Abstract
In maize, 24-nt phased, secondary small interfering RNAs (phasiRNAs) are abundant in meiotic stage anthers, but their distribution and functions are not precisely known. Using laser capture microdissection we analyzed tapetal cells, meiocytes, and other somatic cells at several stages of anther development to establish the timing of 24-PHAS precursor transcripts and the 24-nt phasiRNA products. By integrating RNA and small RNA (sRNA) profiling plus single-molecule and sRNA FISH (smFISH or sRNA-FISH) spatial detection, we demonstrate that the tapetum is the primary site of 24-PHAS precursor and Dcl5 transcripts and the resulting 24-nt phasiRNAs. Interestingly, 24-nt phasiRNAs accumulate in all cell types, with the highest levels in meiocytes, followed by tapetum. Our data support the conclusion that 24-nt phasiRNAs are mobile from tapetum to meiocytes and to other somatic cells. We discuss possible roles for 24-nt phasiRNAs in anther cell types.
In plants, small RNAs (sRNAs) belong to three main categories: microRNAs (miRNAs) involved in target mRNA cleavage, heterochromatic small interfering RNAs (hc-siRNAs) involved in RNA-directed DNA methylation (RdDM), and phased, secondary small interfering RNAs (phasiRNAs). PhasiRNAs were initially reported in male reproductive organs of maize and rice1-6, and more recently in diverse monocots and eudicots7-10; they are found primarily in anthers. In most species, 21-nt phasiRNAs predominate during early anther development, while 24-nt phasiRNAs are abundant during meiosis.
PhasiRNA biogenesis starts with transcription of long noncoding precursors (PHAS transcripts) by RNA polymerase II. Subsequently, the PHAS transcripts are capped and polyadenylated, translocated to the cytoplasm, and then bound by ribosomes11. In the polysomal complex, an Argonaute-associated 22-nt miRNA complementary to sequences in the PHAS transcript binds to this precursor molecule and triggers cleavage. miR2118 is utilized for most 21-nt PHAS precursors and miR2275 for most 24-nt PHAS precursors based on the presence of complementary sequences; a fraction of precursors lacks such complementarity and may be processed utilizing other miRNAs. Then, the 3′ portions of cleaved transcripts are converted to double-stranded RNA by RNA-DEPENDENT RNA POLYMERASE 6 (RDR6), and subsequently processed by Dicer-like proteins (DCL4 for the 21-nt and DCL5 for the 24-nt type) to yield the phasiRNAs1,2,4,12. Despite the expectation of equal stoichiometry of all the products from one PHAS transcript, individual phasiRNA abundances derived from the same transcript can differ by more than 1000-fold indicating differential stabilization6,13. It is possible that only these few abundant products generated from one precursor have functions.
In maize the 21-nt phasiRNAs are distributed throughout immature anther lobes based on in situ hybridization with one or a few examples, although the production of miR2118 is restricted to the epidermis1. Several observations associate phasiRNAs with male sterility. First, the absence of specific 21-nt phasiRNAs are implicated in causing male sterility in rice14,15. Second, in numerous male-sterile mutants of maize, either the 21-nt or 24-nt phasiRNAs are absent reflecting the tight control of the normal timing of their appearance during development1,2,16,17. More specifically addressing the role of 24-nt phasiRNAs, dicer-like5 mutants unable to process most 24-nt PHAS precursors are male-sterile and exhibit aberrant tapetal development under normal growing conditions2. In maize, the expression of PHAS precursors and miR2275 as well as the accumulation of 24-nt phasiRNAs occur mainly in the tapetum at the prophase I stage in anthers1. Male-sterile mutants with tapetal developmental defects evident during early meiotic stages are missing the 24-nt phasiRNAs2,17. Interestingly, the male sterility cases in rice and maize exhibit temperature-dependent male sterility: in appropriate growing conditions, the plants are male-fertile despite the absence of specific or an entire class of phasiRNAs2,15,16. This observation has led to the hypothesis that phasiRNAs may buffer anther development against environmental perturbation, particularly in the tapetum. In addition to the expectation that phasiRNAs could modulate mRNA half-life of still poorly defined targets, the 24-nt phasiRNAs result in increased CHH DNA methylation at 24-PHAS loci in maize, as assessed in whole anthers13. In analysis of isolated meiocytes, 24-nt phasiRNAs have been detected, and there is an increase in CHH DNA methylation, leading to the hypothesis that these phasiRNAs might regulate meiosis18 by an unknown mechanism related to cis methylation of their own loci. To date, the precise role of 24-nt phasiRNAs in regulating maize anther development has remained elusive.
