AtPV42a and AtPV42b Redundantly Regulate Reproductive Development in Arabidopsis thaliana

Background The conserved SNF1/AMPK/SnRK1 complexes are global regulators of metabolic responses in eukaryotes and play a key role in the control of energy balance. Although α-type subunits of the SnRK1 complex have been characterized in several plant species, the biological function of β-type and γ-type subunits remains largely unknown. Here, we characterized AtPV42a and AtPV42b, the two homologous genes in Arabidopsis, which encode cystathionine-β-synthase (CBS) domain-containing proteins that belong to the PV42 class of γ-type subunits of the plant SnRK1 complexes. Methodology/Principal Findings Real-time polymerase chain reaction was performed to examine the expression of AtPV42a and AtPV42b in various tissues. Transgenic plants that expressed artificial microRNAs targeting these two genes were created. Reproductive organ development and fertilization in these plants were examined by various approaches, including histological analysis, scanning electron microscopy, transmission electron microscopy, and phenotypic analyses of reciprocal crosses between wild-type and transgenic plants. We found that AtPV42a and AtPV42b were expressed in various tissues during different developmental stages. Transgenic plants where AtPV42a and AtPV42b were simultaneously silenced developed shorter siliques and reduced seed sets. Such low fertility phenotype resulted from deregulation of late stamen development and impairment of pollen tube attraction conferred by the female gametophyte. Conclusions Our results demonstrate that AtPV42a and AtPV42b play redundant roles in regulating male gametogenesis and pollen tube guidance, indicating that the Arabidopsis SnRK1 complexes might be involved in the control of reproductive development.


Introduction
As sessile organisms, plants are exposed to a constantly changing environment. It is therefore essential for them to sense and integrate endogenous and environmental stimuli to generate suitable cell responses for optimizing growth and development [1]. The control of energy balance is one of the crucial factors for such adaptive processes in plants, which involves a group of plant protein kinases, the SNF1-Related Kinase 1 (SnRK1) family [2].
SnRK1 is a serine/threonine kinase that has a catalytic domain similar to that of Sucrose non-fermenting 1 (SNF1) from yeast and AMP-activated protein kinase (AMPK) from mammals [2,3]. In yeast, SNF1 is one of the main regulators of carbon metabolism and mediates the diauxic shift from fermentative to oxidative metabolism in response to glucose starvation [4,5]. AMPK, the mammalian counterpart of SNF1, is an energy sensor that regulates energy balance by activating the processes that produce energy, while inhibiting those that consume energy [6][7][8]. In plants, SnRK1-type kinases play an important role in the global regulation of metabolism, and are also involved in plant development and stress responses [1]. SnRK1s from different plant species can complement the yeast snf1g mutant phenotype, demonstrating an evolutionary conservation in their function [9][10][11][12][13][14][15].
SNF1, AMPK-a, and SnRK1 serve as the catalytic a-subunits that are associated with other two regulatory subunits (b-type and c-type) in the conserved heterotrimeric kinase SNF1/AMPK/ SnRK1 complexes found in fungi, mammals and plants [1]. Association of the three subunits in SNF1/AMPK/SnRK1 complexes is differentially regulated by various hormonal and environmental signals, cell and tissue types, and developmental stages. In yeast, b subunits anchor a and c subunits, thus directing the kinase complexes into their targets or specific subcellular localizations, while c subunits function in activating the kinase activity of a subunits [16][17][18][19].
Three c-type subunits (AMPKc1, AMPKc2, AMPKc3) in mammals have been identified as being homologous to the c subunit of the SNF1 complex, SNF4, in yeast [20]. Furthermore, phylogenetic analysis of SNF4-like plant proteins has revealed three subgroups of c-type subunits in plants: AKINbc-, AKINc-, and PV42-type proteins [3,15,[21][22][23]. While it has been shown that AKINbc contributes to SnRK1 heterotrimeric complexes in Arabidopsis and is possibly involved in plant-pathogen interactions [23], the biological function of AKINc-and PV42-type proteins remains unclear.
