KOMPEITO, an Atypical Arabidopsis Rhomboid-Related Gene, Is Required for Callose Accumulation and Pollen Wall Development

Fertilization is a key event for sexually reproducing plants. Pollen–stigma adhesion, which is the first step in male–female interaction during fertilization, requires proper pollen wall patterning. Callose, which is a β-1.3-glucan, is an essential polysaccharide that is required for pollen development and pollen wall formation. Mutations in CALLOSE SYNTHASE 5 (CalS5) disrupt male meiotic callose accumulation; however, how CalS5 activity and callose synthesis are regulated is not fully understood. In this paper, we report the isolation of a kompeito-1 (kom-1) mutant defective in pollen wall patterning and pollen–stigma adhesion in Arabidopsis thaliana. Callose was not accumulated in kom-1 meiocytes or microspores, which was very similar to the cals5 mutant. The KOM gene encoded a member of a subclass of Rhomboid serine protease proteins that lacked active site residues. KOM was localized to the Golgi apparatus, and both KOM and CalS5 genes were highly expressed in meiocytes. A 220 kDa CalS5 protein was detected in wild-type (Col-0) floral buds but was dramatically reduced in kom-1. These results suggested that KOM was required for CalS5 protein accumulation, leading to the regulation of meiocyte-specific callose accumulation and pollen wall formation.


Introduction
In higher plants, coordinated interactions between male and female reproductive organs are crucial for fertilization. As the first step of fertilization, pollen grains formed in male anthers are conveyed and attached to the stigmatic papillae cells of female pistils. This pollen-stigma adhesion event is tightly regulated, promoting crossing within species and preventing interspecies crossing. One of the key components in this process is exine, which is the surface coating of pollen grains; purified exine was shown to retain the ability to bind to the stigma [1].
Multiple lines of evidence suggest that callose synthesis is spatiotemporally regulated. In cotton, the SKS-like protein PSP231 triggers and fine-tunes callose synthesis and deposition [25]. In peas, a 145 kDa protein, namely, CFL1, has callose synthase activity and its molecular weight is lower than that predicted from the gene sequence [26,27]. Partial treatment of callose synthase with trypsin from tobacco pollen tubes in the presence of digitonin was shown to cause the activation of CalS activity. CDKG1, which encodes a cyclin-dependent kinase, is associated with the spliceosome and regulates CalS5 pre-mRNA splicing and protein maturation [28]. However, the mechanism by which callose synthase activity and callose production are regulated during microsporogenesis is not fully understood.
We hypothesized that as yet unknown factor(s) contribute to regulating pollen and callose wall development. To identify such factor(s), we aimed to find mutants that were defective in this process. In this article, we report a kompeito-1 (kom-1) mutant that was defective in pollen wall patterning and pollen-stigma adhesion in Arabidopsis thaliana. We then performed genetic and molecular analyses of KOMPEITO (KOM), a novel member of the Rhomboid protein family, to clarify its role in callose accumulation and pollen wall development.

