Three Small Cysteine-Free Proteins (CFP1–3) Are Required for Insect-Pathogenic Lifestyle of Metarhizium robertsii

Unique CFP (cysteine-free protein; 120 aa) has been identified as an extraordinary virulence factor in Beauveria bassiana (Cordycipitaceae), a main source of wide-spectrum fungal insecticides. Its homologs exclusively exist in wide-spectrum insect pathogens of Hypocreales, suggesting their importance for a fungal insect-pathogenic lifestyle. In this study, all three CFP homologs (CFP1–3, 128–145 aa) were proven essential virulence factors in Metarhizium robertsii (Clavicipitaceae). Despite limited effects on asexual cycles in vitro, knockout mutants of cfp1, cfp2 and cfp3 were severely compromised in their capability for normal cuticle infection, in which most tested Galleria mellonella larvae survived. The blocked cuticle infection concurred with reduced secretion of extracellular enzymes, including Pr1 proteases required cuticle penetration. Cuticle-bypassing infection by intrahemocoel injection of ~250 conidia per larva resulted in a greater reduction in virulence in the mutant of cfp1 (82%) than of cfp2 (21%) or cfp3 (25%) versus the parental wild-type. Transcriptomic analysis revealed dysregulation of 604 genes (up/down ratio: 251:353) in the Δcfp1 mutant. Many of them were involved in virulence-related cellular processes and events aside from 154 functionally unknown genes (up/down ratio: 56:98). These results reinforce the essential roles of small CFP homologs in hypocrealean fungal adaptation to insect-pathogenic lifestyle and their exploitability for the genetic improvement of fungal insecticidal activity.


Introduction
Beauveria (Codycepitaceae) and Metarhizium (Clavicipitaceae) in Hypocreales are insectpathogenic fungi deficient of teleomorph and serve as the main sources of environmentfriendly mycoinsecticides [1,2]. Fungal insecticidal activity is an overall output of various cellular processes and events involved in fungus-insect interactions. Insect-pathogenic fungal genomes comprise approximately 15% genes involved in fungal infection and kill action and hence are speculated to be virulence factors [3][4][5][6][7]. Among hundreds of genes characterized in the post-genomic era, however, virulence factors exploitable for genetic improvement of fungal insecticides remain limited [8][9][10].
Fungal insect pathogenicity and virulence are distinct, but related, concepts in a general sense [11]. After conidial germination on insect surfaces, hyphal invasion into the insect body begins under the actions of extracellular (secreted) enzymes involved in proteolysis, chitinolysis and lipolysis [12][13][14], such as subtilisin-like Pr1 family proteases collectively required for cuticular penetration [15], and/or of the mechanic pressure of appressoria formed at the hyphal tips. As specialized infective structures, appressoria often form in the infection course of Metarhizium spp. [16,17] instead of Beauveria bassiana [18], implicating somewhat distinctive mechanisms of the two fungal lineages in the host infection. Upon arrival in the host hemocoel, hyphae must overcome the host immune defense responses [4,13] before the formation of hyphal bodies (thin-wall blastospores), which proliferate in vivo by yeast-like budding to accelerate mycosis development and host mummification to death [19][20][21]. Thus, fungal insect pathogenicity and virulence rely upon the capabilities of normal infection through cuticular penetration and subsequent hemocoel colonization, respectively.
Many small proteins, particularly cysteine-rich proteins, secreted in the host infection course of plant-pathogenic fungi, are identified virulence factors [22][23][24], such as multiple Avr proteins targeted by resistant proteins in the host plant [25,26] and extracellular proteins to stimulate a hypersensitive response of the host plant [27,28]. Such small secreted proteins (SSPs) are also speculated as a large class of key virulence factors in fungal insect pathogens [3,13] despite a scarcity of direct evidence. Revealed by previous time-course transcriptome analyses, hundreds of secreted and non-secreted protein genes were highly induced in the first 48-h infection course of cuticular penetration and hemocoel colonization by B. bassiana against Plutella xylostella [29]. Some of those genes have been identified as virulence factors due to sharply reduced or even abolished pathogenicity in the absence of each gene [30][31][32][33][34]. Among the studied genes, only one encodes small cysteine-rich protein [13 Cys resides in 230 amino acids (aa)] but proved a vacuole-localized protein (VLP4) [30]. In a recent study to characterize 10 SSP genes in B. bassiana, only one encoding small cysteine-free protein (CFP, 120 aa) was identified as an extraordinary virulence factor, and the remaining SSP genes contributed little to virulence irrespective of their encoding cysteine rich or deficient proteins [35]. Overexpression of this CFP gene resulted in a marked increase in the fungal insecticidal activity [36]. The previous studies indicate that cysteine richness is not necessarily a criterion for SSPs to be considered as virulence factors in fungal insect pathogens. Revealed by genome survey, one to three CFP homologs (96-145 aa) exist exclusively in six wide-spectrum insect pathogens of Codycepitaceae (Beauveria brongniartii and Cordyceps fumosorosea) and Clavicipitaceae (Metarhizium anisopliae, M. brunneum, M. guizhouense and M. robertsii) [35]. However, no homolog was found in other ascomycetes, including entomopathogenic (Ascosphaera api, C. militaris, C. confragosa, Moelleriella libera and Sporothrix insectorum) or non-entomopathogenic (Aspergillus, Botrytis, Fusarium, Neurospora and Pyricularia) fungi. Neither was any CFP homolog found in the specialists M. acridum (Orthoptera) and M. rileyi (Lepidoptera). Since the six insect pathogens are all important sources of fungal pesticides [1], their CFP homologs are possible virulence factors to be exploited for improving the insecticidal activities of those fungi as does CFP in B. bassiana. This study sought to test the hypothesis by functional analysis of three CFP homologs (named CFP1-3) in M. robertsii. As presented below, CFP1-3 were evidently required for insect pathogenicity and virulence of M. robertsii, and the most important CFP1 was involved in transcriptional mediation of hundreds of genes associated with virulence-related cellular events.

