A Y-chromosome–encoded small RNA acts as a sex determinant in persimmons

See allHide authors and affiliations

Science  31 Oct 2014:
Vol. 346, Issue 6209, pp. 646-650
DOI: 10.1126/science.1257225

Y male plants affect female development

Although most plants have both male and female organs within a single flower, some produce separate male and female plants. In some cases, such as persimmons, males are determined by a Y chromosome. Akagi et al. examined the gene transcript differences between male and female persimmons. A gene on the Y chromosome regulated a non–sex chromosome–linked small RNA that suppresses female organ development. This small RNA was localized to male flowers and could affect female development in other plant species. The evolutionary history of these genes suggests that they are tied to the origin of the separation of sexes in the persimmon family.

Science, this issue p. 646


In plants, multiple lineages have evolved sex chromosomes independently, providing a powerful comparative framework, but few specific determinants controlling the expression of a specific sex have been identified. We investigated sex determinants in the Caucasian persimmon, Diospyros lotus, a dioecious plant with heterogametic males (XY). Male-specific short nucleotide sequences were used to define a male-determining region. A combination of transcriptomics and evolutionary approaches detected a Y-specific sex-determinant candidate, OGI, that displays male-specific conservation among Diospyros species. OGI encodes a small RNA targeting the autosomal MeGI gene, a homeodomain transcription factor regulating anther fertility in a dosage-dependent fashion. This identification of a feminizing gene suppressed by a Y-chromosome–encoded small RNA contributes to our understanding of the evolution of sex chromosome systems in higher plants.

Sexuality promotes and maintains genetic diversity in eukaryotic organisms. The characterization of sex chromosomes revealed evolutionary mechanisms governing sexuality in animals (13). However, most plant sex chromosomes, which could be present in up to 5% of species (4, 5), remain poorly characterized (58). Dioecy, the separation of sex organs among male and female individuals, can be controlled by a heterogametic male system comparable to that of mammals and based on X and Y chromosomes, or on the X-to-autosome ratio (58). Species with heterogametic females, such as those of birds (ZW system), are less common (8). Studies of Y-chromosome structure and evolution in Silene latifolia (911), papaya (Carica papaya) (1215), and date palm (16) have revealed a heterochromatic nonrecombining region controlling sex determination, a feature shared by loci controlling other sexual characters such as asexual reproduction via apomixis and the inability to self-fertilize via self-incompatibility systems (7). For Y-linked sex determination, as for apomixis, it has been challenging to identify genetic determinants in this heterochromatic context. A theoretical model postulates that two changes must occur during the transition from hermaphroditism to dioecy: a recessive mutation resulting in male sterility and a dominant female-suppressing mutation (8, 17).

The Diospyros genus, within the Ebanaceae (Ericales), contains mostly tree species, including the economically important persimmons (D. kaki, D. virginiana, and D. lotus) and ebony (D. ebenum). Dioecy may predate the divergence of the Diospyros genus (18) and possibly even the origin of the Ebenaceae (35 to 65 million years ago) (1820). Male flowers have fertile stamens but rudimentary, arrested carpels and are organized in a three-flower cyme. Female flowers display developed but defective anthers that normally do not produce pollen grains (fig S1, A to P). Although a single female flower is formed per inflorescence, lateral aborted flower primordia are often visible on the flower pedicel (fig. S1, Q and R).

We used de novo whole-genome sequencing and transcriptome approaches to characterize the sex determination system in the diploid D. lotus, located to a single sex determination (SD) locus on the Y chromosome (21).

To identify male-specific sequences, genomic sequencing libraries were constructed from D. lotus, segregating F1 sibling trees (21). Libraries pooled according to sex were sequenced to an estimated 45 to 50× coverage (Fig. 1A). Every 35–base pair (bp) subsequence (35-mer or k-mer)present in these sequencing reads was cataloged, and reads including significant male-pool–specific k-mers (MSKs) (Fig. 1B) were used for in silico assembly (methods 2 and 3) of 5100 contigs putatively located on the Y chromosome. We discarded those in which potential Y-specific polymorphisms perfectly matched reads from one or more female samples, which suggests that they are probably located in a region of active recombination (fig. S2). Integrating and expanding the remaining contigs resulted in ~800 contigs located on putative male-specific regions of the Y chromosome (MSYs, fig. S3) and amounting to a total length of ~1 Mb (Fig. 1C).

