GDP-D-mannose epimerase regulates male gametophyte development, plant growth and leaf senescence in Arabidopsis

Plant GDP-D-mannose epimerase (GME) converts GDP-D-mannose to GDP-L-galactose, a precursor of both L-ascorbate (vitamin C) and cell wall polysaccharides. However, the genetic functions of GME in Arabidopsis are unclear. In this study, we found that mutations in Arabidopsis GME affect pollen germination, pollen tube elongation, and transmission and development of the male gametophyte through analysis of the heterozygous GME/gme plants and the homozygous gme plants. Arabidopsis gme mutants also exhibit severe growth defects and early leaf senescence. Surprisingly, the defects in male gametophyte in the gme plants are not restored by L-ascorbate, boric acid or GDP-L-galactose, though boric acid rescues the growth defects of the mutants, indicating that GME may regulate male gametophyte development independent of L-ascorbate and GDP-L-galactose. These results reveal key roles for Arabidopsis GME in reproductive development, vegetative growth and leaf senescence, and suggest that GME regulates plant growth and controls male gametophyte development in different manners.

. Expression pattern and subcellular localisation of GME in Arabidopsis. (A-J) Histochemical GUS activity in seedlings (A), inflorescences (B), flowers (C), pollen grains (D) and pollen tubes (E) from T1 transgenic Arabidopsis expressing the GUS reporter gene under the control of the GME promoter (P GME ::GUS) with GUS activity in wild type (I,J) as a control. (K) and (L) RT-PCR analysis (K) and real-time PCR analysis (L) of the GME expression level in Arabidopsis roots (R), stems (S), stem leaves (SL), rosette leaves (RL) and flowers (F). ACTIN2 was used as the normalisation or internal control. (M) Subcellular localisation of GME in epidermal cells from N. benthamiana leaves. GME-fused GFP driven by the 35S promoter (35S::GME-GFP) and 35S::GFP were expressed in N. benthamiana leaves, respectively. GFP fluorescence was detected 50 h after infiltration. Infiltration with buffer was used as a negative control. BF, bright field. (N) Arabidopsis protoplasts transformed without (Control) or with 35S::GME-GFP. Red fluorescence indicates chloroplasts.
We next performed reciprocal crosses to determine which type of gametophyte development was affected in GME/gme-1. When the stigmas of GME/gme-1 heterozygotes were pollinated with WT pollen grains, the progeny segregation ratio (GME/GME:GME/gme-1) was identical to the expected 1:1 (Table 1). Conversely, when pollen grains from GME/gme-1 heterozygotes were crossed onto WT stigmas, all progeny were wild type ( Table 1), suggesting that gme-1 is a null mutation that leads to complete failure of transmission of the male gametophyte, but not the female gametophyte. We also transformed GME/gme-1 heterozygotes with the coding sequence of GME driven by its native promoter (P GME ::GME) to obtain P GME ::GME transgenic plants in a gme-1 background (gme-1 P GME ::GME) ( Fig. 2B and C), demonstrating that GME complemented the transmission defect caused by the gme-1 mutation.
Taken together (Table 1, Fig. 2A-C and Supplemental Figs 1 and 2), these results show that Arabidopsis GME is essential for male gametophyte transmission. GME is required for pollen germination and pollen tube elongation. In Arabidopsis anthers, the microspore mother cells undergo sequential meiosis and mitosis to form tricellular pollen grains containing two sperm cell nuclei and one vegetative cell nucleus. Pollen grains from dehisced anthers are released onto the stigma of a carpel and germinate to form pollen tubes, which elongate and go through the stigma and transmitting tract, thereby delivering the two sperm cells to an ovary for double fertilisation [23][24][25][26] . Any disruption in this process will lead to failed male gametophyte transmission.
We next explored in which stages GME regulates male gametophyte development. Observation of the surface of mature pollen grains from WT, GME/gme-1 heterozygous and gme-2 homozygous plants by environmental scanning electron microscopy showed that all of the pollen grains from the gme mutants were oval-shaped with long indented lines on their surface as in wild type (Fig. 2D).
