Dissecting the transcriptional regulation of proanthocyanidin and anthocyanin biosynthesis in soybean (Glycine max)

Summary

Proanthocyanidins (PAs), also known as condensed tannins, are plant natural products that are beneficial for human and livestock health. As one of the largest grown crops in the world, soybean (Glycine max) is widely used as human food and animal feed. Many cultivated soybeans with yellow seed coats lack PAs or anthocyanins, although some soybean cultivars have coloured seed coats that contain these compounds. Here, we analyse the transcriptional control of PA and anthocyanin biosynthesis in soybean. Ectopic expression of the transcription factors (TFs) GmTT2A, GmTT2B, GmMYB5A or R in soybean hairy roots induced the accumulation of PAs (primarily in phloem tissues) or anthocyanins and led to up‐regulation of 1775, 856, 1411 and 1766 genes, respectively, several of which encode enzymes involved in PA biosynthesis. The genes regulated by GmTT2A and GmTT2B partially overlapped, suggesting conserved but potentially divergent roles for these two TFs in regulating PA accumulation in soybean. The two key enzymes anthocyanidin reductase and leucoanthocyanidin reductase were differentially upregulated, by GmTT2A/GmTT2B and GmMYB5A, respectively. Transgenic soybean plants overexpressing GmTT2B or MtLAP1 (a proven up‐regulator of the upstream reactions for production of precursors for PA biosynthesis in legumes) showed increased accumulation of PAs and anthocyanins, respectively, associated with transcriptional reprogramming paralleling the RNA‐seq data collected in soybean hairy roots. Collectively, our results show that engineered PA biosynthesis in soybean exhibits qualitative and spatial differences from the better‐studied model systems Arabidopsis thaliana and Medicago truncatula, and suggest targets for engineering PAs in soybean plants.

Introduction

Proanthocyanidins (PAs, condensed tannins) are flavonoid compounds consisting of oligomeric or polymeric forms of flavan 3‐ols, usually epicatechin and catechin. PAs protect plants from UV damage and provide resistance to pests (Dixon et al., 2013). Their benefits also include antibacterial and antioxidant activities (Mayer et al., 2008; Zhou et al., 2014). PAs can reduce bloating in ruminants when used in animal feed, leading to lower CO₂ emissions (Mueller‐Harvey et al., 2019), and also increase the protein nutrition of ruminants (Jerónimo et al., 2016). For these purposes, breeding for PA‐rich legumes, such as clovers and alfalfa, has become an important goal in agriculture.

Soybean (Glycine max) is one of world’s most important legume crops that provides protein and oils for human consumption and animal feed. Coloured (black or brown) soybean seeds contain anthocyanins and/or PAs in the seed coat (Ito et al., 2013; Kovinich et al., 2011; Todd and Vodkin, 1993). However, the majority of soybean cultivars have a yellow seed coat, which is likely due to a naturally occurring microRNA mechanism that targets and degrades the chalcone synthase (CHS) family gene transcripts (Cho et al., 2013). This mechanism reportedly occurs exclusively in the seed coat. Soybean cultivars with a yellow seed coat and hilum possess the I (inhibitor) locus on chromosome 8 where several CHS genes are located (Wang et al., 1994; Cho et al., 2019; Tuteja et al., 2004), whereas in soybean cultivars with yellow seed coat and black hilum, such as Williams 82, the organization of the CHS genes in this region, the iⁱ locus, is different and does not generate the microRNAs in the seed coat. The R locus is responsible for anthocyanin production in black soybean seed coats (RR); there is a lack of anthocyanins in the near isogenic line (rr) with brown seeds, where a transposon insertion causes loss of function of R (Zabala and Vodkin, 2014).

Several other important genetic markers are associated with seed coat or flower colour in soybean, including O, T and W1 and W3/W4. O was proposed to be the Anthocyanidin Reductase (ANR) gene (Kovinich et al., 2012). T has been suggested to be the Flavonoid 3′‐Hydroxylase (F3′H) gene, which is critical for cyanidin‐based anthocyanin biosynthesis (Toda et al., 2002). W1 and W3/W4 encode Flavonoid 3′5′ Hydroxylase (F3′5′H) and Dihydroflavonol 4‐Reductase (DFR) genes, respectively, that together determine flower colour (Park et al., 2015; Yan et al., 2014; Zabala and Vodkin, 2007). These colour variations are important phenotypic traits for soybean breeding.

In coloured soybean seed coats, PAs are mostly epicatechin‐based, similar to the situation in seed coats of the model plants Arabidopsis thaliana and Medicago truncatula (Ito et al., 2013). The anthocyanins in black soybean seed coats are mostly cyanidin‐based, with a smaller amount of delphinidin‐based anthocyanins (Choung et al., 2001). The biosynthesis and regulation of PAs and anthocyanins in seed coats of A. thaliana and M. truncatula have been well studied (Dixon et al., 2013). Both pathways share several key enzymes at early stages, such as CHS, chalcone isomerase (CHI), flavanone 3‐hydroxylase (F3H), DFR and anthocyanidin synthase (ANS), leading to anthocyanidin production (Figure 1a). ANR is considered a key enzyme for PA biosynthesis, converting anthocyanidin (e.g. cyanidin) to the corresponding flavan 3‐ol (in this case epicatechin). Loss of function of ANR leads to reduction in the levels of both soluble and insoluble PAs in seed coats, with increased accumulation of anthocyanins (Xie et al., 2003). Loss of function of leucoanthocyanidin reductase (LAR) in M. truncatula seed coats leads to a dramatic increase in the level of insoluble PAs and reduced levels of soluble PAs, consistent with a function of this enzyme in PA polymerization (Liu et al., 2016b, 2016a,2016b, 2016a). Several other proteins have also been linked to PA biosynthesis and transport, including the H⁺‐ATPase AHA10, glutathione S‐transferase (GST) and MATE (multidrug and toxin extrusion) transporters (Baxter et al., 2005; Li et al., 2011; Marinova et al., 2007).

Figure 1.

PA and anthocyanin biosynthesis pathways in soybean and expression and function of key PA pathway genes. (a) Simplified PA and anthocyanin biosynthetic pathway in soybean; (b) Transcript levels of GmTT2A, GmTT2B and GmANR1 during soybean seed (cv Clark) development as determined by qRT‐PCR (based on weight of seed in mg). Values represent average transcript levels ± SD (n = 3 biological replicates); (c) Complementation of the Arabidopsis tt2 mutant with GmTT2A or GmTT2B.

