New steps in mucilage biosynthesis revealed by analysis of the transcriptome of the UDP-rhamnose/UDP-galactose transporter 2 mutant

Abstract

Upon imbibition, epidermal cells of Arabidopsis thaliana seeds release a mucilage formed mostly by pectic polysaccharides. The Arabidopsis mucilage is composed mainly of unbranched rhamnogalacturonan-I (RG-I), with low amounts of cellulose, homogalacturonan, and traces of xylan, xyloglucan, galactoglucomannan, and galactan. The pectin-rich composition of the mucilage and their simple extractability makes this structure a good candidate to study the biosynthesis of pectic polysaccharides and their modification. Here, we characterize the mucilage phenotype of a mutant in the UDP-rhamnose/galactose transporter 2 (URGT2), which exhibits a reduction in RG-I and also shows pleiotropic changes, suggesting the existence of compensation mechanisms triggered by the lack of URGT2. To gain an insight into the possible compensation mechanisms activated in the mutant, we performed a transcriptome analysis of developing seeds using RNA sequencing (RNA-seq). The results showed a significant misregulation of 3149 genes, 37 of them (out of the 75 genes described to date) encoding genes proposed to be involved in mucilage biosynthesis and/or its modification. The changes observed in urgt2 included the up-regulation of UAFT2, a UDP-arabinofuranose transporter, and UUAT3, a paralog of the UDP-uronic acid transporter UUAT1, suggesting that they play a role in mucilage biosynthesis. Mutants in both genes showed changes in mucilage composition and structure, confirming their participation in mucilage biosynthesis. Our results suggest that plants lacking a UDP-rhamnose/galactose transporter undergo important changes in gene expression, probably to compensate modifications in the plant cell wall due to the lack of a gene involved in its biosynthesis.

The transcriptome of urgt2 plants is used to identify compensation mechanisms in response to mutation in the UDP-rhamnose/galactose transporter to discover new genes involved in mucilage biosynthesis and/or modifications.

Introduction

The plant cell wall is a highly dynamic structure composed of hemicellulose, pectins, proteins, cellulose, and aromatic compounds which surround the cell membrane acting as a physical barrier, influencing plant development, cell morphology and also controlling external environmental cues (Verhertbruggen et al., 2013; Levesque-Tremblay et al., 2015; Rasool et al., 2017; Wolf et al., 2017). The tight synergistic regulation between the integrity of cell walls and physiological as well as environmental factors comprises the control of a large repertoire of genes involved in cell wall biosynthesis and metabolism (Höfte and Voxeur, 2017; Barnes and Anderson, 2018; Gigli-Bisceglia and Hamann, 2018). Hence, coordinated changes of the cell wall composition and structure allow plants to grow and react to external factors, indicating that cell walls can functionally adapt to respond to new environmental cues. There are other circumstances where plants harbouring mutations in genes involved in the metabolism of the plant cell wall exhibit changes in its structure and composition, triggering the expression of other genes to compensate the changes produced in the mutant in order to maintain cell wall functionality (Burton et al., 2000; Bischoff et al., 2009; Carpita and McCann, 2015; Guénin et al., 2017). In fact, the existence of complex transcriptional regulatory networks, which control the expression of a huge amount of genes related to the synthesis and/or modification of the cell wall, has been described (Yong et al., 2005; Carpita and McCann, 2015; Taylor-Teeples et al., 2015; Golz et al., 2018). Therefore, the study of single cell wall mutants sometimes makes the observation of obvious phenotypes difficult (Carpita and McCann, 2015). Evidence of compensatory mechanisms, leading to the maintenance of cell wall functionality, has been provided in different studies (Burton et al., 2000; Yong et al., 2005; Bischoff et al., 2009; Guénin et al., 2011; Kim et al., 2015; Saez-Aguayo et al., 2017; Xu et al., 2017). In this regard, a recent study by Guénin et al. (2017) comparing the transcriptome from the hypocotyl of etiolated plants from the wild-type and a mutant in pectin methyl esterase 3 (PME3), showed a down-regulation of genes coding for pectin-degrading enzymes in the mutant, supporting the hypothesis that connects pectin methylesterification and degradation; however, this mutant also showed an increase in the vacuolar CRUCIFERIN, the most abundant storage protein in Arabidopsis seeds (Guénin et al., 2017). How this result is linked to pectin metabolism remains unclear, but it certainly shows that a mutation in a gene involved in pectin metabolism leads to unexpected changes in gene expression.

Mucilage is an extracellular matrix that has been quite suitable to study pectin biosynthesis and its modification. Seed coat mucilage represents 3% of the seed weight and most of it corresponds to pectin, which is easily extruded from mature dry seed upon imbibition in water (Western et al., 2000; Young et al., 2008; Arsovski et al., 2010; Western, 2012; Francoz et al., 2015). Mutations affecting mucilage composition sometimes do not lead to other physiological effects on the plant, making it a good tool to identify proteins involved in the biosynthesis and modification of pectins (Western et al., 2000; Young et al., 2008; Dean et al., 2011; Western, 2012; Saez-Aguayo et al., 2013; Rautengarten et al., 2014; Francoz et al., 2015; Voiniciuc et al., 2015b, c; Ralet et al., 2016; Saez-Aguayo et al., 2017; Shi et al., 2018). Mucilage is constituted by soluble (SM) and adherent (AM) layers. Both of them are composed mainly (>90%) of unbranched rhamnogalacturonan 1 (RG-I), a polymer made of alternating repeats of (1,4)-α-d-GalA and (1,2)-α-l-Rha units (Ridley et al., 2001; Ralet et al., 2016). The SM layer can be removed easily upon water imbibition and, in addition to Rha and GalA, it contains traces of Ara, Xyl, Man, Gal, and Glc (Macquet et al., 2007; Voiniciuc et al., 2015b, c; Ralet et al., 2016). In contrast to the SM, the AM is attached to the seed surface and can be detached from the seed by digestion with rhamnogalacturonan hydrolase (Macquet et al., 2007), by high-speed mechanical agitation (Voiniciuc et al., 2015c), or by ultrasonic treatment (Zhao et al., 2017). In addition to RG-I, the AM contains minor amounts of cellulose, arabinan, galactan, galactoglucomannan, and homogalacturonan (HG; Macquet et al., 2007; Voiniciuc et al., 2015b; Ralet et al., 2016).