In vitro pollen germination assays were performed to explore whether the processes that occur after tricellular pollen grain formation are affected in gme mutants. As shown in Fig. 3A and B, the pollen grains from gme-1 heterozygous and gme-2 homozygous plants exhibited reduced germination rates (~39% for GME/gme-1 and ~65% for gme-2) compared with wild type (~80%). Consistently, a maximum of four pollen grains from the qrt/qrt pollen tetrads could germinate, while no more than two pollen grains from the qrt/qrt GME/gme-1 pollen tetrads could germinate ( Fig. 3D and E). Thus, GME is required for pollen germination.
Moreover, the pollen tube length in gme-2 homozygous plants was much shorter than that in wild type ( Fig. 3A and C), indicating that GME is required for pollen tube elongation. To confirm this conclusion, we performed in vivo pollen tube growth experiments. WT pollen tubes could reach the bottom of the transmitting tract (Fig. 3F), and WT siliques were consistently filled with seeds (Fig. 3F). In contrast, the pollen tubes of gme-2 plants could reach no more than half the length of the transmitting tract, and mature gme-2 siliques possessed few seeds that were mainly located in the upper part of the siliques (Fig. 3G).
Taken together (Fig. 3), these results demonstrate that GME is required for pollen germination and pollen tube elongation. In support of this conclusion, genetic complementation experiments showed that GME with its native promoter (P GME ::GME) could rescue in vitro pollen germination, pollen tube elongation, in vivo pollen tube elongation and seed setting in gme-2 plants and in gme-1/gme-2, which was generated by crossing GME/gme-1 with gme-2 ( Defective male gametophyte development in gme mutant plants is not due to an ascorbate deficiency. As GME is a key enzyme in L-ascorbate biosynthesis, we next explored whether the defects in male gametophyte development in the gme mutant plants was due to a deficiency in L-ascorbate. qRT-PCR analysis confirmed that GME expression was reduced in GME/gme-1 and gme-2 plants (Fig. 4A) and decreased in gme-2 pollen grains (Fig. 4B). The ascorbate contents in GME/gme-1 and gme-2 plants were reduced to 64% and 28% of the WT level, respectively (Fig. 4B), demonstrating that ascorbate biosynthesis was decreased in these Arabidopsis gme mutants. insertion in GME and the core structure of the P GME ::GME vector (GME under the control of the GME promoter). The grey rectangle, black rectangle and triangle represent the UTR, exon and T-DNA insertion site, respectively. The primers indicated by arrows were used to identify the genetic background of the gme-1 mutants in (B). (B) Genotyping of GME/gme-1, Col-0 wild type (WT), and transgenic P GME ::GME in a gme-1 background. gme-1 P GME ::GME was generated by transforming P GME ::GME into GME/gme-1 heterozygous plants. The primer pairs LP1/SailLB3 and LP1/ RP1, respectively, are specific for the gme-1 T-DNA insertion and WT GME. ACTIN2 PCR products were used as a control. (C) Six-week-old Col-0 WT and gme-1 P GME ::GME plants. (D) Environmental scanning electron microscopy of pollen grains at floral stage 13 from WT, GME/gme-1 and gme-2 plants. Bars = 20 µm. (E) The gme-1 mutation does not affect male meiosis, mitosis, pollen viability or pollen vacuole development at the tricellular stage. The panels from left to right show tetrad pollen grains at the tricellular stage stained, respectively, with DAPI, Alexander's stain, fluorescein diacetate/propidium iodide and neutral red. Bars = 20 µm.
We next investigated whether the application of ascorbate could restore male gametophyte development in gme plants. The inflorescences of WT and gme-2 plants were treated with L-ascorbic acid (ASA) and sodium ascorbate (NaSA), respectively. WT plants exhibited good fertility when treated without or with ASA or NaSA, while male fertility in gme-2 could not be rescued by ASA or NaSA (Fig. 4D). Thus, ascorbate application could not rescue male gametophyte development in gme-2.
In vitro pollen germination assays were performed to detect whether ascorbate could rescue pollen germination in qrt/qrt GME/gme-1. As shown in Fig. 4E-G, ASA application could not recover the pollen germination rate in GME/gme-1, and it even inhibited pollen germination at high concentrations. Similar results were obtained for NaSA (data not shown). These results indicate that the defects in male gametophyte development in gme plants cannot be attributed to an ascorbate deficiency.
