Endoplasmic Reticulum– and Golgi-Localized Phospholipase A 2 Plays Critical Roles in Arabidopsis Pollen Development and Germination

The phospholipase A 2 (PLA 2 ) superfamily of lipolytic enzymes is involved in a number of essential biological processes, such as inﬂammation, development, host defense, and signal transduction. Despite the proven involvement of plant PLA 2 s in many biological functions, including senescence, wounding, elicitor and stress responses, and pathogen defense, relatively little is known about plant PLA 2 s, and their genes essentially remain uncharacterized. We characterized three of four Arabidopsis thaliana PLA 2 paralogs (PLA 2 - b , - g , and - d ) and found that they (1) are expressed during pollen development, (2) localize to the endoplasmic reticulum and/or Golgi, and (3) play critical roles in pollen development and germination and tube growth. The suppression of PLA 2 using the RNA interference approach resulted in pollen lethality. The inhibition of pollen germination by pharmacological PLA 2 inhibitors was rescued by a lipid signal molecule, lysophosphatidyl ethanolamine. Based on these results, we propose that plant reproduction, in particular, male gametophyte development, requires the activities of the lipid-modifying PLA 2 s that are conserved in other organisms. Finally, each ampliﬁed PLA 2 - g and PLA 2 - d ORF was introduced into the Bam HI site of this construct. For construction of the Pro35S:PLA 2 :YFP fusionproteinconstruct,YFPwasinsertedintothe Sma I/ Bam HI-digested pFAST vector (Clontech), and the full-length PLA 2 cDNA was subse- quently cloned into the Kpn I-digested pFAST-YFP vector. For PLA 2 :YFP expression in Arabidopsis pollen, PLA 2 promoters (- g and - d ) were substituted for the cauliﬂower mosaic virus 35S promoter in the Pro35S:PLA 2 :YFP vector (designated ProPLA 2 - g :PLA 2 - g :YFP and Pro- PLA 2 - d :PLA 2 - d :YFP , respectively). PCR reactions were performed using Phusion High-Fidelity DNA polymerase (Finnzymes), and all of the constructs above were veriﬁed by DNA sequencing. The primers used for the cloning are described in Supplemental Table 1 online. enterokinase. The reaction mixture for PLA 2 activity contained 0.5 mM Ca 2+ and 0.05% Triton X-100 in a ﬁnal volume of 300 m L (50 mM Tris-HCl; pH 6.5 for PLA 2 - b and pH 8.0 for both PLA 2 - g andPLA 2 - d ).Beforesubstrateswereaddedtotheenzymereaction, and 10 m M manoalide dissolved in ethanol was preincubated with PLA 2 at 37 8 C for 60 min. Unlabeled PE (0.5 m mol) and 35 pmol radiolabeled L -3-phosphatidyl-[ 14 C]-ethanolamine-1,2-dioleoyl (55 mCi/mmol; GE Health-care)wereusedassubstratesforeachreaction.Theenzymereactionwere performed at 37 8 for 30 min. Lipid were separated by thin layer chromatography and developed with chloroform/ methanol/acetic acid/water (85:15:12.5:3.5, v/v/v/v). 14 C-PE and 14 C-lysoPE were detected on BAS-MS Imaging plates (Fujiﬁlm) and quantiﬁed using the Bio-Imaging Analyzer (FLA7000; Fujiﬁlm).


INTRODUCTION
Phospholipase A 2 (PLA 2 ) hydrolyzes the phospholipid molecule at the sn-2 position to produce lysophospholipid and a free fatty acid, both of which are precursors for second messengers of signal transduction pathways and also function as signaling molecules per se. The PLA 2 superfamily is generally categorized into five principal families of lipolytic enzymes based on their functional, structural, and catalytic properties, namely, the secreted PLA 2 s (sPLA 2 ), the cytosolic PLA 2 s (cPLA 2 ), the Ca 2+ -independent PLA 2 s, the platelet-activating factor acetylhydrolases, and the lysosomal PLA 2 s. To date, only two of these PLA 2 families have been reported in plants: the low molecular weight secreted PLA 2 s and the patatin-like PLAs (similar to the Ca 2+ -independent PLA 2 s but showing both PLA 1 and PLA 2 activity) (Schaloske and Dennis, 2006). The plant PLA 2 s are involved in such important physiological processes as senescence, wounding, stress responses, pathogen defense, and the induction of secondary metabolite accumulation (Wang, 2001(Wang, , 2004Ryu, 2004;Lee et al., 2005;Mansfeld, 2009;Scherer, 2010). Most of the physiological data reported to date have been obtained from studies on the effects of either PLA 2 inhibitors or PLA 2 products (Scherer and Arnold, 1997;Paul et al., 1998;Suh et al., 1998) or are limited to nonspecific lipid acyl hydrolases, such as the patatin-related PLA (Holk et al., 2002;Viehweger et al., 2002;Rietz et al., 2004;Scherer, 2010). The involvement of PLA 2 in phospholipid signaling in plant microtubule organization was reported by Gardiner et al. (2008), who observed that root tips treated with PLA 2 inhibitors show anisotropic growth and disorganization of the microtubule arrays. Arabidopsis thaliana PLA 2 -b has also been shown to function in light-induced stomatal opening as well as shoot gravitropism and cell elongation Seo et al., 2008). More recently, Lee et al. (2010) demonstrated that PLA 2 -a modulates PIN-FORMED protein trafficking to the plasma membrane (PM) in the Arabidopsis root, thereby revealing that PLA 2 also plays a role in intracellular membrane trafficking in plants.
A cell has many curved membrane systems, such as the endoplasmic reticulum (ER), Golgi, endosome, and multivesicular bodies. Membrane deformation, as a process of vesicle trafficking, budding, and fusion, is regulated by the interplay between lipids and proteins, which act like wedges in the membrane. Changes in lipid composition, including modification of the membrane via, for example, phospholipid acylation and/or deacylation, is also a potential mechanism of membrane deformation . Brown et al. (2003) proposed that PLA 2 hydrolyzes the phospholipids on one side of a membrane, thereby creating a localized concentration of inverted cone-shaped lysophospholipids that in turn drives the formation of positive membrane curvature, a process considered to be the first step in membrane tubule formation. This proposal was based on results showing that PLA 2 regulates the Golgi complex and membrane tubulation of the trans-Golgi network (de Figueiredo et al., 1998) and modulates membrane-protein trafficking (Choukroun et al., 2000). Staneva et al. (2004) reported a direct role of PLA 2 in the vesiculation processes associated with the intermembrane transport through budding and the fission of giant liposomes. More recently, Gubern et al. (2008) reported that the biogenesis of lipid droplets, which function as storage organelles for energy generation and membrane regeneration, is inhibited by treatment with cPLA 2 inhibitors or by cPLA 2 suppression with small interfering RNA (siRNA). This steadily accumulating body of data suggests the existence of a relationship between PLA 2 and the membrane network. However, most of the information currently available on the function of PLA 2 is based on studies of mammalian cells, and little is known on its role in the plant membrane network.
The intracellular membrane network in the developing pollen grain is characterized by the extensive proliferation of the ER and surface-linked membrane vesicles that serve as lipid precursors to increase the surface area of the PM during pollen tube germination (Piffanelli et al., 1998). Pollen development is also typified by the diverse transformation of vacuoles and the biogenesis and movement of ER and Golgi bodies. Observations of ultrastructural changes in vacuoles, ER, and Golgi suggest that these organelles are linked to the accumulation of metabolites necessary for pollen development, pollen germination, and tube growth (Hesse, 1990;Bedinger, 1992;McCormick, 1993;Yamamoto et al., 2003). H + -ATPases have been reported to be activated by the mobilization of the H + pool and to play key roles in Golgi organization during pollen development (Dettmer et al., 2005). They have also been found to be regulated by lysophosphatidyl choline (LPC), which is generated by PLA 2 , in suspension-cultured cells of California poppy (Eschscholzia californica) (Viehweger et al., 2002). It has been reported that phosphatidylinositol 4,5-bisphosphate, inositol 1,4,5-triphosphate, and phosphatidic acid (PA), as second messengers produced from membrane phospholipids by phospholipase C (PLC) or phospholipase D (PLD), promote pollen tube growth (Malhó , 1998;Monteiro et al., 2005;Dowd et al., 2006). However, the possible roles of the structurally different PLA 2 s with PLC and PLD in pollen tube growth remain to be elucidated.
The aim of this study was to establish the roles of members of the PLA 2 gene family in Arabidopsis. We present evidence that three PLA 2 paralogs (b, g, and d) reside in the ER and/or Golgi, albeit their structural similarity with animal sPLA 2 , and that the suppression of PLA 2 , as evidenced by RNA interference (RNAi), induces pollen degeneration. We also provide biochemical evidence that lysophosphatidyl ethanolamine (LPE), a PLA 2 product, plays a key role in pollen germination and pollen tube growth.

