Conservation and Expansion of Transcriptional Factor Repertoire in the Fusarium oxysporum Species Complex

The Fusarium oxysporum species complex (FOSC) includes both plant and human pathogens that cause devastating plant vascular wilt diseases and threaten public health. Each F. oxysporum genome comprises core chromosomes (CCs) for housekeeping functions and accessory chromosomes (ACs) that contribute to host-specific adaptation. This study inspects global transcription factor profiles (TFomes) and their potential roles in coordinating CC and AC functions to accomplish host-specific interactions. Remarkably, we found a clear positive correlation between the sizes of TFomes and the proteomes of an organism. With the acquisition of ACs, the FOSC TFomes were larger than the other fungal genomes included in this study. Among a total of 48 classified TF families, 14 families involved in transcription/translation regulations and cell cycle controls were highly conserved. Among the 30 FOSC expanded families, Zn2-C6 and Znf_C2H2 were most significantly expanded to 671 and 167 genes per family including well-characterized homologs of Ftf1 (Zn2-C6) and PacC (Znf_C2H2) that are involved in host-specific interactions. Manual curation of characterized TFs increased the TFome repertoires by 3% including a disordered protein Ren1. RNA-Seq revealed a steady pattern of expression for conserved TF families and specific activation for AC TFs. Functional characterization of these TFs could enhance our understanding of transcriptional regulation involved in FOSC cross-kingdom interactions, disentangle species-specific adaptation, and identify targets to combat diverse diseases caused by this group of fungal pathogens.


Genome Partition
The genome partition results for chromosome-level assemblies were retrieved from previous reports for Fol4287 [29], FoII5 [32], Fo5176 [27], and Fo47 [28]. Fo47 has a clear genome partition with 11 core chromosomes and one accessory chromosome, therefore serving as the reference for the genome partition of other F. oxysporum genomes. MUMmer/3.22 was applied to align scaffolds of genome assemblies against 11 core chromosomes of the reference genome Fo47 using default parameters. The scaffolds aligned to the core chromosomes of Fo47 with a coverage larger than 5% were annotated as core scaffolds. The rest of the scaffolds were partitioned as accessory scaffolds. Genes residing on the core and accessory scaffolds were annotated as the core and accessory genes, respectively.

Phylogenetics Analysis
Protein sequences were aligned via MAFFT/7.313 [33]. The iqtree/1.6.3 [34,35] was run on the sequence alignment to generate the phylogeny (by maximum likelihood method and bootstrapped using 1000 replicates) [36]. Visualization was conducted via the Interactive Tree of Life [37] to produce the phylogram. OrthoFinder 2.5.4 [23] was used for orthogroup determination. To build a species phylogram, 500 randomly selected conserved proteins (single-copy orthologs) were aligned first. The alignment was then concatenated, and the phylogeny was determined and visualized using the above methods.

Expansion Index Calculation
To understand genome regulation among FOSC, we developed two expansion index scores. The first uses two yeast lineages as the baseline (EI y ):

EI y =
Average number o f TFs in FOSC + 1 Average number o f TFs in yeasts + 1 In the second index score (EI f ), we directly compared F. oxysporum with its Fusarium relatives to calculate the expansion index as follows: Average number o f TFs in FOSC + 1 Average number o f TFs in FOSC sister species + 1