We next analysed mutants of the ascorbate biosynthetic genes VTC2 and VTC5, which encode two GDP-L-Gal phosphorylases that function redundantly to control Arabidopsis ascorbate biosynthesis 2 , in order to verify that an ascorbate deficiency does not affect male gametophyte development. The double mutant vtc2 vtc5 contained only ~22% of the WT level of ascorbate, and it exhibited severe growth defects (Supplemental Fig. 3), demonstrating that ascorbate biosynthesis in vtc2 vtc5 was severely blocked. We also used vtc2/vtc2 VTC5/vtc5 and vtc5/vtc5 VTC2/vtc2 plants that were homozygous for one allele and heterozygous for the other to perform reciprocal crosses with wild type. As shown in Supplemental Table 1, regardless of whether WT plants or mutants were used as recipients, gametophyte transmission was unaffected, suggesting that the abolishment of ascorbate biosynthesis by the mutation of both vtc2 and vtc5 does not affect male gametophyte development.
Taken together ( Fig. 4 and Supplemental Table 1), our data demonstrate that an ascorbate deficiency is not responsible for the defects in male gametophyte development observed in gme mutant plants.
Boric acid and GDP-L-Gal cannot restore pollen germination and pollen tube growth in gme mutant plants. A previous study showed that the growth defects in GME-silenced tomato plants could be rescued by the application of boric acid, which promotes the boron-mediated in muro cross-linking of cell wall polysaccharides, but not by ascorbate 28 . We thus explored whether boric acid could rescue male gametophyte development in our gme mutant plants. In vitro pollen germination assays using qrt/qrt and qrt/qrt GME/gme-1 supplied with different concentrations of boric acid showed that qrt/qrt displayed high pollen germination rates (~65-67%), while qrt/qrt GME/gme-1 treated with different concentrations of boric acid exhibited low germination rates (~31-33%) (Supplemental Fig. 4A and B). These findings suggest that boric acid cannot restore pollen germination in qrt/qrt GME/gme-1.
Next, inflorescences from WT and gme-2 plants were treated with boric acid. Regardless of whether they were treated with or without boric acid, the WT siliques were large and full of seeds while gme-2 produced small siliques with few seeds (Supplemental Fig. 4C), indicating that boric acid supplementation could not restore male gametophyte development and fertility in gme-2.
As GDP-L-Gal is a precursor of cell wall polysaccharides (e.g., RGII) 19 , we also tested whether the application of GDP-L-Gal could restore pollen germination in GME/gme-1. Our results indicate that GDP-L-Gal was unable to recover pollen germination in qrt/qrt GME/gme-1 (Supplemental Fig. 4D).
Taken together, these data (Supplemental Fig. 4) demonstrate that treatment with boric acid or GDP-L-Gal cannot restore male gametophyte development in gme mutants.
Growth defects in gme mutants. We next investigated whether Arabidopsis GME regulates growth. As shown in Fig. 5, gme-2 homozygous plants exhibited retarded growth in terms of their rosette leaves, height, stem diameter and fertility (e.g., silique length and seed number per silique; Fig. 5). Compared with gme-2, gme-1/gme-2 plants showed much more severe growth defects, including a dramatically reduced rosette leaf size, thinner stems, shorter siliques and fewer seeds ( Fig. 5 and Supplemental Fig. 1). The growth defects of gme-2 and gme-1/gme-2 could be restored by genetic complementation with GME (Fig. 5). Thus, GME plays important roles in vegetative growth.
The growth defects in gme can be rescued by boric acid but not ascorbate. To explore the reason for the growth defects of the Arabidopsis gme mutants, we treated gme-1/gme-2 plants with boric acid, ASA or L-Gal, respectively. As shown in Fig. 6A and B, the growth defects of the gme-1/gme-2 mutant could be rescued by boric acid, but not by ASA or L-Gal, suggesting that the growth defects of the gme-1/gme-2 mutant were due to reduced in muro cross-linking of cell wall polysaccharides. On the other hand, the severe growth defects of the  ascorbate-deficient mutant vtc2 vtc5 could be recovered by ASA and L-Gal, but not by boric acid (Supplemental Fig. 5), suggesting that the growth defects of the vtc2 vtc5 double mutant were due to an ascorbate deficiency rather than in muro cross-linking of cell wall polysaccharides.
In conclusion, these results ( Fig. 6 and Supplemental Fig. 5) demonstrate that the growth defects of gme mutants are caused by reduced in muro cross-linking of cell wall polysaccharides.
