Genome assemblies of 52 Magnaporthe oryzae strains were downloaded from NCBI and their gene/protein models were predicted using Augustus/3.3.2 that was trained using Magnaporthe grisea gene model (Chiapello et al., 2015; Gladieux et al., 2018). A BLAST library of proteomes of all strains were made, and protein sequences of seven diterpene synthases in Chiapello et al. (2015) (five from the BR32 genome and two from the 70–15 genome from NCBI- MGG_01949 and MGG_14722) were queried against the library using blastp (blast/2.7.1+) (Chiapello et al., 2015; Gladieux et al., 2018). The resulting matching sequences were filtered by retaining the matches with specific parameters (identity >30% and e value <1e-20). The filtered sequences were taken as the query again and searched against the BLAST library and the same filtering was applied. The process was iterated three times until no more matching sequences was found. This finally resulted in 321 protein sequences that were further annotated as DiTPS candidates in 52 Magnaporthe oryzae strains. Sequences <600 amino acids were not included in the study. The putative DiTPSs were further checked for the presence of functional catalytic sites (class I and class II active sites) to shortlist bifunctional diterpene synthases. Class II cyclases are characterized by the DxDD motif involved in protonation-initiated cyclization. Class I active site has the characteristic D(D,E)xx(D,E) motif that facilitates ionization initiated cyclization of the class II intermediate (Supplementary Figure 1). A summary of the computational pipeline is depicted (Supplementary Figure 2). For phylogenetic analysis, 14 genomes were selected representing two genomes each from seven pathotypes (Supplementary Table 1). Similarly, for sesquiterpene synthases, the blast search was iterated three times (identity >30% and e value <1e-20) until no more matching sequences were found. This finally resulted in 160 protein sequences and the sequences with motifs- DDxxD, NSE/DTE, and RxR and sequence lengths more than 220 amino acids were used for the phylogenetic analysis. However, for chimeric terpene synthases, iterated blast search with identity >30% didn't result in a limited pool, and therefore a stringent parameter of identity >90% was used and the blast search was iterated four times (e value <1e-20) until no more matching sequences were found. This finally resulted in 194 protein sequences and the sequences with N-terminal motifs- D/ED/ExxD/E, NSE [N/HDxx(S/T)xxxD/E], and C-terminal motifs- DDxxD and DDxxN sequence length more than aa were used for phylogenetic analysis.

The filtered sequences for all three terpene synthases- di-, sesqui, and chimeric terpene synthases were then aligned using Muscle alignment tool followed by construction of maximum likelihood tree with 100 bootstrap repetitions using the software MEGAX (Kumar et al., 2018).

Synthetic codon-optimized diterpene synthase genes for heterologous expression in Escherichia coli were obtained from Thermo Scientific. The genes were amplified by PCR and sub-cloned into pET28b expression vectors between Bam HI and Hind III restriction sites, using In-fusion HD cloning kit (Takara Bio) and primers in Supplementary Table 2. The constructs were confirmed by full-length sequencing. The genes were also cloned into pESC-URA vector between Not I and Spe I restriction sites using In-fusion HD cloning kit and primers for expression in yeast (Supplementary Table 2). The diterpene synthase Class I knock-out mutants were generated by mutating the first aspartate residue of the DDxxD class I motif to alanine by PCR based site-directed mutagenesis of pET28b vectors cloned with diterpene synthase genes (Supplementary Table 2).