Isolation of Kompeito-1 (Kom-1), Which Was Defective in Pollen Wall and Pollen-Stigma Adhesion
We screened mutants that were defective in plant fertility in an ethyl methanesulfonate (EMS)-induced Arabidopsis population. In the M2 population of EMS-treated plants, we found plants with shorter siliques ( Figure 1A,B) and fewer seeds per silique. Pollen grains of this mutant lacked regular, mesh-like pollen wall patterning and had dot-like sporopollenin deposits on their surfaces ( Figure 1E,F), as shown with the Auramine-O staining, which reacts with sporopollenin [29]. This mutant was named kompeito-1 (kom-1) because the shape of the pollen grains resembled a Portuguese candy, namely, kompeito or confeito, in their scanning electron microscopy images ( Figure 1C,D). Transmission electron microscopy (TEM) analyses of pollen exines revealed that the statue-like structures, i.e., tecta ( Figure 1G, arrowheads), were absent from the surface of the kom-1 pollen grains ( Figure 1H). When wild-type (WT) Col-0 pistils were hand-pollinated with WT pollen and washed 10 s after pollination, 71.3% of pollen grains remained on the stigma. In contrast, when WT pistils pollinated with kom-1 pollen grains were washed 10 s after pollination, only 16.9% of pollen grains remained ( Figure 1I). When the same washing experiment was performed 1 h after pollination, which is enough time for pollen tubes to germinate and penetrate inside the stigma, 89.0% of WT and 83.7% of kom-1 pollen grains remained ( Figure 1I). These data suggested that it was difficult for kom-1 pollen grains to be retained on the stigma surface, but once retained, they could germinate normally.
were washed 10 s after pollination, only 16.9% of pollen grains remained ( Figure 1I). When the same washing experiment was performed 1 h after pollination, which is enough time for pollen tubes to germinate and penetrate inside the stigma, 89.0% of WT and 83.7% of kom-1 pollen grains remained ( Figure 1I). These data suggested that it was difficult for kom-1 pollen grains to be retained on the stigma surface, but once retained, they could germinate normally. The vegetative parts of the kom-1 mutant were indistinguishable from the WT plants, and its female gametophytes developed normally (Supplemental Figure S1A,B). Despite The vegetative parts of the kom-1 mutant were indistinguishable from the WT plants, and its female gametophytes developed normally (Supplemental Figure S1A,B). Despite the abnormality of the pollen surface, when the kom-1 pollen grains were pollinated onto the WT pistil, the pistil produced seeds normally. This suggested that the male kom-1 gametophytes were functional (Supplemental Figure S1C-G). When the kom-1 pollen grains or pistils were crossed with the WT plants, all of the F1 progeny had the normal pollen phenotype and no mutant plants were segregated. When the F1 plant was self-pollinated, 24.7% (24 out of 97) of the F2 plants showed a kom-1 mutant phenotype. These data indicated that kom-1 was a recessive, sporophytic, but not gametophytic mutant, and that the wall patterning of the pollen grains was important for pollen-stigma adhesion.

Defective Callose Wall Development in Kom-1
We analyzed the composition and structure of the callose wall in developing kom-1 anthers stained with aniline blue, which is a callose indicator ( Figure 2). Pollen mother cells and microspore tetrads of the WT plants were enclosed with a thick wall of callose ( Figure 2B,C). No callose was stained in the cell wall surrounding the kom-1 pollen mother cells or tetrads ( Figure 2E); however, a moderate amount of callose was observed in the walls separating the kom-1 microspores ( Figure 2F). TEM analyses of the callose wall ultrastructure further confirmed that the peripheral wall surrounding the kom-1 tetrad was thin and lacked callose ( Figure 2J,K) compared with the thick wall of the WT ( Figure 2G,H). TEM images also showed that the walls within and between the microspores were thinner and less electron-dense ( Figure 2K). The pattern of sporopollenin aggregation observed at the tetrad stage in the WT ( Figure 2I) did not form in kom-1 ( Figure 2L). These results suggested that KOM was responsible for the synthesis of the callose wall surrounding the tetrad during meiosis, but was not absolutely necessary for the callose wall that was synthesized in the tetrad stage and secreted between microspores.

Mapping of the Kom-1 Locus and Complementation Using the KOM Genomic DNA
To isolate the gene responsible for the recessive kom-1 mutation, we mapped the locus using polymerase chain reaction (PCR)-based polymorphism detection methods with DNA samples prepared from F2 plants of a cross between kom-1 (Col-0 ecotype) and Landsberg erecta (Ler). A single locus at the bottom of chromosome 1 was linked to the kom-1 phenotype ( Figure 3A). After fine mapping, the kom-1 locus was narrowed down to a 12.5 kb region of BAC F28K19, which contained four open-reading frames. We Since female meiosis also accompanies the formation of callose walls, we wondered whether megasporogenesis was affected by the kom-1 mutation. The amount of cal-lose in the callose walls separating the megaspore mother cells was reduced in kom-1 (Supplementary Figure S2D-F) compared with that of the wildtype, which had thick callose walls (Supplementary Figure S2A-C). These results suggested that KOM was also important for callose accumulation during megasporogenesis. Surprisingly, female kom-1 gametophytes developed to become indistinguishable from those of the WT (Supplementary Figure S1A,B) and were fertile (Supplementary Figure S1G). One possible explanation is that three of the four megaspore mother cells present after meiosis undergo apoptosis during megasporogenesis; thus, callose reduction may not produce defects in female gametogenesis.