FOSC TFome Expansion Resulted from the Acquisition of ACs
We compared 30 ascomycete fungal genomes ( Figure 1 and Table 1) including 15 strains within the FOSC, nine sister species close to F. oxysporum, two yeast genomes (Saccharomyces cerevisiae and Schizosaccharomyces pombe), and four other filamentous fungal species (Neurospora crassa, Aspergillus nidulans, Aspergillus acristatulus, and Magnaporthe oryzae). To maintain consistency, the protein sequences for all of these genomes were retrieved from the MycoCosm portal [19].
We compared 30 ascomycete fungal genomes ( Figure 1 and Table 1) including 15 strains within the FOSC, nine sister species close to F. oxysporum, two yeast genomes (Saccharomyces cerevisiae and Schizosaccharomyces pombe), and four other filamentous fungal species (Neurospora crassa, Aspergillus nidulans, Aspergillus acristatulus, and Magnaporthe oryzae). To maintain consistency, the protein sequences for all of these genomes were retrieved from the MycoCosm portal [19].   For a comprehensive TFome annotation, we used InterProScan (IPR) terms associated with fungal transcriptional regulation [21,22] and curated a mapping with updated IPR classification (interproscan version: 5.38-76.0) [52]. In addition, we searched the IPR classification of the protein families and obtained all other terms related to the transcriptional regulation activity. This resulted in 234 TF-related IPR terms (Table S1). Since most terms were initially defined in the mammalian systems, fungal genomes included in this study were associated with 71 out of the 234 TF-related IPR terms (Table S1, Materials and Methods, and Figure S1A,B for the annotation pipeline). After removing 13 terms for redundancy (two terms describing the identical domain) and 10 terms for minimal presentation (<4 among the 30 genomes), this comparative TFome study focused on 48 IPR terms in 27,967 TFs (Tables S1 and S2). Notably, a quarter of these terms were not reported to be affiliated with fungal transcriptional regulation by either Park et al. (2008) [21] or Shelest (2017) [22] (Table S1).
Comparing the total number of protein-coding genes (x) and the total number of TFs (y) within the same genome, we observed a strong positive correlation (y = 0.07264x − 190.9, r 2 = 0.9361) ( Figure 2A). FOSC TFomes were larger than other genomes included in this study, with an average of 1144 TFs per genome ( Figure 2A, Table 1). After partitioning each FOSC genome into core and accessory regions (see Section 2.3 for details), we observed a positive correlation between the number of TFs encoded in the accessory chromosomal region of each strain (defined as accessory TFs hereafter) with the size of the accessory genomes (Mb) (y = 17.239x + 3.553) ( Figure S2). This suggests that accessory chromosomes contribute directly to the expanded TFome. To understand the genome regulation among the FOSC, we developed an expansion index score using two yeast lineages as the baseline (EIy): Based on this index value, we classified TF families into three major groups (Tables 2 and S1). Group 1 contained 14 TF families with an expansion score of 1, indicating high conservation. Group 2 included four families with an index score below 1, reflecting some level of gene family contraction. Group 3 contained 30 families with an expansion index greater than 1, indicating gene expansion. To understand the genome regulation among the FOSC, we developed an expansion index score using two yeast lineages as the baseline (EI y ):

EI y =
Average number o f TFs in FOSC + 1 Average number o f TFs in yeasts + 1 Based on this index value, we classified TF families into three major groups (Tables 2 and S1). Group 1 contained 14 TF families with an expansion score of 1, indicating high conservation. Group 2 included four families with an index score below 1, reflecting some level of gene family contraction. Group 3 contained 30 families with an expansion index greater than 1, indicating gene expansion.

Conserved TF Families That Are Primarily Associated with General/Global Transcription Factors
Fourteen TF families accounted for 30% of our annotated TF families. Most of these fourteen families had a single ortholog in all genomes included in this study ( Figure 2B; Tables 2 and S1), suggesting their functional conservation across the Ascomycota. These 30% conserved TF families accounted for less than 2% of the total TFomes. Annotation based on S. cerevisiae and other model organisms suggested their involvement in transcription/translation regulation and cell cycle control.

Transcription/Translation Regulation
Nine TF families were annotated to be related to transcription and translational regulation including TATA box-binding protein (TBP), TBP-associated factors (TAFs), RNA polymerase II elongation regulator Vps25, and CCAAT-binding factors (CBFs) related to ribosomal biogenesis.
One of the most conserved TF families, transcription initiation TBP binds directly to the TATA box to define the transcription start and initiate transcription facilitated by all three RNA polymerases. In fact, the function of TBP is so conserved that the yeast homolog can complement TBP mutations in humans [53,54]. Seven conserved TF families are classified as transcription positive/negative regulators, and transcription elongation factors. TAF12 and TAF II 28 are parts of the transcription factor TF II D complex. Interacting with TBP, TAFs form the TF II D complex and positively participate in the assembly of the transcription preinitiation complex [55]. Similarly, TF II H works synergistically with TF II D to promote the transcription [56]. In contrast, negative cofactor 2 (Ncb2) inhibits the preinitiation complex assembly [57]. Other factors include the CNOT1, a global regulator involved in transcription initiation and RNA degradation [58], and Vps72/YL1, which contributes to transcriptional regulation through chromatin remodeling, as reported in yeast [52,59]. Vps25 is a subunit of the ESCRT-II complex, which binds to the RNA polymerase II elongation factor to exert transcriptional control in mammalian systems [60]. One TF family is suggested to be involved in translational regulation. CCAAT box is a common cis-acting element found in the promoter and enhancer regions of genes in the eukaryotes [61,62]. CBFs are necessary for 60S ribosomal subunit biogenesis and are therefore involved in translational control [63][64][65]. This family including Noc3, Noc4, and Mak21 in S. cerevisiae had three members in each genome, and a clear single-copy orthologous relationship could be observed for each member ( Figure S3A).

