Comparative genome analysis and the genome-shaping role of long terminal repeat retrotransposons in the evolutionary divergence of fungal pathogens Blastomyces dermatitidis and Blastomyces gilchristii

Abstract

Blastomyces dermatitidis and Blastomyces gilchristii are cryptic species of fungi that cause blastomycosis, an often severe disease involving pulmonary infection capable of systemic dissemination. While these species appear morphologically identical, differences exist in the genetic makeup, geographical range, and possibly the clinical presentation of infection. Here, we show genetic divergence between the cryptic species through both a Blastomyces species tree constructed from orthologous protein sequences and whole genome single-nucleotide variant phylogenomic analysis. Following linked-read sequencing and de novo genome assembly, we characterized and compared the genomes of 3 B. dermatitidis and 3 B. gilchristii isolates. The B. gilchristii genomes (73.25–75.4 Mb) were ∼8 Mb larger than the B. dermatitidis genomes (64.88–66.61 Mb). Average nucleotide identity was lower between genomes of different species than genomes of the same species, yet functional classification of genes suggested similar proteomes. The most striking difference involved long terminal repeat retrotransposons. Although the same retrotransposon elements were detected in the genomes, the quantity of elements differed between the 2 species. Gypsy retrotransposon content was significantly higher in B. gilchristii (38.04–39.26 Mb) than in B. dermatitidis (30.85–32.40 Mb), accounting for the majority of genome size difference between species. Age estimation and phylogenetic analysis of the reverse transcriptase domains suggested that these retrotransposons are relatively ancient, with genome insertion predating the speciation of B. dermatitidis and B. gilchristii. We postulate that different trajectories of genome contraction led to genetic incompatibility, reproductive isolation, and speciation, highlighting the role of transposable elements in fungal evolution.

Introduction

Blastomycosis can be a serious fungal infection. It often presents as a pulmonary illness, but systemic disease can also occur after hematogenous dissemination. While direct cutaneous inoculation is a potential route of infection, blastomycosis is most often acquired through inhalation of spores from the environment. Once inhaled, spores undergo a thermally dependent morphological transition to yeast cells which cause disease (Pullen et al. 2022). It is estimated that approximately 50% of Blastomyces infections are asymptomatic and self-resolving. However, serious pulmonary infections causing acute respiratory distress syndrome do occur (Pullen et al. 2022). Without treatment the infection is able to disseminate to other body sites including skin, bone, genitourinary tract, and central nervous system (Pullen et al. 2022). Depending on patient demographics, mortality rates range from 4–22% (Khuu et al. 2014).

Historically, blastomycosis infections were ascribed to Blastomyces dermatitidis, which is endemic to the eastern half of North America, particularly in areas along the Ohio and Mississippi River Valleys, and the Great Lakes (Ashraf et al. 2020). Cases of African and Middle Eastern blastomycosis are attributed to newly described species Blastomyces percursus and Blastomyces emzantsi (Schwartz et al. 2021), while Blastomyces helicus has been implicated in a few blastomycosis cases in western Canada and the United States (Schwartz et al. 2019). Several studies demarcated genetic groups among clinical isolates of North American blastomycosis cases (Meece et al. 2010, 2011; Brown et al. 2013; McTaggart et al. 2016), which led to the description of 2 cryptic phylogenetic species, Blastomyces gilchristii and B. dermatitidis (Brown et al. 2013). Incidence rates of blastomycosis caused by B. dermatitidis and B. gilchristii suggest a broad and expanding geographic range of endemicity with localized regions of hyperendemicity, raising increased public health concern in North America regarding these fungi (Seitz et al. 2014; Brown et al. 2018; Ashraf et al. 2020; Mazi et al. 2023). Blastomyces spp. exist in the environment as a filamentous mold which produces spores but little is known about their specific environmental niche(s) owing to the fact that attempts to isolate these fungi from environmental sources have been largely unsuccessful (Pfister et al. 2011). Recent PCR-based soil surveillance has shown promise as an environmental monitoring technique (Jackson et al. 2021).

While morphologically indistinguishable, several differences exist between the cryptic species B. dermatitidis and B. gilchristii. Genetic differences and reproductive isolation have been described based on multilocus sequence typing (Brown et al. 2013) and multilocus microsatellite typing (Meece et al. 2011; McTaggart et al. 2016). The geographic ranges inhabited by the 2 species also differ. Based on genetic analysis of clinical isolates, B. dermatitidis are recovered from cases throughout the North American range, while B. gilchristii appears to be restricted to several Canadian provinces and a few north-eastern US states (McTaggart et al. 2016). While both are capable of causing blastomycosis (Dalcin et al. 2016; Frost et al. 2017; Laux et al. 2020; Kaplan et al. 2021), data suggests that variation in clinical disease presentation may exist. In studies that differentiate the 2 species, B. dermatitidis has been observed to be more likely associated with disseminated disease while B. gilchristii appears to primarily cause pulmonary-exclusive infections (Meece et al. 2013; Frost et al. 2016; Fritsche et al. 2023). Interestingly, in a recent pediatric survey, more than 90% of cases of blastomycosis in children were caused by B. gilchristii compared with 56% of cases in adults (Frost et al. 2017).

Undoubtedly, genomic differences underpin the genetic segregation, reproductive isolation and different clinical manifestations of B. dermatitidis and B. gilchristii. Nonsynonymous substitutions, duplications, or deletions are examples of alterations to functional genetic components, such as protein coding genes or regulatory regions, which could impact fungal physiology. However, a large portion of eukaryotic genomes contains repetitive DNA with no known functional role in the development or physiology of the organism (Palazzo and Gregory 2014). This includes short tandem repeats and transposable elements (TE) such as retrotransposons and DNA transposons. TEs are found in almost all eukaryotic genomes but the relative TE content varies widely from <1% to >80% with large disparities in genomic TE content evident across all levels of taxonomic hierarchy of plants, animals, fungi, and protists (Arkhipova 2018). The evolutionary histories of most organisms likely involved bursts of transposition activity followed by phases of inactivity (Oggenfuss et al. 2021). TEs are less constrained by selective pressures (Palazzo and Gregory 2014; Arkhipova 2018) and have been shown to contribute substantially to the genome differences between fungal species (Grandaubert et al. 2014; Mat Razali et al. 2019; Lorrain et al. 2021) and even lineages of the same species (Faino et al. 2016; Oggenfuss et al. 2021).

