Species Distinction in the Trichophyton rubrum Complex

The Trichophyton rubrum species complex comprises commonly encountered dermatophytic fungi with a worldwide distribution. The members of the complex usually have distinct phenotypes in culture and cause different clinical symptoms, despite high genome similarity. In order to better delimit the species within the complex, molecular, phenotypic, and physiological characteristics were combined to reestablish a natural species concept.

(v) Tween 80 opacity test. Tween 80 agar medium was prepared with 10.0 g Bacto peptone, 5.0 g NaCl, 0.1 g CaCl 2 , and 15.0 g agar per 1,000 ml distilled water (17). After autoclaving, the medium was cooled to about 50°C and 5 ml of autoclaved Tween 80 was added. Culture plates (90 mm) were filled with 25 ml medium, inoculated, and incubated at 24°C for 2 weeks. Fungal growth was examined weekly; the presence of a halo due to calcium precipitation around the colony indicated lipolytic activity.
Molecular methods. (i) DNA extraction. Fungal material was transferred to 2-ml screw-cap tubes filled with 490 l 2ϫ cetyltrimethylammonium bromide (CTAB) buffer (2% CTAB, 100 mM Tris-HCl, 20 mM EDTA, 1.4 M NaCl) and 6 to 10 acid-washed glass beads. Then, 10 l of proteinase K was added and the components were mixed thoroughly on a MoBio vortex mixer (MoBio Laboratories, Carlsbad, CA, USA) for 10 min. After incubation for 60 min at 60°C, 500 l chloroform-isoamyl alcohol (24:1) was added and the mixture was shaken for 2 min; the tubes were centrifuged for 10 min at 14,000 rpm, and the supernatants were collected into new Eppendorf vials. To 400 l of the DNA sample, 2/3 vol (270 l) ice-cold isopropanol was added, the mixture was centrifuged again at 14,000 rpm for 10 min, and the upper layer was dissolved in 1 ml ice-cold 70% ethanol. The tubes were centrifuged again at 14,000 rpm for 2 min, air dried, and resuspended in 50 l TE (Tris-EDTA) buffer. Samples were stored at Ϫ20°C until analysis.
(ii) ITS PCR. Primers ITS1 (TCCGTAGGTGAACCTGCGG) and ITS4 (TCCTCCGCTTATTGATATGC) were applied for amplification, and 10 to 100 ng DNA was added to 25 l PCR buffer. The PCR program was as follows: 95°C for 4 min, followed by 35 cycles of 95°C for 45 s, 52°C for 30 s, and 72°C for 2 min, with a delay at 72°C for 7 min. The PCR products were visualized on 1% agarose gels and sequenced using an ABI BigDye Terminator (v3.1) cycle sequencing kit. The sequencing reactions were done at 95°C for 1 min and 30 cycles of 95°C for 10 s, 50°C for 5 s, and 60°C for 4 min on an ABI 3730XL automatic sequencer (Applied Biosystems, Foster City, CA, USA) with the ABI Prism BigDye Terminator cycle sequencing kit. The sequences were edited and assembled with the SeqMan program (DNAStar, Madison, WI, USA), manually corrected, and aligned using the MAFFT server (www.ebi.ac.uk/Tools/msa/mafft/) with default parameters. A phylogenetic tree was generated using the maximum likelihood method with MEGA (v7.0) software based on the general time-reversible model. T. erinacei IHEM 19618 and T. verrucosum IHEM 5480 were used as outgroups. The sequences of newly sequenced strains were deposited in GenBank; the accession numbers are listed below. The accession numbers of all the strains used in this study are listed in Table S1 in the supplemental material.
(iii) ITS diversity. Polymorphisms in ITS sequences of 175 strains from the T. rubrum complex were detected using DNASP (v5.10) software. The number of polymorphic sites (S), haplotype diversity (Hd), and the average number of pairwise nucleotide differences per site () were calculated (gaps/missing data were included). A haplotype network was constructed in the Network (v4.6.1.0) program (Fluxus-Technology, UK) using the median-joining method.
