At least three families of hyphosphere small secreted cysteine‐rich proteins can optimize surface properties to a moderately hydrophilic state suitable for fungal attachment

Summary The secretomes of filamentous fungi contain a diversity of small secreted cysteine‐rich proteins (SSCPs) that have a variety of properties ranging from toxicity to surface activity. Some SSCPs are recognized by other organisms as indicators of fungal presence, but their function in fungi is not fully understood. We detected a new family of fungal surface‐active SSCPs (saSSCPs), here named hyphosphere proteins (HFSs). An evolutionary analysis of the HFSs in Pezizomycotina revealed a unique pattern of eight single cysteine residues (C‐CXXXC‐C‐C‐C‐C‐C) and a long evolutionary history of multiple gene duplications and ancient interfungal lateral gene transfers, suggesting their functional significance for fungi with different lifestyles. Interestingly, recombinantly produced saSSCPs from three families (HFSs, hydrophobins and cerato‐platanins) showed convergent surface‐modulating activity on glass and on poly(ethylene‐terephthalate), transforming their surfaces to a moderately hydrophilic state, which significantly favoured subsequent hyphal attachment. The addition of purified saSSCPs to the tomato rhizosphere had mixed effects on hyphal attachment to roots, while all tested saSSCPs had an adverse effect on plant growth in vitro. We propose that the exceptionally high diversity of saSSCPs in Trichoderma and other fungi evolved to efficiently condition various surfaces in the hyphosphere to a fungal‐beneficial state.


Summary
The secretomes of filamentous fungi contain a diversity of small secreted cysteine-rich proteins (SSCPs) that have a variety of properties ranging from toxicity to surface activity. Some SSCPs are recognized by other organisms as indicators of fungal presence, but their function in fungi is not fully understood. We detected a new family of fungal surface-active SSCPs (saSSCPs), here named hyphosphere proteins (HFSs). An evolutionary analysis of the HFSs in Pezizomycotina revealed a unique pattern of eight single cysteine residues (C-CXXXC-C-C-C-C-C) and a long evolutionary history of multiple gene duplications and ancient interfungal lateral gene transfers, suggesting their functional significance for fungi with different lifestyles. Interestingly, recombinantly produced saSSCPs from three families (HFSs, hydrophobins and cerato-platanins) showed convergent surface-modulating activity on glass and on poly(ethylene-terephthalate), transforming their surfaces to a moderately hydrophilic state, which significantly favoured subsequent hyphal attachment. The addition of purified saSSCPs to the tomato rhizosphere had mixed effects on hyphal attachment to roots, while all tested saSSCPs had an adverse effect on plant growth in vitro. We propose that the exceptionally high diversity of saSSCPs in Trichoderma and other fungi evolved to efficiently condition various surfaces in the hyphosphere to a fungalbeneficial state.