RESULTS
Expression of PLA 2 -b, -g, and -d Increases with Pollen Maturation Using RT-PCR to compare the expression profiles of all members of the Arabidopsis PLA 2 gene family in detail, we detected PLA 2 -a and PLA 2 -b transcripts in all of the sporophytic tissues tested. Unlike PLA 2 -a, PLA 2 -b was strongly expressed in pollen. The PLA 2 -d transcript was detected exclusively in pollen, while the PLA 2 -g transcript was expressed in floral tissues, such as the bud, open flower, and pollen ( Figure 1A). However, because the bud and open flower samples contained pollen grains inside their anther sac or on the surface of the stigma, we considered the PCR amplifications of PLA 2 -g to originate from pollen in the bud and open flower. Due to its lower expression level, the same number of PCR cycles used to detect PLA 2 -g did not produce enough amplified product of PLA 2 -d to be visible on electrophoresis. However, PLA 2 -d was also detected when more PCR cycles were performed. As shown in the promoter:b-glucuronidase (GUS) analysis ( Figures 1C and 1E), PLA 2 -g was expressed only in pollen grains, as was PLA 2 -d. In a subsequent RT-PCR analysis ( Figure 1B), we used cDNAs enriched for three different pollen developmental stages (microspore, bicellular, and tricellular stages) to determine the precise pollen stages at which PLA 2 genes are expressed. PLA 2 -a expression was not detected in pollen at any of these developmental stages, while PLA 2 -b was expressed continuously during all stages. PLA 2 -g was expressed slightly at the microspore and bicellular stages but strongly at the tricellular stage and in mature pollen. PLA 2 -d expression was found to be initiated in the tricellular stage ( Figure  1B). These results indicate that the expression of PLA 2 -d was relatively low during the early stages of pollen development and subsequently increased considerably at the tricellular stage. There was no difference in the expression pattern of the PLA 2 s in dehydrated and hydrated pollen grains (see Supplemental Figure  1 online). Lee et al. (2003) reported that the PLA 2 -b promoter drives widespread GUS expression in all plant tissues. We found that while the PLA 2 -g and -d promoters drove GUS expression exclusively in the pollen ( Figures 1C and 1E), the GUS expression patterns of all three promoters during flower development were similar in that expression was initially weak at the early developmental stages and gradually increased as the floral organs matured. Staining with 4',6-diamidino-2-phenylindole confirmed that the expression of PLA 2 -b, -g, and -d was low at the early stages of pollen development and increased as the pollen matured ( Figure 1G). GUS signals remained strong in the germinated pollen and pollen tube ( Figures 1D and 1F).

Suppression of PLA 2 s in Pollen
If PLA 2 s are essential for pollen development and pollen germination and tube growth, normal pollen development should be impeded by any alteration in their expression. To test this hypothesis, we searched for T-DNA-tagged knockout (KO) mutants and found SALK_022347.48.40 (designated as pla 2 -g-1) and SALK_033172.37.00 (designated as pla 2 -g-2) for PLA 2 -g but none for PLA 2 -b and -d. The pla 2 -g-1 and the pla 2 -g-2 mutants contain a T-DNA in the second exon and third intron of the PLA 2 -g, respectively. RT-PCR using PLA 2 -g-specific primers confirmed that the pla 2 -g-1 and pla 2 -g-2 homozygous lines are null mutants (see Supplemental Figure 2 online) and that the other PLA 2 isoforms are expressed normally. In addition, the pollen grains of these KO plants did not show any phenotypic differences compared with the wild type. Given their high degrees of similarity at the amino acid level and their strong expression in pollen, we presume that PLA 2 -b, -g, and -d have a functional redundancy during pollen development, even though they individually may play different roles at several pollen developmental stages.
To suppress the expression of all PLA 2 genes in pollen simultaneously, we used the RNAi approach with two different pollen-specific promoters, Lat52 (Twell et al., 1989) and PLA 2 -d. A 576-bp full open reading frame (ORF) from ATG to TAG of PLA 2 -d was used as a trigger of RNAi (E-values of estimating similarities: 1e 2120 , 5e 219 , and 4e 204 with PLA 2 -g, PLA 2 -b, and  (Smyth et al., 1990). The arrowhead indicates the flower bud at the initiation stage of anther yellowing. Bars = 1 mm. PLA 2 -a, respectively) and cloned into the pHANNIBAL vector in both the sense and antisense orientations under the control of the Lat52 promoter (designated as LPRNAi) and the PLA 2 -d promoter (designated as DPRNAi), and then subcloned into the pART27 vector ( Figure 2A). These constructs were transformed into wild-type and pla 2 -g-1 homozygous plants. The normal vegetative tissues and defective male gametophyte of LPRNAi-or DPRNAi-introduced wild-type plants were not phenotypically different from those of pla 2 -g-1 containing LPRNAi or DPRNAi constructs. Consequently, the DPRNAi lines were used in subsequent analyses. Of the 60 DPRNAi lines generated, six lines showing a severely sterile phenotype with highly reduced numbers of seeds in shorter siliques compared with the wild type were chosen for study (Figures 2B and 2C). The anthers of the RNAi plants dehisced normally, but most siliques failed to set any seeds, although partially filled siliques with <10 seeds were occasionally produced ( Figures 2D and 2E). To confirm whether this sterility was caused by the suppression of PLA 2 s in pollen by the RNAi approach, we performed RNA gel blot analysis to detect siRNA using total RNAs isolated from flowers. siRNA was detected in both DPRNAi and pla 2 -g/DPRNAi transgenic plants but not in wild-type plants ( Figure 2F). The real-time PCR analysis using RNAs extracted from pollen also showed that the transcript levels of the three PLA 2 genes, PLA 2 -b, -g, and -d, were reduced ( Figure 2G).
According to Xing and Zachgo (2007), however, at least 10% of RNAi transgenic Arabidopsis plants, regardless of the construct types, produce 20 to ;50% nonviable pollen in an anther as a side effect. We obtained 60 T1 lines from LPRNAi and DPRNAi transgenic plants that survived on kanamycin (kanamycin resistance conferred by the RNAi vector), of which six (10% of transgenic population) showed severe sterility, as shown in Figure 2. Around 80% of the seed sterility in these six RNAi mutants possibly originated from the direct effect of RNAi (50%) combined with the side effects of RNAi on pollen (20 to ;50%).
Approximately 50% of 54 transgenic lines were fertile but with half of their pollen grains collapsed. To verify that the observed pollen disorder was caused by RNAi, we performed RNAi silencing of PLA 2 s in the quartet mutant, in which pollen grains remain in permanent tetrads (Preuss et al., 1994), and also observed that 50% of the pollen grains were aborted (Figures 3A and 3B). We produced and analyzed 10 T1 RNAi transgenic lines in the quartet background. Of these, two produced normal pollen and eight produced defective pollen. Among the eight lines with the defective pollen phenotype, 50% of the pollen was aborted in four lines, 75% in two lines, and <30% in two lines. Although the RNAi transgenic lines in the quartet background were not analyzed as extensively as the RNAi transgenic lines in the wild-type (or pla 2 -g) background, the defective pollen ratio was consistent, demonstrating that the pollen was aborted due to the suppression of PLA 2 s in the pollen grain. To ascertain whether or not pollen abortion causes a reduction of PLA 2 expression in the RNAi lines, we performed quantitative RT-PCR in the RNAi lines using primers of other pollen-specific genes, such as VAN-GUARD1 (VGD1) (Jiang et al., 2005) and sugar transporter (STP11) (Schneidereit et al., 2005). As shown in Supplemental Figure 3 online, these two genes were not downregulated in the RNAi lines. Taken together, these results demonstrate that RNAi suppression of all PLA 2 members in the pollen was successful.

