Seed-coat protective neolignans are produced by the dirigent protein AtDP1 and the laccase AtLAC5 in Arabidopsis

Abstract Lignans/neolignans are generally synthesized from coniferyl alcohol (CA) in the cinnamate/monolignol pathway by oxidation to generate the corresponding radicals with subsequent stereoselective dimerization aided by dirigent proteins (DIRs). Genes encoding oxidases and DIRs for neolignan biosynthesis have not been identified previously. In Arabidopsis thaliana, the DIR AtDP1/AtDIR12 plays an essential role in the 8-O-4′ coupling in neolignan biosynthesis by unequivocal structural determination of the compound missing in the atdp1 mutant as a sinapoylcholine (SC)-conjugated neolignan, erythro-3-{4-[2-hydroxy-2-(4-hydroxy-3-methoxyphenyl)-1-hydroxymethylethoxy]-3,5-dimethoxyphenyl}acryloylcholine. Phylogenetic analyses showed that AtDP1/AtDIR12 belongs to the DIR-a subfamily composed of DIRs for 8-8′ coupling of monolignol radicals. AtDP1/AtDIR12 is specifically expressed in outer integument 1 cells in developing seeds. As a putative oxidase for neolignan biosynthesis, we focused on AtLAC5, a laccase gene coexpressed with AtDP1/AtDIR12. In lac5 mutants, the abundance of feruloylcholine (FC)-conjugated neolignans decreased to a level comparable to those in the atdp1 mutant. In addition, SC/FC-conjugated neolignans were missing in the seeds of mutants defective in SCT/SCPL19, an enzyme that synthesizes SC. These results strongly suggest that AtDP1/AtDIR12 and AtLAC5 are involved in neolignan biosynthesis via SC/FC. A tetrazolium penetration assay showed that seed coat permeability increased in atdp1 mutants, suggesting a protective role of neolignans in A. thaliana seeds.


Introduction
Phenolic compounds including lignins, lignans, neolignans, and flavonoids are major plant secondary metabolites (also known as specialized metabolites; Petersen et al., 2010;Ralph et al., 2019). These metabolites are derived from the phenylpropanoid metabolic pathway and are widely distributed in the plant kingdom (Weng and Chapple, 2010). Phenolic compounds play an important role in the structural enhancement of the plant body, biological defense, environmental stress tolerance, and interactions between plants and other organisms, and also contribute to human health as pharmaceuticals, dietary supplements, flavors, and pigments (Vogt, 2010).
Lignans are a class of dimeric phenylpropanoid metabolites [(C 6 C 3 ) 2 ] with a C8-C8 0 linkage; compounds with all other types of linkages are called neolignans (Moss, 2000;Umezawa, 2003aUmezawa, , 2003b. The biosynthesis of lignans such as pinoresinol, matairesinol, lariciresinol, and secoisolariciresinol is initiated with coniferyl alcohol (CA) in the cinnamate/ monolignol pathway and has been well-studied. CA is oxidized by oxidases such as peroxidases and laccases (LACs) to give rise to the corresponding radical, which is then dimerized stereoselectively at the C8 and C8 0 positions to a lignan, pinoresinol, with the aid of dirigent proteins (DIRs; Davin and Lewis, 2003;Umezawa, 2003a;Suzuki and Umezawa, 2007; Figure 1A). Pinoresinol is further converted to matairesinol via lariciresinol and secoisolariciresinol by pinoresinol/ lariciresinol reductase and secoisolariciresinol dehydrogenase (Davin and Lewis, 2003;Umezawa, 2003a;Suzuki and Umezawa, 2007; Figure 1A). In Arabidopsis (Arabidopsis thaliana), pinoresinol is converted to lariciresinol by pinoresinol reductase (Nakatsubo et al., 2008). The biosynthetic pathway from CA to matairesinol is common to a number of plant species, and a wide variety of lignans are derived from pinoresinol, lariciresinol, secoisolariciresinol, and matairesinol by aromatic substituent modification (Umezawa, 2003a).
In contrast, the biosynthetic pathway for neolignans remains largely unknown. Neolignans are also derived from monolignols generated in the cinnamate/monolignol pathway (Umezawa et al., 2011). The biosynthetic pathways for producing lignan and neolignans diverge at the step for oxidative coupling of monolignols ( Figure 1A). Several studies dealing with crude enzyme preparations mediating the formation of neolignans from phenylpropanoid monomers have been reported (Orr and Lynn, 1992;Katayama and Kado, 1998;Sartorelli et al., 2001;Lourith et al., 2005;Umezawa et al., 2011); however, there are no published reports of genes encoding oxidases and DIRs involved in neolignan biosynthesis.
DIRs belong to a multi-gene family that is classified into six subfamilies (DIR-a, DIR-b/d, DIR-c, DIR-e, DIR-f, and DIRg) based on phylogenetic analyses (Ralph et al., 2007). The DIRs involved in lignan biosynthesis (guiding 8-8 0 coupling) belong to the DIR-a subfamily and are functionally conserved among plant species.