Introduction
Recent studies have shown that besides enzymes and small molecules, the secretomes of filamentous and dimorphic fungi can frequently contain a diversity of small secreted cysteine-rich proteins (SSCPs) unique for fungi (Kubicek et al., 2011(Kubicek et al., , 2019Wang et al., 2019;Feldman et al., 2020;Gao et al., 2020;Huber et al., 2020). This diverse group comprises a few distinct protein families, such as hydrophobins (HFBs) and cerato-platanins (CPs), although most SSCPs remain unclassified. Individual fungal genomes can contain from a few dozens of such proteins in yeasts (Wang et al., 2019;Sun et al., 2020) up to several hundred in mushrooms (Zajc et al., 2013;Krizs an et al., 2019). The SSCPs are characterized by their small size (<300 amino acids, a.a.) (Pellegrin et al., 2015;Kubicek et al., 2019), the presence of a relatively short signal peptide (~20 a.a.) Kim et al., 2016;Feldman et al., 2020) and cysteine enrichment. The latter feature is crucial for their properties and structure but is frequently ambiguous. For example, Kubicek et al. (2019) proposed a minimum threshold of 5% cysteine residues for SSCPs, referencing an average 1.5% content in all fungal proteins. Other authors rely on the presence of an evolutionarily conserved cysteine pattern in the primary structure, which can consist of only four residues in the CPs (Pazzagli et al., 1999;Chen et al., 2013;Gaderer et al., 2014) and up to eight in the HFBs (Wösten, 2001;Aimanianda et al., 2009;Bayry et al., 2012).
Probably the most interesting feature of SSCPs is their abundant secretion in liquid medium (up to 20%-30% of the total secretome) (Pellegrin et al., 2015;Feldman et al., 2020;Gao et al., 2020) and the unusually high polymorphism of their primary structures (Gaderer et al., 2014;Krizs an et al., 2019). Some SSCPs have a long evolutionary history of gene duplications (GDs) and lateral gene transfers (LGTs), making them frequent members of the orphomes (the genome fraction comprising orthologue-less genes) of individual fungal taxa (Przylucka et al., 2017a;Feldman et al., 2019;Gao et al., 2020). Paradoxically, despite the abundance of SSCPs in fungal secretomes, functions have been proposed for only a few, while the role of the majority of them remains unknown (Aimanianda and Latge, 2010;Feldman et al., 2017;Wang et al., 2019;Gao et al., 2020).
The large cluster of SSCP investigations reflects that other organisms use them as an indicator of fungal presence, probably because they are produced in large quantities. Therefore, SSCPs are frequently studied as effectors and elicitors, i.e. from the perspective of the host plants or animals in which the SSCPs either cause toxicity or elicit an immune response (Pellegrin et al., 2015;Dagvadorj et al., 2017;Darwiche et al., 2017;Fang et al., 2019;Li et al., 2019;Wang et al., 2019;Yu et al., 2020). These studies reveal the ecological role of individual SSCPs, but they do not sufficiently explain their function given that orthologous SSCPs are also present in saprotrophic fungi (Feldman et al., 2020;Gao et al., 2020).
The exceptionally high diversity of fungal SSCPs and their ubiquitous presence in biotrophic and saprotrophic fungi suggest their functions are related to aspects of the fungal life cycle other than host interactions (Garrigues et al., 2018;Hajdu et al., 2019;Holzknecht et al., 2020). The SSCP CPL1 in the human pathogenic yeast Cryptococcus neoformans (Tremellales, Basidiomycota) is important for its adhesion to and invasive growth into substrates . Similarly, at least two SSCP families in filamentous fungi (HFBs and CPs) are amphiphilic, having profound or even superior surface activity (Wösten, 2001;Gaderer et al., 2014) and the ability to reduce the force of water tension (van der Vegt et al., 1996). HFBs are particularly small (~100 a.a.) and have a conserved pattern of eight cysteine residues C-CC-C-C-CC-C that efficiently stabilizes their tertiary structure by disulfide bonds (Linder et al., 2005). CPs are slightly larger (~150 a.a.) and have only four single cysteines. Recent physiochemical investigations and expression profile studies have shown that HFBs and CPs are probably functionally complementary (Gaderer et al., 2014). Proteins from both of these families can spontaneously assemble in interfaces, mediate the hydrophobicity of surfaces and reduce water tension (Linder et al., 2005;Gaderer et al., 2014). These properties attract research attention to HFBs for the development of multiple industrial or pharmacological applications, and to CPs for their ability to trigger plant immune responses and their role in plant pathogenicity.
Fungi from the fungivorous genus Trichoderma as well as a few other hypocrealean taxa are considered suitable models for the functional characterization of HFBs and CPs, because the respective gene families are expanded in their genomes (Kubicek et al., 2019;Gao et al., 2020). In contrast to the majority of Ascomycota fungi, which have a few HFBs and usually only one or two CPs , Trichoderma spp. can have seven (e.g. T. reesei and T. citrinoviride) to 15 (e.g. T. harzianum and T. gamsii) HFB-encoding genes (Kubicek et al., 2019). All whole-genome sequenced Trichoderma strains (September 2020) have at least three CPs, and a few species have four . Moreover, Trichoderma spp. are ecologically versatile, and several environmentally opportunistic species can parasitize other fungi, colonize dead wood and other cellulosic substrates, grow in soil, establish in rhizosphere, become endophytes, kill water moulds (Oomycota), control plantparasitic nematodes and attack immunocompromised humans . Thus, the expansion of surface-active SSCPs (saSSCPs) such as the HFBs and CPs in Trichoderma genomes may be linked to the diversity of the biotic and abiotic interactions of these fungi.
This study continues the evolutionary and functional characterization of saSSCPs in Trichoderma (Kubicek et al., 2008;Gao et al., 2020). We detected and characterized a new family of saSSCPs, named here as hyphosphere proteins (HFSs) and the two new Trichoderma HFBs, here named HFB11 and HFB12. The application of recombinant HFB2, HFB4, HFB12, HFS1 and EPL1 proteins (produced by Pichia pastoris) to Solanum lycopersicum (tomato) roots resulted in divergent effects on fungal attachment and in a noticeable phytotoxic effect, while the complete secretome of T. guizhouense stimulated tomato growth. Interestingly, all recombinantly produced saSSCPs (HFS1, HFB2, HF4, HFB12 and EPL1) showed convergent surfacemodulating activity that favoured the attachment of hyphae indicating that the high diversity of saSSCPs in Trichoderma likely evolved for the efficient conditioning of various surfaces in the hyphosphere to a hydrophilic state, making it suitable for attachment or colonization by the fungus.

Results
Genome mining of T. guizhouense revealed a new family of SSCPs and two new HFBs We first performed a broad screening of the wholegenome sequence of T. guizhouense NJAU 4742 (Druzhinina et al., 2018;Kubicek et al., 2019) for saSSCPs such as HFBs and CPs. For this purpose, we expanded the previously available annotations (Kubicek et al., 2019) and collected the small (<30 kDa, <300 a.a.) secreted proteins that contained four to eight single or double cysteine residues (see Experimental procedures for further details). The results, shown in Table 1 and in  the Supporting Information 1 Table S1, confirmed the three CPs characterized in this species by Gao et al. (2020). These data also revealed that there are two putative HFBs described by Kubicek et al. (2019) do not belong to this family. The first (GenBank ID: OPB44529) lacks the signal peptide. A comparison with other genomes revealed that it is likely an orphan Trichoderma protein that is present in section Trichoderma (e.g. T. gamsii, T. atroviride and T. asperellum) and Harzianum clade species but absent in the Longibrachiatum clade. The second putative HFB (GenBank ID: OPB38365) has the HFB-specific cysteine pattern of C-CC-C-C-CC-C, but it harbours six more cysteines (14 total) and has a size equivalent to that of three HFBs (28 kDa, 277 a.a.). Although it is probably a member of the SSCPs, it presents as a single copy protein that is relatively conserved in all mined ascomycetes, differing from the HFB family. Thus, we concluded that OPB38365 possibly represents a separate family of SSCPs and excluded it from this investigation. We also found two novel putative HFBencoding genes that were previously detected by Kubicek et al. (2019), which we name here as hfb11 (OPB42521, encoding HFB11) and hfb12 (OPB45278, encoding HFB12). Thus, the T. guizhouense genome encodes at least 10 confirmed HFBs (Table 1). Moreover, we found three other proteins with primary structures fulfilling our search criteria (Table 1). These proteins (OPB41482, OPB43804 and OPB40307) had a pattern of eight single cysteine residues and were larger than the HFBs (7-16 kDa) and CPs (12-14 KDa), with molecular weights varying from 21 to 28 kDa and sequences contained 219-280 residues.
As the physiochemical surface-active properties of the HFBs and CPs in Trichoderma spp. are relatively well studied (Kubicek et al., 2008;Seidl-Seiboth et al., 2011;Espino-Rammer et al., 2013;Bonazza et al., 2015;Przylucka et al., 2017b;Gao et al., 2020), we compared their hydropathy profiles to those of the newly detected proteins (Fig. 1A). The principal component analysis (PCA) based on the Kyte and Doolittle hydrophobicity coefficient (Kyte and Doolittle, 1982) and the values of the GRAVY scores, which define a protein sequence's average degree of hydrophobicity, showed that the novel proteins had hydropathy profiles similar to those of the HFBs and CPs (Fig. 1B) and therefore had the potential for being surface-active or amphiphilic. In particular, the GRAVY scores suggested these novel proteins were comparable to HFB6 and the pseudo-class I HFB members HFB9a/b in Trichoderma (Seidl-Seiboth et al., 2011). As they have a unique pattern of eight single cysteine residues (see below) and are considerably larger in size, we assigned them to a putative new family of saSSCPs and named them HFSs what is phonetically similar to 'hyphosphere' proteins, see below: HFS1 (OPB41482), HFS2 (OPB43804) and HFS3 (OPB40307).
In general, the primary structure analysis indicated that HFB2, HFB3, HFB4, HFB5 and HFB11 are potentially more hydrophobic than HFB6, HFB9a, HFB9b, HFB10, HFS1 and HFS2, which are somewhat hydrophilic. Proteins such as HFB12, CPs (EPL1, EPL2 and EPL3) and HFS3 have intermediate values. The isoelectric point (pI) analysis showed no correlation with a particular family or hydropathy profile, with ranges varying from 3.83 to 6.71 in the HFBs, 4.37 to 5.78 in the CPs and 4.36 to 6.03 in the HFSs. Only a few proteins, namely EPL1, HFS1 and HFB10, had pIs close to the 6-7 range, while the majority of the others could be considered 'cationic' (Huber et al., 2020), suggesting they would require unique environmental conditions and potentially precipitate under neutral (pH 7; like water) conditions. All 16 proteins have a relatively short signal peptide (16-21 a.a.), which is characteristic of HFBs (Kottmeier et al., 2011).
To confirm the functionality of the saSSCP-encoding genes, we tested their expression during the main stages of the asexual life cycle of T. guizhouense when these proteins are known to be active. This included submerged vegetative growth that associated with high CP expression (Gaderer et al., 2014;Gao et al., 2020), and aerial hyphae formation and conidiation which require large amounts of HFBs (Lugones et al., 2004;Winefield et al., 2007;Cai et al., 2020). The qPCR analysis confirmed the expression of all CP-and HFB-encoding genes except hfb6 and hfb11 (Table 1). The expression of hfb12 was low compared with major hfbs such as hfb4, hfb10, hfb2 and hfb3 (Cai et al., 2020), but showed the hfb-characteristic pattern of a drastic (several hundred folds) increase with the formation of aerial hyphae and conidiation. The three genes encoding HFS proteins showed a relatively low but consistent expression profile that did not change much with the development of the fungus (Table 1). Thus, our data suggest that at least 14 of the detected putative saSSCP-encoding genes are actively transcribed, while the other two could either be putative pseudogenes or become functional under specific conditions not tested here.