Abnormal Pollen Development Is Caused by Disordered Cellular Membranes
As pollen-specific promoters (Lat52 and PLA 2 -d) were used for RNAi construction, the siRNAs are likely to be produced during pollen development; consequently, vegetative growth would be expected to be normal. This expectation is supported by the normal morphologies found in the RNAi plants. To determine the effect of RNAi suppression of all Arabidopsis PLA 2 genes on pollen, we first compared the mature pollen grains of the wildtype and RNAi lines using a scanning electron microscope.
Compared with the normal pollen of the wild type ( Figure 3C), the pollen of the DPRNAi 4-4 line with a 66% pollen abortion rate was partially shrunken ( Figure 3D, arrow) or completely collapsed ( Figure 3D Sanders et al. (1999). Bars = 10 mm.
to 3J) revealed that pollen grains of the RNAi lines at the early microspore stage (i.e., anther stage 8; Sanders et al., 1999) were not phenotypically different ( Figure 3H) from those of the wild type ( Figure 3E). The first signs of abnormal pollen development in the DPRNAi 1-1 line, which had a pollen abortion rate of 67%, such as the detachment of the PM from the cell wall and shrinking cytoplasm, appeared at the vacuolated late uninucleate microspore stage (i.e., anther stage 9; Sanders et al., 1999;Figures 3F and 3I). By the mature pollen stage (i.e., anther stage 12; Sanders et al., 1999), RNAi pollen grains had collapsed completely and were devoid of cytoplasmic content ( Figures 3G and 3J), although the exine layer and all other structures, both inside and outside the anthers, were not different from those in wildtype pollen. As depicted in Table 1, aborted pollen ratios were very similar at the bicellular and tricellular pollen stages (52 to ;72%).
The ultrastructure of the pollen grains was observed by transmission electron microscopy (TEM) to determine the underlying factors causing pollen degeneration. In accordance with the observations of Yamamoto et al. (2003), we observed that the wild-type microspores contained a round nucleus, a distinct nucleolus and nuclear membrane, and ribosome binding rough ER ( Figures 4A, 4C, and 4E). By contrast, the nucleus of the pollen grain of DPRNAi 4-4 line was irregular in shape with an indistinct membrane, and the ribosomes were scattered around the nucleus membrane ( Figures 4B and 4D, arrowheads). Moreover, the RNAi pollen grains failed to form a defined endomembrane system; this was particularly evident for the ER, which had an irregular shape and a fragmented membrane (Figures 4D,arrows). The ribosomes in RNAi pollen were observed to be scattered throughout the shrinking cytoplasm, possibly due to the lack of a defined ER structure and ER fragmentation. While typical Golgi stacks, consisting of five to eight flat cisternae, were observed in the wild-type pollen ( Figure 4E), irregular Golgi stacks with a decreased number of cisternae were observed in the RNAi pollen ( Figure 4F). Also in the RNAi pollen, a number of small vesicles were observed in the region surrounding Golgi stacks ( Figure 4F, asterisks), which suggested that ONO (2-[p-amylcinnamoyl]amino-4-chlorobenzoic acid), a PLA 2 inhibitor, had induced some degree of disintegration of the Golgi stacks and the formation of numerous vesicles in the trans-Golgi face . Finally, the PM in the RNAi pollen was detached from the cell wall, and the electron density of the contents of the shrinking cytoplasm was higher than that of the wild type ( Figures 4A and 4B).
PLA 2 -g Localizes to the ER and Golgi, and PLA 2 -d to the ER Our results to this point suggest the possibility that the pollenexpressing PLA 2 -b, -g, and -d play a critical role in pollen development. Seo et al. (2008) reported that PLA 2 -b localizes to the ER in guard cells and is involved in stomatal opening. Based on our results, as shown in Figure 4, it is possible to conjecture that PLA 2 -g and -d also localize to the ER and/or Golgi and act on the function of membrane deformation and trafficking of these organelles. However, earlier experiments involving the transient expression of green fluorescent protein (GFP)-tagged Arabidopsis PLA 2 proteins in onion epidermal cells demonstrated that PLA 2 -b and -g are secreted to the cell wall/ DPRNAi (1-1) 72% (n = 110) 67% (n = 130) DPRNAi (4-4) 69% (n = 105) 66% (n = 123) pla 2 -g-1/DPRNAi (2-2) 54% (n = 118) 58% (n = 238) pla 2 -g-1/DPRNAi (4-4) 57% (n = 95) 52% (n = 98) Completely collapsed pollen grains were counted. n indicates total counted pollen grains. extracellular space Lee et al., 2003). To resolve this inconsistency, we fused PLA 2 -g and -d with yellow fluorescent protein (YFP) at the C-terminal region under the control of the 35S promoter and introduced the fusion proteins into tobacco (Nicotiana tabacum) leaf epidermal cells by Agrobacterium tumefaciens infiltration. Transient expression of the fluorescence of each fusion protein was observed by confocal laser scanning microscopy. STtmd:cyan fluorescent protein (CFP) (a Golgi marker; Figures 6D). Seventy-two percent of the STtmd:CFP signal overlapped with that of PLA 2 -g:YFP, and 16% of the PLA 2 -g:YFP signal overlapped with that of STtmd:CFP in the full image, but the signals were found to significantly overlap each other (100 and 86%, respectively) in the region containing small punctate signals (region 1 in Figure 5A). This result indicates that PLA 2 -g localized to the trans-Golgi. ERD2:GFP and PLA 2 -g:YFP signals overlapped with each other (92 and 75%, respectively) in full image and overlapped (92 and 90%, respectively) in the region containing small punctate signals and networks (region 2 in Figure 5B), demonstrating that PLA 2 -g localized to cis-Golgi and ER. GFPHDEL and PLA 2 -g:YFP signals overlapped with each other (87 and 66%, respectively) in the full image, but their degree of overlap (1.2 and 2.3%, respectively) was less in the region containing small punctate signals (region 3 in Figure 5C). This result can be considered to be reliable as GFPHDEL is only expressed in the ER. In comparison, the STtmd:CFP and PLA 2 -d: YFP signals overlapped with each other to a lower degree (50 and 12%, respectively), thereby confirming that PLA 2 -d does not localize to trans-Golgi. The relatively high degree of overlap of the STtmd:CFP signal with that of PLA 2 -d:YFP, even though PLA 2 -d localizes to the ER, leads to the conclusion that STtmd:CFP is weakly fluorescent in the ER, as reported by Boevink et al. (1998). The ERD2:GFP and PLA 2 -d:YFP signals overlapped with each other (36 and 73%, respectively), and the GFPHDEL and PLA 2 -d: YFP signals showed a high degree of overlap (78 and 71%, respectively).
To further confirm the localization of PLA 2 -g and -d in pollen, we fused PLA 2 -g and -d with a soluble-modified red-shifted GFP (smRS-GFP) (Davis and Vierstra, 1998) under the control of Lat52 and with YFP under the control of the PLA 2 promoter and then transformed these four constructs into tobacco and Arabidopsis. The GFP signal of ProLat52:smRS-GFP (control) was detected evenly throughout the cytosol and nuclei in the tobacco pollen grain and tube ( Figure 7A). In comparison, the GFP signal of the ProLat52:PLA 2 -g:smRS-GFP tobacco pollen and pollen tube was observed in a spindle-shape membrane boundary structure, which is considered to be the ER fragment (Cheung et al., 2002), (D) Quantitative analysis of the distribution of PLA 2 -g relative to Golgi and/or ER markers. The percentage of signal intensity present in the shared pixels (middle; white) relative to the total signal is measured for both the gray (top; PLA 2 -g:YFP) and black (bottom; STtmd:CFP, ERD2: GFP, and GFPHDEL) channels. Gray and black values are denoted as the difference between 100% and the overlap measured for the corresponding channel. The extent of colocalization signal intensity is quantified in the full image as well as in region 1;3 marked in (A) to (C), respectively. and in small punctate structures, namely, the Golgi ( Figure 7B, arrows). The GFP signal of ProLat52:PLA 2 -d:smRS-GFP pollen was similar to that of ProLat52:PLA 2 -g:smRS-GFP pollen, but there was no fluorescence in the Golgi in the former ( Figure 7C). These results were identical when the PLA 2 -g and PLA 2 -d promoters were used in Arabidopsis ( Figures 7D and 7E), leading to the conclusion that PLA 2 -d localizes to the ER and PLA 2 -g to both the Golgi and ER. These results, together with those of Seo et al. (2008), which show that PLA 2 -b localizes to the ER, suggest that PLA 2 -b, -g, and -d have specific roles in the ER and/or Golgi.