Mapping of the Kom-1 Locus and Complementation Using the KOM Genomic DNA
To isolate the gene responsible for the recessive kom-1 mutation, we mapped the locus using polymerase chain reaction (PCR)-based polymorphism detection methods with DNA samples prepared from F2 plants of a cross between kom-1 (Col-0 ecotype) and Landsberg erecta (Ler). A single locus at the bottom of chromosome 1 was linked to the kom-1 phenotype ( Figure 3A). After fine mapping, the kom-1 locus was narrowed down to a 12.5 kb region of BAC F28K19, which contained four open-reading frames. We sequenced the genomic DNA of this region from the kom-1 plant and found only one G-to-A substitution in the At1g77860 gene. We cloned the full-length cDNAs of At1g77860 from WT plants and the kom-1 mutant. A comparison of the genomic and cDNA sequences revealed that the mutation site was in the splicing acceptor (AG) of the fourth intron, which disabled normal splicing of the KOM gene ( Figure 3B). In kom-1 plants, an alternative splicing site 23 bp downward from the original site was used as a new splice acceptor, resulting in a frameshift that produced a truncated protein with 16 novel amino acids at its C-terminal end. The resulting kom-1 mRNA encoded a truncated protein that was approximately 70 amino acids shorter at the C terminus ( Figure 3C). This observation indicated that the C-terminus of KOM was essential for its biological function.
We obtained two additional lines of evidence to support the conclusion that At1g77860 encodes the KOM gene. First, we performed a complementation test of the kom-1 mutation using a 3.6 kb genomic fragment that spanned the KOM gene, including the promoter region and 3' element. The transgenic plants were restored to WT levels of callose accumulation and pollen wall patterning ( Figure 3E-H). Second, a similar phenotype was observed in an allelic mutant, namely, SALK_016980 (kom-2), which contained a T-DNA insertion in the fourth exon. The F1 plants of the cross between kom-1 and kom-2 did not complement the mutant phenotype (data not shown), suggesting that kom-1 and kom-2 were allelic and loss-of-function mutations. the promoter region and 3' element. The transgenic plants were restored to WT levels of callose accumulation and pollen wall patterning ( Figure 3E-H). Second, a similar phenotype was observed in an allelic mutant, namely, SALK_016980 (kom-2), which contained a T-DNA insertion in the fourth exon. The F1 plants of the cross between kom-1 and kom-2 did not complement the mutant phenotype (data not shown), suggesting that kom-1 and kom-2 were allelic and loss-of-function mutations.

KOM Was Found to Be a Unique Member of the Rhomboid Family of Proteins
The deduced KOM peptide encoded 385 amino acid residues with a signal anchor domain at its N-terminus and 7 predicted transmembrane domains; it belonged to the Rhomboid family of proteins ( Figure 3D), which are conserved in bacteria, animals and plants [30][31][32][33][34][35][36][37]. Unlike many members of the Rhomboid family [31,38], KOM did not contain a conserved His residue at TMD6, which, in all other Rhomboids, is required for the serine protease activity. In addition, KOM had a relatively long tail at the C-terminus [31]. Phylogenetic analysis showed that KOM is a distantly-related member of Arabidopsis Rhomboid proteins (Supplementary Figure S3).

Expression of KOM-1 Was Restricted to the Meiocytes
Reverse-transcription PCR analysis showed that KOM was mainly expressed in flowers ( Figure 4A, lane 6), as well as in the roots and rosette leaves at low levels ( Figure 4A, lanes  2 and 3). This expression pattern was also supported by publicly-available RNAseq data (Supplementary Figure S4). Results of in situ RNA hybridization of floral tissues showed that KOM mRNA expression was restricted to pollen mother cells that had undergone meiosis ( Figure 4C). KOM expression was not detectable in tapetum cells ( Figure 4B) or tetrads ( Figure 4D). KOM mRNA was also present in megaspore mother cells under meiosis (Supplementary Figure S2G-I). This temporal and spatial expression pattern of the KOM gene was consistent with its function in the regulation of callose deposition to the callose wall.