sRNAs can function by travelling within and between organs: from the site of synthesis, they can spread both to neighboring cells and systemically over long distance19-21. Cell-to-cell movement of sRNAs is probably through plasmodesmata (PD), a plasma membrane-lined pore acting as an intercellular channel that connects the cytoplasm of adjacent cells22. To date, the potential movement of 24-nt phasiRNAs from the site of biogenesis to the neighboring cell layers in maize anthers has not been well studied. A major concern with our previous localization analysis for 24-nt phasiRNAs and biogenesis components primarily to the tapetum is that only a few examples were evaluated, and the probes can adhere to the callose coat surrounding the germinal cells, yielding a non-specific signal1. Therefore, in meiotic stage maize anthers, two major questions should be addressed: 1) do meiocytes contain all the components for generating 24-nt phasiRNAs? And, 2) using RNA-seq and small RNA-seq, what is the timing of synthesis and the distribution of these components in anther lobes?
To explore the distribution of 24-nt phasiRNA biogenesis components and the resulting small RNAs, we used laser capture microdissection (LCM) to isolate meiocytes (ME), tapetum (TAP), and other somatic cells (OSC) from two anther developmental stages for low input RNA-seq and sRNA-seq and for a more comprehensive developmental analysis using RT-qPCR. By integrating comprehensive RNA/sRNA profiling and single-molecule or sRNA FISH (smFISH/sRNA-FISH) spatial detection, we demonstrate that TAP contain the most 24-PHAS primary transcripts, whereas ME ultimately accumulate more 24-nt phasiRNAs at meiotic prophase I. We discuss these results in a model in which 24-nt phasiRNAs move from TAP to ME during meiosis.
Results
LCM collections from meiotic stage maize anthers
Control of fertility in maize anthers, the male reproductive floral organs, is important for hybrid seed production as an economic way to improve grain yield. Maize tassels produce hundreds of spikelets, each of which contains an upper and lower floret, and each floret contains three anthers (Fig. 1a). The lower floret stamens develop approximately 1 day later than the upper floret stamens within the same spikelet. At the beginning of meiosis, each anther contains four lobes; each lobe consists of four somatic cell types that are required to support the meiotic cell development (Fig. 1a). The high regularity of maize anther development, numerous anthers, and a detailed staging system allow rapid dissection of carefully staged anthers for cytological and molecular analysis23 (Fig. 1b, c). To explore the spatial distribution of 24-nt phasiRNAs and their pathway components in the W23 inbred (Fig. 2a), LCM was used to isolate three samples (ME, TAP, and OSC) from two stages of meiotic anthers (1.5 mm zygotene and 2.0 mm pachytene) (Supplementary Fig. 1 and Supplementary Table 1). Collections from additional stages were used in RT-qPCR analysis to provide finer detail of the timing of biogenesis of a subset of 24-PHAS precursor and Dcl5 transcripts.
24-PHAS precursors accumulate in tapetal cells
Seventeen RNA-seq libraries were built from ME, TAP, and OSC recovered from 1.5 mm and 2.0 mm stage anthers plus single, fixed anthers (FA) (Supplementary Table 2). To check the reproducibility of the datasets, a correlation heatmap and a Principal Component Analysis (PCA) plot were constructed (Fig. 1d, e). Biological replicates are highly similar and distinct from other cell collections at both stages, confirming the high quality of our LCM sample collections. 126 (TPM > 0) of 176 previously discovered 24-PHAS precursor types were identified in FA and LCM cell sample collections1 (Supplementary Table 3); PHAS precursors that overlapped protein coding genes were excluded. As shown in Fig. 2b, 118 of these loci had sufficient and reliable expression (TPM reads > 5 in at least one sample) for further analysis. Overall, the 24-PHAS precursors were most abundant in TAP at the 1.5 mm stage, showing an expression pattern and high abundance similar to that in 1.5 mm FA (Fig. 2b); therefore, the majority of 24-PHAS precursors in whole anthers are found in the tapetal cells. The average total expression levels of 24-PHAS loci are approximately 65 times lower in ME and 20 times lower in OSC at this zygotene stage (Fig. 2b and Supplementary Table 3). Single cell RNA-seq (scRNA-seq) analysis of individual ME detected a low abundance of 24-PHAS loci during the pre-zygotene/zygotene stages in a A188/B73 maize hybrid stock24. The total average abundance of these 24-PHAS transcripts per zygotene sample in A188/B73 is less than that in the pooled meiocyte LCM samples at 1.5 mm from W23 maize anthers. Furthermore, our previous scRNA-seq data also identified extremely low abundances of 24-PHAS precursors in premeiotic AR and PMC in W23 maize anthers25. These data suggests that the TAP is the main source for 24-PHAS precursors in zygotene stage maize anthers.