Contrary to its ascription as “junk DNA”, transposable elements in particular can produce subtle alterations with profound evolutionary consequences (Mat Razali et al. 2019). TE invasions promote genome size expansion and function as agents of insertional mutagenesis. Counterbalancing the impact of TE invasion, homologous and illegitimate recombination mediate genome contraction in order to constrain genome size (Macas et al. 2015; Faino et al. 2016). Insertions in coding regions may alter protein structure and function while targeting regulatory regions can alter gene expression by exerting epigenetic control (Bourgeois and Boissinot 2019; Mat Razali et al. 2019). Although TE insertions generate genetic variability, the substrate for evolution, most insertions are selectively neutral or even slightly deleterious (Arkhipova 2018). Nevertheless, variation in the TE genomic landscape is purported to play a role in adaptive evolution, influencing host range, pathogenicity, and other aspects of fungal lifestyle (Mat Razali et al. 2019). For example in Fusarium oxysporum, TEs are present in the promoter regions of several genes located on the pathogenicity chromosome that are expressed during plant infection (Schmidt et al. 2013). The genomes of plant pathogens Magnaporthe oryzae and Verticillium dahliae contain lineage-specific regions abundant in both TEs and secreted proteins and genes that mediate virulence, host recognition and evasion of host defenses (Huang et al. 2014; Faino et al. 2016; Mat Razali et al. 2019). TE activity is postulated to drive accelerated evolution in these genome compartments which causes lineage-specific changes in virulence effectors such as the multiple independent lineage-specific losses of the host recognition determinant AveI in V. dahliae (Faino et al. 2016) and the presence/absence polymorphism of 6 avirulence Avr-genes necessary for rice plant host immune response across various strains of M. oryzae (Huang et al. 2014). TEs also shape genome structure by facilitating chromosomal rearrangements and acting as substrates for ectopic recombination (Bourgeois and Boissinot 2019; Mat Razali et al. 2019). TEs drive speciation by promoting reproductive isolation through genome reorganization causing genomic incompatibilities that can lead to pre- or post-zygotic barriers and hybrid inviability (Serrato-Capuchina and Matute 2018).

Despite the severity of blastomycosis and the importance of B. dermatitidis and B. gilchristii as fungal pathogens, only 4 B. dermatitidis and 1 B. gilchristii genome assemblies are publically available (Muñoz et al. 2015). Initial whole genome analysis and characterization of B. dermatitidis and B. gilchristii described highly expanded genomes with large isochore regions of low GC content. These regions had much lower gene density but were enriched in Gypsy long terminal repeat retrotransposons (LTR-RTs). Gypsy LTR-RTs were expanded in B. gilchristii strain SLH14081 compared to B. dermatitidis strain ER-3, contributing to an 8.8 Mb genome assembly size difference (Muñoz et al. 2015). In this study, we have provided additional phylogenomic analysis of B. dermatitidis and B. gilchristii and expanded the genome characterization with 2 additional de novo assembled whole genome sequences of each species. In addition to confirming that the B. gilchristii genome is on average 8.6 Mb larger than that of B. dermatitidis, we compared average nucleotide identity (ANI), proteome functional classifications, and differing LTR-RT landscapes, providing insight into the role of repetitive elements in the evolution of these cryptic species.

Materials and methods

Isolates

Forty isolates were randomly selected from among those cultured from Ontario patient specimens received by Public Health Ontario's Laboratory between February 2016 and July 2022. Using multilocus sequence analysis (Brown et al. 2013), 22 isolates were identified as B. gilchristii and 18 as B. dermatitidis (Supplementary Table 1).

Whole genome sequencing and phylogenomic analysis

Illumina short-read whole genome sequencing was performed on 20 B. gilchristii isolates and 16 B. dermatitidis isolates (Supplementary Table 1). Briefly, DNA was extracted using the DNeasy PowerSoil Pro kit (Qiagen, Germantown, MD, USA). Libraries were prepared using the Illlumina DNA prep kit (Illumina, San Diego, CA, USA) and sequenced on a NextSeq 550 (Illumina) at 2 × 150 bp read length. Fastq files are available in the NCBI Sequence Read Archive (SRA) under Bioproject accession number PRJNA890593 (Supplementary Table 1).

Two isolates of each species were selected for whole genome sequencing using linked reads. Isolates 22,281 (B. dermatitidis) and 23,019 (B. gilchristii) were derived from respiratory specimens, bronchoalveolar lavage and sputum, respectively; while 23,166 (B. dermatitidis) and 22,264 (B. gilchristii) were cultured from tissue. For these isolates, DNA was extracted from fungal material cultured in SAB broth for 5 days at 28°C and 200 rpm. After filtering from broth using vacuum filtration and a sterile filter unit, fungal material was crushed in liquid nitrogen and processed using the MagAttract HMW DNA kit (Qiagen). Next generation sequencing was performed at Canada's Genomic Enterprise (CGEn), The Centre for Applied Genomics (Toronto, ON, Canada) using Chromium Genome library preparation v2 kit (10X Genomics, Pleasanton, CA, USA). Libraries were sequenced on a HiSeq 2500 (Illumina) at 2 × 150 bp read length following the manufacturer's instruction (Supplementary Table 1). For downstream analyses of de novo assembled genomes, fastq files were assembled using supernova software v 2.1.1 (Weisenfeld et al. 2018). Since haploid genomes were sequenced, pseudohap assemblies were used for subsequent analyses requiring de novo assembled genomes. Files are available under NCBI BioProject PRJNA890593 (Supplementary Table 1).

Species tree

For 3 B. dermatitidis, 3 B. gilchristii, 7 B. emzantsi, 1 Blastomyces parvus, 14 B. percursus, and 1 Blastomyces silverae for which genome assemblies were available (Supplementary Table 1), genomes were masked using RepeatMasker v4.1.1 (Smit et al. 2013–2015) with repeat libraries generated using RepeatModeler v2.0. (Flynn et al. 2020). Proteins were predicted using funannotate (https://github.com/nextgenusfs/funannotate) with BUSCOs from Histoplasma capsulatum and protein evidence from UniProt (The UniProt Consortium 2023). A species tree was generated using OrthoFinder (Emms and Kelly 2019) using proteins predicted by funannnotate. Briefly, the consensus species tree was inferred using STAG (Emms and Kelly 2018) from gene trees generated for orthogroups where all individuals were present. The species tree was rooted using STRIDE (Emms and Kelly 2017).

