In flowering plants, male reproductive organs are called stamens, each of which consists of a filament and an anther . Cells in the anther undergo meiosis to produce microspores, which further develop into mature pollen grains . Therefore, anther development is critical to achieve pollen formation and subsequent success of fertilization [3–6]. According to morphological features, anther development can be grouped into two phases and then be further divided into 14 anther stages [5, 7, 8]. At the beginning of phase 1 (anther stages 1 to 8), the stamen primordium has 3 layers, L1-L3 from surface to interior. The L1 cells later become the epidermis and the L3 cells give rise to the vascular and connective tissues. Some of the L2 cells develop into archesporial cells which then divide into parietal cells and primary sporogenous cells. Additional cell division and differentiation in the L2-lineage establish a characteristic four-lobed structure at anther stage 5. Each lobe consists of central pollen mother cells surrounded by outer endothecium, middle layer and inner tapetum. Pollen mother cells undergo meiosis at stage 5-6, producing tetrads at stage 7. Dissolution of the tetrad callose wall releases microspores at stage 8. In phase 2, the microspores undergo mitosis and develop into mature pollen grains during stages 9-12. Meanwhile, pollen wall materials are deposited from both the microspores and the tapetum layer. After the degeneration of tapetum, the mature pollen is released and is able to start pollination.
Previous studies indicated that early anther development depends on transcriptional regulation and cell-cell communication [5, 7–9]. The SPOROCYTELESS (SPL)/NOZZLE (NZZ) gene is one of the earliest genes that regulate anther cell fate determination [10, 11]. SPL/NZZ is activated by AG, a C function gene in the ABC model [12–14]. SPL/NZZ is expressed as early as anther stage 2-5 and a mutation in SPL/NZZ leads to the failure of differentiation of parietal and sporogenous cells, and consequentially blocks the formation of anther wall and microsporocytes [15, 16].
EXCESS MALE SPOROCYTES1 (EMS1) and TAPETUM DETERMINANT1 (TPD1) are also essential for male fertility with a later expression peak at stage 5 . EMS1 is a leucine-rich repeat receptor-like protein kinase (LRR-RLKs) and TPD1 is likely its ligand [15, 18, 19]. In both ems1 and tpd1 mutants, anthers produce more microsporocytes at the expense of the tapetum, indicating that communication between adjacent cell layers determines the cell fate of archesporial cell progenies in order to form normal anther wall . Besides EMS1 and TPD1, other cell-cell communication-related genes are also involved in anther development, such as SOMATIC EMBRYOGENESIS RECEPTORLIKE KINASES1/2 (SERK1/2), and RECEPTORLIKE PROTEIN KINASE2 (RPK2) [20, 21].
Upon the formation of the anther lobes, DYSFUNCTIONAL TAPETUM1 (DYT1) and AMS, encoding two bHLH transcription factors, are required for tapetal functions at subsequent stages [22, 23]. In dyt1, tapetum cells harbor enlarged vacuoles and reduced cytoplasm. The dyt1 meiocytes have comparatively thin callose walls, cannot complete cytokinesis and finally collapse. RNA in situ hybridization experiments showed that DYT1 reaches its peak expression at anther stage 5 to 6 . AMS functions near the time of meiosis, slightly later than that of DYT1. In the ams mutant, the microsporocytes can complete meiosis but the tapetum cells prematurely collapse and microspores are degraded before the first pollen mitosis . Beside these regulators, a large number of other genes are also expressed in the anther, and mutations in some of them lead to male sterility by affecting early anther cell formation, tapetum formation, meiosis or pollen maturation [5, 7, 16, 24–28].
However, due to the functional redundancy of members of many gene families, the subtleties of the phenotypes of single-gene mutants, and possible early phenotypes that obscure anther function, forward genetics has limitations in uncovering anther gene functions . Expression profiling has become increasingly informative and might circumvent the limitation of forward genetics. In recent years, global gene expression profiling by microarray has been used to detect floral gene expression and obtain clues for understanding reproductive development. However, most studies to investigate stamen expression profiles have been conducted by analyzing transcripts from the whole inflorescences of male sterile mutants [30–35], rather than the anther itself . Little transcriptomic information about specific organs is currently available, especially for Arabidopsis whose male reproductive organs are quite tiny [32, 33, 36]. Thus the detection of anther-specific or preferential genes in mixed floral tissues might be hampered by the moderate detection sensitivity of microarray technology. As mentioned above, SPL, EMS1 and AMS have important functions at different stages of anther development, although they have temporal overlap of expression [10, 17, 22, 23]. Therefore, analysis of their shared and distinct effects on the anther transcriptome can shed some light on gene regulatory networks [37–39].
To obtain more information on transcriptomes near the stage of meiosis, we collected anthers at stage 4 to 7 from ams mutants and wild-type Arabidopsis, even though it is time consuming and technically difficult to dissect developing anthers, because we wanted to identify the genes affected by the ams mutation that might be too diluted to detect using RNAs from whole-inflorescences. The ams transcriptome data and comparison with previous data from spl and ems1 anthers  provide detailed information on early anther development. Additionally, with known information of other floral organs in Arabidopsis, we identified genes that function during early anther stage around meiosis. We found that many transcription factor genes were preferentially expressed during early anther development, such as bHLH, MYB, and MADS. Closely related homologs were hypothesized to have either redundant or divergent functions according to phylogenic studies [40–42]. Moreover, further investigation of organ-specific transcriptome revealed the importance of both anther-specific and non-specific transcription factors in early anther development. We propose an expanded gene regulatory network that contributes to the precise regulation of temporal and spatial events during early anther development.