Cell Cycle Control
Five TF families were related to cell cycle control including cell cycle progression, DNA repair, and machinery/cell integrity maintenance.
One conserved TF family, APSES-type HTH, was reported to be involved in cell cycle control and crucial to development [66]. Every genome included in this study encoded four copies of the APSES-type HTH gene ( Figure S3B) that formed four clades of single-copy orthologs in all genomes except yeasts. Genes in Clade 1 included StuA homologs. As a target of the cyclic AMP (cAMP)-dependent protein kinase A (PKA) signal transduction pathway, StuA was reported to be involved in dimorphic switch [67,68], fungal spore development, and the production of secondary metabolites [69]. Genes in Clade 2 and Clade 3 included S. cerevisiae Swi4 and Swi6, which were reported to form a protein complex regulating cell cycle progression from G1 to the S phase [70] as well as meiosis [71]. Genes in Clade 4 included homologs of S. pombe Bqt4, anchoring telomeres to the nuclear envelope [72].
The conserved TF family, DTT, represented by the S. cerevisiae homolog Itc1, is recognized as a subunit of the ATP-dependent Isw2p-Itc1p chromatin remodeling complex and is required for the repression of early meiotic gene expression during mitotic growth [73].
The other conserved TF family, RFX, was reported to be involved in DNA repair. Each strain encoded two orthologous copies, except for F. venenatum encoding two copies within the RFX1 clade ( Figure S3C). Being a major transcriptional repressor of DNA-damageregulated genes in S. cerevisiae, Rfx1 functions in DNA damage repair and replication checkpoint pathways [74]. In F. graminearum, Rfx1 was reported to be essential in maintaining the genome integrity [75]. The other copy, Rsc9 in S. cerevisiae, was reported to be a member of the chromatin structure-remodeling complex RSC, which is involved in transcription regulation and nucleosome positioning [76,77].
The conserved TF family NFYA was reported to bind to the CCAAT box. All strains maintained a single copy of this family. Its yeast homolog, Hap2, has been reported to induce the expression of mitochondrial electron transport genes [78] and its F. verticillioides homolog NFYA Hap2 was reported to be essential for fungal growth and the virulence on maize stalks [79].
The MADS MEF2-like TF family including S. cerevisiae Rlm1 was reported to be a component of the protein kinase C-mediated MAP kinase pathway involved in maintaining cell integrity [80]. Having a paralog from the whole genome duplication in S. cerevisiae, Rlm1 was detected as a single copy gene in all filamentous fungi included in this study. Its member in F. verticilioides, Mef2 has been reported to play a vital role in sexual development [81].

Gene Family Contractions in FOSC Partially Caused by Whole Genome Duplication in Yeast
We detected an expansion score of less than 1 for four TF families, MATalpha_HMGbox, NOT4, MADS_SRF-like, and HSF (heat shock factor), reflecting some level of gene family contraction among members of FOSC compared to the two yeast genomes ( Figure S4).
TF family MATalpha_HMGbox including S. cerevisiae mating type protein alpha 1 has been reported to be a transcription activator that activates mating-type alpha-specific genes [82]. Reflecting the potential heterothallic mating strategy, all F. oxysporum Mat1-1 type strains contained this TF, but not the Mat1-2 strains, even though sexual reproduction has not been observed in FOSC [83].
TF family NOT4 was reported to be a component of the multifunctional CCR4-NOT complex, a global transcriptional repressor of the RNA polymerase II transcription [84]. Most genomes included in this study encoded one copy of this TF family, but some filamentous fungal genomes including A. nidulans, F. redolens, F. oxysporum strains NRRL26365, MRL8666, and PHW726 lost it. The functional implication of this loss remains to be discovered.
