An Update on Trichoderma Mitogenomes: Complete De Novo Mitochondrial Genome of the Fungal Biocontrol Agent Trichoderma harzianum (Hypocreales, Sordariomycetes), an Ex-Neotype Strain CBS 226.95, and Tracing the Evolutionary Divergences of Mitogenomes in Trichoderma

Members of the genus Trichoderma (Hypocreales), widely used as biofungicides, biofertilizers, and as model fungi for the industrial production of CAZymes, have actively been studied for the applications of their biological functions. Recently, the study of the nuclear genomes of Trichoderma has expanded in the directions of adaptation and evolution to gain a better understanding of their ecological traits. However, Trichoderma’s mitochondria have received much less attention despite mitochondria being the most necessary element for sustaining cell life. In this study, a mitogenome of the fungus Trichoderma harzianum CBS 226.95 was assembled de novo. A 27,632 bp circular DNA molecule was revealed with specific features, such as the intronless of all core PCGs, one homing endonuclease, and a putative overlapping tRNA, on a closer phylogenetic relationship with T. reesei among hypocrealean fungi. Interestingly, the mitogenome of T. harzianum CBS 226.95 was predicted to have evolved earlier than those of other Trichoderma species and also assumed with a selection pressure in the cox3. Considering the bioavailability, both for the ex-neotype strain of the T. harzianum species complex and the most globally representative commercial fungal biocontrol agent, our results on the T. harzianum CBS 226.95 mitogenome provide crucial information which will be helpful criteria in future studies on Trichoderma.

Mitogenomes are ideal tools for the evolutionary analysis of eukaryotes due to their (1) small size, (2) high copy numbers, (3) limited recombination, and (4) high mutation rates leading to independent evolutions of nuclear genomes [2,5]. When considering the ecological features of fungi, such as a broad diversity of lineages and lifestyles in nature, fungal mitogenomes can provide crucial information on evolutionary adaptations and divergences of the mitochondria in different environments [1,6]. According to the recent rapid developments of high-throughput sequencing technologies, fungal organelle genome databases have increasingly accumulated, however, these are still less abundant than those of animals and plants [4,7]. Thus far, most studies of fungal mitogenomes have focused on the (1) sequencing and characterization of individuals and (2) the elucidation of their phylogenetic positions within its relatives to support the potentials for one of the significant molecular markers identifying fungal species. Until recently, few studies have reported on comparative analyses of mitogenomes, ranging from a taxa level for specific fungal groups, under evolutionary viewpoints [8][9][10][11].
When considering the species complex of T. harzianum, initially it was suggested to be an aggregate species including several cryptic species [21], genomic studies of T. harzianum can provide more knowledge of their adaptive evolution and functional mechanisms in particular niches with both of systematic-and ecological perspectives [19,20]. In terms of the T. harzianum CBS 226.95, an ex-neotype strain of the T. harzianum species complex and the most globally representative commercial fungal biocontrol agent, the nuclear genome of this fungus has already been sequenced (GenBank accession no. GCA_003025095) [19], but the mitogenome has not. Here, we present a complete de novo mitogenome of the fungus T. harzianum CBS 226.95 with genomic characterization, comparative analyses, and the tracing of time-scaled evolutionary divergences.

DNA Extraction and Genome Sequencing
The ex-neotype strain T. harzianum CBS 226.95 [22] was obtained from the Westerdijk Fungal Biodiversity Institute (formerly the Centraalbureau voor Schimmelcultures (CBS) Fungal Biodiversity Centre, The Netherlands). The fungus was cultured on Difco TM potato dextrose medium (Difco Laboratories Inc., Detroit, MI, USA) at 25 • C, and the obtained mycelia were used as a source of genomic DNA. Total genomic DNA was extracted using a Plant/Fungi DNA Isolation Kit (Sigma-Aldrich Co. Ltd., St. Louis, MO, USA) according to the manufacturer's instructions and followed by further purification using Phenol-Chloroform (Sigma-Aldrich Co. Ltd., USA) [23]. A quality/quantity of the extracted DNA was measured using the Qubit assay in a Qubit TM 3.0 Fluorometer (Thermo Fisher Scientific Inc., Waltham, MA, USA).
The genomic DNA was used to construct a 20 kb insert SMRTbell ® DNA library in a BluePippin TM size-selection system (Pacific Biosciences, Menlo Park, CA, USA). Then, it was sequenced on a single molecule real-time (SMRT) sequencing platform by a PacBio RS-II DNA sequencer with P6 polymerase-C4 sequencing chemistry (Pacific Biosciences, USA) [24] at the Génome Québec Innovation Centre (Montréal, QC, Canada).

