Maize multiple archesporial cells 1 (mac1), an ortholog of rice TDL1A, modulates cell proliferation and identity in early anther development Free

In animals, primordial germ cells or pluripotent reproductive cells arise during embryo development; however, plant reproductive organs and germ cells differentiate late (Bhatt et al., 2001). During flower development, germ cells arise de novo from sub-epidermal somatic cells (Ma, 2005). In ovules, only a single megaspore mother cell develops and only one of its four meiotic products survives to produce the embryo sac. By contrast, numerous pollen mother cells (PMC) differentiate in anthers, and these undergo meiosis to give rise to tetrads of haploid spores. Although these processes are indispensable for plant sexual reproduction, it remains unclear how germ cells acquire their fates and how sporogenesis is coordinated with somatic development.

A mature maize anther contains four anther lobes each with four somatic layers – epidermis (EP), endothecium (EN), middle layer (ML) and tapetum (TA) – that encircle the PMC (Bedinger and Fowler, 2009). The floral meristem of plants consists of three layers, designated L1, L2 and L3 (Poethig, 1987). Within anther lobes, the L1 differentiates as EP, and the L2 forms the PMC and the inner somatic cell layers (Fig. 1). Analysis of anthers in transverse sections from diverse flowering plants demonstrates a very regular pattern of cell division. Based on this regularity and the absence of cell movement, cell lineage models are proposed to explain development (Kiesselbach, 1949; Ma, 2005; Feng and Dickinson, 2010a). In the classical lineage model, a single L2-derived (L2-d) hypodermal cell in each locule divides periclinally to generate one primary sporogenous cell and one primary parietal (somatic) cell. The latter divides to form a complete layer encircling the sporogenous cell with subsequent periclinal divisions, yielding the three rings of somatic cell types. Centrally, the primary sporogenous cell, also known as an archesporial (AR) cell, proliferates to form a mass of PMCs, and then produces microspores through meiosis.

Detailed confocal microscopic analysis of maize anthers is not consistent with a strict cell lineage model. In anthers smaller than 170 μm, dozens of similar sized L2-d cells beneath the EP in a lobe were arranged in a disorganized manner with no evidence for a hypodermal cell or an early division to separate the somatic and germinal lineages. Then, AR cells arise within a group of ~100 L2-derived somatic cells as enlarged germinal cells (Kelliher and Walbot, 2011). A later step in the lineage model was substantiated by confocal reconstruction: the secondary parietal layer divides periclinally to generate the ML next to the EN, and the TA next to the AR cells.

Recent molecular genetic studies have also addressed plant germ cell fate determination (Feng and Dickinson, 2010a; Ma and Sundaresan, 2010). Several Arabidopsis thaliana genes crucial for this process have been isolated and characterized: SPL/NZZ (SPOROCYTELESS/NOZZLE), EMS1/EXS (EXCESS MALE SPOROCYTES1/EXTRA SPOROGENOUS CELLS), TPD1 (TAPETUM DETERMINANT1) and SERK1/2 (SOMATIC EMBRYO RECEPTOR KINASE 1/2) (Schiefthaler et al., 1999; Yang et al., 1999; Canales et al., 2002; Zhao et al., 2002; Yang et al., 2003; Albrecht et al., 2005; Colcombet et al., 2005; Yang et al., 2005; Feng and Dickinson, 2010a). ems1/exs and tpd1 mutants and serk1/serk2 double mutants display the same phenotypes of excess PMCs and absence of the TA (or both TA and ML) (Canales et al., 2002; Zhao et al., 2002; Colcombet et al., 2005; Feng and Dickinson, 2010a). These genes are considered to function in the same pathway to regulate cell fate determination. EMS1/EXS, SERK1 and SERK2 encode membrane-localized leucine-rich-repeat receptor-like protein kinases (LRR-RLK), and TPD1 encodes a putative ligand. EMS1/EXS interacts with TPD1 in vitro and in vivo (Yang et al., 2005; Jia et al., 2008). Based on these findings, cell-cell communication has been proposed as essential for anther cell fate determination and differentiation. It has been speculated that SERK1/2 and EMS1/EXS form heterodimeric receptors (Colcombet et al., 2005), but the ability of TPD1 to bind heterodimers is unknown. The precise timing and cellular location of protein expression of the proposed signaling pathway components are also unclear. Nevertheless, two models to explain anther cell specification have been proposed. In the first model, these proteins specify tapetal fate late, after the secondary parietal division produces tapetal precursors; in mutants, those precursors fail to acquire TA fate and instead differentiate as PMCs (Zhao et al., 2002), resulting in an excess number of PMCs. In the second model, the signaling pathway is proposed to act earlier and in its absence there are excess PMCs resulting from additional cell divisions in the L2. Later, the excess PMC affect somatic cell differentiation (Canales et al., 2002). The coordination mechanism proposed in both models is that TPD1 is secreted by PMC to control the developmental fate of neighboring TA (model 1) or precursor somatic (model 2) cells by binding to the EMS1/EXS receptor (Feng and Dickinson, 2010a; Ma and Sundaresan, 2010). Two major issues remain unaddressed to test these models: (1) there is no experimental evidence to support TPD1 secretion; and (2) the protein localizations of TPD1 and EMS1/EXS are unknown early in anther development.

