Conservation and expansion of a necrosis‐inducing small secreted protein family from host‐variable phytopathogens of the Sclerotiniaceae

Abstract Fungal effector proteins facilitate host‐plant colonization and have generally been characterized as small secreted proteins (SSPs). We classified and functionally tested SSPs from the secretomes of three closely related necrotrophic phytopathogens: Ciborinia camelliae, Botrytis cinerea, and Sclerotinia sclerotiorum. Alignment of predicted SSPs identified a large protein family that share greater than 41% amino acid identity and that have key characteristics of previously described microbe‐associated molecular patterns (MAMPs). Strikingly, 73 of the 75 SSP family members were predicted within the secretome of the host‐specialist C. camelliae with single‐copy homologs identified in the secretomes of the host generalists S. sclerotiorum and B. cinerea. To explore the potential function of this family of SSPs, 10 of the 73 C. camelliae proteins, together with the single‐copy homologs from S. sclerotiorum (SsSSP3) and B. cinerea (BcSSP2), were cloned and expressed as recombinant proteins. Infiltration of SsSSP3 and BcSSP2 into host tissue induced rapid necrosis. In contrast, only one of the 10 tested C. camelliae SSPs was able to induce a limited amount of necrosis. Analysis of chimeric proteins consisting of domains from both a necrosis‐inducing and a non‐necrosis‐inducing SSP demonstrated that the C‐terminus of the S. sclerotiorum SSP is essential for necrosis‐inducing function. Deletion of the BcSSP2 homolog from B. cinerea did not affect growth or pathogenesis. Thus, this research uncovered a family of highly conserved SSPs present in diverse ascomycetes that exhibit contrasting necrosis‐inducing functions.


