The taxonomy of the Trichophyton rubrum complex: a phylogenomic approach

Abstract

The medically relevant Trichophyton rubrum species complex has a variety of phenotypic presentations but shows relatively little genetic differences. Conventional barcodes, such as the internal transcribed spacer (ITS) region or the beta-tubulin gene, are not able to completely resolve the relationships between these closely related taxa. T. rubrum, T. soudanense and T. violaceum are currently accepted as separate species. However, the status of certain variants, including the T. rubrum morphotypes megninii and kuryangei and the T. violaceum morphotype yaoundei, remains to be deciphered. We conducted the first phylogenomic analysis of the T. rubrum species complex by studying 3105 core genes of 18 new strains from the BCCM/IHEM culture collection and nine publicly available genomes. Our analyses revealed a highly resolved phylogenomic tree with six separate clades. Trichophyton rubrum, T. violaceum and T. soudanense were confirmed in their status of species. The morphotypes T. megninii, T. kuryangei and T. yaoundei all grouped in their own respective clade with high support, suggesting that these morphotypes should be reinstituted to the species-level. Robinson-Foulds distance analyses showed that a combination of two markers (a ubiquitin-protein transferase and a MYB DNA-binding domain-containing protein) can mirror the phylogeny obtained using genomic data, and thus represent potential new markers to accurately distinguish the species belonging to the T. rubrum complex.

Data Summary

1. Cornet L and Becker P, European Nucleotide Archive (ENA), project accession: PRJEB45548.

2. Cornet L and Becker P, European Nucleotide Archive (ENA), Biosample accessions: SAMEA8907458 to SAMEA8907475

3. Cornet L and Becker P, European Nucleotide Archive (ENA), sequence reads: ERR6039636 to ERR6039653 (2021).

4. Cornet L and Becker P, European Nucleotide Archive (ENA), Genome accessions: GCA_910591595.1, GCA_910591615.1, GCA_910591655.1, GCA_910591785.1, GCA_910591815.1, GCA_910591845.1, GCA_910591905.1, GCA_910591955.1, GCA_910591995.1, GCA_910592025.1, GCA_910592065.1, GCA_910592095.1, GCA_910592115.1, GCA_910592145.1, GCA_910592165.1, GCA_910592235.1, GCA_910592265.1, GCA_910592315.1 (2021).

Impact Statement.

Dermatophytes of the Trichophyton rubrum complex are anthropophilic fungi intensively studied due to their importance in human health. This complex is currently composed of three species (T. rubrum, T. soudanense and T. violaceum), and phenotypic variants (T. megninii, T. kuryangei and T. yaoundei) for which doubts remain around their status. In this work, we present the first phylogenomic analysis of the T. rubrum complex, conducted on 3105 core genes of 27 strains, of which 18 were newly sequenced from the BCCM/IHEM culture collection. We demonstrate the validity of the currently accepted species and further show that the three analysed morphotypes are genetically separate species. We also provide two potential new gene markers that resolve the T. rubrum complex phylogeny better than conventional barcodes, which will facilitate future identifications of isolates, and studies on the relationships between species belonging to this complex.

Introduction

Dermatophytes (Onygenales, Arthrodermataceae) are a group of closely related, pathogenic, fungi that cause superficial skin infections in both humans and animals. Species belonging to the Trichophyton rubrum complex are strictly anthropophilic dermatophytes and can infect the glabrous skin (tinea corporis and tinea pedis), the scalp (tinea capitis) and the nails (onychomycosis) in immunocompetent patients. Three species are currently accepted in the T. rubrum complex: T. rubrum (Castell.) Sabour. 1911, T. violaceum Sabour. ex E. Bodin 1902 and T. soudanense Joyeux 1912 [1–3]. T. rubrum has a worldwide distribution. It has become an increasingly prevalent dermatophyte in North America, Europe, Australia and East Asia since the 1950s, following a change in habits, such as the use of occlusive footwear. It causes tinea pedis, tinea corporis and onychomycosis, and is characterized morphologically by a fast growth and high sporulation in culture. Trichophyton violaceum is predominant in Middle Eastern countries, East Africa and South China. In vitro colony growth is slow and sporulation is low or absent. It mainly causes infections of the scalp (asymptomatic infections, kerion, favus, scalp penetration and black dot infections). Trichophyton soudanense causes infections of the scalp and is endemic to West African countries. It is also a slow-growing and non- or poorly-sporulating species.