Results
The atdp1 mutants lack putative neolignans We previously reported that the T-DNA insertion mutant designated dp1 (SAIL_60_D04) lacked a compound having a peak at m/z 668 in seeds ). An additional homozygous T-DNA insertion mutant, SALK_062238, was isolated and designated as atdp1-2, and dp1 (SAIL_60_D04) was renamed atdp1-1 (Figure 2A). The T-DNA was inserted between -711 and -702 base pairs of the AtDP1/AtDIR12 promoter region in atdp1-2. Reverse transcription-polymerase chain reaction (RT-PCR) analysis detected AtDP1/AtDIR12 transcripts in atdp1-2, whereas the transcript was not detected in atdp1-1 ( Figure 2B), indicating that atdp1-1 is a null allele of the AtDP1/AtDIR12 gene. No obvious phenotypic differences were observed in the appearances of wild-type and atdp1-1 mutant seeds by scanning electron microscopy ( Figure 2C).
Metabolome analyses showed that a compound with a peak at m/z 506 corresponding to a new putative neolignan compound and the previously reported putative neolignan compound (m/z = 668; Matsuda et al., 2010) were both missing in seeds of the atdp1-1 mutant ( Figure 3). To assess whether the AtDP1/AtDIR12 mutation was responsible for the absence of the two compounds observed in the mutant, we performed a complementation test of the atdp1-1 mutant. The mutant was transformed with two types of AtDP1/AtDIR12 genomic clones (gAtDP1-1 and gAtDP1-2, 3.0 kb/1.9 kb AtDP1/AtDIR12 genomic fragments containing 1.8kb/0.72-kb promoter regions, respectively). Independent transgenic lines carrying gAtDP1-1 and gAtDP1-2 had essentially the same metabolic profile as wild-type plants, although the levels of the target compound were lower in transgenic lines carrying gAtDP1-2 ( Figure 3). These data demonstrated that AtDP1/AtDIR12 is crucial for the biosynthesis of these two putative neolignan compounds in seeds, and that a 1.9-kb AtDP1/AtDIR12 genomic fragment containing 0.72 kb of the promoter region contains at least the minimum region necessary for functional complementation of the metabolic defect in atdp1-1.
To determine the structures of the putatively annotated neolignan compounds missing in the atdp1 mutant, we chemically synthesized authentic standards (racemates) for the two stereoisomers of neolignan 1, namely erythro-SC(4- RT-PCR analysis of transcripts in wild-type (WT, Col-0) and two independent homozygous mutant lines (atdp1-1 and atdp1-2). C, Scanning electron micrograph of epidermal cells from a dry seed of WT (Col-0) and the homozygous atdp1-1 mutant. Scale bars = 100 mm.
O-8)G (1e) and threo-SC(4-O-8)G (1t), as shown in Figure 1C and further described in Materials and methods section. The structural identities of the synthesized compounds were validated by nuclear magnetic resonance (NMR) spectroscopy with 1D 1 H and 13 C and 2D correlation spectroscopy (COSY), heteronuclear single quantum correlation (HSQC), and heteronuclear multiple bond coherence (HMBC) pulse schemes (see Materials and methods section and Supplemental Data Set 1). The stereochemistry of 1e and 1t were based on close comparisons of their NMR chemical shift and coupling constant data with those reported in the literature (Ralph and Helm, 1991;Helm and Ralph, 1992;Wolfram et al., 2010). The retention time on ultra performance liquid chromatography (UPLC; 7.64 min), mass and MS/MS spectra of the missing compound (m/ z = 506) in the atdp1-1 mutant were identical to those of the authentic standard 1e (Figure 4). In addition, we confirmed that MS/MS spectra of the products ( Figure 4) were identical with those of SC(4-O-8)G (1) previously isolated from transgenic Brassica napus seeds (Wolfram et al., 2010). UPLC-MS/MS data suggested that there is no detectable amount of the threo isomer 1t in the Col-0 control seeds. A minor peak (retention time 7.94 min; Figure 4) with MS/MS spectra identical with that of 1e may correspond to the cis-form of 1e because trans-cinnamic acid derivatives are known to isomerize to the cis-form after exposure to sunlight or UV light (Turner et al., 1993;Yang et al., 1999;Wong et al., 2005), and the ratio of the minor peak increased during the purification process in the light.

