Proteomics-Based Prediction of Candidate Effectors in the Interaction Secretome of Trichoderma harzianum and Pseudocercospora fijiensis

Abstract

Microbe–microbe interactions have been explored at the molecular level to a lesser degree than plant–pathogen interactions, primarily due to the economic impact of crop losses caused by pathogenic microorganisms. Effector proteins are well known for their role in disease development in many plant–pathogen pleinteractions, but there is increasing evidence showing their involvement in other types of interaction, including microbe–microbe interactions. Through the use of LC-MS/MS sequencing, effector candidates were identified in the in vitro interaction between a banana pathogen, Pseudocercospora fijiensis and a biological control agent, Trichoderma harzianum. The diverse interaction secretome revealed various glycoside hydrolase families, proteases and oxidoreductases. T. harzianum secreted more proteins in the microbial interaction compared to P. fijiensis, but its presence induced the secretion of more P. fijiensis proteins that were exclusive to the interaction secretome. The interaction secretome, containing 256 proteins, was screened for effector candidates using the algorithms EffHunter and WideEffHunter. Candidates with common fungal effector motifs and domains such as LysM, Cerato-platanin, NPP1 and CFEM, among others, were identified. Homologs of true effectors and virulence factors were found in the interaction secretome of T. harzianum and P. fijiensis. Further characterization revealed a potential novel effector of T. harzianum.

1. Introduction

Pseudocercospora fijiensis is the causal agent of the black Sigatoka disease, which affects species of the genus Musa, notably bananas, one of the world’s most consumed fruits [1]. The hemi-biotrophic filamentous fungus, P. fijiensis, devastates acreages of farmland in tropical and sub-tropical countries, which can result in a 35–100% loss in crop yield [2,3]. The most effective control method used is chemical control in the form of contact fungicides like Mancozeb and systemic fungicides like strobilurins and azoles (DMIs). In many tropical countries, these synthetic chemicals are alternately applied in an intensive manner, often more than 30 times per year [1,4]. The exhaustive use of these pesticides poses human health and environmental risks and also lowers the efficiency of their active ingredients due to fungicide resistance [5,6]. Alternatively, conventional breeding for resistance is challenging in Musa spp. due to the polyploid nature of cultivars, making conventional breeding costly and time-consuming [7]. Biological control is another alternative for disease management that can replace or supplement chemical control but has variable in-field efficacy. Key considerations for the use of microbial inoculants include environmental conditions, the type of formulations used and the method of application [8]. Application quantity and frequency must also be considered as the rhizosphere is a complex environment with many competitors [9]. Consequently, investigating the molecules used by biological control agents may uncover new and more effective tools for disease management compared to the use of the biological control agents themselves.

Trichoderma harzianum is a filamentous, saprophytic fungus and biocontrol agent used for the control of plant-pathogenic fungi [10,11]. More than 60% of the registered bio-fungicides on the commercial market originate from species of the Trichoderma genus [12,13], with T. harzianum being among the most popular biocontrol species. Fungal pathogens responsible for plant diseases of economic importance such as Fusarium spp. [14], Rhizoctonia solani [15] and Botrytis cinerea [16], among others, can be controlled by Trichoderma spp. due to its ability to compete for nutrients, release toxic metabolites, secrete lytic enzymes and induce systemic resistance in plant hosts [17]. Regarding the banana black Sigatoka pathogen, P. fijiensis, T. harzianum has shown promising results in vitro, in part because it grows faster and is able to better compete for space and nutrients [18]. It is likely that other unknown molecules play a role in the antagonism of P. fijiensis. T. harzianum can limit fungal growth of P. fijiensis by more than 60% in vitro [19], lending support to the investigation of T. harzianum and the molecules associated with its biocontrol capacity.

In the search for better alternatives for the control of this pathogen, the investigation of molecular targets called effectors is a viable field of study. Effectors are defined as small molecules that manipulate the structure and function of the host cell, establishing infection (virulence factors) and/or triggering a defense response (avirulence factors) in the host [20]. Thus, effectors play a vital role in plant defense where they can induce susceptibility (effector-triggered susceptibility, ETS) to an infection or induce resistance through complex and coordinated responses known as effector-triggered immunity (ETI). The induction of ETI can lead to a type of programmed cell death at the site of infection to restrict the pathogen’s advancement in the host, termed the hypersensitive response (HR); this phenomenon in the host is also triggered by the recognition of conserved pathogen or microorganism-associated molecular patterns (PAMPs or MAMPs) and the resulting immune response is termed PAMP-triggered immunity (PTI).

A pathogen’s effectorome can be considered its arsenal of biological weapons, largely involved in the suppression of the host’s immune responses [21,22]. More recently, effectors have been defined as molecules of diverse functionality and molecular nature that influence organisms’ interactions with each other, usually to the benefit of the producer organism [23]. The effector narrative has expanded from pathogens to include mutualists [24] and biological control agents [25,26] but the majority of what we know about effectors today comes from the literature on plant–pathogen interactions. An interesting subset of effectors in plant–pathogen interactions also function in microbe–microbe interactions. In the Zymoseptoria tritici–Triticum aestivum interaction, the effector Zt6 is a RNase with phytotoxic activity in the host plant, but also presents activity against the microorganisms E. coli, S. cerevisiae and P. pastoris [27]. In Verticillium dahliae, the effector VdAve1 plays a role in both plant host and microbial interactions; the effector was found to inhibit the in vitro growth of certain Gram-positive bacteria and modify bacterial communities in planta [28]. Regarding Trichoderma spp., the T. atroviride effector Tal6 binds to cell-wall chitin and prevents the recognition of its chitin oligomers by the plant host [29]. Tal6 is expressed during microbial interactions and is involved in Trichoderma’s mycoparasitism, most likely protecting the cell wall against microbial chitinases. Epl1 is a cerato-platanin found in various Trichoderma species that induces the expression of plant defense-related genes [30,31,32]; in a microbial interaction between T. harzianum and B. cinerea, Epl1 was found to downregulate the expression of certain B. cinerea virulence-associated genes [33] and also affects hyphal coiling during the mycoparasitism of R. solani and Sclerotinia sclerotiorum [31].

Most effectors are secreted to the extracellular space since they are generally proteins that execute their functions in the host–whether plant, animal or microorganismple. The sum of all the proteins secreted to the extracellular space by a cell, tissue, organ or organism is referred to as the secretome [34], though the secretome can encompass other secreted molecules such as lipids and free nucleic acids. The protein-based secretome is a valuable reservoir containing many virulence-related proteins, including effector proteins, that can be mined with bioinformatics programs. In the predicted secretome of T. harzianum, 108 effector candidates were first identified [35], followed by 113 candidates [36]. With the improvement in the functionality of fungal effector algorithms, the number of possible effector candidates for T. harzianum has greatly expanded; 307 candidates were identified by EffHunter [37], and a considerable 2693 effector candidates in the complete effectorome were predicted by WideEffHunter [38]. Currently, the functions of only three P. fijiensis effectors (Avr4, Ecp6 and Ecp2) in the plant host are known [39].

Secretome studies have been reported between Trichoderma spp. and plant hosts such as Arabidopsis thaliana [40], maize [41,42] and banana [43], as well as different plant substrates such as cellulose and xylans [44]. The effectors that have been characterized in Trichoderma spp–plant interactions include hydrophobins, glycoside hydrolases and cerato-platanins [45]. Most of these molecules have been described as elicitors or inducers of plant defense pathways [46,47]. Secretome studies of Trichoderma spp. in the presence of fungal cell walls [48] or inactivated mycelium [49] have revealed candidate effectors that play a role in biological control, although effector identification in Trichoderma spp., particularly in microbial interactions, is an understudied theme of investigation. Tal6 and Epl1 are some of the few effectors whose characterization has begun in Trichoderma microbial interactions.

The first in silico-predicted secretomes of P. fijiensis were reported to have between 105 and 172 effector candidates [5,50]. Regarding experimental secretome studies, Escobar-Tovar et al. (2015), using SDS-PAGE and 2D-PAGE, observed 436 spots for a virulent strain of P. fijiensis growing in vitro, while they observed 101 spots for the virulent strain in interactions with the host plant [51]. The protein spots were sequenced by LC-MS and, in total, 90 proteins including virulence-associated proteins were identified. Chuc-Uc et al. 2011 using SDS-PAGE and PAGE characterized the P. fijiensis in vitro secretome which causes necrosis in resistant and susceptible banana cultivars, as well as the non-host Carica papaya [52]; a 30 kDa thermostable protease was also identified as a possible effector in the P. fijiensis–Musa interaction. Burgos-Canul et al. 2019 [53] identified the cell wall proteins of two isolates of P. fijiensis using nano-HPLC–MS/MS. Of the 2686 proteins obtained, 24 shared pathogenicity factors were identified, including Avr4 and Ecp2.

