The zinc finger protein DCM1 is required for male meiotic cytokinesis by preserving callose in rice

Meiotic cytokinesis influences the fertility and ploidy of gametes. However, limited information is available on the genetic control of meiotic cytokinesis in plants. Here, we identified a rice mutant with low male fertility, defective callose in meiosis 1 (dcm1). The pollen grains of dcm1 are proved to be defective in exine formation. Meiotic cytokinesis is disrupted in dcm1, resulting in disordered spindle orientation during meiosis II and formation of pollen grains with varied size and DNA content. We demonstrated that meiotic cytokinesis defect in dcm1 is caused by prematurely dissolution of callosic plates. Furthermore, peripheral callose surrounding the dcm1 pollen mother cells (PMCs) also disappeared untimely around pachytene. The DCM1 protein contains five tandem CCCH motifs and interacts with nuclear poly (A) binding proteins (PABNs) in nuclear speckles. The expression profiles of genes related to callose synthesis and degradation are significantly modified in dcm1. Together, we propose that DCM1 plays an essential role in male meiotic cytokinesis by preserving callose from prematurely dissolution in rice.


Author summary
Meiosis comprises two successive cell divisions after a single S phase, generating four haploid products. Meiotic caryokinesis (nuclear division) has been extensively studied in many organisms, while mechanisms underlying meiotic cytokinesis remain elusive. Here, we identified a novel CCCH-tandem zinc finger protein DCM1 that prevent the premature dissolution of callose both around the PMCs and at the dividing site (callosic plates). Loss of the callosic plates disrupts the meiotic cytokinesis, leading to the random distribution of spindles during meiosis II and aberrant meiotic products. DCM1 interacts with the two rice poly (A) binding proteins, independently of the conserved CCCH domain. Moreover, DCM1 coordinates the expression profiles of genes related to callose synthesis and degradation. We suspect monocots and dicots may adopt distinct meiotic cytokinesis patterns during male gamete generation. Introduction Cytokinesis is the process by which the two daughter nuclei resulting from nuclear division are physically separated by the establishment of a cell plate and/or cell wall. In animal and yeast dividing cells, an actomyosin ring contracts centripetally to separate the daughter cells [1]. In higher plants, however, cytokinesis involves the formation of a cell plate through the fusion of vesicles at its centrifugally expanding periphery [2,3]. Meiosis is a specialized type of cell division consisting of one round of DNA replication and two rounds of nuclear division [4]. In plant male meiosis, two different types of cell plate formation, namely successive and simultaneous cytokinesis, are documented. Each caryokinesis is directly followed by a cytokinetic event in successive cytokinesis (typically in monocots), while the simultaneous cytokinesis occurs only when both nuclear division are finalized (typically in dicots) [5]. Callose, a β-1, 3-glucan polymer with β-1, 6-branches, plays important roles in response to biotic and abiotic stresses, as well as in a variety of developmental processes, especially in cell plate formation and reproductive development in plants [6,7]. During mitosis, callose is deposited at the cell plate during the tubular network stage and is later replaced by cellulose [8]. Mutation of MASSUE/AtGSL8 in Arabidopsis, which encodes a putative callose synthase, leads to seedling lethality and a striking cytokinesis-defective phenotype [9,10]. Moreover, enlarged tetrad 2, which harbors a splice site mutation of AtGSL8, undergoes premeiotic endomitosis due to cytokinetic defects in flowers [11]. These findings indicate that callose is required for plant mitotic cytokinesis.
For pollen mother cells (PMCs), callose is placed both at the division site as well as at the outer cell wall. The peripheral callose may prevent PMCs fusion and cohesion, and appear to participate in the formation of the primexine by providing a mold for pollen exine construction during microsporogenesis [12]. Callose synthase 5 (CalS5), also known as glucan synthase-like 2 (AtGSL2), is responsible for the synthesis of callose deposited at the primary cell wall of PMCs, tetrads and microspores, and is essential for exine formation of pollen wall in Arabidopsis [12,13]. Mutation of OsGSL5, the rice homolog of AtGSL2, also results in defective callose deposition and abnormal pollen exine structure formation [14]. AtGSL1 and AtGSL5, two closely related genes in Arabidopsis, are required for synthesis of the interstitial callose that normally separates microspores. The enlarged pollen grains and multi-nucleate microspores in gsl1-1/+ gsl5-2/gsl5-3 suggest that callose is required for meiotic cytokinesis [15]. In Arabidopsis mpk4, which has a male-specific meiotic cytokinesis defect, PMCs fail to form normal intersporal callose walls after meiosis, and thus cannot complete meiotic cytokinesis [16]. Moreover, TETRASPORE/STUD, a kinesin required for male meiotic cytokinesis in Arabidopsis [17,18], was reported to function upstream of MPK4 in a putative cascade pathway.
CCCH zinc finger proteins are characterized by a zinc finger motif consisting of three cysteines and one histidine coordinated by a zinc cation and have been identified in Arabidopsis, rice [19] and other plant species [20][21][22]. Recent studies have revealed that CCCH proteins participate in the regulation of many developmental processes and environmental responses [23,24]. Although the molecular functions have not been fully characterized, multiple lines of evidence suggest that proteins with CCCH motifs can regulate gene expression through modulation of RNA metabolism. For example, OsTZF1 binds to U-rich sequences in the 3'UTR of two potential target mRNAs in vitro [25], and AtTZF1 can trigger the degradation of AU-rich elements (ARE)-containing mRNA in vivo [26].
Here, we report a novel protein, DCM1 (Defective Callose in Meiosis 1), is required for meiotic cytokinesis in rice. The meiotic cytokinesis defect is caused by the premature dissolution of callosic plates during both telophase I and telophase II. DCM1 interacts with two nuclear poly (A) binding proteins, which are members of the polyadenylation factor family, independently of the conserved tandem CCCH domain. Our study provides insight into the mechanism underlying meiotic cytokinesis in monocots.