The taxonomic relationships of the dermatophyte fungi were revised in 2017 by de Hoog et al. [2], regrouping the species in seven molecularly supported genera. The genus Trichophyton is placed in a derived position on the evolutionary tree and contains both zoophilic and anthropophilic species. The rRNA intergenic transcribed spacer region (i.e. ITS1, 5.8S rRNA and ITS2) is currently the most informative marker available for this genus [1, 2, 4]. Species within the genus Trichophyton are closely related, with only small genetic distances between them. Interestingly, the low genetic variation of the ITS region in the T. rubrum complex contrasts with a high phenotypic variability. This resulted in the description of taxa that were later synonymized with the current three species: T. rubrum (including T. raubitschekii, T. kanei, T. fischeri, T. flavum, T. fluviomuniense, T. pedis, T. rodhainii, T. kuryangei and T. megninii), T. soudanense (including T. circonvolutum and T. gourvilii) and T. violaceum (including T. yaoundei, T. glabrum and T. violaceum var. indicum) [1, 2, 5]. Following the cutoff-value of 99.6% as suggested in Vu et al. 2019 [5], the genetic distance between the ITS region of T. rubrum and T. violaceum is sufficiently large to recognize them as separate species. The distance between T. rubrum and T. soudanense is smaller, but a polyphasic approach, in which a phylogenetic analysis of the ITS gene was combined with data on morphology, physiology, geography and clinical characteristics, has led to the recognition of T. soudanense as a separate species [3, 5]. Finally, phylogenetic analyses also revealed two distinct clusters of strains that do not group with any of the three accepted species, namely a cluster around the kuryangei morphotype and a cluster around the megninii morphotype [3]. However, further analyses are required to determine their exact position and status. Several other methods, such as PCR fingerprinting and Matrix Assisted Laser Desorption/Ionization Time-Of-Flight Mass Spectrometry (MALDI-TOF MS) were indeed unable to distinguish all accepted species within the complex, demonstrating again their close interrelationships [3, 5, 6].

A low degree of genome-wide intra- and interspecific variation is observed in dermatophyte genomes in general [7, 8], including within the T. rubrum complex [4, 9], where the largest genome-wide variation is obtained between T. rubrum and T. violaceum, with 99.38% of sequence identity [4]. The genomes of T. soudanense and the megninii morphotype both exhibit a higher identity to the T. rubrum representative genome, with an average of 99.94% [9]. The intraspecific variation of T. rubrum genomes is extremely low. In the study of Persinoti et al. 2018 [9], seven T. rubrum strains, sampled across the globe, showed a 99.99% identity, which is indicative of the clonal nature of this species. The close genetic distances suggest that the species in the T. rubrum complex diverged in a very short evolutionary time-span. Recent divergence from a common lineage could be incomplete, resulting in species boundaries that are difficult to draw [5].

The T. rubrum complex affects human health on a global scale. Groups with distinct morphology, geographical distribution and clinical aspects are well-known within the complex, despite low genetic variation. Therefore, our study aims to provide additional support for a stable taxonomy through phylogenomic analyses. Hence, we test the confirmation of T. rubrum, T. violaceum and T. soudanense and investigate the possible reinstatement of T. megninii, T. kuryangei and T. yaoundei. Finally, we mine core genes in order to identify new candidate markers for future phylogenetic analyses and species identification.

Methods

Isolates

Twenty-seven strains belonging to the Trichophyton rubrum complex were selected. Nineteen strains were obtained from the BCCM/IHEM fungi collection, of which 18 were newly sequenced in this study. In addition, eight strains and their assembled genomes were downloaded from GenBank. Two strains of Trichophyton interdigitale were selected as the outgroup of the T. rubrum complex. The strains, along with their geographic origin, source of isolation and ENA accession numbers, are listed in Table 1. Isolates were chosen to maximize the diversity within the T. rubrum complex, including representatives of a wide variety of morphotypes (megninii, kuryangei, yaoundei, kanei, fischeri and raubitscheckii).

Table 1.

