HOMEODOMAIN GLABROUS2 regulates cellulose biosynthesis in seed coat mucilage by activating CELLULOSE SYNTHASE5

Abstract

Numerous proteins involved in cellulose biosynthesis and assembly have been functionally characterized. Nevertheless, we have a limited understanding of the mechanisms underlying the transcriptional regulation of the genes that encode these proteins. Here, we report that HOMEODOMAIN GLABROUS2 (HDG2), a Homeobox-Leucine Zipper IV transcription factor, regulates cellulose biosynthesis in Arabidopsis (Arabidopsis thaliana) seed coat mucilage. HDG2 is a transcriptional activator with the transactivation domain located within its Leucine-Zipper domain. Transcripts of HDG2 were detected specifically in seed coat epidermal cells with peak expression at 10 d postanthesis. Disruptions of HDG2 led to seed coat mucilage with aberrant morphology due to a reduction in its crystalline cellulose content. Electrophoretic mobility shift and yeast one-hybrid assays, together with chromatin immunoprecipitation and quantitative PCR, provided evidence that HDG2 directly activates CELLULOSE SYNTHASE5 (CESA5) expression by binding to the L1-box cis-acting element in its promoter. Overexpression of CESA5 partially rescued the mucilage defects of hdg2-3. Together, our data suggest that HDG2 directly activates CESA5 expression and thus is a positive regulator of cellulose biosynthesis in seed coat mucilage.

The transcription factor HOMEODOMAIN GLABROUS2 directly activates expression of CELLULOSE SYNTHASE5 to regulate cellulose synthesis in Arabidopsis seed coat mucilage.

Introduction

In myxospermous plants including Arabidopsis (Arabidopsis thaliana), seed coat mucilage is synthesized and secreted during the seed epidermal cell differentiation process. After completion of mucilage synthesis, additional components including cellulose are deposited in the middle of the mucilage pocket to form the columella structure (Western et al., 2000; Macquet et al., 2007). The radial cell walls between epidermal cells also undergo secondary thickening (Stork et al., 2010; Mendu et al., 2011). Subsequently, the epidermal cells undergo programmed cell death, and the mucilage is dehydrated and deposited in the seed coat (Mendu et al., 2011). Upon hydration, the mucilage swells rapidly and ruptures the seed coat. This results in the formation of a gelatinous, polysaccharide-rich capsule enclosing the seed (Western et al., 2000, 2001).

The Arabidopsis seed coat epidermis has become a model system for studying the synthesis and regulation of cellulose, hemicellulose, pectin, and the interactions between them (Arsovski et al., 2010; Haughn and Western, 2012; Western, 2012; North et al., 2014; Šola et al., 2019). The mucilage surrounding Arabidopsis seeds exists as a nonadherent outer layer and an adherent inner layer (Western et al., 2001). Rhamnogalacturonan I (RG-I) accounts for ∼90% of the polysaccharides in both layers. Small amounts of cellulose, hemicellulose (galactoglucomannan and xylan), and homogalacturonan are present in both layers (Macquet et al., 2007; Voiniciuc et al., 2015). The adherence of the inner layer to the seed surface is largely determined by interactions between cellulose, xylan, and galactoglucomannan (Sullivan et al., 2011; Yu et al., 2014; Voiniciuc et al., 2015; Hu et al., 2016a, 2016b; Ralet et al., 2016; Griffiths and North, 2017).

Genetic and biochemical analyses of Arabidopsis have led to the identification of numerous transcription factors (TFs) that are involved in regulating epidermal cell differentiation and mucilage biosynthesis (Francoz et al., 2015; Golz et al., 2018). These TFs include MYB5/TRANSPARENT TESTA2 (TT2), TT8/ENHANCER OF GLABRA 3 (EGL3), a member of the basic helix-loop-helix (bHLH) family, and the WRKY TF TRANSPARENT TESTA GLABRA 1 (TTG1). Together these TFs constitute a transcriptional complex MYB-bHLH-WD40 (MBW; Zhang et al., 2003; Western et al., 2004; Bernhardt et al., 2005; Gonzalez et al., 2009; Li et al., 2009). The members of the MBW complex, in concert with APETALA2 (AP2), transcriptionally regulate morphogenesis of seed testa cells and subsequent mucilage formation at least partially by regulating the downstream TTG2 and GLABRA2 (GL2) genes. TTG2 and GL2 encode a WRKY and an HD-ZIP TF, respectively (Western et al., 2001; Gonzalez et al., 2009; Li et al., 2009).

Cellulose is synthesized by cellulose synthase complexes (CSC) located at the plasma membrane (McFarlane et al., 2014). Each complex consists of at least 18 cellulose synthase (CESA) monomers and contains at least three CESA isoforms (Gonneau et al., 2014; Hill et al., 2014; McFarlane et al., 2014). CESA1, CESA3, and CESA6 are largely responsible for synthesizing primary wall cellulose, whereas formation of secondary wall cellulose involves CESA4, CESA7, and CESA8 (Desprez et al., 2007; Persson et al., 2007; Endler and Persson, 2011; McFarlane et al., 2014). Several auxiliary proteins, including COBRA (COB), STELLO1 (STL1), and STELLO2 (STL2), are involved in the assembly and crystallization of cellulose (McFarlane et al., 2014; Sorek et al., 2014; Zhang et al., 2016).

Cellulose synthesis in Arabidopsis seed coat mucilage requires CESA3/Isoxaben Resistant 1 (IXR1) and CESA5/MUCILAGE MODIFIED 3 (MUM3). CESA5/MUM3 (hereafter referred to as CESA5) has the dominant role. Mutating CESA5 leads to a substantial reduction in crystalline cellulose content and a decrease in the adhesion of the mucilage inner layer to the seed surface (Mendu et al., 2011; Sullivan et al., 2011; Griffiths et al., 2015). Similarly, disruptions of CESA3/IXR1 (hereafter referred to as CESA3) result in decreased crystalline cellulose content accompanied by defects in ray structure (Griffiths et al., 2015). CESA2, CESA6, and CESA9 have no direct role in seed mucilage cellulose synthesis. Nevertheless, CESA2 and CESA9 together with CESA5 control secondary thickening of the radial walls of the seed coat epidermal cells (Mendu et al., 2011). Several structural proteins, including COBRA-LIKE 2 (COBL2), SALT OVERLY SENSITIVE 5 (SOS5), FEI2, STL1, and STL2 are also involved in seed mucilage cellulose synthesis and assembly (Harpaz-Saad et al., 2011; Ben-Tov et al., 2015; Zhang et al., 2016). For example, mutating COBL2 leads to a decrease in crystalline cellulose content, which results in decreased mucilage adherence (Ben-Tov et al., 2015). Simultaneous mutations in STL1 and STL2 lead to abnormal ray structure and defective adherence in seed mucilage (Zhang et al., 2016).

The factors controlling the transcriptional regulation of the numerous genes involved in cellulose synthesis and assembly remain largely unknown (Griffiths and North, 2017). Recently, SEEDSTICK (STK), a MADS-box TF, was proposed to be involved in modulating cellulose deposition in seed mucilage (Ezquer et al., 2016). Crystalline cellulose is reduced in stk mutants and is accompanied by a reduction in the expression of CESA5 and FEI2. Additionally, the expression of CESA2, which is involved in the formation of the radial cell wall and columella, is increased in the stk mutant (Ezquer et al., 2016). However, it remains to be determined if STK directly regulates these genes in cellulose deposition. Indeed, additional transcriptional regulators are likely to be required to control cellulose biosynthesis and assembly in seed mucilage.

