EAT1 transcription factor, a non-cell-autonomous regulator of pollen production, activates meiotic small RNA biogenesis in rice anther tapetum

Abstract

The 24-nucleotides (nt) phased secondary small interfering RNA (phasiRNA) is a unique class of plant small RNAs abundantly expressed in monocot anthers at early meiosis. Previously, 44 intergenic regions were identified as the loci for longer precursor RNAs of 24-nt phasiRNAs (24-PHASs) in the rice genome. However, the regulatory mechanism that determines spatiotemporal expression of these RNAs has remained elusive. ETERNAL TAPETUM1 (EAT1) is a basic-helix-loop-helix (bHLH) transcription factor indispensable for induction of programmed cell death (PCD) in postmeiotic anther tapetum, the somatic nursery for pollen production. In this study, EAT1-dependent non-cell-autonomous regulation of male meiosis was evidenced from microscopic observation of the eat1 mutant, in which meiosis with aberrantly decondensed chromosomes was retarded but accomplished somehow, eventually resulting in abortive microspores due to an aberrant tapetal PCD. EAT1 protein accumulated in tapetal-cell nuclei at early meiosis and postmeiotic microspore stages. Meiotic EAT1 promoted transcription of 24-PHAS RNAs at 101 loci, and importantly, also activated DICER-LIKE5 (DCL5, previous DCL3b in rice) mRNA transcription that is required for processing of double-stranded 24-PHASs into 24-nt lengths. From the results of the chromatin-immunoprecipitation and transient expression analyses, another tapetum-expressing bHLH protein, TDR INTERACTING PROTEIN2 (TIP2), was suggested to be involved in meiotic small-RNA biogenesis. The transient assay also demonstrated that UNDEVELOPED TAPETUM1 (UDT1)/bHLH164 is a potential interacting partner of both EAT1 and TIP2 during early meiosis. This study indicates that EAT1 is one of key regulators triggering meiotic phasiRNA biogenesis in anther tapetum, and that other bHLH proteins, TIP2 and UDT1, also play some important roles in this process. Spatiotemporal expression control of these bHLH proteins is a clue to orchestrate precise meiosis progression and subsequent pollen production non-cell-autonomously.

Fig 1. Bimodal expression of EAT1 protein at both early meiosis and postmeiosis in anther tapetum.

(A) Anther lobe architecture around meiosis in rice. (B) Diagram of the EAT1pro-EAT1-GFP transcriptional fusion construct. Closed and grey boxes indicate protein coding and untranslated regions, respectively. (C) eat1-4/eat1-4 flowers of T₀ plants carrying EAT1pro-EAT1-GFP (#1, #2) and an empty vector. Bars, 1 mm. Flower images were taken after removal of lemmas. (D) EAT1-GFP signals (green) in developing anther sections from ST.1 to ST.5. In a transgenic plant harboring the EAT1pro-EAT1-GFP. The EAT1-GFP signals were restricted to tapetal nuclei in ST.2, ST. 3 and ST.5 anthers, and not detected in the ST.2 anther from the negative control (n.c., right most panel). About the meiotic events and anther lengths corresponding to the respective stages. See Table 1 and S7 Table. Bars, 20 μm.

Fig 2. The eat1-4 mutation affects meiotic chromosome condensation non-cell-autonomously.

(A) Cross sections of anthers at late meiosis (ST.4). Tapetum and PMC formation of the eat1-4 mutant was almost comparable to that of the wild-type (WT). Bars, 20 μm. (B) Accumulation (ST.2) and degeneration (ST.3) of β-1,4 glucan (green) at tapetal-cell and PMC walls. Nuclei were counterstained with propidium iodide (magenta). Bars, 20 μm. (C) Typical PMCs observed in respective meiotic stages in wild-type (top) and eat1-4 anthers (bottom). Meiotic chromosomes, stained with 4',6-diamidino-2-phenylindole (DAPI), were decondensed frequently in the eat1-4 PMCs. An arrow indicates lagging chromosomes. Bars, 20 μm.

