Phenylphenalenones and Linear Diarylheptanoid Derivatives Are Biosynthesized via Parallel Routes in Musella lasiocarpa, the Chinese Dwarf Banana

Here, we use transcriptomic data from seeds of Musella lasiocarpa to identify five enzymes involved in the formation of dihydrocurcuminoids. Characterization of the substrate specificities of the enzymes reveals two distinct dihydrocurcuminoid pathways leading to phenylphenalenones and linear diarylheptanoid derivatives, the major seed metabolites. Furthermore, we demonstrate the stepwise conversion of dihydrobisdemethoxycurcumin to the phenylphenalenone 4′-hydroxylachnanthocarpone by feeding intermediates to M. lasiocarpa root protein extract.

P henylphenalenones (PPs) are complex polycyclic natural products that play an important role in the chemical defense system of banana and plantain (Musaceae). 1 Previous studies demonstrated the antipathogen activities of PPs 2−11 and linked the increased accumulation of PPs to the increased viability of disease-resistant banana varieties. 8−24 However, most of the intermediates of PP biosynthesis are still unknown and none of the genes involved in the pathway have been identified.
In a recent study, we elucidated the structures of various PPs and linear DH derivatives found in the seeds of the Chinese dwarf banana (Musella lasiocarpa, Musaceae), which were present at later stages (brown and black seeds) but absent at early stages of development (yellow seeds) (Figure 1A). 25 Based on their structural features, we hypothesized that the assembly of both compound classes likely shares the initial steps of dihydrocurcuminoid biosynthesis and involves two different starter substrates (p-coumaroyl-CoA 1 and feruloyl-CoA 2; Figure S1).The biosynthesis of curcuminoids has been well studied in Curcuma longa and involves the sequential action of the type III polyketide synthases diketide-CoA synthase (DCS) and curcumin synthase (CURS) (Figure S3). 26 Here, we perform a de novo transcriptome assembly and compare gene expression across three seed stages.A translated nucleotide search (tBLASTn) using DCS and CURS as query sequences led to the identification of two DCS homologues (MlDCS1 and MlDCS2) and two CURS homologues (MlCURS1 and MlCURS2), which all were highly expressed in yellow seeds (Figure 1A).
The catalytic activities of MlDCS1 and MlDCS2 were established in assays using purified enzymes overexpressed in Escherichia coli.Both enzymes accepted various phenylpropanoid-CoA esters as starter substrates and used one molecule of malonyl-CoA as the extender substrate to catalyze the formation of the corresponding diketide-CoAs (Figure S4).
The benzalacetone derivatives were detected after alkaline hydrolysis using high-resolution electrospray ionization mass spectrometry (HRESIMS).Product identities were confirmed by comparing their retention times and MS/MS fragmentation patterns with those of commercial standards.Interestingly, the substrate specificities of the two enzymes differed.While MlDCS1 clearly preferred p-coumaroyl-CoA 1 as the starter substrate, MlDCS2 favored feruloyl-CoA 2 (Figure 1B).To test the catalytic function of the enzymes in the plant environment, Nicotiana benthamiana leaves were transiently transformed via Agrobacterium-mediated infiltration.HRE-SIMS analysis of alkaline-hydrolyzed leaf extracts revealed that plants expressing MlDCS1 exclusively formed p-coumaroyldiketide-CoA 3, while feruloyldiketide-CoA 4 was the only product formed in plants expressing MlDCS2 (Figure S4).
The enzymatic functions of MlCURS1 and MlCURS2 were assessed in co-incubations with purified MlDCS1 or MlDCS2 in the presence of their corresponding preferred starter substrates, 1 and 2. The two MlCURSs were able to catalyze the decarboxylative condensation of diketide-CoAs 3 or 4 with another molecule of the substrates 1 or 2, respectively, to yield bisdemethoxycurcumin 9 and curcumin 10, respectively (Figure 1C).However, MlCURS1 showed higher activity in combination with MlDCS1 and 1, while MlCURS2 performed better with MlDCS2 and 2 (Figure 1B).These findings were further substantiated through co-infiltration experiments with different combinations of MlDCS and MlCURS in N. benthamiana leaves, in which MlCURS1 and MlCURS2 exhibited the same substrate preferences observed in the in vitro assays (Figures S5 and S6).
Given that curcuminoids 9 and 10 possess a Δ 6 double bond, while the putative PP precursors (Scheme 1) and all isolated linear DH derivatives 25 are saturated at C-6, reduction of the Δ 6 double bond may be a shared step in the biosynthesis of both PPs and linear DH derivatives (Figure S1).It could occur at either the phenylpropanoid-CoA, diketide-CoA, or curcuminoid stage (Figure S7A).We searched for double bond reductases (DBRs) that could act on double bonds in conjugation with carbonyl functions in the de novo transcriptome assembly of M. lasiocarpa.A total of nine homologues were annotated as NADPH-dependent 2-alkenal reductases.However, only one of them, named MlDBR, showed an expression pattern similar to those of MlDCS and MlCURS (Figure 1A).The activity of the recombinant enzyme was initially tested against p-coumaroyl-CoA 1, feruloyl-CoA 2, bisdemethoxycurcumin 9, and curcumin 10.However, no conversion was detected for any of the phenylpropanoid-CoA and curcuminoid substrates.
We next tested diketide-CoAs as potential substrates by incubating MlDBR with MlDCS1 or MlDCS2 in the presence of their respective starter substrates 1 and 2, malonyl-CoA, and NADPH.A mass peak at m/z 237.0765 [M − H] − (calcd for C 12 H 13 O 5 , 237.0768) was detected in assays comprising MlDBR, MlDCS2, substrate 2, and NADPH (Figure S7).The product was isolated from large-scale incubations and its structure was elucidated by NMR spectroscopy (Figures S20− S23 and Table S1), revealing the previously undescribed dihydroferuloyl-β-keto acid 6a (Figure 1D).The corresponding product, p-dihydrocoumaroyl-β-keto acid 5a, was identified in co-incubations containing MlDBR, MlDCS1, substrate 1, and NADPH through detection of an ion at m/z 207.0664 [M − H] − (calcd for C 11 H 11 O 4 , 207.0663; Figure S7).The recorded mass was consistent with the loss of a methoxy group from 6a, and comparison of MS/MS fragmentation patterns further supported this identification (Figure S7C).Spontaneous hydrolysis of dihydroferuloyldiketide-CoA 6 to 6a in pH 7.5 buffer was expected and underlined by a linear correlation between the formation of 6a and the incubation time of the assay (from 1 to 10 h; Figure S7D).