Identification of a mutant with dramatically reduced fertility
We isolated the defective callose in meiosis 1 (dcm1) mutant in a screen of our rice mutant libraries with reduced fertility. The dcm1 mutant exhibited no defects in vegetative growth but was nearly sterile with panicles occasionally bearing a few seeds (Fig 1A). When stained with I 2 -KI, most pollen grains (85.1%, n = 766) were less stained or even shrunken compared with wild-type pollens, indicating the dramatically reduced viability of male gametes. The viability of dcm1 female gametes was evaluated by pollinating its flowers with wild-type pollen grains. The resulting normal seed setting rate indicated that female fertility was not affected in dcm1 (S1 Table).
To investigate the cellular defects of dcm1, we examined transverse sections of both wildtype and mutant anthers at different developmental stages ( Fig 1B). During early meiosis, four layers of anther somatic cells (epidermis, endothecium, middle layer and tapetum) enclosed the anther locule, where PMCs contacted with the tapetal layer. After meiosis, the PMCs gave rise to microspores. Among these stages, no significant differences were observed between wild-type and dcm1 anthers. During the bicellular pollen stage, wild-type pollens were vacuolated with an increased volume and tapetal cells were deeply stained with toluidine blue. By contrast, tapetal cells of dcm1 were larger than normal with reduced staining and started to degenerate at this stage. During the mature pollen stage, all wild-type pollen grains were round and the inner layers of the anther degenerated. However, besides round pollen grains, shrunken empty pollen grains (75%, n = 56) were observed in dcm1. In addition, the middle layer, which was not visible in mature wild-type anther, was swollen in the mutant.
To obtain a more detailed understanding of the abnormalities of the dcm1 pollen grains, mature pollen was visualized using transmission electron microscopy (TEM) (Fig 1C). Ubisch bodies, specialized structures forming along the inner surface of the tapetum, seemed to be normal in dcm1. The wild-type pollen grains had completed pollen exine deposition and exhibited distinctive sublayers, including sexine and nexine. However, only one layer, which seemed to be uniform in component, was observed in dcm1 pollen grains (n = 17), indicating that the formation of pollen exine was disrupted in dcm1.