Here, we have used the Arabidopsis seed coat epidermis to identify HOMEODOMAIN GLABROUS2 (HDG2) as a transcriptional regulator of cellulose biosynthesis. We provide several lines of biochemical and genetic evidence to show HDG2 regulates cellulose synthesis in seed coat mucilage by directly activating the expression of CESA5. These results increase our understanding of the transcriptional regulation mechanisms of cellulose biosynthesis in plants.

Results

HDG2 is coexpressed with CESA5

Previously, we identified a subset of Arabidopsis genes putatively involved in seed coat mucilage polysaccharide biosynthesis and maintenance (Hu et al., 2016a) by re-analyzing microarray datasets generated from laser capture microdissection of selected seed components at different growth stages (Le et al., 2010). Among these genes, the Homeodomain Leucine-Zipper (HD-ZIP) IV subfamily gene HDG2 was identified as a TF that is coexpressed with CESA5 (Supplemental Figure S1).

HDG2 is expressed in developing seed coat

To confirm that HDG2 is expressed in the seed coat, we first used reverse transcription-quantitative PCR (RT-qPCR) analysis (Figure 1A) to examine the transcript abundance of HDG2 in selected tissues including siliques at various growth stages. The expression of HDG2 was much higher in flowers, floral buds, and siliques than in roots and leaves. The transcripts of HDG2 were detected in siliques at 7 d postanthesis (7 DPA), which corresponds to the linear cotyledon stage. The abundance of HDG2 transcripts reached a maximum level in siliques at 10 DPA (bent cotyledon stage). Subsequently, the expression decreased and only relatively low levels of transcript were detected at 13–16 DPA, which corresponds to the mature green stage. The expression pattern of HDG2 in seeds was further studied in wild-type (WT) plants by examining GUS activity driven by the HDG2 promoter. GUS signals were detected exclusively in the seed coat, particularly in the mucilage pocket where mucilage is synthesized (Figure 1B and C).

Figure 1.

Expression of HDG2 in developing seed coat. A, RT-qPCR analysis of HDG2 expression in selected tissues and seeds at various development stages (4–16 DPA). HDG2 expression was normalized to ACT2 and the expression of HDG2 in roots was set as 1.0. Values are mean ± sd of three biological replicates. DPA, days postanthesis. B and C, GUS activity driven by the HDG2 promoter in seed coats. Seeds were dissected at 10 DPA (B) and 13 DPA (C) for the GUS staining assay. C, columella. Bar = 50 µm. D–K, In situ hybridization assay of HDG2 expression in seed coat cells. D, F, H, and J, Hybridization with anti-sense probes. E, G, I, and K, Hybridization with sense probes. D and E, 4 DPA. F and G, 7 DPA. H and I, 10 DPA. J and K, 13 DPA. C, columella, RW, radial cell wall, SG, starch granule. Bars = 20 µm.

We obtained a more detailed expression profile for HDG2 in specific seed coat cells by using an in situ hybridization assay to examine the abundance of its transcript at various seed developmental stages (Figure 1D–K). The hybridization signals for HDG2 were clearly detected from 4 DPA to 13 DPA. Between 4 and 7 DPA, intense signals were present in the epidermal cells of the outer layers of the seed coat, where a large amount of mucilage was being produced (Figure 1D and F). At 10 DPA, the signals accumulated in the cytoplasm of the outermost layer of the seed coat (Figure 1H). At 13 DPA, abundant transcripts of HDG2 were detected in the cytoplasm surrounding the columella and radial cell wall in the outmost layer of the seed coat (Figure 1J). In contrast, the sense probe produced almost no discernible hybridization signals (Figure 1E, G, I, and K). We conclude that HDG2 is expressed in seed coats and its expression corresponds to the timing of mucilage production.

Transcriptional activity of HDG2 is located in the LZ domain

A previous study revealed that HDG2 is a transcriptional activator (Peterson et al., 2013). To determine which domain of HDG2 contributes to transcriptional activity, the conserved domains (homeodomain (HD), leucine-zipper (LZ), steroidogenic acute regulatory protein-related lipid-transfer (START), and START-associated domain (SAD)) were combined in different permutations, fused to the GAL4 DNA binding domain, and transformed into yeast (Saccharomyces cerevisiae) cells (Figure 2A and B). All the transformed yeast cells grew normally on SD (Synthetic Dropout) medium lacking tryptophan (SD/-Trp), a result consistent with successful transformation (Figure 2C). However, only yeast cells containing the LZ domain grew when transferred to SD medium lacking histidine (SD/-His; Figure 2D). Yeast colonies containing the LZ domain construct turned bright blue when grown in the presence of X-α-Gal, indicating that the LacZ reporter gene was activated (Figure 2E). These experiments provide evidence that HDG2 is a transcriptional activator and that the LZ domain is involved in transcriptional activation in plants.

Figure 2.

HDG2 is a transactivation activator. A, Schematic diagram illustrating the different truncations of HDG2 domains fused with GAL4 in pGBKT7. HD, homeodomain. LZ, leucine-zipper. START, steroidogenic acute regulatory protein-related lipid-transfer. SAD, START-associated domain. B to E, Transactivation activity assay of different domain combinations in yeast. B, Graphical representation of the positions of different constructs on the plate. The constructs containing different combinations of HDG2 domains and EV correspond to those shown in A. The recombinant constructs and EV were individually transformed into yeast AH109. C, Growth of yeast transformants containing different constructs on SD/-Trp medium. D, Growth of yeast transformants containing different constructs on SD/-His medium containing 20 mM 3-AT. E, α-galactosidase activity assay of yeast transformants in the presence of X-α-Gal substrate.

HDG2 affects mucilage adherence to the seed surface

To identify functional roles of HDG2 in mucilage biosynthesis and/or modification, we obtained two mutants containing T-DNA inserts in HDG2. RT-PCR experiments confirmed that SALK_127828C (hdg2-2) is a knock-down line whereas SALK_138646C (hdg2-3) is a knock-out mutant, as previously reported (Peterson et al., 2013). Staining the homozygous seeds with ruthenium red revealed that the adherent mucilage of the hdg2 mutants was much thinner than WT (Figure 3A–C and E). It is notable that the mucilage phenotypes of the hdg2 and cesa5/mum3 mutants resemble one another (Figure 3; Sullivan et al., 2011).

Figure 3.

Disruptions of HDG2 lead to defective mucilage adherence. A–D, Ruthenium red staining of seeds after gentle shaking. Seeds of WT (Col-0) (A), hdg2-2 (B), hdg2-3 (C), and cesa5-2 (D) were suspended in water, shaken gently, and then stained for 5 min with ruthenium red. Bar = 150 µm. E, Quantification of the thickness of adherent mucilage (A–D). The average thickness of the adherent mucilage was calculated by measuring 30 seeds for each genotype. Data are mean ± sd. Values marked with different letters indicate significant differences as determined by one-way analysis of variance (ANOVA) with Duncan’s multiple range test (P < 0.05). F, Quantification of mucilage contents. WT and hdg2-3 seeds (100 mg) were sequentially extracted with water and 2M NaOH to obtain nonadherent and adherent mucilage, respectively. Data represent mean ± sd of three biological replicates. Asterisks denote significant differences (P < 0.05) between hdg2-3 and WT based on Student’s t test.

