Introduction

Rice (Oryza sativa) is well known for producing many diterpenoid phytoalexins and has served as model system for investigation of these natural products, which are prevalent in cereal crops more generally (Murphy and Zerbe 2020). One of the early findings from these studies was discovery of biosynthetic gene clusters (BGCs) associated with such metabolism, which were among the first to be discovered in plants (Nutzmann et al. 2016). In particular, functional identification of the consecutively acting pairs of cyclases led to realization of the proximity of the relevant genes (Prisic et al. 2004; Wilderman et al. 2004). Moreover, the two identified regions each also contained a number of genes for cytochrome P450 (CYP) monooxygenases, which were later demonstrated to act on the diterpenes generated by the co-clustered cyclases (Swaminathan et al. 2009; Wang et al. 2011, 2012; Wu et al. 2011, 2013; Kitaoka et al. 2015).

The use of two consecutively acting cyclases is characteristic of the labdane-related nature of almost all rice diterpenoids. Their production is initiated by bicyclization of the general diterpenoid precursor (E,E,E)-geranylgeranyl diphosphate (GGPP) catalyzed by class II diterpene cyclases, which prototypically produce the eponymous labdadienyl/copalyl diphosphate (CPP) as distinct stereoisomers (Peters 2010). For example, all land plants must produce ent-CPP and, hence, contain such a CPP synthase (CPS) for phytohormone biosynthesis. For this purpose, ent-CPP must then be further cyclized to the tetracyclic olefin ent-kaurene by a class I diterpene synthase. In angiosperms the CPS and kaurene synthase (KS) required for gibberellin phytohormone biosynthesis have often given rise to expanded gene families involved in more specialized metabolism, with the latter generally termed KS-like (KSL) (Zi et al. 2014).

In rice, functional characterization of the syn-CPP synthase OsCPS4 and subsequently acting syn-pimaradiene synthase OsKSL4 led to discovery of co-localization their genes on chromosome 4 (Wilderman et al. 2004; Xu et al. 2004). Notably, the only other genes in the region encoded the closely related paralogs CYP99A2 and CYP99A3, and pair of short-chain alcohol dehydrogenases, defining a BGC (c4BGC). Consistent with RNAi knock-down studies (Shimura et al. 2007), it was then shown that CYP99A3 (Wang et al. 2011), and latter CYP99A2 (Kitaoka et al. 2015), react with syn-pimaradiene, both introducing oxygen at carbon-19 (C19), as required for momilactone biosynthesis.

Similarly, functional characterization of the ent-cassadiene synthase OsKSL7 and upstream ent-CPP synthase OsCPS2 dedicated to more specialized metabolism led to discovery of another BGC located on chromosome 2 (Cho et al. 2004; Prisic et al. 2004). This c2BGC contains multiple members of the CYP71Z and CYP76M sub-families, with some of the latter surprisingly involved in biosynthesis of the momilactones associated with the c4BGC, demonstrating interdependent evolution of these two BGCs (Kitaoka et al. 2021; Li et al. 2022). Not surprisingly, the presence of CYP76M sub-family members at the c2BGC locus is then conserved throughout the Oryza genus (Miyamoto et al. 2016), with c4BGC orthologs found even more widely (Wu et al. 2022).

More recently reported was another diterpenoid BGC on chromosome 7 (c7BGC), encoding ent-10-oxodepression production, which seems to be almost entirely restricted to the japonica subspecies (ssp.) and is only rarely found in the other major ssp. (indica) of rice (Zhan et al. 2020). While this report only described the phylogenetically related but functionally distinct CYP71Z2 and CYP71Z21, an independent study found the c7BGC also contains genes encoding CYP71Z22, which is closely related and functionally redundant with CYP71Z21, as well as CYP71Z30, albeit this latter is found as a pseudo-gene even in some cultivars (cv.) from ssp. japonica (Liang et al. 2021).

Oryzalexin S is a phytoalexin against the devastating fungal blast pathogen Magnaporthe oryzae, and is derived from syn-stemar-13-ene via hydroxylation at C2α and C19 (Tamogani et al. 1993). The relevant OsKSL8, acting on the syn-CPP produced by OsCPS4, has been identified, with the encoding gene found on chromosome 11 (Nemoto et al. 2004). In addition, a closely related (both mechanistically and phylogenetically) syn-stemod-13(17)-ene synthase was then identified (Morrone et al. 2006). Although the relevant gene loci was not found in the available genome sequence, the relatively low sequence identity (< 93% nucleotide sequence identity) suggested that it was not an allele of OsKSL8, as other KSL alleles (even with functionally distinct activity and from different ssp.) shared > 99% identity (Xu et al. 2007), and so this was termed OsKSL11. However, it has since been clarified that OsKSL11 is a distinct allele of OsKSL8 derived from ssp. indica rice, which can then be termed OsKSL8i, while the other ssp. japonica derived allele is then OsKSL8j (Toyomasu et al. 2016).

