Host infection by the grass-symbiotic fungus Epichloë festucae requires catalytically active H3K9 and H3K36 methyltransferases

Recent studies have identified key genes in Epichloë festucae that control the symbiotic interaction of this filamentous fungus with its grass host. Here we report on the identification of specific fungal genes that determine its ability to infect and colonize the host. Deletion of setB, which encodes a homolog of the H3K36 histone methyltransferase Set2/KMT3, specifically reduced histone H3K36 trimethylation and led to severe defects in colony growth and hyphal development. The E. festucae ΔclrD mutant, which lacks the gene encoding the homolog of the H3K9 methyltransferase KMT1, displays similar developmental defects. Both mutants are completely defective in their ability to infect the host grass, and mutational studies of key residues in the catalytic SET domains from these proteins show that these phenotypes are dependent on the methyltransferase activities of SetB and ClrD. A comparison of the differences in the host transcriptome between seedlings inoculated with wild-type versus mutants suggests that the inability of these mutants to infect the host was not due to an aberrant host defense response. Co-inoculation of either ΔsetB or ΔclrD with the wild-type strain enables these mutants to colonize the host. However, successful colonization by the mutants resulted in death or stunting of the host plant. Transcriptome analysis at the early infection stage identified four fungal candidate genes, three of which encode small-secreted proteins, that are differentially regulated in these mutants compared to wild-type. Deletion of crbA, which encodes a putative carbohydrate binding protein, resulted in significantly reduced host infection rates by E. festucae. Author Summary The filamentous fungus Epichloë festucae is an endophyte that forms highly regulated symbiotic interactions with the perennial ryegrass. Proper maintenance of such interactions is known to involve several signalling pathways, but much less is understood about the infection capability of this fungus in the host. In this study, we uncovered two epigenetic marks and their respective histone methyltransferases that are required for E. festucae to infect perennial ryegrass. Null mutants of the histone H3 lysine 9 and lysine 36 methyltransferases are completely defective in colonizing the host intercellular space, and these defects are dependent on the methyltransferase activities of these enzymes. Importantly, we observed no evidence for increased host defense response to these mutants that can account for their non-infection. Rather, these infection defects can be rescued by the wild-type strain in co-inoculation experiments, suggesting that failure of the mutants to infect is due to altered expression of genes encoding infection factors that are under the control of the above epigenetic marks that can be supplied by the wild-type strain. Among genes differentially expressed in the mutants at the early infection stage is a putative small-secreted protein with a carbohydrate binding function, which deletion in E. festucae severely reduced infection efficiency.


Introduction
whereas plants infected with the ezhB mutant had a late onset host phenotype characterized by an 80 increase in both tiller number and root biomass [8]. Deletion of the cclA gene, which encodes a 81 homolog of the Bre2 component of the COMPASS (Set1) complex responsible for deposition of 82 H3K4me3, also led to transcriptional activation of the IDT and EAS genes in axenic culture. In 83 contrast, deletion of the kdmB gene, which encodes the H3K4me3 demethylase, decreased in 84 planta expression of the IDT and EAS genes, and reduced levels of IDTs in planta [10]. Plants 85 infected with the cclA or kdmB mutants have a host interaction phenotype that is similar to wild-86 type, demonstrating that loss of the ability to add (DcclA) or remove (DkdmB) H3K4me3 does not 87 impact on the ability of E. festucae to establish a symbiosis. To complete our analysis of the four 88 major histone H3 methyltransferases we examine here the role of E. festucae SetB (KMT3/Set2) 89 in the symbiotic interaction.

90
Set2 was originally identified in yeast as an enzyme that induces transcriptional repression 91 through methylation of histone H3K36 in yeast [11]. The Set2 protein interacts with the 92 hyperphosphorylated form of RNA polymerase II (RNAPII) and catalyses methylation of H3K36 93 during transcriptional elongation [12][13][14]. A role for H3K36 methylation in transcriptional 94 repression has been shown for both yeast and metazoans [11,[15][16][17]. More recent studies showed 95 that presence of H3K36me3 in filamentous fungi is not associated with gene activation [18,19], 96 and has an important role in regulating fungal development and pathogenicity. Deletion of SET2 97 homologs in Magnaporthe oryzae [20], Fusarium verticillioides [21], and Fusarium fujikuroi [19] 98 led to growth defects in culture and reduced host pathogenicity. Culture growth defects were also 99 observed for the corresponding mutants in Neurospora crassa and Aspergillus nidulans [22][23][24][25] 100 highlighting the importance of Set2 for fungal development. Here we show that a catalytically 101 6 functional Set2 homolog is crucial for E. festucae to infect its host L. perenne and establish a 102 mutualistic symbiotic association.

105
SetB is an H3K36 methyltransferase that regulates fungal growth and development 106 The gene encoding the E. festucae homolog of the Saccharomyces cerevisiae H3K36 107 methyltransferase Set2 (KMT3) was identified by tBLASTn and named setB (gene model no.

