Rice ORMDL Controls Sphingolipid Homeostasis Affecting Fertility Resulting from Abnormal Pollen Development

The orosomucoids (ORM) are ER-resisdent polypeptides encoded by ORM and ORMDL (ORM-like) genes. In humans, ORMDL3 was reported as genetic risk factor associated to asthma. In yeast, ORM proteins act as negative regulators of sphingolipid synthesis. Sphingolipids are important molecules regulating several processes including stress responses and apoptosis. However, the function of ORM/ORMDL genes in plants has not yet been reported. Previously, we found that temperature sensitive genetic male sterility (TGMS) rice lines controlled by tms2 contain a deletion of about 70 kb in chromosome 7. We identified four genes expressed in panicles, including an ORMDL ortholog, as candidates for tms2. In this report, we quantified expression of the only two candidate genes normally expressed in anthers of wild type plants grown in controlled growth rooms for fertile and sterile conditions. We found that only the ORMDL gene (LOC_Os07g26940) showed differential expression under these conditions. To better understand the function of rice ORMDL genes, we generated RNAi transgenic rice plants suppressing either LOC_Os07g26940, or all three ORMDL genes present in rice. We found that the RNAi transgenic plants with low expression of either LOC_Os07g26940 alone or all three ORMDL genes were sterile, having abnormal pollen morphology and staining. In addition, we found that both sphingolipid metabolism and expression of genes involved in sphingolipid synthesis were perturbed in the tms2 mutant, analogous to the role of ORMs in yeast. Our results indicated that plant ORMDL proteins influence sphingolipid homeostasis, and deletion of this gene affected fertility resulting from abnormal pollen development.


Introduction
Orosomucoid (ORM) family proteins are ER proteins encoded by ORM and ORMDL (ORM-like) genes. These proteins are highly conserved from yeasts to plants to human. Recently, human ORMDL3 was reported as a genetic risk factor associated with asthma in diverse populations [1][2][3]. In addition, the expression of ORMDL3 gene was shown to be associated with the disease symptoms [4,5]. ORMDL3 is involved in pro-inflammatory diseases by binding and inhibiting sarco-endoplasmic reticulum Ca 2+ pump (SERCA), which reduce ER Ca 2+ concentration and increase unfolded-protein response (UPR), a process believed to induce inflammation [6].
In yeast, deletion of the ORM proteins showed susceptibility to agents that increase protein-misfolding in the ER [7]. In addition, deletion of the ORM proteins increased UPR and slowed ER-to-Golgi transport of the tested proteins though these effects were suppressed by high temperatures [8]. After treating with agents that increase protein-misfolding, ORM2 gene expression was increased significantly, indicating that it is up-regulated by UPR [8]. In addition, genetic interaction profiles of orm2D deletion mutants showed inverse correlation with the interaction patterns of yeast mutants with reduced LCB1 and LCB2 expression. Furthermore, yeast cells with over-expression of ORM1 or ORM2 showed genetic interaction profiles highly correlated with those seen in yeasts with reduced LCB1 and LCB2 expression, indicating that increased expression of ORM reduced LCB 1/2 activity [9]. LCB1 and LCB2 encode serine palmitoyltransferase, the first and rate-limiting enzyme in sphingolipid biosynthesis. Over-expression of ORM1 or ORM2 resulted in reduced LCB levels, while cells with deleted ORM1/2 had highly elevated levels of LCBs [9]. Thus, it was proposed that ORM proteins were negative regulators of sphingolipid synthesis [9]. Yeast cells with ORM1/2 deletions increased flux throughout the sphingolipid pathway, resulting in growth defects [9] though the precise metabolic impact of loss of ORM function on sphingolipid synthesis in yeast is currently ambiguous, since orm1D/orm2D mutants have been reported to have reduced [8] or elevated [9] levels of ceramides. In addition, alteration in ORM gene expression or mutations to their phosphorylation sites perturbed sphingolipid metabolism, leading to the hypothesis that ORMs play a central role in lipid homeostasis [9,10].
Sphingolipids are key cellular membranes and signaling molecules involved in several cellular activities, such as cell proliferation, cell differentiation, apoptosis, and stress responses [11][12][13][14]. In plants, sphingolipids are important for cell growth and the establishment of cell polarity through their contribution to the functional organization of the endomembrane system [15]. Accordingly, sphingolipids were reported to be involved in a trafficking pathway with specific endomembrane compartments and polar auxin transport protein [16]. Sphingolipids are also reported to be involved in responses to ABA and cold temperature [17][18][19], also regulated mineral ion homeostasis and programmed cell death [20,21].