De Novo Assembly and Annotation
Only raw reads with a read-quality higher than 0.85 were used for de novo mitogenome assembly using the hierarchical genome-assembly process (HGAP) [25] in the SMRT TM pipeline (Pacific Biosciences, USA). The de novo mitogenome assembly's accuracy was verified using the P-mapping module [26], and the circularity was confirmed using Gepard [27].
Maximum likelihood (ML) [46] analysis was conducted in RAxML (v.8.2) with 1000 bootstrap replicates [47], and the bootstrap (BS) values were displayed on the nodes of the constructed phylogenetic tree. Bayesian inference (BI) [48,49] analysis was conducted in MrBayes (v.3.2) [50] under the Markov chains Monte Carlo (MCMC) algorithm (three heated and one cold chain, with a heating coefficient 0.1) for 2 × 10 6 generations (sampling for every 400 gen.). After a burn-in for the first 25% of the sampled trees, the remaining trees were used to construct the consensus BI tree with values of Bayesian posterior proba-bilities (BPPs). The stationarity of BI analysis was determined by calculating the average standard deviation for split frequencies (<0.01).
The non-synonymous substitution rate (Ka), synonymous substitution rate (Ks), and Ka/Ks ratio were calculated on the pairwise alignments of each target gene between species using DnaSP (v.6.12) [53], and obtained values were plotted using ggplot2 [40] in R (v.3.4) [41]. The codon sites evolving under the episodic diversifying positive selection were detected using the mixed effects model of evolution (MEME) method [54] in HyPhy (v.2.1) [55], and sites under the p-value threshold of 0.1 (p < 0.1) were considered.

Genomic Features of the T. harzianum CBS 226.95 Mitogenome
After filtering low-quality reads, a total of 1,806,923,812 read bases were used as seeds for the assembling process, and a complete mitogenome of T. harzianum CBS 226.95 was assembled de novo with a length of 27,632 bp (coverage depth, 573.59×; GenBank accession no. MN564945).
A closed-circular DNA molecule of the T. harzianum CBS 226.95 mitogenome was composed of A (36.1%), T (36.4%), G (15.1%), and C (12.4%), with the AT content (72.45%) approximately 2.63 times higher than the GC content. The asymmetric biases of these nucleotide compositions were inferred from the skewness, each for the negative value of AT-skew and the positive value of GC-skew, it showed more frequency of thymine (T) and guanine (G) than adenine (A) and cytosine (C) in the forward strand. The coding and non-coding regions of the entire length of the mitogenome accounted for approximately  79.25% and 20.75%, respectively (Table S1). All genes constituting the coding regions were predicted to be transcribed in a clockwise direction on the same forward strand (Figure 1).
A closed-circular DNA molecule of the T. harzianum CBS 226.95 mitogenome was composed of A (36.1%), T (36.4%), G (15.1%), and C (12.4%), with the AT content (72.45%) approximately 2.63 times higher than the GC content. The asymmetric biases of these nucleotide compositions were inferred from the skewness, each for the negative value of ATskew and the positive value of GC-skew, it showed more frequency of thymine (T) and guanine (G) than adenine (A) and cytosine (C) in the forward strand. The coding and noncoding regions of the entire length of the mitogenome accounted for approximately 79.25% and 20.75%, respectively (Table S1). All genes constituting the coding regions were predicted to be transcribed in a clockwise direction on the same forward strand ( Figure  1). All 14 conserved core PCGs were composed of three ATP synthases (atp6, atp8, and atp9), one apocytochrome b (cob), three cytochrome c oxidases (cox1, cox2, and cox3), and seven NADH dehydrogenases (nad1, nad2, nad3, nad4, nad4L, nad5, and nad6), without any introns capable of splitting each of their coding sequences (CDS) to the exon-intron structure. Two ORFs that were predicted without assigned functions, orf408 and orf170, were also found as a form of non-freestanding ORF. The orf408 gene was entirely located within a coding region of the rps3 gene for ribosomal protein S3 and rrnL gene for a large subunit ribosomal RNA, and the orf170 gene was identified completely within a conserved domain for homing endonuclease (HE). All of the core PCGs and ORFs were initiated with the All 14 conserved core PCGs were composed of three ATP synthases (atp6, atp8 and atp9), one apocytochrome b (cob), three cytochrome c oxidases (cox1, cox2 and cox3), and seven NADH dehydrogenases (nad1, nad2, nad3, nad4, nad4L, nad5 and nad6), without any introns capable of splitting each of their coding sequences (CDS) to the exon-intron structure. Two ORFs that were predicted without assigned functions, orf408 and orf170, were also found as a form of non-freestanding ORF. The orf408 gene was entirely located within a coding region of the rps3 gene for ribosomal protein S3 and rrnL gene for a large subunit ribosomal RNA, and the orf170 gene was identified completely within a conserved domain for homing endonuclease (HE). All of the core PCGs and ORFs were initiated with the standard ATG start codon and terminated with the TAA stop codon, except for three genes: the nad4 gene with the GTG start codon, and atp9 and orf170 with the TAG stop codon. An overlap of one nucleotide base was only found between the TAA stop codon of the nad4L gene and the ATG start codon of the nad5 gene (−1 bp, corresponding to the codon of A at nucleotide position 897) (Table S1).
All 24 tRNA genes (trn genes), accounting for approximately 6.49% of the entire mitogenome, were shown with average lengths ranging from 71 bp to 87 bp. These trn genes were found as single copies, but several were found in multiple copies with different anticodons (two copies for trnR Arg (trnR Arg [ACG] and trnR Arg [TCT]), two copies for trnS Ser (trnS Ser [GCT] and trnS Ser [TGA]), and two copies for trnL Leu (trnL Leu [TAA] and trnL Leu [TAG])). A trnM Met gene coding for methionine was also found in three copies with the same anticodon ((trnM Met [CAT]) (Tables S1 and S2)). Most of the tRNA genes were predicted to have a typical cloverleaf secondary structure. However, some trn genes (positions: trnY Tyr at 13,598, trnS Ser at 13,854 and 16,306, and trnL Leu at 21,911 and 22,462) were found with an extra variable arm (V-arm) ( Figure S1), which was consistent with previous reports about the specificity of extra V-arm to tRNA Leu , tRNA Ser , and bacterial-/eukaryotic organellar tRNA Tyr [56,57].
Two rRNA genes, the rns gene for small subunit ribosomal RNA and the rrnL gene for large subunit ribosomal RNA, accounted for approximately 22.50% of the entire length of mitogenome. The rns gene located between the atp6 gene and the cox3 gene was found as a freestanding ORF (length 1503 bp). Otherwise, the rrnL gene was predicted in a position ranging over two of PCGs, orf408 and rps3 (length 4713 bp). The rps3 gene for ribosomal protein S3 was found to be an internal ORF of the group I intron that was positioned within the rrnL gene [58], and was 1368 bp in size (141 bp larger than orf408) with the ATA start codon ( Figure 1 and Table S1).

