed evenly among the three clusters. The 6 multilocus genotypes in cluster III were shared by 13 isolates (87 %) from horses, 8 strains (24 %) from other animals and 5 human isolates (9 %) (Figs 1 and 2). Cluster II was less variable, although the total number of strains was comparable to that in cluster III (26 strains in cluster III and 24 strains in cluster II). Only 4 multilocus genotypes were shared by 15 animal strains (31 %), among which was 1 isolate from a horse, and 9 (17 %) of the human isolates (Figs 1 and 2). Six out of seven (86 %) human strains of Turkish origin were found in group II (Fig. 1). In group I, a single multilocus genotype was shared by most of the human isolates (74 %), independent of their distant geographical origins (Korea, Austria and Mexico), together with 11 animal strains (23 %) from Germany, including one isolate from a horse (Figs 1 and 2).

(15K): Fig. 2. Association between the three clusters of genotypes and the host animal. The number of strains is given as a percentage. Black bars, horse; grey bars, cat/dog; white bars, human.

After pre-defining the three populations on the basis of the cluster analyses, the repeated linkage disequilibrium analysis did not reject the null hypothesis of random mating for population III (I_A=0.049; P=0.33). Population I did not recombine, as it consisted of a single clone. For population II, the I_A did not reveal a meaningful result as several genotypes were present in one of the loci. Support for population differentiation in M. canis was given by the statistics of theta (θ=0.733, P<0.001).

Data on clinical pictures were obtained from 47 of the 53 patients studied. In total, 22, 18 and 7 cases of tinea capitis, tinea corporis and tinea faciei were revealed, respectively. Whilst strains in cluster I were able to cause all three forms of tinea, cases of tinea faciei and tinea corporis were missing in clusters II and III, respectively (Fig. 3).

(17K): Fig. 3. Association between the three clusters of genotypes and the clinical picture [tinea capitis (grey bars), tinea corporis (white bars) and tinea faciei (black bars)]. The number of cases (37, 7 and 3) for clusters 1–3 are given as a percentage.

The M. canis sample set under study could be subdivided into three populations (I–III), of which I and II in particular had established largely clonal dimensions. By contrast, in population III comprising the majority of animal strains (44 %), the null hypothesis of random mating was not rejected. These results suggested that clonal and recombining population structures in M. canis exist concomitantly and that mating may occur. This finding is not unexpected, as the species is known to mate in the laboratory, but reproduces mitotically by conidia when transmitted between host individuals (Hironaga et al., 1980). Support for the argument that M. canis is composed of heterogeneous populations comes from the statistically significant high values of theta. Low mating competence and a predominantly clonal mode of reproduction are evident for many fungal species (Taylor et al., 1999). The differences between populations are meaningful when associations are found with particular virulence factors, such as the keratinolytic proteinases in dermatophytes (Giddey et al., 2007; Monod et al., 2002).

From our results, we could exclude the possibility that geographical differentiation and allopatric speciation played an important role in structuring the populations of M. canis. The strains from Germany (all animal isolates) did not show any monophyletic clustering, and the human isolates from Europe, South America and Asia could be found jointly as a single cluster.

M. canis is a zoophilic fungus that is only isolated rarely from soil. Hence, we anticipated an association of genotypes with particular host species. However, we revealed cat-associated genotypes in all three populations (I–III). Similarly, equine ringworm or colonization of horse fur could not be linked to a monophyletic group of strains. The set of 15 isolates from Germany comprised 5 multilocus genotypes belonging to all 3 populations of M. canis. Of the horse isolates 10 out of 15 had identical alleles at both loci, indicating that a single clone was involved. It is evident that horse isolates have acquired the ability to cause superficial disease more than once in the course of their evolution. Maintenance of horse isolates as a separate species, Microsporum equinum (Delacroix & Bodin, 1896), is therefore not justified. A similar situation has been published with multilocus genotypes of Histoplasma capsulatum (Kasuga et al., 2003). Molecular analysis of four genes of isolates causing equine histoplasmosis, formerly classified as a separate variety, H. capsulatum var. farciminosum, revealed that strains were distributed over three phylogenetic clades. The authors concluded that maintenance of the var. farciminosum was phylogenetically meaningless.