In rice, MSP1 (MULTIPLE SPOROCYTE1), an LRR-RLK, interacts with TDL1A (TPD1-like 1A) protein, its predicted ligand (Nonomura et al., 2003; Zhao et al., 2008). Rice MSP1 and TDL1A genes are proposed to be homologs of Arabidopsis EMS1/EXS and TPD1, respectively. Interestingly, there are phenotypic differences between rice and Arabidopsis mutants. In rice msp1 plants, both anther and ovule are affected (Nonomura et al., 2003), whereas no female phenotypes were reported in Arabidopsis. Although rice msp1 anthers are phenotypically similar to Arabidopsis ems1/exs and tpd1 mutants, an RNA interference (RNAi) line directed against the rice TDL1A gene only has ovule defects (Zhao et al., 2008). These distinctions suggest that the proposed signaling pathway may have species-specific characteristics.

Maize multiple archesporial cells 1 (mac1) mutants share the key phenotype of extra archesporial cells in both ovule and anther (Sheridan et al., 1996; Sheridan et al., 1999). mac1 anthers contain just a single layer of somatic L2-d cells. Here, we report the cloning of mac1, which encodes a small secreted protein. Based on its sequence similarity and its mutant phenotype, maize mac1 is the ortholog of rice TDL1A. In this study, we focused on the role of MAC1 during early anther development through characterizing the mutant phenotype, defining MAC1 protein localization and conducting double mutant analyses with other mutants that affect either anther development or meiosis. Our data indicate that MAC1 functions very early in anther development, prior to tapetal formation. MAC1 regulates the number of AR cells by preventing their overproliferation after specification. In addition, MAC1 is implicated in promoting periclinal cell division in the adjacent somatic cells, and may also act later, in the specification of tapetal cell fate.

The mac1-1 mutant was provided by Bill Sheridan (University of North Dakota). The mac1, afd1, am1 and ocl4-1 mutants and inbred lines W23 and B73 were grown in Berkeley and Stanford, CA under field irrigation and fertilization or grown under greenhouse conditions. The mac1-Y211 allele was recovered from a directed Mutator (Mu)-tagging experiment (supplementary material Fig. S1).

Genomic DNA was isolated from leaves (Dellaporta, 1994). Ten pairs of primers spanning the putative mac1 gene were used to check the locus in the mac1-1 mutant (supplementary material Table S1). The Southern blot experiment for characterization of mac1-Y211 was performed as described previously (Nan and Walbot, 2009). The full length mac1 cDNA was amplified using GeneRacer (Invitrogen, Carlsbad, CA, USA).

mac1-related gene sequences were obtained from NCBI, and the virtual translation, full-protein sequences were aligned using ClustalW2 (http://www.ebi.ac.uk/Tools/msa/clustalw2/) with default options. The phylogenic tree was built using the maximum parsimony method with bootstrap values using PHYLIP (http://evolution.genetics.washington.edu/phylip.html).