Generation of Targeted Gene Mutants
The nucleotide sequences of cfp1 (447 bp), cfp2 (489 bp) and cfp3 (509 bp) were disrupted by deleting the promoter/coding regions of 438 (46 + 392), 459 (75 + 384) and 362 (16 + 346) bp from the genome of the wild-type strain M. roberstii ARSEF 2575 (designated WT), respectively (Figure S1A-C). The promoter/coding region of each target gene was deleted by homologous recombination of 5 and 3 flanking/coding fragments separated by bar marker in the constructed vector p0380-5 x-bar-3 x (x = cfp1, cfp2 or cfp3); the deleted region was complemented into an identified deletion mutant (DM) of each gene by ectopic integration of a cassette composed of full-length coding and flanking sequences and sur marker in the vector p0380-x-sur. The two vectors of each target gene were constructed and integrated into the respective WT and DM strains via Agrobacterium mediated transformation, as described previously [35]. Putative mutant colonies screened using bar resistance to phosphinothricin (200 µg/mL) or sur resistance to chlorimuron ethyl (10 µg/mL) were identified via PCR and real-time quantitative PCR (qPCR) analyses. All PCR and qPCR primers used for DNA amplification, vector construction and targeted gene detection are listed in Table S1. The identified DM and CM strains (Figure S1D-I) were evaluated together with the WT strain in the experiments, which included three independent replicates to meet a requirement for one-factor (strain) analysis of variance and Tukey's honestly significant difference (HSD) test for phenotypic differences between DM and control (WT and CM) strains.

Assessments of Growth Rates under Different Culture Conditions
All DM and control strains were grown at 25 • C and L:D 12:12 for 7 d on the minimal medium CDA (Czapek-Dox agar: 3% sucrose, 0.3% NaNO 3 , 0.1% K 2 HPO 4 , 0.05% KCl, 0.05% MgSO 4 and 0.001% FeSOa 4 plus 1.5% agar) and CDAs amended with different carbon or nitrogen sources or with deleted carbon or nitrogen source. The colony growth of each strain was commenced by spotting 1 µL of a 10 6 conidia/mL suspension per plate. As an index of radial growth, the diameter of each colony was measured perpendicularly to each other across the center. The same method was also used to initiate radial growth on the plates of CDA alone (control) or supplemented with different stress agents. The tested agents included the osmotic salt NaCl (0.7 M), the osmotic carbohydrate sorbitol (1.5 M), the oxidants menadione (0.02 mM) and H 2 O 2 (2 mM), and the cell wall perturbing chemicals Congo red (12 µg/mL) and calcofluor white (20 µg/mL). After a 7d incubation at 25 • C, each colony diameter was measured as aforementioned. The sensitivity of each strain to each of the tested stresses was assessed as the percentage of relative growth inhibition (RGI) using the formula (d c − d s )/d c × 100, where d c and d s denote the diameters of control and stressed colonies, respectively.