Whole genome single-nucleotide variant phylogenomic analysis

Whole genome single-nucleotide variant (SNV) phylogenomic analysis was conducted using the MycoSNP pipeline (Bagal, Phan, et al. 2022) with reference genome B. gilchristii SLH14081 (GCF_000003855.2) using sequence data generated above and an additional 22 B. dermatitidis and 5 B. gilchristii raw fastq files downloaded from the NCBI SRA (Supplementary Table 1). Prior to MycoSNP analysis, the raw fastq files from the 4 genomes sequenced with the Chromium Genome library preparation (22,264, 22,281, 23,019, 23,166) were processed with the longranger basic pipeline (10X Genomics) for barcode handling and removal. Maximum-likelihood (ML) phylogenetic analysis of concatenated whole genome SNVs generated by MycoSNP was conducted in IQ-TREE 2 with ascertainment bias correction, ModelFinder plus, and 1,000 ultrafast bootstrap approximations (Nguyen et al. 2015; Kalyaanamoorthy et al. 2017; Hoang et al. 2018; Minh et al. 2020). The tree was depicted in R v4.2.1 (R Core Team (2022)) using ggtree (Yu 2020).

Comparison of B. dermatitidis and B. gilchristii

For an in-depth comparison of B. dermatitidis and B. gilchristii genomes, we utilized the 4 de novo assembled genomes generated by the Chromium linked-read technology and B. gilchristii SLH14081 and B. dermatitidis ER-3 (Supplementary Table 1), which represent the most complete genome assemblies available for each of these species.

ANI comparisons and protein functional annotation

To assess inter- and intraspecific ANI of B. dermatitidis and B. gilchristii, pairwise comparisons of these genomes was performed using OrthoANI (Lee et al. 2016). Protein predictions were made using funannotate as described above with additional expressed sequence tag (EST) evidence generated from RNA reads from B. dermatitidis ATCC 26199 (PRJNA185598) assembled using Trinity (Grabherr et al. 2011). Funannotate functional annotation incorporated protein functional predictions from InterProScan v5.59-91.0 (Jones et al. 2014; Blum et al. 2021) and Phobius (Käll et al. 2007). Comparative analysis was performed using the funannotate compare function with Chi-square tests and Benjamini–Yekutieli correction for multiple comparisons (Benjamini and Yekutieli 2001) used to assess differences in functional protein counts between B. dermatitidis and B. gilchristii isolates.

Repetitive sequence analysis

LTR-RT library construction

Using genomes ER-3 and SLH14081, candidate LTR-RT were identified using LTR_Finder (Xu and Wang 2007) and LTRharvest (Ellinghaus et al. 2008) using default parameters. Using these candidates, LTR_retriever (Ou and Jiang 2018) was used to generate a set of 232 full length, intact LTR-RTs in ER-3 and 50 in SLH14081. TEsorter (Zhang et al. 2022) employing the GyDB (Llorens et al. 2011) was used to detect functional domains and to classify each of the full length, intact LTR-RTs as Gypsy, Copia, or unknown. The 2 sets of LTR-RTs were combined to create a redundant Blastomyces LTR-RT library. Using CD-HIT (Huang et al. 2010), the redundant Blastomyces LTR-RT library was clustered according to the 80–80–80 rule (80 bp long, 80% identity, 80% minimal alignment coverage for the longer sequence) (Wicker et al. 2007). A nonredundant Blastomyces LTR-RT library was constructed from representative sequences of each cluster.

Repetitive sequence annotation

RepeatMasker was run on all isolates using the nonredundant Blastomyces LTR-RT library with parameters -no_is -norna -nolow -div 40 -cutoff 225. Then, non-LTR-RT repetitive sequences or each genome were identified using RepeatModeler on the Blastomyces LTR-RT masked genomes from the previous step. The nonredundant Blastomyces LTR-RT library was combined with each isolate-specific RepeatModeler library for a final round of RepeatMasker to include low diversity and simple repeats as well transposable elements (TEs) including LTR-RTs. The statistical significance of comparisons genomic size data and element count data were evaluated by 2-tailed t-test and Chi-square goodness-of-fit test, respectively, with Benjamini–Yekutieli correction for multiple comparisons (Benjamini and Yekutieli 2001).

Annotation of solo-LTRs and complete, truncated, and nested LTR-RTs

We used the software program REANNOTATE (Pereira 2008) to detect solo-LTRs and complete (interrupted or not interrupted), truncated and nested LTR-RTs from all study isolates using the RepeatMasker output from a redundant Blastomyces LTR-RT library scan with parameters -no_is -norna -nolow -div 40 -cutoff 225 and a “fuzzy” file linking equivalent LTR-RT elements based on identical clustering of INT and LTR elements by CD-HIT.

Repetitive sequence divergence and dating

The output (.align files) from the final round of RepeatMasker was further parsed using scripts from https://github.com/4ureliek/Parsing-RepeatMasker-Outputs (Kapusta and Suh 2017) to examine the divergence of repetitive sequence elements compared to the family consensus sequence and employing a fungal substitution rate of 1.05 × 10⁻⁹ nucleotides/site/year (Castanera et al. 2016) to estimate element age.

The insertion ages of intact LTR-RTs of ER-3 and SLH14081 were calculated in LTR_retriever by estimating the divergence time of the 3′ and 5′ LTR regions using the Jukes-Cantor model for noncoding sequences (Jukes and Cantor 1969; Ou and Jiang 2018) and applying the aforementioned fungal substitution rate.

Phylogenetic analysis of reverse transcriptase (RT) domains

Genomic TE sequences were extracted from all 6 study isolates using the output from the final round of RepeatMasker and the out2seqs scripts from the TEsorter package (Zhang et al. 2022). Six-frame translations of the DNA TE sequences were conducted using transeq from EMBOSS (Rice et al. 2000) using the –clean parameter to specify “X” for stop codons. TEsorter and the GyDB were used to classify the translated TE sequences using parameters -st prot -db gydb -p 20 -cov 50 -eval 1e-5 and extract the RT domains. Sequences of Copia or select Gypsy families from each of the 6 study isolates were combined then analyzed by phylogenetic comparison employing MAFFT for alignment (Katoh and Standley 2013) and IQ-TREE 2 for ML phylogenetic analysis employing ModelFinder plus and 1,000 ultrafast bootstrap approximations (Nguyen et al. 2015; Kalyaanamoorthy et al. 2017; Hoang et al. 2018; Minh et al. 2020). Trees were depicted in R v4.2.1 (R Core Team (2022)) using ggtree (Yu 2020).