The conservation among c-type subunits in fungi, mammals and plants partly lies in the four cystathionine-b-synthase (CBS) domains found in these proteins [1,24]. The CBS domain was first discovered in the genome of the archaebacterium Methanococcus jannaschii [24,25]. It is about 60 residues long, and composed of a sheet of three b strands packed with two a helices. CBS domains have been found to bind to metallic ions such as Mg 2+ and adenosyl compounds such as AMP, ATP, and S-adenosyl-Lmethionine, which may trigger a conformational change in the CBS domains, thus regulating the activity of associated enzymatic domains [26][27][28]. The CBS domain-containing proteins comprise a large family of evolutionarily conserved proteins that have been found in all kingdoms of life, among which the mammalian ones are so far the best characterized. In humans, CBS domaincontaining proteins are highly diversified and have been found to undertake various biological roles, ranging from metabolic enzymes and transcriptional regulators to ion channels and transporters [29]. In contrast, very few information is available for the CBS domain-containing proteins in plants. So far, 48 Arabidopsis proteins have been designated as CBS domaincontaining proteins [29], which include c-type subunits of the SnRK1 complex.
In this study, we show that AtPV42a and AtPV42b, the two homologous genes in Arabidopsis, encode CBS domain-containing proteins that belong to the PV42 class of c-type subunits of the plant SnRK1 complexes. They are expressed in different tissues throughout the developmental stages of Arabidopsis. Artificial microRNA-mediated silencing of both AtPV42a and AtPV42b exhibits the defects in late stamen development and pollen tube attraction conferred by the female gametophyte, which results in reduced seed sets. These results suggest that AtPV42a and AtPV42b play redundant roles in regulating male gametogenesis and pollen tube guidance in Arabidopsis.

Results
AtPV42a and AtPV42b are putative c-type subunits of the plant SnRK1 complexes AtPV42a (At1g15330) and AtPV42b (At1g80090) genes are two close homologues in Arabidopsis. The AtPV42a gene consists of 2 exons and 1 intron, while AtPV42b consists of 5 exons and 4 introns ( Figure 1A). A BLAST search against the NCBI protein database revealed that AtPV42a and AtPV42b were two Arabidopsis proteins homologous to PV42 from Phaseolus vulgaris, which is a founding member of the PV42 class of c-type subunits of the plant SnRK1 complexes [3,23]. AtPV42a and AtPV42b shared 60% and 54% amino acid identity with PV42, respectively. Multiple sequence alignment and protein domain analysis revealed that like other c-type subunits of the SNF1/AMPK/SnRK1 complexes, such as SNF4 from yeast, AtPV42a and AtPV42b contained four CBS domains ( Figure 1B). These sequence analyses imply that both AtPV42a and AtPV42b are putative members in the PV42 class of c-type subunits of the Arabidopsis SnRK1 complex.

Expression of AtPV42a and AtPV42b in Arabidopsis
To examine the spatial and temporal expression patterns of AtPV42a and AtPV42b in Arabidopsis, real-time PCR analyses were performed with gene-specific primers using total RNA extracted from various tissues in 28-day-old adult plants and from the whole seedlings at different developmental stages (3-, 8-, 14-day-old). Overall, AtPV42a and AtPV42b exhibited a similar spatial expression pattern in most of tissues examined in adult plants ( Figure 2A). Their expression was relatively high in rosette leaves, cauline leaves, open flowers, and developing silique, but low in stems and floral buds. The only discrepancy was that the relative expression of AtPV42b in roots as compared with other tissues was lower than that of AtPV42a. It is noteworthy that both AtPV42a and AtPV42b were expressed at the highest levels in dry seeds ( Figure 2A). In the seedlings 3, 8, and 14 days after germination, the expression of both genes remained at stable levels, with a slight decrease in transcripts levels concomitant to an increase in seedling age ( Figure 2B).
We further performed in situ hybridization to study the expression of these two genes in developing flowers. Both genes were detectable in the developing septum inside the gynoecia and microspore mother cells in the locules of stage 9 flowers ( Figure 2C, F). Their expression levels were either low or absent in anther cells in the flowers at late stages ( Figure S1), but persistent in developing ovules in stages 11 and 13 flowers. In stage 11 flowers in which integuments were just initiated on the ovules, the expression of both genes was detected in funiculi and ovules ( Figure 2D, E, G, H). In stage 13 flowers in which integuments completely enveloped the nucellus, both genes were also expressed in whole ovules ( Figure 2I, J).
High sequence similarity and comparable gene expression patterns between AtPV42a and AtPV42b indicate that they may play similar roles in Arabidopsis growth and development.