Details of the IHEM strains and public assemblies

Genome accession	Strain n°	Species	Former species	Geographic origin	Source	Isolation date
GCA_910591655.1	IHEM 13979	Trichophyton kuryangei	Trichophyton rubrum (morphotype kuryangei)	Burundi	tinea capitis	1966
GCA_910591595.1	IHEM 4712	Trichophyton kuryangei	Trichophyton rubrum (morphotype kuryangei)	Burundi	tinea capitis	1968
GCA_012184535.1	IHEM 26527T	Trichophyton kuryangei	Trichophyton rubrum (morphotype kuryangei)	Burundi	scalp	1956
GCA_910591615.1	IHEM 13968	Trichophyton megninii	Trichophyton rubrum (morphotype megninii)	Portugal	onychomycosis	1987
GCA_910591905.1	IHEM 13976T	Trichophyton megninii	Trichophyton rubrum (morphotype megninii)	Portugal	tinea corporis	1989
GCA_000616965.1	CBS 735.88	Trichophyton megninii	Trichophyton rubrum (morphotype megninii)	Spain	chin	1988
GCA_910591955.1	IHEM 25556	Trichophyton rubrum	Trichophyton rubrum	Belgium	pachyonychia	2012
GCA_910592265.1	IHEM 26523T	Trichophyton rubrum	Trichophyton rubrum	Netherlands	tinea pedis	1958
GCA_910592115.1	IHEM 26721	Trichophyton rubrum	Trichophyton rubrum	Belgium	tinea pedis	2015
GCA_910591845.1	IHEM 4915	Trichophyton rubrum	Trichophyton rubrum	Belgium	onychomycosis	1989
GCF_000151425.1	CBS 118892	Trichophyton rubrum	Trichophyton rubrum	Germany	onychomycosis	unknown
GCA_000616805.1	CBS 100081	Trichophyton rubrum (morphotype fischeri)	Trichophyton rubrum (morphotype fischeri)	Canada	contaminant	1997
GCA_000616825.1	CBS 288.86	Trichophyton rubrum (morphotype fischeri)	Trichophyton rubrum (morphotype fischeri)	Canada	contaminant	1986
GCA_000616845.1	CBS 289.86	Trichophyton rubrum (morphotype kanei)	Trichophyton rubrum (morphotype kanei)	Canada	buttock	1986
GCA_910592315.1	IHEM 26520	Trichophyton rubrum (morphotype raubitscheckii)	Trichophyton rubrum (morphotype raubitscheckii)	Canada	skin	1997
GCA_000616985.1	CBS 202.88	Trichophyton rubrum (morphotype raubitscheckii)	Trichophyton rubrum (morphotype raubitscheckii)	Canada	foot	≤1988
GCA_910591815.1	IHEM 13459	Trichophyton soudanense	Trichophyton soudanense	Somalia	tinea corporis	1966
GCA_910592065.1	IHEM 19743	Trichophyton soudanense	Trichophyton soudanense	Senegal	tinea capitis	≤1970
GCA_910592025.1	IHEM 19744	Trichophyton soudanense	Trichophyton soudanense	Senegal	tinea capitis	≤1970
GCA_910592235.1	IHEM 19751	Trichophyton soudanense	Trichophyton soudanense	Togo	tinea capitis	1980
GCA_000616865.1	CBS 452.61	Trichophyton soudanense	Trichophyton soudanense	Congo	unknown	1959
GCA_910591785.1	IHEM 13775	Trichophyton violaceum	Trichophyton violaceum	Congo	tinea capitis	1984
GCA_910592165.1	IHEM 25578	Trichophyton violaceum	Trichophyton violaceum	Iran	tinea capitis	2012
GCA_910592145.1	IHEM 26519T	Trichophyton violaceum	Trichophyton violaceum	Netherlands	skin	1992
GCA_001651435.1	CMCC(F)T3l	Trichophyton violaceum	Trichophyton violaceum	China	hair, tinea profunda cysticum	2013
GCA_910592095.1	IHEM 13375	Trichophyton yaoundei	Trichophyton violaceum (morphotype yaoundei)	Kenya	tinea capitis	1973
GCA_910591995.1	IHEM 19041	Trichophyton yaoundei	Trichophyton violaceum (morphotype yaoundei)	Belgium	tinea corporis	2001
GCA_000622975.1	MR816	Trichophyton interdigitale	Trichophyton interdigitale	Germany	toe nail	1996
GCA_012182655.1	UCMS-IGIB-CI14	Trichophyton interdigitale	Trichophyton interdigitale	India	skin, nail	2014

DNA extraction

The strains were cultivated for 7 to 21 days at 25 °C on Sabouraud agar with a cellophane filter for easier harvesting of fungal material. Genomic DNA was extracted using the Qiagen Genomic-tip 20 G⁻¹ kit (Qiagen, Valencia, CA, USA), according to the manufacturer’s instructions and following the protocol for yeast lysis with several adaptations: (i) before lysis, the collected fungal material was placed at −80 °C for at least 30 min and subsequently lyophilized overnight, (ii) samples were immersed in liquid nitrogen and ground with a micropestle prior to enzymatic lysis, (iii) during the washing steps, DNA was washed four times instead of three times, (iv) after precipitation of DNA with isopropanol, samples were centrifuged for 60 min at 15 °C instead of 4 °C. The integrity of the extracted genomic DNA was checked with the 4200 TapeStation System (Agilent, Santa Clara, CA, USA). Highly intact DNA (DNA integrity values >8.5) was used for further analyses.