Here the CYPs catalyzing hydroxylation at C19 and, subsequently, C2α of syn-stemarene to generate oryzalexin S were identified as the closely related CYP99A2/3 from the c4BGC and CYP71Z21/22 from the c7BGC, respectively. Intriguingly, utilizing the recently reported phylogenetically representative set of pan-rice genomes (Zhou et al. 2020), it was found that the relevant syn-stemarene producing OsKSL8j is actually not present in the examined (sub)tropical spp. japonica cultivars, having been replaced by the functionally distinct, syn-stemodene producing OsKSL8i (OsKSL11), the implications of which are discussed.

Materials and methods

Unless otherwise stated, all chemicals were obtained from Fisher Scientific. The genes utilized here were largely synthetic, optimized for expression in Escherichia coli, as previously reported—i.e., sCYP99A2 and sCYP99A3 (Wang et al. 2011), as well as sCYP71Z6 and sCYP71Z7 (Wu et al. 2011), and sCYP71Z22, but the native CYP71Z21 (Liang et al. 2021). These CYPs were expressed as described (Kitaoka et al. 2015), either with just the CYP reductase (CPR) from Arabidopsis thaliana (AtCPR1) for feeding studies, or in a previously described modular metabolic engineering system (Cyr et al. 2007), as adapted for CYP expression (Kitaoka et al. 2015), with expression of a GGPP synthase (AgGGPS), OsCPS4 and sOsKSL8j to produce syn-stemar-13-ene. As needed, metabolic flux was increased as previously described (Morrone et al. 2010). The resulting diterpenoids were extracted and analyzed by GC–MS and, as necessary, purified by flash chromatography and HPLC, using GC–MS analysis to track the targeted compounds, which were then structurally analyzed by NMR and/or utilized for feeding studies, all as has been previously described—e.g., (Swaminathan et al. 2009; Wang et al. 2011, 2012; Wu et al. 2011, 2013; Kitaoka et al. 2015). Gene presence was determined via BLAST searches of the relevant genes from cv. Nipponbare (ssp. japonica) against the cDNA databases for the described (sub)species and, where relevant, cultivars at http://oryza.gramene.org (Tello-Ruiz et al. 2022). Phylogenetic analysis of KSL8 was carried out using CLC Main WorkBench 22 (Qiagen Aarhus A/S).

Results

Based on the previously reported ability of CYP99A3 to catalyze at least C19-hydroxylation of syn-stemodene (Wang et al. 2011), the activity of both this and the closely related paralog CYP99A2 with the oryzalexin S precursor syn-stemar-13-ene (1) was examined. Specifically, these CYPs were each co-expressed with the requisite CPR in E. coli also engineered to produce this putative substrate. In both cases this led to observation of what appeared to be hydroxylated derivatives of 1, along with the accompanying syn-stemod-12-ene also made by OsKSL8j in approximately the same ratio relative to 1 as seen here for these hydroxylated products (Xu et al. 2007), as indicated by the increase in molecular weight (MW) from 272 to 288 Da (Fig. 1Ai, ii and Supplemental Fig. S1). For structural analysis, metabolic flux to isoprenoids was increased and the culture volume increased with CYP99A3 (as this seems to be more efficient than CYP99A2, see Supplemental Fig. S1), enabling purification of sufficient amounts of this diterpenoid for NMR analysis (Supplemental Table S1), which revealed the major product to be the expected C19-hydroxylated derivative syn-stemar-13-en-19-ol (2).