Pollen Germination Is Inhibited by Treatment with PLA 2 Inhibitors
The patterns of GUS expression driven by the promoters of PLA 2 s in the germinating pollen and pollen tube (Figures 1D and 1F) suggest that PLA 2 enzymes are active during pollen germination and tube growth. This led to the question of whether the observed incomplete pollen germination and/or tube growth could ultimately result in the sterility observed in the RNAi lines. However, direct testing of the involvement of PLA 2 enzymes in pollen germination and tube growth in the RNAi pollen was not possible due to (1) RNAi suppression of PLA 2 s being initiated in the unicellular stage, and (2) complete collapse of the pollen grains before the mature pollen stage. Adopting an alternative approach, we treated wild-type Arabidopsis and tobacco pollen grains with seven different PLA 2 inhibitors, manoalide, 4-bromophenacyl bromide (BPB), aristolochic acid (AA), trifluoromethyl ketone (AACOCF3), palmitoyl trifluoromethyl ketone (PACOCF3), bromoenol lactone (BEL), and ONO, and observed their effects on pollen germination. Because various conditions, such as pH, temperature, pollen density, plant age, and flower stage, can greatly affect in vitro pollen germination in Arabidopsis, the bicellular-type tobacco pollen, which is known to be more consistent in terms of in vitro germination (Brewbaker, 1967;Boavida and McCormick, 2007), was used in an initial test to select the most effective PLA 2 inhibitors. Manoalide was found to be the most effective inhibitor, with AACOCF3, PACOCF3, and AA having no inhibitory effect at all on tobacco pollen germination. BEL, ONO, and BPB also had no inhibitory effect on pollen germination at concentrations of <10 mM and only a slight inhibitory effect at higher concentrations (see Supplemental Figure 4 online). We treated the recombinant PLA 2 -b, -g, and -d proteins with manoalide to test for an in vitro enzyme inhibition effect. Manoalide effectively inhibited PLA 2 -g and -d activities and slightly inhibited PLA 2 -b activity ( Figure 8A). AA has been reported to have an inhibitory effect at the biochemical level on recombinant PLA 2 -b activity as well as a physiological effect on stomatal opening and root growth (Gardiner et al., 2008;Seo et al., 2008); however, it did not have an inhibitory effect on pollen germination in our experiment. Thus, we focused on analyzing the effect of AA on in vitro PLA 2 -a, -g, and -d enzyme activities. AA inhibited PLA 2 -a but not -g and -d (see Supplemental Figure 5 online), which supports our hypothesis that PLA 2 -g and -d have specific functions in pollen and important roles in pollen germination. The fact that several inhibitors have different effects on the different PLA 2 isoforms also suggests that these PLA 2 s have different mechanisms at the cellular level. Manoalide was chosen as the inhibitor in subsequent experiments. Pollen germination was inhibited by manoalide at a concentration of as low as 3 mM (19% germination), with more severe inhibition occurring at increasingly higher concentrations (i.e., in a dosage-dependent manner; 8.5% germination at 5 mM manoalide); 10 mM of manoalide completely prevented pollen germination ( Figures  8B to 8F).
To ascertain whether this inhibitory effect of manoalide was due to the absence of PLA 2 -hydrolized products, we supple-mented the pollen germination medium containing 10 mM manoalide with one of three PLA 2 products, namely, lysophosphatidic acid (18:1-LPA), 18:1-LPC, or 18:1-LPE, as well as with one PLD product, phosphatidic acid (18:1-PA), which is known to be an important signaling molecule of pollen germination (Potocký et al., 2003) (Figure 9). Interestingly, only 18:1-LPE effectively rescued the inhibition of pollen germination by manoalide  ( Figure 9D); the same results were also found in the in vitro Arabidopsis pollen germination test ( Figure 9I). There were no differences between pollen cultured under normal conditions (no inhibitors) and those cultured on medium containing 18:1-LPE in terms of pollen germination and tube growth. The average length of the pollen tube after a 3 h incubation under normal conditions and in the LPE treatment was 488.3 6 97.8 and 489.1 6 86.9 mm, respectively; the germination rate was 74.8 and 75%, respectively. All these data suggest that germination inhibition by manoalide is not due to a nonspecific inhibition of other lipases in pollen because (1) the other inhibitors did not prevent pollen germination, (2) AA did not inhibit PLA 2 -g and -d activities in vitro, and (3) the inhibition by manoalide was recovered when the culture medium was supplemented with the PLA 2 product LPE.
The PLA 2 Inhibitor Manoalide Interferes with VHA-c4 Localization It is known that V-ATPases are the key regulators of membrane trafficking as well as energizers of secondary active transport. Based on the results of experiments involving concanamycin A, a specific V-ATPase inhibitor, Dettmer et al. (2006) reported that V-ATPase in the trans-Golgi network is essential for endocytic and secretory trafficking. Concanamycin A affects Golgi morphology, such as the bending of Golgi cisternae, swelling of Golgi ends, fragmentation of Golgi stacks, and accumulation of large vesicles. Furthermore, plant V-ATPase has a critical function in the development of the male gametophyte (Dettmer et al., 2005). In a yeast two-hybrid experiment in which we screened for interactors of PLA 2 -g and -d, we found that PLA 2 -g and -d interacted with VHA-c4, a subunit of the V-ATPase complex (see Supplemental Figure 6 online). Because PLA 2 products are known to activate V-ATPases (Palmgren and Sommarin, 1989;Pedchenko et al., 1990;Viehweger et al., 2002), the binding of these two proteins can provide a clue to the function of both PLA 2 and VHA-c4.