AtDP1/AtDIR12 is expressed in the outer integument cells of seeds
The accumulation pattern of AtDP1/AtDIR12 transcripts in various organs including seeds at nine developmental stages was investigated by reverse transcription quantitative PCR (RT-qPCR). AtDP1/AtDIR12 transcripts predominantly accumulated in seeds at developmental stages 9 to 11, but not in other organs ( Figure 8A). These data are consistent with the seed-specific accumulation of erythro-SC(4-O-8)G (1e) and the expression profile of AtDP1/AtDIR12 archived on the Arabidopsis eFP browser (http://bar.utoronto.ca/efp/cgibin/efpWeb.cgi). The expression profile of anthocyanidin reductase encoded by BANYULS (BAN), a key enzyme in the proanthocyanidin (PA) biosynthetic pathway, was clearly distinct from that of AtDP1/AtDIR12 during seed development. Compared to AtDP1/AtDIR12, BAN transcripts accumulated in seeds across a broader range of developmental stages (stages 3-10; Figure 8A), suggesting that the two phenylpropanoid-derivative biosynthetic pathways for neolignan and PA are distinctively regulated at the transcriptional level in Arabidopsis seeds.

Oxidases for neolignan biosynthesis
Next, we focused on identifying oxidases required for the formation of the monolignol radicals. LACs and/or peroxidases have been suggested to generate the monolignol radicals for lignan/neolignan; however, no specific genes for neolignans were known. Davin et al. reported that a LAClike oxidase mediated the formation of enantioselective ( + )-pinoresinol in the presence of Forsythia DIR (Davin et al., 1997;Gang et al., 1999). Therefore, we gave priority to LACs. We hypothesized that the target LAC(s) would need to be simultaneously expressed with AtDP1/AtDIR12 to perform the function. Among 17 Arabidopsis LACs, positive correlations (r 4 0.3) of coordinated expression were noted between AtDP1/AtDIR12 and two LACs, i.e. AtLAC15/ TRANSPARENT TESTA 10 (TT10; At5g48100, r = 0.708) and AtLAC5 (At2g40370, r = 0.399). AtLAC15/TT10 and AtLAC5 are major LACs expressed in seeds (Schmid et al., 2005). RT-qPCR showed that transcripts of AtLAC15/TT10 and AtLAC5 are abundant in seeds at developmental stages 9 to 11 as with AtDP1/AtDIR12. The abundance of AtLAC5 transcripts was relatively low compared to those of AtDP1/ AtDIR12 and AtLAC15/TT10 ( Figure 8A).

Localization of AtLAC5 in seeds
Experiments using the AtLAC5 promoter (2,358 bp)-GUS fusion construct suggested that AtLAC5 is localized in the replum and abscission zone in siliques and seed embryos (Turlapati et al., 2011). As the AtLAC5 promoter region was estimated to be 4,140 bp (Turlapati et al., 2011), we transformed plants with constructs containing either the 4,140bp or a 2,000-bp promoter region of AtLAC5 fused to GUS (ProAtLAC5-4140-GUS, ProAtLAC5-2000-GUS). The blue coloration by GUS staining was predominantly observed in the replum, septum, and funiculus and to a lesser extent in the valves of transgenic plants, using either promoter (Figure 10, A, B, E, and F). Staining of the seeds was also observed but it was not clear (Figure 10, C, D, G, and H).
TargetP-2.0 (http://www.cbs.dtu.dk/services/TargetP/) predicts the presence of an N-terminal signal peptide in AtLAC5 (likelihood 0.9998), suggesting that AtLAC5 may sort to compartments in the secretory pathway as is also the case for AtDP1/AtDIR12. Transcriptome data obtained from the Arabidopsis eFP browser showed that AtLAC5 transcripts accumulate exclusively in seeds and particularly in seed coats of mature green seeds (Supplemental Figure  5). The Arabidopsis Seed Coat eFP Browser showed that the accumulation level of AtLAC5 transcripts in ap2-7 mutants is lower compared to wild-type (Supplemental Figure 5). AP2 is a transcription factor required for differentiation of the epidermal and subepidermal palisade layers of the outer integument corresponding to the oi2 and oi1 cell layers, respectively (Western et al., 2001;Dean et al., 2011). These data suggest that AtLAC5 is expressed in the outer integument of seeds. The accumulation level of AtDP1 transcripts in the ap2-7 mutants is also lower than that of wild-type (Supplemental Figure 5). The Arabidopsis Seed Coat eFP browser also suggests that AtDP1/AtDIR12 is expressed earlier than AtLAC5, an observation that is somewhat contradictory to our RT-qPCR data showing that the expression of AtLAC5 is similar to that of AtDP1/AtDIR12 (Figure 8). AtDP1/AtDIR12 may be ready for precise radical coupling prior to the appearance of AtLAC5 or tissue dissection may trigger additional AtDP1/AtDIR12 expression.
The Plant Cell, 2021 Vol. 33, No. 1 (Shirley and Chapple, 2003). A mutant defective in the SCT/SCPL19 locus was designated as sinapoylglucose accumulator 2 (sng2). SC levels are decreased and levels of 1-O-sinapoyl-b-D-glucose and choline are increased in ethyl methanesulfonate-induced sng2 mutant seeds (Shirley et al., 2001). We isolated two homozygous sng2 T-DNA insertion mutants, sng2-3 (SALK_053495C) and sng2-4 (SALK_018120C) to compare with a known sng2-2 mutant (SALK_002255C; Lee et al., 2012; Figure 12; Supplemental Data Set 3). The T-DNA in sng2-3 and sng2-4 was inserted in the exon of SNG2 but between positions + 787 to + 808 and + 1884 to + 1894 base pairs, respectively, from the start codon of SCT/SCPL19. The analyses of neolignans and choline derivatives showed that SC-and FC-conjugated neolignans were clearly missing in the three sng2 mutant seeds as with SC and FC (Figure 12; Supplemental Data Set 3). On the other hand, the accumulation level of benzoylcholine in the three sng2 mutants was similar to that in Col-0. This finding is consistent with the previously published data using the sng2-2 mutant (Lee et al., 2012). Collectively, these data support our notions that SCT/SCPL19 also functions as a feruloylglucose: choline feruloyltransferase in planta, and SC/FC are coupled to CA to yield the SC-and FCconjugated neolignans (Figure 11).