KOM Was Localized in the Golgi Apparatus
Most Rhomboid-like proteins are distributed through the secretory pathway [39], although some members act in mitochondria [33,40]. Arabidopsis Rhomboid-like proteins were also experimentally shown or predicted to localize several subcellular organelles [31,41,42]. To investigate the subcellular localization of KOM, we expressed the KOM-GFP fusion protein transiently in Arabidopsis protoplasts and in stable transgenic plants ( Figure 5). The fusion protein was observed as dot-like structures and most of the signals were co-localized with fluorescence signals of RGP1 ( Figure 5A-F), which is a trans-Golgi marker [43], suggesting that KOM was localized in the Golgi apparatus in plant cells.

KOM Was Localized in the Golgi Apparatus
Most Rhomboid-like proteins are distributed through the secretory pathway [39], although some members act in mitochondria [33,40]. Arabidopsis Rhomboid-like proteins were also experimentally shown or predicted to localize several subcellular organelles [31,41,42]. To investigate the subcellular localization of KOM, we expressed the KOM-GFP fusion protein transiently in Arabidopsis protoplasts and in stable transgenic plants ( Figure 5). The fusion protein was observed as dot-like structures and most of the signals were co-localized with fluorescence signals of RGP1 ( Figure 5A-F), which is a trans-Golgi marker [43], suggesting that KOM was localized in the Golgi apparatus in plant cells.

KOM Was Localized in the Golgi Apparatus
Most Rhomboid-like proteins are distributed through the secretory pathway [39], although some members act in mitochondria [33,40]. Arabidopsis Rhomboid-like proteins were also experimentally shown or predicted to localize several subcellular organelles [31,41,42]. To investigate the subcellular localization of KOM, we expressed the KOM-GFP fusion protein transiently in Arabidopsis protoplasts and in stable transgenic plants ( Figure 5). The fusion protein was observed as dot-like structures and most of the signals were co-localized with fluorescence signals of RGP1 ( Figure 5A-F), which is a trans-Golgi marker [43], suggesting that KOM was localized in the Golgi apparatus in plant cells.

KOM Was Required for the Accumulation of CalS5 during Pollen Development
In the Rhomboid family of proteins, several key amino acid residues are conserved throughout organisms [30]. Mutations at the Asn of TMD2, Ser of TMD4 and His of TMD6 abolish the protease activity of Rhomboids in vivo [44][45][46]. Of the three essential residues for the protease activity, the His of TMD6 was not conserved in KOM but was replaced by Asn ( Figure 3D). In addition, two Ser residues were located in tandem in TMD4 in KOM. We investigated whether these residues are required for the biological function of KOM. First, we performed genetic complementation tests of the kom-1 mutation using KOM genomic DNA (Supplementary Figure S3A-F). In this experiment, the Asn of TMD2, His of TMD6, the first Ser of TMD4, second Ser of TMD4 or both Sers of TMD4 were replaced by Ala (N141A, N241A, S187A, S188A and S187A/S188A, respectively). Unexpectedly, all mutated alleles were able to complement the kom-1 mutation (Supplementary Figure S5B-F) and restored the aberrant pollen wall patterning of kom-1 to the WT phenotype (Supplementary Figure S5A). This implied that KOM did not require proteolytic activity for its normal function. Consistent with this, in a standard Rhomboid activity assay, which we used previously to demonstrate the proteolytic activity of Arabidopsis Rhomboid AtRBL2 [31,46], KOM showed no proteolytic activity against classical Rhomboid substrates, including Spitz and Keren of Drosophila (Supplementary Figure S5H,I) and human TGFβ (data not shown). These findings suggest that the role of KOM in callose regulation did not depend on Rhomboid-like proteolytic activity.
The male sterility and callose wall abnormality observed in kom-1 pollen grains closely resembled the phenotype of cals5, which is a mutant lacking the function of male-specific CasS5 [15,16]. To investigate a genetic interaction between KOM and CalS5, we crossed kom-1 with a cals5 mutant. Among the mutant alleles, cals5-5, which has a T-DNA insertion in the third intron, showed a weaker phenotype in callose accumulation and pollen wall patterning [16]. When we crossed kom-1 with cals5-5, the kom-1/cals5-5 double mutant showed disorganized pollen wall patterning, which was similar to those of the severe pollen phenotype of kom-1 ( Figure 6A-C). This result suggested that KOM was epistatic to CalS5 and raised the possibility that KOM may regulate the function of CalS5. It was reported that CalS5 is expressed in the tapetum cells and meiosis to tetrad stages of microspore [17]. In our experimental condition, CalS5 signals were strongly detected in the microspores and weakly detected in the tapetum cells, both in WT and kom-1 ( Figure 6D-F), suggesting that KOM did not affect the CalS5 expression.  (D-F) Analysis of CalS5 expression in anthers using in situ mRNA hybridization. CalS5 mRNA was expressed in pollen mother cells undergoing meiosis in the WT (D), as well as in kom-1 (E). Sense mRNA of CalS5 was used as the negative control (F). (G) Western blot analysis of proteins from floral buds using polyclonal antibodies against CalS5-N and CalS5-L. CalS5 was detected as a 220 kDa protein in the WT flower buds, but it was less prevalent in kom-1 (kom) and absent in the cals5-5 (cs5) null allele. Bars: 20 µm.
To determine whether KOM is required for the function of the CalS5 protein, we attempted to detect CalS5 protein in WT and mutant plants. We raised polyclonal antibodies against the N-terminal region of CalS5 (anti-CS5N) and its central loop region (anti-CS5L) for Western blot analysis. Both anti-CS5N and anti-CS5L detected a 220 kDa band in protein extracted from the WT membrane fraction, which corresponded to the molecular size of CalS5 protein. In contrast, no such band was detected in protein extracted from cals5, and a faint band was detected in those from kom-1 ( Figure 6G). Together, these findings indicated that KOM was required for CalS5 protein accumulation and had an essential function in the formation of the callose wall during pollen development.