Microsporogenesis has been reported to be controlled by sphingolipids in several plant species [22][23][24]. In Arabidopsis, the study of fbr11-2/lcb1-1 mutants showed that the fbr11-2 mutant, an allele of lcb1-1, was transmitted only through female gametophytes, and initiated apoptotic cell death in bi-nucleated microspores [23]. The FBR11/LCB1 expression was confined in microspores during microgametogenesis. These results suggested that SPT modulated programmed cell death plays an important role in the regulation of male gametophyte development [23]. In addition, Arabidopsis LCB2 loss-of-function mutant demonstrated that sphingolipids are important for gametophytic development by alterations in the endomembrane system of pollen [22]. Furthermore, sterile or severely reduced in fertility were observed in antisense or RNAi transgenic plants with repressed expression of dihydrosphingosine C4 hydroxylase 1(DSH 1), a key constituent of sphingolipids, which supported the role of sphingolipids on fertility in plant [25].
Male sterility facilitates hybrid seed production. Rice hybrids showed 20% higher yields than the best inbred varieties [26]. Hybrid rice is a promising alternative to increase food production to meet the demand for future need. Previously, we found that temperature-sensitive genetic male-sterile 2 (tms2) rice line (TGMS) lacked a 70Kb section of DNA that encoded 4 expressed genes, one of which is an ORMDL ortholog [27]. Although there are three ORMDL genes in rice, only one (LOC_Os07g26940) is located in the deleted region. In this report, we found that the LOC_Os07g26940 gene showed strong expression in rice panicles and anthers. In addition, LOC_Os07g26940 showed differential expression under low and high temperatures, corresponding to fertile and sterile conditions, respectively. When RNAi was used to suppress ORMDL genes, either in the deletion region or all ORMDL three genes in rice, the resulting transgenic plants showed abnormal pollen development that resulted in male sterility. In addition, we showed that sphingolipids metabolism and expression of genes in sphingolipids biosynthesis were perturbed in the tms2 mutants. Similar to yeast, our results indicated that the plant ORMDL gene modulates sphingolipid homeostasis, affecting fertility by controlling pollen development. Therefore, understanding the processes that control plant fertility could be useful in generating high yield crops by hybrid or by genetic engineering.

Expression analysis of candidate genes
Seven genes located in the 70 Kb loci associated with the TGMS tms2 phenotype were annotated to encode expressed proteins, four of which expressed in panicles [27]. In this study, we examined the expression of these four genes (putative cytochrome P450 77A3 LOC_Os07g26870, LOC_Os07g26930, ORMDL LOC_Os07g26940, and LOC_Os 07g26974) in anthers of wild type plants by RT-PCR, confirming the identity of the amplicons by sequencing. Our results indicated that only LOC_Os07g26930 and ORMDL LOC_Os07g26940 were expressed in anthers (data not shown). To determine whether LOC_Os07g26930 and ORMDL LOC_Os07g26940 were differentially expressed at low and high temperatures corresponding to fertile and sterile conditions, the expression of these genes was monitored in panicles and anthers of wild type plants grown in controlled growth room at 26 and 32uC, using qRT-PCR. Information from GRAMENE rice database indicated that LOC_Os07g26930 and ORMDL LOC_Os07g26940 have 1 and 3 splicing forms respectively. Primers specific for all these forms were designed and used for expression analysis by qRT-PCR, compared to Actin1 as internal control. The results showed that expression of LOC_Os07g26930 was constant at low and high temperatures in panicle or in anther of wild type plants (Fig. 1A), inconsistent with a role in TGMS. In the case of ORMDL LOC_Os07g26940, we detected all three splicing forms by qRT-PCR. The results showed that expression levels of ORMDL LOC_Os07g26940.1 was similar at low and high temperatures in the panicle. Interestingly in the anther, this splicing form showed much higher expression at low temperature (fertile condition) than that at high temperature (sterile condition) (Fig. 1A). ORMDL LOC_Os07g26940.2 and ORMDL LOC_Os07g26940.3 showed similar levels of expression at low and high temperatures in panicle or in anther, with a little higher at high temperatures in panicle for ORMDL LO-C_Os07g26940.2. While ORMDL LOC_Os07g26940.3 showed a little higher expression in the anther at high temperature (sterile condition) than at low temperature (fertile condition) (Fig. 1A). To further study tissue-specific expression of ORMDL LO-C_Os07g26940, multiple different rice tissues were used for qRT-PCR analyses including leaves, stems, roots, anthers, small panicles, medium panicles, and large panicles. Using elongation factor 1 a, EF-1a as a reference gene, the results indicated that all the three transcripts derived from ORMDL LOC_Os 07g26940 were detected in most of the tested tissues, except in root where we could not detect expression of ORMDL LOC_Os 07g26940.2. The three forms were predominantly accumulated in the panicle and anther (Fig. 1B).