The contractions of the other two TF families, MADS SRF-like and HSF, were primarily caused by the whole genome duplication in yeast. In contrast to the contraction at the global scale, both TF families expanded among some FOSC strains when compared to other filamentous fungi ( Figure S4).
As reported in M. oryzae, MADS SRF-like TF, essential for the transcriptional regulation of growth-factor-inducible genes [85], is important for microconidium production and virulence in host plants [86]. Due to the event of whole genome duplication, the S. cerevisiae genome contained two copies of this TF family, while all filamentous genomes encoded a single copy. However, we detected an average of 2.73 copies among the phytopathogenic FOSC strains. There were six copies in the genome of Fo5176, a pathogen of Brassicaceae plants including A. thaliana (Table S1).
TF family HSF has been reported to activate the production of heat shock proteins that prevent or mitigate protein misfolding under abiotic/biotic stresses [87]. The S. cerevisiae genome contained five copies of the HSF TF family, while all non-FOSC filamentous fungi had three copies. Members of FOSC exhibited some level of expansion to four or five copies (Fo47:4, Fol4287: 5, II5: 4, HDV274: 4, and Fo5176: 4), with 1-2 copies encoded in ACs. Phylogenetically, all HSF TFs were clustered into three major clades, named as Skn7, Sfl1, and Hsf1 ( Figure 3A,B). All AC-encoding HSFs were phylogenetically close to Hsf1 ( Figure 3A). Based on the study in M. oryzae, the family Sfl1 is essential for vegetative growth, conidiation, sexual reproduction, and pathogenesis [88]. Based on a study in F. graminearum, the family Skn7 is involved in regulating the oxidative stress response and is essential for pathogenicity [89]. Our expression data generated during the plant colonization [24] supported the involvement of all three core genes during plant colonization ( Figure 3C). However, the Hsf1 accessory copies of these two strains were distinct, as the Fo47 AC-encoding Hsf1 was upregulated and the Fo5176 AC-encoding Hsf1 was downregulated, post inoculation ( Figure 3C), suggesting that their distinct regulatory function is involved in these two distinct interactions. TF family HSF has been reported to activate the production of heat shock proteins that prevent or mitigate protein misfolding under abiotic/biotic stresses [87]. The S. cerevisiae genome contained five copies of the HSF TF family, while all non-FOSC filamentous fungi had three copies. Members of FOSC exhibited some level of expansion to four or five copies (Fo47:4, Fol4287: 5, II5: 4, HDV274: 4, and Fo5176: 4), with 1-2 copies encoded in ACs. Phylogenetically, all HSF TFs were clustered into three major clades, named as Skn7, Sfl1, and Hsf1 ( Figure 3A,B). All AC-encoding HSFs were phylogenetically close to Hsf1 ( Figure 3A). Based on the study in M. oryzae, the family Sfl1 is essential for vegetative growth, conidiation, sexual reproduction, and pathogenesis [88]. Based on a study in F. graminearum, the family Skn7 is involved in regulating the oxidative stress response and is essential for pathogenicity [89]. Our expression data generated during the plant colonization [24] supported the involvement of all three core genes during plant colonization ( Figure 3C). However, the Hsf1 accessory copies of these two strains were distinct, as the Fo47 AC-encoding Hsf1 was upregulated and the Fo5176 AC-encoding Hsf1 was downregulated, post inoculation ( Figure 3C), suggesting that their distinct regulatory function is involved in these two distinct interactions.  [24]. See Section 2 for details of the data reprocessing and visualization.  [24]. See Section 2 for details of the data reprocessing and visualization.