Several transcription factors (TFs), namely members of the MYB, bHLH and WD40 families, are known to regulate PA biosynthesis. Their functions were initially characterized by genetic and biochemical analysis of A. thaliana transparent testa mutants whose seeds are light yellow in colour due to lack of oxidized PAs in the seed coat. In many plant species, TT2 (AtMYB123) and its homologs are direct activators of genes encoding ANR, LAR and other enzymes in the PA biosynthesis pathway (Nesi et al., 2001). Furthermore, TT2 forms a ternary complex with TT8 (bHLH) and TTG1 (WD40) to activate genes related to PA biosynthesis (Xu et al., 2014). In addition, MYB repressors, both R2R3‐type and R3‐type MYBs, are thought to regulate PA and anthocyanin biosynthesis through distinct mechanisms to control spatial and temporal accumulation (Huang et al., 2014; Jun et al., 2015; Matsui et al., 2008; Yoshida et al., 2015).

Despite the progress made in our understanding of PA and anthocyanin biosynthesis in species such as A. thaliana, M. truncatula and grapevine (Vitis vinifera), far less is known about the regulation of these processes in soybean seeds. Because of the importance of soybean as an animal feed, there is considerable interest in being able to manipulate the levels of anthocyanins and PAs in the seeds, and possibly introduce them also into leaves. A prerequisite for such attempts is a better understanding of the transcriptional control of the biosynthetic pathways. Here, we have used a combination of genetic, bioinformatic and transcriptomic approaches to address the regulation of PA and anthocyanin biosynthesis in soybean. Our results provide new perspectives for engineering these compounds in this important crop.

Results

Soybean TT2 homologs share conserved functions with related MYB TFs from other plants in regulating PA biosynthesis

To initiate the study, we first characterized the soybean homologs of TT2 TFs, which are key regulators of PA biosynthesis in other plant species (Hancock et al., 2012; Liu et al., 2014; Lu et al., 2017; Yoshida et al., 2008). Using TT2 homologs of A. thaliana (AtTT2) and Trifolium arvense (TaMYB14) to BLAST search against the translated soybean genome database (https://phytozome.jgi.doe.gov/pz/portal.html#!search?show=BLAST&method=Org_Gmax), we found two candidate homologs with high sequence similarity, namely GmTT2A (Glyma.13G109100) and GmTT2B (Glyma.17G050500; Figure S1a). Both proteins possess the conserved TT2 domain (highlighted in Figure S1a) that is shared among TT2 homologs in other plant species. Phylogenetic analysis placed GmTT2A and GmTT2B in the same clade with known PA‐specific MYB TFs such as TaMYB14, AtTT2 and PtMYB134 (Figure S1b). We cloned GmTT2A and GmTT2B from young soybean (cv Clark) seed coat RNA, and studied their transcript profiles using quantitative reverse transcriptase polymerase chain reaction (qRT)‐PCR. The transcript levels of both GmTT2A and GmTT2B genes reached a peak at early stages of seed development and declined sharply afterwards, resembling the expression pattern of the GmANR1 gene (Glyma.08G062000; Figure 1b).

To test whether GmTT2A and GmTT2B share conserved functions with AtTT2, we first ectopically expressed them individually in the A. thaliana tt2 mutant (SALK_005260) under control of the constitutive CaMV 35S promoter. The seed colour phenotype of the tt2 mutant could be complemented by either gene (Figure 1c). The seeds of tt2 lines expressing GmTT2A or GmTT2B showed a brown colour similar to wild type. The restored accumulation of PAs in these brown seeds was demonstrated by staining with p‐dimethylaminocinnamaldehyde (DMACA) reagent, which stains PAs purple (Figure 1c).

To evaluate further the ability of GmTT2A and GmTT2B to induce PAs, we expressed the genes in M. truncatula hairy roots, a system that has been used extensively for dissecting PA signalling (Liu et al., 2014; Lu et al., 2017; Pang et al., 2008). DMACA staining suggested elevated PA content in both GmTT2A‐ and GmTT2B‐OX (overexpressing) lines, compared to the control line expressing the β‐glucuronidase (GUS) gene (Figure S2a), and the transcript levels of both MtANR and MtLAR were highly elevated (Figure S2b). Subsequent quantification of soluble and insoluble PAs further demonstrated the increased accumulation of both classes of PA in GmTT2A‐ and GmTT2B‐OX lines (Figure S2c). Phloroglucinolysis followed by high‐performance liquid chromatography (HPLC) showed that the PAs were mostly epicatechin‐based (Figure S3a), consistent with previous studies in which MtMYB14 and GhMYB36 were expressed in M. truncatula root cultures (Liu et al., 2014; Lu et al., 2017). Analysis by normal phase HPLC showed the presence of PA monomers, dimers and higher oligomers (Figure S3b). The compositional data were further confirmed by LC‐MS/MS analysis (Figure S4), which showed a large increase in free epicatechin in the GmTT2‐expressing lines, and epicatechin starter and extension units generated by phloroglucinolysis. Finally, staining of cross‐sections with DMACA indicated that PAs in GmTT2‐expressing roots were present in both phloem and cortex regions (Figure S2d).

Soybean TT2 homologs function in complexes to regulate PA and anthocyanin biosynthesis

To determine how GmTT2A and GmTT2B regulate the expression of genes encoding key enzymes of PA biosynthesis, we performed transactivation assays using an A. thaliana protoplast transient assay system. Given that TT2 functions with WD40 and TT8 as a complex to regulate ANR and LAR expression in several other plant species, we hypothesized that GmTT2A and GmTT2B might also work in a similar manner. Indeed, when combined with GmTT8 (Glyma.02G147800) and GmWD40 (Glyma.06G136900), both GmTT2A and GmTT2B strongly activated the GmANR1 and GmLAR2 (Glyma.10G204800) promoters, and GmTT2A appeared to have a stronger effect than GmTT2B (Figures 2a,b).

Figure 2.

Effects of different combinations of soybean TFs for activation of the promoters of soybean PA and anthocyanin biosynthesis genes. TFs were tested by transfection, along with a promoter/luciferase reporter gene, into A. thaliana protoplasts. Labels on x‐axes represent TFs that were transfected in each assay. (a) Relative firefly luciferase activity driven by the GmANR1 promoter; (b) Relative firefly luciferase activity driven by GmLAR2 promoter; (c) Relative firefly luciferase activity driven by GmANS promoter. Values represent means ± SD (n = 3 biological replicates).

Other TFs, such as MYB5 and anthocyanin‐related TFs, can also promote PA synthesis (Liu et al., 2014; Xie et al., 2006). MYB5 from M. truncatula was able to induce production of PAs in M. truncatula hairy roots (Liu et al., 2014), and R from soybean was proposed to be responsible for the anthocyanin production in coloured soybean seeds (Zabala and Vodkin, 2014). Therefore, we cloned GmMYB5A (Glyma.14G154400, Figure S5) and R (Glyma.09G235100, Figure S6) from soybean seed coat (cv Clark) RNA and tested their ability to transcriptionally activate PA biosynthesis genes. Transactivation assays showed that, in the presence of GmTT8 and GmWD40, GmMYB5A can activate the GmANR1 and GmLAR2 promoters (Figures 2a,b), and R can activate the GmANS (Glyma.01G214200) promoter (Figure 2c).