To further verify our LCM RNA-seq data and to quantify transcripts in more stages, we randomly selected three highly expressed 24-PHAS precursors (PHAS15, PHAS116, and PHAS296) for smFISH and RT-qPCR assays (Fig. 2c). In the LCM RNA-seq data, transcripts for all three loci were significantly enriched in TAP at the 1.5 mm stage compared to ME and OSC (Supplementary Fig. 2). smFISH images confirmed that TAP contain most of these three 24-PHAS transcripts (yellow dots resulting from dye AF647 labeling) at the 1.25 mm and 1.5 mm stages during late leptotene/pre-zygotene and zygotene stages in W23 anthers (Fig. 2d, g and Supplementary Fig. 3a). Quantification counts (transcript count) demonstrated that TAP is significantly higher at both stages (Fig. 2e, h and Supplementary Fig. 3b). Almost no expression was detected at earlier or later stages (Fig. 2d, e, g, h and Supplementary Fig. 3a, b) or in control hybridization assays (Supplementary Fig. 4). For RT-qPCR using W23 anthers and LCM samples, ZmCyanase was used as the internal reference; it is consistently expressed across all anther stages and is expressed in all cell layers based on smFISH detection (Supplementary Fig. 5) compared to the no label control hybridization test (Supplementary Fig. 6). In 0.75-2.5 mm W23 maize anthers, the expression patterns of the three 24-PHAS loci showed a similar pattern, with high expression levels in both 1.25 mm and 1.5 mm anthers and with significant enrichment in TAP compared to ME and OSC (Fig. 2f, i and Supplementary Fig. 3c).
To further solidify the conclusion that TAP is the primary site of 24-PHAS transcript biogenesis, we utilized the dcl5-mu3 mutant. Several dcl5 mutants have been documented to contain normal 24-PHAS precursor levels but lack virtually all the 24-nt phasiRNAs2. We reasoned that in the absence of the processing machinery, the large 24-PHAS precursors would accumulate in the cells where they were synthesized. This was indeed true as the PHAS116 transcripts were found to be highly enriched in TAP at the 1.75 mm stage (this is a slightly later stage than in W23 anthers) in homozygous dcl5 (dcl5-mu03//dcl5-mu03) anthers (mixed inbred background) using smFISH (Fig. 2j).
Overall, these data demonstrate that maize tapetal cells are the main site for 24-PHAS precursor accumulation and that accumulation starts at early meiotic prophase I and peaks during zygotene. We also established that accumulation is already substantial at the initial stages of meiosis in 1.25 mm anthers, approximately one day earlier than detected previously by analysis of 1.0 mm and 1.5 mm W23 whole anthers1.
The spatial distribution of 24-nt phasiRNA biogenesis pathway components
Argonaute (AGO) proteins, binding partners of miRNAs and phasiRNAs, play essential roles in phasiRNAs biogenesis and RNA silencing (RNAi)26,27. AGOs are assumed to be involved at two steps in 24-nt phasiRNA biogenesis and function: a specific AGO as the binding protein for the miR2275 trigger and one or several AGOs as the binding proteins for the diverse 24-nt phasiRNAs (Fig. 2a). We identified transcripts encoding 18 AGO proteins in LCM cell collections (Fig. 3a). Nine AGO protein transcripts were expressed in one or two cell collections, and the other nine AGO proteins were detected in all three cell collections. None of the AGO isotype transcripts were specific to tapetal cells. Fifteen of 18 AGO members were expressed in TAP at the 1.5 mm stage, providing a list of AGOs that could be involved in 24-nt phasiRNA biogenesis (Fig. 3a). Among these, AGO1 family members have already been designated as candidates in previous studies in maize and rice5,7, 28, and now three AGO members (AGO10b, AGO2b, and AGO5c) not expressed in 1.5 mm TAP can likely be excluded as candidates for 24-nt phasiRNA biogenesis (Fig. 3a). In meiocytes, transcripts were found for 11 AGO proteins, many of which were more highly expressed in ME than other cell types, but none were exclusive to ME (Fig. 3a). Several of these AGO types were predicted in previous transcriptome profiling and spatial localization studies to be potential 24-nt phasiRNA binding partners, such as AGO18 family members1, 5, 29 and AGO2b5 based on the timing of their expression. It is noteworthy that we detected abundant AGO5c transcripts, and these were enriched in ME and absent in TAP (Fig. 3a). AGO5c is encoded by the maize male-sterile28 gene and is homologous to rice MEL130, which is the binding partner for a subset of 21-nt phasiRNAs in rice14.
