Two tomato GDP-D-mannose epimerase isoforms involved in ascorbate biosynthesis play specific roles in cell wall biosynthesis and development

Highlight The two tomato GDP-D-mannose epimerase isoforms play specific roles in cell wall biosynthesis and plant development but participate similarly in ascorbate biosynthesis.


Introduction
Plants exhibit variations in vitamin C or ascorbic acid (AsA) content both between plant species and in different tissues. No viable ascorbate-less mutant has ever been reported, indicating that AsA is of crucial importance in plants. AsA is a major antioxidant, which ensures the protection of plant cells against reactive oxygen species (ROS) both generated during normal physiological processes as well as by biotic and abiotic stresses. In addition, ascorbate has been implicated in the regulation of key developmental processes involving cell division and cell expansion (Smirnoff, 2000). A possible link between ascorbate and plant growth control is the interconnection between the major ascorbate biosynthetic pathway first described by Wheeler et al. (1998) and the cell wall biosynthesis. The formation of GDP-D-mannose is the initial step in the pathway of ascorbate biosynthesis, whilst GDP-D-mannose is also a known precursor for the synthesis of D-mannose, L-fucose and L-galactose and therefore for the pectin rhamnogalacturonan-II (RG-II) and for hemicelluloses such as (galacto)glucomannans (Reiter and Vauzin, 2001).
In addition, GDP-D-mannose epimerase (GME) produces GDP-L-galactose from GDP-D-mannose and therefore is the crossroad between L-ascorbate and cell wall polysaccharide biosynthesis. GME is the most highly conserved protein involved in ascorbate biosynthesis (Wolucka and Van Montagu, 2007) and has raised considerable interest in recent years. Change of GME expression may also influence ascorbate synthesis in stress conditions  and adjust the plant balance between ascorbate and cell wall monosaccharide biosynthesis (Wolucka and Van Montagu, 2003). Thus, it has been hypothesized that, in association with VTC2 (GDP-L-galactose phosphorylase), the enzyme catalyzing the following step of ascorbate biosynthesis, GME, constitutes a control point for the regulation of the ascorbate pathway in plants (Laing et al., 2007;Wolucka and Van Montagu, 2007). Experimental evidence in support of these hypotheses are still lacking to date, possibly because knockout mutation of the two genes encoding the GDP-galactose phosphorylase VTC2 and VTC5 identified so far in Arabidopsis is lethal (Dowdle et al., 2007). Two GME genes named SlGME1 and SlGME2 are present in the tomato, Solanum lycopersicum, genome (Zou et al., 2006;Stevens et al., 2007) in contrast with most plants in which GME exists as a single copy (Watanabe et al., 2006). Analysis of tomato transcriptome data (e.g. from the tomato eFP browser, http://bar.utoronto.ca/efp_tomato/cgi-bin/efp-Web.cgi) shows that transcripts of GME1 and GME2 display specificity in their expression profiles according to the developmental stage and the organ considered (e.g. leaf, flower, fruits). Recently, Zhang et al. (2011) showed that the overexpression of both SlGME1 and SlGME2 enhanced AsA synthesis capacity and improved the tolerance of the plant to several abiotic stresses. Previously, Gilbert et al. (2009) demonstrated that simultaneous partial inactivation of the two SlGME genes in RNAi-silenced tomato lines displayed a 40-60% decrease of AsA content as well as growth defects affecting both cell division and cell expansion. Detailed analysis of the cell wall composition of these RNAi-silenced tomato lines revealed changes in the structure and the composition of hemicelluloses and pectins, as well as an alteration of the cell wall monosaccharide content, especially those directly linked to GME activity, such as D-mannose and L-galactose (Gilbert et al., 2009). Subsequently it was shown that in seedlings showing a growth deficiency, the silencing of the two SlGME genes resulted in a 60% decrease in terminal L-galactose in the side chain A of RG-II as well as in a lower capacity of RG-II to perform in muro cross-linking (Voxeur et al., 2011). The restoration of both the wild-type (WT) growth phenotype as well as an efficient in muro boron-mediated cross-linking of RG-II by supplementation of these GME-silenced lines with boric acid and not with L-galactose or AsA strongly suggested that the growth defect in GME-deficient tomato lines was most likely related to the alteration of the D-mannose-and L-galactosecontaining polysaccharides like hemicelluloses, such as mannans and the pectic RG-II, respectively, rather than to an L-AsA deficiency (Voxeur et al., 2011).
Pectins are complex acidic polysaccharides of the primary cell wall containing three distinct domains: homogalacturonan, rhamnogalacturonan-I (RG-I) and RG-II (Carpita and Gibeau, 1993). RG-II is the most structurally complex pectic polysaccharide composed of a homogalacturonan backbone substituted with four structurally different oligosaccharide side chains (A-D) (O'Neill et al., 2004). Despite its high complexity, RG-II represents 1-4% of the pectin-rich primary cell wall. It is evolutionary conserved as is present in the primary cell wall of all vascular plants (O'Neill et al., 2004). In the primary cell wall, RG-II exists predominantly as a dimer that is cross-linked by a borate diester between two apiosyl residues of the side chain A (Kobayashi et al., 1996;O'Neill et al., 1996;Ishii et al., 2001;Perez et al., 2003). This boronmediated cross-linking of RG-II induces the formation in planta of a three-dimensional pectin network that is believed to influence cell wall properties and plant growth (Funakawa and Miwa, 2015;references therein). The link between RG-II dimerization and plant development has been deduced mainly from studies of mutants affected in the RG-II biosynthesis, among them the Arabidopsis mur1 (murus1) mutant deficient in the synthesis of L-fucose (O'Neill et al., 2001).
In the present study, the specific silencing of each of the two GDP-D-mannose epimerases present in tomato leads to the conclusion that both SlGME1 and SlGME2 activities participate in the biosynthesis of AsA as they are functionally complemented in the respective RNAi-lines. Regarding the link of the GME activity with the primary cell wall metabolism, the situation is somehow different. Indeed, our data clearly showed that the two GME proteins contribute to the primary cell wall synthesis, however they strongly suggested the existence of a specialized cell wall-related activity of SlGME1 and SlGME2 according to the considered plant tissue. Several lines of evidence support this hypothesis and are discussed regarding the key role of GMEs in the intimate link between the metabolism of ascorbate and non-cellulosic cell wall polysaccharides.