The diterpene synthase products were determined using a modular metabolic engineering system (Cyr et al., 2007; Morrone et al., 2010). Briefly, the vectors carrying each diterpene synthase gene were individually co-transformed with a GGPP synthase carried on a compatible pACYC-Duet plasmid into C41 Overexpress E. coli cells pre-transformed with pIRS (to increase yield) for recombinant expression of each diterpene synthase. The resulting recombinant strains were grown at 37°C overnight in 5 ml of TB media (supplemented with 100 mM phosphate buffer pH 7.5 and appropriate antibiotics). One ml of overnight culture was used to inoculate 50 ml of TB media and the cultures were incubated at 37°C and 160 rpm till mid-log phase. At this point, the temperature was dropped to 16°C for 1 h and then 1 mM isopropylthiogalactoside (IPTG) was added for induction and 50 mM sodium pyruvate was added as a carbon supplement. The induced culture was further incubated for 3 days at 16°C and 160 rpm to allow product formation. The resulting products were extracted adding equal volume of hexane and the organic layer was separated, dried and re-dissolved in 1 ml of n-hexane for product analysis.

The pESC-URA vectors carrying diterpene synthase genes were co-transformed with mFPS plasmid (avian FPP mutant expressing GGPP) into yeast ZXB competent cells using previously described (Zhuang and Chappell, 2015). The yeast cells were grown and maintained in YPD media supplemented with ergosterol. The positive colonies were obtained using appropriate SC selection media containing ergosterol (SCE media) and were used to inoculate 50 ml of SCE media containing galactose as an inducer for pESC-URA cloned genes. The culture was grown at 28°C and 220 rpm for 6 days before extraction of resultant products using equal volume of acetone, followed by one volume of hexane. The upper organic layer was centrifuged, dried up and redissolved in 1 ml of n-hexane for product analysis.

The product analysis was performed by gas chromatography with mass spectral detection (GC-MS) on Agilent 7890/7000C GC-QQQMS using an Agilent HP-5MS column with 1.2 ml/min helium flow rate. Samples (1 ul) were injected in splitless mode using an Agilent 7693 autosampler. The following temperature program was used: the oven temperature started at 70°C, which was maintained for 1 min, and then increased at a rate of 20°C/min to 300°C, where it was held for another 2 min. The mass spectrum was recorded by mass-to-charge ratio (m/z) values in the range from 60 to 600 starting from 9 min after sample injection until the end of the run. For identification of enzymatic products, the retention time and mass spectra of products were compared to authentic standards (Supplementary Table 3).

The diterpene scaffolds generated by diterpene synthases are further decorated by oxidases, reductases, and transferases to form bioactive diterpenoids. The biosynthetic genes of specialized metabolites in most fungi are typically located next to each other and form biosynthetic gene clusters. To evaluate if MoDiTPS1 and MoDiTPS2 occur as functional biosynthetic gene clusters (BGCs) in M. oryzae genomes, putative BGCs of MoDiTPS1 and MoDiTPS2 were predicted using in-silico analysis tools. Briefly, the putative biosynthetic gene clusters of the two diterpene synthases -MoDiTPS1 and MoDiTPS2 from M. oryzae 70-15 reference genome were first predicted through sorting of genes with oxidase, reductase, transferase, or GGPP synthase annotations, 70 kb upstream and downstream of diterpene synthase genes from Mo 70-15 genome. The synteny of these putative biosynthetic genes among other pathotype genomes was then analyzed by using the sorted genes as query against chromosomes containing diterpene synthase genes of B71, and MZ5-1-6 genomes using SimpleSynteny online tool (Veltri et al., 2016).

Given the potential roles of diterpenoids in fungal development and manipulation of host defense mechanisms, we assessed the genetic potential of the M. oryzae reference genome (Mo70_15) to produce diterpenes. Using BLAST analysis, and presence of characteristic protein domains for published genome analysis we identified two putative diterpene synthases (Chiapello et al., 2015). As Magnaporthe is a species complex that infects other grass species in addition to rice, we then investigated the distribution of diterpene synthases (DiTPSs) among M. oryzae pathotypes that infect different host plants. Utilizing the diterpene synthase mining approach described in the Methods section and depicted in Supplementary Figure 2, we identified putative DiTPSs from 52 M. oryzae proteomes covering 14 pathotypes (Chiapello et al., 2015; Gladieux et al., 2018). As most bifunctional DiTPS sequences are >800 amino acids, sequences <600 amino acids were not included in the study. The putative DiTPSs were further checked for the presence of functional catalytic motifs (in both class I and class II active sites) to create a shortlist of bifunctional diterpene synthases. This DiTPS mining approach revealed presence of an additional enzyme—DiTPS6 in addition to five annotated DiTPS enzymes. The DiTPSs are numbered as MoDiTPS1 through MoDiTPS6. The phylogenetic distribution of DiTPSs among 14 representative pathotypes is illustrated in Figure 1.