We investigated the transcript abundance of Dcl5 and Rdr6 in the LCM RNA-seq data. Dcl5 was significantly higher in TAP at the 1.5 mm stage than any other samples (Fig. 3b). In contrast, Rdr6 was expressed in all three cell collections, with slightly higher expression levels in ME, followed by TAP at both stages (Fig. 3b). To precisely pinpoint the timing of expression, RT-qPCR demonstrated that Dcl5 transcripts were significantly higher in TAP samples (Fig. 3c); furthermore, the expression pattern of Dcl5 is similar to the three 24-PHAS precursors (PHAS15, PHAS116, and PHAS296), with the highest expression levels detected in TAP compared to ME or OSC at both 1.25 mm and 1.5 mm stages (Fig. 3c). As shown in Fig. 3d-f with smFISH detection, Dcl5 transcripts (yellow dots resulting from dye Texas Red labeled smFISH probes) were prominently localized in TAP at the 1.25 mm, 1.5 mm, and 1.75 mm stages in either W23 anthers or heterozygous Dcl5//dcl5 fertile anthers. Far fewer yellow dots were detected in ME or OSC, and almost no signal was detected in the negative control samples (Fig. 3f and Supplementary Fig. 6). An additional smFISH analysis of Dcl5 using AF647 labeled smFISH probes confirmed that Dcl5 transcripts were localized mainly in TAP at 1.25 mm and 1.5 mm (Supplementary Fig. 7). These data suggest that TAP is the main site for Dcl5 accumulation at early meiotic prophase I in maize.
Previous studies have reported that four basic helix-loop-helix (bHLH) factors (Ms23, Ms32, bHLH52, and bHLH122) are required for normal tapetal cell specification and differentiation in maize17,31. Ms23 is the master transcription factor (TF) and is required for expression of 24-PHAS precursors, miR2275, and 24-nt phasiRNAs1. Here, Ms23 and bHLH51 were highly expressed in TAP at 2.0 mm, whereas Ms32 and bHLH122 were more abundant in TAP at 1.5 mm (Supplementary Fig. 8). These TF expression patterns are consistent with our previous work in maize anthers, particularly the absence of bHLH122 transcripts at 2.0 mm17. Importantly, the TAP is significantly higher for these four TFs compared with the other two cell collections (Supplementary Fig. 8). Collectively, the distribution of key TFs by RNA-seq and localization data provide further evidence that TAP is the main site for transcription of PHAS precursors and their processing into 24-nt phasiRNAs. Based on the timing of 24-PHAS accumulation, heterodimers of Ms32 and bHLH122 are candidates for direct regulation of 24-PHAS loci. Because Ms23 is essential for expression of bHLH122, it is also a critical factor.
24-nt phasiRNAs are abundant in all three cell collections
To investigate the spatial distribution of 24-nt phasiRNAs in meiotic stage anthers, we collected ME, TAP, and OSC by LCM from 1.5 mm and 2.0 mm W23 maize anthers for low input sRNA-seq (Supplementary Table 1, 2). We first classified sRNAs by length (18 nt to 30 nt) in FA and LCM cell collections. Confirming previous results1, the 24-nt size sRNA class is the most abundant at both stages (Supplementary Fig. 9a). The 21-nt size sRNA class is also detectable, but with lower expression levels (Supplementary Fig. 9a). Using the set of 118 24-PHAS loci for analysis (Fig. 2b), 24-nt phasiRNAs from these loci were more abundant in all three cell collections at 2.0 mm than at the 1.5 mm stage (Fig. 4a and Supplementary Table 3). The expression patterns of 24-nt phasiRNAs and 22-nt miR2275 family members are very similar, with the highest abundance in ME at 2.0 mm during pachytene, followed by TAP at this stage (Fig. 4a). A previous sRNA-FISH co-localization study reported that miR2275 and one 24-nt phasiRNA had higher abundances in AR/ME and TAP in maize anthers32. Here, two of the miR2275 family members (miR2275a and miR2275b) were highly expressed, indicating they might play key roles in 24-PHAS precursor cleavage (Fig. 4b). Although the precursors are most abundant in TAP cells (Fig. 2b, c and Supplementary Fig. 2), the 24-nt phasiRNAs accumulate to higher levels in the ME (Fig. 4a, c and Supplementary Fig. 9b). Individual 24-nt sRNA abundances from products of the same locus (PHAS15, PHAS116, and PHAS296) are not uniform: only a few of the 24-nt sRNAs show high abundance despite the expectation of equal stoichiometry from a single precursor molecule (Fig. 4d, e and Supplementary Fig. 10a). Furthermore, sRNAs localization studies illustrated that 24-nt phasiRNAs from three 24-PHAS loci (PHAS15, PHAS116, and PHAS296) were expressed in all three cell types, with higher expression levels in ME and TAP, lower in OSC from 1.25 mm to 2.0 mm (Fig. 4f, g and Supplementary Fig. 10b). Almost zero expression, considered to be background, was detected in control hybridization assays (no labeled control and mouse HKH RNA control) (Fig. 4f, g and Supplementary Fig. 10c). These distribution patterns lead us to hypothesize that the TAP is the site for nearly all 24-nt phasiRNAs biogenesis and that the presence of 24-nt phasiRNAs in ME and OSC depends on transport during early meiotic prophase I in maize.