Pollen Wall Patterning Was Required for Pollen-Stigma Adhesion
Pollen-stigma adhesion is the first step of cell-cell interactions during fertilization in plants. Several factors, including pollen wall patterns, pollen coat components and stigmatic proteins, are involved in the pollen-stigma adhesion event [1,16,47,48]. Among these, pollen wall patterning is thought to play a major role in the initial pollen-stigma adhesion [1]. The kom mutation affects pollen sterility, but both female and male gametophytes can function normally. Data obtained from pollen-stigma adhesion assays showed that male sterility is caused by the weak adhesion of kom-1 pollen grains to the stigma ( Figure 1I). The pollen grains of kom-1 contained apparently normal pollen coats but aberrant pollen wall patterning ( Figure 1D,F). Therefore, the pollen wall patterning played a principal role in pollen-stigma adhesion and male fertility observed in the kom mutant.
Another class of pollen wall mutant comprises mutations that affect the synthesis of the callose wall, which provides a mold for pollen wall patterning. Mutations in the CalS5 gene abolish the callose wall, causing male sterility [15,16,28]. In plants with cals11/cals12 double mutations, the callose wall enclosing the pollen mother cells appears normal, but the wall separating tetrad microspores cannot be formed normally due to a lack of callose accumulation [12]. Transgenic plants expressing callase (β-1,3-glucanase) in the anther locule are unable to synthesize the callose wall during microsporogenesis, resulting in the production of pollen grains with aberrant wall patterning [23,24]. In the present study, kom-1 lacked the callose necessary to enclose the pollen mother cells (Figure 2E,F). Because KOM shares no similarity with glucan synthases or glucanases, KOM represents a new class of proteins that modulate the activity of callose synthesis/degradation enzymes.