Evolutionary analysis suggests a long evolutionary history of HFSs in filamentous Ascomycota
The newly described HFB11 in T. guizhouense is 91 a.a. long and has a molecular weight of 7.59 kDa. This protein also has the characteristic C-CC-C-C-CC-C pattern and a short signal peptide (20 a.a.). The evolutionary analysis revealed it to be an orphan protein present in only a few species in the Harzianum clade of the genus Trichoderma ( Fig. 2A), including T. simmonsii, T. afroharzianum, T. harzianum and the putatively new species T. sp. M10 (Cai and Druzhinina, 2021). We did not detect any hfb11 sequences in Harzianum clade species such as T. pleuroti and T. pleuroticola or in the sister clade Virens. The amino acid sequence of HFB11 was 50% similar to HFB5 (hfb5, MF527119; Fig. 2A), which is in agreement with the similar primary structure of these two proteins (Fig. 1).
Another newly described HFB12 has a sequence of 93 a.a. harbouring the typical C-CC-C-C-CC-C cysteine pattern of HFBs that provides for an M w of 7.57 kDa with a 16 a.a. signal peptide. The phylogenetic analysis ( Fig. 2B) showed that among the 42 Trichoderma spp. (August 2020) whole-genome sequences, hfb12 can be found in the Harzianum and Virens clades, T. cyanodichotomus (which is phylogenetically close to T. virens), and in some species of section Trichoderma such as T. asperellum, T. asperelloides and T. hamatum, but not in the species T. atroviride or T. gamsii. The gene is also lacking in the Brevicompactum clade and section Longibrachiatum. Interestingly, the phylogeny of HFB12 was not concordant with the phylogeny of the genus (Kubicek et al., 2019), which indicated the likely scenario of a gene loss (GL) event in section Longibrachiatum and a partial loss in the other clades. No homologues to the hfb12 genes were found in other fungi, though the closest neighbour was <50% similar to the Fusarium oxysporum species complex (Hypocreales, Ascomycota). Phylogenetically, hfb12 was highly divergent compared with the other HFBs in Trichoderma. Thus, we assume hfb12 is an orphan Trichoderma gene that is not present in the core genome of the genus or in other fungi.
To investigate the evolution of the T. guizhouense HFSs, we performed a sequence similarity search (BLASTP) against the NCBI and JGI databases and screened all 42 publicly available Trichoderma genomes [August 2020; (Cai and Druzhinina, 2021)] as well as the genome of the closely related hypocrealean fungus Escovopsis weberi (de Man et al., 2016). The multiple sequence alignment revealed that the pattern of eight single cysteine residues of HFSs was highly conserved in all Trichoderma spp. For example, the HFS2 in T. reesei (XP_006967391) had a structure of C 49 -C 65 N 66 N 67 L 68 C 69 -C 129 -C 152 -C 174 -C 182 -C 194 . In other fungi (see below), the amino acids and length of the sequence between the C residues slightly varied, except for the space between the second and third cysteines, which were always separated by three amino acids; the first two were most frequently asparagine (N) or aspartic acid (D), while the third was an aliphatic amino acid such as leucine (L) or isoleucine (I). The maximum likelihood (ML) phylogram in Fig. 2C demonstrated that the three HFS proteins were homologous and shared the A. Deep purple colour highlights relatively high values of either the hydrophobicity based on protein GRAVY scores, the Kyte and Doolittle hydrophobicity coefficient (Kyte and Doolittle, 1982) for individual residues, or the isoelectric point (pI). Green corresponds to the respective low values. The cluster analysis on the right was applied to the hydropathy profile values using complete linkage and Euclidian distances. B. Principal component analysis (PCA) of the Kyte and Doolittle hydropathy coefficients, which shows a high similarity between proteins from the three families (i.e. both components explain minor variability). statistically supported common ancestor. The ancestor of HFS1 and HFS2 was, in turn, monophyletic with HFS3. In Trichoderma, the similarity between the amino acid sequences of the proteins varied from 62% to 100% for HFS1, 78% to 100% for HFS2 and 71% to 100% for HFS3. Three HFS-encoding genes were harboured in all the screened Trichoderma genomes with the exception of T. citrinoviride, which was missing HFS2. In general, the topologies of the individual HFS subclades were concordant with the topology of the Trichoderma phylogenomic tree (Kubicek et al., 2019), suggesting the vertical evolution and stabilizing selection pressure (Druzhinina et al., 2018;Cai et al., 2020;Gao et al., 2020) that were recorded for the other groups of secreted proteins in Trichoderma.
The topology of the HFS phylogram suggests a GD event in the ancestor of the genus Trichoderma that resulted in paralogous copies of hfs1 and hfs2. To test this, we collected HFS sequences from other fungi and performed NOTUNG and T-Rex analyses (Chen et al., 2000;Boc et al., 2012;Gao et al., 2020), which allowed us to test for possible GD, GL and LGT events in the evolutionary history of the HFSs in fungi. The results showed that HFSs are common in filamentous Ascomycota, but absent in other Dikarya fungi. The analyses confirmed the evolution of HFS1 and HFS2 by GD (Fig. 3), showing that the event most likely took place in the ancestor of hypocrealean fungi. One duplicated copy of hfs1 was inherited by the fungi from most other families, while hfs2 was only maintained in Trichoderma and likely lost in the other families. Interestingly, our analysis showed that hfs1 was putatively laterally transferred to the genome of Tolypocladium ophioglossoides (Hypocreales, Ascomycota), which has its own copy of hfs1 that was obtained vertically. The third gene, hfs3, also evolved as a result of a GD event but only one copy was maintained in Trichoderma. Similar to the evolution of CPs , the evolutionary history of the HFSs in filamentous Ascomycota encountered numerous cases of LGTs, GDs and GLs, all indicating the functional significance of these proteins. Our data show that the genes encoding HFSs were present in parasites feeding on any type of live biomass and in polyphages feeding saprotrophically on different organic matter (Fig. 3). Thus, the evolutionary analysis indicates the functional importance of these genes for filamentous Ascomycota, but it does not suggest that the acquisition of these genes was associated with a particular nutritional strategy or lifestyle of the fungi.
Heterologous production of saSSCPs in P. pastoris In previous studies, we explored the use of a gene deletion strategy for the functional characterization of saSSCPs in Trichoderma (Cai et al., 2020;Gao et al., 2020). However, the high number of SSCPs and the low expression profile of the newly revealed proteins indicated that this approach would not be suitable for their characterization. Therefore, we selected HFB12 and HFS1 for recombinant production in a Komagataella phaffii [syn. Pichia pastoris (Kurtzman, 2009)] host. For the comparison, we used the previously characterized saSSCPs of Trichoderma such as the HFBs [HFB2, (Askolin et al., 2005), HFB4 (Cai et al., 2020)] and a CP [EPL1; (Bonazza et al., 2015;Gao et al., 2020)]. For the construction of mutants in P. pastoris, the production of each protein was confirmed by SDS-PAGE (Fig. 4A) and immunoblotting (Supporting Information 2 Fig. S1). The flask fermentation of the engineered P. pastoris strains yielded at least 0.3-0.5 g L À1 of recombinant protein, varying slightly between the proteins produced and the fermentation batches. The expected size of each recombinant protein, shown in the SDS-PAGE, excluded the possibility of glycosylation in the recombinants.
The behaviour of each recombinant protein (indicated by 'r') in water was estimated using dynamic light scattering (DLS) (Przylucka et al., 2017b). The hydrodynamic diameters of the proteins based on the mass weight (%) are shown in Fig. 4B. The majority (mass > 90%) of the rHFB2, rHFB4 and rEPL1 proteins displayed a hydrodynamic size of 4.3-6.7 nm, which corresponded to that of a monomer or dimer of previously reported proteins (Hakanpää et al., 2004;Przylucka et al., 2017b). The rHFS1 protein had a slightly larger particle size of 7.8 nm. Interestingly, rHFB12 was found mainly (100%) in large aggregates with a dynamic size >70 nm, which could also be seen in the intensity distribution plot (Supporting Information 2 Fig. S2). Therefore, using HFB1 [2-3 nm; (Hakanpää et al., 2004)] from T. reesei as a reference for monomers, rHFB2, rHFB4, rEPL1 and  Table S2. rHFS1 tended to form dimers or oligomers in a water solution, while rHFB12 easily formed aggregates, explaining the retention of this protein in the sample well during the SDS-PAGE analysis (Fig. 4A).