Abnormal meiotic products give rise to pollen grains of varied size in dcm1
The size of the pollen grains of dcm1 seemed to be nonuniform ( Fig 1A and Fig 1B). To confirm this, we observed the mature pollen grains of wild type and dcm1 using scanning electron microscope (SEM). The wild-type pollen grains were round and uniform in size (n = 182). However, the pollen grains of the mutant varied in size, with some pollen grains (41.7%, n = 206) obviously larger than the wild type (Fig 2A).
As the pollen size is always correlated with its ploidy [27], the nuclear DNA of pollen grains was stained with 4', 6-diamino-phenylindole (DAPI) (Fig 2B). In the 238 checked pollen grains of dcm1, 153 normal-sized dcm1 pollen grains (64.3%) had a wild-type-like nuclei configuration and staining intensity. However, 78 enlarged pollen grains (32.8%) with more intensively stained and enlarged nuclei were observed in dcm1, suggesting increased gametophytic DNA content in the mutant. Doubled nuclei were also observed in seven larger pollen grains (2.9%) among the 238 checked pollen grains of dcm1.
PMCs give rise to a group of four haploid spores through meiosis. Each uninucleate microspore undergoes an asymmetric mitotic division to produce the vegetative cell and the generative cell. The generative nucleus completes the second mitotic division, developing into trinucleate pollen. To clarify the time when the defect of dcm1 occurred, before or after the microspore stage, we observed the microspores of both wild type and dcm1 (S1 Fig). Most microspores (98%, n = 200) of wild type were uniform in size, with the diameter ranging from 17 μm to 23 μm. However, the sizes of microspores in dcm1 were variable (n = 277), ranging from 15 μm to 47 μm.
We then checked the meiotic products of wild type and dcm1 by acetocarmine dyeing ( Fig  2C). In wild-type tetrad, four microspores were held together in a tetragonal shape. However, the arrangement of meiotic products in dcm1 varied. In 95 checked PMCs of dcm1, 18 PMCs (18.9%) divided into two daughter cells and each cell contained two nuclei. 43 PMCs (45.3%) divided into three daughter cells with one nucleus in each small cell and two nuclei in the big one. In addition, four daughter cells, each of which had one nucleus (14.7%), and one daughter cell with four nuclei (21.1%) were also observed in dcm1.
Spindle organization during telophase II was investigated by performing immunolocalization with α-tubulin antibody ( Fig 2C). In wild-type PMCs, two sets of spindles were roughly parallel to each other, leading to the formation of four well separated poles at telophase II. However, most checked PMCs (n = 67) in dcm1 had fused spindles (13.4%), tripolar spindles (59.7%) or linear spindles (7.5%). Among them, spindles in seven dcm1 PMCs (10.4%) linked every two nuclei when they were adjacent to each other, which was reminiscent of the spindle arrangement during simultaneous meiotic cytokinesis.

Meiotic cytokinesis is defective in dcm1
The abnormal spindle arrangements observed in dcm1 during telophase II indicated that meiotic cytokinesis might be defective, leaving the spindle physically unseparated and free to move. As a monocot, rice undergoes successive cytokinesis during meiosis, in which the dyad is generated after meiosis I and the tetrad is formed after meiosis II. In the wild-type PMCs, homologous chromosomes segregated and were pulled to opposite poles during meiosis I ( Fig  3). A cell plate was established that insulated the two chromatin groups and cytoplasm at telophase I. The cell plate persisted through metaphase II when the chromosomes aligned at the equatorial plate and anaphase II when the chromatids were pulled toward opposite poles. Two new cell plates, both perpendicular to the first one, were formed during telophase II and tetrads were then produced.
The meiotic stages from metaphase I to anaphase I in dcm1 were roughly the same as those in wild type (Fig 3). However, no cell plate was formed in most PMCs (94.4%, n = 72) during telophase I. The two equatorial plates and spindles were not parallel to each other in dcm1 during metaphase II. During anaphase II, the separating sister chromatids were pulled in varied orientations. Without the cell plate, meiosis II took place in a single cell. We also observed defects in the second meiotic cytokinesis in the mutant, which gave rise to microspores with four nuclei.