Artificial microRNA-mediated silencing of AtPV42a and AtPV42b To investigate the biological roles of AtPV42a and AtPV42b in Arabidopsis, we first attempted to identity insertion mutants from public resources. While insertion mutants of AtPV42a were not available in all public resources searched, a SAIL line (CS823876) containing a T-DNA insertion at the last exon of the AtPV42b gene was obtained from Arabidopsis Biological Resource Center and named as atpv42b-1 ( Figure 1A). The T-DNA insertion and the resulting disrupted transcription of AtPV42b in atpv42b-1 were confirmed by genotyping PCR and RT-PCR using the primers flanking the insertion site, respectively ( Figure S2, Figure S3). atpv42b-1 did not exhibit visible phenotypes under normal growth conditions. This could be due to two reasons: (1) the incomplete AtPV42b transcript produced from the coding region preceding the T-DNA insertion site may still function; (2) functional redundancy between AtPV42a and AtPV42b may exist.
To create knockdown lines for AtPV42a and AtPV42b, we generated amiR-atpv42a and amiR-atpv42b-1 transgenic plants that expressed artificial microRNAs specifically targeting these two respective genes. We obtained 17 and 20 independent transformants for amiR-atpv42a and amiR-atpv42b-1 at the T1 generation, respectively. None of the 17 amiR-atpv42a transgenic lines showed visible phenotype, whereas 9 amiR-atpv42b-1 lines developed shorter siliques and reduced seed sets ( Figure 3A-C). Since AtPV42a and AtPV42b shared high sequence similarity, micro-RNAs designed for either of them could simultaneously affect the expression of the other in transgenic plants. Thus, the expression of both AtPV42a and AtPV42b was examined in rosette leaves of 7 and 10 selected amiR-atpv42a and amiR-atpv42b-1 T1 transformants, respectively. As expected, the expression levels of AtPV42a were dramatically decreased, whereas the levels of AtPV42b were not significantly changed in the 7 amiR-atpv42a transformants ( Figure 2K). However, in most of the 10 selected amiR-atpv42b-1 lines, the transcript levels of both AtPV42a and AtPV42b were significantly downregulated, and such downregulation was relevant to the low fertility phenotype exhibited by these transgenic  lines ( Figure 2L, Table 1). These observations demonstrate that plants are defective in reproductive development only when the expression of both AtPV42a and AtPV42b is significantly compromised, suggesting that AtPV42a and AtPV42b function redundantly in controlling Arabidopsis reproductive development. Therefore, we chose the transgenic line 10 of amiR-atpv42b-1 (hereafter referred to as amiR-atpv42b-1) ( Figure 2L, Table 1), which displayed the strongest phenotypes, for further morphological and molecular characterization.
To confirm the functional redundancy of AtPV42a and AtPV42b, the transgenic line 3 of amiR-atpv42a (hereafter referred to as amiR-atpv42a) ( Figure 2K) was crossed with the T-DNA insertion mutant atpv42b-1 ( Figure 1A). The resulting homozygous progenies, where the expression of both AtPV42a and AtPV42b is disrupted ( Figure S4), exhibited a similar low fertility phenotype to amiR-atpv42b-1 (Table 1). This is in agreement with our suggestion that AtPV42a and AtPV42b function redundantly in reproductive development.
To further test whether a single knockdown of AtPV42b impairs the reproductive process, we generated amiR-atpv42b-2 transgenic plants overexpressing another artificial microRNA targeting AtPV42b. Real-time PCR assay showed that in 6 selected amiR- atpv42b-2 independent lines, AtPV42b was significantly downregulated, whereas the expression of AtPV42a was almost unaffected ( Figure S5). Unlike amiR-atpv42b-1, amiR-atpv42b-2 transgenic plants were morphologically indistinguishable from wild-type plants ( Table 1). These results, together with the phenotype of atpv42b-1, further corroborate that AtPV42b functions redundantly with other factors, such as AtPV42a, to regulate reproductive development.
We also created 18 and 15 independent transgenic lines overexpressing AtPV42a and AtPV42b, respectively. While AtPV42a and AtPV42b were overexpressed in most of these lines, none of them exhibited abnormal phenotypes in reproductive development (data not shown).