Neolignans affect seed coat permeability
Phenolic compounds such as flavonoids are known to be antioxidants that can prevent the generation of free radicals and subsequent oxidative damage to cells (Sano et al., 2016). Arabidopsis transparent testa mutants that are unable to synthesize flavonoids, including PAs, showed a significant increase in seed coat permeability and reduced germination capacity during long-term seed storage (Debeaujon et al., 2000).
To investigate the effect of decreasing the neolignan content of seeds, we tested the permeability of seed coats by the tetrazolium penetration assay (Berridge et al., 1996;Vishwanath et al., 2013Vishwanath et al., , 2014. Tetrazolium salts are colorless compounds, but they are converted to red products called formazans by endogenous NADH-dependent reductases after they penetrate the seed coat (Berridge et al., 1996). The intensity of red coloration is directly proportional to the seed coat permeability (Vishwanath et al., 2014). Quantitative analyses by extraction of the formazans in 95% ethanol showed that accumulation of formazans in atdp1-1 was the highest and that accumulation in lac5-2, lac5-3, wild-type, and the parental line was 73.7% ± 6.4%, 88.8% ± 17.2%, 28.7% ± 12.2%, 32.5% ± 11.8% of atdp1-1, respectively ( Figure 13B; Supplemental Data Set 4). The permeability of the atdp1-1 mutant complemented with genomic AtDP1/AtDIR12 was similar to that of the wildtype.
We also analyzed the PAs and lignin content to test the possibility that loss of function of AtDP1/AtDIR12 or AtLAC5 might also affect the levels of these compounds, thereby resulting in higher seed coat permeability in the mutants. The PA content as determined by the butanol-HCl assay (Tohge et al., 2005;Kitamura et al., 2010) showed no significant difference between Col-0 and the atdp1 and lac5 mutants in which the neolignan content was substantially lower than wild type ( Figure 13C; Supplemental Data Set 4). The relative lignin content of Col-0 and the mutants was evaluated by the yield of lignin monomers released by analytical thioacidolysis, a reaction that cleaves the quantifiable lignin monomers by chemical cleavage of the major b-O-4 linkages in lignin polymers (Lapierre et al., 1986). Consequently, we detected no lignin reduction in atdp1-1, supporting our notion that the decrease in the neolignan content primarily led to the high permeability of the seed coat in this mutant ( Figure 13D; Supplemental Data Set 4). Additionally, the two lac5 mutants (lac5-2 and lac5-3), also had reduced yields of guaiacyl-type lignin monomers, whereas yields of syringyl-type lignin monomers in the lac5 mutants were similar to those of wild type ( Figure 13D; Supplemental Data Set 4). This result suggests that AtLAC5 may be involved not only in FC-conjugated neolignan biosynthesis but also in guaiacyl-type lignin biosynthesis in Arabidopsis seeds. This result supports the hypothesis that the high permeability of the lac5 seed coats may be attributed to a reduced content of both neolignan and lignin. Figure 11 The proposed biosynthetic pathways of neolignans, lignans, and lignins in Arabidopsis. Asterisks indicate the reader's convenience, each phenolic radical is depicted with its resonance form chosen to allow the radical coupling, 8-O-4 0 and 8-5 0 , to be rationalized; the two FC ᭹ structures shown, for example, are simply two resonance forms of the same phenolic radical. DIR, dirigent protein; laccase, LAC; PRX, peroxidase; PLR, pinoresinol/lariciresinol reductase; PrR, pinoresinol reductase; G, guaiacyl moiety; SC, sinapoylcholine; FC, feruloylcholine; CA, coniferyl alcohol.