Syntenic comparison of the MAT locus

The MAT locus of SLH14081 is located on supercontig NW_003101669.1:991747-1048538 bounded by the AP endonuclease 2 (AP2) and SlaB genes. For each isolate, MAT locus region was subset from the assembly contigs based on blast searches against the SLH14081 MAT locus and single-copy genes APN2 and SlaB. Syntenic comparison of the MAT locus was visualized using the Artemis comparison tool (Carver et al. 2005) using comparison files generated from pairwise blast searches. Repetitive element annotation was added from the final round of RepeatMasker using the nonredundant Blastomyces LTR-RT library combined with each isolate-specific RepeatModeler library.

Results

Species tree and whole genome SNV phylogenomic analysis demonstrate that B. dermatitidis and B. gilchristii are genetically distinct

A species tree based on orthogroups constructed by OrthoFinder using protein sequences predicted from genome assemblies of 3 B. dermatitidis, 3 B. gilchristii, 7 B. emzantsi, 1 B. parvus, 14 B. percursus, and 1 B. silverae clearly delineated these 6 Blastomyces spp. (Fig. 1a). Based on 5,238 orthogroups, cryptic species B. gilchristii and B. dermatitidis were more closely related with lower branch support values (0.427, 0.419, respectively) compared to the other species B. percursus (0.733) and B. emzantsi (0.737). However, species tree branch support values for B. dermatitidis and B. gilchristii were higher (0.779) when analyzed independent of the other Blastomyces spp. using an expanded set of 6,962 orthogroups common to B. dermatitidis and B. gilchristii (Fig. 1b).

Fig. 1.

a) Species tree of the genus Blastomyces including B. dermatitidis (3), B. gilchristii (3), B. emzantsi (7), B. parvus (1), B. percursus (14), and B. silverae (1). The species tree was generated by OrthoFinder based on 5,238 orthogroups for which all individuals were present. Branch support values for each bipartition indicate the proportion of individual orthogroup species trees that contained that bipartition. b) OrthoFinder species tree generated for B. dermatitidis and B. gilchristii based on 6,962 orthogroups. c) Phylogenomic analysis of 268,539 whole genome SNVs from 40 B. dermatitidis and 28 B. gilchristii isolates by ML using the transversion model with ascertainment bias correction. A single B. gilchristii isolate from Quebec (*) is genetically divergent from other B. gilchristii isolates from Ontario, Minnesota and Quebec (opaque circle). The tree is drawn to scale with branch lengths measured in the number of substitutions per site. The percentage of trees of 1,000 ultrafast bootstrap approximations in which the associated taxa clustered together is shown next to the nodes.

Whole genome SNV phylogenomic analysis encompassing 268,539 sites separated B. dermatitidis (n = 40) and B. gilchristii (n = 28) into 2 phylogenetically distinct groups (Fig. 1c). The average interspecies genetic distance was 1.185 ± 0.072 which was greater than the average intraspecies genetic distance for either B. dermatitidis (0.304 ± 0.160) or B. gilchristii (0.103 ± 0.122). Genetic diversity among B. gilchristii isolates was lower than among B. dermatitidis isolates. Of note, a single isolate from Quebec was genetically divergent from other isolates from Ontario, Minnesota, and Quebec (Fig. 1c).

B. dermatitidis and B. gilchristii genomes were different sizes with reduced ANI but similar gene content

To further characterize the genetic differences between B. dermatitidis and B. gilchristii, we utilized 4 genomes de novo assembled from linked-read sequences representing 2 clinical isolates of B. dermatitidis (22,281 and 23,166) and 2 clinical isolates of B. gilchristii (22,264 and 23,019) isolates together with 2 well-characterized reference strains B. dermatitidis ER-3 and B. gilchristii SLH14081 (Muñoz et al. 2015). Together, these genomes confirmed the previously documented 8.6 Mb average genome size difference (Muñoz et al. 2015) between B. gilchristii (73.3–75.4 Mb) and B. dermatitidis (64.9–66.6 Mb) (Table 1). The percentage of Onygenales universal single copy orthologs (BUSCOs) detected in the genome assemblies was high (97.3–98.5%) (Table 1). The GC content of the genomes ranged from 35.8–37.3% (Table 1), however, the GC frequency distribution of the short-read sequences exhibited a bimodal distribution as previously described (Supplementary Fig. 1) (Muñoz et al. 2015). Furthermore, the ANI between species was 96.6%, while the intraspecies ANI ranged from 97.1–99.5% for B. dermatitidis and 98.7–99.4% for B. gilchristii (Supplementary Table 2). B. gilchristii genomes had more funannotate-predicted genes (8,467 ± 170) than B. dermatitidis genomes (8,218 ± 178), although this difference was not significant (P = 0.155). Despite these differences, comparison of counts of InterProScan classifications revealed no significant differences between the species (Supplementary Table 3).

Table 1.

Genome characteristics.

Isolate	Total assembly length	No. scaffolds	Scaffold N50	% BUSCOs in genome	GC content (%)	Genes
B. dermatitidis
ER3	66.61 Mb	25	5.55 Mb	98.4%	37.1%	8,175
22,281	64.96 Mb	2,550	0.080 Mb	98.3%	37.2%	8,066
23,166	64.88 Mb	2,739	0.071 Mb	98.3%	37.3%	8,414
B. gilchristii
SLH14081	75.40 Mb	100	2.44 Mb	98.5%	35.8%	8,329
22,264	73.53 Mb	2,981	0.063 Mb	98.1%	35.8%	8,416
23,019	73.25 Mb	3,475	0.048 Mb	97.3%	35.8%	8,657

Gypsy LTR-RTs distinguish B. dermatitidis and B. gilchristii genomes

A large proportion of the Blastomyces genomes was composed of repetitive elements, ranging from 59.4–60.3% (38.6–40.2Mb) of the B. dermatitidis genomes and 63.0–63.6% (46.6–47.5Mb) of the B. gilchristii genomes. The vast majority of the repetitive content was attributed to LTR-RTs, with less than 5% of the genomes ascribed to other transposable elements and simple repeats. Gypsy LTR-RTs were the predominant LTR-RT superfamily, comprising 47.5–52.1% of the genomes (Table 2; Fig. 2). Furthermore, the statistically significant 8.6 Mb average size difference between B. dermatitidis and B. gilchristii genomes (P = 0.0092) was attributed largely to LTR-RTs. LTR-RTs and specifically Gypsy LTR-RTs showed a corresponding 7.5 Mb and 7.1 Mb statistically significant size difference, respectively, between the species genomes (P = 0.0043; 0.0044) (Table 2).

