Natural variation of root lesion nematode antagonism in the biocontrol fungus Clonostachys rosea and identification of biocontrol factors through genome‐wide association mapping

Abstract Biological control is a promising approach to reduce plant diseases caused by nematodes to ensure high productivity in agricultural production. Large‐scale analyses of genetic variation in fungal species used for biocontrol can generate knowledge regarding interaction mechanisms that can improve efficacy of biocontrol applications. In this study, we performed a genome‐wide association study (GWAS) for in vitro antagonism against the root lesion nematode Pratylenchus penetrans in 53 previously genome re‐sequenced strains of the biocontrol fungus Clonostachys rosea. Nematode mortality in C. rosea potato dextrose broth (PDB) culture filtrates was highly variable and showed continuous variation (p < .001) between strains, indicating a polygenic inheritance. Twenty‐one strains produced culture filtrates with higher (p ≤ .05) nematode mortality compared with the PDB control treatment, while ten strains lowered (p ≤ .05) the mortality. The difference in in vitro antagonism against P. penetrans correlated with antagonism against the soybean cyst nematode Heterodera glycines, indicating lack of host specificity in C. rosea. An empirical Bayesian multiple hypothesis testing approach identified 279 single nucleotide polymorphism markers significantly (local false sign rate < 10–10) associated with the trait. Genes present in the genomic regions associated with nematicidal activity included several membrane transporters, a chitinase and genes encoding proteins predicted to biosynthesize secondary metabolites. Gene deletion strains of the predicted nonribosomal peptide synthetase genes nps4 and nps5 were generated and showed increased (p ≤ .001) fungal growth and conidiation rates compared to the wild type. Deletion strains also exhibited reduced (p < .001) nematicidal activity and reduced (p ≤ .05) biocontrol efficacy against nematode root disease and against fusarium foot rot on wheat. In summary, we show that the GWAS approach can be used to identify biocontrol factors in C. rosea, specifically the putative nonribosomal peptide synthetases NPS4 and NPS5.

The genome of C. rosea strain IK726 was first sequenced using the short-read Illumina platform technology  and later on supplemented with the long-read PacBio platform . With the advancement of DNA sequencing technology, genome sequencing has become inexpensive and user-friendly and comparative genomics is now a regular practice in fungal molecular biology . Genome-wide association (GWA) is a powerful technique used to study genetic variants, to identify single nucleotide polymorphisms (SNPs) in different individuals and to pinpoint candidate genes that are associated with a trait of interest in the genome (Bush & Moore, 2012;Korte & Farlow, 2013).

K E Y W O R D S
antagonism, biocontrol, Clonostachys rosea, genome-wide association study, phytopathogenic nematodes, plant growth, wheat In this study, we used a GWA approach to gain insight into the genetic basis of nematode antagonism and biocontrol in C. rosea. We found 279 SNP markers associated with in vitro antagonism of the root lesion nematode P. penetrans, and genes positioned in the vicinity of these SNPs were predicted to encode several membrane transporters, a chitinase and secondary metabolite biosynthesis proteins (including NRPSs) with possible roles in nematode antagonism. By generating C. rosea deletion strains of the nps4 and nps5 NRPS genes, we show that NPS4 and NPS5 play a role in growth, conidiation and biocontrol of fusarium foot rot and nematode root disease of wheat.

| Fungal strains and maintenance conditions
Clonostachys rosea strains , mutants derived from the C. rosea strain IK726 wild type, B. cinerea strain B05.10, F. graminearum strain PH-1 and Rhizoctonia solani strain SA1 were revived from stock cultures stored in 20% (w/v) glycerol at −80°C and maintained on potato dextrose agar (PDA) medium (Oxoid) at 25°C in darkness.

| Antagonism assay using root lesion nematodes
An in vitro assay was performed to determine nematicidal activity in culture filtrates of 53 different C. rosea strains. Fifty millilitres potato dextrose broth (PDB; Oxoid) in 300 ml E-flasks were inoculated with 1 × 10 6 C. rosea conidia/flask and incubated at 25°C on a rotating shaker (120 rpm) under dark conditions. After five days, fungal biomass was separated from the broth by filtering through three layers of filter paper (Ahlstrom-Munksjö). The culture filtrate was further sterile-filtered through a 0.45 μm cellulose membrane (Sarstedt), as described previously (Iqbal, Dubey, McEwan, et al., 2018).
Pratylenchus penetrans nematodes were purchased from the laboratory of Nematology, Department of Plant Sciences, Wageningen University, Netherlands. Eight hundred microlitres of culture filtrate or fresh liquid media (control) was mixed with 200 μl of water containing approximately 100 P. penetrans nematodes in wells of 24-well plastic plates, followed by incubation at 25°C in darkness for 24 hr.
The number of dead nematodes after incubation was determined by washing the nematodes with cold water and by performing the touch assay under a light microscope (Wild M5A Heerbrugg) at 50× magnification (Gill, Olsen, Sampayo, & Lithgow, 2003). The nematode mortality assay was performed on five biological replicates for each C. rosea strain.

| Test of host specificity of antagonism
Based on the results from the in vitro study with P. penetrans, five strains with high antagonism and five with low antagonism were also tested against soybean cyst nematode (SCN; Heterodera glycines) to assess C. rosea host specificity. In addition to testing for nematicidal activity as with P. penetrans, pathogenicity towards SCN eggs, inhibition of egg hatching and juveniles was also investigated.
To assess pathogenicity towards eggs, cysts were extracted from roots of soybean plants grown in the greenhouse, surface-sterilized with 0.5% NaOCl for three min and placed on two per cent water agar as described previously (Chen, Dickson, & Mitchell, 1996).
To assess inhibition of egg hatching, SCN eggs were collected as described previously (Chen & Chen, 2002;Chen et al., 1996) and surface-sterilized with an aqueous solution containing 50 ppm streptomycin, 100 ppm chlortetracycline and 50 ppm Fungin. Following 6 hr of incubation at room temperature, the eggs were collected by centrifugation at 2,500 g RCF for ten min, rinsed with distilled water twice and stored at 4°C until further use.
Two millilitres C. rosea culture filtrates (described above) were supplied into wells of 24-well plastic plates, while fresh medium and water served as controls. One hundred microlitres of water containing 500 eggs was placed on 1-cm-diameter sieves with 35 μm mesh size so that the mesh surface just touched the filtrate solution (Chen et al., 1996). Following incubation at 25°C in darkness for 54 hr, the number of hatched nematodes was determined using an inverted microscope (Nikon Eclipse TS100).
Inhibition of egg hatching was calculated as described previously (Chen et al., 1996). Each fungal strain and control treatment was replicated three times.
Stage 2 (J2) juveniles of H. glycine were collected by placing eggs on three layers of coffee filter paper (Bunn-O-Matic Corporation) and kept on a sterile collecting box. Egg hatching was promoted by moistening the filter paper with 4 mM ZnCl 2 solution followed by incubation at 27°C for 48 hr. Eight hundred microlitres C. rosea culture filtrates (described above) or fresh liquid media (control) was mixed with 200 μl of water containing approximately 100 H. glycines J2 juveniles in wells of 24-well plastic plates, followed by incubation at 25°C in darkness for 24 hr. The number of dead nematodes was determined using an inverted microscope as described above on five biological replicates for each fungal strain.