108
EfM3.042710; [26]). Protein structure analysis identified canonical PreSET, SET, PostSET, WW 109 and SRI domains along with an additional TFS2N domain (Fig 1A), as reported for the N. crassa 110 Set2 homolog [5]. E. festucae setB was deleted by targeted gene replacement using a nptII 111 geneticin resistance cassette for selection. PCR screening of Gen R transformants and Southern blot 112 analysis identified a single ΔsetB strain (S1 Fig). Western blot analysis of total histones showed 113 that H3K36me3 was specifically depleted in this DsetB strain, while levels of H3K36me1/2 were 114 increased ( Fig 1C). Introduction of the setB wild-type allele restored these methylation defects, 115 confirming the role of SetB in H3K36 methylation ( Fig 1C). This ΔsetB strain grew extremely 116 slowly in culture compared to the wild-type strain, a phenotype that was complemented by re-117 introduction of the setB wild-type allele ( Fig 1D). Microscopic analysis of the hyphal morphology 118 revealed that ΔsetB hyphae had a wavy pattern of growth, branched more frequently and had 119 hyphal compartments that were significantly shorter than observed for the wild-type strain (Fig   120   1E).  To study the role of SetB in the symbiotic interaction of E. festucae with its host, L. perenne 138 seedlings were inoculated with ΔsetB and grown for 8-12 weeks, after which time mature plant 139 tillers were tested for infection by an immunoblot assay using an antibody raised against E. 140 festucae. Four independent inoculation experiments revealed that ΔsetB was unable to infect these 141 host plants (Table 1, Fig 2A). The absence of the mutant in these mature plants was confirmed by 142 confocal laser-scanning microscopy (CLSM) (Fig 2B). To determine if ΔsetB failed to infect and 143 colonize leaf tissue, or whether it was lost from the leaf tissue during tiller growth (reduction of 144 endophyte persistence), we also examined the infection status of seedlings at one-and two-weeks 145 post inoculation (wpi) by CLSM. At both timepoints, both endophytic and epiphytic hyphae of 146 wild-type were observed at the infection site, but only epiphytic and not endophytic hyphae of 147 ΔsetB (Fig 2C and D). These observations demonstrate that ΔsetB is completely unable to colonize 148 the host at the site of infection. Re-introduction of the wild-type setB allele into the mutant restored 149 the ability of this strain to infect L. perenne ( Fig 2B, Table 1).

150
Subsequently, we sought to determine whether the phenotypes observed for ΔsetB were 151 caused by the lack of H3K36 trimethylation in this strain. Substitution of a conserved cysteine 152 residue for alanine in the SET domain of Neurospora crassa SET-2 has previously been shown to 153 result in a severe reduction in methyltransferase activity, confirming the importance of this amino 154 acid for enzyme activity [11]. We therefore substituted alanine for the corresponding C254 residue 155 in E. festucae SetB ( Fig 1B) and tested if this allele was able to complement the H3K36 156 methylation, culture growth and host infection defects of ΔsetB. Introduction of this setB C254A allele 9 H3K36me1/2 observed for this strain ( Fig 1C). Likewise, the colony growth ( Fig 1D) and host 159 infection phenotypes (Table 1) were only partially rescued by introduction of this allele. Although 160 the ΔsetB/setB C254A strain was able to infect the host, infection was at a significantly lower rate 161 than the ΔsetB/setB complement or the wild-type strains (  The E. festucae ΔclrD mutant, which is defective in H3K9 mono-, di-and tri-methylation, also has 179 a non-infection phenotype [8,10]. However, that analysis did not differentiate lack of infection 180 from lack of persistence as host plants were only examined for infection at 10-12 wpi. To 181 discriminate between these two possibilities, we examined seedlings infected with mutant or wild-182 type at one-and two-wpi using CLSM. At both timepoints epiphytic and endophytic hyphae were 183 observed for wild-type infected seedlings, but only epiphytic hyphae were observed at the 184 inoculation site for ΔclrD (Fig 3). These observations suggest that ΔclrD, like ΔsetB, is completely 185 incapable of infecting the host plant. It is interesting to note that in addition to this non-infection 186 phenotype, ΔclrD also shares several other phenotypes with ΔsetB, including a slow growth rate 187 on PDA ( Fig 4A) and aberrant hyphal and cellular morphologies ( Fig 4B).    To substantiate the above PCR results, we repeated the co-inoculation study using ΔclrD 249 or ΔsetB strains, constitutively expressing eGFP, together with wild-type expressing mCherry. We 250 were able to observe both eGFP-and mCherry-expressing hyphae in the mature plant tissues by 251 15 CLSM, confirming that both ∆clrD and ∆setB mutants are able to infect if co-inoculated with the 252 wild-type strain (Fig 5). In addition to individual eGFP-or mCherry-expressing hyphae, we also 253 observed hyphae expressing both eGFP and mCherry (Fig 5A), the result of anastomosis between 254 mutant and wild-type hyphae. A hyphal fusion test in axenic culture showed that the ΔsetB and 255 ΔclrD mutants were indeed capable of fusing with the wild-type strain (S5 Fig). However, the 256 presence of individual eGFP-labelled hyphae in the plant indicates that this fusion with wild-type 257 hyphae is not required for the mutants to enter the host, as it is not possible for anastomosed hyphae 258 to re-segregate. Taken together, these results support the hypothesis that ΔsetB and ΔclrD lack 259 permissive factors for infection, which can be supplied by co-inoculation with the wild-type strain. In addition to performing an immunoblot assay using anti-E. festucae antibody to 267 determine infection in the co-inoculation experiment described above, we also performed a replica 268 blot using anti-GFP antibody to determine the presence of the eGFP-labelled mutants in the   Four genes from these DEG sets stood out as candidate infection factors that might 301 contribute to host colonization. These were identified by applying two criteria: first we looked for 302 small secreted protein (SSP)-encoding genes that were upregulated in wild-type in planta at three 303 dpi (this study) relative to axenic culture [9], which suggests they may play important roles during 304 host infection (S2 File). Secondly, we searched for genes exhibiting strong differential expression 305 in one or both of the ∆clrD and ∆setB mutants vs. wild-type. The first gene we identified, named