Prematurely dissolution of callose occurs in dcm1 PMCs
Previous studies have demonstrated that cell plate formation in successive-type PMCs follows a similar pattern as somatic cytokinesis, in which callose plays an essential roles [5,9,10]. So, we monitored callosic plate formation in wild-type and dcm1 PMCs by combined DAPI and aniline blue staining. Aniline blue binds to callose and emits fluorescence under ultraviolet light. In wild-type anaphase I PMCs, callose was absent between the two separating chromosome mass. When the chromosomes reached the opposite poles during early telophase I, only weak callose signals appeared in the center of the PMCs ( Fig 4A). As chromosomes decondensed into chromatin during telophase I, a callosic plate was formed. At the end of meiosis, two additional callosic plates were formed and four chromatin masses were separated.
In dcm1, aberrance was first observed at telophase I, which corresponds to the time when the callosic plate was formed in wild-type PMCs. Although intact callosic plates were occasionally observed in a few cells (3.6%, n = 138), incomplete callosic plates with an intact periphery and a broken inner plane were always observed ( Fig 4A). During meiosis in rice, callosic plate formation occurs in an outward direction (S2 Fig) [28], indicating that the callosic plates with broken inner planes in dcm1 were in the process of dissolution, not formation. After the complete dissolution of the inner plane, only a callose ring was observed between the chromatin groups. This callose ring, which was presented as two dot-like callose signals in anther transverse sections (88.2%, n = 51), was closely associated with the periphery of the parental PMCs ( Fig 4B). Callosic plates with broken inner planes and callose rings were also observed during telophase II, accompanied by a disordered distribution of nuclei. In wild type PMCs, callosic plates persisted to the tetrad stage. Our observations suggested that prematurely dissolution of callosic plates occurred during both telophase I and telophase II in dcm1.
During rice meiosis, callose is deposited both at the cell plate and around the PMCs. To determine whether the callose surrounding the PMCs was also disrupted in the mutant, we compared the callose pattern in wild type and dcm1 by staining anther transverse sections ( Fig  4B). At preleptotene stage, the very beginning of meiosis, callose was detectable neither in wild type nor in dcm1. In wild-type PMCs, callose first appeared in the center of the locule at leptotene and extended to form an intact callose wall at pachytene. However, despite the normal callose deposition at leptotene, no callose signal around dcm1 PMCs (n = 72) was observed at pachytene and later meiotic stages. These observations indicated that peripheral callose is also under untimely dissolution in dcm1 PMCs.
Immunogold assay using antibody to β-1, 3-glucan showed similar callose defect in dcm1 (Fig 5). In wild-type dyad, callose signals were observed both in cell plate and cell wall ( Fig   Fig 3. Meiotic cytokinesis is defective in dcm1. PMCs of wild type and dcm1 at different developmental stages are shown. Cell plates are indicated by arrows. Bars = 10μm. https://doi.org/10.1371/journal.pgen.1007769.g003 5A). In the PMCs of dcm1 during corresponding stage, cell plate was not formed and callose located in cell wall was almost completely lost (Fig 5B and 5C). Corresponding to the dot-like callose signals in semi-thin transverse sections (Fig 4B), immunogold labeling particles were only observed in the region of cell wall to which the hypothetical cell plate attached in dcm1.

DCM1 encodes a protein with five tandem CCCH motifs
The progenies of a dcm1 heterozygous plant segregated for fertile and sterile phenotypes in a ratio of approximately 3:1 (154:48, χ2 = 0.165, P>0.05), indicating that the mutant phenotype was caused by a single recessive nuclear gene mutation. Through map-based cloning (S3 Fig)  and next-generation sequencing, a nucleotide deletion in the first exon of LOC4341610 was found. We knocked out this gene in wild-type plants using a CRISPR/CAS9 gene editing approach. The resulting transgenic lines displayed defective meiotic cytokinesis, mimicking the phenotype of dcm1 (S4 Fig). This confirmed that the mutant phenotype is indeed caused by LOC4341610 dysfunction. We obtained a 6807 bp full-length cDNA of DCM1 by performing rapid amplification of cDNA ends (RACE) (Fig 6A). The open reading frame (ORF) is 6207 bp in length, and the deduced protein contains 2068 amino acids. The DCM1 protein contains five tandem CCCH type zinc finger motifs at its C terminus (Fig 6B). Phylogenetic analysis showed that DCM1 is more closely related to its plant homologs, which constitute an isolated branch in the phylogenetic tree (Fig 6C). Multiple sequence alignments of the tandem CCCH domain from different species revealed that this domain is conserved among different kingdoms (Fig 6D).
The expression profile of DCM1 in different organs was determined by RT-PCR analysis. DCM1 transcripts were detected in all tested organs of wild type and young panicle of dcm1 (Fig 7A). In transgenic plants carrying the pDCM1::GUS construct, GUS signals were detectable in the anther (Fig 7B). To more precisely determine the spatial and temporal patterns of DCM1 expression in the anthers, we performed RNA in situ hybridization on wild-type anther sections. No DCM1 expression was detected in sporogenous cells in the anther during the premeiotic stage (Fig 7C). At the stage when callose began to deposit around the PMCs, DCM1 was specifically expressed in PMCs that adhered to each other. DCM1 expression was also observed in the tapetum during later stages.