| Sequence and phylogenetic analyses of nonribosomal peptide synthetases
Predicted NRPS proteins were retrieved from the C. rosea IK726 genome sequence Karlsson et al., 2015).
Identification of conserved protein modules and features was made using the conserved domain database (CDD) (Marchler-Bauer et al., 2017) and the simple modular architecture research tool (SMART) (Letunic & Bork, 2018). Signal peptides for targeting proteins to the secretory pathway were identified using SignalP ver. 4.1 (Nielsen, 2017).

| Gene expression analysis
For expression analysis of NRPS genes, C. rosea was confronted in a dual plate assay with B. cinerea (Cr-Bc), F. graminearum (Cr-Fg) or with itself (Cr-Cr) as described previously Iqbal et al., 2019). Seven to ten mm of the growing front of C. rosea was harvested at contact with B. cinerea or F. graminearum mycelia. Mycelium harvested at the same stage in a confrontation of C. rosea with itself served as the control treatment.
Each treatment was replicated five times.
RNA extraction was carried out using a Qiagen RNeasy kit according to the manufacturer's instruction (Qiagen). RNA was treated with RNase-free DNase I (Fermentas) and reverse transcribed (RT) in a total volume of 20 μl using the Maxima First-Strand cDNA Synthesis Kit (Fermentas) followed by 10-fold dilution and stored at −20°C. Specific PCR primers targeting the C. rosea NRPS genes nps4 and nps5 (Table S1) were designed using the PrimerSelect software implemented in the DNASTAR Lasergene ver. 10 software package (DNASTAR Inc.). Primer amplification efficiency was determined based on the amplification of serial dilutions of C. rosea genomic DNA. Transcript levels were quantified using an iQ5 qPCR system (Bio-Rad) using the SsoFast EvaGreen Supermix (Bio-Rad) as described previously (Kamou et al., 2016). A melt curve analysis was performed to confirm that the signal was the result of single product amplification. Relative expression levels were calculated in relation to the reference gene β-tubulin (tub) (Mamarabadi, Jensen, Jensen, & Lübeck, 2008) using the 2 -ΔΔCT method (Livak & Schmittgen, 2001) with five biological replicates, each based on two technical replicates.

| Construction of gene deletion cassette, transformation and mutant validation deletion
To study the function of nps4 and nps5, deletion strains were constructed. A three-fragment multisite gateway cloning system (Invitrogen) was used to construct nps4 and nps5 gene deletion cassettes. PCR amplification of ~2 kb of the 5′ flank and the 3′ flank regions of both nps4 and nps5 genes was performed using C. rosea genomic DNA and the gene-specific primer pairs NPS4-ups F/ups R, NPS4-ds F/ds-R, NPS5-ups F/ups R and NPS5-ds F/ds-R (Table S1), respectively, as shown in Figures S1 and S2. Gateway entry clones of the purified 5′ flank and 3′ flank PCR fragments were generated according to the manufacturer's instruction (Invitrogen). We used a gateway entry clone for the hygromycin-resistance cassette (hygB) that was constructed during our previous studies Dubey, Broberg, Sooriyaarachchi, et al., 2013;Dubey, Ubhayasekera, Sandgren, Jensen, & Karlsson, 2012).
The gateway LR recombination reaction was carried out using the entry plasmid of the respective fragments and the destination vector pPm43GW (Karimi, De Meyer, & Hilson, 2005) to generate a deletion cassette following the conditions as described by the manufacturer (Invitrogen).
PCR positive transformants were tested for mitotic stability and then purified by single spore isolation as described previously (Dubey et al., 2014. The length of the nps4 and nps5 genes prevented the construction of gene complementation cassettes, and therefore, all phenotypic analyses were performed with five independent positive transformants to ensure that any phenotypic effect was due to the nps4 and nps5 gene deletion and not due to ectopic insertions.
A dual plate confrontation assay was performed to determine the in vitro antagonistic ability of C. rosea strains against B. cinerea, F. graminearum and R. solani Iqbal et al., 2019).
In addition, a culture filtrate test was used to measure the in vitro antagonistic ability against B. cinerea, F. graminearum, R. solani and P. penetrans nematodes. Culture filtrates were prepared by inoculating 100 ml of PDB in 500 ml E-flasks with 2 × 10 6 C. rosea conidia/ flask, followed by incubation at 25°C in darkness on a rotating shaker (120 rpm) for 2 weeks (Dubey et al., 2014;Fatema et al., 2018;Iqbal et al., 2019). Antagonistic ability against other fungi was determined by measuring their mycelial biomass, while nematicidal activity was determined by counting the number of dead nematodes as described previously (Gill et al., 2003;Iqbal et al., 2019).
To examine the involvement of NPS4 and NPS5 in secondary metabolite production, the culture filtrates mentioned above were further analysed by ultra-high performance liquid chromatography-mass spectrometry (UHPLC-MS). UHPLC-MS analysis was performed using an Agilent 1290 Infinity II UHPLC (Agilent) hyphenated with a Bruker maXis Impact QTOF MS (Bruker Daltonic GmbH.). Samples (1 μl) were analysed on a reversed-phase UHPLC column (2.1 × 50 mm, 1.5 μm, Accucore Vanquish, Thermo Scientific), eluted with a gradient of acetonitrile (MeCN) in water, both with 0.2% formic acid (10%-95% MeCN in 3 min and 95% MeCN for 1.2 min at 0.9 ml/min). The MS was used in a positive mode in the range m/z 50-1,500. Compass Data Analysis 4.3 software (Bruker Daltonics) was used to calibrate mass spectra against sodium formate clusters, to create extracted-ion chromatograms corresponding to compounds known to be produced by Clonostachys and to convert data to mzXML format. Ion-chromatogram peak picking was conducted using the web-based platform XCMS Online (Tautenhahn, Boettcher, & Neumann, 2008). The in vitro antagonism assays and chemical analyses were performed on five biological replicates while mycelial biomass production measurements were carried out with three biological replicates.