Despite the fact that all of the M. canis strains under study were isolated from epidemiologically unrelated individuals, 74 % of the isolates from human patients but only 23 % of the animal isolates shared a single genotype. The over-representation of this genotype 1 among the human isolates suggests that it may have a higher degree of virulence (possibly a proteinase adapted to human keratin) than the ten remaining genotypes, having a higher potential to infect humans when transmitted from animals. The genotype has a pandemic distribution. Strains harbouring genotype 1 were isolated from patients from three continents (Europe, South America and Asia). This seems to be in conflict with the data presented by Cano et al. (2005) using inter-simple-sequence repeat PCR (ISSR-PCR), who demonstrated that numerous strains from humans have a limited distribution, and may even be restricted to a single patient. The authors found a total of 21 genotypes among 24 mainly human isolates from Spain. With the exception of a single genotype (pattern 1), none of the animal genotypes occurred on humans, and vice versa. The conclusions derived from these data could not be explained by DNA loci located between single repeats, such as the flanking regions of microsatellites. Mutation rates of flanking and repeat regions have been analysed in detail in other fungal species, revealing a 2500-fold higher mutation rate for the latter loci (Dettman & Taylor, 2004). Thus, microsatellite loci are likely to be more variable than the flanking regions analysed using ISSR-PCR. The high variability among strains in the study of Cano et al. (2005) may partly be due to the low reproducibility (93 %) of the technique used. In contrast to Cano et al. (2005), the few other studies performing strain typing in M. canis, e.g. randomly amplified polymorphic DNA, reported a very low variability among epidemiologically unrelated strains from cats, dogs and humans, despite their morphological diversity (Brilhante et al., 2005; Faggi et al., 2001). Such results are in agreement with the extensive developmental work we have done while searching for polymorphic markers: 90 % of the microsatellite markers were unacceptable for the population analysis due to the low variability displayed. This is likely to be caused by the high portion of clonal reproduction within the species. In another clonal species, Trichophyton rubrum, only 30 % of the typable microsatellites were polymorphic, whereas in strongly recombining populations of Microsporum persicolor, almost 100 % of the loci were useful for a population genetic study (R. Sharma, S. de Hoog, W. Presber, Y. Gräser & R. C. Rajak, unpublished data). Although our findings suggest that some M. canis strains have an increased infective potential to humans, there was no indication of an association between genotypes and the type of tinea caused by this genotype, as tinea capitis was caused by genotypes of all three clusters. The missing manifestations of tinea corporis and tinea faciei by strains of clusters II and III, respectively, may have been due to the low number of human cases (nine and five) falling in these population groups.

In conclusion, our data cannot rule out the possibility that there are several asexual, separated lineages within M. canis. Systematic sampling needs to be undertaken, particularly in animal strains from geographically remote locations. Additional microsatellite markers need to be developed to support the indication of the presence of recombination events in this fungus.

In addition to epidemiological studies, the markers developed in this project can be applied to other studies. In addition to virulence traits, as shown here, drug resistance may also be associated with particular genotypes. Although the loci under study are probably not based on resistance genes, the largely clonal reproduction in populations of M. canis keeps genes and their associated traits together. The developed microsatellite markers can also be used as diagnostic tools for the rapid and specific identification of species of the M. canis complex directly from clinical specimens. This has been shown already by Kardjeva et al. (2006) for the detection of T. rubrum. Mc(GT)₁₃ is able to discriminate the species M. canis, M. audouinii and M. ferrugineum by agarose gel electrophoresis.

For providing and identifying clinical strains, we thank: K.-H. Böhm and U. Siesenop, College of Veterinary Medicine, Germany; W. Buzina, Laboratory for Mycology and Molecular Biology, ENT University Hospital, Austria; N. Kiraz, Department of Microbiology, Osmangazi University Faculty of Medicine, Turkey; R. Arenas, Department of Dermatology, General Hospital Dr Manuel Gea Gonzale, Mexico; and J.-A. Kim, Department of Dermatology, Seoul National University, South Korea.