(iv) AFLP. AFLP genotyping was performed as previously described (18). Selective primers HpyCH4IV-C (50-fluorophore-GTAGACTGCGTACCCGTC-30) and MseI-TGAG (50-GATGAGTCCTGACTAATG AT-30) were used in the restriction ligation procedure. Fragments were analyzed on an ABI3500xL genetic analyzer (Applied Biosystems, UK); for this purpose, the amplicons were 10ϫ diluted with double-distilled H 2 O (ddH 2 O), and 1 l of this dilution was added to 0.1 l the LIZ600 internal size marker (Promega, Leiden, The Netherlands) and 8.9 l ddH 2 O. Raw data were further processed by using BioNumerics (v7.5) software (Applied Maths, Sint-Martens Latem, Belgium), and a dendrogram was generated based on the unweighted pair group method using average linkages (UPGMA) algorithm.
(v) MALDI-TOF MS. MALDI-TOF MS was performed as described by Packeu et al. (19). Briefly, a reference spectrum database was constructed based on the method of Cassagne et al. (20) and Normand et al. (21) and included reference strains representing all species from the T. rubrum complex available in the BCCM/IHEM collection (Table S2).
The isolates were incubated for 7 days on SGA with chloramphenicol. Some slow-growing strains were tested after 12 to 14 days. Protein was extracted with formic acid-acetonitrile. Reference metaspectra were established based on the mass spectra from four subcultures of the same strain. Samples were analyzed in quadruplicate, i.e., with four parallel analyses of the same extract. Spectra were recorded in the positive linear mode in a mass range of from 2 to 20 kDa using MALDI Biotyper automation control software (Bruker Daltonics, Bremen, Germany). The spectra of each spot were compared with those in the reference library and analyzed with MALDI Biotyper (v3.0) software (Bruker), with log scores (LS) ranging from 0 (no match) to 3 (perfect match). The MS-based identification was considered reliable only if at least three out of the four spots resulted in the same identification with an LS of Ն1.70. The spectra are maintained in a database at Sciensano and are available upon request.
Data availability. The ITS sequences of newly sequenced strains were deposited in GenBank under accession numbers MK806589 to MK806666.

RESULTS
ITS diversity. ITS alignment was performed for 175 strains of the T. rubrum complex; strains T. erinacei IHEM 19618 and T. verrucosum IHEM 5480 were added as an outgroup in the phylogenetic analysis. The total alignment length (with gaps) was 565 bp; the number of invariable sites was 514, and the number of variable sites was 51, among which 40 were gaps/missing data (singleton variable sites, 2; parsimony-informative sites, 9). The results of ITS haplotype analysis are shown in Fig. 1. In total, 12 haplotype patterns were revealed (Hd, 0.7938).
A phylogenetic tree based on the ITS sequences was constructed for 175 strains using the maximum likelihood method (see Fig. S1 in the supplemental material). Three main groups differing by a few single nucleotide polymorphisms (SNPs) were revealed in both haplotype analysis and the phylogenetic tree. Group 1 (haplotype 7 [H7] to H12 in the haplotype network, 57 strains) contained the neotype of T. violaceum, CBS 374.92. Within this group, nine strains (H12), including the type strain T. yaoundei CBS 305.60, deviated in one SNP (T-C) at position 217 and are referred to as group 1A. CBS 141826 (H8) and CBS 141829 (H9) were found to have mutations at position 375 (A-G) and 525 (C-T), respectively. An insertion of C was found at position 117 in IHEM 13375 (H10), and a deletion of A was found at position 236 in CBS 118535 (H11). These indels and SNPs led to small deviations in the tree caused by strains with different haplotypes.
Group 2 contained 39 identical strains (H2), including T. soudanense neotype strain IHEM 19751; among this group, 16 strains (H1) deviated by a deletion of 37 bases. Group 3 was observed at a distance of 4 bp from group 2; it contained 57 isolates, including 6 strains differing by 1 SNP, which were referred to as group 3A (H6).