| Assay for biological control of fusarium foot rot disease
An assay for biological control of fusarium foot rot on wheat (Triticum aestivum, winter wheat variety "Stava") was carried out using five biological replicates. Each replicate consisted of 12-15 plants, following a procedure described previously (Dubey et al., 2014Knudsen, Hockenhull, & Jensen, 1995). In short, surface-sterilized wheat seeds were coated with C. rosea wildtype, Δnps4 or Δnps5 deletion strain conidia (1 × 10 7 conidia/ ml) and were sown in moistened sand in individual plastic pots (5 × 5 × 5 cm). Water-soaked seeds were used as a control. The pathogen was inoculated by placing a 5-mm-diameter PDA plug with F. graminearum mycelium in a hole close to the seeds. A sterile PDA plug without any F. graminearum mycelium was used as a control. After inoculation, holes were covered with moist sand.
Pots were maintained in a growth cabinet under controlled conditions: that is light/dark photoperiod 12 hr/12 hr with 150 μmol/ m 2 s −1 light intensity, a temperature of 15°C ± 1°C and a relative humidity of 80 ± 5%. Seedlings were harvested three weeks postinoculation, and disease symptoms were scored based on a 0-4 scale as described previously (Dubey et al., 2014Knudsen et al., 1995).

| Assay for biological control of nematode root disease
An assay of nematode root disease in wheat was performed with eight biological replicates for each C. rosea strain, with the exception for strains 1,830 and CBS 188.33 where only three biological replicates were included. Naturally, nematode-infested soil containing a diverse community of nematodes was collected from an agricultural field (coordinates: 59°51'10.98"N, 17°40'37.344"E) close to Uppsala, Sweden. Soil was collected from five different locations during fallow conditions and prepared for experiment as described previously (Iqbal et al., 2019;Iqbal, Dubey, McEwan, et al., 2018). Formulations of C. rosea strains (five with high in vitro antagonism and five with low in vitro antagonism against P. penetrans) and C. rosea IK726 wild-type, Δnps4 and Δnps5 strains were prepared. Inoculum concentration was determined by counting colony-forming units (CFUs) and mixing inoculum with soil to reach a final concentration of 2.5 × 10 5 CFU/g of soil (Iqbal, Dubey, McEwan, et al., 2018;Jensen et al., 2000). Plastic pots (9 × 9 × 10 cm) were filled with 500 g of C. rosea-inoculated or uninoculated soil (control). Eight wheat seeds (cv. Stava) were planted in each pot (replicate) and placed in a growth cabinet under controlled conditions as described previously (Iqbal et al., 2019;Iqbal, Dubey, McEwan, et al., 2018). All pots were subjected to an equal amount of water throughout the 7-week experimental period. At the end of the experiment, plants were cut at the root-shoot junction at the surface of the soil, and plant dry shoot weight (g) and shoot length (cm) were recorded. Living nematodes were extracted from fifteen g of soil and one g of harvested roots using a modified Baermann funnel method (Baermann, 1917;Viketoft, Palmborg, Sohlenius, Huss-Danell, & Bengtsson, 2005). Nematodes were counted using a light microscope (Wild M5A Heerbrugg) and were classified to the genus level based on an analysis of morphological characters.

| Statistical analyses
Gene expression and phenotypic data were analysed by analysis of variance (ANOVA) using a general linear model approach implemented in Minitab® Statistical Software ver. 18.1 (Minitab Inc.). Pairwise comparisons were performed using Fisher's least significant difference or the Tukey-Kramer method at the 95% significance level. Statistical analysis of UHPLC-MS data was performed using MetaboAnalyst 4.0, a web-based tool suite for metabolomic data analysis (Chong et al., 2018). Following sample normalization by sum and log transformation, differences between sample groups (wild type and mutants) were analysed by principal component analysis (PCA) and t tests.

| Screening of C. rosea strains for in vitro antagonism against nematodes
After 24 hr of incubation, there were significant (p < .001) differences in P. penetrans nematode mortality in culture filtrates from the different C. rosea strains (Figure 1). Culture filtrates from 21 strains resulted in a significantly (p ≤ .05) higher nematode mortality (25%-66%) compared with the PDB medium control where 18% of the nematodes were deceased (Figure 1). Culture filtrates from ten strains showed significantly (p ≤ .05) lower nematode mortality levels (8%-12%) compared with the PDB control. Nematode mortality in culture filtrates from the remaining 23 strains was not significantly different from the PDB control ( Figure 1).
Culture filtrates of strains with a high in vitro antagonism against P. penetrans also inhibited hatching of SCN eggs (p ≤ .05) compared with the PDB control treatment (Figure 2b). In contrast, strains with a low in vitro antagonism against P. penetrans were not able to inhibit egg hatching. In fact, strains CBS 188.33, 2177 and 1882 increased (p ≤ .05) hatch rate by 42%-69% compared with the PDB control ( Figure 2b). Egg hatching in the PDB control was 55% lower (p ≤ .05) compared with egg hatching in water only.
Culture filtrates of C. rosea strains with a high in vitro antagonism against P. penetrans also resulted in a higher (p ≤ .05) mortality of SCN juveniles, compared with the PDB control treatment (Figure 2c). This was not the case in culture filtrates from strains with a low in vitro antagonism against P. penetrans, where no differences in mortality compared with the control were detected.