AtDP1/AtDIR12 is essential for neolignan biosynthesis
Our results demonstrated that AtDP1/AtDIR12 is essential for the biosynthesis of SC-and FC-conjugated 8-O-4 0 -type neolignans in Arabidopsis (Figures 3, 5). Although the absolute configurations of the AtDP1/AtDIR12-derived neolignans 1e and 4e ( Figure 1C), i.e. whether they have (8S, 7R) and/or (7S, 8R) configurations, are yet to be determined, it is plausible that AtDP1/AtDIR12 guides the regio-and stereo-selective 8-O-4 0 coupling of CA with SC and FC to form optically active products. Intriguingly, the wild-type seeds appeared to accumulate the erythro isomers 1e and 4e exclusively with no detectable amounts of the corresponding threo isomers 1t and 4t (Figure 4; Supplemental Figure 1). This observation strongly suggests that AtDP1/ AtDIR12 may guide not only the radical coupling step but also that in the subsequent rearomatization step in which the quinone methide intermediates are attacked by water during the formation of neolignans 1e and 4e. This intriguing stereochemical mechanism, including which enantiomer of erythro compounds is accumulated and how these reactions are mediated by AtDP1/AtDIR12, should be further addressed together with a detailed three-dimensional protein structure.
To date, several DIRs involved in the stereoselective (enantioselective) 8-8 0 coupling of CA to produce lignan, ( + ) or (-)-pinoresinol, have been identified (Davin et al., 1997;Kim et al., 2002Kim et al., , 2012Kim et al., , 2015Ralph et al., 2007;Pickel et al., 2010;Dalisay et al., 2015), but there have been no reports of DIRs involved in the 8-O-4 0 couplings of monolignols or their analogs for neolignan biosynthesis. Of the six DIR subfamilies, AtDP1/AtDIR12 belongs to the DIR-a family with other DIRs involved in lignan biosynthesis. Phylogenetic analysis of functionally identified DIR-a genes revealed that DIR-a can be classified into two clusters based on their stereoselectivity, ( + )-and (-)-pinoresinol-forming DIRs. AtDP1/AtDIR12, AtDIR13, and AtDIR14 belong to a cluster of (-)-pinoresinol-forming DIRs that include AtDIR6 and AtDIR5 ( Figure 6). Evolutionary relationships of the Arabidopsis DIRa genes proposed by Wang et al. (2013) suggest that AtDIR5 was an ancestral gene and AtDP1/AtDIR12 and AtDIR6 were derived from AtDIR5 by the whole-genome duplication event b or AtDP1/AtDIR12 was derived from AtDIR6 by the whole-genome duplication event a (Supplemental Figure 6; Wang et al., 2013). These data suggest that AtDP1/AtDIR12 may have diverged from (-)-pinoresinol-forming DIRs.
In addition, a variety of (neo)lignans, including aglycones and glycosides, are reportedly stored in Arabidopsis leaf vacuoles (Dima et al., 2015). On the other hand, in flaxseed (Linum usitatissimum), lignans are mainly localized in the secondary walls of sclerite cells in the seed outer integument (Attoumbre et al., 2010). AtDP1/AtDIR12, AtLAC5, and SCT/ SCPL19 that are all involved in seed neolignan biosynthesis were predicted to have signal peptides, suggesting that these proteins may be secreted and sorted into compartments such as vacuoles, tonoplasts, plasma membranes, and cell walls by the secretory pathway; however, the intracellular localization patterns of AtDP1/AtDIR12, AtLAC5, and SCT/ SCPL19 are unknown. All three flaxseed DIRs involved in lignan biosynthesis, LuDIR1, LuDIR5, and LuDIR6, also have signal peptides (Corbin et al., 2018). In Forsythia intermedia stems, DIR(s) are localized to the cell wall (Burlat et al., 2001). Arabidopsis LACs for lignin biosynthesis, AtLAC4 and AtLAC17, are reportedly localized to secondary cell walls (Schuetz et al., 2014), and AtLAC15/TT10 for proanthocyanidin biosynthesis is localized to vacuoles (Pang et al., 2013). Serine carboxypeptidase-like (SCPL) proteins are known to localize in vacuoles (Hause et al., 2002;Mugford et al., 2013). Further investigation is required to elucidate the intracellular localization of AtDP1/AtDIR12, AtLAC5, and SCT/SCPL19 and the sites for neolignan biosynthesis and storage. Figure 13 Seed coat permeability of atdp1 and lac5 mutants. Seed coat permeability was tested by measuring the conversion of tetrazolium salts to red products called formazans. (A) Seeds of wild type (Col-0) and the mutants (atdp1-1 and lac5-3) stained in a tetrazolium salt (TZ) solution or H 2 O after 48 h incubation. (B) The absorbance of ethanol extracts containing formazan from the seeds stained with tetrazolium salts at 485 nm. The intensity of A 485 is directly proportional to the permeability of the seed coat. (C) Soluble and insoluble proanthocyanidins were analyzed by acid hydrolysis as described in Materials and methods section. No significant difference (P 5 0.05) was detected for the insoluble proanthocyanidins. (D) The yield of thioacidolysis-derived p-hydroxyphenyl (H)-, guaiacyl (G)-, and syringyl (S)-type trithioethylpropane monomers released from H-, G-, and S-type lignin polymer units. Data represent the means ± SD (three biological replicates per sample). The means were compared by a one-way ANOVA. Statistically significant differences (P 5 0.05) were identified by Tukey's test and are indicated by lowercase letters to represent differences between groups. n.s., not significant; CWR, cell wall residue.