Materials and methods
Plant material and culture Cherry tomato [Solanum lycopersicum L. cv. West Virginia 106 (WVa106)] plants were grown in a greenhouse or in vitro as described (Alhagdow et al., 2007). The RNAi-mediated silencing of tomato SlGME1 (Solyc01g097340) and SlGME2 (Solyc09g082990) genes was performed by stable transformation of tomato (Alhagdow et al., 2007), using a 253 bp fragment from 1238 to 1491 after the start codon (SlGME1) and a fragment of 285 bp from 1151 to 1436 (SlGME2). Polyploid plants or those with multiple T-DNA insertions were excluded from further analyses. Three lines were selected as homozygous transgenic lines and kept for further studies, they correspond to L-9, L-10 and L-12 for the P 35S :Slgme1 RNAi lines and L-7, L-10 and L-13 for the P 35S :Slgme2 RNAi lines. Two extra P 35S :Slgme1 RNAi lines, L-1 and L-3, were maintained as heterozygous T0 plants by cutting because they remained infertile. For in vitro culture, seeds of the homozygous transgenic lines and WT were treated for 10 min in 3% NaOCl (v/v), washed thoroughly three times for 10 min with sterile MilliQ water and sown on one-half strength Murashige and Skoog (MS) medium (Kalys-Duchefa) containing 3% (w/v) sucrose and 0.15% (w/v) Phytagel (Sigma) under a 16/8 h (light/dark) photoperiod under 400 µmol m −2 s −1 light at 24 °C. For boron supplementation assays on in vitro seedlings, the MS medium was supplemented or not with 1.5 mM boric acid. The same supplementation experiment was performed on 10-week-old plants obtained by taking cuttings from the original lines. On average six cuttings were used for each transgenic line and WT. The watering was carried out three times per week, but only once with a Liquoplant fertilizing solution (1.85 g l −1 , Plantin SARL Courthezon France) supplemented with a range of boric acid from 0.5 to 2 mM and twice using tap water at pH 6.

Flower pollination and aniline blue staining assay
The pollination was carried out by gentle vibration of the fully opened flower at anthesis, corresponding to the auto-pollination process. For the manual and the cross-pollination experiments, flower emasculation was performed on 11 mm closed flowers. Then, pollen grains from flowers of the same plant or another genotype were used and thoroughly applied to the stigma by scraping the inner face of a fully opened anther cone. To ensure the cross pollination, this operation was repeated with three distinct anther cones on the same emasculated flower. The presence of pollen grains on stigma, the growth of pollen tubes in the style and surrounding the ovules inside the ovary were analyzed by aniline blue staining as previously described by Johnson-Brousseau and McCormick (2004). After 6 to 8 h, the pollinated flowers were incubated in the fixing solution containing ethanol/acetic acid/water (70:5:25, v/v/v) for 2 h at room temperature. The fixed flowers were sequentially rehydrated in 70%, 50% and 30% ethanol for 30 min, and finally washed with distilled water three times for 5 min. The rehydrated flowers were deposited on a glass slide and treated overnight in a softening solution composed of 8 M NaOH, at room temperature in a wet chamber. After several washes with distilled water the pistil tissues were stained with decolorized aniline Blue solution [0.1% (w/v) aniline blue in 100 mM K 3 PO 4 buffer, pH 11] for 2 h in the dark (Johnson-Brousseau and McCormick, 2004). The UV illumination was used to detect the aniline blue-stained pollen grains and tubes using an inverted Leica DMI6000B microscope equipped with a Leica DFC300FX camera.