Fig. 1 Identification of male-specific sequences from genomic sequence reads.

(A) Genomic sequencing libraries were created from D. lotus segregating F1 sibling trees. (B) Reads from samples of the same gender were pooled and searched for the presence of gender-specific 35-bp k-mers (see inset graph). Reads including MSKs were assembled to generate Y-linked genomic contigs. (Inset graph) Distribution of gender-specific or -biased k-mers (35 bp). (C) Expected organization of the assembled contigs within the MSY locus and surrounding regions. Different numbers of individuals in (A) and (B) reflect availability through sexual maturity in the sib-family.

To identify genes expressed during the differentiation of male androecia and female gynoecia, RNA-Seq was performed on tissue from mixed buds, including floral organ primordia, from nine males and nine females from the F1 used for the genomic analysis, and their parents. We followed three complementary approaches to interpret these expression data: (i) we mapped the reads to the MSY genomic contigs; (ii) we seeded assembly with the reads, including male-specific cDNA k-mers (methods 2 and 4, fig. S4); and (iii) we assembled a draft of the transcriptome and identified genes that were differentially expressed between male and female individuals (method 6). Most of the expressed (reads per kilobase of sequence per million mapped reads) > 1.0) gene fragments isolated from (i) and (ii) were identical, even though they were derived independently, suggesting that we had successfully identified the majority of the sequences associated with MSYs (fig. S5). After integration of the genes identified using those two approaches, 22 candidate genes underlying the SD locus were identified (table S1).

We identified 62 genes that were differentially expressed between male and female samples [false discovery rate (FDR) < 0.01, table S2], seven of which were MSY-linked. We focused on a pair of class I homeodomain transcription factors, because the first gene, which we named OGI (Japanese for “male tree”), exhibits male-specific expression in developing buds, is a Y-specific SD candidate with no homologous sequence in the female genomic reads, and its coding sequence presents multiple disruptive stop codons. The second gene, which we named MeGI (Japanese for “female tree”), is not MSY-linked and is thus located on an autosomal region, but exhibited female-biased bud- and flower-specific expression (fig. S6, A and B) (FDR = 1 × 10–05, table S2). Moreover, these two genes are monophyletic orthologs of the barley Vrs1 gene (fig. S7). In barley, mutation of Vrs1 lifts the developmental inhibition on the lateral flowers, resulting in fully fertile three-flower inflorescences instead of a single central fertile flower and two rudimentary lateral spikelets (22, 23). This mutant is architecturally similar to the male flowers of D. lotus, whereas wild-type barley architecture resembles that of female D. lotus flowers (fig. S1, Q and R). OGI sequence analysis predicted the presence of a hairpin structure, with high similarity to the homologous region of the MeGI gene (Fig. 2A and fig. S8).

Fig. 2 Sequence and expression analysis of OGI and MeGI.

(A) Comparative structure of the OGI and MeGI genes. White and gray triangles in the OGI coding regions indicate disruptive deletions and insertions, respectively. Homeobox domains are noted. The blue region in OGI bears no homology to the MeGI sequences. (B) PCR amplification of the OGI gene in males (M) and females (F) of Diospyros species. (a) In hexaploid D. kaki, monoecious (B) cultivars bear both male and female flowers. (C) Maximum likelihood–based phylogenetic tree of the OGI (blue) forward (FR) and inverted (IR) repeat sequences and the corresponding region from the MeGI (red) sequence, from various Diospyros species. One-tenth values of bootstraps from 1000 replicates are indicated on the branches. (b) Taxa cover a substantial portion of the Diospyros genus (18, 20). (c) Ks value from the codon frame of the MeGI gene used to estimate (d) the timing of divergence between MeGI and OGI (method 8), to at least 52 million years ago, before the origin of the Diospyros genus.