AtLAC5 is required for the biosynthesis of FCconjugated neolignans
Peroxidases and/or LACs have been reported to be probably involved in monolignol oxidation in lignin and lignan biosynthesis, but no oxidases for neolignan biosynthesis have been identified. Here, we revealed that AtLAC5 is also essential for neolignan biosynthesis. Interestingly, AtLAC5 is involved in the synthesis of the FC-conjugated 8-O-4 0 -type as well as the 8-5 0 -type neolignans but is not involved in the synthesis of SC-conjugated neolignans ( Figure 5). This finding suggests that AtLAC5 is responsible for the oxidation of FC but not for SC; other oxidases may oxidize SC to form SC-conjugated neolignans with CA. LACs are frequently able to catalyze the oxidation of a wide range of substrates in vitro (Sterjiades et al., 1992;Bao et al., 1993); however, the substrate specificity may be more strict in vivo. A seed coatspecific LAC from Cleome hassleriana, ChLAC8, is capable of oxidizing caffeoyl alcohol and sinapyl alcohol but not CA (Wang et al., 2020). This finding is consistent with our hypothesis.
In addition, our lignin analysis data based on thioacidolysis suggested that AtLAC5 is also involved in the oxidation of CA to form guaiacyl-type lignins but is not involved in the oxidation of sinapyl alcohol for syringyl-type lignins in Arabidopsis seed coats ( Figure 13D). These data suggest that AtLAC5 preferentially oxidizes guaiacyl-type substrates, e.g. FC and CA, over syringyl-type substrates, e.g. SC and sinapyl alcohol, although further biochemical studies are needed to support this hypothesis.
The in vitro assays using developing seeds showed that tt10 mutant seeds had a reduced polymerization activity to produce possible 8-O-4 0 , 8-8 0 , and/or 8-5 0 dimers of CA compared to wild-type seeds (Pourcel et al., 2005;Liang et al., 2006). TT10 is localized in oi1 cells (Pourcel et al., 2005); however, no significant changes in the neolignan content were observed in the tt10 (atlac15) mutants. Our results indicated that TT10/AtLAC15 is not involved in neolignan biosynthesis and suggest that the substrate specificity of TT10/AtLAC15 may be also strict in vivo, or TT10/AtLAC15 may be separated spatially from monolignols in the cells.

Physiological roles of neolignans in Arabidopsis seeds
AtDP1/AtDIR12 is specifically expressed in seeds. Analyses using transgenic plants harboring an AtDP1/AtDIR12 promoter fused to AtDP1/AtDIR12 cDNA with YFP or GUS showed that AtDP1/AtDIR12 is localized in the oi1 cells of developing seeds (Figure 8). These results suggest that AtDP1/ AtDIR12-guided neolignans are synthesized in oi1 cells that surround the embryo. The tetrazolium salt assay showed that higher seed coat permeability was observed in the atdp1-1 mutant that is deficient in neolignans ( Figure 5) but has comparable accumulation levels of PAs and lignins as the wild type ( Figure 13). This result suggests that neolignans play a role in protecting seeds against environmental factors. Furthermore, considering the remarkable biological activities of neolignans and lignans as toxins and deterrents for insects and microorganisms (Nitao et al., 1992;Schroeder et al., 2006;Kulik et al., 2014), neolignans in Arabidopsis seeds may also contribute to chemical defense against herbivores and pathogenic microorganisms by their toxicity.
How neolignans affect seed coat permeability is unknown. Arabidopsis mutants unable to synthesize PAs also had a significant increase in seed coat permeability to tetrazolium salts (Debeaujon et al., 2000). PAs accumulate in the innermost layer of the inner integument (endothelium or inner integument 1 layer; Debeaujon et al., 2003). Seeds are firmly protected by phenolic compounds such as neolignans and PAs in at least two cell layers. These compounds, along with lignin in testa cell walls, may form the barriers to penetration.
Our findings herein shed light on new aspects of DIRs and LACs in neolignan biosynthesis. By regulating the stereoselective radical coupling of phenolic compounds, DIRs and LACs are able to produce structurally diverse, specialized phenolic metabolites for chemical defense.