Gain-of-Function among Filamentous Ascomycete Fungi
Three TF families, CP2 (EI y = 2.73), HTH_AraC (EI y = 2), and HTH_Psq (EI y = 3.53), were absent in both yeast genomes, suggesting a gain of function among filamentous ascomycete fungi (Table S1). TF family CP2 was studied in animal and fungal kingdoms with a function related to differentiation and development [90]. Both HTH_AraC and HTH_Psq are part of the helix-turn-helix (HTH) superfamily. HTH_AraC was first reported in bacteria as a positive regulator regulating the arabinose operon [91][92][93]. HTH_Psq, as part of the eukaryotic Pipsqueak protein family, has been reported in vertebrates, insects, nematodes, and fungi to regulate processes involved in cell death [94]. Most FOSC genomes encoded a single copy of HTH_AraC, while the number of HTH_Psq-containing proteins ranged from 0 to 9 in the FOSC and 0 to 3 in other Fusarium genomes. Since the HTH_Psq domain also exists in transposases [94], and ACs in FOSC are transposon-rich, it remains to be studied whether proteins containing the Psq domain are bona fide TFs.

Seven Exceedingly Expanded TF Families
Among the families containing minimally one yeast ortholog, seven TF families had expansion indices greater than 2 (Table 2 and Figure 2B) including Zn2-C6 (EI y = 15.09), bZIP (EI y = 5.80), Znf_C2H2 (EI y = 4.15), Homeobox (EI y = 2.28), PAI2 (EI y = 3.42), NDT80 (EI y = 3.47), and bHLH (EI y = 3.48). Based on the number of increments, the most significantly expanded TF families were Zn2-C6 (44 in yeasts versus 671 in FOSC) and Znf_C2H2 (40 in yeasts versus 167 in FOSC) ( Figure 2C and Table S1). Because of their large expansion, these seven families accounted for more than 75% of the total TFome. All seven families exhibited a gradual expansion, following the pattern yeasts < non-Fusarium filamentous fungi < non-FOSC Fusarium < FOSC (Table S1). Annotating large TF families could be challenging. Here, we described some examples based on the literature.
Zn2-C6, a fungal TF family [95], was detected as the most significant expanded TF family, reaching over 600 members among the FOSC genomes and accounting for more than half of the total TFome. Able to form a homodimer, this group of TFs are also able to bind to the specific palindromic DNA sequence through direct contact with the major groove of the double-stranded DNA molecules [95]. The versatility of this group of TFs can be achieved by domain shuffling and by changing the nucleotide binding specificity. In addition to the well-documented Ftf1 [17,[96][97][98][99], five additional TFs within this family were reported in F. oxysporum including Ctf1 [100], Ctf2 [100], Fow2 [101], XlnR [102] and Ebr1 [103]. Their functions were reported to be involved in the development, metabolism, stress response, and pathogenicity.
The other five families were bZIP, Homeobox, PAI2, Ndt80, and bHLH. The bZIP domain contains a region for sequence-specific DNA binding followed by a leucine zipper region required for dimerization [111]. Three F. oxysporum bZIP TFs were reported including Atf1 [112], Hapx [113], and MeaB [114], all of which are important for fungal pathogenicity. Homeobox is a DNA binding motif with a helix-turn-helix structure. In S. pombe, a homeobox-domain containing protein Phx1 was reported to be a transcriptional coactivator involved in yeast fission. In M. oryzae, the homeobox-domain containing protein Hox played roles in conidiation and appressorium development [115]. The TF family PAI2 is involved in the negative regulation of protease synthesis and sporulation of the Bacillus subtilis [116]. The TF family Ndt80 is essential for completing meiosis in S. cerevisiae [117,118] and Ustilago maydis [119] by promoting the expression of sporulation genes for the fulfillment of meiotic chromosome segregation [120]. The TF family bHLH forms a superfamily of transcriptional regulators found in almost all eukaryotes and is involved in diverse developmental processes [121]. In F. graminearum, a bHLH-domain containing protein Gra2 was reported to regulate the biosynthesis of phytotoxin gramillin [122], while a bHLH-domain containing protein SreA in Penicillium digitatum is required for anti-fungal resistance and full virulence in citrus fruits [123].

Other Families
The other 20 TF families (expanded but with EI y ≤ 2) accounted for 20% of the TFome, with an average 9.6 copies in each genome examined (Table S1).
Four TF families were functionally linked to chromatin remodeling including Bromodomain (EI y = 1.52), CBFA_NFYB (EI y = 1.35), Znf_RING-CH (EI y = 1.11), and ARID (EI y = 1.25). The TF family Bromodomain contained Spt7. As a crucial part of the SAGA complex in yeast, Spt7 has been reported to recognize the acetylated lysines of histones and eventually lead to chromatin unwinding [124]. The CBFA_NFYB domain was found in proteins (e.g., S. cerevisiae Dls1) that regulate RNA polymerase II transcription by controlling the chromatin accessibility (e.g., telomeric silencing) [125]. The TF family Znf_RING-CH also had a functional connection to chromatin modification (e.g., S. cerevisiae Rkr1) [126]. The domain ARID, a 100 amino acid motif, has been found in many eukaryotic TFs [127] such as Swi1 in S. cerevisiae, playing an important role in chromatin remodeling. This domain is also required to transcribe a diverse set of genes including some retrotransposons [128,129].
TFs belonging to the Ste12 family are only found in the fungal kingdom. Except for S. pombe, every genome encodes one copy. Binding to a DNA motif that mediates pheromone response, Ste12 TFs were reported to regulate fungal development and pathogenicity [130] and are involved in mating and pseudohyphae formation [131]. In F. oxysporum, Ste12, downstream of the Fmk1-mediated MAPK cascade, is involved in the control of invasive growth and fungal virulence [110].
Among the others, the Znf_NFX1 domain has been found in NK-X1, a repressor of the human disease-associated gene HLA-DRA [132]. The HMG_box (high mobility group box) in S. cerevisiae is seen in three proteins: Spp41, which is involved in the negative expression regulation of spliceosome components [133]; Nhp6a, which is required for the fidelity of some tRNA genes [134]; and Ixr1, a transcriptional repressor that regulates hypoxic genes [135]. Fep1, an example of Znf_GATA, was reported to be a transcriptional repressor involved in the regulation of some iron transporter genes under high iron concentrations [136]. S. cerevisiae Mbf1, belonging to Cro/C1-type HTH, is a transcriptional coactivator [137].