Library construction and genome sequencing

Genomic library preparation and whole genome sequencing were performed by Eurofins Genomics Europe Sequencing (Konstanz, Germany). Paired-end libraries were constructed based on the NEBNext Ultra II FS DNA Library Prep Kit for Illumina (New England Biolabs, Ipswich, MA, USA) and sequenced at a 100× depth using the Illumina HiSeq 2000 platform (Illumina, Inc., San Diego, CA) with PE125 mode.

Genome assembly and annotation

For the 18 new strains, raw Illumina paired-end reads were trimmed for adaptors and filtered for low-quality reads using fastp v0.19.6 [10], with default settings. Remaining reads were assembled using metaSPAdes v3.10.1 [11], with default settings. The metagenomes obtained were binned using CONCOCT, with default settings [12]. The contamination levels and the completeness of fungal bins were estimated with EukCC [13], with default settings. Contigs resulting from the assembly of the fungal bins were scaffolded on the representative genome T. rubrum CBS 118892 (GCF_000151425.1) [7] using RaGOO v1.1 [14], with default settings. Average Nucleotide Identity (ANI) between genomes was estimated using dRep software [15], with the ‘no quality filtration’ option activated. Proteins were predicted with AMAW v1.0 (Meunier et al., unpublished), a wrapper tool for MAKER [16], with no prior gene model. Annotations are available in the figshare repository (https://doi.org/10.6084/m9.figshare.14762394.v2). However, the proteins of T. rubrum CBS 118892 (GCF_000151425.1) and T. violaceum CMCCT3I (GCA_001651435.1) were provided as supporting evidence, whereas RNA-seq data of T. rubrum (SRR9861013), freshly assembled using Trinity v2.4.0 [17], with default settings, were provided as expressed sequence tag (EST) evidence. Proteins from the 11 NCBI genomes (including two outgroups) were also predicted with AMAW.

Core genes

The protein sequence files were used for orthology inference using OrthoFinder v2.3.3 [18], with default settings. The script classify-mcl-out.pl [19] (available at https://metacpan.org/dist/Bio-MUST-Tools-Mcl) was used to select single-copy genes present in all genomes, resulting in the selection of 3105 core genes. Protein sequence alignments were then back-translated by capturing and aligning the corresponding DNA sequences with the programme leel ([20]; available at https://metacpan.org/dist/Bio-MUST-Apps-FortyTwo).

Phylogenomic analyses

Phylogenomic analyses were conducted on the 3105 core genes with the same jackknife protocol for both DNA and protein jackknife trees. Sequences from the 3105 files were first aligned using MAFFT v7.453 [21], run with the anysymbol, auto and reorder parameters. Conserved sites were then selected using trimAl v1.12 [22] with the gappyout option. One hundred datasets of ca. 100000 conserved positions (either nucleotides or amino-acids) were constructed by randomly combining alignment files using the script jack-ali-dir.pl from Bio-MUST-Core (available at https://metacpan.org/dist/Bio-MUST-Core). The 100 supermatrices were assembled using SCaFoS v1.30k [23], with default settings. Trees were inferred using RAxML v8.1.17 [24] under the fast experimental tree search method. The PROTGAMMALGF and GTRGAMMA model were used for protein-based and DNA-based analyses, respectively. Two consensus trees were built from the two sets of 100 jackknife trees using consense v3.695 (from the phylip package [25], but modified to handle long sequence names), with default settings.

A large protein phylogenomic analysis was performed on the whole set of 3105 protein core genes. After selection of conserved sites with trimAl [22] v1.12, a supermatrix of 29 organisms × 1688754 unambiguously aligned amino-acid positions (0.87% missing character states) was assembled using SCaFoS v1.30k, with default settings. A large DNA phylogenomic analysis was performed on the same set of 3105 protein core genes. After selection of conserved sites with trimAl, a supermatrix of 29 organisms × 5083094 unambiguously aligned nucleotide positions (1.25% missing character states) was assembled with SCaFoS, as for the protein tree. The supermatrices are available in the figshare repository https://doi.org/10.6084/m9.figshare.14762394.v2. The trees were inferred using RAxML v8.1.17 with 100 bootstrap replicates under the PROTGAMMALGF model for the protein tree and GTRGAMMA model for the DNA tree.

Ribosomal RNA region

The ribosomal RNA region was predicted and extracted for both the 18 newly sequenced genomes and the 11 NCBI genomes. RNAmmer v1.2 [26] was used in eukaryotic mode to predict small subunit (SSU) rRNA (18S) and large subunit (LSU) rRNA (28S) coordinates. Internal Transcribed Spacers (ITS) 1 and 2 were predicted using ITSx [27], with default settings. The contig corresponding to the ribosomal RNA region was manually added to the main fungal bin when it had been excluded by CONCOCT (Table S1).