Phylogenetic Analyses of the T. harzianum CBS 226.95 Mitogenome
Excluding an unverified dataset, 47 complete mitogenomes (i.e., 46 from Hypocreales species (belonging to the class Sordariomycetes) and one of N. crassa OR74A (belonging to the order Sordariales that composed of the class Sordariomycetes with the order Hypocreales)) were retrieved from the GenBank and used in the analyses (Table S5).
In both of the molecular phylogenies (ML-and BI tree), all Hypocreales species were well observed with high confidence, under the significant clades supported by seven major families of the order Hypocreales; interestingly, Paecilomyces penicillatus SAAS_ppe1, belonging to the family Clavicipitaceae, was found with a closer relationship with the family Hypocreaceae than those of native family (BS = 100, BPP = 1.00). It may be resolved by improving the Hypocreales phylogeny resolution on accumulations of informative datasets of the mitochondrial genomes from closely related family species (Figures 2  and S2).
The genus Trichoderma, including T. harzianum CBS 226.95, was placed in a sister clade with the genus Hypomyces. Among hypocrealean fungi, a sister relationship between T. harzianum ATCC 26799 and T. gamsii KUC1747 was observed identically in both tree topologies (BS = 100, BPP = 1.00) (Figure 2 and Figure S2).

Comparative Analysis of the Trichoderma Mitogenomes
The complete de novo mitogenome of T. harzianum CBS 226.95 was compared with five complete Trichoderma mitogenomes that publicly available from the genus Trichoderma, excluding an unverified dataset (as of January 2021) ( Table 1).
The entire lengths of mitogenome ranged from 27,632 bp (T. harzianum CBS 226.95) to 42,130 bp (T. reesei QM9414) with an average AT content (72.2%). The intergenic regions ranged from 17.58% (T. reesei QM9414) to 31.73% (T. asperellum B05), and the mitogenome of T. harzianum CBS 226.95 was shown to fall in the middle of this range. The mitogenomes of T. gamsii KUC1747, T. asperellum B05, and T. hamatum were presented with the negative value of AT-skew as those of T. harzianum CBS 226.95. In contrast, only two mitogenomes, from T. reesei QM9414 and T. atroviride ATCC 26799, contained all positive values for both AT-/GC-skew ( Table 1).
The mitogenome of T. atroviride ATCC 26799 was shown to harbor more protein CDS and trn genes than other Trichoderma species (Table 1). Across Trichoderma species, gene lengths for apocytochrome b (cob) and cytochrome c oxidases (cox1, cox2, and cox3) were observed in various sizes due to the presence of introns (Table S6). The T. reesei QM9414 mitogenome, which contains both the most extensive length of entire sequences and restricted proportions of non-coding regions, was observed to carrying more introns than other Trichoderma species. A total of 11 intron loci were detected, mainly in the exonintron structures of the PCGs for apocytochrome b (cob) and cytochrome c oxidases (cox1, cox2, and cox3), which increase the full size of the target gene itself than the total number of annotated genes in the mitogenome sequences, whereas few introns were predicted in the T. gamsii KUC1747 and T. harzianum CBS 226.95 (Table 1 and Table S8). Notably, an IA-type intron (i.e., an intron of the subtype IA) was commonly found with an intronic ORF in the region of the rrnL gene of all Trichoderma species except T. asperellum B05 (i.e., in this study, the likelihood of the rrnL gene being present in the T. asperellum B05 mitogenome was not considered because this gene was not reported in the GenBank dataset of the T. asperellum B05 mitogenome) ( Figure 3A and Table S8). Finally, compared with other Trichoderma species, the mitogenome of T. harzianum CBS 226.95 showed the highest identity of 95.17% to T. reesei QM9414 (Table 1). The tree was generated using concatenated sequences of 13 core genes (atp6, atp8, cob, cox1, cox2, cox3, nad1, nad2, nad3, nad4, nad4L, nad5, and nad6), and the mitochondrial genome of Neurospora crassa OR74A (Sordariales) was used as an outgroup. All Sordariomycetes species that used for the phylogenetic tree are described in Table S5, and Bayesian posterior probabilities (BPPs) values were marked on the nodes. Asterisks for the Neurospora crassa OR74A mitogenome (*) and T. harzianum CBS 226.95 mitogenome (**).
The genus Trichoderma, including T. harzianum CBS 226.95, was placed in a sister clade with the genus Hypomyces. Among hypocrealean fungi, a sister relationship between T. harzianum ATCC 26799 and T. gamsii KUC1747 was observed identically in both tree topologies (BS = 100, BPP = 1.00) (Figures 2 and S2).