Orthologous Survey of TF Families That Were Manually Curated
To further understand expanded TFs and their impacts on transcriptional regulation, we curated a list of 102 TFs reported in the literature focusing on F. oxysporum, F. graminearum, and other phytopathogenic fungi (Table S3 and examples as described in the previous section). Compared to this list of curated TFs using Orthofinder, we defined 80 orthologous groups among the Fusarium genomes (Table S4). Sixty-two out of the 80 orthogroups were identified using the above IPR-annotated pipeline including 17 in Zn2C6, nine in Znf_C2H2, and one containing both the Znf_C2H2 and Zn2-C6 domains (Table S4). This helps add to the functional annotation of these large TF families while also adding additional annotation to 18 TF families (Table S4), accounting for 32 genes per genome (3% of average Fusarium TFome size). These newly annotated TFs include homologs of those without a domain annotation (e.g., disordered proteins F. oxysporum Ren1 [138], M. oryzae Som1 [139], and homologs of those with noncanonical TF domains such as Ankyrin_rpt and WD40_repeat).
We then directly compared F. oxysporum with its Fusarium relatives to calculate the expansion index as follows: The EI f ranged from the highest score of 3.54 (Fug, AreA_GATA) to the lowest score of 0.5 (Fox1, Fork_head) (Table S4). Among these 80 orthogroups, 36 groups were conserved (EI f = 1), with one gene per genome. Ten of these conserved groups were functionally validated in F. oxysporum (Table S4). Twenty four groups had scores less than 1, while 20 groups had a score greater than 1 (Table 3 and Table S4). Expanded groups included Fug1 (AreA_GATA, EI f = 3.54), Cos1 (Znf_C2H2, EI f = 2.8), Ftf1/Ftf2 (Zn2-C6, EI f = 2.7), Ebr1/Ebr2 (Zn2-C6, EI f = 2.5), and Ren1 (disordered, EI f = 2). We also identified PacC (EI f = 1.57) as the second most expanded group within the highly expanded Znf_C2H2 family. We will further discuss these six groups (Table 3). Both Ftf1/Ftf2 and Ebr1/Ebr2, belonging to the Zn2-C6 family, contributes directly to fungal virulence [3,17,97]. The deletion of AC-encoding Ftf1 reduced the pathogenicity of F. oxysporum f. sp. phaseoli [97], highlighting the direct functional involvement of AC TF in virulence. In Fol, deleting either Ftf1 (AC encoding) or Ftf2 (CC encoding) reduced the fungal virulence [96,152]. Constitutive expression of either Ftf1 or Ftf2 induced the expression of effector genes [17]. The core copy Ftf2 was conserved among all Fusarium species, and the AC copy Ftf1 was only found in F. oxysporum and F. redolens (Figure 4). Ebr1 had multiple homologs in F. oxysporum, but a single copy in F. graminearum [103]. The F. oxysporum genome had three AC-encoding paralogs: Ebr2, Ebr3, and Ebr4. Interestingly, these AC-encoding paralogs are regulated by core copy Ebr1 [103]. It is worth noting that the Ebr2 coding sequence driven by the Ebr1 promoter was able to rescue the Ebr1 knockout mutation, indicating some functional redundancy of this family.
CC-encoding PacC gene-dependent as the induction disappeared in the CC-encoding PacC knockout mutant, further supporting a cross-talk between the core and accessory TFs . Similar to EBR1, the expression of the AC-encoding PacC genes was much lower than that of the CC-encoding PacC gene, and knockouts of one AC PacC gene affected a small subset of genes compared with the CC PacC knockout, which had a broader effect on the cellular processes [49].  . Unique expansion of some TFs, driven by ACs, may provide clues to host-specific adaptation. RNA-Seq data were previously described [24,25]. (A) Ftf1, the TF involved in tomato pathogenicity was most significantly expanded (10 copies of accessory FTFs) in the tomato pathogen Fol4287 genome and the expression of eight out of 10 were induced during plant colonization. (B) Ren1 was the most significantly expanded (seven copies of accessory RENs) in the Arabidopsis pathogen Fo5176 genome, and two of them were induced during plant colonization.
Fug1 has a role in pathogenicity (maize kernel colonization) and fumonisin biosynthesis in F. verticillioides [147]. The deletion of Fug1 increased the sensitivity to the antimicrobial compound 2-benzoxazolinone and to hydrogen peroxide, suggesting its role in mitigating stresses associated with the host defense [147]. Neither CC nor AC-encoding copies of these two genes were experimentally examined in FOSC. Ren1, a disordered protein without IPR functional domain, was expanded with a EIf score of 2 among the FOSC. However, the only reported study on its function is in F. oxysporum f. sp. melonis, regulating the development of conidiation [138].