Despite multiple disruptive mutations, the OGI gene sequence and male specificity are highly conserved in the Diospyros genus (Fig. 2B). Phylogenetic analyses suggest that the establishment of OGI predated Diospyros radiation (Fig. 2C and fig. S9) and that suppressed recombination maintained OGI on the Y chromosome for tens of millions of years. The nucleotide substitution patterns within the OGI repeats suggest lineage-independent coevolution (fig. S10), which is consistent with selective pressure maintaining a conserved doubled-stranded RNA structure. Furthermore, sequence analysis of ~150 kb surrounding OGI revealed highly repetitive sequences and the presence of male-specific regions (fig. S11), similar to MSY regions from other species (3, 14). In contrast, divergence between the X and Y alleles of the other 21 SD candidates suggests that they were established more recently than dioecy within this genus (18, 20) (figs. S9 and S12, table S3, and text S1 and S2). In conclusion, evolutionary inference supported OGI as the best SD candidate.

There is no proven method for D. lotus transformation, but it is expected to take several years. Therefore, we elected to functionally characterize OGI and MeGI in other systems instead. Transient coexpression assays in tobacco (Nicotiana benthamiana) demonstrated that overexpression of OGI suppressed the expression of MeGI (fig. S13A) (P = 0.00082, Student’s t test), suggesting that OGI can repress MeGI in plants.

Three of 11 Arabidopsis thaliana plants independently transformed with MeGI driven by the constitutive CaMV35S promoter (table S4) exhibited stunted growth and sterile androecia, occasionally producing nonfunctional pollenlike grains (Fig. 3, A to I), whereas carpels could produce fertile seeds upon cross-pollination (Fig. 3, J and K). The other eight A. thaliana transformants carrying the same overexpression construct were not stunted and were typically male-fertile (Fig. 3L, fig. S14, and table S4) but also displayed MeGI mRNA incomplete splicing, resulting in low expression of the functional MeGI transcript (P = 0.0017, Student’s t test; Fig. 3, M and N). Similarly, in N. tabacum, 5 out of 12 plants transformed with MeGI driven by its native promoter exhibited low pollen germination (P < 0.005, Student’s t test) and shorter androecia, whereas carpels remained normal (fig. S15). Taken together, the phenotypes observed in transgenic A. thaliana and N. tabacum were consistent with the morphology of female flowers in D. lotus (fig. S1). Furthermore, the association of male sterility with higher expression of MeGI in persimmon and in transgenic Arabidopsis suggested that MeGI may be a dosage-dependent growth inhibitor that can sterilize androecia. The OGI RNA hairpin structure suggested a role for RNA interference in MeGI repression. To test this hypothesis, we analyzed small RNA. OGI production of male-specific 21-bp small RNAs was high in buds and low in flowers. MeGI-specific small RNAs accumulated in the buds and flowers of males only, consistent with OGI triggering transitive and persistent small RNA production from MeGI (Fig. 4A and figs. S6C and S13B).

Fig. 3 Phenotypes of A. thaliana plants overexpressing MeGI under the control of the CaMV-35S promoter.

(A) to (K) Severe feminized and stunted phenotypes. (A) Whole plants at 62 days after germination. (B) Defective development of stamens and petals. (C and D) Dissected flowers (without petals and sepals) of 35S-MeGI (C) and control (D). Sg, stigma; DS, defective stamens; St, stamens; At, anthers; Fl, filaments. (E) Pollenlike grains (PL) sometimes produced from defective anthers (DA) in fully mature flowers. (F and G) Positive vital staining of pollen(-like) grain from the 35S-MeGI (F) and control (G) plants. (H and I) Pollen tube (PT) formation assay for 35S-MeGI (H, negative) and control (I, positive) plants. (J and K) Silique elongation and seed (Sd) formation after cross-pollination using wild-type (Col-0) pollen (CP) but not after self-pollination (SP). (L) to (N) Comparison of stunted and feminized (Femin.), and WT-like and hermaphroditic (Herm.) 35S-MeGI lines. (L) Representative phenotypes. (M and N) Polymerase chain reaction (PCR) amplification of genomic MeGI sequence (M), including both introns (Fig. 2A) and quantitative PCR (N) corresponding to fully spliced MeGI. MeGI transcripts in hermaphroditic plants exhibited incomplete splicing (M), resulting in at least a two- to threefold reduction in functional transcript levels (N).