Six strains, including T. kuryangei type strain CBS 517.63 and three T. megninii strains, could not be clearly assigned to any of the three groups and were considered to have an ambiguous affiliation (H3, H4); they were provisionally excluded from the study. Based on the positions of the (neo)type strains in groups 1 to 3, we conclude that group 1 represents T. violaceum (with group 1A being a variant), group 2 represents T. soudanense, and group 3 represents T. rubrum. The T. violaceum group contained T. Phenotype. The strains in group 1/1A (T. violaceum/T. yaoundei) had an average growth rate of 12.98 Ϯ 4.31 mm/2 weeks. After 2 weeks, T. violaceum colonies were invariably glabrous, leathery, and wrinkled and had a white, cream, or yellow color (occasionally with brownish or reddish tinges) at the obverse and a white, cream, or brown color (occasionally with reddish tinges) at the reverse; the colonies of some strains acquired a darker, red-brown or dark purple color (Table 1). Group 2 (T. soudanense) strains had an average growth rate of 14.05 Ϯ 3.69 mm/2 weeks, and the colonies had a similar texture, with the colonies also being wrinkled and usually showing a white or yellow to orange color at the obverse and an orange to orangeyellow color (occasionally with a brownish tinge) at the reverse. Group 3 (T. rubrum) strains had an average growth rate of 18.33 Ϯ 3.94 mm/2 weeks and formed fluffy colonies with a white (occasionally cream) color at the obverse and a white, cream, yellow, or brown (occasionally orange) color at the reverse.
Physiology. The physiological characteristics of each strain are shown in Table 1. Urea hydrolysis was observed in all strains; representative negative, weak, and positive results are shown in Fig. 2A. Weak urea hydrolysis (pale pink medium) was observed in 8.77%, 21.81%, and 3.51% of the strains in group 1, group 2, and group 3, respectively. Almost all group 3 strains (with one exception) showed the ability (minimally weak to excellent) to degrade and assimilate keratin, whereas 54.39% and 52.73% of the strains in groups 1 and 2, respectively, failed to hydrolyze keratin. There was no statistically significant difference in keratin degradation between groups 1 and 2 (P ϭ 0.86), whereas their difference with group 3 was significant (P Ͻ 0.001) (Fig. 2B).
Among randomly selected strains incubated with blond children's hair, none produced perforations (Fig. 2E and F); at most, some hyphae were found to be attached to the hair surface (Fig. 2E), whereas Nannizzia gypsea CBS 130813, used as a positive control, showed localized areas of pitting ( Fig. 2C and D).
The Tween 80 opacity test was used to examine the lipolytic ability of the analyzed dermatophytes. Most group 1 strains (47 out of 57, 82.46%) showed a positive response, manifested by a large halo of precipitate around the colonies (7.89 mm, on average, after 2 weeks of incubation) (Fig. 2H). In contrast, most strains in groups 2 and 3 had no or weak lipolytic activity; only a few of them showed lipolysis (9/57 and 13/55, respectively; average halo, 4.67 and 4.62 mm, respectively) (Fig. 2G). There was no difference between groups 2 and 3 (P ϭ 0.30), but their difference with group 1 was statistically significant (P Ͻ 0.001).
AFLP. A total of 172 strains were typed to obtain an AFLP dendrogram. As was found with the ITS data, strains showed a high similarity in their AFLP fingerprint profiles. Although four patterns (AFLP-A to AFLP-D) were distinguishable in the den-drogram based on an arbitrary cutoff value of Յ95% similarity, the difference between patterns was not significant. Group 1 comprised strains with all four patterns (A, 12.50%; B, 32.14%; C, 3.57%; D, 51.79%), whereas group 2 strains had two patterns (A, 58.49%; C, 41.51%) and group 3 strains had three patterns (A, 33.93%; B, 5.36%; C, 60.71%). Most strains in groups 2 and 3 had patterns A and C, and group 1 strains mainly showed patterns B and D. Pattern D was detected exclusively in group 1. T. kuryangei and T. megninii strains had only patterns A and C (Fig. 3).