The synthesis of hemicellulose and pectins occurs in the Golgi apparatus, where a number of glycosyltransferases (GTs) transfer the sugar residue from an activated sugar nucleotide donor, UDP- or GDP-sugar, to a growing polysaccharide chain. Most GTs are type-II membrane-bound proteins with a catalytic domain facing the Golgi lumen (Wulff et al., 2000; Sterling et al., 2001; Scheible and Pauly, 2004; Liepman et al., 2010). However, most nucleotide sugars used by GTs are synthesized in the cytosol (Bonin et al., 1997; Seifert, 2004; Bar-Peled and O’Neill, 2011); therefore, the Golgi membrane is a physical barrier blocking the access of UDP/GDP-sugars to the active site of the GTs. To cope with this topological problem, nucleotide sugar transporters (NSTs) located in the Golgi membrane transport UDP/GDP-sugars from the cytosol to the Golgi lumen and supply to the GTs the substrates needed for polysaccharide biosynthesis (Reyes and Orellana, 2008; Orellana et al., 2016; Temple et al., 2016). In Arabidopsis thaliana, 44 genes encode putative NSTs and they are similar to those encoding plastidic triose phosphate translocators (TPTs); altogether, they form a gene family of 51 members (Knappe et al., 2003; Rautengarten et al., 2014). To date, a number of these NSTs have been functionally characterized, specifically transporters for GDP-Man, GDP-Fuc, UDP-Gal, UDP-Glc, UDP-Rha, UDP-Xyl, UDP-GalA, UDP-GlcA, and UDP-Araf (Baldwin et al., 2001; Norambuena et al., 2002, 2005; Handford et al., 2004, 2012; Bakker et al., 2005; Rollwitz et al., 2006; Rautengarten et al., 2014, 2016, 2017; Ebert et al., 2015; Saez-Aguayo et al., 2017). Furthermore, more recently URGT2 and UUAT1 were characterized as a UDP-rhamnose/UDP-galactose transporter gene and a UDP-uronic acid transporter gene, respectively, and mutations in both genes led to changes in mucilage composition. Both mutants exhibited less Rha and GalA in the SM layer, suggesting that levels of RG-I are affected in mutants in these transporters (Rautengarten et al., 2014; Saez-Aguayo et al., 2017). However, the urgt2 mutant exhibits a stronger mucilage reduction than the uuat1 mutant, suggesting the predominant participation of URGT2 in providing substrates to GTs during mucilage RG-I biosynthesis. Interestingly, higher levels of methylesterification of HG were also observed in mucilage from the uuat1 mutant in comparison with wild-type plants, suggesting the triggering of compensation mechanisms when genes involved in mucilage biosynthesis are mutated (Saez-Aguayo et al., 2017). Since some of the misregulated genes may be directly involved in cell wall biosynthesis, their identification can lead us to discover new players in this process. Therefore, we investigated further the mucilage phenotypes of urgt2-2 and compared the transcriptomes of both the wild-type (Col-0) and urgt2-2. The results revealed several pleiotropic changes affecting sugar composition and also the organization of seed coat mucilage. In order to identify misregulated genes triggered by the lack of URGT2 during mucilage production, we performed a comparative transcriptome analysis of developing seeds from wild-type and urgt2-2 at 8 days after pollination (8 DAP) using RNA sequencing (RNA-seq). The results revealed changes in the expression of 3149 genes during the stages when the mucilage secretory cells are actively engaged in the production of large amounts of pectin. The results also showed that the lack of URGT2 leads to an up-regulation of genes already described as involved in mucilage biosynthesis. Moreover, we observed changes in the expression of other cell wall-related genes, which can explain some of the phenotypes observed in the mutant. In particular, we found up-regulation of UAFT2 and UUAT3, two genes coding for NSTs not described previously to play a role in mucilage biosynthesis. To confirm their function, we selected mutants and performed biochemical analyses of the mucilage. The results showed that both uaft2 and uuat3 exhibited changes in mucilage composition, confirming their role in the biosynthesis of mucilage.

Materials and methods

Plant growth

Arabidopsis thaliana plants were germinated and grown in a growth chamber using a long-day period (16 h photoperiod); the light intensity was 120 mmol m⁻² s⁻¹ and the temperature varied between 19 °C and 28 °C. For collection of seeds and aerial tissue, plants were grown in soil (Top Crop) supplemented with fertilizer (Top Veg) at a relative humidity of 65%. The plants were germinated in Murashige and Skoog (MS) medium (Duchefa) (2.155 g l^–1), 1% sucrose, and 0.4% agar. T-DNA insertion lines for urgt2-1 (SALK_125196), urgt2-2 (SALK_071647), uaft2-1 (SALK_011583), uaft2-2 (SALK_018646), and uuat3-1 (GK-380D03) were obtained from the ABRC (http://abrc.osu.edu/) using the SIGnAL Salk collection (Alonso et al., 2003).

Tissue collection, RNA isolation, and RNA sequencing

Col-0 and urgt2-2 plants were grown simultaneously under the growth conditions detailed above. Flowers were tagged at the beginning of pollination, which was defined phenotypically as the time at which the flowers were just starting to open, as previously described by Western et al. (2000). For RNA-seq experiments, developing seeds were dissected from the siliques at 8 DAP and total RNA was isolated from the developing seeds of four siliques using the RNeasy Plant Micro kit (Qiagen) including a DNase I treatment (Invitrogen™) following the manufacturer’s instructions. Samples to make the libraries were obtained from three different plants in order to generate biological replicates. Each total RNA sample had a 260:280 nm ratio of at least 1.8 and an RNA quality number (RQN) value >8.0. From 1.0 μg of total RNA, cDNA libraries for each sample were constructed using the TruSeq^® Stranded mRNA kit (Illumina). Quality control and concentration were determined by capillary electrophoresis (Fragment Analyzer^®, AATI). Six cDNA libraries (2 genotypes×3 biological replicates) were sequenced on one lane of Illumina NexSeq500.

Cloning procedures

The URGT2 coding sequence (CDS), without the stop codon, was amplified from cDNA prepared from Arabidopsis leaf RNA, using the primers sense 5'-CACCATGGAGAAAGCAGAGAACGAGA-3' and antisense 5'-TGCTTTATTATTTCCAAGCTCCAT-3'. The resulting PCR products were introduced into the pENTR™/D TOPO^® vector according to standard protocols (Life Technologies) to generate the entry clone pENTR-URGT2. The URGT2 promoter (pURGT2) was amplified from Arabidopsis genomic DNA using the primers 5'-CACCTCATGTGTTGCGAATCTTATTC-3' and 5'-TTGGATTCAAATTAAAAAAATTCGAAATCTGAAATC-3'. The resultant PCR product was introduced into the pENTR 5-TOPO vector (Thermo Fisher Scientific) to generate the pENTR5-URGT2 entry clone. The C-terminal green fluorescent protein (GFP) fusion was obtained by recombining the entry clone with the destination vector R4pGWB504 (Nakagawa et al., 2007) using LR clonase (Thermo Fisher Scientific).

The entire CDS of the UUAT3 gene was cloned from cDNA prepared from Arabidopsis leaf RNA. The sequence without the stop codon was PCR-amplified using the following primers: sense 5'-CACCATGAAGATGGCGACGAATGGC-3' and antisense 5'-ATTTTTGTTTCGTTTCTTGGCTTC-3'. The resulting PCR products were introduced as explained above to generate the entry clone pENTR-UUAT3. The intergenic region (1900 bp) between At5g05820 and At5g05830 was defined as the UUAT3 promoter (pUUAT3) and was amplified from Arabidopsis genomic DNA using the following primers: sense 5'-CGCTTTTCTTCTTCCTAATCCTG-3' and antisense 5'- GCCACTGGGTTTTGGAGTTA-3'. The resultant PCR products were introduced into the pENTR™/5'-TOPO^® vector (Thermo Fisher Scientific) to generate the pENTR5-pUUAT3. The C-terminal GFP fusion was obtained by recombining the entry clones with the destination vector R4pGWB504 (Nakagawa et al., 2007) using LR clonase (Thermo Fisher Scientific).

Analysis of RNA-sequencing

Raw data reads (.fastq files) were trimmed using the Trim Galore Cutadapt (Martin et al., 2011) wrapper using the -paired and -q 25 option. Trimmed reads were aligned against the Tair10 reference genome (Arabidopsis Genome Initiative, 2000) using STAR (Dobin et al., 2013). Mapped read pair counts were calculated with HTseq-count (Anders et al., 2015) and counts were normalized into CPM (Glusman et al., 2013). Analysis of differentially expressed genes (DEGs) was performed using edgeR (Robinson et al., 2010) considering a false discovery rate (FDR) <0.05, and up-regulated and down-regulated genes were used to identify enriched Gene Ontology terms using AgriGO (Du et al., 2010).

Ruthenium red staining

Mucilage released from mature dry seeds was stained directly with 0.01% (w/v) Ruthenium red after imbibition in 0.5 M EDTA, pH 8.0, for 60 min. The coloration of AM seeds was observed with a stereoscopic microscope (LEICA EZ4 HD).