| Correlation between in vitro antagonism and biocontrol of nematode root disease
The five C. rosea strains with high in vitro antagonism against P. penetrans and the five strains with low in vitro antagonism indicated above were further evaluated for their ability to control nematode root disease on wheat using a pot experiment.
All C. rosea strains, except 1701 and 1882, were able to reduce (p ≤ .05) the number of nematodes belonging to plant-parasitic F I G U R E 1 Mortality of Pratylenchus penetrans nematodes in potato dextrose broth culture filtrates from 53 Clonostachys rosea strains. Nematode mortality was assessed in culture filtrates from C. rosea strains and in a control medium (black bar) after 24 hr of incubation at 25°C. Error bars represent the standard deviation based on five biological replicates. Asterisks indicate a significant difference (p ≤ .05) between a C. rosea culture filtrate and the potato dextrose broth control treatment, as determined by Fisher's least significant difference genera (Pratylenchus, Helicotylenchus, Tylenchorhynchus, Merlinus, Boleodorus, Heterodera and Paratylenchus) recovered from soil ( Figure 3a). The average number of plant-parasitic nematodes recovered from pots inoculated with strains showing high in vitro antagonism was 107 ± 33 compared with 169 ± 51 from pots inoculated with strains showing low in vitro antagonism. This difference was significant at p = .052.
All C. rosea strains, except CMI192798, 2177 and 1882, were effective in reducing (p ≤ .05) Pratylenchus and Heterodera nematodes recovered from wheat roots (Figure 3b). The difference between strains with high and low in vitro antagonistic ability (19 ± 6 versus 27 ± 7 nematodes per pot) was not significant (p = .084).
Soil inoculation with C. rosea strains SDT-5-1, 1830, CBS 907.72G, CBS 100000 and CBS 188.33 increased (p ≤ .05) dry shoot weight of the wheat plants by on average 31%, compared with the control treatment ( Figure 3c). However, dry shoot weight of wheat plants did not differ (p = .110) between C. rosea strains with high and low in vitro antagonistic ability.

| Identification of C. rosea SNPs associated with in vitro antagonism to nematodes
The GWA analysis of the 53 C. rosea strains and empirical Bayesian multiple hypothesis correction identified a total of 279 SNPs that were associated with in vitro antagonism against P. penetrans at lfsr ≤ 1 × 10 -10 . Out of these, 193 SNPs were located in intergenic regions, 29 in exonic regions, five in intronic regions and seven in the 5′ or 3′ untranslated regions (Table S2)  . One SNP was located 21 bp upstream of CRV2G00012293, predicted to encode the CHIA4 chitinase . SNPs unitig_1:417,862 and unitig_306:5,647 were associated with the predicted fungal-type transcription factors CRV2G00013596 and CRV2G00021935, respectively. Three SNPs were associated with a predicted TfdA-family oxidoreductase (CRV2G00021724), a predicted decarboxylase (CRV2G00022270) and a predicted cytochrome P450 (CRV2G00019702), respectively, all implicated in biosynthesis of secondary metabolites in other fungal species. SNP unitig_347:5,746 was located in an exon of CRV2G00021286, predicted to encode a NRPS. By relaxing the lfsr cut-off for significance, two additional SNPs associated with predicted NRPSs were identified, unitig_346:7,460 located 1,009 bp from CRV2G00021335 (lfsr = 0.002) and the intronic unitig_246:8,741 in CRV2G00021642 (lfsr = 0.017; Table S2).

| Sequence analyses of NRPS genes
Comparisons between the C. rosea IK726 genome ver. 1  and ver. 2  revealed that CRV2G00021286 and CRV2G00021335 represented partial fragments of the predicted NPS5 NRPS .
CRV2G00021642 represented a partial fragment of the predicted NPS4 NRPS . The open reading frames of

| Gene expression analyses of nps4 and nps5
The primer pair NPS4 F_val, NPS4 R_val and NPS5 F_val, NPS5 R_val (Table S1) specifically amplified a 149 bp and 179 bp amplicon from nps4 and nps5, respectively. The relative expression for both nps genes was twofold higher during the confrontation between C. rosea and B. cinerea (p ≤ .01), but not with F. graminearum, compared with the self-confronting control treatment (Figure 5a,b).

| Generation and validation of nps4 and nps5 deletion strains
The nps4 and nps5 genes were replaced with the hygB selection cassette by homologous recombination using ATMT in C. rosea IK726.
In total, 150 hygromycin-resistant C. rosea colonies were obtained

| Phenotypic analyses
The growth rate of all C. rosea nps4 and nps5 deletion strains was on average 25% and 23% higher (p ≤ .001), respectively, than the wildtype growth rate when grown on PDA plates (Figure 6a). Conidiation of nps4 and nps5 deletion strains was on average 156% and 182% higher (p ≤ .001), respectively, on PDA plates (Figure 6b).
Mycelial biomass of B. cinerea, F. graminearum and R. solani was significantly (p ≤ .019) increased (47%-71%) in culture filtrates from C. rosea Δnps4 and Δnps5 strains compared with filtrates from the wild-type strain (Figure 6c-e). On solid medium, growth rates of B. cinerea, F. graminearum and R. solani did not differ during interactions with C. rosea wild-type, Δnps4 or Δnps5 strains, at either three or five DPI ( Figure S3).
Nematicidal activity of culture filtrates from Δnps4 and Δnps5 strains was significantly weaker (p < .001) compared with the wild type; 24 hr of incubation in culture filtrates from nps4 or nps5 deletion strains resulted in 22% and 23% mortality, respectively, of P. penetrans nematodes compared with 33% mortality in wild-type culture filtrates (Figure 6f). However, nematode mortality in the PDB control was even lower (12%, p ≤ .05) compared with either C. rosea strain (Figure 6f).

The metabolite profiles obtained by UHPLC-MS were different
for the wild-type cultures compared to cultures of nps4 and nps5 deletion strains ( Figure S4), as shown by the clustering obtained F I G U R E 2 Effect of Clonostachys rosea strains on soybean cyst nematode, including egg parasitism, egg hatch inhibition and juvenile mortality. (a) Egg-parasitic index (EPI) was determined on water agar plates and calculated based on a previously established scale (Chen et al., 1996). (b) The number of soybean cyst nematode eggs hatched in C. rosea potato dextrose broth (PDB) culture filtrates. Egg hatching inhibition was assessed after 54 hr of incubation at 25°C. (c) Soybean cyst nematode juveniles (J2) mortality in C. rosea PDB culture filtrates. Juvenile mortality was assessed after 24 hr of incubation at 25°C. Error bars represent the standard deviation based on three biological replicates in the case of soybean cyst nematode EPI (a) and eggs hatched (b) and five biological replicates in case of juvenile mortality (c). Different letters indicate a statistically significant difference (p ≤ .05) between treatments within an experiment as determined by the Tukey-Kramer test. Dark and light grey columns correspond to high and low antagonism strains, respectively, while black column corresponds to control treatment with PCA. However, a more detailed comparison between nps4 or nps5 deletion strains and wild-type culture filtrates failed to identify any compounds that were produced exclusively in the wild-type cultures. One unknown compound, with retention time (t R ) 12 s and m/z 244.118 ([M + H] + ), corresponding to the molecular formula (C 11 H 17 NO 5 ), was produced at significantly higher levels in wildtype cultures than in any of the mutant strains (wild type/nps4: 7.4, p < .001; wild type/nps5 = 5.3, p < .001). The molecular formula and MSMS data do not fit any molecule in the database Dictionary of Natural Products.