Discussion
In anthers, TAP is the innermost of the four somatic layers, adjacent to the germinal cells. The TAP plays essential roles in providing nutrients, enzymes, and pollen wall components to support AR, ME, and microspore (MSP) development. Defects in tapetal function or the abnormal differentiation of TAP usually leads to meiotic arrest or to later pollen abortion2,17,31,33-35. 24-nt phasiRNAs have been reported as abundant in meiotic stage maize anthers from hundreds of unique copy loci1. In TAP-defective, male-sterile mutants lacking key tapetal-expressed TFs, almost all the 24-nt phasiRNAs are missing1,17, which indicates that TAP is likely to be involved in 24-nt phasiRNA biogenesis. Our low-input RNA-seq data for ME, TAP, and OSC demonstrated that the TAP is the main source for the 118 24-PHAS precursor transcripts as well as Dcl5 transcripts in 1.5 mm stage anthers (Fig. 2b, c and Fig. 3b), confirming a previous study in the same W23 background that 1.5 mm anthers were the peak stage for 24-PHAS loci production with 1.0 mm anthers lacking expression1. We further refined the timing using smFISH and RT-qPCR assays, demonstrating that three selected 24-PHAS loci (PHAS15, PHAS116, and PHAS296) and Dcl5 transcripts were initially expressed at 1.25 mm during the leptotene-zygotene transition and were continuously expressed at 1.5 mm or 1.75 mm during zygotene (Fig. 2d-i, 3d-f and Supplementary Fig. 3, 7). Moreover, the earlier timing for 24-nt phasiRNA generation corresponds to the stage just prior to and including the leptotene-zygotene transition (Fig. 4f, g and Supplementary Fig. 10b), which is a key checkpoint in meiosis. A previous study has reported that CHH DNA methylation is elevated in isolated ME, specifically at the phasiRNA-producing loci18. Therefore, a logical hypothesis is that some of the 24-nt phasiRNAs generated in the TAP and transported to ME might induce methylation in the corresponding loci. It is unclear how such cis methylation could contribute to meiotic regulation. Furthermore, in dcl5 mutants, meiosis is normal in the near complete absence of 24-nt phasiRNAs2, indicating that whatever role the 24-nt phasiRNAs play, it is dispensable.
By using LCM-purified cell collections, our data demonstrate that TAP cells are the primary site of 24-PHAS precursor accumulation and 24-nt phasiRNA biogenesis. sRNA-seq data for ME, TAP, and OSC illustrated that 24-nt phasiRNAs and miR2275 family members were found in all three cell collections, with higher expression levels in ME at 2.0 mm during pachytene, followed by TAP at the same stage (Fig. 4a-c). Furthermore, sRNA-FISH detection confirmed that the 24-nt phasiRNAs originating from three 24-PHAS loci (PHAS15, PHAS116, and PHAS296) were present in all three cell layers from 1.25 mm to 2.0 mm (Fig. 4f, g and Supplementary Fig. 10b). Our major new conclusion is the proposal that 24-nt phasiRNAs generated in the TAP move to other cell types. This conclusion is summarized in a model of 24-phasiRNA biogenesis and mobility in maize anthers during meiotic prophase I (Fig. 5). We demonstrated that TAP is the main site of 24-nt phasiRNAs biogenesis, as TAP contain the most of 24-PHAS precursors, Dcl5 transcripts, adequate miR2275, and Rdr6. To understand why ME later contain the highest levels of 24-nt phasiRNAs, we note that there are approximately 30 tapetal cells touching each ME, whereas only approximately 0.6 tapetal cell touch neighboring OSC23. Utilizing the LCM cell numbers we recovered and the total 24-nt phasiRNA reads we generated (Supplementary Table 1, 3b), we calculated that each tapetal cell “donates” approximately 5% of its 24-nt phasiRNAs to ME or OSC during meiotic prophase I in maize anthers. If each TAP cells exports 5% (5 units) of its 24-nt phasiRNAs to ME cells, each ME would have 5 units x 30 = 150 units of 24-nt phasiRNAs, therefore reaching a higher level than the source TAP cells. In contrast, export of 5 units of 24-nt phasiRNAs to the OSC, would result in a low level of 24-nt phasiRNAs in the immediate receiving middle layer cell with further dilution as 24-nt phasiRNAs move throughout the OSC group. A challenge for future research will be documenting the path of 24-nt phasiRNA movement. Other cases of plant intercellular movement of sRNAs are probably through PD, regulating tissue or organ development by genes silencing and RdDM22. PD are membrane-lined intercellular channels that connect the cytoplasms of adjacent cells, such as between TAP and PMC/ME or non-TAP somatic cells in anthers36,37. Therefore, we speculate that 24-nt phasiRNAs move from the biogenesis site in TAP to ME or OSC through PD in meiotic prophase I maize anthers.