Chemical synthesis of authentic standards for SC(4-O-8)G and FC(4-O-8)G
The authentic, racemic standards for neolignans SC(4-O-8)G (1) and FC(4-O-8)G (4) were synthesized as shown in Figure 1C. The starting compounds, i.e. ethyl sinapate 8SC and ethyl ferulate 8FC were synthesized according to methods described in the literature , and 1-(4-acetoxy-3-methoxyphenyl)-1-ethanone was prepared from acetovanillone using standard acetylation procedures with pyridine and acetic anhydride. Other chemicals were purchased from Wako Pure Chemical (Osaka, Japan) or Nacalai Tesque (Kyoto, Japan) and were used as received. Flash chromatography was performed with Redi Sep Rf silica cartridges on a Combi Flash Rf system (Teledyne ISCO, Lincoln, NE, USA) using an ethyl acetate (EtOAc)/hexane gradient as the eluent and a UV detector (at 254 nm). Preparative HPLC was performed with an XBridge BEH C18 OBD Prep column (130 Å , 5 mm, 10 mm Â 150 mm; Waters Co., Milford, USA) on a Shimadzu 20A HPLC system (Shimadzu Co. Ltd., Kyoto, Japan) using an acetonitrile/water gradient as the eluent and a UV detector (at 254 nm). NMR spectra were recorded on a JEOL JNM-LA400MT FT-NMR system (400 MHz, JEOL, Tokyo, Japan) or a Bruker Biospin AVANCE III 800US system (800 MHz, Bruker Biospin, Billerica, MA, USA). The central solvent peaks were used as internal references (d C /d H , acetone-d 6 , 29.84/2.04; chloroform-d, 77.00/7.26; methanol-d 4 , 49.00/3.31). Standard 1D and 2D (COSY, HSQC, and HMBC) NMR experiments were used for structural assignments. Determination of erythro and threo configurations was based on the comparison of NMR chemical shifts and coupling constant data reported for identical or analogous erythro and threo 8-O-4-type dimer models as reported by Ralph and Helm (1991) and Helm and Ralph (1992).

Compound 10SC
To a suspension of compound 8SC (3.48 mmol) and K 2 CO 3 (3.48 mmol) in 10 mL acetone, compound 9 (3.87 mmol) in 20 mL acetone was added dropwise at room temperature and then refluxed for 16 h. After cooling to room temperature, the reaction mixture was filtered and added to EtOAc/ hexane (1:2, v/v). The organic phase was washed with distilled water and brine, dried over Na 2 SO 4 , and evaporated in vacuo to a residue that was purified by flash chromatography to produce a colorless solid of compound 10SC (57.4% yield).

Compound 10FC
Compound 10FC was synthesized from compounds 8FC and 9 using the above described procedure for compound 10SC and was isolated as a colorless solid (73.3% yield).

Compound 11SC
To a stirred suspension of compound 10SC (1.96 mmol) in 30 mL of dioxane, K 2 CO 3 (14 mmol) and 37% (v/v) formaldehyde (3.92 mmol) were added. After 5 h of stirring at room temperature, the reaction mixture was filtered and concentrated in vacuo to a residue that was purified by flash chromatography to produce a colorless solid of compound 11SC (76.2% yield).

Compound 12SC
To a solution of compound 12SC (1.45 mmol) in 25 mL ethanol, NaBH 4 (10.5 mmol) was added at room temperature. The reaction mixture was stirred under nitrogen gas at room temperature for 5 min, and then the excessive NaBH 4 was quenched by adding acetic acid (1 mL). The reaction mixture was concentrated in vacuo and extracted three times with EtOAc (10 mL) over water (10 mL). The combined organic layer was washed with brine, dried over Na 2 SO 4 , and evaporated in vacuo to a crude oil that was purified by flash chromatography to produce a colorless solid of compound 12SC as a mixture of erythro and threo isomers (95.8% yield, erythro/threo isomer ratio = 7:3 as determined by 1 H-NMR).

Compound 12FC
Compound 12FC was synthesized from 11FC using the above described procedure for compound 12SC and was isolated as a mixture of erythro and threo isomers in the form of a colorless solid (86.3% yield, erythro/threo isomer ratio = 6:4 as determined by 1 H-NMR).