Seed immunolabeling

Immunolabeling was performed using four monoclonal antibodies, INRA RU1 (recognizes unbranched RG-I), JIM5 (recognizes poorly methylesterified HG), JIM7 (recognizes highly methylesterified HG), and LM25 (recognizes xyloglucan) (Ralet et al., 2010; Saez-Aguayo et al., 2013, 2017). Double labeling with an antibody plus calcofluor white (0.01%, w/v) or propidium iodide (20 mg ml⁻¹) was performed as indicated for each antibody to observe the seed surface and AM layer. Optical sections were obtained using a Leica TCS LSI spectral confocal laser scanning microscope. A 488 nm argon laser line was used to excite Alexa Fluor 488, a 405 nm diode laser line was used to excite calcofluor white, and a 543 nm neon laser line was used to excite propidium iodide. Fluorescence emission was detected between 504 nm and 579 nm for Alexa Fluor 488, between 412 nm and 490 nm for calcofluor white, and between 550 nm and 725 nm for propidium iodide. For comparisons of the signal intensity within one experiment, the laser gain values were fixed.

Extraction of mucilage layers

To determine the monosaccharide contents from SM, AM, and naked seeds, a sequential extraction procedure was carried out. Soluble mucilage was extracted by incubating 50 mg of seeds with 5 ml of water for 3 h. The adherent layer was detached from the seed by sonication employing an ultrasonic homogenizer sonic ruptor 250 (Omni International^©) for 2 min at 20% ultrasonic power followed by 1 min at 30% ultrasonic power. Upon SM and AM extraction, samples were stabilized by heating at 100 °C for 5 min and lyophilized.

AIR preparation and acid hydrolysis

Naked seeds were ground with liquid nitrogen and then washed three times with 80% ethanol (4 h, at room temperature), twice with methanol/chloroform 1/1 (v/v) (3 h, room temperature), and twice with acetone (1 h, room temperature) for removal of lipids. The final acid-insoluble residue (AIR) was dried out overnight at room temperature. Naked seed AIR and mucilage fractions were hydrolyzed for 30 min with 400 µl of 2 M trifluoroacetic acid (TFA) at 121 °C. TFA was evaporated at 65 °C with gaseous nitrogen and the samples were washed twice in 400 µl of 100% isopropanol and dried with gaseous nitrogen. Hydrolyzed products were suspended with water and clarified by passing through a syringe filter (pore size: 0.45 mm), transferred to a new tube, and used for HPAEC-PAD analysis as described below. Inositol and allose were used as internal controls for TFA hydrolysis.

HPAEC-PAD

A Dionex ICS3000 ion chromatography system, equipped with a pulsed amperometric detector, a CarboPac PA1 (4×250 mm) analytical column, and a CarboPac PA1 (4×50 mm) guard column, was used to quantify sugars. The separation of neutral sugars was performed at 26 °C with a flow rate of 1 ml min^–1 using an isocratic gradient of 20 mM NaOH for 32 min followed by a separation of acidic sugars using 75 mM NaOAc and 150 mM NaOH for 18 min at a flow rate of 1 ml min^–1 at 26 °C, followed by a wash with 200 mM NaOH for 5 min. After every run, the column was equilibrated in 20 mM NaOH for 10 min. Standard curves of neutral sugars (d-Fuc, l-Rha, l-Ara, d-Gal, d-Glc, d-Xyl, and d-Man) or acidic sugars (d-GalA and d-GlcA) were used for quantification.

Determination of PME activity

Total protein extracts were obtained by grinding 100 mg of dry seeds with extraction buffer (1 M NaCl, 12.5 mM citric acid, and 50 mM Na₂HPO₄, pH 6.5). The resulting homogenate was shaken for 1 h at 4 °C and then centrifuged at 14 000 rpm for 15 min, and the supernatant was retained. Protein concentrations were determined using the BCA kit (Pierce™ BCA Protein Assay Kit, Thermo Scientific), and equal quantities of protein (30 μg) in the same volume (20 μl) were loaded into 6 mm diameter wells in gels prepared with 0.1% (w/v) of 85% esterified citrus fruit pectin (Sigma-Aldrich), 1% (w/v) agarose, 12.5 mM citric acid, and 50 mM Na₂HPO₄ pH 6.5. After incubation overnight at 28 °C, plates were stained with 500 μg ml⁻¹ ruthenium red for 45 min and destained with water for >3 h. Measurements of stained areas were performed with ImageJ 1.34S software (Freeware, National Institute of Health).

Methylesterification analysis of mucilage layers and AIR samples

The degree of methylesterification of SM, AM, and naked seeds was measured from methanol released by alkaline de-esterification of extracts with 0.2 M NaOH (for mucilage layers) and with 1 M NaOH (for naked seeds) for 1 h at 4 °C. After the neutralization of extracts with 0.2 M or 1 M HCl, released methanol was measured as described by Saez-Aguayo et al. (2013). All experiments were done using five technical replicates and at least three biological replicates.

Expression analysis by quantitative real-time PCR (qRT-PCR)

Seeds were dissected from approximately four siliques at each DAP (6, 8, 10, and 12 DAP) for further RNA extraction. RNA extractions were performed as described above using an RNeasy Plus Micro Kit (Qiagen). A 1 μg aliquot of total RNA was used as a template for first-strand cDNA synthesis with an oligo(dT) primer and SuperScript II (Thermo Fisher Scientific), according to the manufacturer’s instructions. For reverse transcription–PCR (RT–PCR) analysis, primers described in Supplementary Table S6 at JXB online were used to amplify the entire CDS of URGT2 from single-stranded cDNA in the WT Col-0, urgt2-1, and urgt2-2. The primers used to amplify EF1αA4 were those described by North et al. (2007). qRT-PCR was performed using the Fast EvaGreen qPCR Master Mix kit (Mx3000P; Stratagene). Reactions contained 1 µl of 1:2 diluted cDNA in a total volume of 10 µl. Reactions were performed using primers that have been previously tested for their efficiency rates and sensitivity in a cDNA dilution series.

Relative expression

The quantification and normalization procedures were done using the following equation, as described by Stratagene:

Releative Expression

$${\displaystyle \begin{array}{ll}& \\ =\frac{{(1+\mathrm{E}\mathrm{t}\mathrm{a}\mathrm{r}\mathrm{g}\mathrm{e}\mathrm{t})}^{-\mathrm{\Delta }\mathrm{C}\mathrm{t}\mathrm{t}\mathrm{a}\mathrm{r}\mathrm{g}\mathrm{e}\mathrm{t}}}{\sqrt{\left[{(1+\mathrm{E}\mathrm{N}\mathrm{o}\mathrm{r}\mathrm{m}1)}^{-\mathrm{\Delta }\mathrm{C}\mathrm{t}\mathrm{N}\mathrm{o}\mathrm{r}\mathrm{m}1}\right]\times \left[{(1+\mathrm{E}\mathrm{N}\mathrm{o}\mathrm{r}\mathrm{m}2)}^{-\mathrm{\Delta }\mathrm{C}\mathrm{t}\mathrm{N}\mathrm{o}\mathrm{r}\mathrm{m}2}\right]}}\end{array}}$$

where E corresponds to the efficiency of amplification of target and reference genes, Ct is the threshold cycle, and Norm1 and Norm2 refer to the references or normalizer genes. EF1αA4, and the seed reference gene At4g12590 (Hong et al., 2010) were used as reference genes, and all primers used in this study are described in Supplementary Table S6.

Results

urgt2-2 is a knockout mutant affected in mucilage RG-I composition

The allelic mutants urgt2-1 and urgt2-2 mutated in URGT2 were described by Rautengarten et al. (2014), exhibiting a reduction of ~25% and 50% in Rha and GalA content in SM, suggesting that the content of RG-I was compromised (Rautengarten et al., 2014). RT–PCR and qRT-PCR analysis confirmed that both allelic mutants are knockout (Fig. 1B, C; Supplementary Fig. S3A, B), however urgt2-2 exhibited the greatest changes in mucilage composition (Rautengarten et al., 2014). The monosaccharide composition of both soluble and adherent layers showed that the mutation in URGT2 affects mucilage composition in both layers with a slight decrease in GalA content in the total sugar mucilage amount (Supplementary Fig. S2A). The molecular rescue of the urgt2-2 mutant was obtained by transformation with the proURGT2:URGT2-GFP construct. Several independent transformants were obtained, and GalA and Rha contents of the SM layer of three independent lines were determined. Contents similar to WT Col-0 plants were obtained for both sugars (Fig. 1D), indicating that proURGT2:URGT2-GFP had successfully rescued the mutant. All these results confirm that URGT2 plays a role in SM and AM RG-I biosynthesis.