| Deletion of nps4 and nps5 reduces biocontrol efficacy against fusarium foot rot disease
Coating of wheat seeds with wild-type C. rosea conidia provided adequate protection (p ≤ .05) against fusarium foot rot disease caused by F. graminearum (Figure 7a). Seed coating with conidia from nps4 or nps5 deletion strains also protected against foot rot (p ≤ .05), but to a significantly (p ≤ .05) lower degree compared with the wild type ( Figure 7a). All C. rosea strains (wild type, Δnps4 or Δnps5) increased (p ≤ .05) shoot weight of wheat with on average 75% compared with the fusarium foot rot control treatment ( Figure 7b).

| Deletion of nps4 and nps5 reduces biocontrol efficacy against nematode root disease
Seven weeks after inoculating the soil with C. rosea, the number of nematodes extracted was 32% (wild type), 19% (∆nps4) and 20% (∆nps5) lower, respectively, compared to nematode numbers extracted from the control treatment (p < .001; Figure 7c). Soil application of nps4 and nps5 deletion strains reduced the number of nematodes to a significantly (p < .001) lesser degree (20%) than that of the wild-type treatment (Figure 7c).
Moreover, inoculation of C. rosea wild type also significantly (p < .001) reduced the number of nematodes (Pratylenchus and Heterodera J2) recovered from roots of wheat by 40% compared with the control treatment ( Figure 7d). However, adding nps4 and nps5 deletion strains to soil reduced the number of extracted nematodes by only 23% and 20%, respectively, from wheat roots compared with the control treatment, and to a significantly (p < .001) lower degree (29%) than the wild-type treatment (Figure 7d).

F I G U R E 3
Effect of soil application of Clonostachys rosea strains on nematode populations and shoot weight of wheat. (a) Effect of C. rosea strains on nematode populations recovered from soil including plant-parasitic nematode genera (Pratylenchus,Helicotylenchus,Tylenchorhynchus,Merlinus,Boleodorus,Heterodera (J2) and Paratylenchus) and other trophic groups (bacterivorous, fungivorous and omnivorous). (b) Effect of C. rosea strains on nematode populations recovered from wheat roots. The data include both migratory endoparasitic nematodes from the genus Pratylenchus and endoparasitic nematodes from the genus Heterodera (J2). The control treatment corresponds to plants grown in naturally nematode-infested soil without C. rosea application. (c) Effect of soil application of C. rosea on dry shoot weight (g) of plants grown in soil treated with C. rosea or in soil that was not treated with a C. rosea application (control). Error bars represent the standard deviation based on eight biological replicates, with the exception for strains 1,830 and CBS 188.33 where only three biological replicates were scored. Different letters indicate a statistically significant difference (p ≤ .05) between treatments as determined by Fisher's least significant difference in (a) and by the Tukey-Kramer test in (b) and (c). Dark and light grey columns correspond to high and low antagonism strains, respectively, while black column corresponds to control treatment The inoculation of either C. rosea wild-type or nps4 and nps5 deletion strains in soil significantly increased the shoot length (30%) and dry shoot weight (59%) compared to the control treatment, but there were no differences among the C. rosea treatments (Figure 7e, f). Furthermore, control plants displayed yellowing of leaves and chlorosis but this was not observed in plants grown in soil treated with either C. rosea wild-type or nps4 and nps5 deletion strains.