Transcriptome Analysis to Probe the Essential TFs during Host Colonization
To understand the functional importance of FOSC TFs, we took advantage of two recently reported transcriptomics datasets [24,25] including pathogenic interactions . Unique expansion of some TFs, driven by ACs, may provide clues to host-specific adaptation. RNA-Seq data were previously described [24,25]. (A) Ftf1, the TF involved in tomato pathogenicity was most significantly expanded (10 copies of accessory FTFs) in the tomato pathogen Fol4287 genome and the expression of eight out of 10 were induced during plant colonization. (B) Ren1 was the most significantly expanded (seven copies of accessory RENs) in the Arabidopsis pathogen Fo5176 genome, and two of them were induced during plant colonization.
Both Cos1 and PacC are part of the Znf_C2H2 family. In M. oryzae, Cos1 was reported to be involved in conidiophore development [112] and functions as a negative regulator reducing fungal pathogenicity [153]. PacC has been reported as an important pH-responsive TF in F. oxysporum [18,107]. This TF family was expanded in clinical strains, showing an average accessory copy number of 3.7 of FOSC, whereas the non-clinical strains showed an average accessory copy number of 0.5. All of the Fusarium relatives' genomes examined only contained a single copy of the core PacC. Our previous study using one F. oxysporum clinical isolate revealed that the expression of all PacC genes can be induced with a pH shift from 5.0 to 7.4 (the mammalian physiological pH), indicating a potential role in host adaptation [18]. Interestingly, the induction of AC-encoding PacC genes was CCencoding PacC gene-dependent as the induction disappeared in the CC-encoding PacC knockout mutant, further supporting a cross-talk between the core and accessory TFs. Similar to EBR1, the expression of the AC-encoding PacC genes was much lower than that of the CC-encoding PacC gene, and knockouts of one AC PacC gene affected a small subset of genes compared with the CC PacC knockout, which had a broader effect on the cellular processes [49].
Fug1 has a role in pathogenicity (maize kernel colonization) and fumonisin biosynthesis in F. verticillioides [147]. The deletion of Fug1 increased the sensitivity to the antimicrobial compound 2-benzoxazolinone and to hydrogen peroxide, suggesting its role in mitigating stresses associated with the host defense [147]. Neither CC nor AC-encoding copies of these two genes were experimentally examined in FOSC. Ren1, a disordered protein without IPR functional domain, was expanded with a EI f score of 2 among the FOSC. However, the only reported study on its function is in F. oxysporum f. sp. melonis, regulating the development of conidiation [138].

Transcriptome Analysis to Probe the Essential TFs during Host Colonization
To understand the functional importance of FOSC TFs, we took advantage of two recently reported transcriptomics datasets [24,25] including pathogenic interactions (Fo5176 infecting Arabidopsis and Fol4287 infecting tomato) and endophytic interactions (Fo47 colonizing Arabidopsis) (Supplementary Dataset).
By examining patterns of expression (Table S5), we found that almost all genes within the conserved category (58 out of 60) (Group 1) were consistently expressed (TPM > 1 across all conditions), supporting their general roles in controlling life processes. Within the expanded category (Group 3), the proportion of genes being consistently expressed ranged from 41% to 59% for core TFs and only from 5% to 16% for AC-encoding TFs. With a less strict filter (TPM > 1 at minimum 1 condition), we found that all genes within the conserved category were expressed. Within the expanded category, 93% of core TFs and between 49% and 67% of AC-encoding TFs were expressed. The significant increase in the AC-encoding TFs with a lower stringency further supported their conditional involvement in niche adaptation.
We further reviewed the expression patterns of the reported TFs in Fol4287 (Table S6). Out of the 27 TFs encoded on the core genome, 18 were upregulated (defined by upregulation under at least three out of four in planta conditions compared to the axenic growth) during plant colonization, which is consistent with their reported roles in pathogenicity. The AC-encoding Ftf1 has been reported to play essential functions in fungal pathogenicity [96]. Of the ten accessory Ffs, eight were upregulated during plant colonization.
Using a higher stringency filter, selecting TFs upregulated under all in planta conditions, we searched for: (1) conserved core TFs that may be related to plant colonization and (2) expanded AC TFs that could be related to host-specific pathogenicity. In the Fol4287, Fo5176, and Fo47 genomes, 95, 62, and 44 core TFs were upregulated during plant colonization, respectively. Among them, ten copies were highly conserved (Table S7) as they were single-copy orthologs across all 15 F. oxysporum strains including Fow2 and Sfl1. Fow2, a Zn2C6 TF, is required for full virulence but not hyphal growth and conidiation in F. oxysporum f. sp. melonis [101]. Sfl1 was reported to be essential for vegetative growth, conidiation, sexual reproduction, and pathogenesis in M. oryzae [88].
Fol4287, Fo5176, and Fo47 contained 29, 34, and nine upregulated accessory TFs, respectively, including Ftf1 and Ren1 (Figure 4 and Table S8). Ftfs have been reported to play an essential role in the pathogenicity in Fol4287, although their involvement in other interactions has not. The Fol4287 genome encoded 10 accessory Ftf s and eight were upregulated during plant colonization. The Fo5176 genome encoded six accessory Ftf s, but only one copy was upregulated during plant colonization. Interestingly, eight upregulated Fol4287 and one upregulated Fo5176 Ftf s were clustered together (Figure 4).
The unique expansion with regulatory adaptation (i.e., fine-tuned expression regulation) seemed to be restricted to Fol4287 and not the other pathogenic strain, Fo5176. In contrast, the Fo5176 genome encoded seven accessory Ren1 TF and two were upregulated during plant colonization (Figure 4), while the Fol4287 genome had only one accessory Ren1 not involved in host colonization. While functional validation is needed, strain-specific expansion followed by fine-tuned expression regulation when infecting host species exists and likely contributes to host-specific pathogenicity.