There is an increased number of DiTPS enzymes in pathotypes of grass species, such as wheat (Triticum), brome grass (Bromus), and fingermillet (Eleusine) compared to rice (Oryza) pathotype. MoDiTPS6 is uniquely present in the pathotypes from two grasses –Triticum and Bromus. The observed expansion of DiTPS enzyme family aligns with several studies that have highlighted the fluidity of fungal genomes and expansion of various gene families (Haas et al., 2009; Duplessis et al., 2011). Phylogenetic analysis revealed a tree with two distinct clades where one clade diverged into MoDiTPS2, 3, and 5 while the other clade consists of the remaining MoDiTPS1, 4 & 6 (Figure 1). The divergence of two clades is strongly supported by bootstrap value of 100 on the node of divergence. This observation is in agreement with Fischer et al. (2015) who suggested that fungal diterpene synthases essentially fall into two clades (Fischer et al., 2015). To assess if the observed phylogenetic expansion was distinct to diterpene synthases, a similar gene mining strategy for terpene synthase was employed for sesquiterpenes (C15) and sesterterpenes (C25). Here, the number of terpene synthases were consistent across all pathotypes (Supplementary Figures 3, 4). Thus, within the Magnaporthe terpenome of the currently sequenced strains, phylogenetic expansion seems to be limited to the DiTPSs.

Fungal diterpene synthases have been traced back to an ancient horizontal gene transfer event from plants to Ascomycete (Fischer et al., 2015). We carried out phylogenetic analysis to compare how diterpene synthases from M. oryzae pathotypes relate to characterized diterpene synthases in Ascomycete (Figure 2). The diterpene synthases from genomes of three different pathotypes (70–15 reference genome, MZ5-1-6; fingermillet pathotype genome and B71; wheat pathotype genome) were aligned with previously characterized diterpene synthases. The analysis shows that M. oryzae diterpene synthases among these three distinct pathotypes are closely related to each other in each orthologous group. The phylogenetic tree strongly suggests (100 bootstrap value on the node of divergence) common ancestral origin of ent-kaurene synthases and MoDiTPS1, 4 and 6 clade. MoDiTPS2 and 5 have same node of divergence and do not group with any characterized DiTPS. However, DiTPS3 associates with a known phyllocladanol synthase (Toyomasu et al., 2008). Diterpene biosynthesis in Magnaporthe pathotypes provides a biological system to investigate the importance of chemical diversification in evolution and adaptation to its ecological niche.

To determine the metabolites produced by the two core DiTPSs that are evolutionarily conserved in all pathotypes (i.e., MoDiTPS1 & 2), these were biochemically characterized using a previously described modular metabolic engineering system (Cyr et al., 2007; Morrone et al., 2010). Codon-optimized genes for each were expressed in E. coli engineered to make GGPP. The resultant diterpene scaffolds were confirmed by comparing both mass spectra and retention time to authentic standards (Figure 3). DiTPS1 carries out bifunctional cyclization resulting in manoyl oxide (Figure 3A). DiTPS2 resulted in formation of pimara-8(9),15-diene confirmed via comparison to an authentic standard (Figure 3B) (Karunanithi et al., 2020). The stereochemistry was determined as described below. The co-expression of these diterpene synthase genes with the precursor, GGPP in yeast strain ZXB also resulted in formation of same products (Supplementary Figure 5). Similarly, the expanded array of diterpene synthase from wheat (B71) and finger millet (MZ5-1-6) were expressed in the metabolic engineering system. Although B71 DiTPS1 was found to be nonfunctional in our expression system, interestingly DiTPS6 also produces manoyl oxide similar to DiTPS1, demonstrating functional conservation. DiTPS6 is paralogous to DiTPS1 with an 80% amino acid sequence identity suggesting parallel evolution (Supplementary Figure 6). In contrast, DiTPS3 which is in the same clade as DiTPS2, forms a yet to be identified product demonstrating duplication and neofunctionalization (Supplementary Figure 7). There was no activity observed for the remainder of the genes from wheat pathotype (B71) and fingermillet (MZ5-1-6) (Table 1, Supplementary Figure 6).