Fig. 1
figure 1

A CYP99A2/3 catalyze 19-hydroxylation of syn-stemar-13-ene (1). i) Extracted ion count (EIC) chromatogram from GC–MS analysis of extracts from E. coli co-expressing CYP99A3 and the requisite AtCPR1 as well as engineered to produce 1 via co-expression of a GGPP synthase and OsKSL8j. Peak marked with * is presumed to be 19-hydroxy derivative of syn-stemod-12-ene, which is produced in approximately the same ratio relative to 1 by OsKSL8j as seen here for these hydroxylated products (Xu et al. 2007). ii) Mass spectra of syn-stemar-13-en-19-ol (2). See Supplemental Fig. S1 for verification of CYP99A2 activity and additional GC–MS data supporting assignment of * as syn-stemod-12-en-19-ol. B CYP71Z21/22 catalyze 2α-hydroxylation. i) EIC chromatogram from GC–MS analysis of extracts from E. coli co-expressing CYP71Z22 and the requisite AtCPR1, and then fed syn-stemar-13-en-19-ol (2). See Supplemental Fig. S2 for verification of CYP71Z21 activity. ii) Mass spectra of oryzalexin S (3). iii) EIC chromatogram from GC–MS analysis of extracts from E. coli co-expressing CYP99A3, CYP71Z21 and the requisite AtCPR1 as well as engineered to produce - 1. iv) Mass spectra of syn-stemar-13-en-2α-ol (4). v) EIC chromatogram from GC–MS analysis of extracts from E. coli co-expressing CYP99A3 and the requisite AtCPR1, and then fed syn-stemar-13-en-2α-ol (4) purified from E. coli co-expressing CYP71Z22 and the requisite AtCPR1 as well as engineered to produce 1 (see Supplemental Fig. S3 for EIC chromatogram from GC–MS analysis of extract). C Conservation of Oryzalexin S. i) Biosynthetic pathway, with previously identified diterpene cyclases on top and CYPs identified here below (also indicated by bold text). ii) Conservation of relevant genes (bold italic text indicates those explicitly discussed here)

Full size image

Previous work indicated that CYP71Z6 and CYP71Z7 from the c2BGC act on C2 of specific (ent-CPP derived) labdane-related diterpenes (Wu et al. 2011), producing C2α-hydroxy derivatives (Kitaoka et al. 2015). Accordingly, neither was found to modify syn-stemarene. Here, feeding experiments demonstrated these also do not react with syn-stemar-13-en-19-ol (data not shown). Given the recent discovery of roles for the CYP71Z sub-family members from the c7BGC in diterpenoid phytoalexin biosynthesis, despite their activity with the macrocyclic ent-casbene, which is not a labdane-related diterpene as it is derived from direct cyclization of GGPP by a class I diterpene synthase, these also were investigated. Fortuitously, feeding experiments found that, while CYP71Z2 was unreactive, the closely related CYP71Z21 and CYP71Z22 both efficiently converted syn-stemar-13-en-19-ol to oryzalexin S (Fig. 1Bi, ii and Supplemental Fig. S2). This demonstrates CYP71Z21/22 catalyze the necessary C2α-hydroxylation. In addition, co-expression of CYP71Z22 and CYP99A3 (as well as AtCPR1) in E. coli also engineered to produce syn-stemarene led to production of oryzalexin S (3), along with a substantial amount of an alternative hydroxylated derivative (Fig. 1Biii, iv). To verify this product was the expected syn-stemar-13-en-2α-ol (4), CYP71Z22 was utilized in the metabolic engineering system (Supplemental Figure S3), along with increased metabolic flux to isoprenoids and larger culture volume, enabling purification of sufficient amounts of this diterpenoid to feed to CYP99A3, which was able to convert this to oryzalexin S (Fig. 1Bv). The less efficient production of oryzalexin S (3) from syn-stemar-13-en-2α-ol (4) by CYP99A3, relative to syn-stemar-13-en-19-ol (2) by CYP71Z22, along with accumulation of 4 in the full-pathway containing E. coli, suggests preferential hydroxylation, with initial C19-hydroxylation of syn-stemarene by CYP99A2/3 and subsequent C2α-hydroxylation by CYP71Z21/22 in oryzalexin S biosynthesis (Fig. 1Ci).