Effector identification can be challenging since effector proteins possess different structural characteristics. The most common effector characteristics that have been identified include their small molecular weight, rich cysteine content, the presence of a secretory signal peptide and the absence of transmembrane domains [54]. Many bioinformatic tools are available for the in silico prediction of effector candidates such as Signal P [55], TMHMM [56] and LOCALIZER [57], which identify individual effector structural characteristics, while others combine various individual programs [37], use in-house pipelines [58] or use machine-learning programs [59] for effector identification. WideEffHunter [38] is one of the more recent effector algorithms capable of identifying effectors with the commonly observed structural characteristics (canonical effectors) as well as those that defy one or more of these characteristics; these effectors termed “non-canonical” were previously elusive to other effector identification pipelines that are based on canonical effector characteristics. The identification of effectors is a necessary step for their experimental characterization; the advent of improved effector identification algorithms facilitates high-throughput effector prediction from omics data. Here, we used the WideEffHunter and EffHunter algorithms to analyze the in vitro interaction secretome of T. harzianum and P. fijiensis, following the isolation and identification of interaction proteins by LC-MS/MS. The goal of this study was to identify, for the first time, effectors in the in vitro interaction secretome of T. harzianum and P. fijiensis.

2. Materials and Methods

2.1. Subsection Fungal Strains and Cultivation

Fungal isolates P. fijiensis C1233 and T. harzianum B13-1 were obtained from Musa acuminata cv. Grand Nain leaves from a plantation in Yucatan, Mexico [53], and decomposing banana from a plantation in Tabasco, Mexico [60], respectively. Fungi were cultured on Potato Dextrose Agar (PDA) (DIBICO, Cuautitlan Izcalli, Mexico) and kept at room temperature.

2.2. Fungal Interaction Assay

For the fungal interaction assay, the two fungi were individually inoculated (1 mL mycelial suspension) in 50 mL Potato Dextrose Broth (PDB) (DIBICO, Cuautitlan Izcalli, Mexico). Due to the differing growth rates of the two fungi, the period of cultivation was ten and five days for P. fijiensis and T. harzianum, respectively. During this period, fungi were maintained in shake culture at 25 °C and 100 rpm (LAB-LINE Instruments, Melrose Park, IL, USA). Fungal biomass was obtained by centrifugation (Centrifuge 5810R; Eppendorf, Hamburg, Germany) at 1699× g for 15 min, washed twice with sterile distilled water, and then transferred to 50 mL minimal media (yeast nitrogen base without amino acids and ammonium sulfate) (SigmaAldrich, St. Louis, MO, USA) supplemented with 10 g/L dextrose (Sigma). Conical flasks of minimal media were inoculated with 3 g of each fungus for the interaction, and 6 g of each fungus was used for the control flasks (monocultures of each fungus).

2.3. Isolation of the Interaction Secretome

Culture media on the 5th day of the interaction was collected in 50 mL Falcon tubes and lyophilized. A modified TCA-acetone protocol [61] was used for the precipitation of the proteins present in the interaction and monoculture samples. Briefly, lyophilized material was resuspended in Tris-HCl (Sigma-Aldrich, St. Louis, MO, USA) (pH 7.0) with 0.1 mM phenylmethylsulfonyl fluoride (PMSF, Sigma-Aldrich, USA) and protease inhibitor cocktail (Roche Complete™, EDTA-free Protease Inhibitor Cocktail). Cold 10% w/v TCA (Sigma-Aldrich, USA) in acetone (Sigma-Aldrich, St. Louis, MO, USA) was added per volume of sample and incubated overnight at −20 °C. After incubation, samples were centrifuged at 17,000× g at 4 °C for 45 min. The supernatant was discarded and protein precipitates (pellets) were washed three times with cold 90% acetone. Pellets were left to dry at room temperature for an hour and subsequently resuspended in 500 µL of Tris-HCl (pH 7.0). Proteins were quantified using the Linearized Bradford method [62] measuring the absorbance of the samples at 595 nm and 450 nm. Measurements of absorbance were interpolated on a standard curve based on bovine serum albumin (BSA, Sigma-Aldrich, USA) to calculate the concentration of protein in the samples. Samples were taken in triplicate and pooled for each condition: (a) T. harzianum monoculture, (b) P. fijiensis monoculture, and (c) the interaction. Protein pools were sent for LC/MS-MS analysis (Functional Genomics Center Zurich, Winterthurerstrasse, Zurich, Switzerland).

2.4. LC-MS/MS Analysis of Peptides

Sample processing was performed in duplicate for each sample (T. harzianum monoculture, P. fijiensis monoculture and interaction) according to the manufacturer’s protocol at the Functional Genomics Center Zurich. Briefly, proteins (100 mL) were precipitated with 100 mL of 20% TCA overnight and centrifuged at 17,000× g, and pellets were washed twice with cold acetone and dried at room temperature for 30 min. Pellets were dissolved in 45 mL 10 mM Tris/2 mM CaCl₂ (pH 8.2) and then microwaved for 30 min. Samples were then reduced with 10 mM dithiothreitol (DTT) at 25 °C for 1 h, and protein alkylation was carried out by adding 15 mM iodoacetamide and incubating for 1 h at 25° C in the dark. Protein digestion was conducted using 5 μL trypsin (100 ng/mL in 10 mM HCl) and incubating overnight at 60 °C in the dark. Samples were desalted on C18 ZipTips, eluted with 150 mL 0.1% TFA and 35% acetonitrile, dried, and then dissolved in 20 μL 0.1% formic acid (1 mL was checked for the presence of peptides by MaldiTOF). Positive samples were transferred to autosampler vials for LC/MS/MS; 2 μL of each peptide solution was analyzed using an LTQ-Orbitrap VELOS mass spectrometer (Thermo Fischer Scientific, Bremen, Germany) coupled to an Eksigent-NanoLC-Ultra 1D plus HPLC system (Eksigent Technologies, Dublin, CA, USA). The solvent composition at the two channels was as follows: 0.2% formic acid, 1% acetonitrile for channel A, and 0.2% formic acid, 100% acetonitrile for channel B. Peptides were loaded on a column (75 μm × 80 mm) packed with reverse phase C18 material (AQ, 3 μm 200 Å, Bischoff GmbH, Leonberg, Germany) and eluted with a flow rate of 250 nL/min with an increasing gradient of 0.8% to 4.8% of B in 2 min, 35% at 57 min, 48% at 60 min and 97% at 65 min. The effluent from the HPLC was directly electrosprayed into the mass spectrometer by applying 2.0 kV through a platinum-based liquid junction. Full-scan MS spectra (300–2000 m/z) were acquired in the Orbitrap with a resolution of 30,000 at 400 m/z after accumulation to a target value of 1,000,000. Database searches with the raw data were performed using the Mascot software program version 2.4.1 [63] (SwissProt, all species; Trembl, fungi). Retrieved results were summarized as protein IDs in Scaffold files. The scaffold settings applied were as follows: 1% protein FDR, presence of a minimum of 2 peptides per protein, and 0.1% for peptide FDR. Fixed post-translational modifications: carbamidomethylation on cysteines; variable modifications: acetylation at protein N-terminal and oxidation on methionine.

2.5. Bioinformatic Analysis

Scaffold 4.1 was used to view the sequences identified using the UniProt TREMBl fungal database [64].

For functional annotation of the secreted proteins, sequences were annotated with GO terms using the program OmicsBox version 3.4 (Blast2GO, version 6.0), https://www.blast2go.com/ accessed on 30 April 2025 [65] and annotations were graphed using the online platform WEGO, https://wego.genomics.cn/, accessed on 30 April 2025 [66] for the comparison of control and interaction subset gene enrichment. The sequenced proteins were annotated with information from the InterProScan database version 105.0 (https://www.ebi.ac.uk/interpro/) accessed on 30 April 2025 in the Standalone mode [67].