Fig. 1.

urgt2-2 is a knockout line. (A) Schematic representation of URGT2 structure, as annotated in TAIR (http://www.arabidopsis.org/). The site and orientation of the T-DNA insertion urgt2-2 allelic line are indicated in the scheme. Numbers indicate the positions (in bp) of the start and stop codons and the T-DNA insertion site. Untranslated regions (light gray boxes), protein coding sequences (dark gray), black lines (introns), and LB (left border) are indicated.(B) Analysis of URGT2 expression in the urgt2-2 mutant line. RT–PCR analyses were performed on RNAs isolated from 8 DAP developing seeds from the WT Col-0 and urgt2-2 line using specific primers for the full-length coding sequence of URGT2. EF1αA4 expression was used as a reference gene. (C) qRT–PCR analysis of URGT2 transcript steady-state levels in developing seeds in 8 DAP developing seeds from the WT Col-0 and urgt2-2 mutant line. The expression was calculated relative to EF1αA4 and a seed-specific reference gene (At4g12590). Error bars represent SE values from three biological replicates (n=9). (D) Rescue of the urgt2-2 mutant phenotype using the proURGT2:URGT2-GFP construct. Soluble mucilage content of the WT Col-0, urgt2-2, and the transgenic lines proURGT2:URGT2 L1, L2, and L3. The Rha and GalA content was quantified using HPAEC-PAD. Error bars represent SE values from four technical repeats from three independent lines (n=12). Asterisks indicate significant statistical differences using student t-test (*P<0.01).

Transcriptome analyses in developing seeds from wild-type and urgt2-2 plants

Since the mutation in URGT2 can produce changes in the transcriptome from developing seeds as part of possible compensatory mechanisms, we compared the transcriptome of developing seeds from WT Col-0 and urgt2-2 plants by performing RNA-seq analyses. To perform the experiment, we looked for the time with the highest expression of URGT2; therefore, to determine the moment to collect the seeds for RNA extraction, we determined the URGT2 expression in planta by qRT-PCR (Supplementary Fig. S1). The results showed that URGT2 is highly expressed in developing seeds at 8 DAP, which corresponds to the stage when mucilage is highly produced and accumulated in the seed coat epidermal cells (Western et al., 2004; Macquet et al., 2007). The results also showed a high expression of URGT2 in rosettes; however, no phenotype was observed in this tissue. Based on these results, we isolated RNA from developing seeds at 8 DAP from WT Col-0 and urgt2-2 plants using three biological replicates, and transcriptomic analyses using RNA-seq were performed. We obtained a total of 174 908 519 reads with an average of 29 151 420 reads on each replicate and a minimum and maximum of 24 839 927 and 37 606 112, respectively (Supplementary Table S1). Reads were mapped against the reference genome obtaining 90.2% of mapped reads. Our transcriptomic analysis identified 3149 DEGs between the mutant and the WT Col-0 (FDR <0.01), where 656 were down-regulated and 2493 were up-regulated in urgt2-2. Among the 3149 misregulated genes in urgt2-2, 168 genes encoded proteins predicted to be involved in cell wall biosynthesis and/or modification (Supplementary Table S2).

To validate the expression pattern obtained by the transcriptomic analysis using RNA-seq, qRT-PCR was carried out for three genes showing higher abundance in the mutant (URGT4/At4g39390, UUAT1/At5g04160, and PME58/At5g40180) and three genes showing lower abundance in the mutant (EXPA1/At1G69530, XTH8/At1G11545, and PME53/At5g19730) in urgt2-2. The analysis of the expression of those genes in developing seed at 8 DAP revealed that both qRT-PCR and RNA-seq analysis had the same change of expression trend when urgt2-2 expression was compared with the WT Col-0. Therefore, these results validated the strength of the RNA-seq analysis performed on the urgt2-2 mutant (Fig. 2).

Fig. 2.

qRT-PCR validation of the RNA-sequencing (RNA-seq) experiment. Gene expression patterns from the RNA-seq analysis (in gray) were validated for six representative genes by qRT–PCR (in black). The confirmation includes three genes overexpressed (URGT4/At4g39390, UUAT1/At5g04160, and PME58/At5g40180) and three genes repressed (EXPA1/At1G69530, XTH8/At1G11545, and PME53/At5g19730) in the urgt2-2 mutants, in comparison with the WT Col-0. Error bars represent SE values from three biological replicates (n=9).

High global changes in specific mucilage genes

To demonstrate the occurrence of possible compensation mechanisms that take place in urgt2-2 during mucilage biosynthesis and/or modification, we analyzed the expression profile of genes with a known function in this process. To date, 75 genes acting on epidermal cell differentiation, mucilage biosynthesis, mucilage stabilization, hormone synthesis/perception, mucilage secretion, and mucilage modification have been proposed to play a role (Supplementary Table S7; Kong et al., 2013; Rautengarten et al., 2014; Francoz et al., 2015; Voiniciuc et al., 2015a, b, 2018; Ehlers et al., 2016; Ezquer et al., 2016; Hu et al., 2016; Ralet et al., 2016; Turbant et al., 2016; Griffiths et al., 2017; Saez-Aguayo et al., 2017; Salem et al., 2017; Tsai et al., 2017; Shi et al., 2018; Shimada et al., 2018; Takenaka et al., 2018). From these 75 genes, 37 were differentially expressed in the developing seeds from the WT Col-0 and urgt2-2 mutants. Interestingly, 35 genes were up-regulated in the mutant (Fig. 3). This differential expression affects genes classified as involved in ‘mucilage synthesis’ (i.e. RRT1, GATL5, GAUT11, MUCI70, MUM4/RHM2, MUCI10, IRX7/FRA8, MUCI21/MUM5, IRX14, UUAT1, and CSLA2 among others), ‘mucilage modification’ (i.e. PMEI6, PMEI14, ADK1, FLY1, BLX1, MUM2, and PME58), and ‘cell wall synthesis and modification’ (i.e. CESA10, PRX36/PER36, CESA3/IRX1, and SBT1.7/ARA12). We also observed changes in genes involved in transcriptional pathways (GL2, TTG2, SHP2, MYB5, and STK), mucilage stabilization (COBL2 and MUM3/CESA5), hormone perception (RAPTOR1B and GA1), and genes affecting mucilage secretion, release, or formation (TBA2, MOR1, PRX56, SKD1, and DRC). All these changes in the transcriptome suggest that in order to balance the lack of URGT2, a deep reorganization takes place, leading to changes in the structure and composition of polymers present in mucilage in the mutant.

Fig. 3.

Relative expression of specific mucilage genes in wild-type versus the urgt2-2 mutant line. The log₂FC from the genes described as acting in mucilage-specific processes is presented. From the 75 mucilage-specific genes, 70 genes were detected as being expressed in urgt2-2 by RNA-seq analysis. ROH.1, MYB23, DOF4.2, MYB75, and URGT2 were omitted from the analysis because of the low transcript accumulation. Statistical differences show misregulated genes with *FDR <0.05, **<0.01, ***<0.01.