In Arabidopsis, tapetal-derived 24-nt siRNAs induce DNA methylation in the germline and maintain genome integrity by silencing corresponding genes and transposons38. Here, 24-nt phasiRNAs are generated in TAP and then exported to ME (Fig. 5). These exported 24-nt phasiRNAs are likely required for the elevated CHH DNA methylation at 24-PHAS loci in isolated ME in maize18. In Arabidopsis, the 24-nt siRNAs appear to have a broader role in modifying the ME genome and gene expression than the 24-nt phasiRNAs have in maize. This is particularly true for any impact 24-nt phasiRNAs might have on gene expression, because no mRNA targets for this class of small RNAs have been identified. Nonetheless, it is interesting that small RNAs synthesized in the TAP are transferred to maize ME. Analysis of dcl5 maize mutants establishes that in the absence of 24-nt phasiRNA biogenesis, tapetal cells fail to differentiate properly at normal maize growing temperatures (28 ℃), while meiosis proceeds normally2. These observations indicate that the 24-nt phasiRNAs play an essential role in the TAP -- not in ME -- during the meiotic stages; later, microspore failure renders dcl5 mutants male-sterile, a fact that we attribute to the failure of the tapetum to supply nutrients and wall components to the developing uninucleate microspores. Mutants grown at low temperature (21 ℃) are fertile and tapetal development is slow but sufficient indicating that 24-nt phasiRNAs are not required under these conditions. Considering these observations in toto, it is possible that in maize the transfer of 24-nt phasiRNAs to meiocytes is an evolutionary relic that no longer serves (or fully serves) a required function and that the primary role of the 24-nt phasiRNAs is to support rapid tapetal differentiation.
Methods
Plant materials
Maize (Zea mays) W23 bz2 inbred line plants were grown in Stanford, CA, USA under greenhouse conditions (31 ℃/21 ℃, 14 h day/10 h night light cycle with supplemental LED and UV-A lamps providing ∼50% of summer sun conditions). Anthers were dissected from tassels and then measured using a micrometer (Fisher Scientific).
Homozygous dcl5-mu03//dcl5-mu03 plants are temperature-sensitive male sterile. They are male sterile under optimized growth condition (28 ºC/22 ºC day/night temperature, 14 h light, 400 µmol/m2/s) in the greenhouse. dcl5-mu03//dcl5-mu03 male sterile and dcl5-mu03//Dcl5 fertile siblings were grown in St. Louis, MO, USA under the optimized growth conditions in a greenhouse.
Laser Capture Microdissection (LCM)
ME, TAP, and OSC were isolated from 1.25 mm, 1.5 mm, and 2.0 mm W23 anthers by LCM as described previously39, with minor modification. Fixed anthers were cryoprotected in 10% (2-3 days) and 15% sucrose/PBS (phosphate buffered saline) (5 to 7 days) to better preserve tapetal cell layer morphology. Total RNA was isolated from each cell type using the RNAqueous™-Micro Kit (Invitrogen); RNA quality was evaluated on an Agilent Bioanalyzer using the RNA 6000 Pico Kit (Agilent Technologies).
Cytological staging and imaging
AR/PMC/ME from W23 anthers (from the 0.75 mm through 2.25 mm anther length stages, recovered at 0.25 mm intervals) were extruded, classified using a micrometer, and then appropriately sized anthers were stained with 10 μg/mL Hoechst 33342 (Sigma-Aldrich) according to our pervious report25. Confocal image stacks for the germinal cells were taken with a Leica SP8 microscope using a 93X Glycerol immersion objective. ImageJ40 was used for processing the fluorescent microscopy images. Cytological stages during meiotic prophase I were assigned using established criteria25,41.