Phylogenetic Analysis
LOC_Os07g26940 is a member of the ORMDL protein family, which is highly conserved in eukaryotes. Information from the GRAMENE rice database indicated that ORMDL LO-C_Os07g26940 has three splicing forms, and the results from sequence alignments of these transcripts indicated that the three transcripts of ORMDL LOC_Os07g26940 shared the same sequence for about half of the gene starting from the beginning of the transcripts (Fig. S1). Previously, a BLAST search using the predicted amino acid sequences of ORMDL LOC_Os07g26940 revealed two other sequences (encoded by LOC_Os04g47970 and LOC_Os02g45180) with high percent sequence similarity to ORMDL LOC_Os07g26940, suggesting that at least three ORMDL genes are present in rice [27]. ORMDL LO-C_Os04g47970 has two splicing forms but ORMDL LO-C_Os02g45180 has only one. Similarly, results from comparing the two transcripts of ORMDL LOC_Os04g47970 showed that they share about 75% of the sequences starting from the beginning of the transcripts (Fig. S2). The rice transcripts of ORMDL showed high homology in the coding sequences (Fig. S3). To understand the evolutionary relationships of this gene family in rice and other species across kingdoms, phylogenetic analysis was constructed, using deduced amino acid sequences from the longest transcript of each rice ORMDL. This analysis showed the phylogenetic relationship between rice ORMDL and other ORMDL from different species selected from NCBI Reference Sequence database. The results showed 4 main groups of genes from human, drosophila, yeast, and plants. As expected, rice ORMDL is closer to Z. mays and to S. bicolor than to A. thaliana. Interestingly, ORMDL LOC_Os07g26940 is in a separate clade from the other two rice ORMDL (Fig. 2).

Expression analysis of rice ORMDL genes
To determine the patterns of expressions of rice ORMDL genes in tms2 and wild type rice plants, expression of these genes by qRT-PCR was monitored in anthers of tms2 and wild type plants grown in controlled growth room at 26 and 32uC. Using EF-1a as the reference gene, the results indicated that all the three ORMDL genes in rice showed similar patterns of expression by having higher expression at low temperature (26uC) than that at high temperature (32uC) conditions, and these genes showed higher expression in tms2 than in wild type plants in both conditions. As expected, expression of ORMDL LOC_Os 07g26940.1, used as a representative splicing form of ORMDL LOC_Os 07g26940, was not detected in the tms2 mutant rice plants (Fig. 3).

RNAi transgenic rice plants
To investigate the role of LOC_Os07g26930 and ORMDL LOC_Os07g26940 on pollen development, RNAi technology was used to down-regulate LOC_Os07g26930 or ORMDL LO-C_Os07g26940 expression in rice. For the gene-specific ORMDL LOC_Os07g26940 RNAi construct, sequence present only in the transcripts of ORMDL LOC_Os07g26940, but not in ORDML genes (LOC_Os04g47970 and LOC_Os02g45180) were used (Fig. S3). As this sequence is different from the other two rice ORMDL genes, we predicted that the RNAi only affects ORMDL LOC_Os07g26940 expression. However, because three ORMDL genes are present in rice, knocking down only the ORMDL LOC_Os07g26940 expression may not reveal all phenotypic effects. Thus, we generated an another RNAi construct using conserved sequences present in all rice ORDML genes (Fig. S3), to universally down-regulated all the ORDML gene transcripts. These RNAi sequences were under the control of ubiquitin promoter [28]. These RNAi constructs were transformed in to wild type (WT) Nipponbare plants. We obtained 27, 8 and 6 transgenic lines for LOC_Os07g26930-specificRNAi, LO-C_Os07g26940-specificRNAi, and ORMDL-specific RNAi, respectively. In our experience, it was much more difficult to obtain LOC_Os07g26940-specificRNAi, and ORMDL-specific RNAi, transgenic lines than that from LOC_Os07g26930-specificRNAi construct. In natural condition, at flowering, none of transgenic plants for LOC_Os07g26930-specificRNAi were sterile. For, LOC_Os07g26940-specific RNAi, or ORMDL-specific RNAi, two independent transformed lines from each construct were selected for further studies on the basis of low levels of expression of the target genes and a single copy of the transgene construct. As expected, expressions of all three splicing forms of ORMDL LOC_Os07g26940 were reduced in LOC_Os07g26940-specific RNAi transgenic plants, while expression of the other two ORMDL were not lower than that of wild type plants (Fig. 4A). For ORMDL-specific RNAi, expressions of all three ORMDL were reduced ( Figure 4B). At flowering, the pollen of these plants was different from that of wild type rice plants in both morphology and staining. The wild type pollen was round, showing the presence of starch granules, while pollen of LOC_Os07g26940-specificRNAi, or ORMDL-specific RNAi, was abnormal showing small pear-shaped pollen grains lacking starch (Fig. 4C). In addition, these T0 transgenic plants were sterile when they were grown under glass house conditions. Thus, they were ratooned from one T0 plant to at least 6 plants from each of the original T0 transgenic plant. The plants were grown under glass house conditions for approximately one month. Subsequently, three  Rice ORMDL Regulates Fertility PLOS ONE | www.plosone.org plants from each line were moved to grow in control growth room at 26 and 32uC. At flowering, these plants were sterile, with no plants capable of setting seed. The panicle tissue of the T0 was not used for direct measurement as they were retained to observe seed setting and to allow for T1 seed production.