DCM1 interacts with nuclear poly (A) binding proteins in nuclear speckles
To identify potential interacting partners of DCM1, we screened a rice anther cDNA library using the yeast two-hybrid system. The full-length DCM1 protein can autonomously activate the reporter gene in the absence of a prey protein and thus cannot be used as bait. Therefore, we split the DCM1 protein into two parts and tested their autoactivation activity separately. We found that the N terminus (1-1293) is responsible for autoactivation ( Fig 8A). Therefore, the C terminus of DCM1 from the amino acid 1294 to the end was used as the bait. From this screen, we identified the two nuclear poly (A) binding proteins (OsPABNs) in rice and named them OsPABN1 (LOC_Os02g52140) and OsPABN2 (LOC_Os06g11620) (32.7% of sequenced positive colonies), respectively (Fig 8B). OsPABN1 and OsPABN2 are homologs of human poly (A) binding protein nuclear 1 (PABPN1) and members of the polyadenylation factor family, which are required for mRNA polyadenylation. OsPABN1 and OsPABN2 share high sequence similarity (77% identity) and both contain an RNA recognition motif (RRM) in the central region. The clone with shortest sequence identified in this screen corresponds to a peptide ranging from the 85th amino acid to the C terminal end of OsPABN1, indicating that the coiled-coil domain is not necessary for its interaction with DCM1.
We conducted yeast two-hybrid assays to verify these interactions. Transformants with DCM1-C and OsPABN1, OsPABN2 grew on QDO/X/A media, confirming the interactions between them (Fig 8C). To clarify whether the tandem CCCH domain of DCM1 is required for its interaction with the OsPABNs, we split the DCM1-C into three parts, DCM1-D (1294-1563), DCM1-E (1557-1772), and DCM1-F (1773-2068) and tested their interactions with OsPABNs. The results showed that DCM1-D interacts with OsPABNs, while DCM1-E and DCM1-F, which contain the coiled-coil domain and the tandem CCCH domain, respectively, do not.
We also verified the interaction between DCM1 and the OsPABNs using the bimolecular fluorescence complementation (BiFC) assay (Fig 8E). Interactions between DCM1 and the two OsPABNs reconstituted the cyan fluorescent protein (CFP) in rice protoplasts and the CFP signal was observed in nuclear speckles. Moreover, OsPABN1 and OsPABN2 interact with themselves and between each other (Fig 8D and 8E). The BiFC results indicate that these interactions also take place primarily in the nuclear speckles.