RNA-seq and sRNA-seq library construction and sequencing
Total RNA was isolated from three different LCM cell samples and FA samples at 1.5 mm and 2.0 mm using the RNAqueous™-Micro Kit (Invitrogen). RNA-seq libraries were prepared using CEL-seq242 based on a modified protocol25. Primers were synthesized by phosphoramidite chemistry at the Stanford Protein and Nucleic Acid Facility (Stanford, CA) (Supplementary Table 4). All RNA-seq libraries were sequenced on an Illumina HiSeq 4000 instrument at the Stanford Genome Sequencing Service Center (Stanford, CA) with paired-end 150 bp reads.
For small RNA library preparation, RealSeq®-AC kits (SomaGenics) were used for ultra-low amounts of total RNA starting with 100 ng from whole anthers or 10 ng from LCM cell samples. Sequencing in a single-end mode on an Illumina NextSeq (University of Delaware) yielded 75 bp reads for sRNA-seq.
Data handling and bioinformatics
Paired-end RNA-seq Read1 contained a 10 bp unique molecular identifier (UMI)43 and sample barcodes, and Read2 contained the transcript sequence. Each UMI was processed and attached to the Read2 metadata using fastp44. Barcodes were used to demultiplexed Read2 into a separate file for each library using fastq-multx45. After trimming the adapters by Trim Galore, the reads were mapped to version 4 of the B73 maize genome using Hisat246. RNA-seq reads were normalized by dividing by the total number of transcripts (UMIs) in each sample and multiplying by one million (transcripts per million normalization).
sRNA-seq data were processed as previously described47. In brief, Trimmomatic version 0.3248 was used to remove the linker adaptor sequences. Then, the trimmed reads were mapped to version 4 of the B73 maize genome using Bowtie49. Read counts were normalized to one million to allow for the direct comparison across libraries. PhasiRNAs were designated based on the criteria of a 24-nt length and mapping coordinates within the previously identified 176 24-PHAS loci1 that were updated to version 4 of the B73 genome using the assembly converter tool50.
Quantitative reverse transcription PCR (RT-qPCR)
RT-qPCR for multiple stage anthers (0.75 mm – 2.0 mm; 0.25 mm intervals) and LCM cell sample collections (ME, TAP, and OSC) was performed using the Luna® Universal Probe One-Step RT-qPCR Kit (New England Biolabs). Quantitative PCR was performed using TaqMan primers synthesized by Integrated DNA Technologies (Table S5) on a CFX96 C1000 Touch Real-Time PCR Detection System (BioRad). ZmCyanase expression was used to normalize among biological samples, because this gene is highly expressed at all stages of anther development (Supplementary Fig. 5). Each sample type was tested in two biological and three technical replicates.
Single-molecule fluorescence in situ hybridization (smFISH) for mRNAs and sRNAs
Maize spikelets of various sizes were dissected from the central spike and fixed in 20 mL glass vials using 4% paraformaldehyde in 1x PHEM buffer (5 mM HEPES, 60 mM PIPES, 10 mM EGTA, 2 mM MgCl2 pH 7). Fixation was done three times in a vacuum chamber at 0.08 MPa, 15 min each. After fixation, samples were sent for paraffin embedding at the histology lab in the Nemours/Alfred I. duPont Hospital for Children (Wilmington, DE). Embedded paraffin samples were sectioned using a paraffin microtome and dried on poly-l-lysine-coated 22 × 22 mm #1.5 coverslips (Carl Zeiss Microscopy, LLC, Cat# 474030-9020-000). Samples were then de-paraffinized using Histo-Clear (Fisher Scientific, 50-899-90147) and re-hydrated by going through an ethanol series of 100, 95, 80, 70, 50, 30, 10% (vol/vol) (30 sec each) and finally water (1 min) at room temperature. After protease (Sigma, P5147) digestion (20 min, 37°C), samples were neutralized in 0.2% glycine (Sigma-Aldrich, G8898) for 2 min. After two washes in 1x PBS buffer, samples were dehydrated and then hybridized with smFISH probes. About 20-nt smFISH probes were designed to bind specifically across the length of each target RNA in a non-overlapping series51. For small RNA detection, probes were designed reverse complemented to ∼19 nt of the phasiRNA sequences generated from each locus. Briefly, 23-51 probes (total coverage of about 800 nt for mRNAs or sRNAs) were designed for each target, and each probe was synthesized with a 3’ amino modification (LGC Biosearch Technologies, CA) (Supplementary Table 6). All the probes of one set were pooled and en masse coupled with Alexa Fluor 647 (Thermo Fisher) or Texas Red; the labeled probe fraction was purified using reverse phase HPLC52. The probes were diluted to a concentration of 20 ng/ul in a hybridization buffer containing 10% formamide, 1 mg/mL yeast tRNA, 10% dextran sulfate and 2 mM Vanadyl ribonucleoside complex (New England Biolab, Ipswich, MA, Catalogue Number, S142), 0.02% RNase free BSA (Ambion, AM2618). The biological samples were hybridized with hybridization mix overnight in a humid chamber at 37°C. The samples were then washed two times with 2X SSC buffer (sodium saline citrate) containing 10% formamide, and a final wash was done using 1X TBS buffer (Tris-buffered saline). Samples were mounted using ProLong™ Glass Antifade Mountant with NucBlue™ Stain (ThermoFisher Scientific, P36981).