| INTRODUC TI ON
Some of the most economically important eukaryotic phytopathogens are fungi (Dean et al., 2012). Confined mainly to the Ascomycota and Basidiomycota, these fungi have evolved the means to penetrate plant tissue and sequester valuable nutrients, all at great expense to the host plant. The lifestyles of fungal phytopathogens vary, from obligate biotrophs that are unable to survive outside host tissue, to broad-host necrotrophs that sequester nutrients from necrotized tissue (Oliver and Ipcho, 2004).
Independent of their lifestyles, all phytopathogenic fungi secrete virulence factors, also known as effectors, to aid in the establishment and development of infection within their host(s) (Cook et al., 2015;Lo Presti et al., 2015). Fungal effectors consist of a diverse group of molecules, including toxic secondary metabolites, enzymatic proteins, nonenzymatic proteins, and small interfering RNA molecules (Howlett, 2006;Stergiopoulos and de Wit, 2009;Weiberg et al., 2013;Collemare et al., 2019). Many fungal effectors discovered previously are small, cysteine-rich proteins that are secreted during host infection (Stergiopoulos and de Wit, 2009).
The mechanisms by which proteinaceous effectors influence the host are extremely diverse and have been well described for several biotrophic fungi. The Cladosporium fulvum Avr2 effector actively suppresses the host immune system by inhibiting host proteases that normally function to degrade fungal peptides in the host apoplast (Rooney et al., 2005;van Esse et al., 2008). In a less direct manner, the C. fulvum Avr4 effector suppresses plant chitinase activity by forming a protective coat of protein over the fungal cell wall, preventing chitinases from binding and degrading fungal chitin (van Esse et al., 2007). Ecp6 of C. fulvum scavenges chitin oligosaccharides in order to prevent the elicitation of the host immune system by these molecules (de Jonge et al., 2010). The Ustilago maydis chorismate mutase effector uses its enzymatic activity to coordinate changes to the biosynthesis of antifungal compounds in cells proximal to the infection zone (Djamei et al., 2011).
More recently, proteinaceous effectors of necrotrophic pathogens have also been described (Tan et al., 2010). The majority act to promote host-cell death in accordance with the lifestyles of these plant pathogens (Friesen et al., 2007;Lorang et al., 2012). Wheat pathogens Parastagonospora nodorum and Pyrenophora tritici-repentis both secrete the proteinaceous effector ToxA during host infection (Friesen et al., 2006). ToxA has been shown to interact indirectly with the host's Tsn1 protein to facilitate host-cell death and susceptibility (Faris et al., 2010). Only host genotypes that contain the Tsn1 "sensitivity" gene are susceptible to ToxA-mediated cell death.
Additional Tox proteins of P. nodorum have also been shown to act in conjunction with sensitivity proteins, including SnTox1, SnTox2, SnTox3, and SnTox4 (Liu et al., 2009(Liu et al., , 2012. Traditionally, proteinaceous fungal effectors were identified by their ability to trigger a hypersensitive response in incompatible host tissue (Lauge and De Wit, 1998). More recently, it has become possible to predict putative fungal effectors using bioinformatic analyses. In particular, fungal secretome prediction has become a popular strategy to identify proteinaceous fungal effectors (Amselem et al., 2011;Hacquard et al., 2012;Morais do Amaral et al., 2012;Guyon et al., 2014;Heard et al., 2015;Derbyshire et al., 2017). Secreted fungal proteins contain N-terminal signal peptides that guide these proteins through the classical secretion pathway (Lippincott-Schwartz et al., 2000). Together with transmembrane domain prediction tools, signal peptide sequence prediction analyses have been used to identify fungal secretomes from fungal proteomes (Emanuelsson et al., 2000;Petersen et al., 2011). The identification of putative fungal effectors within a secretome has traditionally involved filtering for small proteins (<200 amino acids) with a high cysteine content (Templeton et al., 1994;Hacquard et al., 2012). The cysteine residues within fungal effector proteins are proposed to form disulphide bonds, which help maintain protein stability within plant tissue (Joosten et al., 1997;Luderer et al., 2002). More recently, effector screening strategies have begun to incorporate complex information, including temporal and tissue-specific gene expression patterns, evidence for positive selection, proteomics, three-dimensional protein structure prediction, and comparative secretome analyses (Pedersen et al., 2012;Guyon et al., 2014;de Guillen et al., 2015;Lo Presti et al., 2015;Sperschneider et al., 2015;Heard et al., 2015;Mesarich et al., 2018).
Proteinaceous fungal effectors are often under strong selection pressure and must constantly evolve at the molecular level to maintain their function (Rouxel et al., 2011;Sperschneider et al., 2014).
To test this hypothesis, we predicted and compared the secretomes of the three necrotrophic fungal phytopathogens Botrytis cinerea, Sclerotinia sclerotiorum, and Ciborinia camelliae. All three of these fungal species are closely related members of the Sclerotiniaceae, sharing the ability to produce sexual fruiting bodies (apothecia) from melanized masses of mycelia (sclerotia) (Whetzel, 1945). Despite their similar necrotrophic lifestyles and taxonomic classification, the number of hosts that each of these pathogens infects varies considerably. It is estimated that B. cinerea and S. sclerotiorum have >1,400 and >400 host species, respectively, including the important crop species Glycine max (soybean), Brassica napus (canola), and Vitis vinifera (grape) (Boland and Hall, 1994;Bolton et al., 2006;van Kan et al., 2017). In contrast, the host range of C. camelliae is restricted solely to the floral organs of some Camellia species and interspecific hybrids (Kohn and Nagasawa, 1984;Denton-Giles et al., 2013). Here, we describe a bioinformatic approach that resulted in the discovery of a new family of conserved SSPs in B. cinerea, S. sclerotiorum, and C. camelliae that were massively expanded in the latter restricted host-range pathogen. We investigated the evolution of these novel SSPs within host-variable pathogens, their putative functions, and their role in fungal virulence.