Sphingolipid metabolism is perturbed in the tms2 mutant and in RNAi rice lines
Recently, yeast ORM proteins were reported as negative regulators of sphingolipid synthesis [8,9]. Since one ORMDL gene was deleted in the rice tms2 mutant, one scenario is that this may affect or alter sphingolipid synthesis by releasing LCB synthesis from homeostatic regulation. To investigate whether the tms2 mutant affects sphingolipid composition, total sphingolipid long-chain bases (LCBs) were analyzed from leafs and panicles of wild-type and tms2 mutant rice plants. Sphingolipids comprise an LCB and an amide-linked a fatty acid moiety, usually also decorated with a sugar head group. The LCB component of sphingolipids is unique to these lipids, and its synthesis in yeast has been shown to be modulated by ORM1/2. Analysis of total LCBs showed that no significant difference in sphingolipid LCB composition was observed in leaves of tms2 mutant and wild type rice plants (Fig. 5A). Importantly, panicles of tms2 mutant rice plants had altered sphingolipid LCB composition, with higher levels of d18:2, t18.1c, and t18:1t but lower levels of t18:0, compared with wild type rice plant (Fig. 5B). The other LCB species were similar in tms2 mutant and wild type rice plants (Fig. 5B). In view of the alteration of LCB composition in the tms2 mutant, and also previous studies in yeast which revealed altered ceramide levels in orm1D/orm2 D mutants [8,9], more targeted analysis was carried out, to measure the ceramide levels in this mutant and also RNAi-suppressed lines. The results showed that both RNAi-suppressed lines and tms2 mutant plants showed a reduced accumulation of ceramides, relative to the WT (Fig. 6A-6B; note that in the case of the RNAi lines, analysis was carried out  on leaf material). This was true for RNAi which suppressed all rice ORMDL transcripts, or specifically just the transcript derived from LOC_Os07g26940 and absent in tms2. Breakdown of these data is shown in table 1 and 2. These data confirm the specific role for LOC_Os07g26940 in modulating ceramide levels in rice.

Expression of genes in sphingolipid synthesis pathway
Rice ORMDL gene LOC_Os07g26940 is proposed, on the basis of homology, to be a regulator of sphingolipid synthesis, and in line with this, deletion of this gene perturbed sphingolipid metabolism in the tms2 mutant. Thus, we reasoned that the deletion might consequently also affect the expression of genes involved in sphingolipid biosynthesis. To determine the expression of genes involved in sphingolipid metabolism, we used genes involved in early steps of sphingolipid synthesis in Arabidopsis, which were homologous to those genes in S. cerevisiae [29], and then searched for homologs of these genes in rice using Arabidopsis and rice database (http://www.arabidopsis.org/; http://www.gramene.org/). A total of 12 rice genes encoded for four enzymes involved in early steps of sphingolipid synthesis were identified (Table 3) and tested for expression analysis in anthers.