Assessments of Conidial Yield and Quality
Since the standard medium SDAY (Sabouraud dextrose agar plus yeast extract) for fungal insect pathogens is too rich for the WT strain's conidiation, cultures used for assessment of conidial yield per strain were initiated by spreading 100 µL aliquots of a 10 7 conidia/mL suspension on the plates (9 cm diameter) of 1/4 SDAY amended with one-fourth (1.0% glucose, 0.25% peptone, 0.25% yeast extract and 1.5% agar) of each SDAY nutrient. During a 15-d incubation at the optimal regime of 25 • C in a light/dark (L:D) cycle of 12:12, three samples were taken from each of 8, 10 and 15 d-old plate cultures using a cork borer (5 mm diameter). Conidia in each sample were washed off into 1 mL of 0.05% Tween 80 through a 10-min supersonic vibration. The conidial concentration in the suspension was quantified using a hemocytometer and converted to an absolute yield (no. conidia/cm 2 plate culture). The quality of conidia from the 15 d-old SDAY cultures was evaluated with two indices, namely median germination time (GT 50 ) assessed at 25 • C and hydrophobicity assessed in a biphasic (aqueous-organic) system, as described previously [35].

Fungal Virulence Bioassays
Standardized bioassay for fungal virulence was initiated by topical application (10-s immersion in 40 mL aliquots) of a 2 × 10 7 conidia/mL suspension to three groups of 35 Galleria mellonella larvae (fifth instar) for normal cuticle infection (NCI) by each strain.
Alternatively, cuticle-bypassing infection (CBI) was commenced by injecting 5 µL of a 5 × 10 4 conidia/mL suspension into the hemocoel of each larva (~250 conidia injected) in each group. All treated groups were maintained at 25 • C and monitored daily for survival or mortality records until larval mortality reached 100% or had no more change. The resultant time-mortality trend in each group was subjected to modeling analysis for the estimation of median lethal time LT 50 (d) as a virulence index.
Total activities (U/mL) of extracellular (proteolytic, chitinolytic and lipolytic) enzymes and Pr1 family proteases crucial for successful NCI were measured from the supernatant of each of three 3 d-old CDB (i.e., agar-free CDA) cultures containing the sole nitrogen source of 0.3% bovine serum albumin (BSA), as described elsewhere [15,20]. The submerged cultures were generated by shaking 50 mL aliquots of a 10 6 conidia/mL suspension in CDB-BSA at 25 • C.
During the period of bioassays, hemolymph samples were taken from the larvae surviving 48-96 h post-CBI and 96-168 h post-NCI and examined to reveal a status of proliferation in vivo. Since hyphal bodies were not directly observable in the samples, 100 µL of hemolymph from three larvae (infected per strain) on each sampling occasion was added to 3 mL of SDBY (i.e., agar-free SDAY) containing kanamycin (100 ng/mL) and ampicillin (200 ng/mL) for bacterial inhibition, followed by a 36-h incubation at 25 • C on a shaking bed (150 rpm). Cells were collected by centrifugation and resuspended in 0.05% Tween 80. Samples from each suspension were then examined under a microscope to reveal the status and abundance of fungal cells and insect hemocytes.

Transcriptomic Analysis
Three cultures (replicates) of the ∆cfp1 and WT strains grown for 3 d on cellophaneoverlaid plates of 1/4 SDAY at the optimal regime were sent to Lianchuan BioTech Company (Hangzou, China) for construction and analysis of cfp3-specific transcriptomes, as described previously [35]. Clean tags generated by filtration of all raw reads from sequencing were mapped to the M. robertsii genome [4]. The significant levels of both log 2 ratio (fold change) ≤−1 or ≥1 and q < 0.05 were used to identify down-or up-regulated genes (DEGs). All identified DEGs were enriched to GO terms in three function classes (p < 0.05) through Gene Ontology (GO) analysis (http://www.geneontology.org/ (accessed on 4 June 2022)) and to various pathways (p < 0.05) through Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis (http://www.genome.jp/kegg/ (accessed on 4 June 2022)).