| D ISCUSS I ON
In the present study, we explored the natural variation in nematode in vitro antagonism among 53 strains of C. rosea to identify potential factors that influence the outcome of biocontrol interactions in C. rosea. We found a significant variation between C. rosea strains in terms of in vitro antagonism against the root lesion nematode P. penetrans, suggesting quantitative and/or qualitative differences in their ability to produce secreted nematicidal compounds or enzymes. Meyer et al. (2004) (Meyer et al., 2004).
Previously, it was shown that C. rosea strain IA produced an epipolysulfanyldioxopiperazine compound with nematicidal activity against Caenorhabditis elegans and Panagrellus redivivus when grown on wheat kernels for 20 days (Dong et al., 2005). Culture filtrates from C. rosea strain IK726 grown in PDB exhibited nematicidal activity against a mixed community of plant-parasitic nematodes (Iqbal, Dubey, McEwan, et al., 2018). Additionally, C. rosea strain 611 grown on PDB produced the extracellular serine protease PrC with nematicidal activity against P. redivivus (Li, Yang, Huang, & Zhang, 2006), showing that not only compounds, but also enzymes, are involved in C. rosea antagonism against nematodes.
An unexpected observation was that the culture filtrates from certain C. rosea strains resulted in lower P. penetrans mortality, but to some extent higher H. glycines egg hatching, compared with the PDB control treatment. A similar stimulatory effect on H. glycines and M. incognita egg hatching was reported previously from a range of different fungi (Meyer et al., 2004) and explained by either F I G U R E 4 Maximum likelihood phylogenetic tree including Clonostachys rosea NPS4 and NPS5 within the context of previously identified clades of NRPSs in fungi (Bushley & Turgeon, 2010). Major groups are shown as collapsed clade including outgroup adenylating enzymes (brown) NPS11/12/ETP (dark blue), cyclosporin synthases (turquoise), ACV synthases (dark green), fungal PKS-NRPS hybrids (light green), ChNPS10 (yellow), alpha-amino adipates (pink), siderophore synthases (red) and the Euascomycete clade (purple box). Both Cr NPS4 and Cr NPS5 fell within the euascomycete clade (purple box), which contained primarily multimodular NRPSs that are expanded in euascomycete fungi. Within the euascomycete clade, the eleven A domains of NPS4 grouped within peptaibol synthases (dark blue branches with C. rosea NPS4 A domains highlighted), suggesting this NRPS may produce an eleven amino acid peptaibol. Three of the C. rosea NPS5 A domains grouped within the ChNPS8/peramine synthase clade (light green with C. rosea NPS5 A domains highlighted). The remaining two A domains of this gene grouped adjacent to the second A domain of PS1 and PS4, two genes involved in the biosynthesis of ergot alkaloids (orange with C. rosea NPS5 A domain highlighted) F I G U R E 5 Expression analyses of nps4 and nps5 genes during interaction with fungal prey. Gene expression of (a) nps4 and (b) nps5 was estimated using RT-qPCR during the confrontation of Clonostachys rosea with Botrytis cinerea (Cr-Bc), C. rosea with Fusarium graminearum (Cr-Fg) and C. rosea with itself (Cr-Cr, control). Relative expression was calculated as the ratio between the nps4 and nps5 genes and tubulin using the 2 -ΔΔCT method. Gene expression during the interaction of Cr-Bc was significantly higher (p ≤ .05) than that during the interaction of Cr-Cr or Cr-Fg, as determined by the Tukey-Kramer test. Error bars represent the standard deviation based on five biological replicates production of specific compounds stimulating egg hatch or fungal utilization of PDB compounds inhibiting egg hatch (Chen, Dickson, & Mitchell, 2000;Meyer et al., 2004). The previously mentioned prC protease gene in C. rosea is induced by the pH-responsive PacC transcription factor, and pacC itself is repressed by low (3-4) pH (Zou, Tu, et al., 2010). As C. rosea can lower the pH of the growth medium (Fatema et al., 2018), we speculate that low nematode mortality in certain C. rosea culture filtrates may be connected with variation in the pH response, although this hypothesis needs to be further investigated. However, the pH itself is reported not to influence egg hatching of H. glycines (Pike et al., 2002) or a direct cause of antagonism against M. incognita (Djian, Pijarowski, Ponchet, Arpin, & Favre-Bonvin, 1991).
Interestingly, egg parasitism also correlated with production of nematicidal compounds or enzymes, suggesting that antibiosis and/ or enzymatic digestion of the egg shell may contribute to direct parasitism. The result from the current study further confirms the ability of C. rosea to directly parasitize nematodes reported previously by Zhang et al. (2008). They showed that pathogenesis starts from attachment of conidia to the nematode cuticle, followed by penetration into the nematode leading to death and degradation of the nematode.
Although high in vitro antagonism was observed in our study, the correlation with biocontrol activity in soil was weak. Although C. rosea strains with high in vitro antagonism against P. penetrans performed better (at p = .052) in reducing plant-parasitic nematode numbers in soil compared with strains with low in vitro antagonism, no significant differences between the groups showing low and high in vitro antagonism were observed in terms of the nematode numbers in roots or in plant biomass. This is not surprising given the complexity of both biological control mechanisms, including not only antibiosis, but also competition for nutrients and space, direct parasitism and induction of plant defence reactions (Harman, Howell, Viterbo, Chet, & Lorito, 2004;Jensen, Karlsson, & Lindahl, 2017), and natural soil environments. However, most C. rosea strains were able to control the populations of plant-parasitic nematodes in the potted plant assay. This decline in nematode populations positively correlated with improved growth of wheat plants and lack of disease symptoms, such as chlorosis and yellowing of leaves, which is usually associated with root damage by nematodes (Jones et al., 2013).
These results are in agreement with previous findings that showed that inoculation of C. rosea strain IK726 into soil reduced the number of plant-parasitic nematodes by 40% -70% and resulted in better growth of carrot and wheat (Iqbal, Dubey, McEwan, et al., 2018).
The significant and continuous variation in nematicidal activity of C. rosea culture filtrates against P. penetrans suggested polygenic inheritance of nematode antagonism and that a GWA of the trait is possible. Using an empirical Bayes large-scale hypothesis testing approach (Stephens, 2016), we were able to identify 279 SNP markers that were significantly associated with in vitro nematode antagonism in C. rosea. Production of, and tolerance to, secondary metabolites is one important feature of antagonism and biocontrol in C. rosea (Demissie, Foote, Tan, & Loewen, 2018;Dubey et al., 2014Dubey et al., , 2016Fatema et al., 2018;Iqbal et al., 2019;Karlsson et al., 2015;Nygren et al., 2018). Therefore, genes predicted to encode proteins involved in secondary metabolite biosynthesis and transport that were physically located close to SNPs segregating with nematicidal activity were considered as candidate biocontrol factors. MFS transporters in family 2.A.1.2.33 have unknown function, and no member of this family has been characterized in detail. Members of the 2.A.1.14.11 family have been shown to be involved in nicotinic acid import across the plasma membrane in Saccharomyces cerevisiae (Llorente & Dujon, 2000), while a member of MFS transporter family 2.A.1.2.85 was shown to be localized in the peroxisomal membrane and is required for transport of isopenicillin N, a precursor of penicillin antibiotic biosynthesis, in the fungus Penicillium chrysogenum (Fernández-Aguado, Ullán, Teijeira, Rodríguez-Castro, & Martín, 