Fasta files with the protein sequences were used as input for the effector identification algorithms.

EffHunter [37] and WideEffHunter [38] were used for the identification of canonical and non-canonical effectors, respectively. For EffHunter, the user-established parameters were amino acid length ≤400 amino acids and ≥4 cysteines (Perl scripts, Linux). Canonical proteins were identified that had 400 amino acids or less with at least 4 cysteines, zero transmembrane domains identified by TMHMM [56] and a signal peptide identified by SignalP 5.0 [55]. The output files contained candidate effector proteins in Fasta format. Subsequently, the original secretome datasets were analyzed with WideEffhunter. The program identified candidates based on the presence of effector motifs ([Li]xAR, PDI, CRN, LysM, EAR, ToxA, RXLR and YFWxC) using associated Regex in its database and homology to the 314 known effectors in its positive database [38]. For the in silico characterization of the proteins, the ProtParam tool, standalone version, which was accessed during 30 January 2025, https://web.expasy.org/protparam/ [68], was used to determine the characteristics of the protein sequences (length and amino acid composition), while the presence of signal peptides was determined by SignalP 5.0 [55], transmembrane domains by TMHMM [56] and subcellular localization using Localizer [57]. Additionally, all candidates were screened for functional domains using the Conserved Domains and Protein Classification in the NCBI CD-Search Tool version v3.21 (https://www.ncbi.nlm.nih.gov/Structure/bwrpsb/bwrpsb.cgi, accessed on 30 January 2025) [69], executing the analysis with default values (threshold value of 0.1 and 500 as the maximum number of hits).

To determine if the identified candidates were unique to the T. harzianum–P. fijiensis interaction and to validate their effector potential, we compared these candidates with those found in previous P. fijiensis–banana and T. harzianum microbial interactions. For P. fijiensis, proteins were compared to those previously reported in the P. fijiensis in planta secretome [51] and the P. fijiensis–Musa interaction transcriptome [70].

T. harzianum candidates identified in this study were also searched for in microbial interactions between Trichoderma spp. and S. sclerotiorum [71], Fusarium oxysporum [48], Guignardia citricarpa [49], R. solani [25], Alternaria alternata and Fusarium graminearum [72], as well as Trichoderma spp.–plant interactions [41,42,73].

Concurrently, all effector candidates identified in the interaction secretome were analyzed against the Pathogen–Host Interactions database (PHI-base v4.17) (http://www.phi-base.org/), accessed on 30 April 2025. PHI-base is a compendium of characterized virulence and pathogenicity factors which classifies the phenotypes of the pathogen’s gene mutants in interactions with plants (reduced virulence, hypervirulence, unaffected pathogenicity and effector plant avirulence determinant) [74].

2.6. Sequence Alignment and Phylogenetic Analysis

Homologs of ThCFEM1 (A0A0F9X6Z0) were searched for using BLASTP in the non-redundant database at GenBank with default settings. Eight hits were chosen to execute an alignment. Subsequently, a multiple alignment was performed using CLUSTALW integrated in the MEGA11 software program, version 11.0.13. The alignment was visualized and analyzed for amino acid residue conservation in the Jalview software version 2.11.4.1. Conserved amino acids were highlighted based on percent sequence identity.

For the phylogenetic analysis of ThCFEM1, an initial BLAST 1.4.0 using default settings identified 100 hits for the class Sordariomycetes. Subsequently, a different approach was taken to identify existing homologs in related and distantly related fungi to allow for a more extensive phylogenetic analysis. A BLAST was performed in different fungal taxonomic families using default settings. All Trichoderma spp. hits were retrieved followed by the top one to five resulting hits in each family analyzed. In total, 42 sequences were used for tree construction; the sequences were aligned using MUSCLE in the MEGA 11 software program. MEGA 11 was used for the construction of the phylogenetic tree using the maximum likelihood method with bootstrap resampling (1000).

2.7. Gene Cloning

The effector candidate (ThCFEM1) without its native signal peptide was cloned using the Gateway^® cloning system (Thermo Fisher Scientific, San Francisco, CA, USA). Using attB-adapted primers, the effector candidate was amplified from cDNA derived from the T-harzianum–P. fijiensis interaction using Phusion Green Hot Start II High Fidelity DNA Polymerase (Thermo Scientific). PCR products were purified using High Pure PCR Product Purification Kit (Roche) and recombined into pDONR™ 221 vector using the BP Clonase II enzyme mix (Thermo Fisher Scientific) in a 10 μL reaction following the manufacturer’s protocol. The product of the BP reaction was transformed into chemically competent E. coli DH5α cells. Selected entry clones grown on LB-kanamycin plates were sequenced using M13 primers to confirm the identity and correct reading frame orientation of effector candidates.

The effector candidate gene cassette flanked by the M13 primer binding sites was amplified from the entry vector for subsequent recombination. The PCR product was purified and then recombined with the destination vector, pEAQ-EGgw [75], using LR Clonase II enzyme mix (Thermo Fisher Scientific) in a 10 μL reaction following the manufacturer’s protocol. Expression clones were produced by transforming electrocompetent E. coli DH5α cells (Bio-Rad Gene Pulser Xcell electroporator, Hercules, CA, USA) with the LR reaction product.

The preliminary confirmation of transformed bacteria was performed by colony PCR using gene specific primers to confirm the presence of the transgene in the entry and expression clones. The insert was sequenced (Macrogen, Inc., Seoul, Republic of Korea) and the resulting data was aligned with the gene of interest to confirm that it was positioned in the correct open reading frame. Plasmid DNA was isolated from the selected positive colonies using the QIAprep^® Spin Miniprep Gene Kit (Qiagen, Germantown, MD, USA) following the manufacturer’s protocol. The final expression construct was purified and dialyzed.

2.8. Transformation of Agrobacterium tumefaciens

For each transformation, 1 μL of dialyzed plasmid DNA (50 ng/μL) was used to transform 40 μL of A. tumefaciens GV3101. A. tumefaciens was transformed using a Bio-Rad Gene Pulser Xcell electroporator following the manufacturer’s default protocol for A. tumefaciens (Voltage 2.4 kV, Capacitance 25 mF, Resistance 200 W). Transformed bacteria were plated on LB agar supplemented with kanamycin (50 mg/L) and incubated at 27 °C for 24–48 h. Positive clones carrying the effector gene of interest were confirmed by colony PCR using gene-specific primers.

2.9. Agroinfiltration Assay in N. benthamiana

Individual colonies of transformed A. tumefaciens were cultivated in 4 mL LB with 50 mg/L kanamycin overnight under shaking incubation (200 rpm) at 27 °C. Five hundred microliters (0.5 mL) of primary inoculum were used to initiate a 20 mL culture of LB with kanamycin (50 mg/L), 10 mM MES (pH 5.6) and 200 μM acetosyringone for 18 h (200 rpm, 27 °C). Eighteen-hour cultures were then centrifuged (2800× g, 20 °C, 10 min) and pellets were resuspended in freshly made infiltration medium (5 g/L MS salts, 20 g/L sucrose, 10 mM MES, 200 μM acetosyringone) to an OD600 of 0.4. Cultures were then incubated in the dark at room temperature for 2–3 h before agroinfiltration.

For the agroinfiltration assay, recombinant bacterial suspensions were loaded into 3 mL needleless syringes and infiltrated into the abaxial surface of young leaves of 5-to-6-week-old plants. Plants were monitored for seven days post-agroinfiltration to evaluate HR development. Four plants were used per infiltration assay, and the experiment was executed three times.

2.10. Cell Death Suppression Assay

Recombinant A. tumefaciens was cultivated as previously mentioned for the agroinfiltration assay, except that bacteria were resuspended to a final OD600 of 0.3 in the infiltration buffer (5 g/L MS salts, 20 g/L sucrose, 10 mM MES, 200 μM acetosyringone). The abaxial sides of leaves were infiltrated with A. tumefaciens expressing the effector candidate or the GFP control, then 24 h later with A. tumefaciens expressing the Phytophthora sojae cell-death inducer, XEG1 [76] was infiltrated at the same site. Lesion development was assessed for 5 days post-infiltration.

3. Results

3.1. Expansions and Contractions in the T. harzianum–P. fijiensis Interaction Secretome

The interpretation of the MS spectra by the Mascot Program revealed more than 300 proteins that belonged to the three experimental conditions: the T. harzianum–P. fijiensis interaction and each fungus grown separately in minimal media as the controls (Figure 1).