Comparative Analysis of the Trichoderma Mitogenomes
The complete de novo mitogenome of T. harzianum CBS 226.95 was compared with five complete Trichoderma mitogenomes that publicly available from the genus Trichoderma, excluding an unverified dataset (as of January 2021) ( Table 1)  The tree was generated using concatenated sequences of 13 core genes (atp6, atp8, cob, cox1, cox2, cox3, nad1, nad2, nad3, nad4, nad4L, nad5 and nad6), and the mitochondrial genome of Neurospora crassa OR74A (Sordariales) was used as an outgroup. All Sordariomycetes species that used for the phylogenetic tree are described in Table S5, and Bayesian posterior probabilities (BPPs) values were marked on the nodes. Asterisks for the Neurospora crassa OR74A mitogenome (*) and T. harzianum CBS 226.95 mitogenome (**).

Divergence Times of the Trichoderma Mitogenomes in Hypocreales
To explore the evolutionary divergence of Trichoderma mitogenomes in Hypocreales, a range for the taxonomic analysis was set to cover major families of the order of Hypocreales with species, such as (1) 10 selected species that represented each significant family in the order Hypocreales (one for Bionectriaceae (Clonostachys rosea 6792), two for Clavicipitaceae (Epichloe typhina E8 and Metarhizium anisopliae ME1), two for Cordycipitaceae (Cordyceps militaris EFCC-C2 and Beauveria bassiana), one for Hypocreales incertae sedis (Acremonium chrysogenum ATCC11550), two for Nectriaceae (Nectria cinnabarina 5175 and Fusarium oxysporum F11), and two for Ophiocordycipitaceae (Hirsutella thompsonii AR-SEF9457 and Ophiocordyceps sinensis)), (2) one strain H. aurantius from the genus Hypomyces, to use as a sub-outgroup for the genus Trichoderma within the Hypocreaceae clade, and (3) one strain N. crassa OR74A, belonging to the order Sordariales that composed of the class Sordariomycetes with the order Hypocreales, to use as an outgroup for all Hypocreales species. Based on the complete mitogenomes of the selected species (Table S5), only exons sequences of the 13 core PCGs were used in the phylogenetic analysis to remove the potential of evolutionary effects by the presence/numbers of introns within the nucleotide sequences of the target gene.  (2) TS  T1  T2  T3  T4  T5 a Based on the GenBank database, NCBI (datasets as of, January 2021). b A number of putative trn genes predicted within the protein-coding genes were indicated inside the parentheses (details in Table S7). c Details of the detected group I intron loci (subtype, position, intronic ORF) are shown in Table S8.
On the time-scaled Bayesian phylogeny, all positions of species were well supported with high BPPs values (BPP = 1.00). Five calibration points were also presented with significant node ages, which were entirely consistent with those of the Bayesian chronogram previously reported using orthologous proteins of Hypocreales genomes (i.e., including Trichoderma genomes) [20].
Under the occurrence of the family Hypocreaceae at 114.83 Mya, a divergence time of the Trichoderma genus was estimated to be 52.46 Mya, and the speciation events that evolved within the genus Trichoderma appear to occur during the following periods:

Divergence Times of the Trichoderma Mitogenomes in Hypocreales
To explore the evolutionary divergence of Trichoderma mitogenomes in Hypocreales, a range for the taxonomic analysis was set to cover major families of the order of Hypocreales with species, such as (1) 10 selected species that represented each significant family in the order Hypocreales (one for Bionectriaceae (Clonostachys rosea 6792), two for Clavicipitaceae (Epichloe typhina E8 and Metarhizium anisopliae ME1), two for Cordycipitaceae (Cordyceps militaris EFCC-C2 and Beauveria bassiana), one for Hypocreales incertae sedis (Acremonium chrysogenum ATCC11550), two for Nectriaceae (Nectria cinnabarina 5175 and Fusarium oxysporum F11), and two for Ophiocordycipitaceae (Hirsutella thompsonii ARSEF9457 and Ophiocordyceps sinensis)), (2) one strain H. aurantius from the genus Hypomyces, to use as a sub-outgroup for the genus Trichoderma within the Hypocreaceae clade, and (3) one strain N. crassa OR74A, belonging to the order Sordariales that composed of the class Sordariomycetes with the order Hypocreales, to use as an outgroup for all Hypocreales species. Based on the complete mitogenomes of the selected species (Table  S5), only exons sequences of the 13 core PCGs were used in the phylogenetic analysis to remove the potential of evolutionary effects by the presence/numbers of introns within the nucleotide sequences of the target gene.
On the time-scaled Bayesian phylogeny, all positions of species were well supported with high BPPs values (BPP = 1.00). Five calibration points were also presented with significant node ages, which were entirely consistent with those of the Bayesian chronogram previously reported using orthologous proteins of Hypocreales genomes (i.e., including Trichoderma genomes) [20].
Under the occurrence of the family Hypocreaceae at 114.83 Mya, a divergence time of the Trichoderma genus was estimated to be 52.46 Mya, and the speciation events that evolved within the genus Trichoderma appear to occur during the following periods: (1) 41.86 Mya between T. harzianum CBS 226.95 and T. reesei QM9414; (2) 8.06 Mya between T. asperellum B05 and T. hamatum; and (3) 3.82 Mya between T. atroviride ATCC 26799 and T. gamsii KUC1747. Interestingly, when considering the divergent times of Trichoderma species that dated on the Bayesian chronogram using orthologous of Hypocreales genomes (i.e., including Trichoderma genomes) [20], the speciation between T. harzianum CBS