Expression level of genes involved in callose metabolism and exine development is altered in dcm1
Based on the observation of prematurely dissolution of callose in the dcm1 PMCs and DCM1 might be involved in mRNA metabolism, we examined the expression profiles of genes involved in callose metabolism in rice anthers during meiosis (Fig 9). There are ten predicted callose synthase genes (OsGSL1-OsGSL10) in the rice genome [29]. Among them, only OsGSL5 and OsGSL8 have been evaluated for their respective functions in male fertility and ovary expansion [14,30]. Compared with wild type, the expression level of OsGSL2 and OsGSL10 increased by 1.49-and 1.97-fold, respectively, in dcm1 ( Fig 9A). However, the expression level of OsGSL3 and OsGSL9 in dcm1 decreased to 58.2% and 72.6% of that in wild type, respectively. OsGSL2 and OsGSL3 are closely related to AtGSL1 and AtGSL5, which synthesize the interstitial callose during meiosis in Arabidopsis [15]. No significant changes in expression level between wild type and dcm1 were observed for the other OsGSL genes.
Osg1 encodes a rice β-1, 3-glucanase that was reported to degrade callose during pollen development [31]. The expression level of this gene in dcm1 increased to 1.30-fold of that in wild type (Fig 9B). The A6 gene, which encodes an O-Glycosyl hydrolases family 17 protein, and its positive regulator AtMYB80 were reported to affect callose dissolution in Arabidopsis [32,33]. The expression patterns of their homologs in rice, OsA6 and OsMYB80, were examined. In the mutant, the expression levels of OsA6 increased to 1.28-fold and OsMYB80 increased to 2.89-fold of wild type. These results suggested that callose degradation process might be altered in dcm1. No significant difference in the expression of UGP1, an UDP-glucose pyrophosphorylase gene that is essential for callose deposition [34], was observed between wild type and dcm1.
We also evaluated the expression level of genes related to exine formation (Fig 9C). OsABCG15 encodes an ATP-binding cassette transporter protein that is essential for pollen exine development in rice [35]. The expression level of OsABCG15 in dcm1 reduced to 13.3% of that in wild type. A fatty acyl carrier protein reductase, DPW, plays a role in the formation of regular exine [36]. The expression level of DPW in the mutant reduced to 29.6% of that in https://doi.org/10.1371/journal.pgen.1007769.g009 DCM1 in male meiotic cytokinesis wild type. These results suggested that exine development process is affected in dcm1. We also compared the expression level of genes related to meiosis. However, no significant differences in the expression of these genes, including OsSPO11-1, OsCOM1, OsRAD17, OsDMC1, OsRAD51, OsMER3 and OsREC8, were observed between wild type and dcm1 (S5 Fig).

Discussion
In this study, we showed that dysfunction of DCM1, a tandem CCCH protein, disturbs the metabolism of callose both around the PMCs and at the cell plate. Without the callosic plate, meiotic cytokinesis in dcm1 is not completed. The defective meiotic cytokinesis during telophase I lead to the random movement of the spindles at meiosis II and the varied arrangement of meiotic products. The aberrant meiotic products develop into pollen grains with varied sizes and DNA contents, which is reminiscent of the phenotypes of atps1 and jason in Arabidopsis [27,37]. In Arabidopsis, instead of the cell plate, an organelle band is formed during meiotic telophase I. The loss of AtPS1 or JASON protein leads to the disruption of organelle band, resulting in the disrupted spindle positioning during male meiosis II [38], just like that observed in dcm1.

The role of callose in meiotic cytokinesis in monocotyledonous plants
The deposition of callose during mitotic and meiotic cytokinesis in plants has been meticulously analyzed by electron microscopy [39,40]. It is possible that the membrane network serves as a trigger for the induction of callose accumulation, since it only appears directly at the cell plate and not in the secretory vesicles en route to the division plane [41]. Samuels et al. (1995) postulated that callose helps to mechanically stabilize the early delicate membrane networks of forming cell plates in tobacco, and to create a rapid spreading force that widens the tubules and converts them into plate-like structures. However, Thiele et al. (2009) revealed that the cell-plate membrane compartment forms and expands, seemingly as far as the parental wall, prior to the appearance of callose. Based on characterization of the gsl8 mutant in Arabidopsis, they speculated that callose might be required for inserting the nascent crosswall at the division site during mitosis.
Callose is only a transient structure during somatic cytokinesis that is replaced by cellulose during the final step. However, this is not the case in successive-type meiotic cytokinesis, which is typically observed in monocots PMCs. Callosic plates are formed after both the first and the second caryokinesis events during meiosis. The callosic plate is preserved when the first cell division is completed and persists through the second cell division. Whether other polysaccharides, such as cellulose and pectin, are deposit on the cell plate during meiotic cytokinesis is still uncertain. It is possible that the callosic plate is the only physical barrier insulating the cytoplasm during meiotic cytokinesis.