2013). Interestingly, all three families evolve under selection for increased paralog gene numbers in C. rosea (Nygren et al., 2018), suggesting that these MFS transporters contribute to ecological niche adaptation (Wapinski, Pfeffer, Friedman, & Regev, 2007). In addition, one SNP was identified in the ABC transporter gene abcb14 (CRV2G00020959) predicted to be involved in multidrug resistance . Several SNPs were associated with genes predicted to be involved in biosynthesis of secondary metabolites in other fungal species; CRV2G00013596 was predicted to be a fungal-type zinc cluster transcription factor with sequence similarity towards the Alternaria cinerariae DHC4 regulator of dehydrocurvularin biosynthesis (Cochrane et al., 2015), CRV2G00022270 showed sequence similarity with a decarboxylase involved in biosynthesis of yanuthone D in Aspergillus niger that acts as an antibiotic against fungi and bacteria (Holm et al., 2014), and CRV2G00019702 had sequence similarity towards the FTMP450-2 fumitremorgin C synthetase involved in alkaloid biosynthesis in A. fumigatus (Kato et al., 2009;Maiya, Grundmann, Li, & Turner, 2006). Involvement of NRPSs in the biosynthesis of nematicidal compounds has been suggested previously in C. rosea (Dong et al., 2005;Iqbal et al., 2019;Karlsson et al., 2015). Therefore, the identified F I G U R E 6 In vitro phenotypic analyses of Clonostachys rosea wild-type, nps4 and nps5 deletion strains. (a, b) show growth and conidiation rates of C. rosea wild-type, nps4 and nps5 deletion strains on potato dextrose agar (PDA). C. rosea wild-type, nps4 and nps5 deletion strains were inoculated on PDA plates and incubated at 25°C. (a) Growth rate was measured daily up to seven days postinoculation (DPI). (b) Conidiation rate was calculated at 14 DPI. (c-f) show antagonistic tests in culture filtrates of C. rosea wild-type, nps4 and nps5 deletion strains against Botrytis cinerea (c), Fusarium graminearum (d), Rhizoctonia solani (e) and the migratory endoparasitic nematode Pratylenchus penetrans in potato dextrose broth. An agar plug of B. cinerea, F. graminearum and R. solani was inoculated in culture filtrates produced by C. rosea wild-type, nps4 and nps5 deletion strains and incubated at 25°C. Mycelia biomass production was determined by weighing the mycelial dry biomass at five DPI. Nematode mortality was assessed in culture filtrates from C. rosea wild-type, Δnps4 and Δnps5 strains and in a control medium after 24 hr of incubation at 25°C. Error bars represent the standard deviation based on five biological replicates. Different letters indicate statistically significant differences (p ≤ .05) between treatments within an experiment based on the Tukey-Kramer test. Light and dark columns correspond to nps4 and nps5 deletion strains, respectively, while black column corresponds to control or wild-type treatment nps4 and nps5 genes were studied in more detail. The predicted modular structures of NPS4 and NPS5 contained all the essential domains, that is A, T and C, which are necessary for a functional NRPSs to carry out peptide synthesis (Conti, Stachelhaus, Marahiel, & Brick, 1997;Keating, Marshall, & Walsh, 2000;Keating et al., 2002;May et al., 2002;Stachelhaus, Mootz, & Marahiel, 1999;Weber, Baumgartner, Renner, Marahiel, & Holak, 2000). The domain structure and phylogenetic placement of NPS4 strongly suggests that it is a peptaibol synthase of eleven A-T-C modules. Peptaibols have previously been shown to be produced by various fungi, including C. rosea (Rodríguez et al., 2011), species of Trichoderma (Degenkolb et al., 2012;Wiest et al., 2002), as well as the potential nematode parasite Purpureocillium lilacinum (Wang et al., 2016), the insect parasite Tolypocladium geodes (Tsantrizos, Pischos, & Sauriol, 1996) and the related truffle parasite T. ophioglossoides (Quandt, Di, Elser, Jaiswal, & Spatafora, 2016). They have canonically been considered to have roles in mycoparasitism or antifungal activity (Schirmböck et al., 1994), and some studies have demonstrated activity against the oomycete pathogen Phytopthora spp. (Sansom, Balaram, & Karle, 1993). Peptaibols are amphipathic compounds, and their antimicrobial function results from formation of ion channels in cell membranes (Sansom et al., 1993). In this study, the increased expression of nps4 in confrontation with the fungal pathogen B. cinerea and reduced disease severity in pot assays with the fungal root rot pathogen F. graminearum confirm a potential role for peptaibols in antagonism and biocontrol of fungal pathogens. While peptaibols have not previously been shown to be involved in activity against nematodes, the presence of large peptaibol synthases in C. rosea IK726 and the nematode egg-parasitic fungus P. lilacinum, coupled with the decreased antagonism towards nematodes observed in vitro in this study, suggests they should be further investigated for activity against plant-parasitic nematodes.
Phylogenetic analysis also placed NPS5 within the euascomycete clade, and three A domains grouped with NPS8 from the corn pathogen C. heterostrophus and peramine synthase. Peramine synthase was first identified in symbiotic Epichlӧe/Neotyphodium endophytes of grasses and has been shown to produce a pyrrolopyrazine compound that has activity against herbivorous insects (Tanaka, Tapper, Popay, Parker, & Scott, 2005). The other two A domains of NPS5 group adjacent to a clade containing the second A domain of two NRPSs (PS1 and PS4) that are involved in the biosynthesis of ergot alkaloids. Like peramine, ergot alkaloids have bioactivity against insects and animals (Cross, 2003), and both their evolutionary history (Spatafora, Sung, Sung, Hywel-Jones, & White, 2007;Torres, Singh, Vorsa, & White, 2008) and experimental evidence (Panaccione et al., 2006;Potter, Tyler Stokes, Redmond, Schardl, & Panaccione, 2008) suggest they may also be involved in protection against herbivory. However, while the presence of Neotyphodium endophytic fungi has been shown to reduce populations of the root lesion nematode, P. scribneri, this effect could not be attributed to ergot alkaloids as genetic knockout strains of ergot alkaloid NRPSs of the endophyte did not lower nematode colonization in planta and only the ergot alkaloid compound ergovaline had any inhibitory effect on nematode mobility (Panaccione et al., 2006). While it is challenging to predict the chemical class of compounds produced by NPS5, it is interesting that it shares an evolutionary relationship with NRPSs producing compounds with activity against other invertebrate pests such as insects.
Higher expression of both nps4 and nps5 during a confrontation with B. cinerea, but not against F. graminearum, suggests a role of both NRPSs during mycoparasitism of B. cinerea. The C. rosea NRPS genes nps1 and nps13 are also induced specifically during interaction with B. cinerea, but not with F. graminearum (Iqbal et al., 2019;Nygren et al., 2018), suggesting that only interaction with specific antagonistic fungi may trigger expression. NRPS gene expression is also reported to be induced in the truffle-parasitizing fungus T. ophioglossoides where the NRPS-like gene TOPH 03459 was highly upregulated in response to secreted material from the truffle host Elaphomyces muricatus (Quandt et al., 2016).