Mucilage from urgt2-2 is not only altered in RG-I monosaccharide composition

To determine whether the changes observed in the transcriptome led to evident changes in mucilage composition, we carried out biochemical and immunological analyses of both SM and AM fractions. First, we analyzed the monosaccharide composition of mucilage (Table 1) and, as expected, we found that Rha decreased by 25% and 13% in the SM and AM, respectively. No changes were observed in seeds where all mucilage was extracted (naked seeds), indicating that changes in urgt2-2 were specific for mucilage. The decrease in Rha was accompanied by a reduction in the GalA content of 15% in the SM and 10% in the AM (Table 1; Supplementary Fig. S2A). Interestingly, we observed changes in other sugars, such as Xyl, which increased 23% in SM and 20% in AM. In addition, Ara increased by 35% in AM but decreased by nearly 60% in SM. Moreover, a small but significant decrease in Fuc was also observed in SM. The decrease in Rha and GalA suggested a decrease in RG-I; thus, we used the antibody INRA-RU1 to assess possible changes in this polymer. Reduced labeling was observed in the AM of urgt2-2 seeds in comparison with WT Col-0 seeds (Supplementary Fig. S2B), thus providing additional evidence of a reduction of RG-I in the mutant. Since we observed an increase in xylose, we also tested antibodies against xylan and xyloglucan on the imbibed seeds. No signal was detected using LM10 and LM11 (xylan); however, using LM25 (xyloglucan), we could detect a thread pattern that was more evident in the urgt2-2 mutant compared with the WT Col-0 (Fig. 4). In contrast, the radial wall xyloglucan labeling observed on the epidermal cells in the WT Col-0 was decreased in the mutant. However, the Glc content in the AM of urgt2-2 did not change, suggesting qualitative but not quantitative changes in xyloglucan in the mutant. Finally, the antibodies against arabinan gave no signal.

Table 1.

Monosaccharide composition of adherent and soluble mucilage layers extracted from wild-type and urgt2-2 dry seeds

Sugars	WT Col-0	urgt2-2
Soluble mucilage (mg g–1 of dry seeds)
Gal-A	11.87 (0.43)	10.13 (045)**
Rha	12.70 (0.46)	9.74 (0.45)**
Fuc	1.27 (0.32)	0.51 (0.15)*
Ara	0.25 (0.03)	0.15 (0.01)*
Xyl	0.98 (0.03)	1.21 (0.05)*
Man	0.18 (0.01)	0.16 (0.02)
Gal	0.40 (0.02)	0.38 (0.01)
Glc	0.25 (0.09)	0.20 (0.06)
Glc-A	0.40 (0.06)	0.27 (0.03)
Total sugars (SM)	28.30 (1.52)	22.74 (1.07)**
Adherent mucilage (mg g–1 of dry seeds)
Gal-A	5.36 (0.25)	4.84 (0.21)**
Rha	6.06 (0.27)	5.36 (0.28)**
Fuc	0.15 (0.01)	0.15 (0.01)
Ara	0.14 (0.01)	0.19 (0.01)**
Xyl	0.62 (0.03)	0.75 (0.03)***
Man	0.23 (0.01)	0.23 (0.01)
Gal	1.00 (0.04)	1.02 (0.05)
Glc	3.06 (0.38)	2.93 (0.55)
Glc-A	0.14 (0.002)	0.14 (0.004)
Total sugars (AM)	16.76 (0.46)	15.61 (0.62)*
Total sugars (AM+SM)	45.48 (1.14)	38.48 (1.24)**
Total sugars (AM+SM) (%)	100 (2.51)	84.60 (2.72)
Naked seeds (mg g–1 of AIR)
Gal-A	17.23 (1.72)	16.09 (1.86)
Rha	14.56 (0.99)	14.59 (1.18)
Fuc	2.41 (0.17)	2.23 (0.18)
Ara	25.38 (1.13)	24.05 (1.31)
Xyl	15.60 (1.05)	14.99 (1.21)
Man	4.08 (1.05)	3.99 (0.29)
Gal	24.21 (1.53)	23.73 (1.81)
Glc-A	2.04 (0.13)	2.03 (0.18)
Total sugars	105.51 (6.73)	101.59 (7.84)

To analyze monosaccharide composition, water and sonication extractions were used to obtain the soluble (SM) and adherent (AM) mucilage, respectively. Sugar content was quantified using HPAEC-PAD from SM, AM, and naked seeds. Means were calculated with data from four biological replicates. SEs are shown in parentheses for two technical replicates each. Asterisks indicate significant statistical differences using Wilcoxon test (* P<0.05; ** P<0.01, *** P<0.001).

Fig. 4.

The urgt2-2 mutants exhibit a higher labeling of xyloglucan domains in the adherent mucilage (AM) layer. Xyloglucan labeling in AM from imbibed seeds of the WT Col-0 and the urgt2-2 mutant lines. Confocal microscopy optical section reconstruction of AM released from imbibed seeds. Asterisks represent differences in labeling. LM25 antibody (cyan) was used to label xyloglucan and propidium iodide (PI) was used to detect the seed surface (purple). dw, distal wall; rw; radial walll and c, columella. Scale bar=50 µm.

The mutation in URGT2 affects homogalacturonan methylesterification in mucilage

The transcriptome analysis showed a number of changes in the abundance of transcripts encoding genes involved in modulating pectin methylesterification, such as PMEI6, PMEI14, PME58, FLY, and SBT1.7 (Fig. 3), suggesting that methylesterification of polysaccharides may be different in the mutant; thus, we assessed the global PME activity of dry seed, measured the methanol release in both mucilage layers, and investigated HG methylation in the AM layer by immunolabeling. The results showed that urgt2-2 had a decrease by 14% in global PME activity in comparison with the WT Col-0 (Fig. 5A) Consistent with the higher PME activity, an increase in methanol released from both layers of mucilage was observed in the urgt2-2 mutant (Fig. 5B). Immunolabeling of the AM layer using JIM7 and JIM5, which recognize highly and partially methylesterified HG, respectively (Fig. 5C), showed an increase of the JIM7 labeling localized in the external AM (yellow) and a decrease in JIM5 labeling (green). Altogether, these results indicate that mucilage from urgt2-2 exhibits higher levels of HG methylesterification.

Fig. 5.

The urgt2-2 mutant is reduced in pectin methylesterase (PME) activity and has increased mucilage methylesterification. (A) Global seed PME activity. Total protein extracts from mature dry seeds of the WT Col-0 and urgt2-2 were used to measure global PME activity. The PME activity was normalized to the average WT Col-0 activity (100%). Error bars represent the SE (n=20 from four biological replicates). Asterisks indicate significant statistical differences using t-test (* P<0.01). (B) Methanol content in the WT Col-0 and urgt2-2 in adherent and soluble mucilage fractions. Error bars represent the SE (n=20, from four biological replicates). Asterisks indicate significant statistical differences using t-test (*P<0.05, **P<0.01). (C) Homogalacturonan labeling in adherent mucilage from dry seeds of the WT Col-0 and the urgt2-2 mutant line. Confocal microscopy optical section reconstruction of AM released from imbibed seeds.#8232;JIM7 (yellow) and JIM5 (green) antibodies were used to label highly methylesterified and poorly methylesterified HG domains, respectively. Propidium iodide was used to detect the seed surface (purple). dw, distal cell wall; rw, radial cell wall; and c, columella: scale bar=50 µm.

Most NSTs are expressed during mucilage formation

Recently evidence was provided that the absence of the UUAT1 transporter produces pleiotropic changes in mucilage (Saez-Aguayo et al., 2017). Due to the changes observed in sugar content in urgt2-2, we decided to investigate whether other NST members could compensate for the lack of URGT2. The transcriptomic analysis revealed that 41 out of the 44 putative NSTs described to date are expressed in developing seed at 8 DAP of urgt2-2, and 17 were significantly up-regulated in the mutant; whereas one was down-regulated (Fig. 6). The increase in NST expression suggested that NSTs play a role in mucilage biosynthesis to compensate for the lack of URGT2 (Fig. 6). Interestingly, among the six URGTs, only the transcript accumulation of URGT4 was more abundant in the mutant than in the WT Col-0. Among the up-regulated NSTs, we found four out of the five members of the uronic acid transporters (UUAT1, UUAT3, UUAT4, and UUAT5, Rautengarten et al., 2014), three out of the four UDP-Araf transporters (UAFT2, UAFT3, and UAFT4,Rautengarten et al., 2017), and also the UDP-Xyl transporter UXT3 (Ebert et al., 2015). The analysis of expression of these NST genes on the seed coat, using the data available on the the Arabidopsis eFP Browser platform (Supplementary Fig. S4), revealed that UUAT1, UUAT3, and UAFT2 were the three genes with the highest expression in the linear cotyledon stage, which has been described as the stage where mucilage is actively synthesized (Western et al., 2000). Therefore, these results suggest that these NST genes may play a role during mucilage biosynthesis.