Two hundred and fifty-six proteins (256) were identified in the interaction between P. fijiensis and T. harzianum, of which sixty-two (62) proteins were found to be exclusive to the interaction (Figure 1) as they were absent in the T. harzianum or P. fijiensis control secretomes. A larger quantity of P. fijiensis proteins were identified in the total interaction secretome; 84 proteins were found in its control and 123 P. fijiensis proteins in the interaction. Meanwhile, for T. harzianum there was a decrease in proteins from the control secretome (202 proteins) to that of the interaction (133 proteins) and most sequences (122 sequences) identified from T. harzianum in the interaction were also found in its large control secretome. Protein sequences for each organism were analyzed for GO annotations with Blast2GO to ascertain which proteins were important to the interaction and their possible functions. Figure 2 shows a comparison of the sequence distribution for the most enriched GO terms between the two fungi in the control and interaction conditions generated using the web-based tool WEGO.

The contraction observed in the T. harzianum interaction secretome was in part attributed to a reduction in proteins related to the GO terms “binding” and “catalytic activity” (Figure 2). Within the “catalytic activity” category, there was a reduction in proteins associated with hydrolase activity in the interaction subset (84) compared to the control (129); however, this difference was not statistically significant. The term “binding” was also more enriched in the control subset compared to the interaction (50 in the control vs. 24 in the interaction) (Figure 2). Moreover, there was a pronounced difference in the number of proteins associated with the term “metabolic process,” favoring the control subset—particularly for “primary metabolic process” (94 vs. 63) and “organic substance metabolic process” (97 vs. 64). In total, there were 101 proteins in the control subset and 68 in the interaction subset associated with “metabolic process”. However, because Figure 2 shows percentages rather than raw values, this difference is not readily apparent due to the larger overall size of the T. harzianum control subset compared to its interaction subset. With respect to the P. fijiensis sequence distribution, there was a notable increase in the number of proteins in the interaction secretome related to “metabolic process” (50 in the control vs. 80 in the interaction) and “Hydrolase activity” (51 in the control vs. 83 in the interaction) which contributed to the expansion of its interaction secretome. Conversely, terms such as “Transporter Activity”, “Localization”, “Small molecule metabolic process” and “Binding” were more enriched in the control secretome. For both T. harzianum and P. fijiensis, enriched Cellular Component terms included “membrane”, “cell” and “cell part”. The percentage of proteins with the annotation “extracellular region” was similar in both conditions for P. fijiensis but higher for the proteins in the T. harzianum control secretome (Figure 2).

3.2. The Diverse Nature of the Interaction Secretome

In order to identify the most abundant protein families secreted by each fungus during the interaction, each protein was annotated using InterProScan. The top 10 most abundant InterPro families for each fungus in the interaction secretome are presented in Figure 3. The glycoside hydrolase (GH) superfamily was the most prominent term observed for both fungi with 17 proteins for T. harzianum and 31 for P. fijiensis. For T. harzianum, GH family 16 (five sequences) and GH families 71 and 43 (two sequences each) were the most abundant among the GH families. GH family proteins were more abundant in the P. fijiensis interaction subset with GH families 3 (eight sequences) and 17 (four sequences), GH 43 (four sequences) and GH 92 (three sequences) being the most numerous GH proteins in the family. After the glycoside hydrolase superfamily, the alpha/beta hydrolase family was the second most abundant family for P. fijiensis (13 sequences) followed by the six-hairpin glycosidase superfamily (11 sequences). With respect to T. harzianum, after the GH superfamily, the most abundant families were aspartic peptidase A1 (11 sequences), glucanosyltransferases (8 sequences) and pectin lyases (7 sequences). With respect to proteases, interesting differences in protease families were observed between the two fungi. Proteases from the aspartic peptidase A1 family were most abundant in T. harzianum (11 sequences) followed by the subtilisin-like serine proteases with 3 sequences. Conversely, the peptidase S8/S53 family serine peptidases were most abundant in P. fijiensis (five sequences) followed by S10 serine peptidases (four sequences).

A closer analysis of the proteins unique to the interaction then followed for the 11 and 58 proteins found for T. harzianum and P. fijiensis, respectively. Regarding the InterPro domain distribution for P. fijiensis, the highest number of proteins was found in the Fibronectin type III-like classification (six sequences) and the glycoside hydrolase family 3 (six sequences), followed by peptidase S53 (three), sedolisin domain (three), peptidase S8/S53 (two) and glycoside hydrolase family 18 (two) (Figure 4). Finally, thirty-seven (37) other domains were identified among the P. fijiensis interaction-specific subset, each with one protein. With respect to T. harzianum, the InterPro domain annotations for its small subset of interaction-specific proteins were carbohydrate binding, WSC (four sequences), glycoside hydrolase families 81 (one sequence) and 16 (one sequence), haem peroxidase (one sequence), peptidase A1 (one), peptidase S9 (one), pectate lyase (one) and a carboxylesterase type B domain (one) (Figure 4).

3.3. The Interaction Secretome Contains Candidate Effector Proteins

Effector identification in the interaction secretome consisted of screening all proteins with the EffHunter and WideEffHunter algorithms. EffHunter was used for the identification of canonical effector candidates, while WideEffHunter focused primarily on non-canonical candidates. In total, 46 and 54 effector candidates were identified for T. harzianum and P. fijiensis, respectively. For P. fijiensis, EffHunter identified 17 canonical candidates and WideEffHunter identified 37 non-canonical candidates. For T. harzianum, 14 canonical candidates and 32 non-canonical candidates were identified by EffHunter and WideEffhunter, respectively. Of the P. fijiensis candidates, 21 were specific to the interaction secretome, while 33 were found in both interaction and control conditions. Among the T. harzianum proteins, 45 candidates were found in both interaction and control conditions while one candidate was found in the interaction-specific subset.

To obtain a better understanding of the effectors that each fungus secreted, the proteins were analyzed to determine sequence commonalities and differences, as well as the presence of effector-associated motifs and domains. All effector characterization data can be found in Supplementary Table S1. The in silico characterization of the effector candidates revealed that their protein lengths ranged from 136 to 1011 for P. fijiensis and 118 to 1117 for T. harzianum, with almost 60% of candidates surpassing 400 amino acids in both fungi. Most candidates (45 for P. fijiensis and 37 for T. harzianum) presented with effector motifs and domains as this is a key criterion used for effector identification with WideEffHunter. The most frequently occurring motifs for P. fijiensis in the interaction were LysM (14 candidates), [LI]xAR (10 candidates), and EAR (9 candidates), while LysM (11 candidates), RXLR (9 candidates) and EAR (6 candidates) were the most observed for T. harzianum. Regarding the abundance of domains, 48 (89%) P. fijiensis candidates and 40 (87%) T. harzianum candidates had identifiable domains with the NCBI CDD tool. Among the effector candidates with annotated domains, Bglx (Beta-glucosidase), Abhydrolase superfamily, Peptidase S53, PnbA and Glyco_hydro superfamily domains were among the most frequently occurring in P. fijiensis, while a wider variety of domains was found among individual T. harzianum candidates including different GH family proteins (13, 47, 61), proteases (Peptidase_M19, Peptidases_S8_5) and various DUF domain proteins (1996, 4965, 6595, 6736). The domains that were unique to the P. fijiensis candidates included Kre9_KNH, S1-P1_nuclease, Nis1 and SurE while those that were unique to the T. harzianum effector candidates included RNase_T2_euk, NPP1 and the LanC_like superfamily.

Regarding homology to known effectors in the positive dataset of WideEffHunter, both fungi had five proteins that were homologous to known effectors. Notable homologs include a protein (A0A139HLB6) which shares 43% identity with ChNis1, an effector of Colletotrichum higginsianum that suppresses PAMP-triggered immunity (PTI), another (M3A9Y6) which shares 30% identity with Bec1040, a haustorium-development related effector in Blumeria graminis, and one (A0A139H3Z9) with 52% identity to the effector Pr1 of Metarhizium anisopliae, a cuticle-degrading protease.

Using PHIB-BLAST v4.17 (http://phi-blast.phi-base.org/) with an e-value cutoff of 1 × 10⁻⁵, the interaction proteins were compared with the proteins in the PHI-base database version 4.17.