The two hdg2 lines exhibited an almost identical mucilage defect phenotype. Thus, hdg2-3 was selected for further analyses. To confirm that the mucilage defect was caused by disrupting HDG2, we transformed the hdg2-3 mutant with the coding sequence of HDG2 fused with 6× MYC under the control of CaMV 35S promoter. Nineteen independent complemented lines were obtained. As expected, overexpression of HDG2 in hdg2-3 rescued the mucilage-defect phenotype of the mutant (Supplemental Figure S2A and B). These data suggest that HDG2 is involved in maintaining the adherence of seed coat mucilage.

Seed coat development and structure is unaltered in hdg2-3

We next compared the morphology of hdg2-3 and WT seed coats from 4 to 13 DPA to determine if the hdg2-3 mucilage defect resulted from abnormal seed coat development. No substantial differences were observed in any of the developmental stages examined (Supplemental Figure S3). Scanning electron microscopy showed that the seed surfaces of hdg2-3 and WT both contained hexagonal-shaped radial cell walls and a protruding central columella. No other morphological differences between hdg2-3 and WT seeds were discernible (Supplemental Figure S4). These results indicate that disrupting HDG2 does not affect the differentiation of seed coat epidermal cells.

Distribution of mucilage between adherent and nonadherent layers is altered in hdg2-3

To determine if the composition of seed coat mucilage was affected by mutating HDG2, the nonadherent and adherent mucilage layers were solubilized and their monosaccharide compositions were determined. In hdg2-3, the amounts of nonadherent mucilage increased whereas the amounts of adherent mucilage decreased. There was no significant difference in the total amounts of mucilage in hdg2-3 and WT seeds (Figure 3F). These results suggest that even though hdg2-3 and WT seeds produce similar amounts of mucilage, the abundance of mucilage in the nonadherent and adherent layers of the mutant is altered.

The nonadherent and adherent mucilage of hdg2-3 and WT were shown by monosaccharide composition analyses to be composed mainly of rhamnose (Rha) and galacturonic acid (GalA). The amounts of Rha and GalA in hdg2-3 and WT did not differ significantly. Moreover, the amounts of xylose (Xyl) and mannose (Man) were similar in the nonadherent and adherent layers. By contrast, the amounts of galactose (Gal) were significantly higher in both mucilage layers of hdg2-3 compared to the WT (Supplemental Figure S5). Thus, the HDG2 mutation does not interfere with the biosynthesis of pectic components (i.e. RG-I) or major hemicellulosic components (i.e. xylan) in seed mucilage.

We next performed whole mount immunohistochemistry to visually compare the distribution of polysaccharide epitopes in WT and hdg2-3 mucilage (Figure 4). Immunolabeling of WT seeds with CCRC-M36, which recognizes RG-I, revealed a thick and dense fluorescence halo around the seed surface (Figure 4A, C, D, and F). hdg2-3 seeds also gave a fluorescence halo but the surface at the outer periphery was relatively irregular in shape compared to that of the WT (Figure 4G–L). Moreover, the diameter of the halo was somewhat smaller in hdg2-3 than in WT, consistent with a smaller adherent mucilage layer in the mutant seeds. It is also notable that the CCRC-M36 labeling patterns of hdg2-3 were similar to that of the cellulose biosynthesis mutant cesa5-2 (Figure 4M–R). An immunoblotting assay of the nonadherent and adherent mucilage spotted onto a nitrocellulose membrane using the RG-I specific monoclonal antibodies CCRC-M14 and CCRC-M36 revealed that both layers of mucilage of hdg2-3 were distributed in a more diffuse form compared to that in the WT (Supplemental Figure S6). These results are in agreement with the mucilage defect phenotype as revealed by ruthenium red staining of whole seeds (Figure 3B and C).

Figure 4.

Whole mount immunolabeling of RG-I in the hdg2-3 and WT mucilage. A, D, G, J, M, and P, Immunolabeling of RG-I by CCRC-M36. B, E, H, K, N, and Q, Staining of cellulose with S4B. C, F, I, L, O, and R, Merged images of double labeling with CCRC-M36 and S4B. D–F, J–L, and P–R represent the magnified views of specific regions corresponding to A–C, G–I, and M–O, respectively. A–F, WT seeds. G–L, hdg2-3 seeds. M–R, cesa5-2 seeds. Bars = 50 µm.

Crystalline cellulose content is decreased in hdg2-3

Previous studies indicate that crystalline cellulose together with hemicelluloses is required to maintain seed coat mucilage organization (Yu et al., 2014; Voiniciuc et al., 2015; Hu et al., 2016a, 2016b; Ralet et al., 2016). As we found no significant changes in hdg2-3 mucilage pectin or hemicellulose, we next investigated whether the crystalline cellulose content was affected by the HDG2 mutation.

We first performed whole-mount immunolabeling of hdg2-3 seeds with a family 3 carbohydrate-binding module (CBM3a) that binds to crystalline cellulose (Figure 5). The fluorescence signal was uniformly arranged into ray-like structures emanating from the surface of WT seeds (Figure 5A and D). In contrast, the fluorescence signal of hdg2-3 was substantially attenuated and appeared as irregular shapes at the peripheral edge of the seed (Figure 5G and J). Moreover, counter-staining with Pontamine Fast Scarlet 4B (S4B), a dye specific for cellulose, revealed discernible differences between hdg2-3 and WT in the patterns of S4B fluorescence (Figure 5B, E, H, and K). The S4B fluorescence was intense and diffuse and originated from the inner surface of WT seeds. Moreover, intense signals were observed at the tips of columella (Figure 5B and E). By contrast, in hdg2-3 S4B fluorescence was narrower and the halo width was substantially reduced. The diffuse staining of cellulose was much reduced in hdg2-3, and the signal primarily accumulated in areas adjacent to the innermost seed surface (Figure 5H and K). The pattern of cellulose fluorescence of hdg2-3 resembled that of the cesa5-2 mutant, albeit the diffuse distribution of cellulose in the ray structure of hdg2-3 was less pronounced than cesa5-2 (Figure 5N and Q).

Figure 5.

In situ immunolabeling of cellulose in mucilage of hdg2-3 and WT. A, D, G, J, M, and P, Immunolabeling of crystalline cellulose with family 3 carbohydrate-binding module (CBM3a). B, E, H, K, N, and Q, Detection of cellulose stained with S4B. C, F, I, L, O, and R, Merged views of fluorescence of CBM3a and S4B. D–F, J–L, and P–R, the enlarged views of specific regions in A–C, G–I, and M–O, respectively. A–F, WT seeds. G–L, hdg2-3 seeds. M–R, cesa5-2 seeds. Bars = 50 µm.

Hydrated WT and hdg2-3 seeds were examined with a polarized light microscope for the birefringence generated by crystalline cellulose in mucilage. A bright halo surrounding the WT seed was observed, indicating that the WT mucilage contains crystalline cellulose (Figure 6A). In contrast, only an extremely faint halo could be observed near the innermost seed surface of hdg2-3 and cesa5-2 (Figure 6B and C).

Figure 6.