All saSSCPs convert surfaces to a moderately hydrophilic state
The measurements of the water contact angle (WCA) on initially hydrophobic [poly(ethylene-terephthalate), PET] and hydrophilic (glass) surfaces coated with the recombinant saSSCPs showed that all five proteins were surface-active and provided a remarkably consistent result (Fig. 5). Irrespective of the initial hydrophobicity of the material, all proteins converted both surfaces to a moderately hydrophilic state, which corresponds to a WCA between 30 and 45 , excluding rHFB12 on PET. Thus, rHFB4 and rHFB12 increased the surface hydrophobicity of glass (30%-44%, p < 0.05) and decreased that of PET (37%-51%, p < 0.05). The rHFB2, rEPL1 and rHFS1 coatings on glass surfaces did not differ from the control, but the hydrophobicity of PET was reduced by 70% (p < 0.05). As predicted by the primary structure analysis, the newly detected SSCPs (HFS1 and HFB12) have potent surface activity similar to that of HFBs and CPs. The aggregation of HFB12 in water (see above) did not abolish its surface activity, although it was slightly reduced compared with that of the other tested saSSCPs. Considering the shared evolutionary history of the three HFSs and the similarity of their hydropathy profiles, they can be assigned as a family of saSSCPs (Fig. 6). More interestingly, the results indicate that T. guizhouense secretes an arsenal of saSSCPs that are all capable of transforming the surfaces surrounding hyphae to a moderately hydrophilic state. The consistency of the results allows us to hypothesize that hydrophilic surfaces corresponding to a 30 -45 WCA are optimal for the fungus. Fig. 3. Evolution of the hyphosphere (HFS) proteins in fungi. A maximum-likelihood phylogenetic tree was constructed using IQ-TREE 1.6.12 (ultrafast bootstrap, N = 1000). The Bayesian Information Criterion was applied when searching for the best amino acid substitution model with ModelFinder (integrated into the IQ-TREE program). Statistically supported nodes (bootstrap values >75) are marked by dots. Arrows indicate putative cases of lateral gene transfer as revealed by NOTUNG and T-Rex; 'D' indicates the putative gene duplication events revealed through the same methods. Vertical bars denote the taxonomic positions of the corresponding species.
saSSCPs assist Trichoderma hyphal attachment to hydrophilic surfaces As T. guizhouense has the ecophysiological features of an aero-aquatic fungus (Cai et al., 2020) and is also rhizosphere competent (Cai et al., 2013Zhang et al., 2016), we hypothesized that saSSCPs that are frequently expressed during submerged growth (Table 1) and transform surfaces to a moderately hydrophilic state (similar to that of glass) could also assist mycelial attachment to such modified hydrophilic surfaces in water. (Note that due to the abundance of saSSCPs, all hydrophobic surfaces would ultimately become hydrophilic.) To model such surfaces, we used glass wool (artificial) and the roots of Solanum lycopersicum (tomato) seedlings (natural substrate). The results showed that the T. guizhouense strain had a higher affinity to the glass wool, while attachment to roots was less successful (Fig. 7). The addition of 1 μM of any rsaSSCPs produced a significant (p < 0.05) and multifold increase in hyphae attachment to the glass wool. As for the attachment to roots, three rHFBs produced a similar result, improving colonization, while rEPL1 and rHFS1 had a reverse impact, preventing hyphae attachment. To see whether these functions were species specific, we performed the same experiment with another Trichoderma strain, T. harzianum CBS 226.95, which is related to T.
guizhouense (Kubicek et al., 2019) but has a different ecologic profile (Cai et al., 2020). The results for T. harzianum repeatedly showed that rsaSSCPs can assist in the attachment of the fungus to glass wool, but none of the selected rsaSSCPs from T. guizhouense NJAU 4742 could prime the colonization of T. harzianum CBS 226.95 on tomato roots (Fig. 7). This indicates that root attachment is a complex process that is probably regulated by both partners, and the addition of saSSCPs does not necessarily have an impact.