Phylogenetic analysis
Deduced amino acid sequences of DIRs or LACs were aligned using the CLUSTAL W program in MEGA X (version 10.1, https://www.megasoftware.net/; Kumar et al., 2018). A phylogenetic tree was constructed using the Neighbor-Joining method (Saitou and Nei, 1987) in MEGA X with the following parameters: bootstrap test (500 replicates), Poisson model, uniform rates, and complete deletion. Machinereadable files of the phylogenetic analyses used to for Figures 6 and 9 are provided as Supplemental Files 1 and 2, respectively.
Generation and analysis of YFP/GUS reporter lines A 720-bp fragment of the AtDP1/AtDIR12 promoter region or the promoter region including the AtDP1/AtDIR12 coding region was amplified by PCR using primers, At4g11180_CCAC + promoter F and At4g11180_promoter R or At4g11180_R (Supplemental Table 2). Amplified fragments were cloned into the pENTR/D-TOPO vector (Invitrogen) as the entry vector. The resulting plasmid was sequenced to confirm the absence of PCR errors. pHGY, pH35GY derivatives from which the CaMV 35S promoter sequence is deleted (Endo et al., 2009) and pBGGUS (Funakoshi, Japan) were used as the destination vectors. The LR reactions for the binary vectors pKYS405 (pHGY/ ProAtDP1-AtDP1CDS-YFP) and pKYS415 (pBGGUS/ ProAtDP1-GUS) were catalyzed using the Gateway LR Clonase TM II enzyme mix (Invitrogen).
Confocal laser scanning microscopy, light microscopy, and electron microscopy Fluorescent samples were observed using a confocal laser scanning microscope system (Model LSM510 META, Axioplan2 imaging, a Plan-Apochromat lens (63X/1.4 oil DIC; optical slice, 1 lm); Carl Zeiss) with 488-nm excitation and a 505-530 nm band-pass filter for YFP, 488-nm excitation and a 650-nm long-pass filter for chlorophyll autofluorescence and a 25-mV argon laser. Composite figures were prepared using the Zeiss LSM Image Browser software.
Surfaces of dry seeds were observed using a scanning electron microscope (Hitachi TM-1000).

Coexpression analyses
Coexpression analyses were conducted as described previously (Yonekura- Sakakibara et al., 2007Saito et al., 2008). The genes encoding LACs (r 4 0.3) were extracted from the list of genes co-expressed with AtDP1/AtDIR12.

Seed coat permeability test
Seed coat permeability was tested by a tetrazolium penetration assay (Vishwanath et al., 2013(Vishwanath et al., , 2014. Dried seeds (10 mg) were incubated in 250 lL of an aqueous solution with or without 1% (w/v) tetrazolium red (2,3,5-triphenyltetrazolium chloride, Wako, Japan) at 30 C for 48 h in the dark. After incubation, the samples were washed with water, resuspended in 1 mL 95% (v/v) ethanol, and finely ground with a mortar and pestle to extract the formazans. After centrifugation at 15,000g for 3 min, the supernatant fraction was recovered and the absorbance at 485 nm was measured. Three biological replicates were used for the analyses.
Proanthocyanidin analysis PA extraction and acid hydrolysis were performed in triplicate as described previously (Tohge et al., 2005;Kitamura et al., 2010). Mature seeds (10 mg) were homogenized in 0.75 mL of 70% acetone containing 5.26 mM Na 2 S 2 O 5 in a mixer mill (Qiagen Retsch MM300 TissueLyser) for 1 min at 20 Hz, followed by sonication for 20 min. After centrifugation at 15,000g for 5 min, the supernatant fraction was evaporated and resuspended in 1 mL of HCl: butanol: 70% acetone (2:10:3). The absorbance of the solutions before/after hydrolysis at 95 C for 60 min was measured at 545 nm and the difference was treated as the soluble PA fraction. The pellet after extraction with 70% acetone was also evaporated, suspended in the HCl: butanol: 70% acetone solution and hydrolyzed as the insoluble PA fraction.

Supplemental data
The following materials are available in the online version of this article.
Supplemental Figure 2. Multiple alignment of DIRs in the DIR-a family.
Supplemental Figure 3. Multiple alignment of Arabidopsis LACs and functionally identified LACs.
Supplemental Figure 6. A proposed model for gene duplication of the Arabidopsis DIR-a gene family.
Supplemental Table 1. Metabolites detected in seeds of wild-type and the tested mutants.
Supplemental Data Set 2. One-way analysis of variance for Figure 5.
Supplemental Data Set 3. One-way analysis of variance for Figure 12.
Supplemental Data Set 4. One-way analysis of variance for Figure 13.
Supplemental File 1. Alignment corresponding to the phylogenetic analysis in Figure 6.
Supplemental File 2. Alignment corresponding to the phylogenetic analysis in Figure 9.