amiR-atpv42b-1 is defective in late stamen development
To uncover the developmental events responsible for the reduced fertility in amiR-atpv42b-1, we compared the morphology of amiR-atpv42b-1 and wild-type plants at different developmental stages. amiR-atpv42b-1 appeared normal during the vegetative phase, floral transition, and early stages of flower development (data not shown). In amiR-atpv42b-1 floral buds at stage 12, floral organs were still morphologically normal ( Figure 3D, E, H, I), whereas at the anthesis stage (flower stages 13 and 14), fewer pollen grains were produced ( Figure 3F, G, J, K). Moreover, the filaments in amiR-atpv42b-1 did not elongate enough to position the locules above the stigma at the flower stage 14 ( Figure 3G, K).
Scanning electron micrograph (SEM) revealed that as compared with those in wild-type plants, more than 50% of the pollen grains from amiR-atpv42b-1 were severely shrunken and of irregular shapes at the anthesis stage ( Figure 4A, B, Table S1). Thus, we examined the transverse sections of wild-type and amiR-atpv42b-1 anthers under the light microscope to investigate the changes at the cellular level. We found that even at anther stage 10 (approximately flower stages 11-12) when tapetum degeneration initiated, there was no visible difference in microspores and other anther tissues between wild-type and amiR-atpv42b-1 ( Figure 4C), indicating that microsporogenesis in amiR-atpv42b-1 may be normal [30]. However, at anther stage 11 when pollen mitotic divisions occur in wild-type plants, most of the microspores in amiR-atpv42b-1 seemed to lack dense cell contents ( Figure 4C). At anther stage 12, when microgametogenesis was completed and microspores developed into tricellular haploid pollen grains, many pollen grains in amiR-atpv42b-1 anthers were shrunken with irregular shapes (Figure 4C), which is consistent with the SEM results ( Figure 4A, B).
We further performed transmission electron microscopy (TEM) to compare the finer cellular structures of the developing pollen cells in wild-type and amiR-atpv42b-1 anthers. Although the transverse sections of amiR-atpv42b-1 anthers at anther stage 9 showed normal-looking microspores under the light microscope ( Figure 4C), TEM revealed that the plasma membrane was often withdrawn from the cell wall in amiR-atpv42b-1 pollen grains ( Figure 5). This is an indication of plasmolysis, implying the abnormal cellular osmotic homeostasis inside the amiR-atpv42b-1 pollen grains. At anther stage 12, shrunken pollen grains in amiR-atpv42b-1 lost most of the cell contents ( Figure 5), indicating that there is a continuous improper cellular osmotic homeostasis from anther stage 9 to 12. These TEM results explain why the microspores at anther stage 11 in amiR-atpv42b-1 demonstrate reduced amount of cell contents ( Figure 4C).
Taken together, these observations suggest that downregulation of AtPV42a and AtPV42b in amiR-atpv42b-1 results in the production of some collapsed pollen grains particularly at late stages of anther development. This might partly contribute to the reduced fertility observed in amiR-atpv42b-1.
Female gametophytes of amiR-atpv42b-1 are defective in pollen tube reception To determine whether the defective stamen development is the only cause of the low fertility found in amiR-atpv42b-1, reciprocal crosses between amiR-atpv42b-1 and wild-type plants were performed. After saturated pollination with either wild-type or amiR-atpv42b-1 pollen grains, the majority of wild-type ovules were normally fertilized (Table 2). However, more than 50% of amiR-atpv42b-1 ovules were unfertilized regardless of the father plants ( Table 2). These results indicate that female tissues in amiR-atpv42b-1 are partly defective. As the morphology of amiR-atpv42b-1 carpels during flower development was almost normal ( Figure 3H-K), we then examined the ovules inside the carpels before and after fertilization using SEM. amiR-atpv42b-1 ovules were morphologically comparable to wild-type ones prior to fertilization ( Figure 6A, D). SEM of wild-type pistils two days after saturated pollination with either wild-type or amiR-atpv42b-1 pollen grains revealed that most of the ovules have been fertilized ( Figure 6B, C), which is in agreement with the observation on seed sets in wild-type plants ( Table 2). This implies that some healthy pollen grains produced in amiR-atpv42b-1 could function normally during fertilization. On the contrary, many amiR-atpv42b-1 ovules were not fertilized after saturated pollination regardless of the source of pollen grains ( Figure 6E, F), demonstrating the fertilization defect in amiR-atpv42b-1 female gametophytes.