Table 2.

Characterization of repetitive elements in Blastomyces genomes including sequence contributions of repetitive elements and TE superfamilies and number of LTR-RT elements characterized as complete, nested, truncated or solo-LTRs.

	B. dermatitidis				B. gilchristii
	ER-3	22,281	23,166	Avg. Bd	SLH14081	22,264	23,019	Avg. Bg	P adj
Genome size	66.61	64.96	64.88	65.48	75.40	73.53	73.25	74.06	0.0092
Total repeat content	40.19	38.69	38.59	39.16	47.70	46.67	46.49	46.95	0.0110
	(60.3%)	(59.6%)	(59.5%)	(59.8%)	(63.3%)	(63.5%)	(63.5%)	(63.4%)
LTR-RT	37.51	35.88	35.83	36.41	44.65	43.59	43.42	43.89	0.0043
	(56.3%)	(55.2%)	(55.2%)	(55.6%)	(59.2%)	(59.3%)	(59.3%)	(59.3%)
Gypsy	32.40	30.92	30.85	31.39	39.26	38.11	38.04	38.47	0.0044
	(48.6%)	(47.6%)	(47.5%)	(47.9%)	(52.1%)	(51.8%)	(51.9%)	(51.9%)
Copia	3.98	3.85	3.84	3.89	4.05	4.16	4.09	4.10	0.0981
	(6.0%)	(5.9%)	(5.9%)	(5.9%)	(5.4%)	(5.7%)	(5.6%)	(5.5%)
DNA transposon	0.20	0.23	0.22	0.22	0.35	0.35	0.35	0.35	0.0221
	(0.3%)	(0.4%)	(0.3%)	(0.3%)	(0.5%)	(0.5%)	(0.5%)	(0.5%)
RC-helitron	0.05	0.06	0.14	0.08	0.10	0.29	0.29	0.23	0.5028
	(0.1%)	(0.1%)	(0.2%)	(0.1%)	(0.1%)	(0.4%)	(0.4%)	(0.3%)
Unclassified	1.67	1.75	1.61	1.68	1.84	1.65	1.70	1.73	1
	(2.5%)	(2.7%)	(2.5%)	(2.6%)	(2.4%)	(2.2%)	(2.3%)	(2.3%)
Simple/low complex	0.75	0.77	0.78	0.77	0.76	0.77	0.77	0.77	1
	(1.1%)	(1.2%)	(1.2%)	(1.2%)	(1.0%)	(1.1%)	(1.1%)	(1.0%)
No. LTR-RT elements	22,075	25,715	25,863	24,551	26,608	30,427	30,573	29,203	<0.0001
No. complete LTRsa	540	329	332	400	476	409	392	426	0.2896
No. nested LTR-RTs	13,863	11,253	11,164	12,093	17,087	14,114	13,562	14,921	<0.0001
No. truncated LTR-RTs	20,001	23,648	23,810	22,486	24,165	27,759	27,899	26,608	<0.0001
No. solo-LTRs	2,585	3,092	3,100	2,926	3,329	3,659	3,704	3,564	<0.0001

P _adj values of statistically significant comparisons between B. dermatitidis and B. gilchristii are in bold. All sizes are in megabases (Mb).

^a Includes interrupted and non-interrupted LTR-INT-LTR combinations.

Fig. 2.

Amount of genome sequence contribution for classes of repetitive elements and non-repetitive content of B. dermatitidis and B. gilchristii genomes. The quantity (Mb) and percentage of each genome comprised of Gypsy LTR-RTs is indicated.

To facilitate LTR-RT annotation and comparison, we constructed a Blastomyces nonredundant LTR-RT library based on 232 intact LTR-RTs from B. dermatitidis ER-3 and 50 intact LTR-RTs of B. gilchristii SLH14081 which was used for identification of LTR-RT elements in each for the 6 genomes. Examination of LTR-RT elements comprising on average >10,000 bp in the genome showed that B. dermatitidis and B. gilchristii possessed the same LTR-RT elements but the quantities of several elements differed between the species. Of the 125 elements examined, statistically significant differences were observed among 19 Gypsy, 12 Copia, and 4 unknown LTR-RT elements. Sixteen elements were on average statistically more abundant in B. dermatitidis contributing an average of 0.97 million more bases to the B. dermatitidis genomes compare to the B. gilchristii genomes (Fig. 3). Conversely, 19 elements were statistically more abundant in B. gilchristii adding on average 7.97 million more bases to the B. gilchristii genomes compared to the B. dermatitidis genomes (Fig. 3). Of note, the greatest size differential was observed with elements Bder197_INT, Bgil46_INT, and Bder129_INT, which exhibited on average 1.48, 1.68, and 2.39 million more base pairs, respectively, in the B. gilchristii genomes compared to the B. dermatitidis genomes.

Fig. 3.

Average genome sequence contribution of 125 LTR-RT (72 Gypsy, 41 Copia, 12 unknown) elements to B. dermatitidis and B. gilchristii genomes. Analysis included elements with an average of at least a 10,000 bp genome contribution in at least 1 species. Elements with a statistically significant difference in the amount of genome sequence contribution between the 2 species are designated with black bars along the x-axis.

LTR-RTs were additionally characterized by estimating the number of complete (interrupted and non-interrupted), nested, and truncated LTR-RTs and solo-LTRs in each of the Blastomyces genomes. This approach involved re-annotating the LTR-RTs in the genomes using the redundant Blastomyces LTR-RT library, which contains information to associate the LTR and internal components of the LTR-RTs, and a linker file to identify equivalent LTR-RT elements based on clustering of INT and LTR elements by CD-HIT. Between 22,075 and 30,573 LTR-RT elements were annotated per genome (Table 2). Few elements comprised complete (interrupted and non-interrupted) LTR-RTs. Instead, an overwhelming number of LTR-RT elements were present as nested and/or truncated. In addition, the number of solo-LTRs per genome ranged from 2,585–3,704. Statistically significantly more nested and truncated LTR-RTs and solo-LTRs were detected in the B. gilchristii genome compared to B. dermatitidis (Table 2).