Fig. 6.

Nucleotide sugar transporter (NST) expression in the urgt2-2 mutant line in comparison with the wild-type. The log₂FC from the NST genes is shown. The NST clades are indicated as described by Rautengarten et al. (2014). From the 44 NSTs genes, 41 genes were detected as being expressed in the urgt2-2 RNA-seq analysis. URGT2, GONST5, and At1g53660 were omitted from the analysis because of the low transcript accumulation. The triose phosphate clade is not shown. Statistical differences show misregulated genes with *FDR <0.05, **<0.01, ***<0.01.

Assessing the role of NSTs differentially expressed in urgt2-2 in the biosynthesis of mucilage

The transcriptome analysis showed that UUAT1, UUAT3, and UAFT2 are highly expressed in the mutant. In order to confirm that this up-regulation occurs on both allelic lines, we assessed the transcript levels of both UUAT3 and UAFT2 in urgt2-1 and urgt2-2, and they were higher than in the WT Col-0 with the exception of UAFT2 in urgt2-2 where we could not see statistical differences from the WT Col-0 (Supplementary Fig. S5A, B). We already know that mutants in UUAT1 have a mucilage phenotype (Saez-Aguayo et al., 2017); therefore, to determine if UUAT3 and UAFT2 are also involved in mucilage formation, we identified two mutant alleles for UAFT2 (uaft2-1 and uaft2-2) and one mutant allele for UUAT3 (uuat3-1). Since we could not identify more knockout mutants for UUAT3, a molecular rescue of uuat3-1 by transforming this mutant with the proUUAT3:UUAT3-GFP construct was performed. The histochemical analysis of uaft2 mutants revealed less ruthenium red staining and RG-I labeling in the AM (Fig. 7). These results correlated with the monosaccharide composition measured on SM and AM by HPAEC-PAD in uaft2 mutants (Fig. 9; Supplementary Table S4). The two uaft2 mutant lines showed a reduction of almost all monosaccharides in the AM layer, which led to a reduction of 17.9% and 20% of total monosaccharide content for uaft2-1 and uaft2-1, respectively. The content of arabinose was significantly reduced in AM (35.3% and 55.3% in uaft2-1 and uaft2-2, respectively) (Fig. 9). Interestingly, no change in Ara was observed in SM; however, a slight increase in Rha and GalA was observed in SM. Furthermore, an increase of 15.7% and 18.9% in the total monosaccharide content of the SM fraction in uaft2-1 and uaft2-2, respectively, has also been observed.

Fig. 7.

uaft2 mutants exhibit an altered adherent mucilage (AM) phenotype. (A) Ruthenium red staining of the AM phenotype in seeds of uaft2-1 and uaft2-2 in comparison with the WT Col-0. Seeds were imbibed for 1 h in 0.5 M EDTA and stained with ruthenium red (0.01%). Scale bar=100 μm. (B) Rhamnogalacturonan-I (RG-I) labeling in adherent mucilage from WT Col-0 seeds and the uaft2 mutant lines. Confocal microscopy optical section reconstruction of AM released from imbibed seeds. INRA RU1 antibody (green) was used to label RG-I epitopes, and propidium iodide was used to stain the seed surface (purple). Scale bar=100 μm.

Fig. 9.

Mucilage phenotype of urgt3 and uaft2 mutant lines. GalA, Rha, and Ara content in soluble (left) and adherent (right) mucilage of uuat3-1, uaft2-1, and uaft2-2 T-DNA mutant lines in comparison with their WT Col-0. Sugar content was obtained using HPAEC-PAD from soluble and adherent mucilage. Error bars represent the SE (n=8–12) from 2–3 biological replicates. Asterisks indicate significant statistical differences by Wilcoxon test (*P<0.05; **P<0.01).

In the case of uuat3-1, a higher intensity of ruthenium red staining and stronger RG-I labeling on the AM layer were observed (Fig. 8B, C). Three rescued lines showed ruthenium red staining and RG-I labeling similar to the WT Col-0 confirming that the phenotype was due to the lack of UUAT3. Further biochemical analysis of uuat3-1 revealed that the mutant had a reduction of 20.1% of the total monosaccharide content in the SM layer. This reduction was mainly due to a reduction in Rha, GalA, and Xyl (Fig. 9; Supplementary Table S5). However, no significant changes of total monosaccharide content of the AM layer were observed in the uuat3 mutant, although there was a slight increase in the content of GalA, Xyl, and Gal.

Fig. 8.

Molecular rescue of the uuat3-1 mutant mucilage phenotype using the proUUAT3:UUAT3-GFP construct. (A) Expression of UUAT3 in the uuat3-1 mutant and the molecular rescue lines. Determination of UUAT3 mRNA steady state in three uuat3-1 rescue lines (proUUAT3:UUAT3-GFP-L6, -L18, and -L23). The expression was normalized in relation to EF1αA4 expression. Error bars represent SE values from three biological replicates (n=9). (B) The Ruthenium red adherent mucilage (AM) phenotype is restored in T₃ seeds of three independent uuat3-1 transformant lines expressing the UUAT3 CDS fused to GFP under the control of the UUAT3 promoter (proUUAT3:UUAT3-GFP). Seeds were imbibed for 1 h in 0.5 M EDTA and stained with Ruthenium red (0.01%). proUUAT3:UUAT3-L6, L-18, and -L23 lines recovered the wild-type AM staining when compared with the uuat3-1 mutant. Scale bar=100 μm. (C) Rhamnogalacturonan-I (RG-I) labeling in AM from WT Col-0 seeds and the uuat3-1 mutant line. Confocal microscopy optical section reconstruction of AM released from imbibed seeds. INRA RU1 antibody (green) was used to label RG-I epitopes, and propidium iodide was used to stain the seed surface (purple). proUUAT3:UUAT3-L6, L-18, and -L23 lines restored the uuat3-1 labeling changes. Scale bar=100 μm.

Discussion

URGT2 is a UDP-rhamnose/UDP-galactose transporter gene involved in the biosynthesis of mucilage RG-I (Rautengarten et al., 2014). A deficiency in this gene produces changes in the content of monosaccharides in mucilage, suggesting that adaptation mechanisms are triggered in the mutant. In order to analyze the changes that take place in the mutant, we performed a transcriptomic analysis of the urgt2-2 mutant in developing seeds using RNA-seq. We are aware that changes in transcript abundance do not necessarily represent changes in the protein content; however, certainly it is an excellent approach to obtain an overview of the changes triggered in a mutant. Transcriptomic data of urgt2-2 seeds revealed a misregulation of 3149 genes, a rather large number of genes in comparison with the numbers obtained in recent studies addressing changes in the transcriptome of mutants in genes involved in cell wall metabolism (Guénin et al., 2017; Faria-Blanc et al., 2018). The analysis of mutants in genes involved in secondary cell wall metabolism found few changes in the transcriptome probably due to the fact that cells producing thickening of the secondary cell entered into a cell death program (Faria-Blanc et al., 2018). In contrast, the tissue used in this study for the transcriptomic analysis corresponds to seeds from 8 DAP, which are actively developing and engaged in producing large amounts of mucilage. A different study, comparing the transcriptome of a mutant on AtPME3 and the wild-type showed misregulation in 464 genes (Guénin et al., 2017). The transcriptome analysis was performed in etiolated hypocotyls, a tissue that also might be less complex in terms of gene expression than 8 DAP developing seeds. On the other hand, there are technical aspects that may also contribute to explain the differences between our study and those looking for changes in the transcriptome of mutants in other cell wall genes. This is the approach utilized to collect the transcriptomic data, since we used RNA-seq, whereas the other studies used microarray approaches. In fact, this was also the case for a recent study, where the transcriptome analysis of the mur3 mutant performed by digital gene expression analysis (DGE-seq) revealed the misregulation of 1423 genes (Xu et al., 2017). Finally, it is important to highlight that knocking out a single gene leads to an unexpected change in the transcriptome. Whether this is because a change in RG-I or other rhamnosylated molecule needs to be well compensated, or the imbalance in the cellular pools of UDP-Rha triggers this response, remains to be answered.