Table 1. Effector candidates in the Trichoderma harzianum–Pseudocercospora fijiensis secretome, which shares identity with known effectors and other virulence-associated proteins.

(A) Trichoderma harzianum effector candidates
Condition	Uniprot ID	Size (aa)	Domain (CDD NCBI)	PHI-Base Virulence and Avirulence Factors * Top Hit	% Identity of Effector Homologs (WideEffHunter)
Int only	A0A0F9X4L9	220	Cupredoxin domain-containing protein	MoPtep1 of Magnaporthe oryzae (32.4%)	-
Int+ Control	A0A0G0ARJ7	395	Glycoside hydrolase family 16 protein	Crh1 of Botrytis cinerea (Plant avr determinant) (46.7%)	-
Int+ Control	A0A0F9ZFB8	458	Glycoside hydrolase family 16 protein	Crh1 of Botrytis cinerea (Plant avr determinant) (45.6%)	-
Int+ Control	A0A1T3C5B5	454	Aspergillopepsin-like protease	BbepnL-1 of Beauveria bassiana (35.6%)	Acp (28.9%) of Candida tropicalis, Sap2 (28.1%), Sap3 (27.1%) and Sap1 (25.8%2) of Candida albicans
Int+ Control	A0A0F9XKY7	298	Cutinase	SsCut1 of Sclerotinia sclerotiorum (30.1%)	CutA (26.5%) of Fusarium solani f. sp. cucurbitae
Int+ Control	A0A0F9XAH1	515	Pepsin-like aspartic protease (SAP-like)	N/A	Acp (25.3%) of Candida tropicalis, Sap2 (24.3%), Sap1 (23.7%) and Sap3 (23.7%) of Candida albicans
Int+ Control	A0A0F9WZ69	530	Pepsin-like aspartic protease (SAP-like)	N/A	Sap1 (34.4%), Acp (34.1%), Sap2 (33.3%) and Sap3 (31.8%) of Candida albicans
Int+ Control	A0A0F9XAQ4	624	Pepsin-like aspartic protease (SAP-like)	SAP1 of Candida albicans (35.9%)	Sap1 (35.9%), Sap2 (35.3%) and Sap3 (34.5%) of Candida albicans and Acp (33.2%) of Candida tropicalis
Int+ Control	A0A1T3CKS4	154	Copper/zinc superoxide dismutase	SOD1 of Fusarium graminearum (87.6%)	-
Int+ Control	A0A0F9ZKW4	552	Bifunctional metallophosphatase/5′-nucleotidase, UshA Superfamily	AdsA of Staphylococcus aureus (26.2%)	-
Int+ Control	A0A0F9ZHK4	324	Transaldolase B	TalA of Francisella tularensis (52.1%)	-
Int+ Control	A0A0F9XJM3	372	Pepsin/retropepsin-like aspartic protease family protein	ACP of Candida tropicalis (26.13%)	- -
Int+ Control	A0A0F9XK41	429	Lytic polysaccharide monooxygenase (LMPO) auxiliary activity family 9 protein	PsFP1 of Phytophthora sojae (36.2%)	- -
Int+ Control	A0A0F9XJ17	595	Histidine phosphatase family protein	Aph1 of Cryptococcus neoformans (34.4%)	-
Int+ Control	A0A0F9ZZN6	541	Glyco_hydro superfamily	Gas1 of Fusarium oxysporum (60.15%)	-
Int+ Control	A0A0F9ZXC9	1117	WSC domain-containing protein; GO-like_E_set	GLX of Fusarium graminearum (44.5%)	-
Int+ Control	A0A0G0AHN3	617	DUF1996 superfamily	GLX of Fusarium graminearum (37.6%)	-
Int+ Control	A0A0F9X6Z0	193	CFEM domain-containing protein	FGSG_02077 of Fusarium graminearum (51.22%)
Int+ Control	A0A0F9XM39	138	Cerato-platanin superfamily	Sm1 of Trichoderma virens (91.3%)	-
Int+ Control	A0A0F9ZCK3	463	Alpha amylase (AmyAc_euk_AmyA domain)	MalS of Salmonella enterica (26.82%)	-
(B) Pseudocercospora fijiensis effector candidates
Condition	Uniprot ID	Size (aa)	Domain (CDD NCBI)	PHI-Base Virulence and Avirulence Factors * Top Hit	% Identity of Effector Homologs (WideEffHunter)
Int only	A0A139HLB6	147	Nis1 family; Necrosis-inducing secreted protein 1	-	ChNIS1 (43.3%) of Colletotrichum gloeosporioides
Int+ Control	M3AIE7	518	Pepsin/retropepsin-like aspartic protease family protein	N/A	Sap1 (29.7%), Sap3 (28.2%) and Sap2 (28.1%) of Candida albicans and Acp (26.4%) of Candida tropicalis
Int+ Control	A0A139H3Z9	408	S9 Family peptidase	N/A	Pr1 (52.3%) of Metarhizium anisopliae
Int only	M2ZK33	423	Alpha/beta hydrolase family protein	FGSG_03846 of Fusarium graminearum (56.7%)	-
Int+ Control	M3A2P3	354	Aspergillopepsin-like	BbepnL-1 of Beauveria bassiana (36.4%)	Acp (27%) of Candida tropicalis and Sap1 (28.2%), Sap2 (26.6%) and Sap3 (25.4%) of Candida albicans
Int+ Control	M2ZEM9	590	GMC family oxidoreductase	GOX of Aspergillus carbonarius (30.28%)	-
Int+ Control	M3AVS4	608	GMC family oxidoreductase	BCIN_03g01540 of Botrytis cinerea (48.19%)	-
Int+ Control	M3A5F8	581	GMC family oxidoreductase	Bab2_0277 of Brucella abortus (25.5%)	-
Int only	M3A9Y6	196	-	BEC1040 of Blumeria graminis (30.5%) CEP2 avr effector of Magnaporthe oryzae (32.9%)	Bec1040 (30.5%) of Blumeria graminis
Int+ Control	N1QC30	308	Glycoside hydrolase family 17 protein (Scw11 domain)	BGL2 of Candida albicans (40.6%)	-
Int+ Control	M2ZG05	548	Carboxylesterase/lipase family protein (PnbA domain)	FGSG_03243 of Fusarium graminearum (39%)	-
Int+ Control	M3B3X0	573	Carboxylesterase/lipase family protein (PnbA domain)	FGSG_03243 of Fusarium graminearum (42.05%)	-
Int only	M3ASU3	434	Peroxidase	MoHPX1 of Magnaporthe oryzae (34.51%)	-
Int only	A0A139IQP5	545	Peptidase S10 superfamily protein	Rs-scp-1 of Radopholus similis (32.2%)	-
Int+ Control	M3AEN1	328	Alpha/beta hydrolase family esterase (LpqC domain)	FAED1 of Valsa mali (34.5%)	-
Int+ Control	M3BB20	1011	Glycosyl hydrolase family 31	BcBGL5 of Botrytis cinerea (57.7%)	-
Int+ Control	M3AMA8	461	Glycoside hydrolase family protein	GEL2 of Aspergillus fumigatus (54.7%)	-
Int only	M3AE52	1002	Glycosyl hydrolase family 31	BcBGL5 of Botrytis cinerea (35.3%)	-
Int only	M2YTH3	569	Beta-N-acetylhexosaminidase	HEX1 of Candida albicans (30%)	-
Int only	M3B352	820	Beta-glucosidase (Bglx)	BcBGL1 of Botrytis cinerea (57.6%)	-
Int+ Control	N1QB78	878	Beta-glucosidase (Bglx)	BcBGL3 of Botrytis cinerea (62.3%)	-
Int only	M2ZWF9	761	Beta-glucosidase (Bglx)	BcBGL1 of Botrytis cinerea (44.8%)	-
Int+ Control	M2ZLK3	553	GMC family oxidoreductase (BetA)	Bab2_0277 of Brucella abortus (23.9%)	-

PHI-base virulence and avirulence factors *: Only top hits with a “reduced virulence” mutant phenotype were considered for virulence factors. N/A: Top hit is neither a virulence nor avirulence factor. Int: Interaction.

presents the top hits from the PHIB-BLAST. Notably, three T. harzianum candidates were similar to effector hits in the PHI-base with the annotation “effector plant avirulence (avr) determinant”: A0A0F9X4L9 shares 32.4% identity with effector MoPtep1 of Magnaporthe oryzae and candidates A0A0G0ARJ7 and A0A0F9ZFB8 share 46.6% and 45.5% identity, respectively, with the effector Crh1 of B. cinerea. One P. fijiensis candidate, M3A9Y6, shares 32.9% identity with the Avr effector Cep2 of M. oryzae. Many of the candidates that had hits in the PHI-base corresponded to proteins with the annotation “reduced virulence”; 23 and 20 candidate effectors of P. fijiensis and T. harzianum, respectively, share identity with bona fide virulence factors in the database (Table 1).