We further investigated the growth of pollen tubes inside the pistils after reciprocal crosses to explore the underlying mechanism of the fertilization defect in amiR-atpv42b-1. At 20 hours after saturated pollination, pistils were collected, fixed, cleared, stained with aniline blue, and observed under UV microscope ( Figure 6G-J). Likewise, the pollen tube growth was independent of the source of pollen grains, but mainly relied on the maternal plants used. In wild-type pistils pollinated with either wild-type or amiR-atpv42b-1 pollen grains, pollen tubes elongated longitudinally through the stylar transmitting tract from the stigma, and then elongated laterally to the ovules for fertilization ( Figure 6G, H). However, in amiR-atpv42b-1 pistils, although the longitudinal pollen tube growth through the transmitting tract was similar to that in wild-type pistils, the lateral pollen tube growth towards the ovules was largely abolished (Figure 6I, J). These results demonstrate that the fertilization defect in amiR-atpv42b-1 pistils is mainly due to abnormal pollen tube attraction by ovules.
As synergid cells are responsible for guiding pollen tubes to the embryo sac by secreting chemical attractants [31,32], we tested the expression of several synergid cell-specific genes in amiR-atpv42b-1 and found that the expression of LORELEI (LRE) was altered ( Figure S6). It has been reported that LRE is involved in proper pollen tube reception [33]. In lre mutants, pollen tubes reaching the embryo sac frequently continue to grow inside the embryo sac, resulting in the failure of fertilization. In amiR-atpv42b-1, where pollen tube attraction was blocked, the expression level of LRE was significantly increased in open flowers ( Figure S6), indicating that disruption of pollen tube guidance to the embryo sac in amiR-atpv42b-1 might be relevant to the upregulation of LRE.

Discussion
Plant SnRK1-type kinases, which serve as the catalytic asubunits and are associated with non-catalytic b-type and c-type subunits in the SnRK1 complexes, play important roles in the global regulation of metabolism in response to cellular and environmental signals [1]. Studies in a wide range of plant species have shown that SnRK1s are involved in the regulation of various developmental processes. For example, disruption of the SnRK1 kinase in transgenic barley plants results in abnormal pollen development and male sterility [34]. The rice SnRK1A protein kinase acts as an important intermediate in the sugar signaling cascade and plays a key role in regulating seed germination and seedling growth [35]. In the moss Physcomitrella patens, downregulation of SnRK1 kinases shows pleiotropic phenotypes including developmental abnormalities [36]. Although these studies have shed light on the function of the catalytic a-subunits in the SnRK1 complexes, the biological function of c-type subunits and their effects on plant development are so far unknown.
In this study, we have characterized AtPV42a and AtPV42b, two close homologues encoding CBS domain-containing proteins in the PV42 class of c-type subunits of the SnRK1 complexes in Arabidopsis. Downregulation of AtPV42a and AtPV42b not only disturbs late stamen development, but also impairs pollen tube reception conferred by the female gametophyte, which eventually results in the low fertility. The function of AtPV42a and AtPV42b in the reproductive development is consistent with their expression in floral organs. Both genes are expressed in the developing septum inside the gynoecia and microspore mother cells in stage 9 flowers, and later in developing ovules. These results suggest that the noncatalytic c-type subunits in the SnRK1 complexes play an important role in mediating plant reproductive growth. This is in accordance with a previous finding that shows the involvement of an a-type subunit of the SnRK1 complexes in pollen development in barley [34].