In vitro pollen germination and characterization of pollen grains
In vitro pollen germination was carried out according to Ye et al. (2009). Pollen grains were isolated from anther cones of newly opened flowers that were torn open, pinched and disseminated over a plate containing 3 ml of germination medium [20 mM MES pH 6, 3 mM Ca(NO 3 ) 2 , 1mM KCl, 0.8 mM MgSO 4 , 1.6 mM H 3 BO 3 , 2.5% sucrose and 24% PEG 4000]. The plate was incubated at 25 °C for 3 and 6 h. Pollen grains and pollen tubes were observed using an Axioplan microscope (Zeiss). Digital images were acquired with a Moticam 2300 CCD camera (Motic Deutschland GmbH, Germany) using Motic Images Plus 2.0 software. To calculate the germination rate, a minimum of 200 pollen grains were counted in each experiment and at least three independent experiments were conducted. For defining the germination percentage, it has been considered that protrusions of the pollen tube tip from the pollen grain aperture were regarded as positive germination (Gibbon et al., 1999). To estimate the pollen density presents in the stamens, one fully opened flower was vibrated three times over a microplate containing 1 ml of germination medium. Pollen grains were counted using a Malassez slide and this operation was repeated with at least six flowers of each genotype.

Histological analysis
Histological analysis was carried out according to Brukhin et al. (2003). Flower buds were fixed in FAA solution (4% paraformaldehyde, 50% ethanol, 5% acetic acid in 1× PBS), placed under vacuum for 10 min and incubated overnight at 4 °C, dehydrated in alcohol and embedded in paraffin (Paraplast plus, Sigma). At least ten buds were sampled and 8 µm-thick sections were stained with 0.05% toluidine blue and slides were observed under a microscope (Zeiss-Axioplan).

Ascorbate content
Mature leaves, fruits (20 DPA, red ripe) and seedlings were harvested at 12:00 am UT and thoroughly frozen in liquid nitrogen. Ascorbate content was measured according to Stevens et al. (2006) using 30 mg FW and 100 mg FW for the leaves and fruits, respectively. Table S1 at JXB online) were performed on several tomato organs, namely leaves, fruits, seedlings and flowers for both SlGME1 and SlGME2 genes as previously described by Gilbert et al. (2009). The efficiency of the two sets of primers for GME1 and GME2 are 95% and 93.5%, respectively, making applicable the expression comparison in all the tomato organs.

Isolation of RG-II
Dry seeds or freeze-dried seedlings were ground into liquid nitrogen and 100 mg or 10 mg of powder was mixed with 1 ml of 70% EtOH in a 1.5 ml Eppendorf, respectively. The homogenate was centrifuged at 13 000 g for 15 min and the pellet was washed with 70% ethanol until the supernatant was clear. This step was repeated three times. The pellet was then treated overnight at room temperature and under agitation with 750 µl of a mix of methanol/chloroform (1:1). An additional washing was performed in 100% acetone for 1 h. The resulting alcohol insoluble residue (AIR) was dried and weighted before being treated with 1 M Na 2 CO 3 at 4 °C for 16 h and then rinsed with water to reach a neutral pH. The pellet was suspended in 500 µl ammonium acetate buffer (0.05 M, pH 4.8) with 5 U endopolygalacturonase from Aspergillus niger (Megazyme, Illinois, USA) and incubated at 37 °C for 16 h. The solubilized material was concentrated in a centrifugal vacuum concentrator. The RG-II was suspended in ammonium acetate buffer.
Polyacrylamide gel electrophoresis of RG-II Polyacrylamide gel electrophoresis was performed according to Chormova et al. (2014) with few changes. A 10% polyacrylamide stacking gel was added to a 26% polyacrylamide gel. 8 µl of RG-II samples (corresponding to 4-7 mg and 1-1.5 mg AIR for seeds and seedlings, respectively) were mixed with 2 µl sample buffer [0.63 M Tris-HCl containing 0.25% (w/v) bromophenol blue and 50% (v/v) glycerol, pH 8.8]. 80 µg of RG-II extracted from lemon fruits was used as control. For monomerization, lemon RG-II was incubated for 16 h at room temperature in 0.1 M HCl (v/v). Before loading, the sample was neutralized with 0.1 M NaOH (v/v). Electrophoresis was conducted at 220 V for 130 min. The gel was then fixed in ethanol/ acetic acid/water (4:1:5) for 30 min and washed three times in 30% ethanol for 10 min and three times in water for 10 min each. Silver staining was performed as described by Chormova et al. (2014).