Marker gene analyses

The 3105 aligned genes, post-BMGE alignments, were used to compute single-gene trees using RAxML v8.1.17 with 100 bootstrap replicates under the GTRGAMMA (for DNA, one partition) and PROTGAMMALGF (for proteins) models. These trees were then compared to the large protein phylogenomic analysis (taken as a reference tree) by computing Robinson-Foulds distances (R-F [28]), as well as ‘information-theoretic’ R-F distances, adjusted for phylogenetic information content (R-F info [29]) using the R package TreeDist [29] for R v3.6.3 [30]. The two sets of R-F distances were computed either on the original 3105 trees or on trees after collapsing branches <1.1e-6 substitution per site, defined hereafter as R-F polytomies, using the di2multi function of the R package ape [31]. This second approach was used to test the potential effect of inflated incongruence caused by strictly-bifurcating RAxML trees resolving identical sequences into randomly ordered soft polytomies. Splits matching between trees were visualized using the VisualizeMatching function of the TreeDist package. The 11 most congruent genes in DNA trees (R-F info polytomy cut-off of 0.4) were considered in various combinations to assemble 18 new concatenations using SCaFoS v1.30k [23] with default settings. Genes were first concatenated in ascending order, beginning from the best gene according to collapsed R-F values and adding genes one at a time to obtain a final concatenation of 11 genes. In the opposite, descending way, starting from a concatenation of 11 genes, genes were removed one by one, again beginning with the best gene according to collapsed R-F values. R-F distances were then computed on the corresponding additional trees, inferred as above.

Results

Genome assembly and annotation

The 18 newly sequenced IHEM strains (5× T. rubrum; 3× T. violaceum; 2× T. yaoundei; 4× T. soudanense; 2× T. megninii; 2× T. kuryangei) of this study were all highly complete after the assembly process, with completeness >99.5%, as estimated by EukCC (Table S1). Ribosomal RNA subunits (SSU, LSU) and the ITS region were generally lacking in the 11 public genomes while they were mostly present in the genomes assembled in the present study (see Table S1 for details). The representative genome used for scaffolding, T. rubrum CBS 118892 (GCF_000151425.1), was composed of 36 scaffolds and featured an N50 of 2.15 Mb. The number of contigs and N50 of the 18 newly assembled genomes are lower, with scaffold numbers ranging from 17 to 19, and N50 ranging from 1.8 to 2.1 Mb. Pairwise ANI values computed between all genomes, including public genomes, indicated that the strains are closely related, with ANI values >99.5% between taxa of the T. rubrum complex.

Phylogenomic analyses

To accurately reconstruct the phylogeny of the T. rubrum complex, we used two independent methods to evaluate statistical support, namely bootstrapping and jackknifing (Fig 1). The three maximum likelihood analyses based on the 3105 core genes, i.e., the DNA jackknife analysis, the protein jackknife analysis and the large protein phylogenomic analysis, all show similar topologies (Figs 2 and 3). In the first bifurcation within the ingroup, the strains are split into two highly supported groups. The first group includes T. violaceum and its morphotype yaoundei. The T. violaceum strains and those from the morphotype yaoundei are each placed into their own clade, which both receive full bootstrap support, as well as high jackknife support (T. violaceum: 80; T. yaoundei: 86). Jackknife support values will always be lower considering that jackknife replicates are much smaller (100000 positions) and thus contain less phylogenetic signal than the longer bootstrap replicates (1688754 positions). On average, the jackknife proportions are lower in the DNA jackknife tree compared to the protein jackknife tree. The second group includes T. rubrum, T. soudanense and the T. rubrum morphotypes megninii and kuryangei. Within this second group, the T. rubrum clade is highly supported by both jackknife and bootstrap proportions. The clade comprising T. soudanense and the morphotypes megninii and kuryangei is fully supported in the bootstrapping analysis, but receives lower support in jackknife (60). The T. soudanense clade is well-supported by both jackknife and bootstrap proportions. The clade containing morphotypes megninii and kuryangei is well supported in both analyses. Each morphotype is placed into its own clade, which both receive full support in bootstrapping and lower support in the jackknife (megninii: 61; kuryangei: 60). The branching patterns within the different taxa show some variation between the different ML analyses, with the exception of the megninii group, which has stable internal relationships for all its strains. The evolutionary distances, as represented by branch lengths, are generally short (Fig. 3), with slightly longer lengths for the T. violaceum – yaoundei group segregating from the other taxa.

Fig. 1.

Graphical abstract. We used 18 newly sequenced BCCM/IHEM strains and 11 public genomes assemblies, of which one BCCM/IHEM strain, to reconstruct the phylogeny of the Trichophyton rubrum complex on a dataset composed of 3105 core genes. We used two independent methods, namely bootstrapping and jackknife, to assess the robustness of our phylogenomic results. Robinson-Foulds (1981) distance comparison allowed the identification of two new single-gene markers. Based on our phylogenomic results, we propose to consider morphotypes of this complex as separate species.