| RE SULTS
2.1 | Comparative analysis of the C. camelliae, B. cinerea, and S. sclerotiorum secretomes reveals an enrichment of cysteine-rich SSPs in C. camelliae A total of 14,711 nonredundant C. camelliae protein sequences were predicted from genomic and transcriptomic data. Secretome prediction was performed for all three fungal species as outlined in Figure   S1. Protein sequences were screened for the presence of signal peptides, cellular localization signals, and the absence of transmembrane domains. A total of 749 C. camelliae, 754 B. cinerea, and 677 S. sclerotiorum secreted protein sequences were predicted.
To determine the level of conservation between the secretomes of S. sclerotiorum, B. cinerea, and C. camelliae, individual proteins from each species were independently aligned to the secretomes of the other two species. The single best alignment to each species was identified and amino acid identity information was plotted on a twodimensional scatterplot, producing a spatial representation of secreted protein conservation (Figure 1a). Each of the three scatter plots pro- To determine which types of proteins were conserved or divergent, all the secreted proteins were annotated using BLAST2GO (Conesa and Götz, 2008). A total of 76%-80% of proteins within each secretome were assigned to gene ontology (GO) categories of predicted protein (30%-35%), CAZyme (20%-25%), oxidoreductase (7%-10%), SSP (5%-10%) or protease (5%) (Figure 1b). The remaining proteins were distributed among 27 additional categories ( Figure 1b and Table S1).
The most striking difference between the secretomes of all three species appeared within the SSP category. C. camelliae had substantially more SSPs compared to B. cinerea and S. sclerotiorum  Table S3). Collectively these proteins will be henceforth referred to as Ciborinia camelliaelike small secreted proteins (CCL-SSPs).
All 75 CCL-SSP family members were predicted to contain an N-terminal signal peptide (Table S3). The predicted signal cleavage site was followed by a domain containing five conserved cysteine residues, which includes the conserved amino acid motif CTYCQCLFPDGSHCC. All 10 of the cysteine residues were predicted to form disulphide bonds and predicted disulphide connectivity patterns were conserved for all 75 proteins (Table S3). The conservation of cysteine residues suggests that the CCL-SSPs are likely to maintain a robust secondary structure.

| CCL-SSP homologs are present across fungal classes
To search for additional cross-species homologs, all 75 of the CCL-SSPs were aligned to the nonredundant protein database using BLASTP. A total of 23 additional fungal species were identified as having at least one CCL-SSP homolog ( does with another C. camelliae CCL-SSPs (Table 1). An alignment of all known homologs (n = 113) indicated that cysteine residues located from cysteine positions 2 to 8 were highly conserved (≥95%) within the greater CCL-SSP family ( Figure S4).

| C. camelliae CCL-SSPs are expressed during early infection
To determine whether the expanded family of CCL-SSP genes in F I G U R E 1 Prediction and comparative analyses of the secretomes of Ciborinia camelliae (pink), Botyrtis cinerea (blue), and Sclerotinia sclerotiorum (green). (a) Scatterplot analysis of the secretomes of C. camelliae, B. cinerea, and S. sclerotiorum. Predicted secreted proteins from each fungal pathogen were aligned to predicted secreted proteins from the other two fungal pathogens using BLASTP. Each query sequence produced two "best hit" amino acid (AA) identity scores. Three graphs were independently generated and were overlaid for comparison. I, a cluster of highly conserved proteins; II, a dominant cluster of C. camelliae-specific proteins. (b) A comparison of the annotated fungal secretomes of C. camelliae, B. cinerea, and S. sclerotiorum. Raw counts represent the number of proteins in each gene ontology category. The top 10 most common categories are shown. Nuc., nucleic acid modification proteins; Pri., primary metabolism proteins; Lip., lipases; Phos., phosphatases; Sec., secondary metabolism proteins; Pro., proteases; SSP., small secreted proteins; Ox/Red., oxidoreductases; CAZ., carbohydrate-active enzymes; Pred., predicted proteins. (c) A histogram displaying the distribution of SSPs for each species based on their cysteine content. Asterisks indicate statistical differences (Fisher's exact test using a 3 × 2 contingency table) (p < .001)

| Recombinant CCL-SSP family members induce host-cell necrosis
To gain insights into the cellular function of the CCL-SSP family, were selected for cloning and recombinant protein expression in Pichia pastoris. The 10 C. camelliae CCL-SSP genes are spread across the phylogenetic spectrum of the C. camelliae CCL-SSP family and include the two homologs that share the greatest amino acid sequence conservation with BcSSP2 and SsSSP3 (Figure 2a). Filter-sterilized culture filtrates were collected for each recombinant protein and F I G U R E 2 (a) Phylogenetic analysis of the conserved small secreted protein (SSP) family. Relative in planta transcript abundance was calculated for each of the conserved Ciborinia camelliae SSPs. Arrows indicate the 12 SSPs that were cloned and expressed as recombinant proteins and include those that were chosen for quantitative reverse transcription PCR (RT-qPCR) analysis (thick arrows). (b) RT-qPCR data for a subset of nine of the 73 C. camelliae SSP genes. All data were normalized to the two fungal housekeeping genes NAD and TUB. Histogram bars from lightest to darkest represent the expression of CcSSP37, CcSSP94, CcSSP36, CcSSP31, CcSSP93, CcSSP33, CcSSP41, CcSSP43, and CcSSP81. Relative expression data were normalized to 6 hr post-inoculation to allow for comparisons between genes. Error bars = ±1 SD infiltrated into host petal tissue. Nine of the 10 culture filtrates that likely to be a heat-stable protein.