3.4. The Interaction Secretome Contains Broad-Host Range Effectors

To determine if the identified candidates were unique to the T. harzianum–P. fijiensis interaction and to validate their effector potential, we compared these candidates with those found in previous P. fijiensis–banana and T. harzianum microbial interaction reports (

Table 2. Effector candidates of Pseudocercospora fijiensis and Trichoderma harzianum involved in other interactions (plant–pathogen, microbe–microbe).

(A) Pseudocercospora fijiensis effector candidates identified in other interactions
Condition	Uniprot ID	Size (aa)	Domain (CDD NCBI)	EffHunter	WideEffHunter	Interaction	Notes
Int only	M3B352	820	BglX	no	yes	P. fijiensis–Musa acuminata [70]	Upregulated in leaf tissue
Int only	M3B2Z2	681	Peptidases_S53	no	yes	P. fijiensis–Musa acuminata [70]	Upregulated in leaf tissue
Control + Int	M3A4I0	175	-	no	yes	P. fijiensis–M. acuminata [70]	Upregulated in leaf tissue
Control + Int	M3BB20	1011	Glyco_hydro_31	no	yes	P. fijiensis–M. acuminata [70]	Down regulated in leaf tissue vs. mycelium
Int only	M2ZY96	195	-	yes	no	P. fijiensis–M. acuminata [51]	Secreted in vitro and in planta by both avirulent and virulent P. fijiensis isolates
(B) Trichoderma harzianum effector candidates identified in other interactions
Condition	Uniprot ID	Size (aa)	Domain (CDD NCBI)	EffHunter	WideEffHunter	Interaction	Notes
Control + Int	A0A0F9XHX6	429	BglC superfamily	no	yes	T. harzianum TR274–S. sclerotiorum [71]	Upregulated in presence of S. sclerotiorum cell walls vs. glucose (12, 24, 36 h)
Control + Int	A0A2N1LN60	118	-	yes	no	T. harzianum TR274–S. sclerotiorum [71]	Upregulated in presence of S. sclerotiorum cell walls vs. glucose at 24 and 36 h
Control + Int	A0A0F9WYR7	515	Glyco_hydro_47	no	yes	T. harzianum ALL42–F. solani [48]	Secreted in culture media with F. solani cell walls
Control + Int	A0A0F9X6H8	689	DUF4965	no	yes	T. harzianum ALL42–F. solani [48]	Secreted in culture media with F. solani cell walls
Control + Int	A0A0F9XM39	138	Cerato-platanin superfamily	yes	no	T. harzianum ALL42–F. solani [48] T. virens–Zea mays [41] T. harzianum–Phaseolus vulgaris [31]	Secreted in culture media with F. solani cell walls and culture media with only glucose Secreted in similar proportions in interactions with maize and when alone; most abundantly secreted protein
Control + Int	A0A0F9ZTU1	281	NPP1 superfamily	yes	no	T. harzianum ALL42–F. solani [48]	Secreted in both culture media with F. solani cell walls
Control + Int	A0A0G0A4H5	878	Peptidases_S8_5	no	yes	T. harzianum ALL42–F. solani [48]	Secreted in both culture media with F. solani cell walls
Control + Int	A0A0F9ZZN6	541	Glyco_hydro superfamily	no	yes	T. harzianum ALL42–F. solani [48]	Secreted in minimal medium with F. solani cell walls
Control + Int	A0A0F9X6Z0	197	CFEM	yes	yes	T. virens–Zea mays [41] T. harzianum–F. solani [77]	Up-accumulated in T. virens control secretome vs. T. virens–maize interaction secretome Upregulated in the presence of F. solani cell walls
Control + Int	A0A0F9XF89	776	GH55_beta13glucanase-like	no	yes	T. harzianum TR274–Phaseolus vulgaris [73]	Up-accumulated in interactions with Phaseolus vulgaris
Control + Int	A0A1T3CKS4	154	Sod_Cu	no	yes	T.virens–Zea mays [42]	Secreted by T. virens after 5 days in contact with maize roots

A). Three P. fijiensis effector candidates from the T. harzianum–P. fijiensis interaction were identified which were upregulated in planta during the late phase of infection in the banana host; a beta glucosidase (M3B352) and serine peptidase (M3B2Z2) which were secreted only in the presence of T. harzianum, and an uncharacterized protein (M3A4I0) which was found in both control and interaction secretomes. These three candidates were all non-canonical effectors. Conversely, an uncharacterized canonical effector (M2ZY96) was found in similar quantities in the control and interaction secretomes by both virulent and avirulent P. fijiensis isolates in another study but was secreted only in the presence of T. harzianum in this study (Table 2A). It is worth mentioning that this candidate shares identity with two effectors, BEC1040 (30.5%) and CEP2 (32.9%) (Table 1). The presence of these effectors in both banana and T. harzianum interactions suggests that microbial interactions can serve as a valuable tool for identifying virulence and pathogenicity factors beyond traditional pathosystems. It also highlights the importance of these effectors to the attack and/or defense strategies of P. fijiensis against different hosts.

T. harzianum candidates identified in this study were also searched for in microbial interactions between Trichoderma spp. and other fungi (Table 2B). T. harzianum effector candidates identified in this study were identified in other microbe–microbe interactions between T. harzianum and S. sclerotiorum and F. solani; two T. harzianum proteins were upregulated in the presence of S. sclerotiorum cell walls: an uncharacterized protein (A0A2N1LN60) and a beta-glucosidase (A0A0F9XHX6). Seven proteins were identified in T. harzianum interactions with F. solani cell walls; one of the proteins (A0A0F9XM39) was also identified in at least two other studies including an interaction with a plant host (Phaselous vulgaris). This is a known cerato-platanin effector of Trichoderma sp. that is constitutively expressed, though its role in plant interactions appears to be a positive one as a plant defense elicitor. Interestingly, a T. harzianum non-canonical effector candidate (A0A0F9XF89) with a beta-glucanase domain is also up-accumulated in an interaction with a plant host, Phaseolus vulgaris. Lastly, a CFEM domain-containing protein in the interaction, A0A0F9X6Z0, was also identified in a plant interaction where it was downregulated compared to the control while it was upregulated in the presence of F. solani cell walls. The effectors pinpointed in various interactions are high-priority candidates for further characterization and as such, further in silico and experimental analysis was performed for the T. harzianum candidate, A0A0F9X6Z0.

3.5. Characterization of ThCFEM1 (A0A0F9X6Z0), a Novel Broad Host-Range Effector from T. harzianum

The T. harzianum candidate A0A0F9X6Z0 was chosen for further characterization as one of the candidates with antecedents in various interactions (T. harzianum–P. fijiensis, T. harzianum–F. solani, T. virens–maize). The protein was identified in the interaction secretome and also in the control secretome of T. harzianum in monoculture; it is a predicted effector according to EffHunter, WideEffHunter as well as the machine-learning classifier, EffectorP. Further in silico characterization with WideEffHunter revealed that the candidate is an extracellular cysteine-rich canonical effector with a YFWxC effector motif (

Table 3. Characteristics of ThCFEM1.