Since we were interested in the identification of networks that may be related to the function of URGT2, which provides one of the substrates for the synthesis of RG-I, the most abundant polysaccharide in mucilage, we focused our attention on misregulated genes that are related to cell wall metabolism. This analysis was combined with the biochemical and morphological characterization of mucilage.

The mucilage reduction of urgt2-2 mutants was mainly due to a decrease in the content of Rha and GalA in both the AM and SM layers. Since URGT2 does not transport UDP-GalA, it is likely that the decrease in GalA could be due to an impairment in the synthesis of the RG-I backbone, which is composed of repeating disaccharide units (GalA–Rha)_n. A similar phenomenon is observed in a mutant in the UDP-uronic acid transporter 1 (UUAT1) described in Saez-Aguayo et al. (2017), where a decrease in Rha is associated with a concomitant reduction in GalA. These results indicate that both UUAT1 and URGT2 are involved in the biosynthesis of RG-I, and the lack of either of these transporters leads to a decrease in both sugars in RG-I. Moreover, there have been three genes described which are responsible for the formation of mucilage RG-I, rhamnosyltranferase 1 (RRT1) and two putative galacturonosyltransferases, GATL5 and MUCI70 (Kong et al., 2013; Takenaka et al., 2018; Voiniciuc et al., 2018). These three genes were up-regulated in the urgt2-2 seed transcriptome (Fig. 3); however, while qRT-PCR analysis for GATL5 confirmed this finding, the result for RRT1 was not conclusive in support of a higher content of RRT1 in urgt2-2 (Supplementary Fig. S5C, D). Despite this finding, the overall results tend to support the idea that the RG-I biosynthetic machinery is up-regulated in urgt2-2, suggesting that the decrease of UDP-Rha availability could be a limiting factor in the synthesis of this polymer.

Other changes were observed in mucilage from urgt2-2 plants, which are difficult to reconcile with an impairment in the transport of UDP-Rha. For instance, seeds from the mutant accumulated more Ara and Xyl in the AM layer, while the SM layer showed a strong increase in the content of Xyl, but also a decrease in the content of Fuc and Ara. It has been described that mucilage is composed of low amounts of heteroxylans and xyloglucan (Voiniciuc et al., 2015b, c). However, the increase in Xyl content on both mucilage layers of urgt2-2 was not accompanied by an increase of the Glc content, suggesting that urgt2-2 mucilage did not accumulate xyloglucan and instead this change could be explained by an increase in xylan. Surprisingly, seeds from urgt2-2 showed higher labeling of xyloglucan epitopes in the adherent mucilage, which could be explained by more accessible xyloglucan epitopes on the urgt2-2 mucilage due to a reduction in the monosaccharide content in this layer. The accumulation of xylose in the AM layer correlated with the up-regulation in the mutant seeds of UXS3 (Kuang et al. 2016) and UXT3 (Ebert et al., 2015), genes involved in the biosynthesis and transport of UDP-Xyl into the Golgi. Moreover, the increase in urgt2-2 of transcripts from IRX14 and MUCI21, two putative xylosyltransferase genes that are essential for xylan elongation and substitution in the seed coat mucilage (Voiniciuc et al., 2015c), support the idea that the increase in Xyl is due to an accumulation of mucilage xylans. In fact, xylan domains have been proposed as mucilage stabilizers due to their high affinity for cellulose and are involved in RG-I attachments to the mucilage cellulose fibers (Voiniciuc et al., 2015c; Ralet et al., 2016); thus, the changes observed in urgt2-2 may seek a higher stabilization of mucilage in the mutant.

The urgt2-2 mutant also exhibited changes in the content of Ara, which was slightly higher in the AM and lower in the SM layer in comparison with the wild-type. However, the total content of Ara in mucilage was not altered, with a mean of 0.37 (SD±0.03) mg g^–1 of seeds for WT Col-0 versus 0.38 (SD±0.03) for mg g^–1 of seeds for urgt2-2, suggesting changes in the partitioning of arabinans between mucilage layers. The increase in arabinans in the AM could be explained by the up-regulation of the UDP-Ara transporter group (UAFT2, UAFT3, and UAFT4; Rautengarten et al., 2017) and higher levels of the BXL1 transcript, which codes for a bifunctional β-d-xylosidase/α-l-arabinofuranosidase, an enzyme necessary to modify the arabinan present in seed mucilage (Arsovski et al., 2009), and also the putative arabinosyltransferase RRA1 (Supplementary Table S4), which has been associated with the synthesis of an Ara-containing polymers such as arabinogalactan proteins (AGPs), which may be in close association with cellulose fibers present in the AM (Egelund et al., 2008). Indeed, it could not be discarded that AGPs could have a role in maintaining the adherence of seed mucilage. In the past few years, it has been described that the AGP protein SOS5/FLA4 played a role in mucilage adherence of Arabidopsis seeds (Harpaz-Saad et al., 2011; Griffiths et al., 2014, 2017). However, SOS5/FLA4 was not significantly misregulated in our RNA-seq study, but, as the authors suggested, other AGPs could also play a role in seed mucilage adherence (Griffiths et al., 2016). Surprisingly, the analysis of the relative expression of AGPs in urgt2-2 revealed the misregulation of 21 AGP genes from the 151 already described (Ma et al., 2017), 16 of which are up-regulated (Supplementary Table S3). Interestingly, among the urgt2-2 up-regulated AGPs, ARABINOXYLAN PECTIN ARABINOGALACTAN 1 (APAP1) has been proposed as a candidate to participate in the control of seed mucilage adherence (Höfte, 2015).

The transcriptomic analysis of urgt2-2 revealed the overexpression of 17 NSTs. Among them, UUAT3 and UAFT2 were NST genes that are the most highly expressed in seed coat cells. Given this feature and to investigate further the possible role of these NSTs in mucilage formation, we decided to study the mucilage obtained from mutant seeds. The study on uuat3 revealed a decrease in the content of Rha and GalA in SM, similar to what was already observed in uuat1 (Saez-Aguayo et al., 2017), indicating that UUAT3 plays a role in mucilage biosynthesis. In contrast, the content of Rha did not change significantly, whereas GalA had a minor increase in the AM in uuat3, suggesting no changes in the content of RG-I in this portion of the mucilage and a slight increase in HG. Interestingly, greater staining with ruthenium red was observed in the AM in uuat3, as well as higher immunodetection with the anti-RG-I antibody. The most likely explanation for these results is a higher exposure to carboxyl groups, thus leading to an increased ruthenium red staining. This change could also be responsible for an increase in the immunoreactivity of the anti-RG-I antibody, producing greater labeling in the AM. The activity of UUAT3 has not been assessed yet, but, given the close sequence similarity to UUAT1 and the decrease in GalA, it is likely that it may also transport UDP-uronic acids. In the case of uaft2 mutants, both allelic lines exhibited a reduction in the content of Ara in the AM. In addition, we observed changes in the distribution of RG-I between the mucilage layers, being less abundant in AM and more abundant in SM. Furthermore, it seemed that the reduction of Ara in the AM of uaft2 mutants led to the accumulation of a more soluble RG-I in seed coat mucilage, reinforcing the hypothesis that Ara-containing polymers, probably AGPs, could play a role in mucilage adherence (Harpaz-Saad et al., 2011; Griffiths et al., 2014, 2017).