Overexpression of GmTT2A, GmTT2B, GmMYB5A and R promotes accumulation of PAs and/or anthocyanins in soybean hairy roots

To test whether the above TFs can stimulate production of PAs and anthocyanins in soybean, we first expressed their open reading frames in soybean hairy roots. GmTT2A, GmTT2B, GmMYB5A, R and GUS were ectopically expressed in both black seed (cv Clark, PI 547438) and yellow seed (Williams 82, PI 518671) hairy root cultures. Compared to GUS control lines, GmTT2A‐, GmTT2B‐ and GmMYB5A‐OX transgenic lines showed no obvious phenotype. However, when R was expressed, black‐seed soybean hairy roots showed a purple colour, and yellow seed soybean hair roots showed a red colour (Figures 3a,c and S7). This difference in colour may be caused by formation of blue delphinidin‐based anthocyanins, in addition to red cyanidin‐based anthocyanins, in black‐seed soybean hairy roots (Figures S8 and S9). Delphinidin is not produced in yellow seed soybean hairy roots where the F3′5′H gene is a null allele (Zabala and Vodkin, 2007). GmTT2A‐ and GmTT2B‐OX lines showed an enhanced purple colour after DMACA staining, whereas no obvious staining was found in GmMYB5A‐OX, R‐OX or GUS control lines. On dissecting the hairy roots, it appeared that the staining was localized exclusively to the phloem region, with no staining in epidermal or cortical cells (Figure 3b). Activity staining of whole roots showed expression of GUS throughout the root, suggesting that the DMACA staining was not limited by 35S‐promoter tissue specificity (Figure 3d). Further analysis showed that both soluble and insoluble PA levels were significantly higher in GmTT2A‐OX and GmTT2B‐OX lines compared with GUS controls (Figure S10a,b). Accurate mass LC‐MS analysis showed that in GUS controls, only free epicatechin monomers could be detected. In GmTT2A‐OX and GmTT2B‐OX lines, whereas epicatechin (potential starter unit) concentrations were reduced compared with GUS controls, epicatechin‐phloroglucinol (extension unit) levels were significantly increased (Figure S10c,d).

Figure 3.

Phenotypes of soybean hairy roots transformed with 35S promoter‐driven GUS, GmTT2A, GmTT2B, R or GmMYB5A. (a) Representative images of transgenic soybean hairy roots (cv Clark); (b) Root phenotypes and cross sections of hairy roots showing PAs accumulating exclusively in vascular tissues of the GmTT2A‐OX line; (c) Cross sections of soybean (W82 and Clark) hairy roots transformed with R. Red to purple colour indicates accumulation of anthocyanins; (d) GUS staining of hairy roots transformed with GUS and GmTT2A, as a control for 35S promoter specificity. Scale bar = 1 mm.

Transcript levels of the anthocyanin biosynthetic genes F3′H, DFR and ANS were significantly higher in GmTT2‐OX and R‐OX lines than in the GUS control, as determined by qRT‐PCR (Figure 4). GmF3′5′H expression was highly induced in R‐OX roots, but only very weakly induced in GmTT2A‐ and GmTT2B‐OX roots, whereas the transcript level in GmMYB5A‐OX was not significantly different from that in the GUS control line (Figure 4). Compared to the GUS control lines, GmANR1 expression was significantly upregulated in GmTT2A‐ and GmTT2B‐OX lines, slightly induced in GmMYB5A‐OX lines, and not detectable in the R‐OX lines. Surprisingly, however, GmLAR2 expression was only induced in the GmMYB5A‐OX lines (Figure 4). GmANR2 (Glyma.08G062100) expression was not detectable in any of the lines, and expression of GmLAR1 (Glyma.20G185700) in GmMYB5A‐OX lines was much lower than that of GmLAR2 (Dataset S5), consistent with the expression patterns of the two soybean LAR genes during seed development (https://phytozome.jgi.doe.gov/pz/portal.html). It therefore appears that GmANR1 and GmLAR2 are the dominantly expressed alleles in their gene families.

Figure 4.

qRT‐PCR analysis of transcript levels of genes involved in soybean PA and anthocyanin biosynthesis in different transgenic soybean hairy root lines. Values represent means ± SD (n = 3 biological replicates). Asterisks indicate significant difference compared to the GUS control at P < 0.05 as determined by Student’s t‐test.

Transcriptome analysis of transgenic soybean hairy roots expressing TFs

To gain further insights into the transcriptional regulatory network of PA and anthocyanin biosynthesis in soybean, we performed RNA‐sequencing analysis using transgenic soybean hairy roots. RNA samples were extracted from black‐seed soybean (cv. Clark) hairy roots overexpressing either GmTT2A, GmTT2B, GmMYB5A, R or GUS (three or four replicates for each line), and RNA‐Seq analysis was performed using an Illumina Nextseq 500 platform.

More than 30 million raw reads were generated from each library, and 88%–94% of trimmed reads could be mapped to the soybean reference genome (Dataset S1). Compared to the GUS control, 1775, 856, 1411 and 1766 genes were upregulated in GmTT2A‐, GmTT2B‐, R‐ and GmMYB5A‐OX lines, respectively, and 2299, 1518, 1690 and 2774 genes were downregulated in these lines, respectively (Figure 5a; Dataset S2); these genes are involved in many aspects of plant development according to gene ontology analysis (Dataset S3). Less than half of the genes differentially expressed in GmTT2A‐OX lines were similarly expressed in GmTT2B‐OX lines, suggesting that GmTT2A and GmTT2B, though sharing conserved functions, are not simply redundant but may also directly or indirectly regulate other pathways or responses. Further categorization of the differentially expressed genes showed that 50 genes were upregulated in all transgenic lines compared to GUS‐expressing lines, and 159 genes were downregulated in all transgenic lines (Dataset S4), the majority encoding proteins outside of the anthocyanin and PA biosynthetic pathways. These results suggest that the regulatory networks encompassing PA and anthocyanin biosynthesis might be partially overlapping in soybean, or that MYB TF over‐expression results in some common off‐target effects in soybean hairy roots.

Figure 5.

RNA‐Seq analysis of soybean hairy roots transformed with PA and anthocyanin pathway TFs. (a) Venn‐diagram showing numbers of genes that are upregulated (upper panel) or downregulated (lower panel) in hairy root lines transformed with GmTT2A, GmTT2B, R or GmMYB5A, compared to the GUS line; (b) Heatmap showing expression levels of selected genes involved in PA or anthocyanin biosynthesis in the different lines.