Sequence Features of CFP1-3
The BLASTp search resulted in the identification of CFP1-3 homologs (EFY97868, EFY97692 and KHO10933, respectively) previously annotated as hypothetical proteins in the M. roberstii genome [4]. CFP1 (128 aa) is close to the query of B. bassiana CFP in molecular size but slightly smaller than CFP2 (141 aa) or CFP3 (145 aa). Sequence alignment analysis revealed 40-43% identities of CFP1-3 to the query ( Figure 1). The three homologs share the same sequence features with the query, including no Cys residue, no predictable domain and the presence of a signal peptide (residues 1-17) at N-termini. In addition, one or two NLS motifs were predicted from the sequences of CFP1 (residues 49-78), CFP2 (residues 24-56 and 105-136) and CFP3 (residues 58-91 and 110-144). Overall, the CFP1-3 homologs in M. roberstii are similar to the query in molecular size and sequence feature.

Limited Roles of cfp1-3 in Asexual Cycle In Vitro
Compared to the WT strain, the Δcfp1 and Δcfp3 mutants showed minor defects in radial growth (suppressed by 10-20%) on CDA or CDAs amended with some of the tested carbon or nitrogen sources (Figure 2A) at the optimal regime of 25 °C and L:D 12:12. Despite little change in radial growth on the tested media, the Δcfp2 mutant was slightly (7%) more sensitive to osmotic stress induced by sorbitol (1.5 M) among several chemical stressors added to CDA, contrasting to a slightly increased (~9%) tolerance of Δcfp1 and Δcfp3 to cell wall stress induced by calcofluor white (20 μg/mL) and of Δcfp1 to oxidative stress induced by menadione (0.02 mM) ( Figure 2B). All DM and control strains were equally responsive to the stresses induced by NaCl (0.7 M), Congo red (12 μg/mL) and H2O2 (2 mM). Moreover, conidial yields were measured from the cultures of 1/4 SDAY during a 15-d incubation at the optimal regime of 25 °C and L:D 12:12. As a result, conidial yields were reduced by 27-41% in Δcfp2 during the period but not significantly affected in the Δcfp1 and Δcfp3 mutants compared to the WT strain ( Figure 2C). Median germination time (GT50) of conidia at 25 °C was moderately shortened in Δcfp2 alone ( Figure 2A) while conidial hydrophobicity was lowered by ~6% in Δcfp2 and Δcfp3 ( Figure 2E).
The above phenotypes were generally restored by targeted gene complementation. These data demonstrated differential, but very limited, roles for cfp1-3 in radial growth, aerial conidiation and stress responses associated with the fungal asexual cycle in vitro.

Limited Roles of cfp1-3 in Asexual Cycle In Vitro
Compared to the WT strain, the ∆cfp1 and ∆cfp3 mutants showed minor defects in radial growth (suppressed by 10-20%) on CDA or CDAs amended with some of the tested carbon or nitrogen sources (Figure 2A) at the optimal regime of 25 • C and L:D 12:12. Despite little change in radial growth on the tested media, the ∆cfp2 mutant was slightly (7%) more sensitive to osmotic stress induced by sorbitol (1.5 M) among several chemical stressors added to CDA, contrasting to a slightly increased (~9%) tolerance of ∆cfp1 and ∆cfp3 to cell wall stress induced by calcofluor white (20 µg/mL) and of ∆cfp1 to oxidative stress induced by menadione (0.02 mM) ( Figure 2B). All DM and control strains were equally responsive to the stresses induced by NaCl (0.7 M), Congo red (12 µg/mL) and H 2 O 2 (2 mM). Moreover, conidial yields were measured from the cultures of 1/4 SDAY during a 15-d incubation at the optimal regime of 25 • C and L:D 12:12. As a result, conidial yields were reduced by 27-41% in ∆cfp2 during the period but not significantly affected in the ∆cfp1 and ∆cfp3 mutants compared to the WT strain ( Figure 2C). Median germination time (GT 50 ) of conidia at 25 • C was moderately shortened in ∆cfp2 alone (Figure 2A) while conidial hydrophobicity was lowered by~6% in ∆cfp2 and ∆cfp3 ( Figure 2E).
The above phenotypes were generally restored by targeted gene complementation. These data demonstrated differential, but very limited, roles for cfp1-3 in radial growth, aerial conidiation and stress responses associated with the fungal asexual cycle in vitro.