Some saSSCPs are toxic for plants and can trigger an immune response
As T. guizhouense itself and its secretomes have considerable growth-promoting effects on plants [Supporting Information 2 Fig. S3 and Cai et al., 2013, we tested the effect of the rsaSSCPs on the growth of tomato seedlings using 0.1 and 1 μM of a purified protein suspension. Surprisingly, with the exception of HFB12, these treatments had a significantly adverse effect on plant growth that varied in strength and range between the different rsaSSCPs (Fig. 8). One micromole of rHFS1 and rEPL1 significantly (p < 0.05) reduced plant growth to 62% and 64% respectively, of the control biomass without the protein addition. The rHFB2 and rHFB4 proteins had a less severe effect (73%-78%) on plant growth, though it was still statistically significant (p < 0.05) when compared with the control, and the plant  height was reduced accordingly. The rHFB12 protein had a neutral effect on the plants, which may be due to its strong aggregation in water (Fig. 4). The growth inhibition effects of the protein application increased when the concentration was raised from 0.1 to 1 μM (Fig. 8).
Since saSSCPs are frequently considered effector-like proteins (Gaderer et al., 2014;Ruocco et al., 2015;Gao et al., 2020), we also tested whether the immune system of the tomato seedlings was activated by the addition of the rsaSSCPs. To this end, we monitored the expression of biomarker genes for the jasmonic acid (JA)-and salicylic acid (SA)-defence pathways in plants. The results ( Table 2) show that 0.1 μM of rHFB2 and rHFB12 slightly but significantly (p < 0.05) activated the JA-and SAmediated defence pathways, as shown by the upregulation of the respective genes LoxA (11-13-fold) and PAL (~two folds). The rEPL1 protein did likewise. However, the elicitation effect of these proteins was not seen when the protein concentration was increased to 1 μM. In contrast, neither rHFB4 nor rHFS1 was able to trigger the plant systemic defence pathways at the two applied concentrations, likely because under natural conditions, HFB4 is associated with conidiation (Cai et al., 2020) and thus secreted in very low amounts in the rhizosphere, and the expression of HFS1-encoding gene is generally low (Table 1).