The expression of 9 genes encoding for four enzymes were detected (Fig. 7A-7E), while the primers specific for the other three genes including serine palmitoyltransferase LOC_Os 01g70370, serine palmitoyltransferase LOC_Os 01g70380, and ceramide synthase LOC_Os 02g49590 were unable to provide good amplification. Two rice orthologs (LOC_Os 02g56300 and LOC_Os 10g11200) of the LCB1 subunit of serine palmitoyltransferase (SPT), showed higher expression at the non-permissive (i.e. sterile) temperature (32uC) in the tms2 mutant compared with the permissive (fertile) temperature (26uC), and these LCB1 transcripts were higher than those of wild type plants. Another rice ortholog of the LCB1 (serine palmitoyltransferase LOC_Os 03g14800) showed similar expression levels in 26uC and 32uC both in wild type and mutant rice plants, and similar levels of expression in wild type and mutant rice plants (Fig. 7A). Interestingly, a rice ortholog (serine palmitoyltransferase LOC_Os 11g31640) of the LCB2 showed higher expression at permissive (fertile) temperature (26uC) compared with the non-permissive (i.e. sterile) temperature (32uC) in both tms2 mutant and in the wild type plants, and expressions of this gene were much higher in the mutant than those in the wild type plants at both conditions (Fig. 7B). A rice ortholog (3-ketosphinganine reductase LO-C_Os02g47350) of TSC10, showed similar levels of expression at permissive (fertile) temperature (26uC) compared with the nonpermissive (i.e. sterile) temperature (32uC) in tms2 mutant, but this gene showed higher expression at permissive (fertile) temperature (26uC) compared with the non-permissive (i.e. sterile) temperature (32uC) in wild type plants (Fig. 7C). A rice ortholog (sphingolipid-C4-hydroxylase LOC_Os06g12250) of SUR2 showed higher expression at the non-permissive (i.e. sterile) temperature (32uC) compared with the permissive (fertile) temperature (26uC) in the mutant, but in wild type, this gene showed higher expression at permissive (fertile) temperature (26uC) compared with the nonpermissive (i.e. sterile) temperature (32uC). Surprisingly, the other tested rice ortholog (sphingolipid-C4-hydroxylase LO-C_Os02g51150) of SUR2 showed higher expression at permissive (fertile) temperature (26uC) compared with the non-permissive (i.e. sterile) temperature (32uC) in both tms2 mutant and in the wild type plants but expression of this gene were much higher in the mutant than those in the wild type plants at both conditions (Fig. 7D). For ceramide synthase, a rice ortholog (LO-C_Os02g37080) of LAC1/LAG1, showed higher expression at the non-permissive (i.e. sterile) temperature (32uC) in the tms2 Table 1. Ceramide composition in panicles of Indica WT and tms2 rice lines.  mutant compared with the permissive (fertile) temperature (26uC), and this expression was much higher than that of wild type plants.
In wild type, this gene showed similar levels of expression, with very low level of expression at both conditions. The other tested rice ortholog (ceramide synthase LOC_Os02g15750) of LAC1/ LAG1 showed higher expression at permissive (fertile) temperature (26uC) compared with the non-permissive (i.e. sterile) temperature (32uC) in both tms2 mutant and in the wild type plants, and this gene showed similar levels of expressions in tms2 mutant and in the wild type plants for each condition (Fig. 7E).

Discussion
The temperature-sensitive genetic male sterility (TGMS) system provides a powerful tool for the production of hybrid rice, in turn leading to greater yields of this vital crop. However, the molecular mechanism(s) of TGMS are currently unknown, precluding the facile incorporation of this trait into germplasm. A previous study reported a TGMS phenotype of Ugp1-cosuppressed plants resulting from aberrant transcripts, which undergo temperaturesensitive splicing in florets [30]. A rice ORMDL, LO-C_Os07g26940 was predicted to have three splicing variants (http://www.gramene.org). Accordingly, we detected all the three transcripts by qRT-PCR. Interestingly, one splice variant showed strongly differential patterns of expression in anther at low and high temperatures, implying a role for ORMDL LO-C_Os07g26940 in response to temperatures in anther. Importantly, this gene showed high expression in panicle and anthers which, in addition to modulation by temperature, coincides with the TGMS phenotype of tms2.
The TGMS phenotype of tms2 mutant rice plants was reported to be controlled by a recessive gene located in chromosome 7. Previously, the tms2 mutant plants were reported to contain at least 70 kb deletion including seven genes annotated as expressed proteins [27]. However, only two of these genes (LO-C_Os07g26930 and ORMDL LOC_Os07g26940) were strongly expressed in panicles and anthers and only the ORMDL gene LOC_Os07g26940 showed differential expression in fertile and sterile conditions in anthers, suggesting a function in this tissue and in TGMS. Accordingly, knock down function of this gene or all ORMDL genes including LOC_Os07g26940, LOC_Os04g47970, and LOC_Os02g45180, by RNAi, affected pollen development resulting in male sterility. Reduction of only ORMDL gene expression in chromosome 7 (LOC_Os07g26940) or all ORMDL genes affected pollen development in both morphology and staining. Pollen grains of LOC_Os07g26940 RNAi and ORMDL RNAi were different in sizes, shapes and staining, with none of these plants setting seeds. Interestingly, the male sterility of these RNAi plants was not modulated by temperature, with LO-C_Os07g26940 RNAi and ORMDL RNAi transgenic plants sterile at both low and high temperatures, suggesting that at least one other factor is needed for the conditional male sterility observed in the TGMS phenotype of tms2. Thus, deletion of ORMDL LOC_Os07g26940 and this (currently unknown) factor are critical for the TGMS phenotype. Although genetic analysis indicates that the TGMS phenotype of tms2 mutant plants was controlled by a single gene [31,32,27], it is possible that this phenotype is controlled by two closely linked genes not resolved by the current mapping population. Thus, a bigger population with greater recombination maybe required to further study this at the genetic level. Alternatively, the apparent inability of RNAi suppression of ORMDL to completely phenocopy the tms2 TGMS may indicate subtle requirements for the total absence of this gene to deliver conditional fertility.