Uniprot ID	Size (aa)	Peptide Signal	Localization	Molecular Function	Domain	PHI-Base	Effector Motif	Taxonomic Distribution
A0A0F9X6Z0	193	Yes	Extracellular, Cell membrane	Metal ion binding	CFEM	FGSG_02077 of F. graminearum (51.22%)	YFWxC	Related species in the order Hypocreales

). The protein also has a GPI anchor as predicted by PredGPI version 1.0 (http://busca.biocomp.unibo.it/predgpi/), accessed on 30 April 2025. A CFEM domain was identified in the sequence by InterProScan and NCBI CDD (amino acids 20–84) and the candidate was thereafter named ThCFEM1. ThCFEM1 also shares 51% identity with a virulence factor reported in the PHI-base, FGSG_02077. To determine its taxonomic distribution, a BLAST search (e value 1 × 10⁻⁵) revealed that the ThCFEM1 sequence is highly conserved among related species of the order Hypocreales and the class Sordariomycetes (Figure 5). This was also observed in the phylogenetic tree constructed for the homologs of ThCFEM1 where the largest observed clade consisted of Sordariomycetes (Figure 6). Among the Sordariomycetes, homologs were predominantly found in different Trichoderma sp. (mycoparasites and saprophytes) and Fusarium sp. (plant pathogens) and to a lesser extent in Metarrhizum sp. (entomopathogens), Ilyonectria sp. (root endophytes and opportunistic fungi) and the nematophagous fungi Pochonia chlamydosporia and Purpureocillium sp. A sequence alignment was performed with ThCFEM1 and seven related species: three from the genus Trichoderma and four others of the same taxonomic class. The alignment in Figure 5 shows a high degree of conservation in the amino acid region 1 to 96 among the sequences. This region encompasses the CFEM domain which can be found between positions 20 and 90 in the amino acid sequences for most of the homologs.

To determine the role of ThCFEM1, we characterized this effector candidate in the non-host model, N. benthamiana. The transient expression of ThCFEM1 through agroinfiltration of N. benthamiana did not exhibit typical HR-related cell death; rather, a slightly chlorotic phenotype was observed when individual infiltration was performed (Figure 7B). The cell death suppression assay followed, to determine whether ThCFEM1 was a suppressor instead of an inducer of host immunity. The agroinfiltration of ThCFEM1 at the same position 24 h before the cell-death inducer XEG1 led to smaller lesions than those induced by the agroinfiltration of GFP followed by XEG1 (Figure 7A). Taken together, these results show ThCFEM1’s role as a possible suppressor of the plant’s immune response.

4. Discussion

The secretome is a rich and diverse source of virulence-related molecular targets that can be exploited. More recently, secretomes of various microorganisms have been investigated with emphasis on the discovery of effectors and virulence-related proteins [78,79]. Comparably, we established an interaction between P. fijiensis and a biological control agent, T. harzianum, then isolated and identified the proteins in the interaction secretome. Subsequently, these proteins were screened with effector prediction algorithms. The previous literature has shown that nutrient-poor media can induce the expression of virulence-related proteins [80], and as such, minimal media (MM) was used for the T. harzianum–P. fijiensis interaction and collected after five days of interaction in a shake culture. Interaction studies of Trichoderma spp. and plant pathogens have shown maximum enzymatic activity of certain chitinases, beta-glucanases and proteases of Trichoderma spp. between days 3 to 5 of the interaction [81,82]. Consequently, the day 5 secretome was selected as an adequate representation of both T. harzianum and P. fijiensis interaction proteins and was subjected to LC-MS/MS sequencing.

The composition of the proteins in the secretomes of the two fungi revealed domains of carbohydrate-active enzymes, proteases, alpha-beta hydrolases, oxidoreductases and proteins with no known domains. Among the GO-annotated proteins in the interaction secretome, more enzymes with hydrolase activity were identified for P. fijiensis than T. harzianum. Certain families such as GH 43, GH17 and GH16 that have been previously identified in P. fijiensis interactions with the host plant Musa [51,70] were also identified in the T. harzianum–P. fijiensis microbial interaction. Interestingly, the P. fijiensis secretome expanded in response to T. harzianum, and many of its proteins were unique to the interaction secretome including GH 18, 20, 31 and 43 proteins, S10 and S53 proteases and hypothetical uncharacterized proteins. Extracellular proteases, specifically the S10 group, have been upregulated in in planta interactions [70], and along with GH proteins, can have functions in cell wall degradation and remodeling, enzyme degradation, detoxification and nutrient metabolism [83,84]. As such, these proteins usually constitute a large proportion of the fungal secretome. P. fijiensis also secreted more proteins related to metabolic processes than T. harzianum in the interaction. Such proteins are involved in the synthesis, catabolism and modification of macromolecules, oxidation-reduction processes and the regulation of metabolic processes. The contraction of the T. harzianum interaction secretome, compared to the control, was largely due to a reduction in proteins associated with metabolic processes, particularly primary and organic substance metabolism. A similar reduction in these proteins was seen in the T. harzianum interaction with the citrus pathogen, G. citricarpa [49], and the presence of many of the same hydrolytic enzymes in control and interaction conditions can be attributed to their dual roles in cell wall metabolism, as well as mycoparasitism. Within the small proportion of uniquely secreted proteins in the interaction by T. harzianum, GH 55, GH 16 and GH 81 family proteins, aspartic protease (A1 domain), carboxylesterase and WSC-domain-containing proteins were identified. It could be inferred that these proteins are particularly important to the interaction and are involved in T. harzianum mycoparasitism. This is a phenomenon which is marked by the production of hydrolytic enzymes and secondary metabolites and the creation of antagonistic environments where nutrients and space are competed for.

The most abundant protein families secreted by T. harzianum were the glycoside hydrolase superfamily, aspartic A1 domain-containing peptidase, glycosyltransferase, pectin lyase and concanavalin A-like lectin/glucanase families. Trichoderma sp. have large arsenals of proteolytic enzymes and extracellular aspartic proteases are common in Trichoderma sp. mycoparasitic interactions [85,86] where they may be involved in cell-wall degradation [87] and enzyme inactivation [88]. Other cell-wall-degrading enzymes secreted by T. harzianum included endo-glucosidases and GH 16 family proteins. Endo-1,3- beta- glucosidases degrade cellulose and beta-1,3-glucan are commonly upregulated during T. harzianum mycoparasitism where they likely attack the host’s cell wall [89,90]. Proteins associated with oxidative stress were found in both fungi but were more abundant in the T. harzianum secretome. These proteins consisted of peroxidases, thioredoxins and superoxide dismutases, which are commonly associated with scavenging reactive oxygen species [91], cell signaling and gene expression [92] and virulence [93].

Effector screening using EffHunter and WideEffHunter uncovered 100 candidates in the T. harizanum–P. fijiensis interaction. WideEffHunter was used to identify non-canonical candidate effectors, while EffHunter, which is strictly based on traditional effector features was used to identify canonical effectors in the secretome. As expected, WideEffHunter widened the scope for effector retrieval; almost 70% of the identified effectors were non-canonical for both fungi. WideEffHunter’s algorithm identifies effector motifs and domains to predict if a protein is an effector. The most common effector motifs identified were LysM, RXLR and EAR. LysM domains facilitate binding to chitin and peptidoglycan compounds and have been frequently reported in other fungal effectors associated with the suppression of chitin-triggered immunity [94,95]. EAR motifs were also common to both fungi and are associated with transcriptional repression [96]. The RXLR motif is a predominantly oomycete-related N-terminal domain associated with cytoplasmic effectors. Many RXLR effectors induce host susceptibility to infection [97,98] and the domain may serve as a proteolytic cleavage site for unconventional effector secretion [99]. The domain was recently found in various fungi by WideEffHunter. [LI]xAR was a common domain found in the P. fijiensis candidates but not as common in T. harzianum; the motif is associated with effectors of the hemibiotroph, M. oryzae, but its role is unknown.

Regarding the effectors that share identity with other effectors compiled in the positive database of WideEffHunter, five T. harzianum effectors shared identify with Acp (C. albicans acid protease), CutA (B. cinerea cutinase) and Sap1, Sap2 and Sap3 (C. albicans aspartic proteinases). Similarly, five P. fijiensis candidates shared identity with the effectors Bec1040 (B. graminis haustorium-development effector), Acp, Sap1, Sap2, Sap3, ChNis1 (C. higginsianum necrosis-inducing protein and inhibitor of PTI) and Pr1 (M. anisopliae serine protease) according to WideEffHunter.