KOM May Represent a New Function for Rhomboid-like Protein
It was reported that CDKG1, which encodes a cyclin-dependent kinase, is associated with a spliceosome to regulate CalS5 mRNA splicing and protein maturation [28]. We did not observe a splicing defect of CalS5 in kom-1 mutant but observed a CalS5 protein accumulation defect ( Figure 6G). Therefore, KOM may regulate CalS5 activity in a different way than CDKG1.
Since the role of Rhomboid proteins was first demonstrated in the Drosophila EGF signaling pathway [46,53], the most characterized function of Rhomboid proteases conserved through organisms, including bacteria, animals and plants, is the proteolytic cleavage of membrane proteins [30,31,54]. Later, this family of proteins was implicated in diverse cellular processes, including protein homeostasis, viral susceptibility, mitochondria membrane fusion and parasite-host interaction [34,39,40,[55][56][57][58][59][60]. The genetic and biochemical data presented in this study suggest that KOM acts not as a Rhomboid protease but as a member of the growing sub-family of proteolytically inactive Rhomboid-like proteins (Supplementary Figure S3) [61]. These pseudoproteases have a very wide range of biological functions, although current evidence suggests that a common mechanistic theme is that they interact specifically with TMDs, thereby regulating membrane proteins, often by affecting their stability. Although KOM is unlikely to have proteolytic activity itself, the loss of KOM showed the clear phenotype of CalS5 with callose accumulation. This is the first evidence that catalytically inactive Rhomboid-like protein has a function in plants.
Interaction with EGF ligands was characterized in catalytically inactive Rhomboid proteins [62]. Interestingly, the C-terminal lumenal domain of KOM is longer than those of other Arabidopsis Rhomboid-like proteins that show detectable serine protease activity against classical Rhomboid substrates from Drosophila [31]. Because CalS5 is a membrane protein and its accumulation is reduced in kom-1, KOM may act, for example, to stabilize CalS5 on a membranous component in meiocyte (Supplementary Figure S6). Further characterization of KOM will provide new insights into the regulatory mechanisms of the Rhomboid family of proteins, callose accumulation and microspore development in plants.

Plant Material and Growth Conditions
Columbia-0 (Col-0) accession of Arabidopsis thaliana was used as wild-type plants (WT). The kom-1 mutant was originally isolated from an EMS-mutagenized M2 population of Col-0. The mutant line used in this study was generated after two backcrosses to Col-0. T-DNA insertion mutants, namely, SALK_016980 (kom-2) (this study) and SALK_072226 (cals5-5) [16], were obtained from the Arabidopsis Biological Resource Center at Ohio State University (Columbus, OH, USA). The hemizygous kom-1/kom-2 plant showed the same phenotypes as kom-1 and kom-2 homozygous plants. All the qualitative data presented in this manuscript are representative of the phenotypes of these two alleles.
Seeds were sown on the surface of vermiculite in small pots and incubated for four days at 4 • C. Plants were grown under continuous white light at 22-24 • C.

Microscopy
For transmission electron microscopy (TEM), flowers were fixed in a 50 mM sodium cacodylate buffer containing 4% paraformaldehyde (Sigma-Aldrich, Tokyo, Japan) and 1% glutaraldehyde (Sigma-Aldrich, Tokyo, Japan) for 12 h. Samples were kept under gentle vacuum conditions for the first hour at room temperature and were then maintained at 4 • C. After fixation, samples were washed with 50 mM sodium cacodylate buffer twice and postfixed in 50 mM sodium cacodylate buffer containing 1% osmium tetroxide (Sigma-Aldrich, Tokyo, Japan) for three hours at 4 • C. After washing with 50 mM sodium cacodylate buffer twice and distilled water once, they were dehydrated with a graded ethanol series and kept in 100% ethanol overnight at 4 • C. They were subsequently treated with propylene oxide (Sigma-Aldrich, Tokyo, Japan) containing 5,10,15,20,30,50 and 80% Spurr's resin and then maintained in 100% Spurr's resin (Funakoshi, Tokyo, Japan) overnight at 4 • C. Anthers were then dissected and embedded in Spurr's resin. Ultra-thin (60-80 nm) sections were cut with a diamond knife, stained with uranyl acetate and lead citrate (Sigma-Aldrich, Tokyo, Japan), and observed with an SM-100SX transmission electron microscope (JEOL, Tokyo, Japan). For the pollen observation, flowers were fixed in a 9:1 ratio of ethanol to acetic acid overnight, treated with 90% and 70% ethanol for 20 min each, cleared in Hoyer's solution (7.5 g gum Arabic (Sigma-Aldrich, Tokyo, Japan), 100 g chloral hydrate (Tokyo-Kasei, Tokyo, Japan), 5 mL glycerol (Sigma-Aldrich, Tokyo, Japan) in 30 mL water) and observed with Nomarski optics. For histological analyses, inflorescences were fixed in FAA solution (formaldehyde/acetic acid/alcohol) overnight, replaced with 50, 70, 80, 90, 99 and 100% ethanol, and embedded in Technovit 7100 resin (Kulzer, Heraeus, Germany). Sections that were 5 µm thick were double-stained with 0.1% Toluidine blue (Cosmo-Bio, Tokyo, Japan) and 1% aniline blue (Chroma, Münster, Germany) and observed under UV illumination.