Compared to the GUS control, the most upregulated genes in GmTT2A‐ and GmTT2B‐OX lines are similar and include F3H, F3′H and ANS. In R‐OX lines, the top genes in the list were ANS, F3H, UFGT (UDP‐glucosyl transferase), MATE, AOMT (S‐adenosyl‐L‐methionine‐dependent methyltransferase), F3′5′H and GST. For GmMYB5A‐OX lines, the most upregulated genes are flavonol synthase (FLS), ANS, AHA10, AOMT and CHI (Dataset S2). Surprisingly, we did not find the signature genes ANR and LAR in the list of significantly upregulated genes. This is likely due to variation of FPKM values in the GUS‐expressing control groups, which influenced the statistical analysis (Dataset S5). However, the heatmap clearly shows the elevated expression levels of GmANR1 transcripts in GmTT2A‐ and GmTT2B‐OX lines, and GmLAR2 transcripts in the GmMYB5A‐OX line (Figure 5b), and this was further confirmed by qRT‐PCR analysis of the same samples used for RNA‐seq.

The transcript data collected from soybean hairy roots showed some striking differences from selected transcript profiles from M. truncatula hairy roots. Expression of GmTT2A and GmTT2B greatly increased transcript levels of GmANR1 but barely induced GmLAR1 or GmLAR2 expression in soybean hairy roots (Figure 4), whereas MtANR and MtLAR transcript levels were elevated by GmTT2A or GmTT2B to similar or comparable levels in M. truncatula roots (Figure S2b). A previous study indicated that MtMYB5 induced MtANR expression to a much higher level than MtLAR in M. truncatula (Liu et al., 2016b, 2016a,2016b, 2016a). Our results showed that in soybean, conversely, GmMYB5A increased expression of GmLAR1 and GmLAR2 to a greater level than GmANR1 (Dataset S5). Compared with GUS‐OX lines, GmANR1 expression was 3.2‐fold, 224‐fold and 179‐fold higher in GmMYB5A‐OX, GmTT2A‐OX and GmTT2B‐OX lines, respectively (Dataset S5). Thus, regulation of ANR and LAR in soybean appears to be via a mechanism distinct from that in M. truncatula.

Along with the genes involved in anthocyanin biosynthesis, two apparently homologous MATE transporter genes, GmMATE82 (Glyma.14G078000) and GmMATE98 (Glyma.17G247400), were highly induced in R‐expressing soybean hairy root lines. Their expression in GmTT2A‐, GmTT2B‐ or GmMYB5A‐ OX lines was much lower than in R‐OX lines, and this was further confirmed by qRT‐PCR (Figure S11). These results suggest potentially unique roles in anthocyanin transport. GmAHA10 expression was induced in GmTT2A‐, GmTT2B‐ and GmMYB5A‐OX lines, but not in the R‐OX line (Figure S11), suggesting specialized functions of downstream genes in PA and anthocyanin biosynthesis which might involve regulation by different TFs.

Analysis of the transcriptome data indicated that a sieve element occlusion protein gene, GmSEOr (Glyma.20G204400), was highly induced in GmTT2A‐ and GmTT2B‐OX lines, but not in R‐OX lines. This was further confirmed by qRT‐PCR (Figure S11). According to an earlier study (Rüping et al., 2010), this family of genes (encoding so‐called P‐proteins) is expressed exclusively in the phloem region, consistent with the location of PAs in GmTT2A‐ and GmTT2B‐OX lines. However, Arabidopsis protoplast transactivation assays using GmTT2 + GmTT8 + GmWD40 together with the GmSEOr promoter indicated that this gene is not under the direct regulation of the PA‐regulatory ternary complex. Induction of GmSEOr expression is therefore likely a secondary response, perhaps caused by accumulation of PAs.

We next compared our results with published transcriptome analysis in M. truncatula overexpressing MtMYB5, MtMYB14 and MtLAP1 (Bond et al., 2016; Liu et al., 2014). A total of 82/45 homologous genes was found to be upregulated in both MtMYB14‐OX and GmTT2A/GmTT2B‐expressing lines, including F3H, ANS and WRKY44; in GmMYB5A‐expressing and MtMYB5‐OX lines, 115 homologous genes were upregulated, including FLS, GL2, WRKY44, TT8 and GmMYB177/203/211/170. There were 31 upregulated homologous genes in R‐expressing and MtLAP1‐OX lines, including GST, MATE82/98, ANS, F3H and TT8 (Dataset S6). These homologous genes upregulated in both M. truncatula and soybean may reflect responses conserved across plant species.

MYB repressors regulate PA and anthocyanin biosynthesis in soybean

MYB repressors are also involved in the regulation of PA and anthocyanin biosynthesis. For example, overexpression of AtMYBL2 (Dubos et al., 2008) or MtMYB2 (Jun et al., 2015) in A. thaliana and M. truncatula, respectively, led to reduced PA and anthocyanin production. A hypothetical model proposes that these repressors occupy the MYB binding sites on TT8 or other bHLH proteins, thus preventing TT2 or MYB5 from forming the normal MYB‐bHLH‐WD40 complex (Jun et al., 2015).

In our soybean hairy root RNA‐Seq data, we found that one R2R3 MYB repressor‐coding gene, GmMYB211 (Glyma.07G216000; Du et al., 2012), was moderately induced in both R‐ and GmMYB5A‐OX lines, whereas another, GmMYB203 (Glyma.14G069100), was highly induced in GmMYB5A‐OX lines and slightly upregulated in R‐OX lines. Each of these two repressor genes has a homologous gene, GmMYB170 (Glyma.02G247100) and GmMYB177 (Glyma.20G013000), both of which were upregulated in GmMYB5A‐OX lines (Dataset S2). Further qRT‐PCR analysis confirmed that GmMYB203 and GmMYB211 transcript levels were higher than GmMYB170 and GmMYB177 transcript levels in R‐OX lines whereas, in GmMYB5A‐OX lines, the transcript levels of the two pairs of homologous genes (GmMYB203 and GmMYB170, GmMYB211 and GmMYB177) were upregulated to a similar level (Figure 6a). According to the soybean tissue‐specific expression database (http://bar.utoronto.ca), GmMYB170 and GmMYB203 are highly expressed in flowers, whereas GmMYB177 and GmMYB211 are highly expressed in seeds, suggesting different roles in flavonoid biosynthesis in different tissues. In addition, ETC1 (Glyma.13G307300), an R3‐type MYB repressor, was induced at high levels in GmTT2A‐, GmTT2B‐ and GmMYB5A‐OX lines, but not in R‐OX lines, suggesting a potential role in regulating PA biosynthesis (Figure 6a). MYB repressors have been shown previously to control the temporal and spatial accumulation of PAs through competition with MYB activators for binding to the promoters of target genes (e.g. Jun et al., 2015).

Figure 6.