SlGME1 and SlGME2 display tissue-and temporalspecific expression patterns
To extend our understanding on the respective roles of the two GME proteins in tomato plants, analysis of their expression patterns was carried out in three distinct tomato organs during their development. Firstly, during the early period of the vegetative growth after germination, SlGME1 and SlGME2 display distinct expression patterns, as the expression of SlGME2 is the highest at the earlier stages (D+3, D+6) and then decreases slowly, whereas SlGME1 expression decreases as early as D+6 and thereafter is gradually reduced up to 20% of the initial level at D+3 (Fig. 1A). In flowers the scenario is different. During the initial stages of floral bud development, SlGME1 is 3-4-fold more expressed than SlGME2 and then it declines progressively to reach 20% of its initial level at stage S11 just before anthesis (Fig. 1B). On the contrary, during the same developmental phase, the expression of SlGME2 increases gradually by 2.5-fold up to stage S11. In fully opened flower at anthesis, both SlGME1 and SlGME2 genes display expression patterns that vary depending on the organ (Fig. 1C). Moreover, SlGME2 expression is always higher than SlGME1 in all the floral organs, especially in stamens, style and ovary, and at its weakest level in petals and sepals. Finally, in fruits, SlGME1 and SlGME2 expression is also different during the active growth phase of the tomato fruit from anthesis through the division and expansion phases (Fig. 1D). At anthesis, SlGME2 expression is still twofold higher than SlGME1 expression, but thereafter SlGME2 expression declines gradually up to almost the end of the expansion phase to reach 20% of its initial expression. In contrast, after anthesis SlGME1 expression displays a 4.4fold increase with a peak in the middle of the division phase at 4 DPA, after which its expression decreases in parallel with SlGME2 up to the end of the expansion phase at 27 DPA, returning to its initial value.

SlGME1 and SlGME2 participate equally in AsA biosynthesis
According to the distinct expression patterns of SlGME1 and SlGME2, it was tempting to hypothesize that the two GME proteins play specific functions in tomato, a question all the more relevant given that tomato displays two GME genes (Watanabe et al., 2006) whereas GME exist as a single gene in most plants studied so far. To further investigate this question, specific RNAi lines were generated for each SlGME gene using a specific RNA interference sequence fragment under the control of the CaMV35S promoter (named P 35S :Slgme RNAi transformants). From 15 independent primary P 35S :Slgme1 RNAi and P 35S :Slgme2 RNAi transformants (named T0) showing the presence of a single copy of the transgene, several independent lines for each gene were selected for further analyses on the T1 generation plants. However, for two lines among the selected P 35S :Slgme1 RNAi lines -L-1 and L-3 -this was not possible as they were infertile, producing small seedless fruits. Consequently, these plants were kept and propagated as T0 plants by taking cuttings. Analysis of SlGME transcripts in leaves and 20 DPA fruits from the T0 P 35S :Slgme RNAi lines showed, as expected, that the expressions of SlGME1 and SlGME2 were strongly reduced in the corresponding P 35S :Slgme RNAi lines, leading to 20% and 5% of residual expression in P 35S :Slgme1 RNAi and P 35S :Slgme2 RNAi lines, respectively (Fig. 2). The silencing specifically targeted the expected SlGME gene since in P 35S :Slgme1 RNAi lines the expression of SlGME2 was not reduced and vice versa. It is noteworthy that this is also the case during flower development and to a lesser extent in developing plantlets, since no modification of the SlGME1 expression level was observed in these P 35S :Slgme2 RNAi lines, and vice versa ( Supplementary  Fig. S1). Moreover, this analysis revealed that in organs of the P 35S :Slgme2 RNAi lines the expression of SlGME1 was not increased, suggesting the absence of any compensatory effect for the reduction of the SlGME2 transcript level ( Fig. 2A, Supplementary Fig. S1). This observation was also true for SlGME2 transcripts in the P 35S :Slgme1 RNAi lines. In contrast with the previous work of Gilbert et al. (2009) in which the two SlGME genes were RNAi-silenced resulting in a 50-60% reduction of the AsA biosynthesis capacity, in the present analysis no change in AsA content was measured either in the leaf or fruits in the two types of transgenic lines (Fig. 2B) and this suggests two possibilities. First, each of the GME proteins taken alone has the capacity to maintain a WT-like AsA biosynthesis, thus corresponding to functional redundancy between GME1 and GME2. Second, together they share this capacity but even with a reduced expression of 10-20% compared to the WT, the low protein level could be sufficient to maintain the AsA biosynthesis.