DISCUSSION
Taxonomic entities among anthropophilic dermatophytes were shown to be similar in their molecular characteristics (6,22); however, our present data show the existence of three species in the Trichophyton rubrum complex, judging from the ITS rDNA data. Although none of the physiological, morphological, geographical, MALDI-TOF MS, or AFLP data were unambiguously diagnostic, we revealed different trends that were statistically significant. Groups 1, 2, and 3 contained the type strains of Trichophyton violaceum, T. soudanense, and T. rubrum, respectively; being the oldest epithets in these clusters, they provide correct names for the three groups. Group 1 contained a slightly deviating group 1A around the type strain of T. yaoundei from the Congo, but more data are required to establish whether it can be considered a separate taxon. Group 3 contained type strains of T. fischeri, T. fluviomuniense, T. kanei, T. raubitschekii, and T. rodhainii, all of which, consequently, can be regarded as proven synonyms of T. rubrum. T. violaceum and T. soudanense are prevalently found on the scalp (80.85% and 71.43% of strains from human sources, respectively), whereas T. rubrum is mostly found on glabrous skin (6.98% of strains from human sources). The geographical origin of the strains is somewhat difficult to trace back because of the increased traveling and migration of humans. Disregarding the isolates from Western countries, it was found that T. rubrum and T. violaceum have a global distribution, whereas T. soudanense is limited to Africa. Cases of T. soudanense infection reported in the United States were also observed among patients of African descent (23).
Colony appearance has been classically used to distinguish species in the T. rubrum complex, and our results confirmed previous data. Colonies of T. violaceum and T. soudanense are glabrous and grow slower than those of T. rubrum; most isolates of T. violaceum lack microconidia, which are generally present in T. soudanense and T. rubrum. Rather unexpectedly, reflexive branching appeared to be a diagnostic marker, as it is very common for T. soudanense and rarely or never observed in the other two species. Macroconidia occur only in T. rubrum, but as they are easily lost after repeated subculturing, this phenotypic trait has limited value. In the reverse, most colonies of T. violaceum are colored cream, those of T. soudanense are yellow-orange, and those of T. rubrum had brown tinges. The colony coloration observed in the present study was obviously influenced by prolonged culturing, which usually leads to the loss of pigmentation; nevertheless, a trend was detected. A naphthaquinone derivative, xanthomegnin, the main pigment synthesized by the members of the T. rubrum complex, was first isolated from a strain identified as T. megninii and later from the other strains of the complex (24)(25)(26); it could also be detected in human skin and nails colonized by T. rubrum (27). The pH-reversible naphthaquinone pigment xanthomegnin is the main pigment responsible for the observed colony colors. The darker tinges in T. rubrum are possibly associated with higher metabolic activity, leading to higher ammonium production and alkaline pH.
Almost all analyzed strains were positive for urea hydrolysis at 24°C, indicating urease expression. Urea broth and agar have been reported to be useless for species identification within the T. rubrum complex, although T. rubrum tends to hydrolyze urea slower than T. mentagrophytes (28). Urease activity should no longer be considered a criterion for differentiation of T. rubrum var. raubitschekii, as both taxa showed positive results (29).
Hydrolysis of Tween 80, derived from polyethoxylated sorbitan, and oleic acid is used as an indicator of the production of lipolytic enzymes (17). Lipases might be associated with different types of hair invasion of dermatophytes. Trichophyton violaceum and T. soudanense are mostly involved in tinea capitis; clinical forms of superficial infections vary from asymptomatic carriage to kerion, favus, scalp penetration, and black dot infection (30)(31)(32). Glabrous skin and the scalp differ in hair size and density and the abundance of sebaceous glands secreting oily material into hair follicles (33). According to the Tween 80 opacity test, T. violaceum had a higher lipolytic ability than the other species. We might speculate that T. rubrum on glabrous skin is directly involved in degradation of the epidermis of the skin, whereas T. violaceum and T. soudanense grow into the hair follicle through the sebaceous gland, reaching the medulla entering the more lipid-rich central hair shaft. This hypothesis is consistent with the fact that the prevalent hair infection type caused by the latter species is endothrix, resulting in the short, broken hairs clinically observed in tinea capitis (34,35).
The reduced Tween 80 hydrolysis in T. soudanense may be associated with a drier hair type prevalent in Africa; however, this speculation requires further experimental confirmation. T. violaceum has been thought to affect children more frequently than adults, because of the possible fungistatic activity of long-chain fatty acids in sebum secreted by sebaceous glands, whose activity increases with age (36). In addition, the higher incidence of tinea capitis in children might be also linked to underdeveloped immunity. The dual function of sebum in fungal infections should be further investigated.