Pollen-Stigma Adhesion Assay
Col-0 or kom-1 stamens were immobilized on a table with double-sided tape. Col-0 pistils that had been emasculated to prevent self-pollination were carefully touched to the stamens. Pollinated pistils were vortexed in a buffer and the numbers of pollen grains detached into the buffer and remained on the stigma were counted 10 s or 1 h after pollination. Average values were obtained from five individual assays.

Mapping and Cloning of the KOM Gene
Approximately 4000 F2 plants from a cross between kom-1 and Ler were used for mapping the KOM locus. DNA markers used for positional cloning were based on an SSLP (simple sequence length polymorphism) between ecotypes Col-0 and Ler. Information about T5M6-1, F28K19-1, F28K19-2 and F28K19-3 markers was obtained from the MON-SANTO Arabidopsis Polymorphism and Ler Sequence Collection (http://www.Arabidopsis. org/Cereon/index.jsp (accessed on 28 March 2022)).

Complementation Test
A 3.6 kb genomic fragment spanning the KOM gene, corresponding to region 24824-28559 of the BAC clone F28K19, was amplified using PCR (primer sequences 5 -GAGCAATGATTTT CTCCTTGAGAGATGC-3 and 5 -GAGATACAACTCTTCGGAAAGG-3 ) and cloned in pBluescript II SK (+) (Toyobo, Japan). This genomic fragment included 664 bp of the 5 region, 1042 bp of the 3 region and the whole ORF of the At1g77860 gene. The fragment was digested with Kpn I and cloned into binary vector pPZP211 [63] to generate pPZP211-KOM. Agrobacterium tumefaciens C58C1 containing pPZP211-KOM was used to transform kom-1 plants using a vacuum infiltration procedure. Transgenic plants were selected on an agar medium containing 30 µg/mL kanamycin (Sigma-Aldrich, Tokyo, Japan) and 100 µg/mL carbenicillin (Sigma-Aldrich, Tokyo, Japan). Pollen phenotypes were examined for complementation.
For the cite-directed mutagenesis of KOM genes, the 3.6 kb KOM genomic DNA was amplified via PCR (using ATGTCGAATTCATCTGCGAGTCTGAACTTC and CAACTGTCG ACAAAACAGA-GTAGGTCTTCG) and cloned into pBluescript II SK (+) at EcoR I-Sal I site. This vector was used as a template of the PCR to generate site-directed mutation alleles of N141A (using TTTCATCTATTCATAGCTCTTGGGAGTTTG and TAATC-CACTGTGCAGCCATGGAG), N241A (CTTTGCAGCTATTGGTGGTTTCATATCAGG and TTGTCTATGAAAGGG-AGAAAGCCTA), S187A (CATCAATCGCTTCTGGTGCTGC and GGATGTTCCGAACA-AACAACACAGC), S188A (CATCAATCTCTGCTGGTGCTGC and GGATGTTCCGAAC-AAACAACACAGC) and S187A/S188A (CATCAATCGCTGCTG-GTGCTGC and GGATGTTCCGAACAAACAACACAGC). PCR products containing the mutated KOM genes were self-ligated and cloned into the binary vector pPZP211 at the EcoR I-Sal I site. The constructs were used to transform kom-1 plants via the Agrobacterium tumefaciens C58C1 strain-mediated floral dip approach. Transgenic plants were selected on an agar medium containing 30 µg/mL kanamycin and 100 µg/mL carbenicillin. Pollen phenotypes were examined for complementation in at least three independent lines.