SlGME1 silencing resulted in the reduction of fruit size
Among the P 35S :Slgme1 RNAi transgenic plants, lines L-1 and L-3 exhibited severe fruit growth defects, which resulted in the development of very small and seedless fruits compared to the line L-9 and WT ( Fig. 3A; Supplementary Fig. S2). In addition, AsA concentration of fully developed flowers (S11) and fruits of L-1, L-3 and L-9 remained unchanged and equal to that of the WT fruits (Figs 2, 3A-C). The more severe fruit size phenotypes observed in lines L-1 and L-3 in comparison with line L-9 were not correlated with a stronger repression of GME1 gene (Fig. 3B). In contrast, L-9, L-10 and L-12 plants were able to produce bigger fruits containing seeds but still fewer than WT fruits and their size were 30-40% smaller than fruits from P 35S :Slgme2 RNAi lines L-7, L-12 and L-13 and the WT (Supplementary Fig. S2). The same decrease in fruit size has been described in fruits from the P 35S :Slgme RNAi plants in which the two SlGME genes were silenced (Gilbert et al., 2009). Thereafter, the male or female origins of these phenotypes was investigated through cross-pollination experiments using P 35S :Slgme1 RNAi lines L-1, L-3, L-9 and WT plants ( Fig. 4A; Supplementary Table S2). It appeared clearly that the defect in fertilization originated from the male gametophyte rather than the female counterpart. Indeed, in the two extreme P 35S :Slgme1 RNAi lines L-1 and L-3, manual pollination with their self-pollen led to the development of small fruits without seed (Fig. 4A), as observed in natural pollination conditions (Fig. 3A). When WT pollen was used to pollinate the flowers of the three P 35S :Slgme1 RNAi lines, fruit development was restored in term of fruit size even though the seed content was still low compared to the WT (Fig. 4A). When the P 35S :Slgme1 RNAi L-1 or L-3 line was used as pollen donor in the cross-pollination, fruit development was not restored, thus indicating a strong defect of the male reproduction organs in these P 35S :Slgme1 RNAi lines (L-1 and L-3). In contrast, fertilization was not significantly decreased in cross-pollination experiments using L-9 moderate P 35S :Slgme1 RNAi line as parent, except in the auto-pollination trial, which confirms also some fertilization defect in this genotype. Consistent with these results, the germination rates in comparison to the WT were significantly reduced in lines L-1 and L-3 but not in line L-9 (Fig. 4B). On the contrary, the values for pollen density differed little between these three P 35S :Slgme1 RNAi lines, but both were very low compared to the WT (Fig. 4B). Analysis by fluorescence microscopy of the pollination process showed that the pollen tube growth was lacking in lines L-1, whereas it was normal in line L-9 ( Supplementary  Fig. S3), thus confirming fertilization data from Fig. 4.   Fig. 3. Analysis of red-ripe fruits and expression of the SlGME genes of flowers from P 35S :Slgme1 RNAi lines and control. (A) Several traits of the red-ripe fruits produced by the P 35S :Slgme1 RNAi lines L-1, L-3, L-9 and control plants were measured: fruit diameter, seed number and ascorbate concentration (expressed in µmol g −1 FW). Data are the means ±SD of a total of ten fruits from each transgenic line and control. Asterisks indicate values that are significantly different from those of the control plants (t-test, P<0.05); ns, infertile fruits with no seed or containing tiny seeds unable to germinate. Scale bar, 1 cm. (B) The relative abundance of SlGME1 transcripts was determined in fully developed flowers (S11) of the P 35S :Slgme1 RNAi lines L-1, L-3, L-9 and compared to control (WT). Data are the means ±SD of three biological replicates with two technical repeats each and were obtained by quantitative RT-PCR normalized against EIF4a. (C) Ascorbate concentration assayed in flowers (S11). Data are the means ±SD of a total of 12 flowers from each transgenic line and control. The pollination of the fully opened flower at anthesis was carried out manually as described in 'Materials and methods' section. Thus, three pollen cones (male part) from flowers at anthesis of the same plant or from another genotype were used on emasculated flowers (female part). In this crosspollination method the first floral part named corresponded to the female acceptor and the second is the male donor (e.g. 'female' GME1-1 × 'male' WT). All data are the means of fruits from ten manually pollinated flowers. (B) Pollen content and germination level. Estimated pollen content per flower (n=10) of the P 35S :Slgme1 RNAi line L-9, the most affected P 35S :Slgme1 RNAi lines L-1, L-3 and the WT plants was determined as described in 'Materials and methods' section. The germination percentage was calculated by counting at least 200 pollen grains after 3 h incubation in 3 ml germination medium. Asterisks indicate values that are significantly different from those of the control (t-test, P<0.05).
Finally, histological analyses of flower buds from the most affected P 35S :Slgme RNAi lines L-1 and L-3 clearly confirmed the very low pollen density in the pollen sacs and the high proportion of pollen grains that remained at the tetrad stage, suggesting an impaired pollen development in comparison with the WT (Supplementary Fig. S4). All these results demonstrate that silencing the SlGME1 gene affects many aspects of the fertilization process. In both P 35S :Slgme1 RNAi lines L-1, L-3 and L-9 this defect originates from pollen quality, quantity and its ability to germinate ( Fig. 4; Supplementary  Fig. S4). Additionally, the deficiency of pollen tube growth observed in the P 35S :Slgme1 RNAi lines L-1 and L-3 may accentuate these defects and lead to the generation of very small seedless fruits (Supplementary Fig. S4). This is not the case for line L-9 where pollen tube growth is similar to the WT, and fruit growth is almost normal and like the WT (Supplementary Fig. S4). It is well known that once pollen set onto the stigma papillae, the pollen tube begins its growth through the transmitting track to reach and fecundate the ovules. This fecundation triggers an auxin burst that corresponds to fruit set and the start of seed development and fruit growth (Gillaspy et al., 1993).
According to Voxeur et al. (2011) who demonstrated that the growth defect of the P 35S :Slgme RNAi plants in which the two SlGME genes were RNAi-silenced could be reverted by addition of boric acid at 1.5 mM in the germination medium, boron supplementation trials were carried out with the most affected P 35S :Slgme1 RNAi L-1 and L-3 lines to reverse the fruit growth phenotype. Ten-week-old plants having two to three fully developing trusses were weekly watered with the boronsupplemented nutritive solution using a range of boric acid concentrations from 0.5 to 2 mM. After one month of treatment, no recovery was observed as the fruits produced never reached the size of the WT fruits whatever the boric acid concentration (data not shown). Additionally, long-term exposure to boron appeared to be toxic and resulted in plant death by defoliation, which was especially rapid at 2 mM boric acid.