Recently, ORMDL was reported as a negative regulator of sphingolipid synthesis by inhibiting serine palmitoyltransferase (LCB1/LCB2/TSC3) [9]. Over expression of ORM1 and ORM2 in yeast reduced LCB and ceramide levels. Conversely, deletion of ORM1/2 elevated the levels of LCB and ceramides [9], although very similar studies have also reported decreased ceramides (in conjunction with elevated LCBs) in yeast mutants lacking ORMs [8]. Since a rice ORMDL gene is deleted in the tms2 mutant, an alteration to sphingolipid homeostasis might be predicted. Accordingly, we found that sphingolipid synthesis was perturbed in the tms2 mutant. Panicles of tms2 mutant rice plants had higher d18:2, t18.1c, and t18:1t but lower t18:0 than the wild type rice plants, although the overall level of LCBs was not significantly altered. Closer examination of the discrete ceramide species present in the panicles of WT and tms2 mutant confirmed a concomitant decrease in ceramides levels, in agreement with total ceramide analysis, but also showed some discrete variation in the accumulation of individual ceramides. For example, the predominant C20:1/d18:2 ceramide species is reduced by ,30% in the tms2 mutant. This is particularly noteworthy, given the total LCB analysis of the tms2 mutant revealed a moderate increase in this accumulation of this sphingoid base. However, since it is currently unclear if LCB modifying enzymes such as the sphingolipid D8desaturase and the C4-hydroxylase utilise free LCBs and/or ceramides as substrates, it is not possible to make causal associations between these two different metabolic pools, or how these are impacted by the tms2 mutation. It is interesting to note that the rice LAG1/LAC1 ceramide synthase ortholog LO-C_Os02g37080 showed upregulation in tms2, and although the substrate-specificity of this enzyme remains to be determined, it may play a role in the synthesis of a discrete subset of ceramides as seen in Arabidopsis [16]. More strikingly, and in clear agreement with the yeast studies of Han et al [8], we observed a significant reduction in the ceramide levels in rice plants with perturbation to the expression of the ORMDL gene LOC_Os07g26940 -this was true for both the tms2 mutant and also lines in which RNAi was used to suppress transcript accumulation. On the basis of these two independent interventions, we conclude that LOC_Os07g26940 is specifically involved in the sphingolipid homeostasis, most likely in a manner analogous to that proposed by Han et al. [8].
In addition, the deletion in tms2 mutant increased expression of LCB1 in high temperature condition compared to that in wild type plants, and increased LCB2 both in low and high temperature conditions in agreement with its previously established role in yeast [9]. Furthermore, the tms2 deletion also affected expression of other genes involved in early steps of sphingolipid synthesis such as SUR2 and LAC1/LAG1. Our results indicated that most of the genes involved in early steps of sphingolipid synthesis showed higher expression in the mutant than that in the wild type plants, particularly under high temperature treatment. In Arabidopsis, LBC1 and LCB2, encoding subunits of serine palmitoyltransferase (SPT), were reported as important genes for male gametophyte development, involved in programmed cell death during pollen mitosis [23]. In addition, Arabidopsis LCB2 null mutants resulted in gametophytic lethality by altering the endomembrane system of pollen and loss of pollen viability, which occurred early in pollen development during transition from the uni-nucleate microspore to the bicellular pollen grain [22]. Deletion of ORMDL gene in tms2 mutant rice plants and RNAi suppression of ORMDL gene expression affected pollen development, resulting in sterility. The precise mechanism by which ORMDL LOC_Os07g26940 modulates aspects of male fertility still remain to be resolvedbased on our data, it appears that, like their yeast orthologs, higher plant ORMs are involved in sphingolipid homeostasis. However, the consequences of release from ORM regulation (reduced ceramides, and an altered LCB profile) is distinct from that observed in yeast, pointing the way for future studies. Given the significantly larger sphingolipidome of plant compared with S. cerevisiae, it is perhaps unsurprising that higher plants have evolved additional tiers of regulation for the control of these bioactive lipids.