Fig. 2.

Comparison between consensus jackknife phylogenomic trees. Phylogenomic analyses were conducted on 3105 core genes in DNA and in proteins. 100×100000 concatenation of genes were produced with SCaFoS and the corresponding trees computed with RAML under the models GTRGAMMA for DNA and PROTGAMMALGF for proteins. T. rubrum is in red, T. kuryangei in orange, T. megninii in purple, T. soudanense in brown and T. violaceum - T. yaoundei is in blue. Jackknife support values are shown at the nodes.

Fig. 3.

Large-scale protein analysis of 27 strains belonging to the T. rubrum complex. The maximum likelihood tree was inferred on 3105 core genes under the PROTGAMMALGF model with RAxML from a supermatrix of 29×1688754 unambiguously aligned amino-acid positions. T. rubrum is in red, T. kuryangei in orange, T. megninii in purple, T. soudanense in brown and T. violaceum is in blue and - T. yaoundei is in green. Bootstrap support values are shown at the nodes. Branch length of outgroup are divided by five.

Marker genes

The tree resulting from the large phylogenomic analysis of all core genes was used as a reference to investigate the presence of potential marker genes within the core gene dataset. All sets of R-F distances for single-gene protein trees were lower than R-F distances for single-gene DNA trees. The single-gene protein trees were thus selected for systematic comparison. R-F distances for DNA and protein trees are available in Table S2.

With a cutoff of 0.40 in ‘info R-F polytomies’ values, the 3105 genes can be reduced to 11 genes (Table 2). R-F values were ≥0.38 (median=0.48, IQR=0.08), and ‘R-F polytomies’ values ≥0.33 (median=0.39, IRQ=0.025). ‘info R-F’ values were ≥0.38 (median=0.43, IQR=0.0975), and ‘info R-F polytomies’ values ≥0.28 (median=0.38, IQR=0.05). None of these 11 genes presented values <0.30, except for the best gene (Ubiquitin-protein transferase) with an ‘info R-F’ value of 0.28 after branch collapse. The optimal combination of these genes was tested in two ways, ascending or descending (see Methods for details). The best association is the concatenation of eight genes using the descending method (Table 2). This association presents R-F values and ‘info R-F values’ of 0.27 and 0.26, respectively, after branch collapse. Low R-F values can also be obtained by combining only two genes (a ubiquitin-protein transferase and a MYB DNA-binding domain-containing protein), using the ascending method. Concatenation of these two genes gives R-F values of 0.42, ‘R-F polytomies’ of 0.33, ‘info R-F’ of 0.34 and ‘info R-F polytomies’ of 0.21.

Table 2.

Robinson-Foulds values for best core genes. The four R-F values were computed with the ape software package. R-F values are given for the 11 best core genes and for various concatenations of them (see Methods for details). The two best concatenations are shown in red

Gene	Individual gene				Ascending					Descending
	R-F	R-F poly	Info R-F	Info R-F poly	#gene	R-F	R-F poly	Info R-F	Info R-F poly	#gene	R-F	R-F poly	Info R-F	Info R-F poly
Ubiquitin-protein transferase	0.54	0.45	0.41	0.28	1	0.38	0.35	0.4	0.27	11	0.38	0.31	0.38	0.31
MYB DNA-binding domain-containing protein	0.42	0.40	0.38	0.32	2	0.42	0.34	0.33	0.21	10	0.35	0.30	0.35	0.30
hybrid PKS-NRPS enzyme	0.38	0.33	0.44	0.32	3	0.46	0.33	0.33	0.21	9	0.31	0.27	0.31	0.27
hypothetical protein	0.42	0.37	0.42	0.37	4	0.46	0.35	0.46	0.35	8	0.27	0.26	0.27	0.26
hypothetical protein	0.62	0.52	0.50	0.38	5	0.42	0.31	0.33	0.21	7	0.31	0.28	0.31	0.28
kynureninase	0.50	0.38	0.50	0.38	6	0.42	0.32	0.42	0.32	6	0.31	0.28	0.31	0.28
hypothetical protein	0.42	0.39	0.42	0.39	7	0.38	0.27	0.38	0.27	5	0.38	0.32	0.38	0.32
multidrug resistance protein; ATPase activity	0.50	0.39	0.50	0.39	8	0.38	0.28	0.38	0.28	4	0.35	0.30	0.35	0.30
hypothetical protein	0.46	0.35	0.47	0.40	9	0.42	0.33	0.42	0.33	3	0.42	0.35	0.42	0.35
nonribosomal peptide synthase	0.50	0.40	0.50	0.40	10	0.42	0.33	0.42	0.33	2	0.38	0.31	0.38	0.31
hypothetical protein	0.38	0.40	0.38	0.40	11	0.38	0.31	0.38	0.32	1	0.50	0.49	0.50	0.49