Early leaf senescence in gme mutants. Further observation showed that gme-2 exhibited early senescence compared with wild type, while gme-1/gme-2 displayed a much more severe early senescence phenotype  Fig. 7A). As a decreased chlorophyll content is a typical physiological marker for senescence in plants, we measured the chlorophyll contents of our gme mutants. As shown in Fig. 7B, the chlorophyll contents of 15-day-old gme-2 and gme-1/gme-2 plants were similar to that in wild type, but the levels decreased more quickly than in wild type at later stages of growth (e.g., days 20, 25 and 30).

Discussion
GME converts GDP-D-mannose to GDP-L-Gal and GDP-L-gulose, which are intermediates of L-ascorbate biosynthesis [13][14][15] . GDP-L-Gal is also a precursor of the cell wall polysaccharide RGII 16,17 . Previous studies showed that the knock-down of both tomato GMEs (SIGME1 and SIGME2) by RNAi increased the level of mannose, decreased the contents of the precursors Gal and L-ascorbate, reduced the amount of RGI galactan side chains and down-regulated the cross-linking of RGII and methyl esterification of pectins in stems, resulting in retarded plant growth, leaf bleaching, fragility and reduced fruit size 19,28 . SIGME1 and SIGME2 control reproductive development and vegetative growth separately 34 . In this study, through analysis of various T-DNA insertional mutants of Arabidopsis GME/gme-1, GME/gme-2, gme-2, gme-1/gme-2, and the transgenic complementation lines gme-1 SCiEnTifiC REPORTS | 7: 10309 | DOI:10.1038/s41598-017-10765-5 P GME ::GME, gme-2 P GME ::GME and gme-1/gme-2 P GME ::GME, we show that GME controls male gametophyte transmission, plant growth and senescence in Arabidopsis.
Although both tomato SIGME1 and Arabidopsis GME control male gametophyte development, they control different stages of male reproductive development. Firstly, SIGME1-RNAi plants exhibited reduced pollen fertility 34 , while pollen grains containing gme-1 from the heterozygote GME/gme-1 failed to germinate and transmit (Figs 2 and 3, and Table 1), demonstrating that Arabidopsis GME controls haploid gametophyte development. Secondly, pollen grains carrying the gme-1 mutation developed to the tricellular stage but were unable to germinate; in comparison, pollen grains carrying the gme-2 mutation could germinate, but they produced short pollen tubes (Figs 2 and 3). On the other hand, SIGME1-RNAi resulted in a reduced density of pollen grains, which usually arrested at the tetrad stage but displayed germination rates above 60% 34 .
Interestingly, the defects in pollen germination and pollen tube elongation were not rescued by application of L-ascorbate or GDP-L-Gal in the gme mutants ( Fig. 4 and Supplemental Fig. 4), even though boric acid was able to rescue the in muro cross-linking capacity of cell wall polysaccharides and restore the growth defects of the Arabidopsis gme-1/gme-2 mutant (Fig. 6). Consistently, the absence of ascorbate (in the vtc2 vtc5 mutant) also had no effect on male gametophyte transmission (Supplemental Table 1). These findings suggest that GME regulates pollen germination and pollen tube elongation independent of both ascorbate biosynthesis and the in muro cross-linking of cell wall polysaccharides. The formation and modification of cell wall pectins, including RGI and RGII, affect pollen tube elongation [35][36][37] . It would be interesting to analyse the cell wall components of pollen grains from gme mutant plants in order to identify those components that are essential for GME-regulated pollen germination and pollen tube elongation.
nuclei by fluorescence microscopy, Alexander staining solution 38 or 0.5 µg/µL of fluorescein diacetate and 1 µg/µL of propidium iodide for pollen viability testing and with 0.02% neutral red for vacuole analysis.
Aniline blue staining of germinated pollen grains in pistils was performed as described previously 39 . Pollinated pistils were collected 16 h after pollination, fixed in a solution of ethanol:acetic acid (3:1) for 2 h, washed with distilled water three times, further softened with 8 M NaOH overnight and then washed with distilled water three times. The softened pistils were incubated with an aniline blue solution (0.1% aniline blue in 0.1 M K 2 HPO 4 -KOH buffer, pH 11) for about 3 h in the dark, and then observed with a Zeiss fluorescence microscope (LSM710; Carl Zeiss AG, Oberkochen, Germany).