Crystalline cellulose is reduced in the hdg2-3 mucilage. A–C, Birefringence of crystalline cellulose in mucilage of WT (A), hdg2-3 (B), and cesa5-2 (C) seeds. Bar = 150 µm. D, Determination of crystalline cellulose in seeds, demucilaged seeds, and mucilage of hdg2-3, cesa5-2, and WT. Bars represent sd (n = 3). Asterisks indicate significant differences from the WT by Student’s t test (P < 0.05).

We next determined the amounts of crystalline cellulose in seeds and mucilage of hdg2-3 and cesa5-2. There was a significant reduction in the crystalline cellulose content of seeds and mucilage of hdg2-3 and cesa5-2 in comparison to the WT. The crystalline cellulose content was slightly lower in the cesa5-2 mucilage compared to that of hdg2-3 (Figure 6D). Together, these results demonstrate that in the hdg2-3 mucilage the amounts of crystalline cellulose are reduced significantly.

HDG2 activates the expression of CESA5 by binding to its promoter

To ascertain whether HDG2 is involved in signaling pathways governing cellulose metabolism in seed mucilage, we used RT-qPCR analysis to examine the expression of genes likely involved in mucilage cellulose biosynthesis and assembly in hdg2-3 and WT. Our results showed that CESA5, which is responsible for mucilage cellulose biosynthesis (Sullivan et al., 2011), was significantly down-regulated in hdg2-3 siliques from 4 DPA to 13 DPA (Figure 7A). We also detected a significant reduction in CESA5 expression in dissected seed coats of hdg2-3 at 10 and 13 DPA (Supplemental Figure S7A). Our results also showed that FEI2 expression was significantly reduced in seed coats of hdg2-3 (Supplemental Figure S7B). By contrast, the expression of SOS5 and COBL2 were similar in hdg2-3 and WT (Supplemental Figure S7C and D).

Figure 7.

HDG2 directly activates CESA5 expression. A, RT-qPCR analysis of CESA5 expression in hdg2-3 siliques at 4–13 DPA relative to WT. Data are mean ± sd of three biological replicates. Asterisks indicate significant differences from the WT by Student’s t test (P < 0.05). B, Specific binding of HDG2 to the L1-box cis-element in the CESA5 promoter by EMSA. GST served as the control protein. Unlabeled probes with 30-fold molar excess were used as the competitor. C–E, Binding of HDG2 to the L1-box cis-element in a Y1H assay. C, Schematic diagram illustrating the reporter and effector constructs. D, Sequences of three tandemly repeated L1-box and mutated L1-box (mL1-box) cis-elements. E, Y1H assay showing the binding of HDG2 to the L1-box cis-element in CESA5 promoter. Transformants in a 10-fold series were grown on SD/-Leu-Trp and SD/-Leu-Trp-His (with 20 mM 3-AT) medium. F–G, Activation of CESA5 promoter by HDG2 in protoplasts. F, Schematic diagram illustrating the constructs used in the transactivation assay. G, HDG2 activated the GUS reporter driven by the CESA5 promoter. GUS activity was normalized to the LUC activity (internal reference). The EV served as the negative control. Data represent mean ± sd from three biological replicates. Asterisks indicate significant differences from the WT by Student’s t test (P < 0.05). H–I, Binding of HDG2 to CESA5 promoter in planta by ChIP-qPCR assay. H, Schematic representation of the distributions of L1-box cis-element and control region in CESA5 promoter. I, ChIP-qPCR showing significant enrichments of CESA5 promoter fragment containing the L1-box cis-element in immunoprecipitated DNA of proHDG2::HDG2-HA hdg2-3 siliques. Data represent mean ± sd of three replicates. Asterisks indicate significant differences from the WT control (Student’s t test, P < 0.05). J, Overexpression of CESA5 rescued the mucilage defects of hdg2-3. Bars = 150 μm.

We next examined the expression of CESA2, CESA3, CESA6, and CESA9, which are involved in the formation of the radial cell wall and also act redundantly with CESA5 in cellulose biosynthesis (Stork et al., 2010; Mendu et al., 2011). Transcript abundance of the CESA2, 3, 6, and 9 genes was not significantly altered in hdg2-3 (Supplemental Figure S7E–H). We also found no significant changes in the expression of genes involved in xylan synthesis, including IRX7, IRX14, and MUM5 in hdg2-3 (Supplemental Figure S7I–K).

HDG2 binds to L1-box (5′-TAAATGTA-3′) and HAHR-box (5′-CATTAAATG-3′) cis-acting elements (Peterson et al., 2013). We first searched for these cis-acting elements in the promoter sequences of genes that were significantly down-regulated in hdg2-3 and thus may be the potential target genes of HDG2. No HAHR-box cis-acting elements were present in these potential target genes. However, an L1-box motif was present –808 bp upstream of the transcription initiation codon (ATG) in the CESA5 promoter. An electrophoretic mobility shift assay (EMSA) was first performed to determine if HDG2 binds to the L1-box cis-element in the CESA5 promoter. Our results showed a discernible reduction in the electrophoretic mobility of the HDG2 protein, suggesting that the HDG2-GST protein binds to the L1-box in vitro. By contrast, the GST protein control alone exhibited no mobility shift when reacted with the L1-box cis-element. Moreover, addition of the unlabeled competitor probe significantly attenuated the binding affinity of HDG2 to the L1-box cis-element, indicating that the binding of HDG2 to the L1-box is specific (Figure 7B).

A yeast one-hybrid (Y1H) assay was next performed to verify the binding of HDG2 to the L1-box cis-element in the CESA5 promoter. All successfully transformed yeasts grew normally on SD/-Leu-Trp medium. However, only transformants harboring a combination of HDG2 and the L1-box cis-element grew on SD/-Leu-Trp-His medium containing 20 mM 3-amino-1,2,4-triozole (3-AT), indicating that HDG2 must bind to the L1-box cis-element to activate the expression of the HIS3 reporter gene (Figure 7C–E). This result provides further evidence that HDG2 binds to the L1-box cis-element of the CESA5 promoter in vivo.

To determine if HDG2 directly activates CESA5 expression in vivo, we performed a transactivation assay using Arabidopsis mesophyll protoplasts. Different combinations of effector, reporter, and reference constructs were cotransformed into the protoplast and the GUS activity was normalized to the luciferase activity (Figure 7F). The GUS activity in protoplasts cotransformed with CaMV 35S::HDG2 and proCESA5::GUS constructs was activated 3.5-fold relative to the control construct (Figure 7G).

We next performed chromatin immunoprecipitation (ChIP) coupled with quantitative PCR (qPCR) analysis using siliques at 10 DPA of proHDG2::HDG2-HA hdg2-3 transgenic lines to examine the interactions of HDG2 and CESA5 in planta. The ChIP-qPCR analysis revealed an approximately six-fold enrichment of HDG2 in the CESA5 promoter fragment containing the L1-box cis-element compared to the control promoter fragment (Figure 7H and I). These results corroborate the Y1H and transactivation assays and confirm that HDG2 transcriptionally activates the expression of CESA5 by binding to its promoter in planta.