Blastomyces TEs, including LTR-RTs, are highly divergent

Divergence of TEs was tabulated by comparing TE family members to their respective consensus sequences. The TE landscape graphs presented as stacked histograms of percent divergence suggested that Blastomyces TEs, including LTR-RTs, were highly diverged. Across all genomes, more than 50% of TE content was highly divergent (≥12%), while less than 10% exhibited a high degree of similarity (≤5% divergence) (Fig. 4). Applying a fungal substitution rate of 1.05 × 10⁻⁹ nucleotides/site/year (Castanera et al. 2016) suggested that the vast majority of Blastomyces TEs were ancient (Fig. 4) and predate the separation of B. dermatitidis and B. gilchristii estimated at 1.9 million years ago (MYA) (95% HPD 0.5–4.4 MYA) (McTaggart et al. 2016). The peaks of the histograms of the TE divergence landscapes of the B. dermatitidis (Fig. 4, a–c) isolates were shifted left compared to those of B. gilchristii (Fig. 4, d–f), suggesting that the B. gilchristii TEs were generally more divergent compared to those of B. dermatitidis.

Fig. 4.

Divergence landscape of TEs of B. dermatitidis ER-3 a) 22,281, b) and 23,166 c) and B. gilchristii SLH14081, d) 22,264, e), and 23,019 f). Stacked histograms express the amount of genome sequence categorized into bins of percent divergence of TE family members compared to their respectively library consensus sequence. Element dating (million years ago, MYA) is calculated using a fungal substitution rate of 1.05 × 10⁻⁹ nucleotides/site/year (Castanera et al. 2016).

We also estimated the insertion ages of intact LTR-RTs in the 2 most contiguous genomes, B. dermatitidis ER-3 and B. gilchristii SLH14081. Since the 3′ and 5′ LTR regions of an LTR-RT are identical at the time of insertion, the age of an element can be estimated based on the divergence of these sequences (Ou and Jiang 2018; Rao et al. 2018). Of note, 232 intact LTR-RTs were identified in B. dermatitidis ER-3 while only 50 were detected in the B. gilchristii SLH14081 genome. The insertion ages of the intact LTR-RTs of B. gilchristii were on average older (median: 45 MYA; mean: 41 MYA) than that of B. dermatitidis intact LTR-RTs (median: 38 MYA, mean: 34 MYA) (Fig. 5). Furthermore, there were 28 intact LTR-RTs (2 Gypsy, 21 Copia, 5 unknown) with an insertion age < 5 MYA in the B. dermatitidis ER-3 genome but only 4 (2 Copia, 2 unknown) in B. gilchristii SLH14081 (Fig. 5). Based on TEsorter analysis, none of the intact LTR-RTs in either species contained all of the functional domains required for transposition.

Fig. 5.

Estimated insertion age (million years ago, MYA) of a) 232 intact LTR-RTs of B. dermatitidis ER-3 and b) 50 intact LTR-RTs of B. gilchristii SLH14081. Insertion age is calculated based on the sequence divergence between 5′ and 3′ LTR regions of the elements.

Phylogenetic analysis of reverse transcriptase domains of B. dermatitidis and B. gilchristii LTR-RT families cluster together

Phylogenetic analysis was used to analyze the evolutionary relationships between reverse transcriptase domains of the LTR-RTs. Due to the sheer number of LTR-RT elements, phylogenetic analysis was only feasible on select subsets of elements. Copia LTR-RT reverse transcriptase domains from all isolates were selected for analysis (Fig. 6a) because there were a relatively large number of young (<5 MYA), intact Copia LTR-RT elements present in ER-3 and SLH14081. As well, we analyzed reverse transcriptase domains from 3 sets of Gypsy elements [Bgil46_INT (Fig. 6b), Bder129_INT (Fig. 6c), Bder197_INT (Fig. 6d)] previously shown to be present in statistically significantly greater quantities (Mbp) in the B. gilchristii genomes compared to the B. dermatitidis genomes. For all sets of reverse transcriptase domains examined, the phylogenies did not delineate according to species. Although some small species-exclusive clades were observed, most clades contained sequences from both B. dermatitidis and B. gilchristii (Fig. 6).

Fig. 6.

Cladograms showing ML phylogenetic relationships inferred from alignments of reverse transcriptase domains of Copia LTR-RT elements a) and 3 families of Gypsy LTR-RT elements Bgil46_INT b), Bder129_INT, c) and Bder197_INT d). Analyses of Copia and Gypsy Bgil46_INT elements employed reverse transcriptase domains from all 6 isolates (a and b). For simplicity, analyses of Gypsy Bder129_INT and Bder197_INT utilized reverse transcriptase domains from SLH14081 and ER-3 only (c and d).

Presence of TEs around the mating-type locus is variable

Blastomyces spp. are heterothallic with the mating type determined by the presence of either the alpha-box (MAT1-1) or HMG domains (MAT1-2) at the MAT locus, located between the APN2, cytochrome c oxidase subunit 6a (COX13) and SlaB genes (Li et al. 2013). Syntenic analysis of the MAT locus revealed a size variation at the MAT locus due to the presence of transposable element sequence content as previously documented (Li et al. 2013) with B. gilchristii isolates possessing a much larger MAT locus due to the presence of Gypsy LTR-RT sequence content compared to the B. dermatitidis isolates (Supplementary Fig. 2).

Discussion

In this study, we provide a genomic analysis and comparison of B. dermatitidis and B. gilchristii, the causative agents of the serious fungal disease blastomycosis, complementing the paucity of publically available genome assemblies and genomic studies available for these organisms (Muñoz et al. 2015; Carignan et al. 2021; Bagal, Ireland, et al. 2022). Both a species tree based on orthogroups and whole genome SNV phylogenomic analysis demonstrate the phylogenetic separation between B. dermatitidis and B. gilchristii. However, there were no significant differences in gene functional classifications, which is consistent with the identical morphologies of these cryptic species. The most substantial genomic different between the 2 species is the quantity of TE content, particularly LTR-RT content, yielding a B. gilchristii genome that is ∼8 Mb larger than the B. dermatitidis genome. We demonstrate that the LTR-RTs are relatively ancient and posit that the 2 species experienced different evolutionary trajectories of genome contraction which facilitated and/or maintained reproductive isolation.