Discussion
For a soilborne pathogen with strong host specificity like FOSC, the adjustment of growth and cell cycle control in response to environmental cues is likely to be essential for survival. Expanded TF families likely contribute to the enhanced functions related to niche adaptation as these TF families play important roles in transmitting external and internal signals and regulating complex cellular signaling responses to the sensed stimuli. Therefore, it is not surprising that the genomes of FOSC had larger TFome sizes than the other fungi included in the study. The expansion of TFs among the FOSC resulted in a positive correlation between the total number of proteins and the size of the fungal TFome, which was also observed in other instances [22].
A total of 14 TF families that control the global transcriptional event such as TBP are highly conserved within the ascomycete fungal lineages. Conserved regulatory mechanisms revealed through this study suggest that the plant colonization process could be a common process among FOSC strains regardless of their host-specific pathogenesis. This notion is also supported by recent studies that highlighted the ability of FOSC strains as root colonizers regardless whether of they cause disease or function as endophytes [25,154].
In contrast to these stable TFs, 30 families were expanded to various degrees, and the most significant expansions occurred in the Zn2-C6 and Znf_C2H2 TF families among the FOSC genomes. The number of Zn2-C6 TFs increased significantly (with the highest expansion score) and made up most of the TFs (56.7%) found within the FOSC TFome. For example, Ftf1, a TF belonging to Zn2-C6 and is involved in tomato pathogenicity, was most significantly expanded to 10 copies of accessory Ftfs in the tomato pathogen Fol4287 genome. Eight out of 10 were induced during plant colonization.
The unique expansion of some TFs, driven by ACs, may provide clues as to hostspecific interactions. Acquiring additional TFs will modify existing regulatory pathways, and this will require fine-tuning existing networks for this group of organisms to successfully adapt to different hosts under diverse environments. A previous survey of kinome (the complete set of protein kinases encoded in an organism's genome) among the FOSC and other Ascomycetes also revealed a positive correlation between the size of the kinome and the size of the genome [48], identical to what we reported here for TFomes. As kinases and TFs are key regulators that modulate all important signaling pathways and are essential for the proper functions of almost all molecular and cellular processes, strong correlations between kinome and TFome suggest the ordered recruitment and establishment of ACs among FOSC genomes.
This realization further emphasizes the importance of additional functional studies. Reverse genetics is a powerful tool in defining the functional importance of a TF. For example, TF Ren1, a disordered protein, was identified by genetic and molecular characterization [138]. This TF is most significantly expanded (seven copies of accessory Rens) in the Arabidopsis pathogen Fo5176 genome and is involved in plant colonization. High throughput approaches such as chromatin immunoprecipitation sequencing (CHIP-Seq) and DNA affinity purification sequencing (DAP-Seq) [155] can be used to profile the cis-regulatory elements globally for a better understanding of transcriptional regulation in the fungal model F. oxysporum. Gene regulatory networks [156,157] can add more resolution to these complex regulatory processes. However, the ultimate understanding of the regulatory roles of each TF will come from careful molecular and biochemical characterization.
Our study offers a comprehensive look at the regulation from the evolutionary perspective while also providing an easily implemented computational pipeline to compare TFs and other functional groups in fungi. A better understanding of functions of TFs will not only inform Fusarium biology [158], but can also be extrapolated to other filamentous fungal systems.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/jof9030359/s1, Table S1. TF-type DNA-binding domains used to filter TFome. Table S2. TFome annotations across 30 genomes. Table S3. Curated TFs in phytopathogens. Table S4. Ortholog copy number of characterized TFs of phytopathogens across 30 genomes. Table S5. Number of TFs being expressed. Table S6. Orthologs of the reported TFs in Fol4287. Table S7. Highly conserved TFs that are constantly upregulated during plant colonization. Table S8. Strain-specific accessory TFs that are upregulated during plant colonization. Figure S1. TFome annotation pipelines. Figure S2. ACs contribute to the FOSC TFome expansion. Figure S3. Phylograms of three conserved families. Figure S4. Minimal gene family contractions in FOSC partially caused by whole genome duplication in yeast. Figure   Data Availability Statement: Unprocessed RNA-Seq data were retrieved from the public source as described in Section 2.2. All processed data supporting our results and conclusions are available in the Supplemental Materials.