Overexpression of CESA5 rescues the mucilage defect of hdg2-3

To provide genetic evidence that HDG2 affects cellulose biosynthesis in seed mucilage through the direct activation of CESA5 expression, we overexpressed CESA5 under the control of CaMV 35S promoter in hdg2-3. A total of 30 independent complemented lines were obtained and the expression of CESA5 in nine randomly selected lines was examined by RT-qPCR. The result showed that CESA5 expression in the complemented lines was substantially upregulated compared to WT (Supplemental Figure S8A). Ruthenium red staining revealed that compared to hdg2-3, the thickness and area of the adherent mucilage increased substantially in 25 out of 30 transgenic lines. However, these values were not completely restored to WT levels (Figure 7J; Supplemental Figure S8B). Staining with S4B of two representative lines showed that the cellulose in the ray structures was present in a diffuse form similar to WT. In addition, the fluorescence halo of CBM3a in the complemented lines was considerably larger in width than that of hdg2-3 and was similar to WT (Supplemental Figure S8C). These results indicate that overexpressing CESA5 largely rescues the mucilage defects of hdg2-3, which further supports our conclusion that CESA5 is a direct target of HDG2 in cellulose biosynthesis in seed mucilage.

Cross talk between HDG2 and other known TFs

A subset of TFs (e.g. AP2, TTG1, TTG2, GL2, and MYB5) are involved in controlling seed coat differentiation and mucilage production (Koornneef, 1981; Jofuku et al., 1994; Rerie et al., 1994; Johnson et al., 2002; Li et al., 2009). To identify potential relationships between HDG2 and these TFs, we examined their expression in 10 DPA siliques of hdg2-3 by RT-qPCR. Our result showed that the expression of GL2 was significantly reduced in hdg2-3 compared with WT. There was no significant difference in the expression of the other TFs (Supplemental Figure S9A). We also investigated if HDG2 expression is regulated by any of these TFs. To this end, we measured the expression of HDG2 in 7–8 DPA siliques of ap2-1, ttg1-1, ttg2-1, gl2-8, myb5-2, myb61-7, and stk. The transcript abundance of HDG2 was significantly decreased in ttg1-1 and myb5-2 compared to WT. No significant alterations in HDG2 expression were observed in the other mutants (Supplemental Figure S9B). These results indicate that TTG1 and MYB5 act upstream of HDG2, while HDG2 activates the expression of GL2.

HDG2 positively regulates the expression of GL2

GL2 directly regulates the expression of CESA5 in Arabidopsis roots (Tominaga-Wada et al., 2009). This prompted us to examine whether GL2 also regulates CESA5 expression in seed mucilage. Our RT-qPCR result showed that CESA5 expression was not significantly affected in gl2-8 siliques at 10 DPA (Supplemental Figure S10A). We next explored whether HDG2 acts upstream of GL2 and directly regulates its expression. To this end, we examined the expression of GL2 in hdg2-3 siliques at various developmental stages from 4 DPA to 13 DPA by RT-qPCR. Our data showed that GL2 expression was significantly repressed in hdg2-3 siliques at the four stages examined (Supplemental Figure S10B). Examination of the GL2 promoter identified an L1-box cis-element (5′-TAAAATGTA-3′) with an additional Adenine (A) at –1,375 bp upstream of ATG. The results of a Y1H assay suggest that HDG2 does not bind to the TAAAATGTA sequence in the GL2 promoter (Supplemental Figure S10C–E), indicating that GL2 may not be under the direct control of HDG2.

Discussion

Cellulose is a minor component of seed mucilage. Nevertheless, it provides major mechanical support for mucilage structure through its interactions with galactoglucomannan and xylan (Sullivan et al., 2011; Yu et al., 2014; Voiniciuc et al., 2015; Hu et al., 2016a, 2016b; Ralet et al., 2016). Here, we have established that HDG2 is a transcriptional regulator that directly activates the expression of CESA5 in mucilage cellulose biosynthesis.

HDG2 positively regulates cellulose biosynthesis via CESA5 in seed mucilage

RT-qPCR, GUS activity assay, and in situ hybridization results revealed that HDG2 was preferentially expressed in the seed coat with the maximum abundance at 10 DPA, which is consistent with the developmental stage of mucilage biosynthesis and modification (Figure 1). Disruptions of HDG2 led to severe defects in mucilage adherence to the seed surface, which highly resembles the phenotype of cellulose biosynthesis or assembly defective mutants (e.g. cesa5, cobl2, and fei2). Correspondingly, crystalline cellulose content was significantly reduced in hdg2-3 mucilage (Figures 5 and 6). These results indicate that HDG2 may be involved in cellulose biosynthesis or assembly in seed coat mucilage. Furthermore, the Y1H, EMSA, protoplast transactivation assay, and ChIP-qPCR analysis revealed that HDG2 could directly bind to the L1-Box cis-element in the CESA5 promoter and activate its expression in vitro, in vivo, and in planta (Figure 7B–I). In addition, we provided genetic evidence supporting that CESA5 is a direct target of HDG2 as overexpression of CESA5 largely rescued the mucilage defects of hdg2-3 (Figure 7J; Supplemental Figure S8). Taken together, our results indicate that HDG2 positively regulates cellulose biosynthesis in seed coat mucilage by directly activating the expression of CESA5. Apart from the MADS-box TF STK, HDG2 represents another type of TF that regulates cellulose biosynthesis in seed mucilage.

Cellulose biosynthesis is governed by different transcriptional schemes in different tissues

Complexes comprised of different CESAs are required for cellulose synthesis in primary and secondary cell walls (Endler and Persson, 2011; McFarlane et al., 2014). For example, in Arabidopsis CESA1, CESA3, and CESA6 are required to synthesize cellulose for the primary wall, whereas cellulose synthesis for secondary walls involves CESA4, CESA7, and CESA8. The Arabidopsis seed coat mucilage is a specialized cell wall, which is composed of linear RG-I together with smaller amounts of cellulose and hemicellulose (Macquet et al., 2007; Voiniciuc et al., 2015). It has been reported that CESA3 and CESA5 are mainly responsible for the biosynthesis of cellulose in mucilage (Sullivan et al., 2011; Griffiths et al., 2015).

Substantial progress has been made in deciphering the transcriptional regulation mechanisms of cellulose biosynthesis, particularly in secondary cell wall formation. For example, in Arabidopsis MYB46 directly regulates the expression of CESA4, CESA7, and CESA8 (Kim et al., 2013). VASCULAR-RELATED NAC-DOMAIN6 (VND6) is a direct regulator of CESA4, while VND7 regulates the expression of CESA4 and CESA8 (Ohashi-Ito et al., 2010; Yamaguchi et al., 2014). OsMYB58/63 is a direct transcriptional regulator of OsCESA7 involved in secondary cell wall biosynthesis in rice (Oryza sativa;Noda et al., 2015). In addition, STK has been proposed to transcriptionally activate the expression of CESA5 and FEI2 in seed mucilage, but it remains unknown whether this is direct regulation (Ezquer et al., 2016). In the current study, we have shown that HDG2 is a direct activator of CESA5 and affects cellulose biosynthesis in seed coat mucilage. Given that different CSC are involved in cellulose biosynthesis in different cell types, it is not unexpected that different transcriptional schemes governing cellulose biosynthesis separately operate in stem secondary cell wall and seed mucilage.