To obtain more direct and reliable evidence for the link between PLA 2 and VHA-c4, we designed an experiment to monitor changes in the localization of VHA-c4 in pollen caused by the PLA 2 inhibitor. We generated transgenic Arabidopsis plants that expressed VHA-c4:smRS-GFP under the control of the Lat52 promoter. The shape of VHA-c4:smRS-GFP in the apex of pollen tubes revealed a typical ER morphology, as described previously (Lovy-Wheeler et al., 2007), suggesting that VHA-c4 localizes to the ER. To obtain evidence supporting the localization of VHA-c4 to the ER, we performed a transient assay using VHA-c4:YFP with the ER marker in tobacco epidermal cells and generated the Pro35S:VHA-c4:YFP transgenic Arabidopsis lines. We found that VHA-c4:YFP colocalized with ERD2:GFP to the ER network of the tobacco epidermal cells but not to the Golgi vesicles (see Supplemental Figure 7A online). The VHA-c4:YFP signal localized to the ER networks, ER body membrane, and small vesicles (see Supplemental Figure 7C online). The morphology of the organelles was similar to those in the hypocotyl cells of the GFPHDEL-expressed line (see Supplemental Figure 7B online). Based on the interaction of VHA-c4 with PLA 2 -d in our yeast two-hybrid assay, we concluded that VHA-c4 colocalized with PLA 2 -d in Arabidopsis. Transgenic lines coexpressing VHA-c4:CFP and PLA 2 -d:YFP showed that these two signals colocalized to the ER network and ER bodies (see Supplemental Figure 7D online). The VHA-c4:XFP signal was particularly strong at the ER body membrane. Taken together, these results strongly suggest that VHA-c4 mainly localizes to the ER, as do other subunits, such as VHA-c99 and VHA-e2 (Seidel et al., 2008). Pollen grains obtained from ProLat52:VHA-c4:smRS-GFP transgenic Arabidopsis were germinated for 1 h, and the pollen tubes were subsequently stained with the ER tracker as an ER marker or FM4-64 as an endocytic tracer. In the pollen tube that was not subjected to manoalide treatment, the VHA-c4:smRS-GFP fusion protein colocalized with the ER tracker but not with FM4-64 ( Figure 10A; see Supplemental Figures 8A and 8B online). Even although VHA-c4:smRSGFP did not colocalize with FM4-64, we did not rule out the possibility that VHA-c4:smRSGFP can localize to endosomal structures. We believed that this phenomenon is very similar to the previous report on ARA6, which is broadly used as a maker for endosomes. When pollen tubes were incubated with BFA, ARA6 was not correlated with the aggregation of FM4-64 because of the heterogeneity of the endosomal populations (Zhang et al., 2010). In the manoalidetreated pollen tube, pollen tube growth and reverse fountain streaming in apex were blocked. After 10-min treatment with manoalide, VHA-c4:smRS-GFP signals were aggregated throughout the pollen tube, as were FM4-64 signals, but the latter did not colocalize with the aggregated VHA-c4:smRS-GFP signals ( Figure 10B). One hour after treatment with manoalide, VHA-c4: smRS-GFP signals and FM4-64 signals were significantly aggregated in the cytosol, and the structural changes were visible using normal light microscopy. FM4-64 signals in the PM faded 1 h after treatment with manoalide ( Figure 10C), similarly to the previous reports (Liao et al., 2010;Zhang et al., 2010). This phenomenon may be explained by two possibilities. One is that the synthesis of the PM is inhibited in the growing pollen tube, due to blocking of both endosome-to-PM recycling and ER-Golgi-vesicle membrane trafficking by the PLA 2 inhibitor. This possibility is supported by a previous report that PLA 2 -a is required for PIN protein trafficking of the PM in the Arabidopsis root . Another possibility is that the blocking of membrane trafficking may provoke cell degeneration and eventually induce the breakdown of PM. This is supported by our data that show that in a TEM image of RNAi at the bicellular stage, the bilayer structure was indistinguishable and appeared to be disassembled (see Supplemental Figure 9 online) and that pollen grains of the DPRNAi 1-1 line were completely collapsed and devoid of cytoplasmic content at the tricellular stage ( Figure 3J).
To confirm the specificity of the PLA 2 inhibitor manoalide on VHA-c4 trafficking, we exposed growing pollen tubes to two well-characterized pharmacological inhibitors, U-73122 and 1-butanol. The PLC inhibitor U-73122 inhibits and depolarizes pollen tube growth (Helling et al., 2006), and the PLD inhibitor 1-butanol blocks both pollen tube growth and germination (Potocký et al., 2003). In our experiment, pollen tubes treated with 1-butanol stopped growing, but the VHA-c4:smRS-GFP signal was not different between pollen tubes treated or not with 1-butanol (see Supplemental Figure 10A online), indicating that VHA-c4 trafficking is not influenced by PLD. The appearance of VHA-c4:smRS-GFP in pollen tubes treated with U-73122 was quite different from that in pollen tubes treated with manoalide. The blocking of endocytic membrane recycling by U-73122, as reported by Helling et al. (2006), resulted in the FM4-64 signal remaining only in the PM and not accumulating in the apical region; in addition, the VHA-c4:smRS-GFP signal was spotted and scattered (see Supplemental Figure 10B online). Because PLC activity is required for membrane deformation, including the endocytic and exocytic pathway (Goñ i and Alonso, 2000;De Matteis and Godi, 2004;McMahon and Gallop, 2005;Helling et al., 2006), it is possible that U-73122 caused disordered VHA-c4 localization. This perturbation in the localization of VHA-c4 by the PLA 2 inhibitor possibly indicates that PLA 2 s have roles in ER membrane deformation or VHA-c4 trafficking, even although the link between VHA-c4 and PLC remains to be determined.