Discussion
Studies on HFBs and CPs have frequently focused on the effect of these proteins on other organisms. It is not a surprise that plants and animals recognize these massively produced proteins as indicators of fungal proximity, and they respond by activating their defence mechanisms (Frías et al., 2011;Wang et al., 2019;Gao et al., 2020). In plant-pathogenic fungi, numerous HFBs and CPs have been reported as effector-like proteins or elicitors (Rohe et al., 1995;Whiteford et al., 2004;Jeong et al., 2007;Frías et al., 2011). However, our evolutionary analyses and the study of Gao et al. (2020) suggest that guizhouense NJAU 4742 were imaged on a cellophane surface covering a potato dextrose agar plate 18 h after inoculation (incubated at 25 C in the dark). The stripes in the hyphosphere correspond to unequal light reflection, likely due to the assembly of different saSSCPs. The right and left images depict a hyphosphere of multiple (left) and a single (right) hyphae imaged using similar setups. Fig. 7. Effect of recombinantly produced surface-active small secreted cysteine-rich proteins (saSSCPs) on the ability of Trichoderma hyphae to attach to biotic (tomato roots) and abiotic (glass wool) surfaces. A. Sample images obtained for T. guizhouense and T. harzianum with or without addition of recombinantly produced HFB4. Images were obtained using a root scanner after 48 h of incubation. Scale, 1 cm. B. Quantification of fungal attachment to biotic (tomato roots) and abiotic (glass wool) surfaces using a qPCR determination of the single copy constitutively expressed house-keeping gene tef1 (Cai et al., 2020;Gao et al., 2020). their role in plant virulence is likely a secondary function of the saSSCPs, because genes encoding HFBs and CPs are also maintained in the genomes of saprotrophic and animal-associated fungi.  Besides the detection of a new family of saSSCPs (see below), the results of this study allow us to offer one possible explanation to the overall diversity of HFB-and CP-encoding genes in the genomes of fungi with different lifestyles. We propose that the critical property of HFBs, CPs, the newly detected HFSs and probably other still undetected SSCPs is their ability to modify the interfaces around hyphae (i.e. in hyphosphere). We show that all T. guizhouense saSSCPs can convert the tested abiotic surfaces to a moderately hydrophilic state and facilitate hyphae attachment to these surfaces. We propose that T. guizhouense and a dozen other environmentally opportunistic Trichoderma spp. have a rich reserve of various surface-active proteins that can secure their attachment to virtually any surface, thus contributing to the nutritional and ecological versatilities of these fungi. This study showed that several HFBs, CPs and HFSs had very similar surface-modulating abilities. They all transformed the highly hydrophobic surface of PET (which can resemble a plant or insect cuticle) and hydrophilic glass to the same moderately hydrophilic state. However, the 16 tested genes encoding saSSCPs from different families had unequal expression and regulation profiles. This brings us to the hypothesis that for fungi that perform most of their life-sustaining functions directly through their body surfaces, suitable interface hydrophobicity in hyphosphere is a critical issue. The regulation of saSSCPs is likely fine-tuned to particular environmental conditions around the hypha, making its surface and the surfaces around it sufficiently hydrophilic for nutrient absorption or coverage in a hydrophobic 'raincoat' to protect it from stressors. This hypothesis can be tested in future studies.
This study showed that all saSSCPs improved the attachment of Trichoderma hyphae to a hydrophilic glass surface. Due to the fact that fungi always secrete a diversity of saSSCPs, tests with initially hydrophobic surfaces would not be feasible, as all such surfaces are converted to a hydrophilic state by the activity of such proteins. However, attachment to roots appeared to be a more complicated process. The recombinant HFBs facilitated the attachment of T. guizhouense to tomato roots, while EPL1 and HFS1 prevented attachment. With or without saSSCPs, T. harzianum could not attach to tomato roots. Similar results of improved colonization were obtained when the CP epl1 gene was deleted from T. guizhouense .
Interestingly, we found several other reports of improved root colonization by fungal mutants lacking one or another CPs  and cases when the lack of HFBs was associated with a reduced ability to attach to roots (Viterbo and Chet, 2006;Guzm an-Guzm an et al., 2017). Together, this suggests that it is not EPL1 or HSF1 that prevents the attachment to roots, but rather that plants counteract the attachment process, which we explain by the actions of plant defence mechanisms. The surfaces of roots, in particular root hairs, must be suitable for the efficient absorption of water. The abundant surface-modulating proteins secreted directly into the rhizosphere by fungi such as Trichoderma can alter the proper state of the root surfaces and impede plant growth. In this study, we demonstrated that the addition of recombinant saSSCPs to hydroponic medium has a significant adverse effect on plant growth. We also showed that those proteins that prevented the attachment of T. guizhouense to roots (EPL1 and HFS1) also triggered an immune response in the tomato plants, while the HFBs went unrecognized by the tomato immune system. The usual silence of these genes during submerged growth may explain the lack of a tomato response to HFBs (Cai et al., 2020). This fits the possibility that roots are usually confronted with CPs and HFSs but not HFBs. As a result, they may lack an efficient system for HFB recognition. Interestingly, the complete secretomes of T. guizhouense and the fungus itself were beneficial for tomato growth, likely due to the presence of such secondary metabolites as plant hormones (Vinale et al., 2009;Cai et al., 2013) and other small molecules (Harman et al., 2004;Li et al., 2015). We assume that under natural conditions, the outcome of fungal-plant interactions for the partners is determined by the tradeoffs between fungal-beneficial and plant-beneficial processes and that the balance is determined by the environmental conditions, which in turn, trigger the secretion of a suitable saSSCP cocktail.
We also detected a new family (HFSs) of saSSCPs with an amphiphilic nature similar to that of the HFBs and CPs. Their hydropathy profiles resemble those of relatively hydrophilic HFBs such as HFB6, HFB9a, HFB9b and HFB10. The evolutionary analysis performed for the specific HFS-encoding genes (Fig. 3) showed that they are commonly present in all Pezizomycotina fungi genomes and have no specificity towards a particular lifestyle. The evolutionary history indicated numerous LGT events and duplications, supporting the functional necessity of these proteins in fungi. Similar results were obtained when the phylogeny of the CPs in Pezizomycotina was studied .
In conclusion, we would like to highlight that SSCPs in general, and the surface-active members (saSSCPs) of this group in particular, are unique fungal proteins. Therefore, they should be considered from the perspective of hyphal needs rather than the response of other organisms to their presence. We propose that the saSSCPs studied here constitute a part of a larger group of hyphosphere-specific proteins that modulate hyphal attachment to surfaces and interactions with other organisms. This work provides initial results in the investigation of a protein-based surface recognition system in fungi that results in a sophisticated cocktail of surface-active proteins altering the environment around the hyphae.