The possible mechanism by which DCM1 coordinates the metabolism of callose
The most prominent feature of DCM1 is the five tandem CCCH domain at its C-terminus, which is conserved between different kingdoms. The homolog of DCM1 in Arabidopsis was named SOP1 (SUPPRESSOR OF PAS2 1), the mutation of which suppressed the developmental defects of a splicing-defective allele of PASTICCINO2 [42]. Red5, the homolog of DCM1 in fission yeast, is essential for efficient elimination of specific meiotic mRNAs during vegetative growth and the proper splicing of meiotic genes [43][44][45], indicating that DCM1 might be involved in mRNA processing or elimination. OsPABN1 and OsPABN2, the two nuclear poly A binding proteins in rice, were identified as interacting partners of DCM1. In fission yeast, Red5 was co-immunoprecipitated with Pab2, the orthologue of OsPABN1 and OsPABN2 [43,46]. In addition, dZC3H3, the homolog of DCM1 in flies, also associates with PABP2 [47]. Recently, the direct interaction between Red5 and Pab2 was detected in a proteome-wide interactome study [48]. Therefore, the interactions between proteins containing five tandem CCCH domain and nuclear poly (A) binding proteins might be conserved in fungi, animals and plants.
Studies of DCM1 homologs in the other species mentioned above indicate that they are involved in nuclear exosome functions, and RNA degradation by the exosome is the main pathway for the removal of unwanted RNA in all kingdoms [49]. Moreover, studies in fission yeast have demonstrated that Pab2 and the nuclear exosome subunit, Rrp6, are the main factors involved in a polyadenylation-dependent pre-mRNA degradation pathway, and inefficient splicing is important to dictate susceptibility to this process [50]. The retained PAS2 transcript in sop1 harbors a nucleotide substitution before a GC splicing site that mimics a pre-mRNA with low splicing efficiency [42], indicating that the tandem CCCH zinc finger protein and the poly (A) binding protein could be involved in conserved biological processes across species.
During meiosis, callose metabolism must be precisely controlled to ensure the completion of meiotic cytokinesis by its deposition and release of microspores by its dissolution. In dcm1, callose around PMCs was deposited normally during leptotene and disappeared during pachytene ( Fig 4B). Similarly, prematurely dissolution of callosic plates occurred during telophase I and telophase II in the mutant. It seems that the dynamic changes in callose (deposition and dissolution) in dcm1 reflect the antagonism between callose synthesis and degradation. We speculate that the defective callose metabolism in dcm1 is caused by unwanted RNAs that are eliminated in the presence of DCM1. In Arabidopsis, the expression of some callose-related genes is affected by the mutation of CALLOSE DEFECTIVE MICROSPORE1 (CDM1), a gene encoding a tandem CCCH-type zinc finger protein [51]. CalS5, the Arabidopsis gene responsible for callose deposition surrounding the PMCs, tetrads and microspores, is regulated in a post-transcription manner [52]. We checked the expression level, splicing and polyadenylation site of OsGSL5 gene and no significant differences were found between dcm1 and wild type. Besides Osg1, whose function was demonstrated by RNA interference assay [31], few callase genes (encoding β-1, 3-glucanase) have been functionally characterized in rice meiosis. Thus, further studies to identify the mRNA targets of DCM1 will help to uncover the mechanisms by which DCM1 regulates callose metabolism during rice meiosis.

Plant materials and grow conditions
The dcm1 mutant was isolated from a collection of mutants induced by 60 Co-γ-ray irradiation on indica rice (Oryza sativa) variety, Guangluai 4. The variety Guangluai 4 was used as the wild type in all experiments. All plants were grown in the paddy field.

Map-based cloning of the DCM1 gene
When a line segregated 1:3 for sterile and fertile plants, the fertile plants was selected to cross with the japonica rice variety Wuyunjing 8. Using the sterile plants that segregated in the F2, F3 and F4 population (299 plants in total), the mutated gene was mapped to a 138.5kb interval between markers C6-25.86 and C6-26.003. Based on the MSU Rice Genome Annotation Project Database and Resource (http://rice.plantbiology.msu.edu/), there are 23 predicted genes in this region and 1bp deletion in the gene LOC4341610 was detected between the mutant and wild type by sequencing. Indel (insertion-deletion) markers used for mapping were designed based on the sequence differences between indica variety 9311 and japonica variety Nipponbare according to the data published (http://www.ncbi.nlm.nih.gov). Primers used were listed in the supporting information (S2 Table).

Knock out DCM1 by CRISPR/Cas9 gene editing approach
The CRISPR/Cas9 gene editing approach in rice was performed according to a protocol described previously [53]. Target sequences (TTGGATTTGACTTTGCTCTT) were selected according to the following criterions: (1) Sequence located in the exon. (2) Followed by the PAM (NGG). (3) Be unique among the genome to avoid multiple knock out sites. Designed primer were incubated at 100˚C for 5 min and subcloned into vector pC1300-Cas9-1gRNA with a hygromycin resistance marker. The resulting constructs were transformed into Agrobacterium tumefaciens EHA105 and then into calli of Yandao 8, a japonica variety. T 0 plants were genotyped to verify whether the editing site was modified. Target sequences used in this study were listed in the supporting information (S2 Table).