The addition of MlDBR to the previously conducted coincubation assays of MlDCS1/MlCURS1 and MlDCS2/ MlCURS2 led to the formation of the expected dihydrocurcuminoid products dihydrobisdemethoxycurcumin 7 and dihydrocurcumin 8, respectively (Figure 1B and Figure S5).No traces of curcuminoids 9 and 10 were observed in the assays, indicating the efficiency of this reduction step.These results were consistently replicated through co-infiltration experiments in N. benthamiana using the same enzyme combinations (Figures S5 and S6).We therefore characterize MlDBR as an unusual diketide-CoA-accepting member of the zinc-independ-Scheme 1. Proposed Biosynthetic Pathway of Phenylphenalenones Organic Letters ent medium-chain dehydrogenase/reductase superfamily, containing two conserved NADPH-binding motifs AXXGXXG and GXXS (Figure S8). 27To the best of our knowledge, no DBR accepting diketide-CoAs as substrates has been reported so far.
The PP skeleton was previously proposed to be formed from 4′,3″,4″-trihydroxy linear DH 12 via a reactive ortho-quinone intermediate that undergoes intramolecular Diels−Alder-like cyclization to yield PPs (Figure S9). 29Of the intermediates formed in the above reactions, we considered 7 is a likely precursor of 12 because of its ring substitution pattern (Figure 2A).The conversion of 7 to 12 could proceed via reduction and dehydration to the linear DH 11 with an α, β, γ, δunsaturated ketone function, followed by hydroxylation on the A ring at the C-3″ position (Figure 2A).To demonstrate this conversion, we employed protein extracts from M. lasiocarpa.Due to the difficulties in preparing cell-free homogenates from the hard seeds, we explored roots as a possible alternative.The outermost root layer proved to be another rich source of aromatic metabolites, including detectable amounts of 7 (Figure S10).Incubation of 2′,3′,5′,6′-deuterium-labeled 7  S11B).No activity was observed in control assays lacking NADPH or using heatinactivated proteins.To determine the identity of 11 and 12, these compounds were isolated in greater quantities from root methanolic extracts and structural determination was carried out by NMR (Figures S44−S53).The MS/MS spectra of the isolated compounds matched those obtained from the corresponding products 11 and 12 formed when 7 or d 4 -7 was incubated with root protein extract.An additional peak observed on incubation with 7 and d 4 -7 had the same retention time as 12 but was found to be hydroxylated on the other aromatic ring (12a) (Figure S11).Since 12a did not occur in the methanolic plant extract and could not be further converted into corresponding PPs, we suspect it to be an artifact obtained only by feeding potential intermediates to root extracts.S3) and the PP structure 4′-hydroxylachnanthocarpone 14 (Figures S54− S57 and Table S2).Neither 14 nor 15 has been previously characterized.
The idea that monocyclic DHs may be biosynthetic intermediates between linear DHs and PPs was previously raised after a compound related to 13 was isolated from Fusarium oxysporum-infected Musa acuminata. 7Although no biochemical evidence was provided at the time, the authors were able to chemically transform the compound into a PP structure (Figure S14).
To confirm the final cyclization step in PP biosynthesis, isolated 13 and 15 were incubated separately with the crude root protein extract.HRESIMS analysis of assays containing 13 revealed an increase in two peaks with MS/MS signatures corresponding to those of 4′-hydroxylachnanthocarpone 14 and 3′-hydroxy monocyclic DH 15 (peaks a and b; Figure S15).In contrast, no increase of the PP product peaks was detected in incubations with 15.A possible explanation could be the altered reactivity of 15 caused by the additional hydroxy group at the D-ring.
The identified two-step cyclization of 12 likely proceeds via a sequential 1,4-/1,6-intramolecular Michael addition, initiated by the oxidation of the linear DH substrate to an activated ortho-quinone (Figure S16).Given the importance of the keto−enol tautomerized position 6 in the proposed reaction mechanism, it was expected that PP structures without a 6-OH group, which are commonly found in the root extract (Figure S10), are formed in the final step of the pathway.Interestingly, neither the incubation of 14 nor that of any of the upstream intermediates with the crude enzyme extract yielded the expected product, 4′-hydroxyanigorufone R-1 (Figure S10).This finding may be due to either the limitations of the crude protein extract or the existence of an alternate biosynthetic route to anigorufone-type PPs.The NMR data of 13 and 15 showed that they exist as equilibrium isomeric mixtures, with the Δ 3 double bond in either E (13 and 15) or Z configuration (13a and 15a, Table S3).The Z isomers 13a and 15a are likely intermediates in the formation of a group of diarylheptanoids with a rare bicyclic tetrahydropyran motif.Such diarylheptanoids are known constituents of M. lasiocarpa and Musa × paradisiaca (musellarins A−E; Figure S17). 28,29he present study provides enzymatic and biochemical insights into the biosynthesis of PPs and linear DH derivatives in M. lasiocarpa.Based on the divergent substrate specificities of four type III polyketide synthases and a novel diketide-CoAaccepting medium-chain dehydrogenase/reductase, two distinct assembly lines are responsible for the production of 7 and 8, the dihydrocurcumin-type precursors of PPs and linear DH derivatives, respectively (Scheme 2A).The stepwise transformation of 7 to 4′-hydroxylachnanthocarpone 14 appears to proceed via two sequential intramolecular Michael additions and constitutes the first complete pathway to a PP scaffold (Scheme 2B).