Originally differentiated by MLST of 7 nuclear loci (Brown et al. 2013), we show a phylogenetic separation between B. dermatitidis and B. gilchristii using more robust genetic methods, namely a consensus Blastomyces species tree generated from phylogenetic analysis of individual orthogroups and phylogenomic analysis of whole genome SNV data from a dataset of 68 sequenced B. dermatitidis and B. gilchristii isolates. Average interspecies genetic distance based on the SNV data was greater than intraspecies genetic distances, demonstrating the genetic differentiation and divergence between species and supporting their status as separate, albeit cryptic species. Although genetic diversity among B. gilchristii isolates was lower than among B. dermatitidis isolates, a single isolate from Quebec suggests that the genetic diversity among B. gilchristii isolates may prove to be greater as more sequence data from isolates from geographically diverse location becomes available.

To facilitate more detailed genomic comparison and analysis, we sequenced and assembled genomes of 2 B. dermatitidis (22,281, 23,166) and 2 B. gilchristii (22,264, 23,019) isolates using Illumina short-read sequencing with the 10X Genomics’ Chromium technology, which uniquely barcodes fragments of the next generation sequencing libraries to aid in de novo assembly (Weisenfeld et al. 2018). Although the 10X Genomics’ Chromium libraries produced genomes that were more fragmented than the Sanger-sequenced genomes of ER-3 and SLH14081 (Muñoz et al. 2015), the genome sizes were similar and the representation of core eukaryotic genes was high, suggesting that the genomes are very nearly complete. Illumina linked reads have been used to generate high quality genome assemblies for other species (Ott et al. 2018; Ozerov et al. 2018; Wang et al. 2020). Although long-read sequencing technologies such as PacBio and Nanopore are an attractive alternative to Illumina linked-read assemblies, they have both a higher cost and error rate. In the future, a hybrid approach involving both short-read and long-read sequencedata would greatly assist the generation of more complete genome assemblies for Blastomyces spp.

The genomes of B. dermatitidis and B. gilchristii were distinctly different based on genome size and ANI. We confirmed a notable genome size difference between species, with B. gilchristii genomes (73.3–75.4 Mb) approximately 8 Mb larger than B. dermatitidis genomes (64.9–66.6 Mb). We also verified the bimodal GC frequency distribution of the short-read fragments of both species suggestive of large isochore-like genomic regions of high and low GC content (Muñoz et al. 2015). Likewise, intraspecies ANI were higher than interspecies ANI values. The interspecies ANI values (96.53–96.60%) were slightly higher than the recommended cutoff of 95–96% for species demarcation (Lee et al. 2016), which is consistent with their status as cryptic species. Diminished whole genome synteny between B. dermatitidis ER-3 and B. gilchristii SLH14081 was described by (Muñoz et al. 2015), largely confined to the GC-poor, repeat-rich regions of the genome.

Gene annotation and functional classification revealed that B. dermatitidis and B. gilchristii displayed no significant differences in gene functional quantification, which was expected given their similar morphologies, ecology, and clinical manifestations. Although not examined in this study, phenotypic variation between species can be caused by genetic alterations to regulatory regions that impact gene expression. Interestingly, variability in the promoter region of virulence factor BAD1 has been documented between the species (Meece et al. 2010; Kaplan et al. 2021). In addition, many SNV differences exist across the genomes of B. dermatitidis and B. gilchristii. Likewise, nonsynonymous SNV differences in coding regions may impact gene function, which could also lead to subtle phenotypic differences between the species.

TEs are ubiquitous in fungal genomes although their genomic proportion varies widely (Muszewska et al. 2011), typically ranging from 0.02–30% (Castanera et al. 2016) but can be as high as 74% (Frantzeskakis et al. 2018). Blastomyces spp. are unique in that a large proportion (∼60%) of their genomes are comprised of TEs. Similar to other fungi, the vast majority of TEs of both species examined in this study consisted of class I LTR-RT elements, particularly Gypsy elements (Muszewska et al. 2011; Grandaubert et al. 2014; Castanera et al. 2016; Kirkland et al. 2018; Rao et al. 2018; Oggenfuss et al. 2021). Most elements were remnant fragments with few complete LTR-RTs detected. Degraded TE remnants are common in fungal genomes (Muszewska et al. 2019). The genome size difference between B. dermatitidis and B. gilchristii was attributed almost completely to the presence of excess LTR-RTs in the B. gilchristii genome. The 2 species possessed the same LTR-RT elements but the quantities of certain elements contained in their respective genomes differ; some elements were more abundant in B. dermatitidis while others were more abundant in B. gilchristii.

Several studies describe genomic TE invasions in fungi which increased genome size and impacted evolution and speciation (Grandaubert et al. 2014; Castanera et al. 2016; Faino et al. 2016; Oggenfuss et al. 2021). Whereas many studies describe relatively young TEs as a results of recent transposition (Castanera et al. 2016; Horns et al. 2017; Rao et al. 2018; Lorrain et al. 2021; Oggenfuss et al. 2021), the TEs of B. dermatitidis and B. gilchristii described in this study were relatively ancient. We hypothesize that TE genome invasion and expansion accompanied or was possibly instrumental in the evolution of a B. dermatitidis/gilchristii common ancestor, differentiating it from other members of the genus and from other Ajellomycetaceae and Onygenales. This is supported by the observation that, despite approximately the same number of genes (Muñoz et al. 2015), the genomes of B. dermatitidis and B. gilchristii are more than twice the size of other closely related species: B. parvus (27.8Mb), B. percursus (32.3Mb) (Dukik et al. 2017), B. silverae (30.4 Mb), Emmonsia crescens (30.7 Mb) (Muñoz et al. 2015), B. emzantsi (34.05 Mb) (Schwartz et al. 2021), H. capsulatum (32.5 Mb) (Voorhies et al. 2022), Paracoccidioides brasiliensis (30.0 Mb), Paracoccidioides lutzii (32.9 Mb) (Desjardins et al. 2011), Coccidioides immitis (28.9 Mb), and Coccidioides posadasii (27 Mb) (Neafsey et al. 2010).