cDNA Cloning and RT-PCR
Total RNA isolation and first-strand cDNAs synthesis were done as described previously [31]. KOM cDNA was obtained by both 5' and 3' RACE using the SMART RACE cDNA Amplification Kit (TaKaRa, Tokyo, Japan). The accession number of KOM cDNA is AB161192.
For RT-PCR, cDNAs were synthesized from 0.9 µg of total RNA with the SuperScript II First-Strand Synthesis System (Thermo Fisher Japan, Tokyo, Japan). The tissues used were the aerial parts from seedlings 10 days after germination, roots from seedlings 10 days after germination, rosette leaves, cauline leaves, stems, siliques and inflorescences. KOM-specific primers (CATTATGGGTATGCTGTTTGC and CTTGTGAGATATTGGTGAAAGG) and actin (ACT8)-specific primers (An et al., 1996) were used to amplify KOM and ACT8 cDNAs using PCR for 35 cycles.

In Situ RNA Hybridization
Expression of KOM and CalS5 in developing flowers was detected using in situ mRNA hybridization as described previously [65,66]. Inflorescences were fixed with 4% paraformaldehyde in PBS. Paraffin sections (8 µm thick) were hybridized with digoxygenin-labeled probes. Antisense and sense probes of KOM and CalS5 were prepared using fragments of their cDNAs that had been amplified using PCR and cloned into pBluescript II SK (+) at the Sal I site. The probe sequences used for hybridization corresponded to KOM cDNA between 59 and 569 bp and CalS5 cDNA between 4864 and 5414 bp.

Subcellular Localization Analysis
To make GFP-tagged constructs, a PCR fragment of G3GFP was used to replace the GUS gene of a binary vector pBI121 to generate p35SG3GFP. KOM cDNA was inserted between the CaMV 35S promoter and G3GFP at an Xba I site to generate p35S-KOM-GFP. For transient expression in Arabidopsis protoplasts, the Hind III-EcoR I fragment of p35S-KOM-GFP, which contained the CaMV 35S promoter, KOM-GFP and NOS-terminator, was cloned into pBluescript II SK (+). Transformation of Arabidopsis Col-0 suspension culture cells [67] was performed as described previously [68]. Transformed protoplasts were incubated under gentle agitation at 23 • C for at least 8 h in the dark. Immunofluorescent staining was performed as described [69,70]. A 500-fold dilution of anti-RGP1 antibodies [43] was used to stain the RGP1 protein, which is a Golgi marker. Alexa-Fluor TM -546-conjugated goat antibodies against rabbit IgG (100-fold dilution; Molecular Probes, Eugene, OR, USA) were used as secondary antibodies. Subcellular localization of proteins was observed with a confocal laser microscope system (LSM510, ZEISS, Jena, Germany) with the 488 nm line of an Ar/Kr laser for GFP and the 543 nm line of a He/Ne laser for Alexa Fluor TM 546. The transient expression assay was repeated more than five times to confirm the protein localization.

Western Blotting
Inflorescences (1.0 g) from Col-0, kom-1 and cals5-5 were ground to a fine powder in liquid nitrogen. Powders of the samples were homogenized in extraction buffer with proteinase inhibitor (20 mM Tris-HCl, pH 7.5, 50 mM NaCl, 1 mM EDTA, 1 mM 2-mercaptoethanol, 0.5 mM phenylmethylsulfonyl fluoride (Signa-Aldrich, St. Luis, MO, USA)). The homogenates were centrifuged at 100,000× g for 60 min to separate the soluble and total membrane fractions. The same amount of proteins in each sample was dissolved with an SDS sample loading buffer and resolved on SDS-PAGE. Western blot analysis was performed using an anti-CalS5-N antibody (corresponding to CalS5 amino acid position 80-180) and anti-CalS5-L (corresponding to amino acid positions 898-1025).

Conclusions
Here, we present genetic and molecular data showing that mutation in the KOMPEITO gene caused pollen wall morphology and pollen-stigma adhesion. KOM falls within a subclass of Rhomboid-like proteins that have lost protease activity. KOM was predominantly expressed in the meiocyte and loss of KOM expression affected CalS5 protein accumulation. Taken together, KOMPEITO was required for proper callose wall formation via affecting CalS5 protein function in the meiocyte, as well as for pollen wall formation and plant fertilization.