Given the previously reported ssp. specific occurrence of both OsKSL8j and the c7BGC, especially the variation observed for the associated CYP71Z sub-family members, it seemed prudent to investigate distribution of the genes associated with oryzalexin S biosynthesis. For this purpose, the recently reported phylogenetically representative set of Asian rice genome sequences (Zhou et al. 2020), along with a broader set of domesticated and wild rice relatives across the Oryza genus, up to and including Oryza punctata, which served as an outgroup (Stein et al. 2018), were examined here (Fig. 1Cii). As previously reported, the c4BGC is widely conserved (Miyamoto et al. 2016), with OsCPS4 and CYP99A2/3 present in all these genomes, and the c7BGC is found in all ssp. japonica, although at least CYP71Z21 is more widely distributed, including other species (Zhan et al. 2020). While the O. rufipogon landrace examined here (W1943) does not contain CYP71Z21/22, which clarifies the basis for the previously reported absence of oryzalexin S therein (Toyomasu et al. 2016), it was previously reported that 12/13 O. rufipogon landraces contain at least CYP71Z21 (Zhan et al. 2020), hence the “±” indication at this point. More interestingly, substantial variation is observed with KSL8. Note that while the originally identified OsKSL8j was cloned from cv. Nipponbare, only a partial cDNA was found in the database examined here. Thus, while apparently partial cDNA are found for a few other O. sativa cultivars in this database (as indicated by appended “p”), these OsKSL8 may be functional and were included in the phylogenetic analysis, enabling assignment of relevant allele. Despite association of the syn-stemarene producing allele with ssp. japonica, the corresponding OsKSL8j is only found in the temperate representative cv. Nipponbare, as well as the wild-rice relative O. rufipogon. The other (tropical and subtropical) representatives contain alleles phylogenetically closer to OsKSL8i (OsKSL11), with full-length open-reading frame (ORF) sequence identities > 95.5% versus < 92.5%, respectively. Consistent with retention in only the temperate cultivars of ssp. japonica, OsKSL8j was found in the temperate cv. KitaakeX (Jain et al. 2019), while OsKSL8i is found in the tropical cv. Carolina Gold (Vaughn et al. 2021). OsKSL8 from the more distant species of rice were included in this phylogenetic analysis (Supplemental Figure S4), allowing tentative assignment of allele in almost all the AA-genome rice, although not the more distant (BB-genome) O. punctata, in which the most closely related cDNA shared 85–88% ORF sequence identity with the other full-length KSL8 examined here. Strikingly, the only other exception is found in the O. sativa ssp. basmati representative, where the most closely related cDNA shared < 82% ORF sequence identity with all other KSL8 examined here (including that from O. punctata).

Discussion

Here the CYPs catalyzing hydroxylation of syn-stemarene at C19 and, as shown here, the subsequent hydroxylation at C2α required for oryzalexin S biosynthesis were identified as the closely related paralogous pairs CYP99A2/3 and CYP71Z21/22 (respectively). The activity of CYP71Z21/22 is somewhat surprising given the differences between the macrocyclic ent-casbene that these react upon for ent-10-oxodepressin production versus the labdane-related (multicyclic) syn-stemar-13-en-19-ol these react with for oryzalexin S biosynthesis. Regardless, the encoding pair of genes are found in the previously identified c4BGC and c7BGC (respectively), with the former also containing that for the syn-CPP synthase OsCPS4, which is required for production of oryzalexin S (Zhang et al. 2021). By contrast, the gene for the relevant syn-stemarene synthase OsKSL8(j) is located elsewhere (chromosome 11), and then cross-stitches together these two BGCs to form the oryzalexin S biosynthetic pathway.

Given both OsKSL8j and the c7BGC are specifically associated with ssp. japonica (Toyomasu et al. 2016; Zhan et al. 2020), and, especially, the discovery that OsKSL11 is a functionally distinct ssp. indica derived allele (more accurately termed OsKSL8i), a wider examination of the genes relevant to oryzalexin S biosynthesis was carried out here. The results were consistent with the previous suggestion that such metabolism was present in the progenitor of ssp. japonica and O. rufipogon (Toyomasu et al. 2016), with possibly more ancient origins suggested by the presence of the relevant genes more widely in the Oryza genus (although this latter hypothesis needs further investigation). Strikingly, in O. sativa ssp. japonica it appears the OsKSL8j allele has only been retained in temperate cultivars, with those from (sub)tropical climates containing OsKSL8i (OsKSL11) instead. While gene flow via hybridization and introgression from ssp. japonica into ssp. indica is well known (Wing et al. 2018), these results indicate introgression of this functionally distinct allele in the opposite direction (i.e., from ssp. indica into ssp. japonica), with the accompanying disappearance of oryzalexin S. Indeed, an even more extreme case is found in ssp. basmati, where a highly divergent allele appears. Thus, it will be of interest to determine the product of this OsKSL8 variant. In addition, although CYP99A3 from the widespread c4BGC has been reported to catalyze oxygenation at C19 of the syn-stemod-13(17)-ene product of OsKSL8i (OsKSL11) (Wang et al. 2011), this is not sufficient for diterpenoid solubility and bioactivity, which invariably requires at least two spatially separated oxy groups (Wu et al. 2013). However, syn-stemod-13(17)-en-19-ol is not further transformed by CYP71Z21/22 (data not shown). Accordingly, no syn-stemod-13(17)-ene derived phytoalexin has yet been identified, so it will be of interest to investigate any additional oxygenation of this to examine bioactivity and determine the basis for introgression of OsKSL8i into (sub)tropical ssp. japonica, as well as the more divergent allele found in ssp. basmati.