Notwithstanding the genome size difference between B. dermatitidis and B. gilchristii primarily attributed to Gypsy LTR-RT elements, several lines of evidence suggest that the insertion of these TEs predates the species’ divergence, negating LTR-RT transpositional proliferation as the cause of the genome size difference between the species. The LTR-RT families are highly divergent suggesting that they are ancient, predating the speciation of B. dermatitidis and B. gilchristii, estimated at 1.9 MYA (95% HPD 0.5–4.4 MYA) (McTaggart et al. 2016). The ancient age of Blastomyces LTR-RTs was estimated by 2 methods: (1) divergence from the repeat family consensus sequence applied to all elements and (2) divergence of the 3′ and 5′ LTR regions of intact LTR-RTs only, with similar results. Both dating methods are commonly applied to LTR-RTs for age estimation, although shortcomings are noted for both methods (Maumus and Quesneville 2014; Kapusta and Suh 2017; Jedlicka et al. 2020). The age estimates for the LTR-RTs are dependent on both the sequence divergence and the substitution rate parameter and binned in 5 MYA increments to account for error. Consequently, the age estimates should be interpreted with caution because LTR-RTs are likely subject to higher mutation rates than the neutrally evolving sequences since error-prone transcription and reverse transcription precede new insertions. Likewise, LTR-RTs are susceptible to repeat-induced point mutation, which would elevate the mutation rate. Hence, the age of these TEs may be overestimated.

Nevertheless, additional observations support the hypothesis that the proliferation of Gypsy LTR-RTs predated the speciation of B. dermatitidis and B. gilchristii. Few of the LTR-RT elements in are ER-3 and SLH14081 are intact and none contain all of the functional domains for transposition, suggesting that none are currently active. Incomplete and nonfunctional LTR-RTs are common in fungi (Muszewska et al. 2011). Furthermore, TE content in B. dermatitidis and B. gilchristii is compartmentalized into large isochore regions of high and low GC content (Muñoz et al. 2015). This suggests a period of time since insertion for chromosomal rearrangements and purifying selection of deleterious TE insertions in exonic sequences to establish TE compartmentalization, and it contrasts with the lack of compartmentalization observed in genomes such as Blumeria graminis with a recent TE expansion (Frantzeskakis et al. 2018; Oggenfuss et al. 2021). Finally, phylogenetic analysis of reverse transcriptase domains of 3 Gypsy LTR-RT elements and the Copia LTR-RT elements shows no strong phylogenetic distinction between species. The analysis suggests that transposition occurred before species divergence, followed by independent evolution of the domains in each of the species.

Genome size is determined by the opposing forces of genome expansion often mediated by TE transposition and genome contraction that occurs when sequences are excised by homologous ectopic recombination between repeat copies or nonspecific DNA loss through double-strand break repair (Devos et al. 2002; Macas et al. 2015; Faino et al. 2016). We postulate that the difference in genome size and LTR-RT content between the B. dermatitidis and B. gilchristii genomes is due to different evolutionary rates and trajectories of genome contraction. This is consistent with the observation that the species contain differing amounts of the same LTR-RT elements. Although genome contraction is difficult to detect through whole genome sequence analysis, the abundance of solo-LTRs in genomes of both species is evidence for extensive genome contraction because solo-LTRs are the byproduct of homologous recombination of the 3′ and 5′ LTR regions either within or between repeat copies leading to the excision of the internal sequence and 1 of the LTR regions (Macas et al. 2015). Different rates of genome contraction may also explain why the LTR-RT elements appeared more divergent and older in B. gilchristii. In a study of TEs in different bird lineages, flightless birds had larger genomes and older TEs compared to flying birds due to relatively slower removal of DNA from their genomes (Kapusta et al. 2017). TEs are known to mediate reproductive segregation leading to speciation by creating chromosomal rearrangements through homologous recombination and alternative transposition that prevent effective meiotic recombination between lineages (Gray 2000; Grandaubert et al. 2014; Serrato-Capuchina and Matute 2018). The MAT locus in Blastomyces is an example of alternate genome locus structure in B. dermatitidis and B. gilchristii caused by variable TE presence (Li et al. 2013). Thus, we hypothesize that following TE and specifically LTR-RT expansion concurrent with the evolution of a B. dermatitidis/gilchristii common ancestor, different trajectories of genome contraction contributed to the evolution B. dermatitidis and B. gilchristii by facilitating and/or maintaining reproductive isolation.

Phylogeographic analysis suggests that B. dermatitidis is present throughout the eastern half of North America whereas B. gilchristii is restricted to select Canadian provinces and some northern US states (McTaggart et al. 2016). With species divergence estimated at 1.9 MYA (95% HPD 0.5–4.4 MYA) during the Pleistocene epoch, it has been postulated that North America glaciations during this period created biotic refugia, providing an impetus for speciation (McTaggart et al. 2016). Under this scenario, repeated glaciation would have caused geographic isolation of subsets of populations creating the opportunity for alternate trajectories of genome contraction leading to speciation. The more southerly B. dermatitidis ancestors may have had more opportunity for recombination necessary for genome contraction (Devos et al. 2002) while the northern B. gilchristii ancestors remained frozen in ice for thousands of years. This may explain why the B. dermatitidis genomes are smaller than those of B. gilchristii.

In conclusion, we characterize genomic differences between cryptic species B. dermatitidis and B. gilchristii, the causative agents of blastomycosis. Future research involving additional and more complete genome sequences will be invaluable in understanding the genetic traits that correlate with the clinical outcomes and ecological variations of these 2 important pathogens. De novo sequencing and analysis of multiple isolates of each species confirms that ancient TE content, in particular Gypsy LTR-RT content, is a major genomic difference between B. dermatitidis and B. gilchristii. Differential TE genomic components likely reinforce reproductive isolation of the 2 species and highlight the importance of the “repeatome” in fungal speciation. However, TEs are also postulated to modulate gene expression usually through TE-mediated transcriptional repression or silencing (Castanera et al. 2016; Torres et al. 2021). Future studies on genes proximal to TEs represent a promising avenue of research to further elucidate the full impact of the differential TE content on species-specific adaptations, fungal lifestyle traits, and virulence.

Data availability

Sequencing raw reads, de novo assemble genomes and genome annotations are available in the NCBI Sequence Read Archive and GenBank under Bioproject number PRJNA890593. Accession numbers are listed in Supplementary Table 1.

Supplemental material available at G3 online.

Funding

Sequencing service for 4 Blastomyces isolates using the Chromium Genome library preparation for linked-read Illumina sequencing was provided in-kind by Canada's Genomic Enterprise (CGEn), The Centre for Applied Genomics (Toronto, ON, Canada) as part of the CanSeq150 program.