P 35S :Slgme2 RNAi lines displayed a growth defect that can be rescued by boron supplementation
Analysis of plant development revealed that the size of the P 35S :Slgme2 RNAi plants after 6 weeks was reduced, whereas the P 35S :Slgme1 RNAi plants developed similarly to the WT (Fig. 5A). The growth delay was observed very early after the germination and was maintained from then on (Fig. 5B). It is noteworthy that the growth delay of the P 35S :Slgme2 RNAi plants was also comparable to that observed in the P 35S :Slgme RNAi lines (e.g. L-108) ( Fig. 5; Gilbert et al., 2009). As demonstrated previously, the growth defect of the P 35S :Slgme RNAi transformants (e.g. line L-108) originated from altered composition and structure of several cell wall polymers, including the hemicellulose mannans and the pectic RG-II (Gilbert et al., 2009). Voxeur et al. (2011) demonstrated that boric acid supplementation during germination could revert this growth defect, notably by restoring the RG-II dimerization state. Several studies also confirmed that supplementation with boric acid may restore RG-II dimerization, cell elongation and growth (Wimmer et al., 2009;Voxeur et al., 2012;Smyth et al., 2013). This effect of boron was also observed in the present study as 1.5 mM boric acid could compensate for the growth delay of the P 35S :Slgme2 RNAi L-7 and L-13, without affecting the growth of the P 35S :Slgme1 RNAi and WT plantlets ( Fig. 5B; Supplementary Fig. S5).
Consequently, the RG-II dimerization state was analyzed by PAGE in seeds and seedlings of the P 35S :Slgme1 RNAi and P 35S :Slgme2 RNAi lines. Alteration of the RG-II dimerization state was only visible in seeds of the P 35S :Slgme2 RNAi line, as shown by the level of monomeric form that was significantly increased, whereas in the P 35S :Slgme1 RNAi , like in the WT, the dimeric form of RG-II remained predominant and unchanged (Fig. 6). Thereafter, in the growing seedlings, no significant RG-II dimerization change was detected in P 35S :Slgme2 RNAi lines and PAGE patterns were similar to those of the P 35S :Slgme1 RNAi and WT. Thus, it seems that the seed germination and growth delay observed in P 35S :Slgme2 RNAi lines could partially originate from the alteration of the dimerization of RG-II. Finally, to investigate further to what extent SlGME2 may be the major form acting during the first stages of plant growth, the analysis of the abundance of the SlGME1 and SlGME2 transcripts was performed in the elongation zone of the tomato hypocotyl seedlings germinating and growing under light or darkness. When seedlings were germinated in darkness the fast growth led to etiolated hypocotyls in which the expression of SlGME1 strongly declined 4-fold, whereas that of SlGME2 remained high and similar to that under light conditions (Supplementary Fig. S6).

Discussion
In plants, GME is a control point between AsA biosynthesis and the non-cellulosic cell wall metabolism. This was definitively established in tomato by the previous works of Gilbert et al. (2009) andVoxeur et al. (2011). Indeed, silencing the two tomato SlGME genes led to a decrease in the plant capacity to synthesize AsA as well as to perform normal growth (Gilbert et al., 2009). This growth phenotype resulted from defects in the structure and composition of hemicelluloses (e.g. galactoglucomannans) and more specifically in the composition and the dimerization of the pectin RG-II (Gilbert et al., 2009;Voxeur et al., 2011). Gene expression analysis of the two SlGME genes in various tomato organs (Fig. 1) further suggested their possible specialization in the tomato plants. Such specialization of GME could be driven by a high demand for AsA and/or GME-related cell wall precursors in a specific tissue and/or developmental stage.
The present work addresses the question of the sub-functionality of the two tomato SlGME genes regarding these two distinct metabolisms. Towards this end, each of the two SlGME genes has been specifically repressed using an RNAi-silencing approach, thus allowing the study of their respective roles in AsA and cell wall biosynthesis.
SlGME1 and SlGME2 share an equal prevalence in the AsA biosynthesis pathway Repression of either SlGME1 or SlGME2 resulted in no change in AsA content in whole plants (Fig. 2), thus suggesting that in tomato both SlGME1 and SlGME2 participate in the production of AsA in all plant tissues. Recently Qin et al. (2015) revealed that in rice, among the three GMP genes encoding the protein that precedes GME in the Wheeler-Smirnoff pathway, only two participate in AsA biosynthesis. Moreover, each of the two GMP activities exhibits organ specificity in leaves and roots (Qin et al., 2015). In the same manner, Maruta et al. (2008) studied the activity of phosphomannose isomerase (PMI), a step upstream to GMP in the AsA pathway. Two PMI genes exist in Arabidopsis and their expressions display differences notably according to light intensity. Only RNAi-silencing of one of the two AtPMI genes affected AsA content in the plants. In most tomato organs, SlGME2 appeared as the major SlGME expressed gene (Fig. 1). Nevertheless, our data also showed that whatever the organ considered and the magnitude of the repression of the SlGME1 or SlGME2 gene expression, no compensatory change in expression of the other non-silenced SlGME gene could be observed. The absence of respective compensation between SlGME1 and SlGME2 genes strengthens the hypothesis that both SlGME1 and SlGME2 are involved in AsA biosynthesis and that each one is sufficient to provide enough AsA for correct functioning of all tomato organs in the conditions examined.