DISCUSSION
We demonstrated that Arabidopsis PLA 2 -b, -g, and -d are expressed differentially during pollen development. RNAi resulted in pollen lethality, and treatment with PLA 2 inhibitors inhibited pollen germination. We also demonstrated that PLA 2 paralogs localize to the ER and/or Golgi, where they play critical roles in pollen development, most likely by modulating membrane deformation and thereby enabling membrane trafficking. PLA 2 s also play critical roles in pollen germination and tube growth by providing LPE. Our results therefore enlarge the body of knowledge on the roles of PLA 2 s in plant development and reproduction.

PLA 2 s Function in Pollen Development by Regulating Intracellular Trafficking and Membrane Deformation Events
The promotion of membrane deformation by PLA 2 has been shown in giant liposomes, where PLA 2 treatment induced budding and fission events (Staneva et al., 2004). Recent studies in animal cells have shown that phospholipid remodeling enzymes have an important function in the organization of the Golgi complex. The functional Golgi organization is mediated by various proteins that interact with specific lipid components to form membrane curving for vesicle formation and tubule fission (Bard and Malhotra, 2006;De Matteis and Luini, 2008;Jackson, 2009). Phosphoinositides recruit proteins for regulating diverse trafficking events, and lysophospholipids, PA, and diacylglycerol may facilitate vesicle or tubule formation by changing the composition of the membrane (van Meer and Sprong, 2004;Gallop and McMahon, 2005;Frost et al., 2009). Diacylglycerol and PA, which are produced by PLC and PLD, respectively, have functional roles in Golgi membrane fission (Yang et al., 2008;Asp et al., 2009). Lysophospholipid, which is generated by PLA 2 , may be involved in Golgi retrograde trafficking and Golgi cisternal structure by modifying membrane tubule formation (de Figueiredo et al., 1998;Brown et al., 2003). LPA acyltransferase (LPAAT), which converts LPA into PA, directly regulates mammalian Golgi complex structure and function (Schmidt and Brown, 2009).
As in animal systems, the Golgi apparatus in plants is of major importance in the intracellular trafficking of proteins, which is affected by V-ATPase (Matsuoka et al., 1997). Membrane fusion and formation events, such as the invagination of vacuoles and the fusion of small vacuoles and the PM, as well as intracellular trafficking by the Golgi apparatus are active processes during successful pollen production (Yamamoto et al., 2003). Dettmer et al. (2005) reported that V-ATPase is essential for Golgi organization and that its KO causes complete pollen lethality; consequently, severe morphological changes of the Golgi stacks and Golgi vesicles in developing pollen would be the first indication of cell degeneration. However, it remains to be addressed how the lack of V-ATPase activity brings about these morphological changes.
Based on the results of our experiments on the localization of PLA 2 -g and -d in the ER and/or Golgi in tobacco epidermal cells (Figures 5 and 6) and the detachment of the PM from the cell wall (shrunken cytoplasm; Figures 3I and 4B), the lack of nuclear and ER membrane, ER fragmentation, irregular Golgi stacks, and the formation of vesicles near Golgi in PLA 2 -suppressed pollen grains ( Figures 4B, 4D, and 4F), we suggest that plant PLA 2 s have functions equivalent to those of their animal counterparts (de Figueiredo et al., 1998;Choukroun et al., 2000;Staneva et al., 2004) in membrane deformation and trafficking. This conjecture is supported by the fact that both the VHA-c4:smRS-GFP signal in the ER and the FM4-64 signal in the endosomes were aggregated by treatment with the PLA 2 inhibitor manoalide ( Figure 10). These physiological and pharmacological results could provide an important clue toward the elucidation of the function of PLA 2 s in membrane deformation for the following reasons. Membrane deformation is known to mainly occur at a very specific membrane region, called the lipid raft, by budding and fission formation. The lipid raft region is enriched in cholesterol, sphingolipids, and an abundance of membrane proteins. PLA 2 s are also localized to this raft region, where they trigger membrane curving by hydrolyzing the sn-2 bond of glycerophospholipids (Lipowsky, 2002;Staneva et al., 2004). This leads to the rather safe assumption that PLA 2 -g and -d interact with VHA-c4 in the lipid raft region, triggering the formation of curvature for membrane deformation and trafficking by hydrolyzing the phospholipid bond. An alternative explanation is that the aggregation of ER in pollen by manoalide disturbs VHA-c4 trafficking, inhibiting pollen tube growth by an improper membrane trafficking event. The former mechanism is supported by the following experimental results: (1) PLA 2 products are known to activate V-ATPases (Palmgren and Sommarin, 1989;Pedchenko et al., 1990;Viehweger et al., 2002), (2) subunits of VHA localize to the ER and trans-Golgi network (Dettmer et al., 2006;Seidel et al., 2008), and (3) VHA-c4:smRS-GFP colocalized with ER tracker but not with FM4-64, which is a well-known endocytic tracer, suggesting that the effect of VHA-c4 is more dominant in the ER than in the other organelles. However, we cannot exclude the possibility that VHA-c4 has an endosome recycling effect because the FM4-64 signal was aggregated following treatment with manoalide. Recent reports of PLA 2 -a functioning in root hair development by regulating endocytic recycling of PIN  and of V-ATPase activity being required for endosome recycling in the root (Dettmer et al., 2006) suggest that PLA 2 -b, -g, and -d may also have some function in the endocytic pathway with VHA-c4. Because PLA 2 -g colocalized with STtmd-CFP as a trans-Golgi marker, the possibility that PLA 2 functions in the Golgi as a modulator of the endocytic pathway remains to be determined.

PLA 2 -g and -d Play a Key Role in Pollen Germination by Producing LPE
Several studies have reported that auxin-activated PLA 2 increases the concentration of free fatty acids and lysophospholipids, which in turn can activate H + -ATPase to induce apoplastic acidification, leading to cell elongation (Scherer and Arnold, 1997;Paul et al., 1998;Scherer, 2002;Rietz et al., 2004;Ryu, 2004;Lee et al., 2005). Since pollen germination and tube growth are also a specialized type of cell elongation, it is possible that lysophospholipids are active in this process. We therefore examined whether the inhibition of pollen germination by a PLA 2 inhibitor could be rescued by different lysophospholipids, namely, 18:1-LPA, 18:1-LPC, and 18:1-LPE. As shown in Figure  9, LPE, which is the major hydrolytic product of PLA 2 -g and -d (Lee et al., 2005), only effectively rescued pollen germination specifically inhibited by manoalide. This result is consistent with the specific expression of PLA 2 -g and -d in pollen and also indicates that LPE in particular is a key molecule in pollen germination and tube growth. The most common fatty acids in the sn-1 and sn-2 positions of phosphatidylethanolamine (PE) in the leaf and pollen are 18:2-linoleic acid and 18:3-linolenic acid, and only rarely is 18:1-oleic acid in these positions (Browse et al., 1986;Wolter et al., 1992;Van Aelst et al., 1993). Consequently, our observation that 18:1-LPE can rescue the inhibition of pollen germination caused by manoalide is most interesting. However, exogenous 18:1-LPE had no effect on pollen germination under normal conditions. These results suggest that only a very minimal amount of 18:1-LPE is needed to regulate pollen germination and tube growth. One possible explanation for this is that PLA 2 -g and -d gene expression level is very low in pollen. Whether 18:1-LPE has a direct function in membrane deformation or works as a lipid signal molecule has not yet been elucidated, and the detailed mechanism of how 18:1-LPE regulates pollen germination and tube growth remains to be solved.