Comparison of the trees produced by these two concatenations to the reference tree shows only differences among strains within species-level clades (i.e. shallow internal nodes), the backbone being the same for all trees (Fig. 4). The ‘8-gene’ concatenation shows differences within the T. soudanense and T. rubrum species while The ‘2-gene’ concatenation shows differences in internal nodes for all taxa, with the exception of T. megninii (Fig. 4). The sequences of these two genes are available in the figshare repository (https://doi.org/10.6084/m9.figshare.14762394.v2).

Fig. 4.

Comparison of the reference tree with marker gene phylogenies. The reference tree is the one of Fig. 3. Marker gene trees were computed on DNA sequences under the GTRGAMMA model. Nodes are enumerated according to the reference tree.

Discussion

The taxonomic units within the T. rubrum complex have been changed and debated many times over the years, and phylogenetic studies on this complex have been unable to fully resolve the relationships between the different species [2, 3, 5]. These species remained largely polyphyletic in phylogenetic trees, resulting in species boundaries drawn with the support of other characteristics such as morphology, physiology, clinical manifestation and geographic distribution. In this study, phylogenomic analysis on 3105 single-copy genes resulted in a highly resolved tree with high bootstrap support for the individual species-clades. The species T. rubrum, T. violaceum and T. soudanense are confirmed in their taxonomic status. The T. rubrum morphotypes kuryangei and megninii and the T. violaceum morphotype yaoundei each group in a distinct clade with high support, suggesting their re-establishment as separate species following the phylogenetic species concept. These conclusions are also supported by both DNA and protein-based jackknife analyses.

Trichophyton kuryangei Vanbreuseghem & Rosenthal was first described in 1961 [32]. This is the oldest name and therefore the epithet with priority. Its type specimen is CBS 517.63^T. This species mostly causes tinea capitis (endothrix, sometimes with a sheath of spores at the base of the hair shaft) and is endemic to Central Africa (Burundi). The colonies are white and have a downy to cottony texture with radial grooves. The reverse is rather pale with colours ranging from white and cream to yellow. Microconidia are pyriform and can also be elongated and divided by two to three septa, resembling pseudomacroconidia.

Trichophyton megninii Blanchard was first described in 1896 [33]. Cultures were not made at that time, therefore no authentic material exists. Also, the protologue is not detailed enough to assure that this taxon is identical to our clade. In 1902, Bodin described a similar species, named T. roseum, which might be a synonym of T. megninii [34]. However, authentic material has been lost and the detailed description mentions both anthropophilic and zoophilic (chicken) strains. Similarly, in 1909, a species called T. rosaceum Sabouraud was described, also mentioning anthropophilic and zoophilic (bird) strains [35]. It is therefore unsure whether T. roseum and T. rosaceum are synonyms of T. megninii [1]. The neotype designated as T. megninii in this study is IHEM 13976=RV 67086; it originates from tinea corporis (tinea faciei), and was isolated in 1989 in Lisbon, Portugal, by de Sequeira. Trichophyton megninii mainly causes tinea corporis and tinea barbae (endothrix, sometimes with a sheath of spores at the base of the hair shaft). Geographically, it is restricted to the Mediterranean area (Portugal, Spain, Sardinia), but can occasionally be found in African regions (Burundi, Somalia). The colonies of T. megninii resemble those of T. rubrum. They are velvety to cottony and have radial grooves. The colony colour is white and the reverse is wine-red to brownish-red. Microconidia are clavate to pyriform. A unique and diagnostic feature of this species is its requirement of l-histidine for growth [1, 36]. In addition, a pink hue can be seen in the aerial mycelium of primary isolates of T. megninii. Nevertheless this feature is quickly lost with subculturing [36].

The original description of Trichophyton yaoundei was done by Cochet & Doby-Dubois in 1957 [37]. The type strain for this species is CBS 605.60^T. Trichophyton yaoundei primarily causes tinea capitis (endothrix) and is endemic to sub-Saharan Africa (Cameroon, Congo) and southeastern Africa (Mozambique). Primary cultures are cream to yellow or tan, becoming chocolate brown with time, and this is often accompanied by the diffusion of a brown pigment in the surrounding agar. The brown pigment is lost in subcultures. Repeated subculturing also leads to loss of pigmentation in the rest of the colony. Colonies are slow-growing and, in primary cultures, the texture is glabrous or waxy, becoming heaped and folded over time. Secondary cultures are often covered by a white, velvety aerial mycelium. Microconidia are rare, if formed they are pyriform. Macroconidia are not formed. Chains of chlamydospores are often present, resembling those of T. violaceum.