Fig 3. Identification and characterization of EAT1-dependent and early meiosis-enriched expression of 24-PHAS precursor transcripts.

(A) Scatter plots of FPKM values for transcripts of 38,311 coding genes, 6,097 noncoding genes and 15,723 TE-like sequences, compared between the wild-type and eat1-4 ST.2 anthers. EAT1-d and EAT1-i indicate that the transcripts show EAT1-dependent and EAT1-independent expressions, respectively. DCL5 had slightly larger standard deviation of FPKM values in wild-type ST.2 (mandarin in left plot). In all plots, dark and faint gray spots represent transcripts whose FPKM values were ≥2-fold different between wild-type and eat1-4 anthers, respectively. (B) Heatmaps representing the expression level of 113 loci encoding 24-PHAS precursor transcrips (left), and of 24-nt siRNAs (right) derived from the corresponding 24-PHAS loci (left). Each experiment includes three biological replicates. The leftmost dendrogram indicates the result of clustering of 24-PHAS expression patterns by R package, gplots. Asterisks indicate that the loci were silent through ST.1 to ST.4 stages (black) or showed EAT1-independent expression (grey). (C) Box plots representing 24-PHAS RNA density per locus (left) and 24-nt phasiRNA density per locus (right) in ST.1, ST.2 and ST.4 anthers of wild-type (green boxes) and eat1-4 (brown boxes). *** indicate that difference is significant at P = 0.001 in Student's t-test. (D) qRT-PCR results of five 24-PHAS transcripts (chr5-20, chr6-97, chr10-100, chr10-101 and chr12-83) in wild-type (black lines) and eat1-4 anthers (gray lines). The bottom numbers correspond to anther developmental stages in Table 1. Relative expression values and standard errors were calculated by using three biological replicates.

Fig 4. Characterization of 24-PHAS loci on rice genome.

(A) A genome-wide distribution of 24-PHAS loci. From top to bottom, the numbers of 24-PHAS loci (101 green triangles correspond to 24-PHAS loci showing an EAT1-dependent and ST.2-enriched expression, 9 gray triangles are previously reported 24-PHAS loci silent through ST.1 to ST.4 and 3 blue triangles are those showing EAT1-independent expression), the amounts of 24-nt sRNA-seq (red), mRNA-seq reads (blue) rated by subtraction of eat1-4 values from wild-type values (see Methods), and frequencies of repetitive sequences including TEs (gray charts). The horizontal length of each box corresponds to the physical distance of respective rice chromosomes. (B) A conserved sequence logo found in upstream of ninety-three 24-PHAS loci detected by MEME program [35], which are potentially targeted by miR2275. The arrow indicates the predicted cleaved position by DCL1 and miR2275 complex [10]. (C) Frequency of repetitive sequence (grey), gene coding region (blue) and miR2275 targeted site (red) around 24-PHAS loci. The data was examined in 93 24-PHAS transcripts with conserved miR2275 targeted sites. The reason why a small peak of miR2275 target site appeared at the 3’ end of 24-PHAS is that some 24-PHAS loci were relatively small in length (~ 500 bp). (D) Characterization of three 24-PHAS loci. From the top to the bottom, the graphs indicate the mapping results of mRNA-seq and 24-nt sRNA-seq reads (gray histograms), the 24-nt phasing pattern (green and orange charts), and the plot of read counts from the degradome-seq using young panicles of indica variety, 93–11 [38]. The degradome analysis revealed that the cleavage of three 24-PHAS transcripts frequently occurs at the position shown in (B), within the predicted miR2275 sites (red dots), while few degradome-seq reads were mapped onto both sense and antisense strands of other regions (gray dots). Reads were depicted by IGV [78]. (E) An example of distribution of EAT1-dependent 24-PHAS-loci cluster (green boxes) on the long arm of chromosome 12, with the context of surrounding genes (blue) and repetitive sequences (black).