Additionally, PHIB-BLAST was used to identify characterized orthologs of the effector candidates in other organisms. The terms “effector protein” and “proteinaceous elicitor” were searched for in the PHI-base hits under gene function or gene name. Four T. harzianum candidates shared identity with effectors in the Pathogen–Host Interaction database: the effectors were Sm1 of T. virens (92.4%), Crh1 of B. cinerea (46.6% and 45.5%) and MoPtep1 of M. oryzae (32.4%). Crh1 and Sm1 are plant defense elicitors while MoPtep1 is a virulence factor of M. oryzae that suppresses PTI. Sm1 is a proteinaceous elicitor of host resistance in Trichoderma–plant interactions with a cerato-platanin domain. Certain cerato-platanins, e.g., BcSpl1 [100] and MgSM1 [101] can induce plant cell death. Other homologs of plant cell-death inducers were also found among the interaction proteins of T. harzianum (NPP1 and Crh1). The roles of cell-death inducers in plant hosts are difficult to infer in microbe–microbe interactions, and they likely have dual functions or roles outside of their cell-death-inducing activity, for example, NPP1 proteins are also expressed during vegetative growth and influence conidial germination and production [102,103], while cerato-platanins are also involved in spore maturation, hyphal growth and cell wall remodeling [104,105]. Crh1 triggers cell death in tobacco and tomato hosts but is also an elicitor of PTI and induces host resistance against B. cinerea; the protein is a highly conserved transglycolase (GH16 family) involved in fungal cell wall biogenesis. MoPtep1 is an effector with a cupredoxin domain which localizes to the plant peroxisome and prevents PTI-related cell death [106]. The homolog of MoPtep1 in T. harzianum is a non-canonical effector and was identified among the interaction-specific proteins of T. harzianum. A P. fijiensis candidate (M3AY96) shared identity with an effector in the PHI-base, Bec1040 (30.5%) of B. graminis, which is linked to haustorium development. The same candidate also shares identity with the M. oryzae effector, MoCEP2, that causes HR-related cell death in N. benthamiana and is vital for fungal pathogenicity [107]. The identification of candidates in this interaction which share identity with Bonafide effectors suggest their possible functions in the interaction as they are yet to be characterized.

BLASTp identified the functional domains present in the effectors and the majority were hydrolytic enzymes with GH family domains (e.g., GH16, 18, 20) and protease/peptidase domains (e.g., S8, S53) as was observed in the characterization of the global secretome proteins. In the interaction, the GH effectors are likely involved in sensing/recognition of the microbial counterpart, nutrient acquisition, cell wall remodeling and protection and possible detoxification of antimicrobial compounds [108]. The dependence on these types of effectors is understandable as the extracellular space represents the frontline of the battle between T. harzianum and P. fijiensis. Protease effectors were also common among both fungi; proteolytic degradation of host proteins can result in the disruption of cell signaling, defense, and physiological processes in the host [109]. These effectors can also affect the integrity of the cytoskeleton [110] and the cell wall [48,71,77]. WideEffHunter confirmed that both fungi secreted protease type effectors similar to the known effectors Acp and Sap1, Sap3 and Sap3.

A review of interaction-based literature uncovered several T. harzianum–P. fijiensis effector candidates involved in other microbe–microbe interactions as well as microbe–plant interactions. Five P. fijiensis candidates identified in the interaction were also identified in P. fijiensis–banana interactions: a S53 peptidase, a betaglucosidase, a GH31 family protein and two candidates with unknown domains. One of the uncharacterized proteins was also upregulated in leaf tissue and WideEffHunter identified a LysM domain. The uncharacterized LysM protein identified by WideEffHunter and upregulated in banana tissue is a high priority effector candidate for further experimental validation. As expected, effector candidates of T. harzianum also play a role in other microbe–microbe interactions. Among the interaction effectors, CFEM, GH47, Bglc, DUF4965, NPP1 superfamily, Peptidase_S8_5, Glyco_hydro superfamily and cerato-platanin domain-containing proteins were identified in other Trichoderma microbial interactions. Regarding Trichoderma–plant interactions, candidates with cerato-platanin, superoxide dismutase, CFEM and GH55_beta13glucanase-like domains were found. One of the broad host range candidates containing a CFEM domain, ThCFEM1, was further characterized in N. benthamiana.

The CFEM (“Common in Fungal Extracellular Membranes”) domain is associated with extracellular membrane proteins of fungi, linked to fungal pathogenicity and virulence. Previously, CFEM effectors have been shown to both induce the hypersensitive response (HR) [111,112] and activate host resistance against pathogens [113], while others are suppressors of plant immunity and the HR, which encourage host colonization [114,115,116]. The domain has been found in various effectors with diverse roles. These proteins can be involved in nutrient (iron) acquisition [117], fungal pathogenicity/virulence [114,115], conidiogenesis [118], appressorium differentiation [117], biofilm formation [119,120], cell wall stability [121] and stress tolerance [118,122].

Many CFEM proteins of pathogens are also actively involved in subverting plant defense, contributing to virulence [123,124]. They can regulate plant immunity by suppressing host defense responses, for example, programmed cell death induced by PAMPs or effectors. ThCFEM1 was shown to interact with the non-host, N. benthamiana, in a similar manner, suppressing cell death induced by XEG1, a protein considered a PAMP and effector in Phytophthora spp. [76]. In this study, the protein was identified in both interaction and control subsets. In the presence of F. solani cell walls, the protein was upregulated [77], while it was downregulated in T. atroviride in the presence of maize roots compared to the control [41]. In interactions with plant hosts, the protein may aid in the initial suppression of host immunity to foster Trichoderma sp. colonization. In microbial interactions, ThCFEM1 could be involved in iron acquisition or sequestration [125,126] and adhesion of its hyphae to the microbial counterpart [127]. The protein may also have a protective function for the T. harzianum cell wall, maintaining cell wall integrity against oxidative and other stresses [121]. Further characterization of ThCFEM1 will shed light on its role in microbial interactions, particularly its contribution to T. harzianum mycoparasitism.

The number of effectors found coincides with the global protein landscape of the interaction secretome; the response of T. harzianum to its counterpart P. fijiensis involves many of the same effectors that it secretes on its own in minimal media. Interestingly, its presence induces P. fijiensis’ secretion of many interaction-specific proteins, including effectors, some of which have been identified in host (Musa spp.) interactions. The T. harzianum response did not involve the secretion of many new proteins in interactions with P. fijiensis compared with its vegetative growth. This is possibly indicative of a “less is more” way of life that has allowed to it to become the renown biological control agent it is today. It is likely able to dominate multiple types of prey because of the low metabolic cost associated with its mycoparasitism. On the other hand, its superior ability may be due to the efficiency of a select few of its effectors with broad-host range activity such as ThCFEM1.

Using a gel-free proteomics approach, effector candidates were uncovered in a microbial interaction where the pathogen, P. fijiensis, was challenged with a biological control agent, T. harzianum. This microbial interaction highlighted possible effector candidates involved in biological control. Further characterization of T. harzianum effectors will reveal robust biocontrol molecules for the control of P. fijiensis, the causal agent of the black Sigatoka disease. Furthermore, interactions between banana and P. fijiensis can be a challenge to establish if the right environmental conditions of high humidity and warm temperatures are not met. The added time taken to propagate plants and for them to reach a suitable infection age (at least 5–8 months) followed by disease progression (first visible symptoms at 2–3 weeks in field conditions and may take longer in vitro) must also be considered. Devising a simpler strategy based on a microbial interaction has proven effective for effector identification in a pathosystem that is relatively difficult to study.

WideEffHunter estimates that 5179 and 2693 effectors comprise the effectoromes of P. fijiensis and T. harzianum, respectively. With more extensive effectoromes being identified, the bottleneck has shifted from effector identification to effector prioritization for further characterization. Microbial interactions can provide a means to prioritize certain candidates for functional characterization in pathosystems where virulence factors have already been identified. Finally, screening for effectors in biological control agents like T. harzianum can reveal both interaction-specific and broad-spectrum effectors, valuable for innovative disease management strategies. Effectors can be employed as molecular markers to select superior biocontrol species or to engineer organisms with improved biological control activity. Alternatively, the production of recombinant proteins may prove viable in the field for the candidates displaying antimicrobial and/or elicitor activity.

5. Conclusions

The T. harzianum–P. fijiensis interaction secretome consists of 256 proteins, of which 100 were identified as candidate effectors. The identified candidates, both canonical and non-canonical, represent a broad range of molecular targets for combating the banana pathogen P. fijiensis, the causal agent of black Sigatoka disease. Those candidates present in other interactions are possible broad-host range candidates while coming investigations will determine which candidates remain specific to this T. harzianum–P. fijiensis interaction. Establishing an in vitro microbial interaction is a relatively rapid and efficient method for the identification of effector candidates for their characterization and future use in the development of disease control strategies.