Fluorescence was detected by spectral unmixing of autofluorescence spectra using laser scanning confocal microscopy on a Zeiss LSM 880 multiphoton confocal microscope. We used 30% power of the 633 nm laser and an alpha Plan-Apochromat 100x/1.46 oil lens to acquire RNA in situ images. Pure dye was used as positive control, and autofluorescence from maize tissue was used as the negative control for spectral bleed-through. After images were acquired, each image was spectra unmixed. The brightness and contrast of images in the same figure panel were adjusted equally and linearly in Zen 2010 (Carl Zeiss). For quantification, the numbers for each localization event were calculated using ImageJ (SpotQuant). Two replicates for each Z-stack of images were used for calculating the copy number for each gene’s transcripts.
Author contributions
X.Z., B.C.M, and V.W. conceived and designed the project. X.Z. performed most of the experiments and bioinformatics. K.H. and M.B. designed and synthesized smFISH probes. K.H. performed smFISH and imaging analysis. T.C. constructed sRNA libraries and performed raw data processing. X.Z. and V.W wrote the manuscript with editing by B.C.M. and K.H. and with contributions from all co-authors.
Competing interests
The authors declare no competing interests.
Supplementary information
Supplementary Figures
Supplementary Fig. 1. Sequential application of LCM to isolate ME, TAP, and OSC in 1.5 mm and 2.0 mm W23 anthers.
Supplementary Fig. 2. Transcript abundances of three 24-PHAS loci (PHAS15, PHAS116, and PHAS296) in LCM cell sample collections in RNA-seq data.
Supplementary Fig. 3. PHAS15 localization in maize anthers.
Supplementary Fig. 4. Negative control hybridization assay.
Supplementary Fig. 5. ZmCyanase localization in maize anthers.
Supplementary Fig. 6. No labeled control hybridization assay.
Supplementary Fig. 7. Dcl5 localization in maize anthers.
Supplementary Fig. 8. The distribution of four tapetal-enriched TFs (Ms23, Ms32, bHLH51, and bHLH122) in LCM cell sample collections at 1.5 mm and 2.0 mm in RNA-seq data.
Supplementary Fig. 9. sRNA distribution in maize anthers.
Supplementary Fig. 10. 24-nt phasiRNAs localization in maize anthers.
Supplementary Tables
Supplementary Table 1. List of LCM cell sample collections from 1.5 mm and 2.0 mm maize anthers.
Supplemental Table 2. Summary of RNA-seq and sRNA-seq libraries prepared from FA and LCM cell sample collections.
Supplementary Table 3. Coordinates and abundance of the 126 24-PHAS loci and their corresponding 24-nt phasiRNAs in FA and LCM cells.
Supplementary Table 4. Primer sequences for CEL-seq 2 library preparation.
Supplementary Table 5. Primer sequences for RT-qPCR by the TaqMan assay.
Supplementary Table 6. Probes sequences (5′-> 3′) used for smFISH or sRNA-FISH.
Acknowledgements
We thank M. Zhang for initial suggestions on project design and for instructing X. Zhou in the LCM technique; B. Nelms for help with low input RNA-seq library construction and raw data processing; Y. Chan and Y. Liu for help in data handling; J. Caplan for helpful advice on smFISH; J. Zhan assisted retrieval and organization of the sequences of 24-PHAS precursors and 24-nt phasiRNAs; and J. Dinneny for use of the SP8 confocal microscope. This project was supported by National Science Foundation award 17540974.
Footnotes
↵* e-mail: bmeyers{at}danforthcenter.org; walbot{at}stanford.edu
During the last submission, we found the authors order was not correct. So, we modified the authors order as same as the ones in the manuscript during this submission. Thanks!