Fig 5. EAT1 and TIP2 bind E-box motifs upstream of 24-PHAS loci and DCL5 gene.

(A) Schematic illustrations of genomic compositions of the 5ʹ upstream regions of two 24-PHAS loci, chr5-20 and chr6-97, in addition to the coding region of the Ubiquitin gene as a negative control. Open boxes indicate the position of consensus E-box motifs. The number at the top of each motif shows a distance (bp) from the transcription start site (TSS). Regions underlined were used in the ChIP-qPCR assay. Grey and closed boxes in the Ubiquitin represent untranslated and coding regions, respectively. (B) ChIP-qPCR results of 24-PHAS promoters using transgenic (TG) plants expressing EAT1-GFP. IgG and non-TG plants were used as negative controls. n.s.; not significant. * and **; significant at P = 0.05 and P = 0.01 in Student's t-test, respectively, less than the leftmost positive ChIP result in each graph. (C) qRT-PCR results of DCL5 mRNA in wild-type and eat1-4 anthers. Relative expression values and standard errors were calculated by three biological replicates. The bottom numbers correspond to anther developmental stages in Table 1. (D) Genomic composition of the 5ʹ upstream region of the DCL5 gene. (E) ChIP-qPCR results of DCL5 promoters using TG plants expressing EAT1-GFP. (F) ChIP-qPCR results of 24-PHAS promoters using TG plants expressing YFP-TIP2. (G) ChIP-qPCR results of DCL5 promoters using TG plants expressing YFP-TIP2. In ChIP-qPCR analyses, relative abundance and standard errors were calculated by two or three biological replicates each subjected to three PCR replications.

Fig 6. EAT1 and TIP2 activate the promoter activity of 24-PHAS loci and the DCL5 gene in interaction with UDT1.

(A) The results of the transient expression assay. Any one or two effector plasmids encoding EAT1 (E1), TIP2 (T2), UDT1 (U1) and TDR (TD) proteins were cotransfected with the reporter constructs into rice protoplasts. The reporter carries a 2-kbp promoter region of the 24-PHASs (chr5-20, chr6-97) or DCL5, fused with the firefly Luciferase. The configuration of all constructs were shown in S10A Fig. The number above each bar is the fold change of the Luciferase activity compared to the negative control without the promoter (leftmost bars). *, ** and ***; the significant fold changes at P = 0.05, 0.01 and 0.001 in Student’s t-test, respectively, compared to the negative control. Error bars indicated standard deviation of three biological replicates. The significant >2 fold changes were in bold. (B) The BiFC results of EAT1-UDT1 an TIP2-UDT1 cotransfections.

Fig 7. The MEL1 Argonaute protein associates with EAT1-dependent 24-nt phasiRNAs in male meiocytes.

(A) MEL1-GFP was specifically expressed in the male germline in ST.2 anthers in MEL1pro-MEL1-GFP transgenic plants. Bars, 10 μm. (B) Pie-charts representing the ratios of 24-nt MEL1-associating siRNAs (masiRNAs) originated from 24-PHAS loci, protein-coding genes, intergenic regions except for 24-PHAS loci and repetitive regions, in wild-type samples through ST.1, ST.2 and ST.4 stages. The numbers with parentheses indicate the read counts of 24-nt masiRNAs extracted from MEL1-IPseq results. (C) The mapping mode of 24-nt masiRNAs on two 24-PHAS loci, for example. Tandem arrays of open box-arrows (top) represent the 24-nt phased interval pattern on both strands of each PHAS locus. Green box-arrows are 24-nt masiRNAs exactly fitting to the interval. Red arrowheads indicate conserved miR2275 targeted sites. Each bar graph (bottom) indicates RPM values of the 24-nt masiRNA (masiRNA_u_0815 or _1708) in wild-type (WT), mel1-1 and eat1-4 anthers. The numbers at the top of bars represent a total read counts of 24-nt masiRNAs with two biological replicates.