Experimental procedures
Microbial strains, plant materials and cultivation conditions Trichoderma guizhouense strain NJAU 4742 (Zhang et al., 2016;Zhang et al., 2019;Cai et al., 2020;Gao et al., 2020;Pang et al., 2020), which is commercially available as a biofertilizer agent in China, was used as the wild type throughout this work. The sibling species T. harzianum (CBS 226.95) was also included as a reference strain when performing the colonization experiment. To express the Trichoderma genes in yeast, the Pichia pastoris strain KM71H from the EasySelect™ Pichia Expression Kit (Invitrogen) was used, with the plasmid pPICZαA used as the backbone vector. Pichia pastoris strains PpEPL1 , PpHFB2 and PpHFB4  which produce a recombinant EPL1, HFB2 and HFB4 respectively, were generated in our previous works. If not otherwise stated, the Trichoderma strains were maintained on potato dextrose agar (BD Difco, USA) at 25 C. The Pichia strains were maintained on yeast extract peptone dextrose agar (Sigma-Aldrich, USA) and fermented in a buffered minimal medium containing glycerol. Protein production was induced by methanol when needed, as described previously (Zhang et al., 2016;Gao et al., 2020). All strains used in this research are listed in Supporting Information 1 Table S3. Tomato seedlings of Solanum lycopersicum cv. HEZUO903 were cultivated in a hydroponic system filled with a 30% Murashige and Skoog basal salt mixture medium (MS; Sigma-Aldrich), which was maintained under controlled conditions of 25 C with cycled illumination (light:darkness = 12 h:12 h).

Plant growth promotion assay
The effect on plant growth from inoculating T. guizhouense NJAU 4742 and its metabolites was evaluated by adding 0.4 ml of 10 6 ml À1 spores or 0.4 ml of 5-day 30% MS (plus 1% glucose) fermenting broth (without fungal cells) to 40 ml of MS medium supplemented with 1% glucose in the hydroponic system. In this test, 3-week-old tomato seedlings showing equal growth were selected and treated as described, and untreated plants were used as the control. The seedlings were allowed to grow for another 3 weeks. Data regarding the plant height and fresh weight were recorded at the end of the experiment. The root growth parameters were determined with a root scanner (Epson Perfection V700 photo, Seiko Epson, Japan).

Standard molecular techniques
The RNeasy Plant Mini Kit (Qiagen, Germany) was used to extract total RNAs from fungal or plant materials according to the manufacturer's instructions. The Rev-ertAid™ First Strand cDNA Kit (Thermo Scientific, USA) was used to synthesize cDNAs with an oligo (dT) 18 primer. The qPCR reactions (20 μl) were performed with 0.5 μM of each primer, 100 ng cDNA and 10 μl of iQ™ SYBR Green PCR super mix (BioRad, USA). The thermal program was implemented in a qTOWER real-time PCR system (Jena Analytics, Germany) with a 3 min initial denaturing at 95 C followed by 40 cycles of 15 s at 95 C and 30 s at 60 C. The specificity of each qPCR reaction was verified by running a melting curve from 55 to 95 C. PCR reactions (20 μl) contained 0.5 μM of each primer, 100 ng of template DNA, 1Â Phanta Max Buffer mix and 2 U Phanta Max Super-Fidelity DNA Polymerase (Vazyme Biotech, China). The thermal cycling parameters consisted of a 3 min initial denaturing step at 95 C; 30 cycles of 15 s at 95 C, 15 s at 59 C and 30 s at 72 C; and a 5 min final extension at 72 C. Supporting Information 1 Table S4 provides the primer information.

Expression of Trichoderma genes in Pichia pastoris
The encoding region of the gene of interest was amplified from T. guizhouense NJAU 4742 cDNA. The signal peptides were predicted by SignalP 5.0 (http://www.cbs.dtu. dk/services/SignalP/) and excluded from the amplicons. Each amplified fragment was cloned into the pPICZαA plasmid between the EcoR I and Xba I sites. The recombinant proteins coded by this construction harbour the yeast α-factor signal sequence at the N-terminus and a 6Â His epitope at the C-terminus. The resulting plasmids were transformed into P. pastoris KM71H by standard electroporation transformation, and one positive transformant expressing the designated recombinant protein was chosen for all subsequent experiments. Detailed mutant construction and confirmation are given in Supporting Information 2 Figs S4 and S5. The recombinant proteins were purified as described in Przylucka et al. (2017b), concentrated via Vivaspin ultrafiltration devices (Sartorius, Germany) and re-suspended in 100 mM potassium phosphate buffer (PB; pH 5.8) or milli-Q water.

Protein biochemical assays
Protein samples were analysed by 15% SDS-PAGE followed by silver staining using the SilverQuest™ Silver Staining Kit (Life Technologies, Germany). The proteins separated by SDS-PAGE were further confirmed by blotting onto an Immobilon ® -P SQ Transfer Membrane (PVDF, 0.2 μm; Merck Millipore, USA) and immunologically visualized by binding with a mouse His-tag horseradish peroxidase antibody (GenScript, China) as described by Gao et al. (2020). The protein concentration was quantified with the BCA protein assay kit (Thermo Fisher Scientific, USA) according to the manufacturer's instructions.

Protein biophysical assays
The soluble state of each recombinant protein in milli-Q water was measured by DLS in a Malvern Zetasizer Nano-ZS (Malvern, UK) at 25 C. Proteins were prepared at a concentration of 0.6 g L À1 and filtrated through a Whatman ® 0.22 μm filter (Sigma-Aldrich). The mass weight (%) diameters of the proteins were obtained from three measurements. Milli-Q water without proteins was used as the control.
The surface activity (modification of a material's surface hydrophobicity) of the recombinant proteins was determined by WCA measurements using a Krüss EasyDrop DSA20E (Krüss GmbH, Germany). Four replicates of each surface material (glass or PET) were prepared as described by Espino-Rammer et al. (2013) with small modifications. Briefly, the surface materials were trimmed to a size of 1.5 Â 0.5 cm 2 and defatted with 70% (vol./vol.) ethanol. They were then washed in sequence with 5% (wt./vol.) Triton X-100, 100 mM Na 2 CO 3 and deionized water, with each wash step performed for 30 min at 50 C. The surfaces were coated with the recombinant proteins by applying a 10 μM (dissolved in PB) concentration of the protein and incubated at 25 C for 12 h. The coated materials were washed twice with deionized water and dried before the WCA measurements.