In vitro pollen germination assay. In vitro pollen germination assays were performed as described previously 40 with modifications. Pollen grains were collected from flowers that had been dehydrated at room temperature for about 1 h, spread on the surface of agar medium (0.01% boric acid, 5 mM CaCl 2 , 5 mM KCl, 1 mM MgSO 4 , 10% sucrose and 1.5% low-melting agarose, pH 7.5), germinated at 22-24 °C for 12 h and then observed under a light microscope with a CCD imaging system. At least 500 pollen grains of each genotype were analysed for pollen germination rate and pollen tube length. The pollen germination medium was added with the indicated concentrations of boric acid, ASA and GDP-L-Gal sodium salt to test their effects on gme mutant pollen germination.
GUS staining. The −2990 bp promoter region of GME was amplified from Arabidopsis genomic DNA and inserted into pBI121 using HindIII and XbaI to generate P GME ::GUS. The primers used to generate the construct are listed in Supplemental Table 2. The construct was transformed into Arabidopsis by the Agrobacterium-mediated floral dip method. Histochemical staining for GUS activity in the P GME ::GUS transgenic plants was performed as described previously 41 .
Generation of GME transgenic plants. The P GME ::GME construct was generated by replacing the GUS gene in P GME ::GUS with the coding sequence of GME. The primers used to generate the construct are listed in Supplemental Table 2. The P GME ::GME construct was introduced into GME/gme-1 and GME/gme-2 heterozygous plants by the Agrobacterium-mediated floral dip method to generate P GME ::GME homozygous transgenic plants in gme-1 and gme-2 backgrounds. Next, gme-1 P GME ::GME was crossed with gme-2 P GME ::GME to generate gme-1/gme-2 P GME ::GME. Subcellular localisation of GME. The coding sequence of GME was cloned into pEGAD for fusion with GFP. Agrobacterium cells containing pEGAD or pEGAD-GME were incubated, harvested, resuspended in infiltration buffer (0.2 mM acetosyringone, 10 mM MES and 10 mM MgCl 2 ), infiltrated into Nicotiana benthamiana leaves with a needleless syringe 42 and incubated at 24 °C for about 50 h before observation for GFP fluorescence. The coding sequence of GME was cloned into pEZS to fuse it with GFP. Arabidopsis protoplasts were transformed with pEZS or pEZS-GME as described previously 43 and observed for GFP fluorescence with a Zeiss microscope (LSM710; Carl Zeiss AG). The primers used to generate the constructs are listed in Supplemental Table 2.
Ascorbate content measurement. Leaves of 5-week-old Arabidopsis plants were homogenised in 6% TCA (approximately 0.2 g FW mL −1 ) and centrifuged at 12,000 x g for 5 min. The ascorbate and dehydroascorbate contents were determined by iron (III) reduction 44 . Total ascorbate represents the sum of the ascorbate and dehydroascorbate contents.
Chlorophyll content measurement. For chlorophyll content measurement, the fifth leaves of WT, gme-2, gme-1/gme-2, gme-2 P GME ::GME and gme-1/gme-2 P GME ::GME plants at different growth stages (15, 20, 25 and 30 days) were harvested and measured as described previously 45 . qRT-PCR and RT-PCR analyses. For the analysis of GME expression in different plant tissues, roots, stems, rosette leaves, stem leaves and flowers from ~5-week-old Arabidopsis plants were harvested for RNA extraction, reverse transcription and subsequent qRT-PCR and RT-PCR analyses. For the qRT-PCR analysis of senescence-associated genes, leaves from 6-week-old Arabidopsis plants were harvested and used for real-time PCR. For the qRT-PCR analysis of GME in WT and gme mutant plants, 4-week-old plants and pollen grains from plants at floral stage 13 were collected, respectively. qRT-PCR analyses were performed with an ABI 7500 real-time PCR system as described previously 46 . The primers used for qRT-PCR and RT-PCR are listed in Supplemental Table 2. ACTIN2 was used as a normalisation or internal control. Accession numbers. The Arabidopsis Genome Initiative numbers for the genes mentioned in this article are as follows: GME (AT5G28840), ACTIN2 (AT3G18780), QRT1 (AT5G55590), VTC2 (AT4G26850), VTC5 (AT5G55120), CAB1 (AT1G29930), CAB2 (AT1G29920), RBCS (At1g67090), SAG13 (AT2G29350), SAG21 (AT4G02380) and SEN4 (AT4G30270). Data availability. All data generated or analysed during this study are included in this published article and the Supplementary Information files.