Phylogenomics has been used in the past to elucidate evolutionary history in fungi [38–40], but to our knowledge, this is the first phylogenomic study conducted on the T. rubrum complex. The clarification of the relationships in the T. rubrum complex is important for correct identifications in clinical settings and to better understand the epidemiology of these pathogens. Classical morphological analyses remain widely used for the identification of dermatophytes clinical isolates, but require a lot of expertise and are prone to errors. MALDI-TOF MS (matrix-assisted laser desorption/ionization time-of-flight mass spectrometry) is increasingly used for their identification even if there are currently some limitations when identifying closely related species. Packeu et al. (2019), 3] showed that MALDI-TOF MS can distinguish between T. rubrum, T. violaceum, T. soudanense and T. yaoundei, yet T. kuryangei could not be distinguished, whereas T. megninii was not tested. The universal barcode ITS region, which is considered the most informative gene for dermatophytes so far, is not sufficient for species delimitation. Potentially, two new protein-coding markers that were found in the set of core genes, a ubiquitin-protein transferase and a MYB DNA-binding domain-containing protein, could be more suitable to reconstruct the real species relationships. Further research is however required to test the usefulness of these markers.

The 18 newly sequenced genomes cover the five species and several phenotypic variants of the T. rubrum complex and were scaffolded on T. rubrum CBS 118892 (GCF_000151425.1). Although the latter assembly lacks the whole ribosomal region, it is the genome that is indicated as representative in the NCBI reference sequence database (RefSeq), and we therefore used it as the template for scaffolding our own genomes [41]. The completeness of the genomes that were assembled in this study, estimated by EukCC [13], is above 99.5% and comparable to the publicly available genomes. We further performed a metagenomic binning to avoid including any residual bacterial contamination of our assemblies.

The ribosomal region, consisting of the subunits 18S, ITS1, 5.8S, ITS2 and 28S, is present in all new genome assemblies, with some exceptions. Interestingly, this ribosomal region was absent in almost all of the publicly available genomes. This absence in public genomes could be the result of metagenomic binning, as this process is known to discard the region from assemblies due to variation in k-mer frequencies and in coverage [42]. This was also true for our assemblies, in which we had to manually add the ribosomal region with the main fungal bin for ten out of our 18 new assemblies.

The phylogenomic analyses showed a topologically stable segregation of the six taxa. Nevertheless, variations in the branching patterns within taxa, in internal nodes, were observed. These variations happened not only between the DNA and protein jackknife trees, but also between the protein jackknife trees and the tree obtained with the large protein phylogenomic analysis. This suggests that the signal present in the 3105 core genes might not be sufficient to resolve the phylogenetic relationships between strains within each species. These discrepancies might be due to incongruence between individual gene histories, which possibly differ from each other. This can be due to incomplete lineage sorting [42, 43], gene duplication and loss [44], or hidden paralogy [45]. This can also be due to the very close genetic distances between strains of the T. rubrum complex since the ANI between morphotypes is >99.5%. These discrepancies are only found within the six taxa and did not impact the conclusion of this study. The large DNA phylogenomic analysis (Fig. S1) shows the same topology as the large protein phylogenomic analysis and the jackknife trees, with high support values (bootstrap >90%). Hence, congruence between all the trees inferred in this study confirms our conclusions. Nevertheless, for future analysis of the T. rubrum complex, focussing on intra-taxa relationships, pangenomic analyses using accessory genes and identifying key traits of strains [46] should be considered.

The comparison of phylogenomic analyses (using bootstrap and jackknife support value), with single-gene phylogenies has already been used with success to define marker genes [47]. The low phylogenetic resolution of the core genes explained why a threshold of 0.40 in R-F values led to the selection of only 11 best genes. Within these 11 genes, concatenations were needed to obtain a sufficient phylogenetic signal in comparison to our reference tree. The best approach was to use a combination of eight genes, which is not convenient for marker genes that have to be Sanger-sequenced individually. The accurate relationships between species can however be obtained with the phylogeny of a concatenation of just two genes, namely a ubiquitin-protein transferase and a MYB DNA-binding domain-containing protein. These two markers correctly reflect the interspecific relationships, as deduced from the large phylogenomic analysis. Their systematic use can support and enhance the molecular identification of these taxa.

Conclusion

A new phylogeny of the T. rubrum complex, based on genomic data, suggests that six species should be recognized in this complex, namely T. rubrum, T. kuryangei, T. megninii, T. soudanense, T. violaceum and T. yaoundei. Genome-wide analyses also revealed two potential new gene markers to facilitate molecular differentiation of these dermatophytes species.