Pollen-Expressed PLA 2 Members May Have a Preferential Role in Pollen Development and Germination
All three PLA 2 s (-b, -g, and -d) studied here have primarily the same biochemical function; they catalyze the hydrolysis of phospholipids at the sn-2 position to generate a lysophospholipid and a free fatty acid (Lee et al., 2005). Their role in pollen development, however, appears to be redundant, since no phenotypic differences were found in a single KO mutant. Based on our results and those reported previously, we suggest that PLA 2 -b, -g, and -d have preferential roles in pollen development and germination. We base this proposal on the following. First, these enzymes were expressed differentially during pollen development. In pollen, PLA 2 -b was expressed continuously during all stages, while PLA 2 -g and -d were expressed predominantly from the tricellular pollen stage onwards, although they were expressed very weakly at earlier stages ( Figures 1B and 1G). Second, the subcellular localizations of these three enzymes were slightly different: PLA 2 -b and -d localized to the ER, while PLA 2 -g localized to the Golgi and ER (Figures 5 and 6; Seo et al., 2008). Third, all of these enzymes have different acyl group preferences and head-group specificity. PLA 2 -b shows a preference for palmitoyl acyl chains over linoleoyl acyl chains and slightly prefers PE over phosphatidylcholine (PC), while PLA 2 -g and -d have an almost exclusive preference for PE but do not exhibit any significant acyl group preference (Lee et al., 2005). Finally, the extent of enzymatic inhibition by the different PLA 2 inhibitors was subtly different: PLA 2 -b was strongly inhibited by AA, slightly inhibited by manoalide, but not inhibited by ONO, while PLA 2 -g and -d were strongly inhibited by manoalide, slightly inhibited by ONO, but not inhibited by AA ( Figure 8A; see Supplemental Figure 5 online; Seo et al., 2008;Lee et al., 2010). We have shown here that different kinds of PLA 2 inhibitors had diverse inhibition effects on pollen germination. Similarly, several PLA 2 inhibitors have been shown to affect root hair growth in different ways. In PIN-overexpressing plants, root hair growth was conspicuously recovered by ONO, only slightly recovered by AA-COCF3 and PACOCF3, and not recovered at all by HELSS . Gardiner et al. (2008) reported that AA inhibits root growth by disrupting cortical microtubule arrays in Arabidopsis. In our experiment, however, AA had no inhibitory effect on pollen germination. This difference implies that AA inhibits PLA 2 -a and/or -b but not -g and -d. The results of our in vitro inhibition assay of PLA 2 by AA confirmed that AA inhibited PLA 2 -a but not -g and -d. Such pharmacological evidence supports the belief that each PLA 2 isoform has a specific function at the cellular level and is plant tissue dependent.
We therefore infer that PLA 2 -b may play a more vital role in pollen development than PLA 2 -g and -d, primarily because membrane trafficking and deformation by the ER and Golgi occur as early as the uninucleate late microspore stage (Yamamoto et al., 2003) when only PLA 2 -b was found to be expressed ( Figures  1B and 1G). PLA 2 -g and -d were primarily expressed at the pollen maturation stage, suggesting that they may function in processes of pollen germination and tube growth rather than in processes associated with pollen development. This hypothesis is supported by our results showing that pollen germination was inhibited by manoalide and that this inhibition could be rescued by LPE but not by LPC or LPA (Figure 9). Plant Low Molecular Weight PLA 2 s Structurally Resemble Animal sPLA 2 s but Are Functionally Similar to Animal cPLA 2 The low molecular weight PLA 2 s in plants have been classified as secreted PLA 2 s based entirely on their structural features, such as a large number of disulfide bonds, two antiparallel central helices, a catalytic His-Asp dyad, a hydrogen binding network between the interfacial binding site and the catalytic site, and a calcium binding loop (Berg et al., 2001;Lee et al., 2005;Mansfeld et al., 2006;Mansfeld, 2009). The results of a computer-based homology modeling analysis suggest that these plant sPLA 2 s are similar to those of sPLA 2 s from bovine pancreas and bee venom in terms of their tertiary structures of the active site and calcium binding loop (Mansfeld et al., 2006) In addition, a transient expression analysis of GFP in onion epidermal cells showed that PLA 2 -b and -g are secreted to the cell wall/extracellular space Lee et al., 2003).
However, we found several pieces of inconsistent experimental data against this classification based on an analogy of sequence homologies. The first consideration is the subcellular localization of PLA 2 s. The extracellular space in plants is known to be acidic, with a pH 5 to ;6. However, the optimum pH for activities of several plant PLA 2 s has been determined to be pH 6.5 to ;10 (Stå hl et al., 1998;Lee et al., 2005). The pH in the ER is near neutral, namely, 7.2 to ;7.4 (Kim et al., 1998), and the pH of Golgi is more acidic from cis-to trans-Golgi (;6.2) (Caldwell and Howell, 2008). Our confocal microscopy results demonstrate that PLA 2 -g and -d localized to the ER and/or Golgi, which essentially unravels the inconsistency between the subcellular localization and the optimum pH required. Contrary to the extracellular space, the ER and/ or Golgi provide the ideal environments for the PLA 2 s to carry out proper enzyme activity. Second, the head-group specificities of plant sPLA 2 s are quite different from those of animal sPLA 2 s. Most of the animal sPLA 2 s prefer anionic phospholipids, but plant enzymes have a preference for zwitterionic phospholipids, such as PE and PC (Mansfeld et al., 2006;Mansfeld and Ulbrich-Hofmann, 2007). According to Mansfeld (2009), these differences may be due to the evolutionary adaptation of plant PLA 2 specificity to the difference in the natural phospholipid composition between plants and animals. Therefore, our results provide us with a possible explanation of why typical cPLA 2 s have not been found in plants and provide experimental evidence for classifying these so-called plant sPLA 2 s as functional cPLA 2 s.
In conclusion, we suggest that the PLA 2 -b, -g, and -d expressed in pollen play critical roles in the physiological processes involved in pollen development most likely by modulating membrane trafficking and deformation as well as by providing the lipid signaling molecules for pollen germination and tube growth. We also suggest that the plant low molecular weight PLA 2 is functionally similar to cPLA 2 rather than to the sPLA 2 found in animals.

Vector Constructions
The PLA 2 -g and PLA 2 -d promoters were isolated from Arabidopsis thaliana genomic DNA. A total of 660 bp of the PLA 2 -g 59 flanking region (ProPLA 2 -g; 252 bp of the 39 untranslated region [UTR] of the preceding gene + 402 bp of the promoter region + 6 bp of the 59 UTR) and 550 bp of the PLA 2 -d 59 flanking region (ProPLA 2 -d; 224 bp 39 UTR of the preceding gene + 223 bp of promoter region + 103 bp of 59 UTR) were PCR amplified. The PCR products were digested with HindIII and BamHI and inserted into the HindIII/BamHI sites of the pBI121 vector (Clontech). To make the RNAi constructs, the 576-bp full-length ORF from ATG to TAG of PLA 2 -d was cloned into the pHANNIBAL vector (CSIRO) for both sense (KpnI) and antisense (BamHI) orientation under the control of the Lat52 promoter (Twell et al., 1989) (designated as LPRNAi) and the PLA 2 -d promoter (designated as DPRNAi), respectively, and both fusion proteins were then subcloned into the pART27 binary vector (Gleave, 1992) at a NotI restriction site. To identify the subcellular localization of PLA 2 -g and -d in pollen, ProLat52:PLA 2 -g:smRS-GFP and the ProLat52:PLA 2 -d: smRS-GFP were constructed as follows. First, Lat52 promoter, PLA 2 -g ORF, PLA 2 -d ORF, smRS-GFP ORF, and NOS terminator were amplified with the specific primer sets (see Supplemental Table 1 online). Second, the amplified Lat52 promoter was digested with HindIII and BamHI and then cloned into the HindIII/BamHI-digested pPZP221 (Hajdukiewicz et al., 1994), and the NOS terminator was inserted into pPZP221 using SacI/EcoRI. Third, the PCR-amplified smRS-GFP product was digested with BamHI and SacI and cloned into the BamHI/SacI-digested construct. Finally, each amplified PLA 2 -g and PLA 2 -d ORF was introduced into the BamHI site of this construct. For construction of the Pro35S:PLA 2 :YFP fusion protein construct, YFP was inserted into the SmaI/BamHI-digested pFAST vector (Clontech), and the full-length PLA 2 cDNA was subsequently cloned into the KpnI-digested pFAST-YFP vector. For PLA 2 :YFP expression in Arabidopsis pollen, PLA 2 promoters (-g and -d) were substituted for the cauliflower mosaic virus 35S promoter in the Pro35S:PLA 2 :YFP vector (designated ProPLA 2 -g:PLA 2 -g:YFP and Pro-PLA 2 -d:PLA 2 -d:YFP, respectively). PCR reactions were performed using Phusion High-Fidelity DNA polymerase (Finnzymes), and all of the constructs above were verified by DNA sequencing. The primers used for the cloning are described in Supplemental Table 1 online.