Indispensable Roles of cfp1-3 in Insect Pathogenicity and Virulence
In the standardized bioassays, the control strains caused 100% mortality of G. mellonella larvae within 17 d after NCI or 7 d after CBI ( Figure 3A). In contrast, the larvae inoculated with Δcfp1, Δcfp2 and Δcfp3 for NCI still survived at 76%, 62% and 52% at the end of 17 d, respectively. All larvae inoculated with each DM for CBI also survived longer. Particularly, 37% of tested larvae survived the Δcfp1 CBI at the end of 10 d. Consequently, the average LT50 estimates were 8.35 (±0.36) d for the control strains against the model insect via NCI and 3.72 (±0.14) d via CBI ( Figure 3B). However, LT50 was unavailable (not robustly assessable) for all DM strains (unreliable estimates: 28.3, 19.2 and 16.7 d) via NCI,

Indispensable Roles of cfp1-3 in Insect Pathogenicity and Virulence
In the standardized bioassays, the control strains caused 100% mortality of G. mellonella larvae within 17 d after NCI or 7 d after CBI ( Figure 3A). In contrast, the larvae inoculated with ∆cfp1, ∆cfp2 and ∆cfp3 for NCI still survived at 76%, 62% and 52% at the end of 17 d, respectively. All larvae inoculated with each DM for CBI also survived longer. Particularly, 37% of tested larvae survived the ∆cfp1 CBI at the end of 10 d. Consequently, the average LT 50 estimates were 8.35 (±0.36) d for the control strains against the model insect via NCI and 3.72 (±0.14) d via CBI ( Figure 3B). However, LT 50 was unavailable (not robustly assessable) for all DM strains (unreliable estimates: 28.3, 19.2 and 16.7 d) via NCI, and prolonged by 82% for ∆cfp1, 21% for ∆cfp2 and 25% for ∆cfp3 via CBI in comparison to the mean from the control strains. and prolonged by 82% for Δcfp1, 21% for Δcfp2 and 25% for Δcfp3 via CBI in comparison to the mean from the control strains.  Extracellular enzymes (ECEs) and Pr1 family proteases are collectively required for cuticle degradation during NCI [14,15]. For insight into blocked NCI of each DM, total activities of ECEs and Pr1 proteases were quantified from the supernatants of 3 d-old CDB-BSA cultures. The two quantities were significantly reduced by 58% and 15% in ∆cfp1, 57% and 28% in ∆cfp2 and 34% and 5% in ∆cfp3 relative to the WT strain, respectively ( Figure 3C).
Next, we tried to directly observe the status of yeast-like budding proliferation in vivo in the hemolymph samples of infected larvae under a microscope as often observed in the samples of larvae infected by B. bassiana [35,36]. Unexpectedly, typical hyphal bodies were not observable in the samples of the larvae infected by control strains in either mode regardless of being stained with or without a fungal cell wall-specific dye (cottonblue or calcafluor white), suggesting a difference in the in vivo status between M. robertsii and B. bassiana. Alternatively, every 100 µL of hemolymph added to 3 mL SDBY containing kanamycin (100 ng/mL) and ampicillin (200 ng/mL) for bacterial inhibition was shaken for 36 h at the regime of 25 • C and 150 rpm. Revealed by microscopic examination of the resultant cells resuspended in aqueous 0.05% Tween 80, the WT and CM cells released from aggregated hemocytes were readily observed in the samples taken 48 h post-CBI but the ∆cfp1 cells were not observable until 96 h post-CBI (lower panels in Figure 3D). Most host hemocytes were lysed in the samples at 48 h after CBI by the control strains but remained intact in the larvae injected with ∆cfp1 at the same time. Moreover, the release of the ∆cfp1 cells from host hemocytes was largely delayed in the samples of surviving larvae post-NCI and very limited even at the end of 144 h post-NCI (upper panels in Figure 3D). On the same sampling occasion, almost all released cells of the control strains were proliferating by yeast-like budding, which concurred with the disappearance of intact hemocytes. These observations implied that the fungal capability of proliferation in vivo was compromised in the absence of cfp1.
Altogether, three CFP genes were all indispensable for fungal insect pathogenicity and virulence via NCI, which was blocked or severely compromised in the absence of each CFP gene. The indispensability relied upon their roles in the processes of host infection through cuticular penetration and subsequent hemocoel colonization. The results also demonstrated a greater role of cfp1 than of cfp2 and cfp3 in the infection cycle of M. robertsii.