Characterization of soybean MYB repressors of PA and anthocyanin biosynthesis. (a) Transcript levels of four potential R2R3 MYB repressors (left panel) and one R3 MYB repressor (GmETC1, right panel) in transgenic soybean hairy root lines expressing R, GmTT2A/B or GmMYB5A. Values are means ± SD (n = 3 biological replicates). Asterisks indicate significant difference compared to the GUS control at P < 0.05 as determined by Student’s t‐test; (b) Arabidopsis protoplast transactivation assays showing relative firefly luciferase activity driven by the GmANR1 or GmGST26 promoters. All reactions in the left, middle and right panels include GmTT2A + GmTT8 + GmWD40, GmMYB5A + GmTT8 + GmWD40, and R + GmTT8 + GmWD40, respectively. Different dosages (molar ratios relative to GmTT2A, GmMYB5A, or R) of added MYB repressors are shown by the numbers on the x‐axes. Reactions without added repressors are used as control groups and shown as 100%, as indicated by numbers on y‐axis; (c) Arabidopsis protoplast transactivation assays with the GmANR1 or GmGST26 promoters, examining mutated versions of the above repressors. All reactions in left, middle and right panels include GmTT2A + GmTT8 + GmWD40, GmMYB5A + GmTT8 + GmWD40, and R + GmTT8 + GmWD40, respectively. Wild type and mutated MYB repressors are shown by the labels on the x‐axis and are added in each reaction at a 1 : 1 ratio to MYB activators, respectively. Reactions without added repressors are used as control groups and shown as 100%, as indicated by numbers on y‐axis. Values are means ± SD (n = 3 biological repressors).

Several key amino acid sequence motifs, notably near the C‐terminal, determine how these MYB repressors function (Jun et al., 2015). Protein sequence alignment shows that GmMYB170 and GmMYB203 are considered homologs of MtMYB2, and GmMYB177 and GmMYB211 are possible homologs of MtMYB530 (Medtr4g485530; Figure S12). GmMYB177 and GmMYB211 both possess LXLXL‐type (m2) and TLLLFR‐type EAR (ethylene‐responsive factor, m3) motifs, whereas GmMYB170 and GmMYB203 lack the TLLLFR‐type motif. Instead, GmMYB170, like MtMYB2, has a semi‐conserved DIDLN motif. To test the functions of these soybean MYB proteins as transcriptional repressors, we performed transactivation assays in A. thaliana protoplasts targeting the activation of the GmANR1 promoter by the GmTT2A‐GmTT8‐GmWD40 complex (Figure 6b). At 1 : 1 ratio of transfected repressor to GmTT2A, together with GmTT8 and GmWD40, GmMYB170 or GmMYB203 could essentially eliminate the luciferase activity driven by the GmANR1 promoter, GmMYB177 reduced the luciferase activity to 10% of the no‐repressor control, and GmMYB211 reduced the activity to roughly 50%. Similar results were observed when we replaced GmTT2A with GmMYB5A or R (the luciferase reporter gene was driven by the GmGST26 promoter when using R; Figure 6b), although GmMYB211 did not inhibit the activity of GmMYB5A or R.

To further characterize these soybean repressors, we performed mutational analyses. Generally, under similar transactivation assay conditions, repressors mutated from LNLDL to ANADA in the m2 motif (Dataset S7), or mutated from TLLLFQ/DIDLN to TAAAFQ/DADAN in the m3 motif (except for GmMYB203), significantly increased the luciferase activity driven by the GmANR1 or GmGST26 promoters when compared to the control groups with native repressors (Figure 6c). The only exception was GmMYB211, mutation of which did not relieve, but even further increased, the inhibition of GmTT2A activity.

To confirm the proposed functions of the soybean MYB repressors as regulators of PA and anthocyanin biosynthesis, we ectopically expressed them in wild‐type A. thaliana plants. The levels of anthocyanins were reduced in hypocotyls of 5‐day‐old transgenic seedlings germinated on MS medium supplied with 1% sucrose, and the PA contents were also decreased in transgenic mature seeds, as indicated by DMACA staining (Figure S13).

Other TFs are differentially expressed in soybean hairy roots accumulating PAs and anthocyanins

In addition to the MYB repressors, we identified other genes encoding TFs that were either upregulated or downregulated in soybean hairy roots overexpressing GmTT2A, GmTT2B, GmMYB5A or R. The total number of genes encoding TFs affected by GmMYB5A, GmTT2A, GmTT2B or R overexpression was 450, 403, 200 and 294, respectively (Dataset S8). The six most affected TF gene families in all transgenic lines included bHLH, MYB, ERF and WRKY families, followed by NAC and C2H2 families. Notably, the two alleles of GmTTG2 (Glyma.03G176600 and Glyma.19G177400), which has previously been shown to be involved in PA biosynthesis (Gonzalez et al., 2016), were significantly upregulated in all transgenic lines. Other TFs with transcript levels upregulated in all transgenic lines include AGAMOUS‐like MADS‐box protein and ZINC FINGER CW‐type coiled‐coil domain protein (Figure S14).

Overexpression of GmTT2B and MtLAP1 promotes accumulation of PAs and anthocyanins in soybean plants

Finally, to test whether the results obtained in soybean hairy roots are translatable to the whole plant, we ectopically expressed GmTT2B and MtLAP1 (a M. truncatula homolog of R) genes in soybean cultivar Williams 82 through Agrobacterium‐mediated transformation, both driven by the 35S promoter. Similar to the effects of R reported above, MtLAP1 induces anthocyanin production in M. truncatula and other legume species (Peel et al., 2009). In total, 22 and 11 independent transgenic lines were obtained overexpressing GmTT2B (ST401) and MtLAP1 (ST402), respectively (selected lines shown in Figure 7a). In ST401, at least six independent lines showed strong expression of GmTT2B in leaves based on qRT‐PCR, and overexpression of GmTT2B upregulated GmANR1 but did not induce GmLAR2 (Figure 7b), consistent with our results from soybean hairy roots. The hilum of transgenic lines with high transcript levels of GmTT2B was grey rather than the normal black colour (Figure 7a). DMACA staining of transgenic seeds overexpressing GmTT2B showed intense purple staining in the hilum region and light staining in the neighbouring seed coat region, whereas in wild‐type seeds and transgenic lines with low GmTT2B expression, the staining was lighter and limited to the hilum region (Figure 7a). This is probably due to the naturally occurring microRNA mechanism in yellow seeded soybean (Cho et al., 2013). To better compare results with wild‐type seeds, we extracted PAs from the pigmented hilum region and non‐pigmented seed coat without the hilum region separately, as described in Cho et al. (2017). Levels of both soluble and insoluble PAs were increased in both dissected hilum and non‐hilum regions of seed coats of GmTT2B‐OX transgenic soybean compared to wild‐type plants (Figure S15a), and LC‐MS analysis confirmed increased levels of epicatechin monomer, epicatechin dimer (procyanidin B2) and epicatechin extension units following phloroglucinolysis (Figure S15b,c). Furthermore, DMACA staining in leaf tissue suggested PA was present in GmTT2B‐OX but not in W82 plants (Figure 7c), although the staining appeared to be predominantly found in vascular tissues, similar to what was found in hairy root samples.

Figure 7.