In conclusion, our results indicate that tms2 mutant plants lacking one member of the rice ORMDL gene family has altered sphingolipid composition. In addition, seven out of the nine expressed genes involved in sphingolipid synthesis pathway showed higher expression in the mutant compared to wild type plants, in agreement with the role of ORM as a homeostatic regulator of sphingolipid biosynthesis [9]. Our results indicate that tms2 mutant plants lacking ORMDL LOC_Os07g26940 affected the expression of several genes involved in early step of sphingolipid synthesis pathway, although the tested genes (orthologous for LCB1, SUR2, and LAC1/LAG1) had discrete expression patterns, suggesting an aspect of sub-functionalization most likely associated with plant-specific sphingolipid homeostasis. Our findings demonstrate an emerging role for ORMDL genes in plants as regulators of ceramide accumulation and also provide support for the important role of sphingolipids on pollen development.

Plant materials and growth conditions
Wild type rice (Oryza sativa ssp japonica cv. Nipponbare, and indica ssp cv. Pathumthanee1) and tms2 mutant (in indica background and backcrossed to Pathumthanee1, BC2F4) were planted in natural green house, and about one month before flowering some of these plants (at least 5 plants/lines/condition) were moved to controlled growth rooms at temperature 26uC or 32uC under 80% relative humidity, 12 h light/12 h darkness with a day until flowering.

Expression analysis by quantitative RT-PCR
For expression of LOC_Os07g26930 and ORMDL LO-C_Os07g26940, specific primers for LOC_Os07g26930 and each transcript of ORMDL LOC_Os07g26940 were used for expression analysis in panicle and anther of wild type plants (Pathumthanee 1) grown in 24 and 32uC control growth rooms. Primers specific for LOC_Os07g26930 and for each transcript of ORMDL LOC_Os07g26940 were designed using Primer-BLAST (http://www.ncbi.nlm.nih.gov/tools/primer-blast/). In addition, primers specific for each transcript of ORMDL LO-C_Os07g26940 were used to test expression in other tissues of wild type plants such as root, stem, panicle, leaf, and anther. For expression of all the three ORMDL genes in tms2 and wild type (Pathumthanee 1), primers specific for each ORMDL genes, were used and total RNAs were extracted from anther of tms2 and wild type plants grown in 24 and 32uC control growth rooms. For expression of ORMDL genes in RNAi transgenic rice plants, total RNA were isolated from leaf, and qRT-PCR were performed using specific primers for each transcript of Os07g26940 or primers specific for each of the other two ORMDL genes. For expression of genes involved in early steps of sphingolipid synthesis, primers specific for 12 rice genes identified as homologous genes of Arabidopsis, which are homologues to known genes involved in sphingolipid synthesis in S. cerevisiae were used for qRT-PCR using total RNAs extracted from anther of tms2 and wild type plants (Pathumthanee 1) grown in 24 and 32uC control growth rooms. For all the data presented, one sample was a mixture of 3 plants for each genotype. Total RNAs were isolated using TRIZOL TM reagent (Invitrogen) following the manufacture's instructions. First-strand cDNA was synthesized using cDNA synthesis kit (Fermentas) and cDNAs transcribed from the total RNA (20 ml reaction volume). Gene specific primer were designed for amplicons about 200 bp. qRT-PCR experiments were performed on the BIO-RAD CFX 96 real time PCR systems, using SsoFast EvaGreen Supermix (Bio-Rad, Singapore). At the end of each experiment a melting curve was determined for each primer pair at a temperature stage from 60uC to 95uC to check the specificity of annealing. Primers targeting Actin1 (Os05g36290) were used to normalize the expression data of each gene in wild type plants. Because tms2 mutation affects expression of Actin1, thus as recommended [33] primers targeting Elongation Factor 1 a, EF-1a (Os03g08020) were used to normalize the expression data of genes compared expression in the mutant and wild type plants. For each gene, at least two biological replicates were performed, each with three technical replicated. Similar results were obtained in the repeated experiments. The average Ct values and standard deviation were calculated from three technical replicates. The quantification of gene expression was performed using the relative quantification method (22DDCT) [34]. All the tested primers were listed in Table S1.

Phylogenetic Analysis
Phylogenetic Analysis was performed using protein sequences encoded by the longest splicing forms of each rice ORMDL obtained from GRAMENE rice database and ORMDL from other species selected from NCBI Reference Sequence database. The tree was generated by neighbour-joining (NJ) algorithm with p-distance method and pairwise deletion of gap, employing MEGA version 5 [35]. A bootstrap statistic analysis was performed with 1000 replicates to test the phylogeny.