Putative tRNAs Overlapped with Protein-Coding Genes
Considering the tRNA genes that fully or partially overlapped with PCGs, some studies of metazoan/nematode mitogenomes have noted that it may allow (1) avoiding the loss of tRNA genes or (2) keeping gene functions by the co-evolution of two overlapping genes, on the mitogenome evolving to compacted lengths [56,59]. Recently, several putative trn genes, fully positioned within the PCGs of Trichoderma mitogenomes, have been described in our study of the T. atroviride ATCC 26799 mitogenome (e.g., tRNA Val overlapped with the 3 end region of nad6 gene, tRNA Met integrated within the 5 end region of nad2 gene) [11]. Similarly, a putative trnV Val gene, fully overlapped with the 3 end region of the nad6 gene, was also found in the complete de novo mitogenome of T. harzianum CBS 226.95 (Tables S2  and S7).
The tRNA genes of mitochondria are expected to have evolved through various processes, such as (1) duplications or degenerations of tRNAs, (2) transpositions of tRNAs to the nucleus, causing the reduced tRNAs in the mitogenome itself, and (3) mutations of the tRNA anticodon, leading a remodeling of the tRNA secondary structure, however, the molecular-/functional aspects of these mitochondrial tRNAs have not yet been fully understood [60,61]. Interestingly, among the Trichoderma mitochondrial tRNAs, which are responsible for 20 standard amino acids in the translation system of mitochondrial coding proteins, a putative trnV Val gene was only found for the amino acid valine across all Trichoderma species (Table S7). This feature requires further study concerning (1) the expression of putative trnV Val as a form of functional tRNA, (2) whether it is co-expressed within the mature mRNA of PCGs, and (3) whether the mitochondria require a nucleusencoded trnV Val from the cytosol for the mitochondrial translation system (i.e., because the putative trnV Val , which is positioned within the protein coding sequences, might already be recognized as a pseudo-tRNA by the tRNA degenerations). Given these questions, all putative trn genes that were observed in the PCGs (trnV Val and tRNA Met ) actively support the likelihood of complex and dynamic evolutions of tRNAs in Trichoderma mitogenomes.