HDG2 does not regulate other cellulose biosynthesis-related genes directly

In addition to CESA genes, several other genes including COBL2, FEI2, and SOS5 have important roles in cellulose biosynthesis and/or assembly in seed mucilage (Harpaz-Saad et al., 2011; Griffiths et al., 2014; Ben-Tov et al., 2015; Griffiths et al., 2016). Mutations in these genes lead to defects in mucilage adherence that is similar to those of cesa5 and hdg2 mutants. FEI2 and SOS5 mediate cellulosic ray structure formation and mucilage adherence through a separate pathway independent of cellulose formed by CESA5 and COBL2 (Griffiths et al., 2014, 2016). To clarify the genetic relationship between these genes and HDG2, we examined their expression in hdg2-3 seed coats. The expression of FEI2 was significantly down-regulated in hdg2-3 seed coats at 10–13 DPA, while no significant changes in COBL2 and SOS5 were observed in hdg2-3 compared to the WT (Supplemental Figure S7B–D). However, no L1-box or HAHR-box cis-acting elements recognized by HDG2 were present in the promoter sequence of FEI2. Thus, we suggest that the transcriptional regulation of FEI2 is not under the direct control of HDG2. Nevertheless, the possibility could not be excluded that HDG2 regulates the expression of FEI2 by binding to a yet un-recognized cis-acting element. The expression of CESA2, CESA3, CESA6, and CESA9, which are involved in the formation of the radial cell wall and columella in the seed coat (Mendu et al., 2011), was not significantly altered in hdg2-3 (Supplemental Figure S7E–H). This is not unexpected as mutations of HDG2 did not affect the morphogenesis of the radial cell wall and columella (Supplemental Figures S3 and S4).

HDG2 does not regulate xylan biosynthesis-related genes

Mutations affecting three key glycosyltransferases (IRX7, IRX14, and MUM5) involved in xylan backbone biosynthesis and side-chain modification in mucilage also result in severe defects of mucilage adherence to the seed surface (Voiniciuc et al., 2015; Hu et al., 2016a, 2016b; Ralet et al., 2016). The phenotypes of irx7, irx14, and mum5 resemble that of hdg2 and cesa5. However, HDG2 is unlikely to regulate xylan biosynthesis since no significant alterations in Xyl content were detected in the nonadherent or adherent mucilage of hdg2-3 (Supplemental Figure S5A and B). Moreover, the expression of IRX7, IRX14, and MUM5 was not significantly altered in hdg2-3 siliques at various seed coat development stages (Supplemental Figure S7I–K). Nevertheless, an L1-box cis-element is present in the promoter sequences of IRX7 (–1,592 bp) and IRX14 (–2,112 bp). Therefore, it remains to be clarified whether HDG2 binds to the L1-box cis-element and regulates the expression of IRX7 and IRX14 in vivo.

HDG2 acts upstream of GL2 and positively regulates its expression

GL2 directly regulates the expression of CESA5 in roots (Tominaga-Wada et al., 2009). Since mutations in GL2 lead to significantly reduced mucilage production rather than defects in mucilage adherence to the seed surface (Shi et al., 2012), it seems unlikely that GL2 directly regulates cellulose biosynthesis in seed coat mucilage. The reduced mucilage production in gl2 mutants is largely attributed to the defects in differentiation of seed coat epidermal cells, wherein a large amount of mucilage is not synthesized. Unexpectedly, the expression of CESA5 was not significantly affected by gl2-8 siliques (Supplemental Figure S10A). Besides the participation in cellulose biosynthesis, CESA5 also acts redundantly with CESA2 and CESA9 in promoting the secondary thickening of radial walls of seed coat cells (Mendu et al., 2011). Thus, it remains an open question whether GL2 regulates the expression of CESA2 and CESA9 during radial cell wall formation of the seed coat.

Both HDG2 and GL2 belong to the HD-ZIP IV subfamily. Members of HD-ZIP proteins are prone to form homodimers or heterodimers and coordinately regulate the expression of target genes. HDG2 and GL2 individually bind to the promoter of CESA5 and activate gene expression in seed mucilage and roots, respectively (Figure 7; ominaga-Wada et al., 2009). GL2 expression was down-regulated in hdg2-3 siliques, indicating that HDG2 acts upstream of GL2 and positively regulates its expression (Supplemental Figures S9A and S10A). Nevertheless, our Y1H assay showed that HDG2 did not bind to the high consensus L1-box cis-element (5′-TAAAATGTA-3′) in the GL2 promoter (Supplemental Figure S10C and D), suggesting that HDG2 does not directly regulate GL2 expression.

A proposed transcriptional model of HDG2 in seed mucilage production

A hierarchical model comprised of at least three tiers of transcriptional regulators has been proposed for seed coat mucilage synthesis and modification (Golz et al., 2018; Li et al., 2020). This model is comparable to the transcriptional regulatory network proposed for secondary cell wall formation in Arabidopsis stems (Zhong and Ye, 2015). Based on our results, we have developed a model for the transcriptional regulation of cellulose synthesis in seed coat mucilage by HDG2 (Figure 8). APETALA2 (AP2) and TFs of the MWB complex are the first layer of regulators, and regulate TTG2 and GL2 expression in the second layer. HDG2 is a third layer transcriptional activator in seed mucilage production. HDG2 activates the expression CESA5 and FEI2, two genes involved in cellulose synthesis and assembly (Sullivan et al., 2011; Harpaz-Saad et al., 2012). The expression of CESA5 and FEI2 is also activated by STK (Ezquer et al., 2016). The expression of HDG2 itself is activated by TTG1 and MYB5, two transcriptional regulators of the MWB complex (Li et al., 2009; Golz et al., 2018) in the first layer of our model. Moreover, HDG2 also activates GL2 expression which itself regulates mucilage production (Rerie et al., 1994; Shen et al., 2006). However, it is noteworthy that these RT-qPCR analyses were carried out using siliques at 7–10 DPA, which are relatively late stages of development when HDG2 transcripts are abundant, while many of these seed coat differentiations associated TFs play predominant roles at the earlier stages of seed development (4–7 DPA). Therefore it remains to be determined if HDG2 and these TFs regulate each other at the earlier stages of seed development.

Figure 8.

Proposed transcriptional regulatory network involving HDG2 in seed mucilage. AP2 and the MBW complex (TTG1, EGL3/TT8, and MYB5/TT2) are the first layer regulators controlling seed coat differentiation and/or mucilage production. They directly or indirectly activate the expression of TTG2 and GL2 at the second layer. HDG2 is a transcriptional regulator at the third layer. The expression of HDG2 is activated by TTG1 and MYB5. In turn, HDG2 activates the expression of GL2. HDG2 also activates the expression of CESA5 and FEI2 involved in mucilage cellulose production as does STK. Solid lines indicate direct regulation while dashed lines indicate in-direct regulation. The model is based on our results as well as the data of others (Western et al., 2004; Gonzalez et al., 2009; Li et al., 2009; Ezquer et al., 2016).

Elucidation of the role of HDG2 in cellulose biosynthesis increases our understanding of the transcriptional regulation of cell wall biosynthesis in plants. Further investigations are required to elucidate the genetic relationship between HDG2, TTG1, MYB5, and GL2, and to identify transcriptional regulators of the other cellulose biosynthesis and/or assembly genes including CESA3, FEI2, and COBL2 in seed coat mucilage.

Materials and methods

Plant materials and growth conditions

Arabidopsis (Arabidopsis thaliana) Columbia-0 (Col-0) seeds were surface sterilized with 70% (v/v) ethanol containing 0.05% (v/v) Tween-20 and then sown on 1/2 strength MS medium containing 0.7% (w/v) agar. Seeds were stratified for 3 d at 4°C and germinated under a long-day (16 h light vs. 8 h dark) photoperiod with a light intensity of 110 mmol m⁻² s⁻¹. The temperature was kept at 21°C ± 1°C and the relative humidity was maintained at 60%–70%. Seedlings were transferred to soil after 7 d and grown in a growth chamber under the same conditions.