■ ASSOCIATED CONTENT
Dr. Ryan Alam (MPI-CE) for helpful discussions.Seed collection and RNA sequencing were supported by the National Natural Science Foundation of China (32070360).This work was supported by the Max Planck Society.

(d 4
-7) with NADPH in the crude root protein extracts resulted in the appearance of two [M + H] + ions at m/z 299.1578 and 315.1520, matching the expected masses of d 4 -11 (calcd for C 19 H 15 D 4 O 3 , 299.1580) and d 4 -12 (calcd for C 19 H 15 D 4 O 4 , 315.1529) (peaks a and b; Figure

Figure 1 .
Figure 1.Biosynthesis of dihydrocurcuminoids in M. lasiocarpa seeds.(A) Expression profiles of the identified genes representing the FPKM values in the three developmental seed stages.(B) The HRESIMS peak area of products detected in assays using the indicated combinations of purified recombinant enzymes and substrates.Diketide-CoA esters were detected after alkaline hydrolysis.Data are means ± s.e.m. (n = 3).(C) The biosynthetic pathway for the formation of dihydrocurcuminoids (7 and 8) and curcuminoids (9 and 10).(D) Extracted ion chromatograms (EIC) of the spontaneously hydrolyzed dihydrodiketide-CoA products, 5a and 6a.Key COSY and HMBC correlations of 6a are shown above.Compound names in blue, R = H; compound names in green, R = OCH 3 .

Figure 2 .
Figure 2. Metabolic conversions catalyzed by protein extracts from M. lasiocarpa roots.(A) The substrates 7, 11, 12, and 13 were individually incubated with the crude protein extract.Depicted are the HRESIMS peak areas of products detected in incubations containing the compound immediately to the left.Data are means ± s.e.m. (n = 3).All products formed are presented in a unified biosynthetic scheme on the left.Detailed EIC and MS/MS data are shown in Figures S11, S13, S15, and S18.(B) Key COSY and HMBC correlations for 14 and 15/15a.