Transcriptomic Insight into Indispensable Role of cfp1 in Infection Cycle
For in-depth insight into the indispensability of cfp1 for the fungal virulence, three 3 dold cultures (replicates) of the ∆cfp1 and WT strains grown on cellophane-overlaid plates of 1/4 SDAY at the optimal regime were used for transcriptomic analysis. Up to 10,682 genes were detected in the transcriptome and mapped to the fungal genome [4]. Of those, 604 DEGs (up/down ratio: 251:353, the same meaning for ratios mentioned below) were identified at significant levels of log 2 R ≤ −1.00 or ≥1.00 and q < 0.05 (Table S2) 4). The class also comprised more small terms, such as cell wall (1:1) and membrane raft (0:2). The low up/down ratios in these terms implied that cellular composition and membranes crucial to extracellular activity were compromised in the ∆cfp1 mutant.
Illustrated in Figure 4C were There were many more DEGs associated with or responsible for the ∆cfp1 mutant's main phenotypes. Listed in Table S7 were 39 DEGs (8:31) encoding varieties of hydrolases involved in cuticle degradation, 15 DEGs (3:12) involved in cell wall integrity and hydrocarbon epitopes and 43 DEGs (11:32) involved in cellular transport and homeostasis vital for manifold cellular processes and events. The majority of these genes were repressed at the transcriptional level. In addition, dozens of DEGs were involved in direct or indirect transcription mediation, including the coding genes of 22 putative transcription factors (9:13) and of 53 proteins/enzymes involved in posttranslational modifications (5:8) or RNA/DNA processing and events (20:20). However, the key developmental activator genes were not present in the list of identified DEGs, coinciding well with an insignificant change in the mutant's conidiation level.
Altogether, our transcriptomic data revealed an important role for CFP1 in genomewide gene expression related to a great loss of the mutant's insect pathogenicity via NCI and a marked reduction in its virulence via CBI. Nearly two-thirds of 154 DEGs coding for hypothetical proteins were repressed, suggesting an essential role of CFP1 in transcriptional activation of those genes functionally unknown but likely crucial to host infection and hemocoel colonization of M. robertsii.