SlGME1 and SlGME2 exhibit organ-specialized cell wall-related roles
Regarding the role of GME in the cell wall biogenesis, the link between AsA metabolism and growth process remains complex and its regulation is far from well established. Several studies have described a possible role of AsA during growth, including cell division and cell expansion processes (Liso et al., 1988, Arrigoni et al., 1989Citterio et al., 1994;Innocenti et al., 1990). However, very little is known about the exact mechanisms by which AsA regulates cell growth in plants (Smirnoff, 2000;Foyer and Noctor, 2005). In 2 000, Smirnoff hypothesized a model where AsA may act as a redox buffer in the cell wall compartment in relation with apoplastic ascorbate oxidase activity and hydrogen peroxide signaling. More recently, Nunes-Nesi et al. (2008) proposed that the interaction between AsA and growth may involve AsA per se in combination with photosynthesis and respiration processes. Ascorbate may also participate in the crosstalk between redox and phytohormone signaling thus affecting plant growth (Considine and Foyer, 2014). However, in the Fig. 6. Effect of SlGME RNAi silencing on the RG-II dimerization detected by polyacrylamide gel electrophoresis (PAGE). Visualization of the polysaccharides was carried out by silver staining. RG-II was extracted from seeds and 4-8-day-old seedlings of WT, P 35S :Slgme1 RNAi L-9, and P 35S :Slgme2 RNAi lines L-7. Each lane on the gel corresponds to Alcohol Insoluble Residue (AIR) saponified and digested with the endopolygalacturonase. RG-II dimer extracted from lemon fruits treated or not with 0.1M HCL were used as controls, mRG-II: monomeric RG-II; dRG-II: dimeric RG-II. present context a causal relationship between AsA and plant growth alterations must be considered with caution. Indeed, supplementation with L-galactose or L-ascorbate of tomato lines silenced for both SlGME genes was shown to rescue WT levels of AsA but not to restore a WT growth phenotype (Gilbert et al., 2009). Moreover, Voxeur et al. (2011) demonstrated that supplementation of GME-silenced seedlings with boron restored WT growth, and this correlated with the recovery of the RG-II dimerization state. It is thus likely that the growth phenotypes observed in either P 35S :Slgme1 RNAi or P 35S :Slgme2 RNAi plants arise from a defect in cell wall biosynthesis, as previously described for SlGME-silenced plants (Gilbert et al., 2009;Voxeur et al., 2011).
GDP-D-mannose and GDP-L-galactose, respectively substrate and product of the GME enzyme, are not only used for L-ascorbate synthesis but also for the synthesis of noncellulosic cell wall polysaccharides and protein glycosylation (Smirnoff, 2000). Besides, GDP-D-mannose is also a precursor of L-fucose (Reiter and Vauzin, 2001). These three nucleotide sugars are used to incorporate D-mannose, L-galactose and L-fucose into several non-cellulosic cell wall polymers (Baydoun and Fry, 1988). Several lines of evidence demonstrate that in tomato the two GME proteins participate in cell wall biogenesis. Firstly, targeted silencing of each SlGME gene affects plant development but not ascorbate level. In addition, growth alterations are specific to each SlGME gene and take place in the organ in which differential expression of the corresponding gene is maximal, i.e. SlGME1 in developing flower and fruit and SlGME2 in growing seedlings and etiolated hypocotyl (Figs 1, 3, 5; Supplementary Fig. S6). Secondly, no compensatory expression was observed between SlGME1 and SlGME2 in the respective P 35S :Slgme RNAi plants ( Fig. 2; Supplementary. Fig. S1). Third, detailed analysis of RG-II in seeds and seedlings, in which SlGME2 deficiency leads to noticeable growth delay, indicates that developmental alteration is linked to a defect in RG-II dimerization (Fig. 6). Altogether, these results suggest specialization of each of the two tomato SlGME genes in specific organs and tissues. This sub-functionalization likely occurs at the transcriptional level and takes place in tissues or conditions in which there is a high demand for cell wall precursors (Matas et al., 2011;Kutschera and Niklas, 2013), e.g. pollen formation and early fruit development for SlGME1 or hypocotyl elongation and seedling growth for SlGME2. The prevalence of SlGME2 expression during dark-induced fast growth in etiolated hypocotyls strengthens this hypothesis ( Supplementary Fig. S6).