In amiR-atpv42b-1 where AtPV42a and AtPV42b are downregulated, defective stamen development is mainly attributed to abnormal microgametogenesis. Histological analysis of anther transverse sections and SEM and TEM of pollen cells demonstrate that anther defects in amiR-atpv42b-1 mainly occur in microspores at the mitotic phase. Pollen development is divided into two phases, microsporogenesis and microgametogenesis [37][38][39]. During microsporogenesis, pollen mother cells undergo meiosis and form tetrads of microspores within the pollen sacs. During microgametogenesis, separate microspores undergo mitosis and differentiate into mature pollen grains. In this process, enzymes Figure 5. Transmission electron microscopy of wild-type and amiR-atpv42b-1 pollen grains. Wild-type and amiR-atpv42b-1 pollen grains at anther stages 9 and 12 are shown. In amiR-atpv42b-1, during plasmolysis occurring at anther stage 9, the plasma membrane (arrows) of the pollen grain is withdrawn from the cell wall, while at anther stage 12, the cytoplasm is almost invisible in the severely shrunken pollen grain. Bars, 2 mm. doi:10.1371/journal.pone.0019033.g005 Table 2. Phenotypic analysis of reciprocal crosses between wild-type and amiR-atpv42b-1 plants. Ovule  are abundant and metabolism is highly active in microspores, which obtain nutrients and water from degenerated tapetum and undergo an asymmetric mitotic division to generate a large vegetative cell and a small generative cell. The generative cell further undergoes a second mitosis to form two generative cells. It is noteworthy that active carbohydrate metabolism not only provides energy for cell division and differentiation during microgametogenesis, but also stores essential substances required for subsequent pollen germination and pollen tube growth. As the SnRK1 complex is the global regulator of carbohydrate metabolism [40], it may also be involved in microgametogenesis. This hypothesis is partly supported by our observation that disruption of AtPV42a and AtPV42b, which encode the c-type subunits of the SnRK1 complexes, compromises microgametogenesis, thus resulting in abnormal pollen grains. amiR-atpv42b-1 also exhibits impaired pollen tube attraction by the female gametophyte. In Arabidopsis, synergid cells accompanying the egg cell are primarily responsible for the pollen tube guidance to the female gametophyte and the release of sperm cells [31,32,41]. Intracellular metabolism is highly active in synergid cells, which uptake and transport metabolites into the embryo sac, and secrete some chemical substances into the filiform apparatus that is involved in pollen tube guidance and reception [42][43][44][45]. So far the molecular mechanisms underlying the function of synergid cells still remain largely unknown. Several genes specifically expressed in synergid cells, including MYB98, ZmEA1, and LRE, have been suggested as important regulators for pollen tube guidance [32,33,43]. Our results have shown that LRE is upregulated in amiR-atpv42b-1, which might be partially responsible for the defect in pollen tube attraction. LRE encodes a putative plant-specific GPI-anchor protein (GAP), a eukaryotic protein that functions as a lipid raft in cell-cell signalling. LRE is required for proper pollen tube guidance to the embryo sac in Arabidopsis and silencing of LRE results in continuous pollen tube growth inside the embryo sac [33]. It is possible that the SnRK1 complex plays a role in the intracellular metabolism of synergid cells. Thus, downregulation of AtPV42a and AtPV42b may affect normal metabolism of synergid cells, which is relevant to the altered expression of key regulators, such as LRE. This eventually results in the impaired pollen tube attraction by the female gametophyte.
Taken together, our results suggest that AtPV42a and AtPV42b play redundant roles in regulating male gametogenesis and pollen tube guidance. It will be interesting to further investigate how they interact with other subunits of the SnRK1 complex to regulate metabolic responses and contribute to the regulation of Arabidopsis development. In addition, the highest expression levels of AtPV42a and AtPV42b are detected in dry seeds (Figure 2A). It has been reported that disruption of the a-type subunit of the SnRK1 complex retards seed germination in rice [35]. Thus, besides the roles in regulating microgametogenesis and pollen tube guidance, AtPV42a and AtPV42b might also play a role in mediating the function of SnRK1 complexes in post-fertilization processes, such as seed development. This will be another intriguing topic to be further studied.

Plant materials and growth conditions
All Arabidopsis plants used in this study were in the Columbia (Col-0) background. For plants grown on soil, seeds were stratified at 4uC for 2 to 3 days and then transferred onto compost soil in shallow trays. The trays were placed in an environmentally controlled growth room at 22uC with a light cycle of 16 h light/ 8 h dark.

Arabidopsis transformation
Agrobacterium tumeficians GV3101 was used for floral dipping according to the published method with minor modifications [48]. Transgenic plants were screened for herbicide resistance against Basta (300 mg/l).

Expression analysis
Total RNA from different organs was isolated using the FavorPrep TM Plant Total RNA Mini Kit (Favorgen) and reverse transcribed using the SuperScript TM RT-PCR System (Invitrogen).
Real-time PCR was carried out using the Power SYBRH Green PCR Master mix (Applied Biosystems) as previously reported [49]. Primers used in real-time PCR amplifications are listed in Table S2. Non-radioactive in situ hybridization and synthesis of RNA probes were carried out as previously published [49].

Aniline blue staining of pollen tube growth
Aniline blue staining of pollen tube growth in pistils was carried out as previously described [50].