In Vitro Pollen Germination Assay
The different pollen grains were harvested on a large scale as described by Johnson-Brousseau and McCormick (2004), with minor modifications, and germinated in vitro as described by Boavida and McCormick (2007) on solid germination medium (0.01% H 3 BO 3 , 1 mM MgSO 4 , 5 mM KCl, 5 mM CaCl 2 , 10% sucrose, and 1.5% low-melting agarose, pH 7.5) at 228C and 100% humidity in the dark. The germination conditions for Arabidopsis and tobacco pollen were the same. Pollen tube germination rates were computed by dividing the total number of germinated tubes by the number of grains. The PLA 2 inhibitors manoalide (Biomol), BPB (TCI), AACOCF3 (Biomol), PACOCF3 (Biomol), bromoenol lactone (BEL; Cayman), and ONO-RS-082 (Biomol) were dissolved in DMSO, and the PLA 2 inhibitor AA (Biomol) was dissolved in water. Each inhibitor was used at concentrations of 3, 5, 7, 10, 20, and 50 mM. For the complementation assay by lipids, 18:1-LPA, 18:1-LPC, 18:1-LPE, and 18:1-PA dissolved in solvents were purchased from Avanti Polar Lipids. Fresh stock solutions were prepared by evaporating the solvent in nitrogen gas and then dissolving the lipids in water by sonication. The lipid stocks were added to the pollen germination medium to a final concentration of 50 mg/mL.

RT-PCR and Quantitative Real-Time PCR
Total RNA was extracted using a Nucleospin RNA plant extraction kit (Macherey-Nagel). A total of 5 mg RNA was reverse transcribed using SuperScript III (Invitrogen) with the oligo(dT) primer according to the manufacturer's instructions. For the RT-PCR analysis, 23 cycles were used for eIF4-a1 as an internal control and 30 cycles for the PLA 2 s. For the real-time PCR, the SYBR Green PCR master mix (Roche) and LightCycler 480II (Roche) were used according to the manufacturer's instructions. The PCR conditions were 958C for 10 min, followed by 40 cycles of 958C for 10 s, 608C for 10 s, and 728C for 20 s. The primers used for RT-PCR and real-time PCR are listed in Supplemental Table 1 online.

siRNA RNA Gel Blot Analysis
Total RNA was extracted from the buds and the open flowers, and 40 mg of total RNA was fractionated in a 15% polyacrylamide gel. The probe for the detection of siRNA was labeled with 32 P-dCTP using a Random Primer DNA labeling kit (Takara). DNA oligomers of 23 and 27 nucleotides were used as molecular size markers. Hybridization signals were detected using a BAS-2500 Bioimaging analyzer (Fujifilm).

Scanning Electron Microscopy and TEM
For scanning electron microscopy, the pollen grains from open flowers were mounted on stubs over double-sided carbon tape and coated with gold particles using a sputter coater (SEM Coating System; Bio-Rad). Specimens were observed with a scanning electron microscope (JEOL5300; Jeol) at an accelerating voltage of 25 kV. The images were transferred to a digital image by Semafore (Jeol), a digital image recording and processing system for scanning electron microscopy. For the semithin sections and TEM, flowers were first immersed in a fixative (2.5% glutaraldehyde and 2% paraformaldehyde in 0.5 M cacodylate buffer, pH 7.5), and the sepals and petals were then removed to allow the fixative to reach the anther, and the samples were vacuum infiltrated (25 Pa for 30 min). After overnight fixing at 48C, the samples were rinsed with 0.5 M cacodylate buffer, postfixed with 1% OsO 4 in 0.5 M cacodylate buffer, pH 7.5, overnight at 48C and then washed again with 0.5 M cacodylate buffer. The samples were dehydrated at 48C stepwise through an ethanol series (10% increments, 20 min per step, 10 to ;100%), and then transferred successively to 3:1, 1:1, and 1:3 (v/v) mixtures of ethanol and Spurr's resin (Ted Pella; 6 h at each step), and then finally transferred to 100% Spurr's resin and left overnight. Each sample was cured in the flat embedding mold for 2 d at 708C. Semithin sections (1 mm) were cut on an ultramicrotome model MTX (RMC) using a glass knife and then stained with 1% (w/v) toluidine O. For TEM, ultrathin sections (80-nm thick) were collected in copper grids (200 mesh), stained with 2% (w/v) uranyl acetate and Reynols's lead citrate, and analyzed by JEM 1010 (Jeol) at 80 kV.

Subcellular Localization Using the Confocal Laser Scanning Microscope
To express transient fluorescent fusion proteins in tobacco leaves, we modified published agroinfiltration methods (Bendahmane et al., 2000;Sparkes et al., 2006). The infiltration solution contained 10 mM MgCl 2 , 10 mM MES, pH 5.7, 0.5% glucose, and 200 mM acetosyringone (59-dimethoxy-4'-hydroxy acetophenone; Sigma-Aldrich). Three or four days after infiltration, the abaxial leaf surface was observed with a confocal laser scanning microscope (LSM510 Meta; Carl Zeiss). For CFP, YFP, and GFP, the excitation wavelengths were 458, 514, and 488 nm, respectively, and the emitted fluorescence was collected with a bandpass filter at 470 to ;500, 530 to ;600, and 505 to ;550 nm, respectively. Imaging colocalization of the GFP and YFP constructs was according to the method described by Brandizzi et al. (2002). The excitation lines of an argon ion laser (GFP, 458 nm; YFP, 514 nm) were used alternately with line switching, using the multitrack facility of the LSM510 Meta confocal microscope. Fluorescence was detected using a 458/514-nm dichroic beam splitter and band-pass filters of 470 to ;500 nm for GFP and 530 to ;600 nm for YFP. For the quantitative colocalization analysis, spectral bleed-through was eliminated. Threshold values were selected to remove background. FM4-64 and ER-tracker (ER-Tracker Red; Invitrogen) were detected using 488/>615 and 543/ >620 excitation/emission filter sets, respectively. The argon ion laser line power was set at 10%. All confocal images were processed with LSM5 software version 4.0 (Carl Zeiss).

Supplemental Data
The following materials are available in the online version of this article. Supplemental Table 1. Primers Used in This Article.