Is RG-II a key component connecting GME activity and specific organ development in tomato?
In plants, the role of the RG-II in growth as well as the role of boron in the dimerization of RG-II have been well documented (Funakawa and Miwa, 2015;references therein). Alteration of the expression of genes involved in RG-II biosynthesis was reported to impair male fertility (Delmas et al., 2008;Deng et al., 2010;Kobayashi et al., 2011;Liu et al., 2011), as well as the growth and stability of the pollen tube cell wall (Dumont et al., 2014). The availability of several mutants, mainly in Arabidopsis, harboring altered cell wall composition and structure allowed the determination of the function of RG-II in plants (Funakawa and Miwa, 2015;references therein). Importantly, all the alterations of the monosaccharide composition within the side chain A that is implicated in the crosslinking of RG-II by borate have been shown to result in plant dwarfism as well as impairment of pollen development (O'Neill et al., 2001(O'Neill et al., , 2004Ahn et al., 2006;Delmas et al., 2008, Kobayashi et al., 2011Voxeur et al., 2011;Dumont et al., 2014). As previously shown in Voxeur et al. (2011), GME deficiency leads to the decrease in terminal L-Gal content in the side chain A of RG-II as well as in a lower capacity of RG-II to perform cross-linking. Furthermore, supplementation of GME-silenced lines with boric acid was able to restore an efficient in muro boron-mediated cross-linking of RG-II. The characterization of several monosaccharide synthases and transferases involved in RG-II synthesis also demonstrated that boron is a key player in normal RG-II assembly through the maintenance of the RG-II structural integrity and stability (Dumont et al., 2015). There is therefore an intricate link between plant growth and RG-II cross-linking, which is impaired in plants showing deficiency in L-galactose in the RG-II side chain A and/or in boron.
In the plant kingdom, the mineral boron is predominantly found in the primary cell wall associated with RG-II (Kobayashi et al., 1996). During pollen development, the borate-RG-II complex is crucial for pollen tube elongation (Dumont et al., 2014), and in most plant species boron is highly present in the tip of the growing pollen tube (Iwai et al., 2006). Moreover, as recently suggested by Funakawa and Miwa (2015) and Dumont et al. (2015), boron could also act as regulator of mechanisms involved in cell growth and/or cell wall integrity-sensing. Beyond its role of physical crosslinking of the pectic network, the borate-RG-II complex might be perceived as a boron reservoir in the primary cell wall (Dumont et al., 2015). Thus, any alterations in the RG-II structure and composition, especially regarding its dimerization status, may impair this sequestrated-boron reservoir and consequently could impact cell wall elongation and integrity. This could be specifically the case in rapid growth processes such as germination and hypocotyl elongation (Matas et al., 2011;Kutschera and Niklas, 2013), as observed for the P 35S :Slgme2 RNAi lines ( Supplementary Figs S4-6). In Capsicum annuum, another species from the Solanaceae family, a study of the distribution of cell wall components using an immunological approach has revealed the preponderance of RG-II in walls of differentiating cells such as during pollen development and embryogenesis compared to proliferating cells (Bárány et al., 2010). Growth rescue by boron was impossible to prove in developing fruits, but several processes including pollen development, density and germination, as well as pollen tube growth, were impaired ( Fig. 4; Supplementary Fig  S3, S4). The RG-II structure and dimerization within the pollen wall have not been investigated in the present study, however the results clearly demonstrate the functional role for the SlGME1 protein during pollen development and fertilization, although further work is needed to define that function. Finally, it is tempting to assume that decreasing the SlGME1 activity may diminish the efficiency of in muro boron-mediated cross-linking of RG-II during pollen development and germination, and finally fertilization.
These findings emphasize the diversity of functions of RG-II in plant development, and provide a frame to explain why tomato SlGME genes underwent sub-functionalization in specific tissues. Actually, no RG-II-and/or ascorbate-deficient GME mutants have been described so far in plants, presumably because strong mutations would be lethal when the plant species harbors a single GME gene. As shown in this study, GME plays a crucial role in cell wall and AsA metabolism in tomato plants. The availability of GME-silenced plants in which the role of GME in cell wall biosynthesis in specific plant organs and conditions can be studied independently from its role in ascorbate biosynthesis will undoubtedly prove invaluable in the future.

Supplementary data
Supplemantary data are available at JXB online. Table S1. Sets of PCR primers used to amplify specific regions of the two SlGME genes and the reference gene EIF4a. Table S2. Flower pollination and fertilization. Figure S1. SlGME1 and SlGME2 expression in plantlets and flowers. Figure S2. Size of red-ripe fruits of P 35S :Slgme1 RNAi , P 35S :Slgme2 RNAi lines and control plants. Figure S3. Flower pollination, pollen tube germination and elongation. Figure S4. Histological sections of floral buds. Figure S5. Effect of exogenous boron supply on the growth of P 35S :Slgme RNAi transgenic and control plants. Figure S6. SlGME1 and SlGME2 expression in hypocotyls of WT tomato seedlings.