Characterization of transgenic soybean plants overexpressing GmTT2B (ST401) or MtLAP1 (ST402). (a) Seeds of wild type and transgenic soybean. Top panel, mature soybean seeds; middle and bottom panels, DMACA stained soybean seed coat from longitudinal and cross‐sections; (b) Transcript levels of GmTT2B, GmANR1, MtLAP1 and GmDFR2 as determined by qRT‐PCR. Values are means ± SD (n = 3 biological replicates). Asterisks indicate significant difference compared to the W82 control at P < 0.05 as determined by Student’s t‐test; (c) DMACA stained soybean leaves from W82 and ST401 plants; (d) W82 control and ST402 seedlings showing anthocyanin accumulation; (e) Cross‐sections of leaf tissues of ST402 plants. Red to purple colour indicates accumulation of anthocyanins; (f) Dissections of W82 and ST402 developing seeds at the 400–500 mg stage. Red to purple colour indicates accumulation of anthocyanins.

Expression of MtLAP1 did not lead to upregulation of GmANR1 or GmLAR2 in transgenic plants (data not shown), but induced the genes encoding anthocyanin‐related enzymes such as GmDFR2 (Figure 7b), consistent with our RNA‐Seq data from hairy roots. In eight of the 11 transgenic lines, expression of MtLAP1 led to dark‐coloured seeds (Figure 7a). Young transgenic seedlings also showed coloured hypocotyls and roots (Figure 7d). Cross‐sections of leaves of MtLAP1‐OX lines showed that anthocyanins accumulated in both epidermal and parenchyma cells (Figure 7e), and dissection of seeds showed that the dark colour was a result of anthocyanin accumulation in both embryos and seed coats (Figure 7f). Unlike the grey hilum colour in GmTT2B‐OX lines, the hilum colour remained black in all MtLAP1‐expressing lines (Figure 7a). Levels of both soluble and insoluble PAs were significantly increased in the hilum of seed coats of MtLAP1‐expressing lines compared to wild‐type seed coats (Figure S15), and anthocyanin levels were also significantly increased in leaves of MtLAP1‐expressing plants (Figure S9). Similar to the anthocyanidin profiles in R‐expressing W82 hairy roots, these plants did not accumulate delphinidin‐based anthocyanins (Figure S9).

Discussion

PA biosynthesis is regulated by a complex transcriptional machinery. To engineer soybean plants for enhanced PA production, it is necessary to have a comprehensive understanding of the underlying regulatory mechanisms. Therefore, we set out to gain insights into the transcriptional regulatory network for PA biosynthesis in soybean. While conserved TFs were identified, some distinct functional roles of these TFs in soybean PA biosynthesis were revealed.

GmTT2A, GmTT2B, GmMYB5A and R play different roles during PA and anthocyanin biosynthesis

TT2 and MYB5 are two major PA regulators conserved in many plant species. Loss of function of TT2 or MYB5 leads to reduced PA content in A. thaliana and M. truncatula seeds (Gonzalez et al., 2009; Liu et al., 2014); on the other hand, overexpressing TT2 or MYB5 in M. truncatula hairy roots induces accumulation of PAs (Liu et al., 2014; Pang et al., 2008). MtLAP1 and its soybean homolog R are MYB TFs regulating anthocyanin biosynthesis, which partially overlaps with the PA biosynthesis pathway. These MYB TFs interact with bHLH and WD40 proteins to form ternary complexes (MBW), which primarily induce expressions of Late Biosynthetic Genes (LBG), including DFR, ANS, BAN, TT12 and AHA10 (Xu et al., 2014). In transgenic soybean hairy roots, CHI, F3′H and ANS were among the most upregulated genes in all transgenic lines expressing potential PA or anthocyanin regulatory TFs. Another gene, UDP‐GLYCOSYLTRANSFERASE 71B2‐RELATED (Glyma.03G032500) with unknown function, was also upregulated in all transgenic lines. In M. truncatula, one UDP‐glycosyltransferase, UGT72L1, catalyses the formation of epicatechin 3′‐O‐glucoside (Pang et al., 2008). Whether the soybean UDP‐GLYCOSYLTRANSFERASE 71B2‐RELATED has a similar function remains to be determined.

In M. truncatula, loss of function of ANR results in loss of soluble and insoluble PAs and increased anthocyanin levels, whereas loss of function of LAR leads to loss of soluble PAs and increased insoluble PAs (Liu et al., 2016b, 2016a,2016b, 2016a), suggesting different roles of these two key reductases in PA biosynthesis. In soybean during seed development, GmANR1 and GmLAR2 were expressed at higher levels than GmANR2 and GmLAR1 and, likewise, GmANR1 and GmLAR2 were induced to higher levels in soybean hairy roots overexpressing GmTT2A, GmTT2B or GmMYB5A, suggesting that GmANR1 and GmLAR2 are the dominantly expressed alleles. In soybean hairy roots and transgenic plants, GmANR1 and GmLAR2 were induced by different TFs; GmANR1 was induced by GmTT2A and GmTT2B, whereas GmLAR2 was much more strongly induced by GmMYB5A. In contrast, overexpressing only the TT2 homologs TaMYB14, MtMYB14 or PtMYB134 in white clover, M. truncatula and poplar, respectively, strongly induced expression of both ANR and LAR. Furthermore, expression of R in soybean hairy roots resulted in significantly higher transcript levels of genes shared between PA and anthocyanin biosynthesis pathways, such as ANS and DFR2, than in lines expressing the soybean homologs of the LBG TFs. Some anthocyanin‐specific genes, such as UFGT, GST and MATE, are also highly induced in R‐OX lines.

According to a previous phylogenetic study, GmMATE82 and GmMATE98 are not orthologs of the anthocyanin‐related MtMATE1 (GmMATE81) or MtMATE2 (GmMATE60 and GmMATE117; Liu et al., 2016b, 2016a,2016b, 2016a), suggesting that other transporters might also be involved in anthocyanin transport in soybean. A study by Bond et al. (2016) showed that MtMATE12, the homologous gene of GmMATE82 and GmMATE98 from M. truncatula, was induced on transient overexpression of the legume anthocyanin regulating MYB TF MtLAP1 in leaf tissue. Interestingly, GmUGT78K1 (Glyma.07G183200), which is strongly induced during anthocyanin biosynthesis during black soybean seed development and shows enzymatic activity for anthocyanidins but not flavan 3‐ols (Kovinich et al., 2010), was significantly induced in R‐OX lines, while strongly repressed in GmTT2A/GmTT2B‐ and GmMYB5A‐OX lines. It is possible that in TT2‐OX lines, potential substrates such as cyanidin are preserved for ANR1 thus leading to PA production, whereas in R‐OX lines increased UGT expression facilitates anthocyanin biosynthesis. Together, these data suggest that, in soybean, GmTT2A/B, GmMYB5A and R/MtLAP1 are all involved in inducing LBG genes involved in PA and anthocyanin biosynthesis pathways to maximize PA accumulation.