To determine the diterpene scaffold intermediates generated by class II active site of the bifunctional diterpene synthases, the class I active site of the diterpene synthases were made inactive by mutation of first aspartate from the DDxxD motif to alanine (retains only the ability to make CPP in the class II active site). Subsequent co-expression of the class I mutants with GGPP resulted in production of the expected intermediate 8α-hydroxy-CPP by DiTPS1 and CPP for DiTPS2 (Supplementary Figure 8). To probe the substrate specificity of the class I active site and, hence, presumably also the stereochemistry of the corresponding class II product, the class II site was inactivated by mutation of the middle aspartate of the DxDD motif to alanine to eliminate CPS activity and fed various CPP (Figure 3) (Kawaide et al., 1997; Toyomasu et al., 2004, 2008; Bromann et al., 2012; Xu et al., 2017). The MoDiTPS1 class I active site did not react with normal 8α-hydroxy-CPP, nor any of the normal, ent or syn stereoisomers of CPP (Supplementary Figure A). To assess whether the intermediate might be enantiomeric, we co-expressed the MoDiTPS1 class I mutant with stereospecific class I enzymes—i.e., an ent- specific kaurene synthase (AtKS) or normal specific abietadiene synthase (AgAS) (Mafu et al., 2015). Since only AtKS reacts with the MoDiTPS1 class II product ent-8α-hydroxy-CPP (Supplementary Figure A); MoDiTPS1 produces ent-manoyl oxide. The DiTPS2 class I active site reacts with normal CPP to produce pimara-8(9),15-diene, indicating that this is relevant stereoisomer of CPP produced by its class II active site. The DiTPS2 class I active site also reacts with syn-CPP resulting in unidentified products but does not form any products with ent-CPP (Supplementary Figure 8B). These results are consistent with previously characterized diterpene synthases from Ascomycete (Figure 2B).

From our analysis the putative MoDiTPS1 cluster possesses GGPP synthase (MGG_01971), oxidases/cytochromes P450 (MGG_01947, MGG_01950, MGG_01960), ligase (MGG_01951), and could possibly include a UDP glycosyl transferase (MGG_01961) (Supplementary Figure 9). The synteny results highlight that the BGCs are not conserved across the pathotypes as the cluster lacks a conserved transcription regulator (MGG_01946). The MoDiTPS1 cluster in genomes B71 and MZ5-1-6 may have variable regulator genes. The putative DiTPS2 cluster shows the presence of a complete and conserved biosynthetic gene cluster among M. oryzae pathotypes with precursor GGPP synthase (MGG_00026), oxidases and/or cytochromes P450 (MGG_00023, MGG_00024 & MGG_00031), dehydrogenase (MGG_00017); transcription factor (MGG_00032), and transporter (MGG_00040) (Supplementary Figure 9). The biochemical characterization also showed conservation of product formation [pimara-8(9), 15-diene] and functionality of MoDiTPS2 among M. oryzae pathotypes (Supplementary Figure 5). These results suggest a conserved role of MoDiTPS2 product in M. oryzae pathotypes. In addition, the MoDiTPS2 cluster is located next to another specialized metabolite cluster annotated as non-ribosomal peptide synthase, bassianolide synthase. It would be interesting to explore if a further functional correlation exists.