Discussion
Based on our experimental data, CFP1, CFP2 and CFP3 were all validated virulence factors in M. robertsii in comparison to the unique CFP discovered in the time-course transcriptomes of B. bassiana infecting P. xylostella [29] and characterized as sole virulence factors among 10 SSPs studied [35]. Importantly, CFP1 was closer to CFP than CFP2 and CFP3 in both molecular size and function because the cfp1 deletion resulted in blocked NCI and greatly attenuated virulence via CBI. For this reason, the ∆cfp1 mutant was chosen for transcriptomic analysis and comparison to the previous ∆cfp-specific transcriptome. While the dataset helped us understand the mutant's main phenotypes, genome-wide gene expression was less affected by the deletion of cfp1 in M. robertsii than of cfp in B. bassiana [35]. Nonetheless, the previous and present studies reinforce that CFP and its homologs in M. robertsii serve as extraordinary virulence factors in the wide-spectrum insect pathogens of Cordycepitaceae and Clavicipitaceae in Hypocreales, respectively.
First of all, all CFP homologs are putative SSPs due to the existence of an N-terminal signal peptide consisting of the first 17 amino acids in their sequences, but do not necessarily act as secreted proteins. This was evidenced by the fact that fluorescence-tagged CFP fusion protein was localized exclusively in the nucleus under normal conditions, partially migrated to cytoplasm in response to stress cues, but was undetectable at the extracellular level [35]. The predicted NLS motifs hint that CFP1-3 could localize in the nucleus and take part in nuclear events as did CFP in the previous study. The involvement of each CFP homolog in nuclear events was clarified by the dysregulation of 604 genes in the present ∆cfp1 mutant and of 1818 genes (1006:801) in the previous ∆cfp mutant. Perhaps due to the existence of three homologs, the deletion of cfp1 in M. robertsii resulted in only one-third of those dysregulated genes seen in the ∆cfp mutant. In particular, the counts of DEGs involved in direct or indirect genome-wide gene regulation were much smaller in the absence of cfp1 than of cfp. In the ∆cfp mutant, for instance, there were 58 DEGs (34:24) encoding transcription factors involved in direct gene mediation and many more DEGs (126:49) encoding proteins/enzymes involved in RNA/DNA processing and events, translation, and posttranslational modifications [35]. The counts of dysregulated genes involved in nuclear events suggest a comprehensive effect of either CFP1 or CFP on the fungal insect-pathogenic lifecycle.
Aside from essential roles in insect pathogenicity and virulence, cfp1, cfp2 and cfp3 were dispensable for most of examined phenotypes in M. robertsii. This is largely different from the pleiotropic effects of cfp on the lifecycle in vitro and in vivo of B. bassiana, including aerial conidiation, submerged blastospore production, spore quality, and cellular responses to oxidative, osmotic, cell wall perturbing and heat stresses. Due to minor or insignificant changes in most phenotypes, counts of DEGs associated with the ∆cfp1 phenotypes were quite limited in comparison to those involved in the previous ∆cfp mutant's cell wall integrity (13:16), response to heat shock (2:6), vacuolar/cellular homeostasis and cellular transport (113:34), and signal transduction (23:9) vital for asexual development, multiple stress responses and/or virulence. The differences in both phenotypic and transcriptomic changes between ∆cfp1 and ∆cfp hint that M. robertsii and B. bassiana are distinctive in genetic background and regulatory mechanisms underlying lifecycle in vitro and in vivo. Such differences have also been revealed in the previous analysis of the master transcription factor Msn2 in the two fungi [37] and in the reviews on fungal adaptation to insect-pathogenic lifecycle and host habitats [8][9][10]38,39]. The differences could be likely due to different evolutionary histories of their insect pathogenic lifestyles [3][4][5][6][7].
Finally, it remains elusive how CFP homologs enable genome-wide gene mediation. Previously, electrophoretic mobility shift assays (EMSAs) revealed no shifting signal for the binding of purified CFP samples to any samples of promoter DNA fragments of 12 genes sharply repressed in the absence of cfp, excluding a possibility for CFP to act as a transcription factor [35]. Thus, we did not try to repeat EMSAs in this study since both CFP1 and CFP have no predictable domain as a clue to DNA-binding activity for further exploration. Like CFP, cytoplasmic Rei1 as a pre-60S subunit export factor and nuclear Ssr4 as a co-subunit of chromatin-remodeling SWI/SNF and RSC complexes were evidently involved in the fungus-insect interaction [29] and characterized as virulence factors and genome-wide gene regulators in B. bassiana, but displayed no DNA-binding activity in EMSAs [33,34]. The regulatory roles of Ssr4 and Rei1 were postulated to be associated with their involvements in the SWI/SNF or RSC complex, which can bind the binding sites of transcription factors [40,41], and the dissociation and recycling of nucleocytoplasmic pre-60S factors [42][43][44], respectively. The ∆cfp1-specific transcriptome contained dozens of DEGs involved in direct or indirect mediation of gene expression, suggesting possible linkages of those DEGs with the regulatory role of CFP1 in M. robertsii. Therefore, we still speculate that small CFP homologs existing exclusively in wide-spectrum insect pathogens might be components of certain nuclear or nucleocytoplasmic protein complexes to mediate expressions of manifold genes involved in the fungal insect pathogenic lifecycle, warranting further exploration.
Supplementary Materials: The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/jof8060606/s1, Figure S1: Generation and identification of cfp1, cfp2 and cfp3 mutants in M. robertsii; Table S1: Paired primers designed for manipulation of cfp1, cfp2 and cfp3 in M. robertsii; Table S2: Differentially expressed genes identified from the transcriptome of M. robertsii ∆cfp1 versus WT strain; Table S3: GO analysis of differentially expressed genes in M. robertsii ∆cfp1 versus WT strain; Table S4: Summation of GO function classes and terms enriched to M. robertsii ∆cfp1 versus WT strain; Table S5: KEGG pathway analysis of differential expressed genes in M. robertsii ∆cfp1 versus WT strain; Table S6: Summation of KEGG pathways enriched to M. robertsii ∆cfp1 versus WT strain. Table S7: Differentially expressed genes related to phenotypes and genome stability of M. robertsii ∆cfp1 versus WT strain.