T-DNA insertional lines of hdg2-2 (SALK_127828C), hdg2-3 (SALK_138646C), cesa5-2 (SALK_099008), ap2-1(CS29), ttg1-1 (CS89), ttg2-1 (CS277), gl2-8 (SALK_130213), myb5-2 (SALK_105723C), myb61-7 (SALK_106556C), and stk (SALK_206790C) were obtained from the Arabidopsis Biological Resource Center. Homozygous lines were identified by PCR using the primers listed in Supplemental Table S1.

Generation of transgenic plants

The coding region of HDG2 without the termination codon was ligated downstream of the CaMV 35S promoter at the EcoRI restriction sites in a modified pBI221-MYC vector to generate the overexpression construct. The 1,800 bp promoter of HDG2 was inserted into the pK7GWF2 vector by Gateway recombination reactions to generate the proHDG2::GUS construct. To generate the proHDG2::HDG2-HA construct, the HDG2 promoter (1,800 bp) was inserted upstream of HDG2 fused with a 3 x HA tag at the PstI and BamHI restriction sites in a modified pCAMBIA1300 vector. The constructs were transformed into hdg2-3 and Col-0 by the floral dip method (Clough and Bent, 1998). Transgenic lines from the T2 generation were used for all subsequent analyses.

In situ hybridization

Paraffin sections and in situ hybridization were performed as previously described (Hu et al., 2016a). Briefly, siliques from 4 DPA to 13 DPA were fixed overnight at 4°C in 4% paraformaldehyde. Samples were the dehydrated using a graded series of ethanol and xylene, and embedded at 60°C in Paraplast Plus paraffin (Sigma-Aldrich). HDG2-specific probes were prepared using a 196 bp fragment of the HDG2 coding sequence amplified and ligated into the pGM-T (TIANGEN) vector using primers listed in Supplemental Table S1. The digoxigenin (DIG)-labeled anti-sense and sense probes were transcribed in vitro using a DIG DNA labeling and detection kit (Roche). Silique samples were cut into consecutive sections (8 μm in thick) with an RM2235 microtome (Leica), then mounted onto polysine adhesive slides and dried overnight at 37°C. In situ hybridization was performed using the antisense and sense probes. The hybridization signal was detected using the nitro-blue-tetrazolium/5-bromo-4-chloro-3-indolyl-phosphate alkaline phosphatase color reactions. Sections were observed using a BX51 light microscope (Olympus) and representative images were captured.

RT-qPCR analysis

Arabidopsis siliques were collected at 4, 7, 10, and 13 DPA. Seed coats at 10 and 13 DPA were dissected by crushing them between two glass slides to remove the contained materials. Total RNA was isolated using an optimized TriZol-based method (Meng and Feldman, 2010). First-strand cDNA was synthesized with a RevertAid First Strand cDNA Synthesis Kit (Thermo Fisher Scientific). RT-qPCR reactions were carried out on a LightCycler 480 real-time PCR system (Roche) using the SYBR I Premix ExTaq (TaKaRa). The reactions were performed with three biological replicates. The expression of target genes was normalized against ACTIN2 (ACT2, AT3G18780) using the 2^–ΔΔCt method (Livak and Schmittgen, 2001). All primers are listed in Supplemental Table S1.

Microscopy and histology

SEM analysis was carried out using mature dry seeds and an S4800 scanning electron microscope (HITACHI) with an accelerating voltage of 20 KV. Sputter coating with platinum gold was performed on an E1045 ion sputter coater (Hitachi).

For resin sectioning, developing siliques were dissected and pre-fixed overnight at 4°C in 2.5% (w/v) glutaraldehyde then postfixed for 1 h with 1% (v/v) osmium tetroxide. Samples were dehydrated using a series of increasing concentrations of acetone (30%, 50%, 70%, 80%, 90%, 95%, and 100%) each for 1 h. The acetone was gradually substituted with Spurr’s resin, and then embedded in Spurr’s resin and polymerized at 70°C. The fixed tissue was cut into 1 μm thick sections and stained with 0.5% (w/v) toluidine blue. Seed coat morphology was observed with a BX51 light microscope (Olympus).

To observe the birefringence of crystalline cellulose under polarized light, dry seeds were hydrated with water and then mounted onto glass slides and examined with an Eclipse E600 POL microscope (Nikon).

To detect GUS activity in developing seeds, siliques at 10 DPA were dissected into small segments to facilitate permeation. Chlorophyll was removed by overnight soaking in 95% (v/v) ethanol, followed by overnight incubation with X-Gluc (20 mg mL⁻¹). Images were captured using an SZ61 stereomicroscope (Olympus).

For ruthenium red staining, fully matured dry seeds were immersed in ruthenium red solution (0.01%, m/v) and mixed by gentle pipetting several times. After staining for 5 min, seeds were mounted onto glass slides and observed with a BX51 light microscope (Olympus).

Immunohistochemistry assay

Whole-mount immunolabeling was performed using CCRC-M36 (specific for RG-I) and CBM3a (for crystalline cellulose) as described (Hu et al., 2016a). Donkey anti-mouse secondary antibody conjugated to AlexaFluor488 (Invitrogen) was used. Seeds were counter-stained with the cellulose-specific dye S4B (Sigma-Aldrich) before the observation. Images were acquired using a FluoView FV1000 Laser Scanning Confocal Microscope (Olympus) with excitation at 488 nm and emission at 500–530 nm.

Monosaccharide composition analysis

Mucilage was sequentially extracted from mature dry seeds (100 mg) with water and 2 M NaOH containing 3 mg mL⁻¹ NaBH₄. The 2 M NaOH fraction extract was neutralized with acetic acid and then both extracts dialyzed for 24 h against running tap water and then freeze-dried. Monosaccharide compositions were determined as described (Hu et al., 2016a). Briefly, ∼0.5 mg mucilage was hydrolyzed for 2 h at 121°C in 2 mL of 2 M trifluoroacetic acid (TFA). The released monosaccharides were derivatized for 30 min at 70°C with 0.5 M 1-phenyl-3-methyl-5-pyrazolone in methanol and 0.3 M NaOH. After cooling on ice, the derivatives were neutralized with 0.3 M HCl and then extracted into chloroform. The monosaccharide derivatives were separated on a high-performance liquid chromatography system (Waters) equipped with a Hypersil ODS-2 C18 column (Thermo Fisher Scientific). Rha, GalA, Man, Gal, Glc, and Xyl were used as standards. Detection of monosaccharide was carried out by monitoring the UV fluorescence at 245 nm.