Evolution of Mitochondrial Introns in Trichoderma
In our recent study of the T. atroviride ATCC 26799 mitogenome [11], Trichoderma mitochondrial introns were characterized with (1) the similarities of intron sequences between Trichoderma species, though not clustered in most within the range of the Trichoderma genus, (2) potentials of the horizontal transfer of introns between inter-/intraspecific fungal diversities over the genus level of Trichoderma, and (3) several gain/loss events of introns on the reconstructed ancestral states of these introns [11]. In contrast to previous studies based on the most homologous introns of the core PCGs, in this study on the mitogenome of T. harzianum CBS 226.95, only two intron loci (all subtype IA) were observed regardless of the core PCGs: one was found in the rrnL gene with two internal ORFs (orf408 and rps3) as an rrnL-i (i.e., intron placed within the rrnL gene), another was also detected with multiple intronic ORFs that were composed by embedding the orf170 gene in an 864 bp-sized HE harboring a GIY-YIG motif ( Figure 3A, Tables S1 and S8). Nevertheless, compared among six Trichoderma species, all features of the Trichoderma mitochondrial introns that were described in our previous study are still meaningful through the constant patterns of intron distributions in Trichoderma species (e.g., most of the intron loci were detected in the exon-intron structures of PCGs for apocytochrome b (cob) and cytochrome c oxidases (cox1, cox2, and cox3), but not in the two species of T. gamsii KUC1747 and T. harzianum CBS 226.95 that harbor all core PCGs on the intronless structures) ( Figure 3A and Table S8).
Since mitochondrial endosymbiosis occurred between the earliest ancestor of mitochondria (pre-mitochondrial alphaproteobacterium) and an archaeal-derived host (Asgard archaea), the ancestral mitochondria genome had experienced massively reduced mitogenomes, similar to modern mitochondria [1]. However, despite its ultimately compact size that was driven by reductive evolutionary processes (e.g., mainly, the loss of redundant/unnecessary genes and/or endosymbiotic gene transfers to the nucleus) [1], the structural intricacy of the current mitogenome varies immensely across eukaryotes [4,62]. In fungal mitogenomes, the group I intron, which is capable of self-splicing, causes the architectural complex of the mitogenome [32,63,64]. Among six Trichoderma mitogenomes, the existence of mitochondrial introns was involved in the structural complexity of the mitogenome itself ( Figure 3A), however, strictly not leading to the increase in the entire lengths of their mitochondrial sequences. Excluding T. reesei QM9414, a correlation between the numbers of introns present and the increase in mitogenome size could not be typically defined across Trichoderma species, suggesting the significance of the location and size of the intron locus rather than the intron's existence itself within mitogenome sequences (Tables 1 and S8).
Including our previous study of the T. atroviride ATCC 26799 mitogenome [11], several recent studies of fungal species have demonstrated the evolutionary dynamics of mitochondrial introns, such as multiple-gains/losses/degenerations of introns, by tracing the ancestral history of these intron loci within the relevant diversity of the target fungus [9,65]. In this study, ancestral states of intron events across all six Trichoderma species were not constructed due to the absence of homologous intron loci of the core PCGs within the T. harzianum CBS 226.95 mitogenome ( Figure 3A and Table S8). However, when comparing phylogenies, six Trichoderma mitogenomes were identically arranged in two kinds of BI topologies, which were obtained either by using full genes (including introns) of the core PCGs (Figure 2) or by only using exon sequences (excluded introns) of the core PCGs ( Figure 3B), such as (1) a clade by T. atroviride ATCC 26799 and T. gamsii KUC1747, (2) a clade by T. asperellum B05 and T. hamatum, and (3) a clade by T. harzianum CBS 226.95 and T. reesei QM9414. In particular, the clustering between T. reesei QM9414 (the largest sized one with the highest number of intron loci) and T. harzianum CBS 226.95 (the smallest lengths harboring few introns) deserved attention related to the divergence time of this clade that was estimated to be much older (~41.86 Mya) than the other two clades (8.06-3.82 Mya) (Figures 2 and 3B). Given the theory of genomic streamlining [66], some, but not all, species of lichenized fungi are speculated to have experienced multiple intron loss events for mitogenome contractions [10]. However, in the evolutionary view of eukaryogenesis, the origin and evolution of introns have been considered using the comprehensive perspectives of two competing theories: the 'introns-early' theory (i.e., introns first existed at the earliest stages of life's evolution, then evolved toward intron loss) and the 'introns-late' theory (i.e., introns emerged/increased in eukaryotes during eukaryote evolution)' [67]. All Trichoderma clustering discovered in this study seems to be insufficient to infer whether the Trichoderma mitogenomes evolved to reduce mitogenome sizes due to intron loss and degeneration under genome streamlining, but clearly imply the continuation of varied-gain and loss of mitochondrial introns during Trichoderma evolution (Figures 2 and 3B).
When considering the identical arrangements of Trichoderma clustering between two BI topologies that were obtained with/without intron sequences of the core PCGs (Figures 2 and 3B), it actively supports a shared evolutionary history between the core PCGs and their introns in Trichoderma. Meanwhile, IA-typed rrnL-i was commonly detected across all Trichoderma species except T. asperellum B05 (i.e., in this study, the likelihood of the rrnL gene being present in the T. asperellum B05 mitogenome was not considered because this gene was not reported in the GenBank dataset of the T. asperellum B05 mitogenome (GenBank accession no. NC_037075)) ( Figure 3A and Table S8). Although it should be comprehensively reflected in the unique features of other intron loci, such as (1) the intron locus harboring multiple intronic ORFs, or (2) the coexistence of different intron subtypes within a single coding gene (Table S8); nevertheless, the rrnL-i is expected to play an essential role in revealing the dynamic history of Trichoderma mitochondrial introns, as one of the primitive factors that is assumed to be inherited from the most recent common ancestor of the Trichoderma genus.