| C-terminus-tagged CCL-SSPs have reduced necrosis-inducing function
The native recombinant proteins BcSSP2 and SsSSP3 induce strong host-cell necrosis. However, it is unclear whether the lack of necrosis from the native C. camelliae CCL-SSP proteins was due to a lack of synonymous function or reduced concentrations of soluble protein.
To facilitate the determination of CCL-SSP protein concentration, a c-Myc 6 × His-tag was included at the C-terminus of CcSSP43 T , CcSSP37 T , CcSSP92 T , BcSSP2 T , and SsSSP3 T . The presence of tagged CCL-SSP proteins in culture filtrate was confirmed by western blot using antibodies raised against the c-Myc tag (Figure 4a). To normalize for variations in protein concentration, the concentration of each tagged protein was semiquantified using a chemiluminescencebased quantification method. All tagged proteins were present in culture filtrates at a higher concentration than the 5-fold diluted SsSSP3 T protein (Figure 4b). Only SsSSP3 T undiluted and diluted (10-fold) culture filtrates were able to induce a host-cell necrosis response (Figure 4c,d). Compared to the native protein assays, the SsSSP3 T host-necrosis phenotype was delayed in its response and never completely necrotized the infiltrated area. BcSSP3 T also failed to induce any visible host-cell necrosis response, suggesting that the addition of the c-Myc 6 × His-tag to the C-terminus of the native BcSSP3 and SsSSP3 proteins perturbs necrosis-inducing function.

| The C-terminal region of SsSSP3 is essential for necrosis-inducing activity
To determine which regions of the CCL-SSPs contribute to the necrosis-inducing phenotype, chimeric proteins were created by swapping  Figure 5d). These data suggest that a specific conformation of the C-terminal half of the SsSSP3 protein is essential for necrosis-inducing ability, which may explain why C-terminal tagged SsSSP3 T and BcSSP2 T proteins had reduced necrosis-inducing ability.
The CCL-SSP family members share many characteristics with the NLP protein family. NLPs are virulence factors that are conserved in oomycetes, fungi, and bacteria (Qutob et al., 2006). They universally share the sequence motif GHRHDWE, which is part of a 24-amino acid motif that is thought to be recognized by nonspecific pathogen recognition receptors (Ottmann et al., 2009;Oome et al., 2014). These proteins are common in hemibiotrophic and necrotrophic microorganisms, and often exist as large gene families (Gijzen & Nurnberger, 2006).
Although the CCL-SSPs do not have the GHRHDWE motif, characteristics that are like the NLPs include the ability of the proteins to maintain function after heating (Oliveira et al., 2012), their nonhost-specific necrosis-inducing function (BcSSP2 and SsSSP3), and their inclusion in the proteomes of multiple unrelated species. Like our observations of the bcssp2 mutant strains, B. cinerea bcnep1 and bcnep2 NLP knockout strains have been shown to retain virulence (Arenas et al., 2010).
Future experiments should assess whether CCL-SSP necrosis-inducing function is dependent on secretion into the host apoplast (i.e., through the production of proteins that lack a signal peptide), a feature demonstrated previously for oomycete-derived NLPs (Qutob et al., 2006).

Nonhost-specific SSPs have recently been reported for
Zymoseptoria tritici (Kettles et al., 2017). These SSPs were shown to induce nonhost cell necrosis when transiently expressed in N. benthamiana. Kettles et al. (2017) concluded that the nonhost necrosis phenotype was due to an interaction between pathogen recognition receptors (PRRs) that detect nonadapted pathogen proteins. The necrosis-inducing phenotype observed for BcSSP2 and SsSSP3 may be a result of a nonspecific interaction with a PRR. The reduction of the necrosis-inducing function observed for C-terminus tagged recombinant BcSSP2 T and SsSSP3 T proteins suggests that slight changes in conformation affect the ability of these proteins to function in planta. We hypothesize that the necrosis-inducing CCL-SSPs described here are recognized by PRRs and act as microbe-associated molecular patterns (MAMPs), comparable to what has been described for the nonhost-specific Z. tritici SSPs and NLP superfamily members (Qutob et al., 2006;Oome et al., 2014;Kettles et al., 2017).
To maintain virulence, fungal phytopathogens must constantly respond to selection pressure from the immune system of their host (Jones and Dangl, 2006). This evolutionary pattern is particularly true for host-specific phytopathogens like C. camelliae. The birth-and-death evolution model has previously been used to describe the evolution of fungal effectors (Stergiopoulos et al., 2012) whereby genes duplicate and diversify in response to host selection pressure (Nei and Rooney, 2005). Based on results presented here, the 73 C. camelliae CCL-SSPs conform to the birth-and-death model. Evidence that C. camelliae CCL-SSPs have increased their numbers through gene duplication includes their proximity to each other in the C. camelliae draft genome, their nucleotide conservation, and their conserved exon/intron structure.
A similar scenario has been reported for SSPs in the biotrophic fungus U. maydis, where 12 genomic clusters of two to five SSP genes were discovered (Kämper et al., 2006).