Full-length cDNA cloning of DCM1
Total RNA extraction was conducted using the TRIzol reagent (Invitrogen), as described by the supplier, from rice young panicle. Reverse transcription was performed using SuperScript III First-Strand Synthesis System for RT-PCR kit (Invitrogen). The cDNA was amplified using the 5'and 3'-Full RACE kits (TaKaRa). PCR using primers CDS-F and CDS-R was performed to amplify the open reading frame. The sequences were sequenced and then spliced together to obtain the full-length cDNA. Primers mentioned above were listed in the supporting information (S2 Table).

Quantitative RT-PCR assay, promoter::GUS analysis of DCM1 expression pattern and RNA in situ hybridization
Real-time PCR analysis was performed using the Bio-Rad CFX96 real-time PCR instrument and Hieff qPCR SYBR Green Master Mix (No Rox Plus) (YEASEN). All PCR experiments were conducted using 40 cycles of 95˚C for 10 s, 60˚C for 30 s. All reactions were performed in triplicate, with Ubiqutin as the normalized reference gene for all comparisons. The primers for qRT-PCR were listed in S2 Table. A 2.4Kb upstream region of the DCM1 gene was amplified and cloned into pCAMBIA1301. The construct was introduced into the japonica rice variety Nipponbare. Spikelets at different developmental stages from transgenic lines were incubated with X-Gluc (5-bromo-4-chloro-3-indolyl-β-D-glucuronide) solution. The spikelets were cleared in 75% (v/v) ethanol and photographed with a stereoscope.
For in situ hybridization, sense and antisense probes were synthesized with T7 RNA polymerase using the digoxigenin RNA labeling kit (Roche). Tissue fixation and hybridization were performed as previously described [54]. Primers used were listed in S2 Table. Cytology Rice anthers at various developmental stages were fixed in 4% paraformaldehyde overnight. The samples were washed in PBS, dehydrated in a graded ethanol series, and embedded in Technovit 7100 resin (Heraeus Kulzer). Microtome sections (4μm thick) were stained with 0.25% toluidine blue to stain the cells and 0.1% aniline blue to stain the callose, respectively. The slides were photographed with an Olympus BX51 microscope and a digital camera.
DAPI staining of mature pollen was performed according to Liu [55]. Young panicles were fixed in Carnoy Solution for at least 24 h at room temperature. Anthers at the proper developmental stage were then squashed in acetocarmine. Meiotic chromosome preparation and immunofluorescence analysis were performed as previously described [56]. Aniline blue (0.1% in PBS) was applied to slides with DAPI to co-stain the callose and DNA. Images were captured under a Zeiss A2 fluorescence microscope with a micro CCD camera. Immunogold assay was conducted using antibody to β-1, 3-glucan at dilution of 1:100 (Biosupplies). The secondary antibody is goat anti-mouse IgG-15 nm gold (Abcam) at dilution of 1:20.

Two-hybrid cDNA library screening and yeast two-hybrid (Y2H) assay
Total RNA was extracted from wild-type anthers during meiosis. Highly purified and intact mRNA was isolated using mRNA Purification Kit (Life technology, NO. 61006). cDNA library construction was performed using Make Your Own Mate & Plate Library System (Clontech NO.630490) according to the manufacturer's instructions. The Y2H assay was conducted with Yeast Transformation System 2 (Clontech NO.630439). The yeast strain Y2HGOLD was used for library screening and Y2H assays. All primers used were listed in S2 Table. BIFC assay The BiFC assay was performed as previously described [57]. The full-length coding sequences of DCM1, OsPABN1 and OsPABN2 were subcloned into vector pSCYNE(R) and pSCYCE(R), respectively. The plasmid pairs were co-transfected into protoplasts of young rice seedling by 40% PEG (polyethylene glycol). Transfected protoplasts were incubated in the dark at 28˚C overnight, and observed using a laser scanning confocal microscopy. Primers used were listed in S2 Table. Supporting information