Refining the criteria for meiotic gene identification by profiling studied genes that function in meiosis
68 genes in Arabidopsis have been identified and characterized with function in meiosis using forward and reverse genetic approaches (Additional file 1, Table S1). The 68 known Arabidopsis genes with functions in meiosis can be assigned to four functional categories: 1) genes which function in homologous chromosome pairing and recombination (e.g., AtSPO11-1, AtDMC1, AtRAD51, AtRAD51c, AtXRCC3, AtMSH4, AtMER3/RCK) [40, 45, 46, 55]; 2) genes that encode structural proteins such as cohesin, histone, centromere proteins and proteins for synaptonemal complex assembly, e.g., SYN1/DIF1, SMC1, ZYP1, and ASY1 [62–66]; 3) genes which function in chromosome spindle organization and movement (e.g. ATK1, AESP, ATK5/AtKin14B, AtPS1) [58, 67–70]; and 4) genes which encode regulatory proteins such as MMD1/DUET, SDS, TAM, and ASK1 [54, 71–74]. By profiling the gene expression of all 68 genes in meiocytes and control tissues, meiosis-specific candidate genes can be efficiently identified by following the two criteria: 1) genes are expressed at twofold or greater in meiocytes versus anthers; and/or 2) genes are expressed at twofold or greater in both meiocytes and anthers versus seedlings with the exclusion of genes that are expressed at fourfold or greater in anthers versus meiocytes.
Three genes, MND1, AtSRP2, AtSRP3, were expressed at very low levels and only had 0.7 read per million reads in meiocytes, which suggests more sequence data or a lower cutoff point is needed in order to cover all important meiosis-specific genes for these datasets [33, 35, 36]. Only 4 genes (AtSPO11-2, MS5, MND1 and AtSRP2)[32, 33, 36, 75] meet the first criterion with preferential expression in meiocytes, and 29 genes were expressed twofold or greater in both meiocytes and anthers comparing to seedlings, which include key meiotic recombination genes, such as AtSPO11-1, AtDMC1, ASY1, AtMLH3, AtRAD51C, AtXRCC3, AtMSH4, AtMSH5, AtMER3/RCK, PTD, AtMUS81, and SDS [49, 76–78] (Additional file 1, Table S1). Genes that do not meet the two criteria are unlikely to be meiosis-specific; for example, ATK5/KIN14B may also have important roles in mitotic cell division [58, 68]. AtRAD51 has a meiosis-specific function in Arabidopsis, but it is expressed in both meiocytes and non-meiotic somatic cells , which is consistent to the RAD51 gene expression in other organisms, such as mice B cells . Most of the meiosis-specific genes, especially for those in the meiotic recombination pathways could be identified by comparative analysis of transcriptome profiles of meiocytes, anthers and seedlings [6–8, 55] (Additional file 1, Table S1).
A visual representation of all statistically different genes from the ANOVA is presented as in the Ward Hierarchical Clustering, which suggests a number of genes that are uniquely expressed in the anther as well as the meiocyte that may be unique candidate genes for control of expression (Figure 3). As shown in Figure 4, with a cutoff point of 5 reads per million reads in at least one sample, more than 1,000 candidate meiosis-specific genes were identified through this approach with an additional 607 genes that are preferentially expressed in meiocytes (Figure 4).
The main purpose of profiling concentrated meiocytes is to eliminate genes that are expressed and function in anther wall development, which is critical, because those genes would be included in the candidate gene pools for entire anther development if transcriptomic analysis is performed using anther materials . A criterion we suggest for meiotic gene identification is to exclude genes that are expressed at fourfold or greater in anthers versus meiocytes, although the gene expression level may be up-regulated in both meiocytes and anthers versus seedlings. For example, DYT1 was found in a gene pool by profiling anther transcriptome compared to other organs, and functions in regulating anther wall development [16, 80]. Here we show that DYT1 was read at 7.8, 89.3, 0.0 reads per million reads in meiocytes, anthers, and seedlings, respectively. The differential expression indicates that DYT1 is specifically expressed in anthers/meiocytes with a significantly preferential expression in anthers, which is consistent to its function in anther wall development . Another example is the ATA1 gene that was reported to be highly expressed in tapetum , and the RNA-Seq results read at 33.4, 564.6, 0.0 reads per million reads in meiocytes, anthers and seedlings, respectively. Both DYT1 and ATA1 were preferentially expressed in anthers versus meiocytes, which implies the feasibility of excluding the anther wall genes by comparative analysis of transcriptome profiling of meiocytes, anthers and seedlings. As indicated in criterion 2 for meiosis-specific gene identification, genes that are expressed at fourfold or greater in anthers versus meiosis should be considered as non-meiosis-specific candidate genes, or candidate genes for anther wall development.