Time-Scaled Bayesian Phylogeny with the Strain T. harzianum HB324
In addition to the six complete mitogenomes of Trichoderma species (including T. harzianum CBS 226.95 of this study), there is additional mitogenomic information from the group of T. harzianum species; the 32,277 bp mitogenome of T. harzianum HB324 (Gen-Bank accession no. MT263519). Nonetheless, defined annotations of this mitogenome are unavailable in GenBank because it was publicly deposited as an unverified dataset (as of January 2021). In a recent study of this mitogenome that described mitogenomic features for the first time [68], some coding genes were suggested with multiple terminations of their coding regions. For these reasons, in this study, the core PCGs of the T. harzianum HB324 mitogenome were re-annotated to remove the positions of stop codons that were multiply predicted in the middle of coding genes and re-defined their gene lengths (Table S9). Then, the evolutionary divergences of Trichoderma mitogenomes were re-traced with T. harzianum HB324 harboring newly modified annotations, under the same analytical conditions that applied for the BI phylogeny in Figure 3B. On the newly constructed time-scaled Bayesian tree, all fungal species were well-positioned with high BPPs values (BPP = 1.00), and five calibration points were also indicated to be compatible with those of the Bayesian chronogram using orthologous of Hypocreales genomes (i.e., including Trichoderma genomes) [20]. However, despite belonging to the same species of T. harzianum, two strains of CBS 226.95 and HB324 were placed separately ( Figure 4A).
As shown in the two BI topologies (Figures 2 and 3B), phylogenetic replacements of the complete six Trichoderma mitogenomes (i.e., T. asperellum B05, T. atroviride ATCC 26799, T. gamsii KUC1747, T. hamatum, T. harzianum CBS 226.95, and T. reesei QM9414) were congruent with those of the chronogram using 638 core orthologous from Hypocreales genome [20], such as (1) a clade by T. harzianum and T. reesei, and (2) a clade consisting of subclades by four Trichoderma species (T. asperellum, T. atroviride, T. gamsii, and T. hamatum). However, as shown in Figure 4A, among the seven Trichoderma mitogenomes (including strain HB324), T. harzianum HB324 was positioned as an outgroup to all Trichoderma taxa. At the same time, T. harzianum CBS 226.95 was clustered with a subclade comprised of the five remaining Trichoderma species. According to the joining of strain HB324, Trichoderma speciation was also predicted more divergently. The divergence of T. harzianum HB324 was estimated at 87.28 Mya and followed by T. harzianum CBS 226.95 at 55.69 Mya. Furthermore, T. reesei QM9414 appeared later than T. harzianum CBS 226.95 at 44.77 Mya ( Figure 4A). mitogenome were re-annotated to remove the positions of stop codons that were multiply predicted in the middle of coding genes and re-defined their gene lengths (Table S9). Then, the evolutionary divergences of Trichoderma mitogenomes were re-traced with T. harzianum HB324 harboring newly modified annotations, under the same analytical conditions that applied for the BI phylogeny in Figure 3B. On the newly constructed timescaled Bayesian tree, all fungal species were well-positioned with high BPPs values (BPP = 1.00), and five calibration points were also indicated to be compatible with those of the Bayesian chronogram using orthologous of Hypocreales genomes (i.e., including Trichoderma genomes) [20]. However, despite belonging to the same species of T. harzianum, two strains of CBS 226.95 and HB324 were placed separately ( Figure 4A).  Compared to T. harzianum CBS 226.95, the mitogenome of T. harzianum HB324 was approximately 4645 bp larger and exhibited a sequence identity of 92.23% (query coverage 94%). The GC content was similar, but skewness patterns were different for all positive values in T. harzianum HB324. As described in the beginning of this discussion, putative trn genes entirely positioned within the coding genes (e.g., tRNA Val of the nad6 gene, tRNA Met of the nad2 gene) were found across all Trichoderma species (Table S7)-however, not in the T. harzianum HB324 (Table S9). In addition, when comparing genetic distances using exon sequences of the 13 core PCGs (atp6, atp8, cob, cox1, cox2, cox3, nad1, nad2, nad3, nad4, nad4L, nad5, and nad6; the atp9 gene was excluded because it has not been reported in the GenBank dataset of the T. gamsii KUC1747 mitogenome (GenBank accession no. KU687109)), T. harzianum HB324 was assumed to be more divergent from T. harzianum CBS 226.95 (p-distance value, 0.099) than other Trichoderma species (p-distance values, 0.055-0.057) ( Figure 4B and Table S10). All these aspects showed a presence of mitogenomic differentiation between two T. harzianum strains, CBS 226.95 and HB324. Nevertheless, it seems that one additional point should be also contemplated together as below.
Since the first description of the name Trichoderma [69], phylogenetic concepts of the genus Trichoderma have been developed to systematically identify Trichoderma taxonomy [70]. Particularly, in earlier studies, some Trichoderma strains were misidentified, and even the name of T. harzianum has been misused for many different species due to the species complex of Trichoderma [71][72][73][74]. To overcome these difficulties, species of Trichoderma have recently been (re-)identified for the most accurate discrimination between Trichoderma using combined approaches, such as (1) morphological phenotypes and (2) simultaneous applications using multiple-DNA barcodes on regions of rDNA internal transcribed spacers (ITS), elongation factor 1-α (tef1), and RNA polymerase II subunit (RPB2) [74,75]; consequentially, a taxonomy of the T. harzianum species complex was revised to include 14 species [74]. For instance, the strain P1 (ATCC 74058), strain T22, and strain ATCC 26799, widely known as representatives of T. harzianum species in earlier studies, were re-determined to be T. atroviride P1 [71], T. afroharzianum T22 [74], and T. atroviride ATCC 26799 [75], respectively. Meanwhile, the fungus T. harzianum HB324, isolated from the leaves of rubber tree (Hevea brasiliensis), was reported to be identified using a DNA barcode only for the ITS region (used PCR primers for ITS4 and ITS5) [76].
In a recent study of the T. harzianum HB324 mitogenome [68], a time-scaled maximum likelihood (ML) tree was constructed using only three coding genes (atp8, atp9, and cox3). On this ML tree, the mitogenome of T. harzianum HB324 occurred at 27.40 Mya, placed apart from the clade covering all other Trichoderma species [68]. In contrast, though disagreements of analysis conditions must be considered (e.g., tree platform, applied gene sets), the clustering of Trichoderma species on the time-scaled BI tree obtained in this study was congruent with those of the chronogram using nucleus orthologous proteins of Hypocreales [20] ( Figure 3B). Nevertheless, when considering the newly constructed BI chronogram by including strain HB324 ( Figure 4A), it is evident that the mitogenome of T. harzianum HB324 has distinctive features, which influenced the Trichoderma divergences dynamically. Taxonomic revisions of Trichoderma species are still ongoing to correctly identify members of the T. harzianum complex [77]. Obtaining improved resolutions of Trichoderma topologies, including new isolates such as the strain HB324, will be more useful for the elucidation of Trichoderma mitogenome evolutionary relationships.