Analysis of sphingolipid long-chain bases
LCBs were liberated from 10mg of lyophilized leaves or panicle by alkaline hydrolysis, extraction method [36]. Briefly, this is preformed using 10% BaOH and dioxane 1:1 v/v. The tubes are then capped and placed in a dry block at 110uC for 20 hours. TheLCBs are extracted in chloroform/dioxane/water (8/3/8, v/ v/v). The LCBsare converted to dinitrophenyl derivatives with 0.2 ml of 0.5% (v/v) methanolic 1-fluoro-2, 4-dinitrobenzene and 0.8 ml of 2 M boric acid/KOH, at 60uC for 30 mins. LCBs are then extracted by phase partitioning with CHCl3/methanol/ H2O, 2:1:1 (v/v/v). The organic phase is removed and washed with an equal volume of 0.1 M KOH and 0.5 M KCl. The organic phase is then blown down and resuspended with methanol, and analyzed by reversed-phase HPLC. Separation was performed using a C18 RP 25064-mm column with a flow rate of 1 ml/min and a concave gradient from 80 to 100% methanol/acetonitrile/2-propanol, 10:3:1 (v/v/v), against water in 45 min. The elution was monitored with an Agilent 1200 DAD measuring at 350 nm and by ESI-MS/MS MRM on an ABSciex 4000 QTRAP. Three replicate extractions and LCB analysis were performed of each extraction method. Data presented as average (+/2 standard error) in ng LCB per mg of dry weight.

Analysis of ceramide
Samples extracted were ground in liquid N2 to a fine powder. 10 mg of each sample was used for each replicate analysis. Analysis of the ceramide fraction was performed using LC-MS/ MS as previously described [37]. Briefly, the ceramides were extracted in the lower phase of isopropanol/hexane/water (55:20:25 v/v/v) at 60 degrees C. Samples were dried down under a stream of nitrogen. The crude extract was de-esterified in 2 ml of 33% methylamine solution in ethanol/water (7:3 v/v) at 50 degrees for 1 hour. After hydrolysis the samples were dried down under a stream of nitrogen. The ceramides were resuspended in 1 mL of tetrahydrofuran (THF)/methanol/water (2:1:2 v/v/v) and 0.1% formic acid and 50 mL was analyzed on a 4000 QTRAP LC-MS/MS system (ABSciex) after HPLC using an Agilent 1200 fitted with a 100-mL sample loop. Separation was achieved on a SUPELCOSIL ABZ+Plus column 15063 mm, 5mm particle size, The sample was eluted at 1 mL min21 with a binary gradient system consisting of solvent A, THF/methanol/ 5 mM ammonium formate (3:2:5v/v/v), and 0.1% formic acid, and solvent B, THF/methanol/5 mM ammonium formate (7:2:1v/v/v), and 0.1% formic acid. held at 40uC. The sample was eluted at 1 mL min 21 with a binary gradient system consisting of solvent A, THF/methanol/5 mM ammonium formate (3:2:5v/ v/v), and 0.1% formic acid, and solvent B, THF/methanol/5 mM ammonium formate (7:2:1v/v/v), and 0.1% formic acid. The gradient started at 40% B rising to 75% B over 10 min. At the end of the gradient, the %B was increased to 100% over 1 min and held at 100% B for an additional 1 min to ensure complete elution of all compounds from the column. The probe was vertically positioned 11 mm from the orifice and charged with 5000 V. The temperature was held at 650uC, GS1 was set at 90 p.s.i., GS2 at 50 p.s.i., curtain gas at 20 p.s.i., and the interface heater was engaged. A minimum of 3 replicate extractions and analysis were performed. Data presented as average (+/2 standard deviation). A t-test was performed to determine statistical significance, and in all cases data presented has a p-value of 0.05. Figure S1 Sequence alignment of the three transcripts of Os07g26940. Sequences of rice ORMDL were obtained from GRAMENE rice database. (DOC) Figure S2 Sequence alignment of the two transcripts of Os04g47970. The sequences were obtained from GRAMENE rice database. (DOC) Figure S3 Sequence alignment of the representative three transcripts of ORMDL genes. Highlight in yellow indicates sequences specific for Os07g26940. Highlight in gray indicates sequences used to make RNAi to knockdown all ORMDL genes. Green and red indicates primer sequences used for Os07g26940R-NAi and ORMDL RNAi, respectively. The sequences were obtained from GRAMENE rice database. (DOC)