Some TF encoding genes, such as GmTTG2 (Glyma.03G176600), were induced to a comparable level in transgenic soybean hairy root lines overexpressing GmTT2A/B, GmMYB5A or R, whereas others exhibited dramatically different expression in various lines. For example, in GmMYB5A‐OX lines, several MYB repressor genes, such as GmMYB170, GmMYB203, GmMYB177 and GmMYB211, were upregulated to much higher levels than in other transgenic lines. GmMYB203 and GmMYB211 transcripts were increased to moderate levels in R‐OX lines but barely expressed in GmTT2A‐ or GmTT2B‐OX lines. GmMYB170 and GmMYB203 are the homologous genes of MtMYB2, which is one of the most highly induced repressor genes in M. truncatula hairy roots overexpressing MtMYB5 (Jun et al., 2015; Liu et al., 2014). In GmMYB5A‐OX soybean hairy root lines, GmMYB170 and GmMYB203 were expressed at a higher level than GmMYB177 and GmMYB211. Based on our in vitro transactivation assays, these diversified soybean MYB repressors seem to share conserved functions and regulatory machinery with their M. truncatula homologs, although mutational analysis suggests potentially different mechanisms of action among the repressors. Figure S16 presents a model of the transcriptional control of PAs and anthocyanins in soybean based on the above data. Co‐induction of repressors that can compete with MYB TFs for promoter binding might allow for tight spatial control of PA accumulation (Jun et al., 2015).

Engineering PAs in soybean

Using master TFs to turn on an entire pathway has been developed as strategy to promote accumulation of different natural products in various plant tissues (Hancock et al., 2012; Peel et al., 2009). Indeed, our current results and others (Dixon et al, 2013; James et al., 2017) show that, in many plant species, overexpression of key TFs (e.g. TT2, MYB5 and LAP1/R) regulating the PA and partially overlapping anthocyanin synthesis pathways is sufficient to induce the accumulation of PAs. Compared to M. truncatula hairy roots, soybean hairy roots produce much lower levels of PAs in response to expression of TFs. The PAs in soybean hairy roots accumulate primarily in the phloem region, dramatically different from their more constitutive accumulation in M. truncatula hairy roots.

Our immediate ongoing work is focused on crossing GmTT2‐OX and MtLAP1‐OX plants to further increase PA level in transgenic plants. However, one concern of using TFs for engineering secondary metabolites in plants is the undesired regulation of genes involved in other biological processes and the consequent influence on plant growth and development (Dasgupta et al., 2017). Our RNA‐seq analysis suggests that, besides anticipated genes in PA and anthocyanin biosynthesis, over 200 TFs spanning several families, many of which have unknown functions, are affected in each of the transgenic lines. Further studies are required to understand how other metabolic pathways and biological processes are affected and how this relates to growth impacts.

The demonstration of the effectiveness of LAP1 in inducing anthocyanins in soybean suggests that introduction of LAP1 into GmTT2 overexpressing lines will result in enhanced PA accumulation, as shown for co‐expression of AtTT2 and the anthocyanin regulator PAP1 in Arabidopsis (Xie et al., 2006). In transgenic soybean plants, the accumulation of PAs in the seed coat is limited to small regions and shows a scattered pattern which may be due to the endogenous microRNA mechanism in the yellow soybean seed coat. It will be necessary to develop strategies to address the bottleneck caused by microRNA‐induced gene silencing in yellow seeded soybeans. Currently, these are the varieties that are amenable to direct genetic transformation, and implementation of genetic engineering strategies to direct PAs to the seeds of black or brown seed coat varieties will require crossing from the yellow seeded primary transformants. Despite the challenges in engineering PAs in soybean, our study identifies important players for future studies on PA biosynthesis and engineering in this species.

Experimental procedures

Chemical standards

Catechin, epicatechin, procyanidin B2, cyanidin chloride, kuromanin chloride and 4‐(dimethylamino)‐cinnamaldehyde were purchased from Millipore Sigma (St. Louis, MO).

Plant growth conditions

Arabidopsis plants were grown in growth chambers at 22 °C with a 16 h light/8 h dark regime. Soybean plants were grown in the greenhouse under long day condition (14 h light/ 10 h dark), at 25 °C light/23 °C dark.

Isolation of Arabidopsis protoplasts for transient expression and quantification of luciferase activity was performed as previously described (Yoo et al., 2007).

RNA extraction, quantitative RT‐PCR and gene cloning

RNA was extracted from soybean seed coats (cv Clark) as described (Wang and Vodkin, 1994) and was subsequently used for cloning of GmTT2A, GmTT2B and R open reading frames. RNA extraction for hairy roots and leaf samples was performed using PureLink™ Plant RNA Reagent (Invitrogen) following the manufacturer’s instructions. The iScript™ Select cDNA synthesis Kit (Bio‐Rad) was used to synthesize cDNA, and PCR was performed using Phusion^® High‐Fidelity DNA polymerase (New England Biolabs).

Quantitative RT‐PCR was performed using PowerUp™ SYBR^® Green Master Mix (Thermo Fisher) with an Applied Biosystems QuantStudio 6 Flex Real‐Time PCR system following the manufacturer’s instructions. Sequences of all primers used are given in Dataset S9. ATP‐binding cassette transporter (CONS4) was used as the reference gene in qRT‐PCR (Libault et al., 2008).

Mutated versions of GmMYB170, GmMYB177, GmMYB203 and GmMYB211 were synthesized at Gene Universal Inc (Newark, DE).

M. truncatula and soybean hairy transformation

Open reading frames of the candidate genes GUS, GmTT2A, GmTT2B, GmMYB5A and R were first cloned into the pENTR/D‐TOPO vector and then inserted in the destination vector pB7WG2D using LR clonase™ II enzyme mix (Invitrogen). Plasmids were then transformed into Agrobacterium rhizogenes strains Aquar1 and K599 for M. truncatula and soybean hairy root transformation, respectively, which was then performed as previously described (Cho et al. 2000; Liu et al., 2014).

Arabidopsis complementation assays

Plasmids GmTT2A‐pB7WG2D and GmTT2B‐pB7WG2D were transformed into Agrobacterium tumefaciens strain Gv3101 and transformed in tt2 mutant plants using the floral dip method as described (Clough and Bent, 1998).

RNA‐seq and data analysis

RNA from 3 biological replicates (for GmTT2A‐, GmTT2B‐, and R‐OX and corresponding GUS‐expressing) or four replicates (for GmMYB5A‐OX and corresponding GUS‐expressing) of soybean hairy root lines (cv Clark) was extracted, and libraries were constructed using the TruSeq stranded mRNAseq prep kit and run on an Illumina Nextseq 500 platform (2 × 75 bp).