Although this study has not included a parallel transcriptome study of microspore/gametophyte, the life after meiosis, a comparison of meiocytes (transition from diploid sporophyte to haploid gametophyte) should advance our understanding of the molecular connections between the two key processes of reproduction development, as well as promoting the means of identification of meiosis-specific genes. Previously, pollen transcriptome profiling using microarrays have found 7,235 genes expressed in Arabidopsis Landsberg erecta with 387 pollen-specific  and 6,587 expressed in Arabidopsis Col-0 ecotype. Since the meiocytes we collected included tetrads, there is likely to be a significant overlap between meiocytes and microspores. A further deep transcriptome sequencing using staged meiosis-I meiocytes is currently being performed.
Transposable element genes in meiosis
While transposable elements (TEs) make up to 14% of Arabidopsis genome , the majority of TEs were silenced during plant development since there were a lack of mRNAs, but higher levels of small RNAs were detected . TEs' activities were usually limited in just one or a few of developmental stages, in which TEs were expressed . In comparing the transcriptomes of anther and meiocytes, we observed a large set of TEs that were expressed preferentially or specifically in meiocytes versus anthers. At a cutoff point of one read per million reads, a total of 1,271 TE genes were expressed in meiocytes, which is about 32.5% of 3,907 TE genes reported or annotated in Arabidopsis  [http://www.arabidopsis.org]. Relatively smaller numbers of TE genes were expressed in controls: 379 and 138 in anthers and seedlings, respectively. With 1,036 TE genes up-regulated and 81 TE genes down-regulated in meiocytes compared to anthers, TE genes may play unique roles in meiosis. In addition, 1,165 TE genes were up-regulated and 48 TE genes down-regulated in meiocytes as compared to seedlings, which are consistent with the comparison between meiocytes and anthers, and demonstrated more significant deviation between meiocytes and seedlings (Additional file 5, Table S3, and Additional file 8, Table S4).
The abundant TE expression in meiocytes suggests substantial activities of TEs in meiosis. It is believed that TEs affect recombination in all meiotic eukaryotes . Recent studies on postmeiotic gametophyte development have found that TE genes are unexpectedly reactivated and transpose only in the vegetative nucleus, but not in the sperm cells of pollen , which suggest that small interfering RNAs (siRNAs) from TEs activated in the vegetative nucleus can target silencing those in gametes . In addition, small RNA pathways have been found to be present and functional in the angiosperm male gametophytes . During female gamete formation, AGO9 was found to preferentially interact with siRNAs derived from TEs and the activity of AGO9 is required to silence TEs in female gametes and their accessory cells . The AGO9 gene (At5G21150) was preferentially expressed in anthers versus meiocytes (M/A = 78.24/172.37) in this study, which is consistent with the discovery of a postmeiotic function. In contrast, the two AGO genes, At1G31290 (AGO3) and At5G21030 (AGO8) were preferentially or specifically expressed in meiocytes (At1G31290: M/A = 11.54/3.30; At5G21030: M/A = 3.03/0.71), which suggest that molecules regulating gene silencing and DNA modification in meiosis differ from those of postmeiotic gametophyte development, both in the male and the female. In the postmeiotic gametes, the gene expression map has demonstrated the similarities between plants and animals , which may also be true in meiosis. To date, it still remains largely unknown how TEs function in meiosis. It is possible that a large number of TEs are activated in meiosis and then silenced after meiosis through siRNA machinery and/or modification of heterochromatin.