Another important change observed in the urgt2-2 mutant mucilage was the methylation of HG. Similar to what was described previously in the study on UUAT1 (Saez-Aguayo et al., 2017), urgt2-2 dry seed presented a decrease of global PME activity accompanied by an increase in methylesterification. These observations suggest that the reduction of RG-I observed in both UUAT1 and URGT2 mutants triggers a compensation mechanism that led to pleiotropic changes in HG methylation (Saez-Aguayo et al., 2017). The phenotype related to changes in HG methylation correlated with the misregulation of genes that have been proposed to be directly involved in the control of HG methylesterification: PMEI6, PMEI14, PME52, FLY1, and SBT1.7 (Rautengarten et al., 2008; Saez-Aguayo et al., 2013; Voiniciuc et al., 2013; Turbant et al., 2016; Shi et al., 2018). Recently, it has been proposed that the transcription factor gene MYB52 activates the expression of PMEI6, PMEI14, and SBT1.7 (Shi et al., 2018). Surprisingly, the expression of MYB52 was not significantly changed in urgt2-2 in comparison with the wild-type. However, PMEI6, PMEI14, and SBT1.7 were up-regulated in urgt2-2. These results indicated that the mucilage modification and stabilization seemed to be controlled by a complex network regulated by transcription factors (Bui et al., 2011; Huang et al., 2011; Saez-Aguayo et al., 2013; Francoz et al., 2015; Ezquer et al., 2016; Golz et al., 2018; Shi et al., 2018). This hypothesis was further strengthened by the finding in the urgt2-2 mutant line of the up-regulation of the transcription factor genes, which have been previously related to the control of the seed coat formation, TTG2, GL2, SHP2, NARS/NAC2, MYB5, and STK (Western et al., 2000; Windsor et al., 2000; Johnson et al., 2002; Li et al., 2009; Francoz et al., 2015; Ehlers et al., 2016; Ezquer et al., 2016; Tsai et al., 2017). Indeed, it has been reported that the STK transcription factor gene positively regulates the expression of PMEI6 and represses the expression of SBT1.7 (Ezquer et al., 2016), suggesting the participation of STK in the control of HG methylesterification Moreover, the up-regulation of PMEI6 and PME14, both genes coding for PME inhibitors involved in the regulation of mucilage PME activity (Saez-Aguayo et al., 2013; Shi et al., 2018), correlated with the reduction in PME activity and the concomitant reduction in HG methylesterification in urgt2-2 (Fig. 5). In addition, six putative PMEs were found to be down-regulated in urgt2-2, suggesting that they are good candidates to control the HG methylesterification in the seed mucilage of Arabidopsis (Supplementary Table S2). Finally, urgt2-2 seeds exhibited an up-regulation of PME58, a PME gene which has been related to the control of the distribution of HG in the adherent mucilage (Turbant et al., 2016). The overexpression of PME58 could not correlate with the global PME activity in the urgt2-2 mutant, but it had been reported that PME58 could be the target for PMEI6, SBT1.7, and FLY1 (Turbant et al., 2016), all of them overexpressed in urgt2-2; thus, this finding suggested that the inhibition of PME58 may result in an up-regulation of its expression.

The observation of pleiotropic changes in HG methylesterification in urgt2-2 is similar to that in mucilage from uuat1 seeds (Saez-Aguayo et al., 2017), where a mucilage phenotype is also observed; thus, these results suggest the triggering of compensatory mechanisms aiming to maintain the functional integrity of the cell wall (Hamann, 2015a, 2015b; Höfte, 2015; Voxeur and Höfte, 2016; Chebli and Geitmann, 2017; Bacete et al., 2018). Detection of an altered cell wall composition or sensing altered turgor pressure in epidermal cells due to differences in the composition of the mucilage pocket could be recognized by different plasma membrane receptors which can lead to a compensation response (Hamann, 2015b; Voxeur and Höfte, 2016; Wolf, 2017). In Arabidopsis, a family of receptors, RLK1-Like kinases, have been described to be related to the maintenance of the cell wall homeostasis, perception of mechanical changes, cell wall integrity, and the control of cellular growth (Voxeur and Höfte, 2016; Wolf, 2017). Among members of this family are the receptors THESEUS1 (THE1) and FERONIA (FER), which are found as part of protein complexes with the co-receptors LORELEI and LORELEI-LIKE GPI-ANCHORED proteins (Wolf, 2017). Interestingly THE1, FER, and LORELEI-LIKE GPI-ANCHORED PROTEIN 2 were overexpressed in the urgt2-2 mutant, suggesting that they may play a role in the reorganization set up in the mutant. Nevertheless, LORELEI-LIKE GPI-ANCHORED PROTEIN 1 was not significantly altered and LORELEI was not detected (Supplementary Table S8). FER is an important element of the cell wall integrity (CWI) signaling pathway as this receptor activates a rapid alkalinization of the apoplast, which could be a key process in cell wall remodeling, as some enzymes are very sensitive to pH changes, such as the PME activity which has a high pH optimum (Boyer, 2009). Two mechanisms for how FER receptor activated rapid alkalinization of the apoplast have been described. The first one is to activate the rapid alkalinization factor 1 (RALF1) which is itself a receptor for other RALF and RALF-like proteins (Haruta et al., 2014). The second one involves the association of FER with a plasma membrane H⁺-ATPase (AHA2), a protein that is believed to be responsible of the alkalinization of the apoplast. In the urgt2-2 mutant lines, AHA2 was overexpressed but RALF1 was no significantly overexpressed. Nevertheless, we found another two members of the RALF protein family (At4g14020 and RALF23) overexpressed in urgt2-2.

Another important sensing module controlling cell wall homeostasis is the brassinosteroid (BR) signaling pathway (Höfte, 2015; Voxeur and Höfte, 2016; Wolf, 2017). The BR receptor complex is formed by the leucine-rich repeat RLKs (LRR-RLKs), BRASSINOSTEROID INSENSITIVE 1 (BRI1) and the co-receptor BRI1-associated kinase 1 (BAK1) (Wolf et al., 2012). The BAK1 protein bound to receptor-like protein 44 (RLP44), which plays a role in the integration of the mechanosensing of cell wall changes in a way independent of the BR sensing module (Wolf, 2014). In fact, it has been proposed that RLP44 senses mechanical cell wall changes, triggering the phosphorylation/dephosphorylation cascade of several members of the BR signaling pathway, leading to the dephosphorylation of the BR-responsive transcription factors BRASSINAZOLE-RESISTANT 1 (BZR1) and BRI1-EMS suppressor 1 (BES1/BZR1) (Sun et al., 2010; Yu et al., 2011). Those transcription factors regulate the expression of several BR-responsive genes, including cell wall biosynthetic genes, such as PMEs and expansins (Sun et al., 2010). Our results showed an up-regulation of BAK1, RLP44, BZR1, and BRI1 like 3 in urgt2-2 in comparison with the wild-type. As there was no prior evidence of the existence of a CWI signaling network that senses the changes in the mucilage structure and composition, the changes in CWI signaling genes suggest the possible existence of a sensing and monitoring mechanism of the CWI during seed development in order to promote correct epidermal cell elongation and mucilage deposition. Therefore, it seems that small changes in the mucilage composition, due to the lack of URGT2, trigger a range of changes in the expression of cell wall/mucilage polymer synthesis and/or modification genes; which, in turn, were highly regulated in a transcriptional way by the CWI signaling cascade. These results show the plasticity of the cell wall which is crucial to maintain the cell wall functionality.

Author contributions

SS-A, AL-G, and AO designed the research; JPP-R, SS-A, JC, AL-G, DS, and HT performed the experiments; TC and CM performed informatics analysis; SS-A, JPP-R, AL-G, and AO analyzed the data; and SS-A, AL-G, FCR, and AO wrote the article.