Our result indicates that the metabolites produced by NPS4 and NPS5 may be toxic for the fungus itself and exhibit an inhibitory effect on C. rosea, as deletion of nps4 and nps5 resulted in increased growth and conidiation. Alternatively, deletion of nps4 and nps5 may lead to redistribution of resources, from secondary metabolite biosynthesis to growth and conidiation (Calvo, Wilson, Bok, & Keller, 2002), that is a trade-off between allocating resources to growth versus secondary metabolite production. Our results are in agreement with previous finding that deletion of the nps1 NPRS gene in C. rosea strain also increased growth rate and conidiation (Iqbal et al., 2019). Similar results were also reported in the mycoparasitic fungus T. virens where deletion strains of NRPS genes, that is tex7, tex8, tex10 and tex13, resulted in increased growth rate compared with the wild type (Mukherjee, Buensanteai, Moran-Diez, Druzhinina, and Kenerley (2012); Mukherjee et al., 2018;. Our results are also in line with previous findings where deletion of the polyketide synthetase genes pks22 and pks29 in C. rosea resulted in F I G U R E 7 Effect of Clonostachys rosea wild-type, nps4 and nps5 deletion strains on biocontrol of fusarium foot rot, nematode root disease and plant growth. (a) In vivo bioassay to test the biocontrol efficacy of C. rosea wild-type, nps4 and nps5 deletion strains against Fusarium graminearum foot rot disease on wheat. Wheat seeds were coated with C. rosea conidia (wild-type, Δnps4 or Δnps5 strains) and planted in moist sand together with an F. graminearum agar plug. Seedlings were harvested three weeks postinoculation, and disease symptoms were scored on a 0-4 scale. (b) Effect of C. rosea wild-type, nps4 and nps5 deletion strains on dry shoot weight (g) of plants treated with C. rosea strains or plants that were not treated with any C. rosea strain (control). (c) Effect of soil applications of C. rosea wildtype, Δnps4 and Δnps5 strains on the nematode population recovered from soil, including plant-parasitic nematode genera (Boleodorus, Helicotylenchus, Heterodera (J2), Merlinus, Paratylenchus, Pratylenchus, Rotylenchus and Tylenchorhynchus) and other trophic groups (bacterivorous, fungivorous and omnivorous). (d) Effect of C. rosea wild-type, nps4 and nps5 deletion strains on the nematode population recovered from wheat roots. Data include both migratory endoparasitic nematodes belong to genus Pratylenchus and endoparasitic nematodes from genus Heterodera (J2). The control treatment refers to plants grown in soil without C. rosea application. (e and f) show effect of soil applications of C. rosea wild-type, nps4 and nps5 deletion strains on shoot length and shoot weight of wheat. (e) Average shoot length (cm) and (f) average dry shoot weight (g) of plants grown in soil treated with C. rosea wild-type, nps4 and nps5 deletion strains. The control treatment refers to plants grown in soil without C. rosea application. For a and b, error bars represent the standard deviation based on five biological replicates and each replicate contained the mean of 12-15 plants while c-f error bars represent the standard deviation based on eight biological replicates. Different letters indicate statistically significant differences (p ≤ .05) between treatments within an experiment as determined by Fisher's least significant difference or Tukey-Kramer test. Light and dark columns correspond to nps4 and nps5 deletion strains, respectively, while black columns correspond to control or wild-type treatment increased conidiation, indicating that secondary metabolites may regulate growth and development in C. rosea (Fatema et al., 2018). The significant increase in B. cinerea, F. graminearum and R. solani mycelial biomass and reduced nematode mortality when grown in culture filtrates of Δnps4 and Δnps5 strains compared with the wild type suggest that NPS4 and NPS5 biosynthesize metabolites with broad-spectrum antagonistic properties against these organisms. Previously, nine verticillin-type epipolysulfanyldioxopiperazine (gliocladin A-E, verticillin A, 11-deoxyverticillin A, Sch52900 and Sch52901) compounds were isolated from C. rosea strain 1A and all compounds showed toxicity against C. elegans and P. redivivus nematodes and more than 50% mortality rate was observed following 24 hr of incubation (Dong et al., 2005). Disruption of the hybrid PKS/NRPS zeamine (zmn) from the gram-negative bacterium Serratia plymuthica strain A153 resulted in attenuation of the nematicidal activity against C. elegans (Hellberg, Matilla, & Salmond, 2015). Culture filtrates from the Δnps4 and Δnps5 strains still exhibited a certain degree of nematicidal activity that might be due to the production of auxiliary compounds or enzymes, such as serine proteases, contributing to the nematicidal activity. As mentioned before, C. rosea strain 611 produced the extracellular serine protease PrC during growth on PDB that caused 80% mortality of P. redivivus during incubation for 48 hr . NPS4 and NPS5 are predicted to contain eleven and five NRPS modules, respectively, suggesting the corresponding nonribosomal peptides to be constructed from eleven and five building blocks, respectively. When comparing wild-type cultures with nps4 and nps5 deletion strains, the corresponding peptides should be absent in the mutant strains, but present in wild-type cultures. However, no compound was found to be exclusively produced in wild type and not present in the nsp4 or nsp5 deletion strains. The compound with molecular formula C 11 H 17 NO 5 at t R 12 s was found at significantly higher concentrations in wild-type cultures compared to mutant cultures, but the formula does not support a peptide type of structure, and it was present in all mutant strains. There are several possible explanations to why the expected peptides were not observed in the wild-type cultures, for example that nsp4 and nps5 may not be highly expressed during normal laboratory conditions but only during biotic or abiotic stress. This is supported by gene expression data, which showed nsp4 and nsp5 to be upregulated during interaction with B. cinerea. The peptides may also be produced at very low levels and thus escape detection, and this could be the case if only very low concentrations of the putative peptides are needed to obtain the desired effects on an interacting organism such as B. cinerea or nematodes.
NPS4 and NPS5 are required for full biocontrol ability of C. rosea to control foot rot disease caused by F. graminearum and nematode root disease in wheat. Together with previous result where deletion of the nps1 NRPS gene in C. rosea attenuated the antagonistic and biocontrol potential of C. rosea (Iqbal et al., 2019), this emphasizes the importance of NRPSs for the biocontrol ability of C. rosea.
However, soil inoculation of the nps4 and nps5 deletion strains still resulted in increased shoot length and shoot weight of wheat plants in accordance with previous work (Iqbal, Dubey, McEwan, et al., 2018), possibly indicating that the plant growth-promoting effect in C. rosea (Ravnskov et al., 2006;Roberti et al., 2008) is mechanistically separated from biocontrol of pathogens.
In the present study, we demonstrated that intrinsic differences exist among C. rosea strains in terms of in vitro antagonism against P. penetrans nematodes. We further showed that these strain differences were also reflected in antagonism against H. glycines, suggesting a broad-spectrum mechanism. We also showed that GWA can be used to identify candidate genes for biocontrol-related phenotypes, such as in vitro antagonism, reduced disease symptoms and reduced pathogen populations. Finally, we showed that the nps4 and nps5 gene products partially contribute to in vitro antagonism and biocontrol in C. rosea.

ACK N OWLED G EM ENTS
The authors thank the Swedish Research Council for Environment, Agricultural Sciences and Spatial Planning (FORMAS, grant numbers 220-2014(FORMAS, grant numbers 220- -289 and 942-2015(FORMAS, grant numbers 220- -1128, the Plant Protection Platform and the SLU Centre for Biological Control (CBC) at the Swedish University of Agricultural Sciences for financially supporting this study. We thank Annika Gustafsson for preparing the C. rosea formulations.

CO N FLI C T O F I NTE R E S T
None declared.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data generated or analysed during this study are included in this manuscript and its supporting files.