| Plant and fungal material
Camellia 'Nicky Crisp' (Camellia japonica × Camellia pitardii var. pitardii) shrubs were maintained in a glasshouse at ambient temperature.  (Amselem et al., 2011). The bioinformatic pipeline used for secretome prediction is outlined in Figure S1. Predicted secretome proteins were annotated using BLAST2GO v. 2 (Conesa and Götz, 2008). GenBank BLAST annotations and BLAST2GO enzyme codes were used to manually group the proteins into 32 common categories (Table S1).
Proteins were conservatively annotated as SSPs if they were shorter than 200 amino acids in length and had ≥4% cysteine content (Kim et al., 2016).
The three secretomes of C. camelliae, B. cinerea, and S. sclerotiorum were compared to each other using BLASTP (Altschul et al., 1997). Alignments that produced an E value ≤10 −3 and included at least 10% of the length of the query sequence were designated as matches. Unsuccessful matches were assigned an amino acid identity of 0%. Alignments of <20% amino acid identity were intrinsically not considered as matches by the BLASTP program and were given an amino acid identity of 0%. Each query sequence produced two amino acid identity scores (from the two species to which it was compared) which were graphed using the two-dimensional scatterplot "smoothScatter" function in R v. 3.2.2 (R Foundation for Statistical Computing; http://www.r-proje ct.org/). Scatterplots were independently generated for each of the three secretomes and were overlaid for comparison using the "Z project" function in ImageJ v. 1.48.

| Phylogenetic tree analysis
The SSP family maximum likelihood phylogenetic tree was created in Geneious v. 6.1.5 using the CLUSTALW alignment tool and the PHYML plugin to build the tree from 1,000 bootstrap samples (Kearse et al., 2012).

| In silico characterization of SSPs
The 46 homologous SSPs identified in the C. camelliae secretome were used to screen the C. camelliae draft genome for additional family members using BLASTN. An alignment E value cut-off of ≤10 −5 identified 27 additional nonredundant SSP genes (Table S3). The full coding sequences of all 73 C. camelliae SSP genes were manually deduced from the draft genome (Table S3). Disulphide bond and connectivity predictions were performed using DISULPHIND v. 1.1 (Ceroni et al., 2006). MEME v. 4.9 (Bailey et al., 2006)

| B. cinerea virulence assays
Arabidopsis virulence assays were performed on 4-week-old plants.
Rosette leaves were inoculated with 5 µl of a 5 × 10 5 suspension of wild-type or mutant conidia in half-strength potato dextrose broth.
Lesion area was measured at 72 hr post-inoculation using ImageJ v. 1.48. N. benthamiana and S. lycopersicum assays were performed on detached leaves of 4-week-old plants. Leaves were placed in a Petri dish on moist filter paper and inoculated with 5 µl of a 5 × 10 5 suspension of wild-type or bcssp2 conidia. Lesion area was measured 48 hr post-inoculation.

ACK N OWLED G EM ENTS
Dr Jan van Kan for providing the Botrytis cinerea B05.10 and advice re-

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are openly available in GenBank at https ://www.ncbi.nlm.nih.gov/genba nk/, reference number PRJNA289037 (C. camelliae genome) and SRS2024035 (in planta RNA-Seq data).

S U PP O RTI N G I N FO R M ATI O N
Additional supporting information may be found online in the Supporting Information section.

FIGURE S1
The bioinformatic pipeline used for fungal secretome