Microscopy examination of T. violaceum-infected hair reveals endothrix accompanied by multiple fungal spores inside the hair, although this species is usually nonsporulating in vitro. The hair perforation test was consistently negative for all analyzed strains, confirming earlier findings (28,37) and indicating that fungi are incapable of degradation of the keratinous hair cuticle to reach the softer cortex in vitro. T. rubrum showed the highest keratin azure degradation, which was either absent or weak in the other species. The keratin azure test was first performed in fungi by Scott and Untereiner (16), who reported that T. rubrum had weak dye release after 6 weeks. Currently, keratin azure is widely used as a substrate to reveal keratinase activity (38). We modified the protocol by applying the agar into normal tubes instead of square bottles for better observation of the results for these slow-growing fungi. Most T. rubrum strains exhibited blue dye release after 1 month of incubation, and almost half of T. violaceum strains and one-third of T. soudanense strains showed some ability to degrade keratin. The nonspecific serine proteases subtilisin 3 (Sub3) and Sub4, detected in T. rubrum culture supernatants (39), were confirmed to degrade keratin azure, and both were predicted to be expressed in T. rubrum and T. violaceum by whole-genome analysis (40). Expression of keratinases, such as Sub3 and Sub4, and concomitant keratin degradation appear to be rather variable within a single species.
MALDI-TOF MS could separate most T. violaceum strains from T. rubrum and T. soudanense but had an insufficient discriminatory power to unambiguously discriminate between T. rubrum and T. soudanense. A recent study on dermatophyte identification using MALDI-TOF MS suggested that inclusion of T. soudanense in the database potentially leads to the misidentification of T. rubrum (41). Besides, some strains failed to be identified with MALDI-TOF MS because of poor growth on agar plates with an extended culturing time.
AFLP genotyping revealed a high degree of similarity among groups 1 to 3, indicating their close genetic relationship. Notably, much larger differences were found using the same amplification system between species of Sporothrix (42) and Cryptococcus (43), which, until recently, were considered species complexes, and even within the single species Hortaea werneckii (44). Approximate AFLP groups A to D were distinguished on the basis of the total profiles sorted by UPGMA clustering of the total profiles. Trichophyton violaceum (ITS group 1) contained strains divided over all AFLP patterns, A to D, confirming an ancestral position, as was noted with the ITS data ( Fig.  1), to T. rubrum and T. soudanense, which shared exclusively profiles A and C. On the basis of the total profiles, including minor bands, three strains clustered in AFLP group B, but on the basis of major bands only, IHEM 13801, IHEM 13801, and HS 215-28581 would more appropriately be classified in group A. The derived characteristics of T. rubrum may explain the above-described poor performance of MALDI-TOF MS, with T. rubrum showing 43.9% mismatches, while T. soudanense and T. violaceum were recognized nearly correctly.
In conclusion, T. violaceum, T. soudanense, and T. rubrum show coherent differences in independent parameters of clinical features, morphology, physiology, and genetics, although none of these parameters is strictly diagnostic. Genetically, the entities are very similar, suggesting a very short time of evolution. Combined with phenotypic differences and clinical predilection, in the absence of sexuality, we conclude that these are separate entities in sympatric evolution but with incomplete lineage sorting, as has also been observed in other recently evolving dermatophytes (45). In view of the clinical significance of these fungi, which has long been recognized in dermatology, we recommend maintaining the entities at the species level. The species, with a confidence level of Ͼ90%, cause different types of infection and show distinct colony morphology and microscopic features. Physiological tests on keratin degradation and lipolysis indicate differences in ecological specialization but are not practical as identification criteria, and AFLP and MALDI-TOF MS do not have sufficient discriminatory power to distinguish between all species reliably. Clinical manifestations remain a primary criterion for identification, which should be best confirmed by ITS sequencing. For laboratories lacking access to ITS sequencing capabilities, clinical manifestations could be combined with morphological characteristics of the strains for identification. T. violaceum and T. soudanense are mostly involved in tinea capitis. The distance in the ITS barcoding gene between T. violaceum and T. rubrum is 6 bp (1.06%), which is beyond the generally applied limit to be accepted as different species. Genetic differences with T. soudanense are smaller, but its endemism in northern Africa as a cause of tinea capitis, supplemented with its phenotypic differences, makes distinction of this entity clinically meaningful.

SUPPLEMENTAL MATERIAL
Supplemental material for this article may be found