Quantification of crystalline cellulose content

Dry seeds and c seeds (20 mg) were ground to a fine power in liquid nitrogen. The powders were sequentially extracted with ethanol (80% and 100%) and acetone three times, vacuum-dried, and then suspended in 400 μL of distilled water. The suspension was treated for 6 h at 70°C with 7 μL of α-amylase (Megazyme), and for 1 h at 50°C with 7 μL of amyloglucosidase (Megazyme), followed by washes with 80% ethanol and acetone. The destarched alcohol insoluble residue (AIR) was collected by centrifugation for 5 min at 12,000 rpm. Five micrograms of AIR was hydrolyzed for 2 h at 121°C in 0.5 mL of 2 M TFA. The hydrolysate was chilled on ice and centrifuged at 12,000 rpm at 4°C for 10 min. The pellets were dissolved in the Updegraff reagent (HoAc:HNO₃:H₂O, 8:1:2, v/v/v, 1 mL) and kept for 30 min at 100°C to hydrolyze any remaining lignin, pectin and hemicellulose. After centrifugation at 12,000 rpm for 10 min, the pellets were washed three times with acetone and vacuum-dried. The residue, which is comprised almost entirely of crystalline cellulose, was suspended in 0.2 mL of 72% H₂SO₄ to dissolve the cellulose. The amounts of Glc were then determined colorimetrically using the phenol-sulfuric acid method. A standard curve was plotted using a Glc solution (0.1 mg mL⁻¹) and the dehydration factor was set to 0.9. The crystalline cellulose content in seed mucilage was calculated by subtracting the demucilaged seeds from seeds.

Electrophoretic mobility shift assay

The coding sequence of HDG2 was ligated into pGEX-4T-1 at the SmaI and XhoI restriction sites to generate the HDG2-GST fusion construct. The construct was introduced into Escherichia coli strain BL21 (DE3). The transformants were cultured to the exponential growth period, and protein expression was induced by adding 0.3 mmol L⁻¹ Isopropyl-β-d-thiogalactoside at 16°C for 12 h at 180 rpm. The bacteria were collected by centrifugation at 6,000 rpm, resuspended in 10 mM Phosphate buffer solution, followed by sonication with an ultrasonic processor (Sonics). The protein was purified using Glutathione Sepharose beads (4Fast Flow, GE Healthcare) and eluted with 10 mM glutathione. In vitro binding of the HDG2-GST fusion protein and the biotin-labeled CESA5 promoter containing the L1-Box (5′-TAAATGTA-3′) cis-element was performed using a LightShift Chemiluminescent EMSA Kit (Thermo Fisher Scientific) following the protocol provided by the manufacturer. Briefly, labeled DNA probes were incubated 30 min with 100 ng of HDG2-GST in binding buffer (10 mM Tris pH 7.5, 50 mM KCl, 1 mM DTT, 2.5% glycerol, 5 mM MgCl₂, 0.05% Nonidet P-40 and 100 ng μL⁻¹ poly (dl-dC)). The unbound probes were separated by polyacrylamide gel electrophoresis. The DNA-protein complex was transferred onto a nitrocellulose membrane and the hybridization signal detected by chemiluminescence. The GST protein was used as the negative control and unlabeled probes in 30-fold molar excess were used for the competition binding assay.

Transactivation assay

The full coding region and different domains of HDG2 (i.e. HD, ZIP, START, and SAD) were individually amplified (Supplemental Table S1) and inserted into pGBKT7 at the NdeI and XmaI sites to generate different fusion proteins with the GAL4 DNA binding domain. The recombinant constructs were separately transformed into yeast (S. cerevisiae) strain AH109. The transformed yeast cells were sequentially dropped onto SD/-Trp and SD/-His (with 20 mM 3-AT) medium. α-Glucosidase activity was determined using 5-bromo-4-chloro-3-indoxyl-α-d-galactopyranoside (X-α-Gal, 20 mg mL⁻¹) as the substrate.

Y1H assay

The entire coding sequence of HDG2 was amplified and inserted into the pGADT7 vector at the EcoRI site to generate the prey construct. Three tandem repeats of L1-box cis-element in the CESA5 promoter were synthesized, denaturized, and ligated into the pHIS2.1 vector at the EcoRI site as the bait construct. The prey and bait constructs were cotransformed into yeast strain Y187. The transformants were sequentially cultivated on SD/-Leu-Trp and SD/-Leu-Trp-His (with 20 mM 3-AT) medium. The empty vector (EV) was used as the negative control.

Transient expression assay

The CESA5 promoter (1,800 bp) was inserted upstream of the GUS reporter gene in a modified pBI121 vector (CaMV 35S removed) to generate the reporter construct. The coding sequence of HDG2 was inserted downstream of the CaMV 35S promoter in a modified pBI221 vector (GUS removed) to generate the effector construct. The LUC reporter gene driven by the CaMV 35S promoter was used as the internal reference control. The effector and reporter constructs were cotransformed into Arabidopsis leaf protoplasts via polyethyleneglycol-mediated transformation. After incubation overnight in the dark, the fold-change of reporter gene expression was determined based on the GUS activity relative to the LUC intensity. The GUS activity of protoplasts transformed with the empty effector vector was used as the negative control. The experiment was carried out in three biological replicates.

ChIP-qPCR

Siliques of proHDG2::HDG2-HA hdg2-3 transgenic lines were collected at 10–13 DPA. Samples were cross-linked with 1% formaldehyde and quenched by 0.125 M glycine under a vacuum. Chromatin was extracted with nuclei extraction buffer and then sonicated to fragment the DNA to ∼200–1,000 bp fragments following a previously described protocol (Yamaguchi et al., 2014). Immunoprecipitation was carried out with HA antibody (Merck) to obtain the protein-DNA complex. DNA not incubated with the HA antibody was used as the input control. The DNA fragment was recovered with a DNA purification kit (Qiagen) following cross-linking and removal of proteins. The precipitated DNA was used as a template for qPCR amplification. The fold enrichment of the target gene fragment relative to the control promoter region in the immunoprecipitated DNA was calculated in proHDG2::HDG2-HA hdg2-3 lines and WT plants. The oligonucleotide sequences used for ChIP-qPCR analyses are listed in Supplemental Table S1.

Accession numbers

Sequence data from this article can be found in the Arabidopsis Genome Initiative under the following accession numbers: AT1G05230 (HDG2); AT5G09870 (CESA5); and AT1G79840 (GL2).

Supplemental data

Supplemental Figure S1. HDG2 is coexpressed with CESA5.

Supplemental Figure S2. HDG2 overexpression complements the mucilage defect of hdg2-3.

Supplemental Figure S3. Differentiation of seed epidermal cells in developing seeds of hdg2-3 and WT.

Supplemental Figure S4. Scanning electron microscopy of mature hdg2-3 and WT seeds.

Supplemental Figure S5. Monosaccharide compositions of the nonadherent and adherent mucilage of hdg2-3 and WT.

Supplemental Figure S6. Immunodetection of RG-I in hdg2-3 and WT mucilage by dot blotting.

Supplemental Figure S7. Expression analysis of other mucilage-related genes in hdg2-3.

Supplemental Figure S8. Phenotypic characterization of CESA5-complemented hdg2-3 lines.

Supplemental Figure S9. Cross talk between HDG2 and TFs associated with seed coat differentiation and mucilage production.

Supplemental Figure S10. HDG2 activates the expression of GL2.

Supplemental Table S1. List of primers used in this study.

 

R.H., Y.K., and G.Z. conceived the project; Y.K., S.P., Y.W., Y.X., and X.W. performed experiments; R.H., Y.K., and G.Z. analyzed data; R.H. wrote the article; and Y.K. and G.Z. revised the article.

The author responsible for distribution of materials integral to the findings presented in this article in accordance with the policy described in the Instructions for Authors (https://academic.oup.com/plphys) is Ruibo Hu (hurb@qibebt.ac.cn).