Fig. 4 Functional analysis of OGI and MeGI.

(A) Small-RNA sequencing of developing male and female buds and read mapping to OGI (top) and MeGI (bottom) sequences. The OGI gene model and MeGI cDNA are shown on top. Coverage distributions and read mapping in the forward (pink) and reverse (blue) direction are shown for male reads for OGI and both male and female reads for MeGI. No female reads mapped to OGI. (B) Chromosomal map of OGI and MeGI, with a model of their interaction and potential function in female (left) and male (right) flower development. RNAi, RNA interference. The model omits a mechanism for male suppression of gynoecia. A second, dominant Y-linked gene could exist. Alternatively, under a single sex-determination model, OGI-mediated repression of MeGI might limit the gynoecia.

Our data suggest a model for sex determination in Diospyros, in which OGI, or Oppressor of MeGI, represses the expression of the feminizing Male Growth Inhibitor gene, MeGI, in male flowers (Fig. 4B). OGI’s role as a dominant suppressor of feminization could be consistent with the SuF gene model (17) for the establishment of dioecy. There is no evidence that MeGI promotes gynoecia formation. Instead, evidence from MeGI expression in hermaphroditic Arabidopsis supports a feminizing role through androecia sterilization. Mechanisms affecting gynoecia formation in Diospyros are still missing in this model, and an additional Y-linked locus may be required, as postulated by the two-mutation model for the evolution of dioecy (8, 17). Alternatively, a single master regulator of sex determination, such as SRY in humans, may be sufficient (2, 8, 24). Together with findings that a W-encoded Piwi-interacting RNA determines sex in silkworms (25), our results suggest that RNA-based sex-determining mechanisms may be present in other systems as well. The involvement of small RNAs in sex determination might explain why they have been difficult to identify so far. Our discovery in Diospyros of a small-RNA–based repressor encoded by the Y chromosome and acting on a transcription factor that affects anther fertility provides a plausible sex-determination mechanism and an exciting starting point for comparison with other dioecious systems.

Supplementary Materials

Materials and Methods

Supplementary Text S1 and S2

Figs. S1 to S15

Tables S1 to S7

References (2646)

References and Notes

  1. Acknowledgments: We thank M. Lieberman and K. Ngo for bioinformatics support, the University of California Davis Genome Center DNA Technologies Core for sequencing support, and the U.S. Department of Agriculture’s Agricultural Research Service for providing plant samples from various Diospyros species. Supported by a Grant-in-Aid from the Japan Society for the Promotion of Science (JSPS) fellows (to T.A.); a Grant-in-Aid for Young Scientists (A) (no. 26712005 to T.A.) and for Challenging Exploratory Research (no. 26660025 to R.T.) from JSPS; a University of California Davis Genome Center seed grant (to I.M.H.); and the Office of Science (Biology and Environmental Research) of the U. S. Department of Energy (grant DE-SC0007183 to L.C.). All sequence data generated in the context of this manuscript have been deposited in the appropriate NCBI database: Illumina and PacBio reads in the Short Read Archives (SRA) database (SRA ID SRP045872), RNA-Seq reads and associated statistics in the Gene Expression Omnibus (GEO: GSE61386) database (BioProject ID PRJNA261435), and the OGI and MeGI genomic sequences obtained from various Diospyros species were submitted to GenBank (IDs KM408638 to KM408642). The Whole Genome Shotgun (WGS) project has been deposited at DDBJ/EMBL/GenBank under the accession no. JRBH00000000. The version described in this paper is version JRBH01000000. The Transcriptome Shotgun Assembly (TSA) project has been deposited at DDBJ/EMBL/GenBank under the accession no. GBSJ00000000. The version described in this paper is the first version, GBSJ01000000. The authors declare no competing financial interests. T.A., I.M.H., and L.C. designed the experiments. T.A. performed the experiments. T.A. and I.M.H. analyzed the data. R.T. initiated the study and maintained the KK persimmon sib-family. T.A. drafted the manuscript. All authors participated in data interpretation, edited the manuscript, and approved the final manuscript. Correspondence and requests for materials should be addressed to R.T. ( and L.C. (
View Abstract

Stay Connected to Science

Navigate This Article