In vitro protein bioactivity assays
The in vitro effects of the selected saSSCPs on plant growth and immune response were determined using the hydroponic system as described above. The proteins were applied to 3-week-old tomato seedlings at concentrations of 0.1 and 1 μM. In the plant growth promotion test, the seedlings were allowed to grow for an additional 3 weeks, while the seedlings used for testing the immune response were sampled for RNA extraction 48 h after applying the protein. The expression of genes corresponding to different plant defence pathways was monitored with RT-qPCR as described above. The bioactivity of each SSCP to assist Trichoderma (T. guizhouense NJAU 4742 and T. harzianum CBS 226.95) colonization was investigated by quantifying the tef1 gene copy number of each strain colonized on roots and on an artificial material, i.e. glass wool (see detailed methodology in our previous work; ).

Genome mining and evolutionary analysis
Genome mining for HFB-encoding genes and the new SSCP family -HFS-encoding genes in the 42 Trichoderma whole-genome sequenced strains (listed in Supporting Information 1 Table S2) -was performed using RapidMiner (version 8.2, USA) with the HFBspecific cysteine sequence pattern of C-CC-C-C-CC-C or the HFS-specific pattern of C-CXXX-C-C-C-C-C-C (X represents any possible amino acid). In addition, EffHunter (Carre on-Anguiano et al., 2020) was used to search for effector proteins that were then manually verified by the presence of a secretory signal peptide (SignalP 5.0). The CP sequences were obtained from our previous report . The computation of the theoretical isoelectric point (pI) and molecular weight (M w ) of the SSCPs from T. guizhouense was performed using the online Compute pI/Mw tool (https://web.expasy. org/compute_pi/). The grand average of hydropathy (GRAVY) value for each protein sequence was calculated using the GRAVY Calculator (http://www.gravycalculator.de/index.php). Detailed hydropathy profiles were calculated using https://web.expasy.org/protscale/. The possibility of a transmembrane region for each protein was checked using an online prediction program (http://www.detaibio.com/tools/index.php?r=transmembrane %2Findex). The sequences of each protein family were aligned using the MUSCLE algorithm integrated into AliView 1.23 (Edgar, 2004;Larsson, 2014). A ML phylogenetic tree was constructed for the new proteins using IQ-TREE 1.6.12 (ultrafast bootstrap, N = 1000; (Nguyen et al., 2015)). The Bayesian Information Criterion was applied when searching for the best amino acid substitution model with ModelFinder (integrated into the IQ-TREE program; (Kalyaanamoorthy et al., 2017)). An evolutionary analysis that included the NOTUNG and T-Rex programs was performed as described in our previous works (Druzhinina et al., 2018;Gao et al., 2020). In the NOTUNG analysis, the costs of GL, GD and LGT were set at a ratio of 1:3:9, which strictly constrains the possibility of GD and LGT events.

Statistical analysis
The relative expression fold changes of the genes of interest were calculated according to the 2 -ΔΔCt method using tef1 as the housekeeping gene (Cai et al., 2020;Gao et al., 2020). The mean and standard deviation were calculated and statistically analysed by ANOVA and Tukey multiple comparison tests at p < 0.05 using STATISTICA 6 software (StatSoft, USA). The bar plot, heatmap and PCA were performed with R version 3.2.2 (https://cran.r-project.org/bin/windows/base/old/3.2.2/). Unless otherwise stated, the data regarding plant growth, root colonization and immune response were obtained from at least three biological replicates.

Supporting Information
Additional Supporting Information may be found in the online version of this article at the publisher's web-site: Table S1. Small secreted cysteine-rich protein (SSCP) sequences. Table S2. Strains used in this study. Table S3. Primers used in this study. Table S4. Genomes used in this study.   S2. The hydrodynamic diameter of each recombinant protein in milli-Q water. The mass weight (%) diameter was measured by dynamic light scatting (DLS) in a Malvern Zetasizer Nano-ZS (Malvern, UK) at 25 C. Proteins were prepared in a concentration of 0.6 g L À1 . Milli-Q water without proteins was used as the control. Fig. S3. Impact of Trichoderma guizhouense NJAU 4742 inoculation and its secretome on plant growth. To see, whether saSSCPs impact the plant-growth promoting effect of T. guizhouense NJAU 4742 we collected the culture filtrate (secretome, 1% Sec) corresponding to the submerged stage growth stage when most of saSSCPs were expressed ( Table 2). The spore suspension of T. guizhouense NJAU 4742 (10 4 spores ml À1 ) (Tri) and the cell-free Sec to the roots of tomato seedlings cultivated in a hydroponic system. Values (mean AE Sd) labelled with the same letter do not statistically significantly differ (ANOVA, p < 0.05). Fig. S4. Schematic diagram of expressing SSCP-encoding genes under a methanol-inducible promoter (P AOX1 ) in P. pastoris. α-Factor, native Saccharomyces cerevisiae α-factor secretion signal peptide; T AOX1 , native transcription termination from AOX1 gene of P. pastoris; Sh ble cassette, ble gene from Streptoalloteichus hindustanus driving resistance to zeocin. Heterologous expression of SSCP-encoding genes from T. guizhouense NJAU 4742 in P. pastoris strain KM71H was performed as described in Gao et al. (20202020). The amplified gene of interest (hfb12 and hfs1) (without signal peptide or intron sequences) was respectively inserted into the position between the restriction site of EcoR I and Xba I of plasmid pPICZαA. The electroporation resulted in numerous positive transformants on zeocin-contained (100 μg ml À1 ) YPD plates (shown in Supporting Information 2 Fig. S4 with three randomly selected mutants). Fig. S5. PCR verification of Pichia mutants expressing hfb12 or hfs1. PpVEC, represents the mutant obtained by transforming the original vector pPICZαA (without a hfb, epl or hfs gene) to P. pastoris cells; PpHFB12, represents mutants harbouring hfb12; PpHFS1, represents mutants harbouring hfs1. DL2000 DNA marker (Vazyme, China) was used in the gel electrophoresis. PCR products were further confirmed by sequencing.