321
Two of these, crbA and sspZ are subtelomeric (S9 Fig). Interestingly, crbA appears to be a 322 component of a four-gene cluster present in some other Epichloë species (S1 Table,

351
The culture colony morphology of all mutants was similar to the wild-type strain (S12 Fig).

352
To test if the strains were disrupted in their ability to infect the host, they were inoculated into 353 perennial ryegrass seedlings and the infection status of the plants was determined at nine wpi by 354 immunoblotting. All mutants were observed to infect the host (S13 Fig); however, infection rates 355 of the ΔcrbA mutant, and to a lesser extent the ΔdmlA mutant, were significantly lower (S2 Table). while CLSM showed that fusion between mutant and wild-type hyphae did occur, hyphae with 406 just one fluorescent marker were also present, highlighting that fusion with the wild-type hyphae 407 is not a prerequisite for host colonization by mutant hyphae after co-inoculation.

408
Since the completion of this study, another E. festucae mutant with a non-infection that are infected with these mutants do not reproduce the incompatibility phenotypes observed for 476 plants co-infected with wild-type and ΔclrD or ΔsetB. This is not surprising given that H3K9 and 477 H3K36 methylation defects would affect the expression of a large number of genes, resulting in 478 more dramatic phenotypes for these mutants. However, these host colonization and symbiosis 479 defects appear to be specific to these two H3 marks as mutations that affect H3K4 and H3K27

487
The growth defects of the ΔclrD and ΔsetB mutants observed in this study are consistent with 488 those observed in the N. crassa and Aspergillus nidulans mutants for these genes [22][23][24][25]. 489 Interestingly, deletion of the set2 homolog ash1 in Fusarium fujikuroi led to a more severe 490 developmental phenotype than deletion of set2, which is characterised by instability of In conclusion, we show here that ClrD-catalysed H3K9 and SetB-catalysed H3K36 498 methylation are crucial in regulating the ability of a fungal symbiont to infect its host. The results 499 of this study also underscore the importance of further analysis into the symbiotic roles and mode 500 of action of the small secreted protein encoded by the crbA gene, which appears to be important 501 for E. festucae infection efficiency.  506 Bacterial and fungal strains, plasmids, and plant material used in this study are listed in S3 Table. 507 E. festucae strains were grown at 22°C on 2.4% (w/v) potato dextrose agar or broth with shaking 508 at 200 rpm. E. festucae protoplasts were prepared and transformed as previously described [68,  Table). Plasmid pYL23, which 516 contains the setB replacement construct, was generated by Gibson assembly [72] from DNA 517 fragments containing the 5' and 3' regions flanking setB that were amplified from an E. festucae 518 Fl1 genomic DNA template using primer pairs YL286F/R and YL288F/R, respectively; a PtrpC-519 nptII-TtrpC gene expression cassette that confers geneticin resistance that was amplified from 520 pSF17.1 using primers YL287F/R; and a NdeI-linearised pUC19 vector sequence. A linear 521 fragment was excised from pYL23 by PacI/SpeI digestion and used for transformation of wild-522 type E. festucae strain Fl1 protoplasts to generate ∆setB strains. A 4.3 kb DNA fragment covering 523 the setB gene including promoter and terminator sequences was amplified from wild-type E. Histone extraction and western blot were performed as previously described [10]. In brief, fungal 569 tissues were ground to a fine powder in liquid nitrogen, and nuclei were isolated by glycerol 570 gradient centrifugation, sonicated, and histones were subsequently isolated by acid extraction.

609
A core set of highly up-and down-regulated genes for the ΔsetB and ΔclrD mutants were 610 identified using R. Genes were only included if their differential expression was in the same  Table. 616 33 Similarly, the hypothesis that genes in ΔsetB and ΔclrD mutants are differentially expressed in 617 planta compared to axenic culture was tested by logistic regression, using previously published 618 RNAseq data [9]. Analysis of L. perenne gene expression was carried out as for the Epichloë 619 mutant gene expression analysis using the same quality-controlled RNAseq data, except that 620 reads were aligned to the previously published L. perenne gene model set [28].