Raw reads were trimmed and filtered using in‐house perl scrips as described previously (Rao et al., 2014). Clean reads were mapped to the Glycine max Wm82.a2.v1 genome (Schmutz et al., 2010) downloaded from Phytozome v12 (http://phytozome.jgi.doe.gov/). Bowtie v2.3.2.0 (Langmead and Salzberg, 2012) and TopHat v2.1.1 (Kim et al., 2013) with default parameters were used for mapping reads to the reference genome. Cufflinks v2.2.1 (Trapnell et al., 2010) was used to estimate the gene expression levels across all samples based on the FPKM (fragments per kilobase of exon model per million mapped fragments) method. Cuffdiff in the cufflinks package (Trapnell et al., 2013) was used to analyse differential gene expression between two samples. The threshold was set as adjusted P‐value ≤ 0.05 to identify significantly differentially expressed genes. TFs were identified and classified based on their Arabidopsis homologs annotated from the Plant Transcription Factor Database v3.0 (Jin et al., 2014).

Genetic transformation of soybean

Soybean transformation was performed at the Plant Transformation Facility at Iowa State University. The open reading frames of GmTT2B and MtLAP1 were cloned in the pENTR/D‐TOPO vector and inserted in the destination vector pB2GW7 using LR clonase™ II enzyme mix (Invitrogen). The recombined vectors were then transformed into Agrobacterium tumefaciens strain EHA101 for transformation of Williams 82 cultivar.

PA and anthocyanin quantification and HPLC analysis

PAs and anthocyanins in hairy roots and seeds of M. truncatula and soybean were extracted, quantified and analysed for degree of polymerization and starter/extension unit composition using methods previously described (Lu et al. 2017).

LC‐MS analysis of PAs

Metabolite analysis was carried out using an Exion ultra high‐performance liquid chromatography system coupled with a high resolution TripleTOF6600 + mass spectrometer from AB Sciex. Compounds were separated using a C18 Acquity UPLC HSS T3 (100 × 2.1 mm, 1.8 µm) column from Waters. The column compartment and the autosampler were kept at 42 and 15 °C, respectively. Analytes were eluted using a gradient of 0.1% formic acid in water (Solvent A) and 0.1% formic acid in methanol (Solvent B) at a flow rate of 0.4 mL/min. The following gradient was applied: 0–1.0 min, 5.0% B; 1.0–2.0 min, 5.0%–10.0% B; 2.0–7.0 min, 10.0%–28.2% B; 7.0–11.0 min, 28.2%–70.0% B; 11.0–11.1 min, 70.0%–95.0% B; 11.1–13.0, 95% B; 13.0–13.1 min, 95.0%–5.0% B; 13.1–15.0 min, 5.0% B.

The mass spectrometer was set to scan metabolites from m/z 250–1000 amu in negative mode with an ion spray voltage of 4000 V. Accumulation time was 100 msec. Declustering potential and collision energy were 50 and 10 V, respectively. The curtain gas (nitrogen), nebulizing and heating gas were fixed at 30, 60 and 60 psi, respectively. The temperature of the source was 500 °C. An APCI negative calibration solution was delivered by a calibrant delivery system every 10 samples to correct for any mass drift that may occur during the run. MS spectra were acquired using Analyst TF 1.8.1 software. Analysis of the data was performed using Sciex OS software.

Phylogenetic and statistical analysis

Multiple protein sequence alignments were performed using the ClustalW program and phylogenetic trees were constructed by the MEGA6.0 software using the neighbour‐joining method with 1000 bootstrap replicates (Tamura et al. 2013).

Significance analysis was performed using Student’s t‐test, and P < 0.05 was accepted as significant between two groups.

Accession numbers

Genbank Accession numbers for TFs mentioned in this manuscript are GmTT2A (XM_003541253.4), GmTT2B (NM_001355658.1), R (NM_001370267.1), MtLAP1 (FJ199998.1) and GmMYB5A (NM_001254563.1).

The RNA‐seq datasets supporting the results of this article are available in the NCBI Sequence Read Archive (SRA) repository, NCBI SRA accession no. PRJNA604625.

Conflict of interest

The authors declare they have no conflicts of interest.

Authors’ contributions

NL and RAD designed experiments; NL and YL performed experiments; NL, XR, JHJ and RAD analysed data; NL and RAD wrote the manuscript.

Supporting information

Figure S1 Sequence analysis of GmTT2A, GmTT2B and other similar transcription factors.

Figure S2 Ectopic expression of GmTT2A and GmTT2B in M. truncatula hairy roots leads to increased accumulation of PAs.

Figure S3 HPLC analysis of PAs in M. truncatula hairy roots transformed with GmTT2A or GUS.

Figure S4 Analysis of PA monomers, dimers and PA composition in M. truncatula hairy roots transformed with GmTT2A, GmTT2B or GUS.

Figure S5 Amino acid sequence alignment of GmMYB5A, AtMYB5 and MtMYB5.

Figure S6 Amino acid sequence alignment of AtPAP1, R and MtLAP1.

Figure S7 Phenotypes of soybean cv W82 hairy roots transformed with GUS, GmTT2A, GmTT2B or R.

Figure S8 HPLC analysis of anthocyanins extracted from black seed and yellow seed soybean hairy roots transformed with R and GUS.

Figure S9 Analysis of anthocyanin quantity and composition in soybean hairy roots and transgenic plants.

Figure S10 Analysis of PA quantity and composition in soybean hairy roots.

Figure S11 Transcript levels of GmAHA10, GmTT12, GmSEOr and GmMATE82/GmMATE98 in transgenic soybean hairy roots expressing PA/anthocyanin regulators.

Figure S12 Amino acid sequence alignment of MtMYB2, MtMYB530, GmMYB170, GmMYB177, GmMYB203 and GmMYB211.

Figure S13 Phenotypes of Arabidopsis seedlings and seeds over‐expressing GmMYB170, GmMYB203, GmMYB177 or GmMYB211.

Figure S14 Numbers of different classes of TFs that are upregulated or downregulated in transgenic soybean hairy root lines expressing GmTT2A/B, GmMYB5A or R.

Figure S15 Analysis of PA quantity and composition in wild‐type and transgenic soybean seeds.

Figure S16 Transcriptional regulation of PA and anthocyanin biosynthesis in soybean.

Data S1 Summary of RNA‐Seq data of transgenic soybean hairy root lines.

Data S2 List of upregulated and downregulated genes in different transgenic soybean hairy root lines.

Data S3 Gene ontology analysis of differentially expressed genes in different transgenic hairy root lines.

Data S4 List of common upregulated or downregulated genes in all transgenic soybean hairy root lines.

Data S5 FPKM values of select genes in RNA‐Seq data of transgenic soybean hairy root lines.

Data S6 Comparison between soybean hairy root RNA‐Seq data with previously published Medicago transcriptome data.

Data S7 Gene sequences of mutated MYB repressors used in this study. Mutated sequences are highlighted in red.

Data S8 List of different families of transcription factors upregulated or downregulated in different transgenic soybean hairy root lines.

Data S9 Primer sequences used in this study.

Lu, N. , Rao, X. , Li, Y. , Jun, J. H. and Dixon, R. A. (2021) Dissecting the transcriptional regulation of proanthocyanidin and anthocyanin biosynthesis in soybean (Glycine max). Plant Biotechnol. J., 10.1111/pbi.13562