Scanning electron microscopy (SEM)
Mature pollen grains collected from wild-type and amiR-atpv42b-1 flowers after anther dehiscence were spread onto the surface of adhesive tapes, and directly observed under a JSM-6360LV scanning electron microscope (JEOL) at an accelerating voltage of 20 kV. Pistils were collected 1 to 2 days after pollination, and carefully opened with a sharp needle. SEM of pistils was performed as previously reported with some modifications [51]. Pistils were fixed with FAA (50% ethanol, 3.7% formaldehyde, and 5% acetic acid) for 2 h, and dehydrated through increasing concentrations of ethanol (30%, 50%, 70%, 80%, 90%, and 100%). After replacing absolute ethanol with the transfer liquid (amyl acetate), pistils were critical-point dried using CO 2 , mounted for sputter coating with gold palladium for 25 s, and observed under JSM-6360LV.

Transmission electron microscopy (TEM)
TEM was performed as previously reported [52]. Wild-type and amiR-atpv42b-1 inflorescences were collected, fixed overnight with 2.5% (v/v) glutaraldehyde in 16PBS (0.1 M, pH 7.4), washed with 16PBS for five times, and post-fixed in 1% OsO4 for 16 h at 4uC. The specimens were subsequently washed with 16PBS for five times, dehydrated in a graded ethanol series, replaced by epoxy dimethylmethane, and embedded in Epon812 resin. Semithin sections 2-4 mm in thickness were obtained using glass knives, stained with 0.5% toluidine blue in 16PBS, and then examined under a light microscope to confirm the developmental stages of microspores. Selected sections were further cut into sections 80-90 nm in thickness using a diamond knife with Leica Ultra-S microtome, stained with 2% uranyl acetate for 90 min and 6% lead citrate for 15 min, and then observed under a JEM-1230 transmission electron microscope (JEOL) at 90 kV. Figure S1 In situ hybridization of AtPV42a and AtPV42b in a stage 11 wild-type flower. A transverse section was hybridized with the antisense AtPV42a (A) or AtPV42b (B) probe. There are hybridization signals inside the gynoecia (arrows), while no signals are detectable in anther cells (arrowheads). Bars, 50 mm. (TIF) Figure S2 Genotyping of atpv42b-1 mutants using PCR analysis. Lanes 1, 3: PCR products amplified with the left (CS823876_LP) and right (CS823876_RP) primers flanking the AtPV42b genomic region. Lanes 2, 4: PCR products amplified with the T-DNA left border primer (LB2_SAIL) and the right primer for AtPV42b (CS823876_RP). We detected the amplification of a T-DNA fragment (lane 4), but not the AtPV42b genomic region (lane 3) in atpv42b-1, indicating that atpv42b-1 is a homozygous T-DNA insertion mutant. (TIF) Figure S3 Transcript levels of AtPV42b in rosette leaves (RL) and open flowers (OF) of wild-type and atpv42b-1 plants. AtPV42b expression is undetectable in atpv42b-1 as compared to that in wild-type plants. Transcript levels were determined by real-time PCR using a pair of primers flanking the T-DNA insertion site and are shown relative to TUB2 expression. Values are the mean 6 standard deviation from three replicates. (TIF) Figure S4 Transcript levels of AtPV42a and AtPV42b in rosette leaves of wild-type and amiR-atpv42a atpv42b-1 plants. The expression of both AtPV42a and AtPV42b is very low in amiR-atpv42a atpv42b-1 as compared to that in wild-type plants. Transcript levels were determined by real-time PCR and are shown relative to TUB2 expression. Values are the mean 6 standard deviation from three replicates. (TIF) Figure S5 Transcript levels of AtPV42a and AtPV42b in rosette leaves of 6 selected amiR-atpv42b-2 independent transgenic lines at the T1 generation. AtPV42a expression is not significantly affected in these transgenic plants as compared to that in wild-type plants, whereas AtPV42b expression is greatly decreased. Transcript levels were determined by real-time PCR and are shown relative to TUB2 expression. Values are the mean 6 standard deviation from three replicates. (TIF) Figure S6 Expression of LRE in open flowers of wild-type and amiR-atpv42b-1 plants. LRE expression is much upregulated in three independent amiR-atpv42b-1 lines than in wild-type plants. Transcript levels were determined by real-time PCR and are shown relative to TUB2 expression. Values are the mean 6 standard deviation from three replicates. (TIF)