Evolutionary Pressures for the Trichoderma Mitogenomes
Related with the adaptive evolution of Trichoderma mitogenomes, selective pressures were investigated using the codon alignments of 13 core PCGs (atp6, atp8, cob, cox1, cox2, cox3, nad1, nad2, nad3, nad4, nad4L, nad5, and nad6; the atp9 gene was excluded because it has not been reported in the GenBank dataset of the T. gamsii KUC1747 mitogenome (GenBank accession no. KU687109)). Compared to the Ka/Ks ratios, most of the core PCGs were assessed with values lower than one (<1), which is supposed to be under purifying selection. However, interestingly, the cox3 gene was measured with a value greater than one (1>) in a pairwise comparison between T. harzianum CBS 226.95 and T. reesei QM9414, implying a positive selection potential. When considering the Trichoderma clustering of the two BI topologies (Figures 2 and 3B), the cox3 gene might be a site for adaptive changes that influenced the evolutionary speciation events between two strains, CBS 226.95 and QM9414, which were closely placed in a sister relationship within the same clade.
Another remarkable thing is that the nad6 gene of T. harzianum HB324 was shown with a Ka/Ks ratio over one (1>) against those of all other Trichoderma species, even including H. aurantius belonging to the genus Hypomyces that consisted of the family Hypocreaceae. It seems to support the potential of evolutionarily diversifying for the fungus HB324, apart from the other species within the clade of all Trichoderma taxa as shown in Figure 4A ( Figure 5 and Table S11). Meanwhile, using the MEME approach on the phylogenetic lineages of Hypocreales ( Figure 4A), the cox1 gene was predicted with higher individuals of 20 codon sites that were assumed to have evolved under a positive diversifying selection, in contrast to the non-sites in the atp8 gene (Table S12).
Under intense selective pressures, organisms may evolve adaptations for their survival in certain ecological conditions, which would be achieved via site-specific substitutions of nucleotides/amino acids affecting the protein structure, further leading to changes in protein functions [78]. In a previous study of the evolution of Trichoderma habitat preferences, T. reesei was found to favor decaying plant materials rather than soil [14]. The OXPHOS pathway is strictly dependent on oxygen to generate cellular energy [79]. When considering the various habitats where Trichoderma have been isolated (e.g., soils, wood, and living plants/fungi) [14], environmental conditions, particularly concerning oxygen, might have affected the evolution of Trichoderma mitogenomes during their adaptive diversification. Our results can be proposed as clear signals of the evolving effects on the essential genes of OXPHOS, further expanding the viewpoint of molecular-and metabolic evolutions of Trichoderma mitochondria under the inherited habits of their geographic origins.
ing H. aurantius belonging to the genus Hypomyces that consisted of the family Hypocreaceae. It seems to support the potential of evolutionarily diversifying for the fungus HB324, apart from the other species within the clade of all Trichoderma taxa as shown in Figure 4A ( Figure 5 and Table S11). Meanwhile, using the MEME approach on the phylogenetic lineages of Hypocreales ( Figure 4A), the cox1 gene was predicted with higher individuals of 20 codon sites that were assumed to have evolved under a positive diversifying selection, in contrast to the non-sites in the atp8 gene (Table S12).  Table S11. Ka, nonsynonymous substitution rate; Ks, synonymous substitution rate. Figure 5. Analysis of Ka/Ks ratios of the core PCGs. Obtained values are described in Table S11. Ka, nonsynonymous substitution rate; Ks, synonymous substitution rate.

Conclusions
Here, a complete de novo mitogenome of an ex-neotype fungus of the T. harzianum species complex, T. harzianum CBS 226.95, is presented for the first time. Trichoderma mitogenomes were comprehensively compared to update the genomic understanding of Trichoderma mitochondria, the evolutionary divergences of Trichoderma mitogenomes were also exclusively traced using the mitochondrial core PCGs and discussed with latent genomic features triggered mitogenomic evolution in Trichoderma. Our results demonstrate that the adaptive evolution of Trichoderma mitogenomes has been actively ongoing with regard to the complex effects of various genetic elements. All of this information will allow for a better understanding of the distinct behaviors of Trichoderma species in nature, thus further expanding our insight into Trichoderma to make them more effective fungal biocontrol agents.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/microorganisms9081564/s1, Figure S1. Predicted secondary structure of transfer RNA genes in the T. harzianum CBS 226.95 mitogenome.; Figure S2. Phylogenetic tree of the Hypocreales species (Sordariomycetes) mitochondrial genomes based on ML (maximum likelihood) analysis; Table S1. Genomic organization of the T. harzianum CBS 226.95 mitogenome;  Table S11. Ka/Ks ratios for the 13-core protein-coding genes in the Hypocreaceae mitogenomes; Table S12. Prediction of codon sites evolving under positive diversifying selection on the MEME approach.