Total maize RNA was isolated with TRIzol Reagent (Invitrogen), treated with DNase and then purified using an RNeasy kit (Qiagen, Valencia, CA, USA). For quantitative PCR, three biological samples for each tissue were assayed with three technical replicates, with the exception of pollen, silks and a 16-day embryo in which only a single biological sample was available. cDNA was synthesized using the Superscript III First Strand Synthesis System (Invitrogen) and quantitative PCR was conducted using the 2× Supermix qRT kit (Ambion, Austin, TX, USA) with primers F14 and R15 (supplementary material Table S1). Amplification of cyanase transcript was used as an internal control. PCR Miner (Zhao and Fernald, 2005) was used to calculate Ct values.

Total proteins were extracted using the TCA/acetone precipitation method and dissolved in Protein Extraction Reagent (Sigma, St Louis, MO, USA). A partial mac1 cDNA, corresponding to amino acids 91-190 of the virtual translation, was used as an antigen to produce a rabbit polyclonal antibody using Genomic Antibody Technology (SDIX, Newark, DE, USA). The antibody was further affinity purified using a peptide column with amino acids 92-108 and used (1:100) in the western blot analysis.

Spikelets were fixed in 3:1 (ethanol:acetic acid) and embedded with LR White resin (London Resin Company, UK). Semi-thin sections (1 μm) were stained with 1% Toludine Blue O in 1% Borax. For the ocl4; mac1 double mutant, anthers were fixed in Bouin fixative, embedded and sectioned as described previously (Vernoud et al., 2009). The cytology of ovules was performed as described previously (Sheridan et al., 1996). Confocal microscopy and EdU incorporation were carried out as described previously (Kelliher and Walbot, 2011).

Spikelets were fixed in 4% formamide in 1× PBS and embedded in paraffin. Serial sections (10 μm) were placed on ProbeOn Plus slides (Fisher, Fairlawn, NJ, USA). The staining procedure was performed using affinity-purified MAC1 antibody (1:50) and anti-rabbit antibody (1:100) conjugated with alkaline phosphatase (Sigma).

The full-length coding region and the predicted N-terminal signal peptide region were cloned into the vector pAVA120 (von Arnim et al., 1998). Plasmid DNA (5 μg) was bombarded into onion epidermal cells using a Biolistic PDS-1000/He device (Bio-Rad). To visualize GFP signal in the cell wall, samples were incubated with 20 mM PIPES-KOH (pH 7.0) for 8 hours before observation (Scott et al., 1999). Plasmolysis was induced by 0.8 M mannitol for 30 minutes.

The original mac1 mutant was isolated from an active Mutator stock and shown to be a monogenic recessive mutation on chromosome arm 10S. It was mapped between the breakpoint of the T9-10b translocation line and oy1 (Sheridan et al., 1996). This interval is predicted between markers TIDP8865 and IDP8390, and comprises 900 kb (Fig. 2A). Among 14 predicted genes within this interval, the GRMZM2G027522 gene model shares similarity to the rice TDL1A and Arabidopsis TPD1 genes with loss-of-function alleles causing phenotypes similar to the mac1 mutant.

DNA sequence analysis uncovered a 1263 bp deletion in the mac1 allele, starting within intron 2 and extending into exon 4 of GRMZM2G027522 (Fig. 2B). Attempts to sequence the transcript in the mac1 mutant showed that intron 2 is not spliced, probably reflecting absence of the splice acceptor site. The resulting transcript has an in frame stop codon in the unspliced intron 2. The predicted encoded protein contains only the first 32 amino acids of the full-length functional protein, suggesting that mac1 is a null mutation.

The mac1-Y211 allele was isolated from a directed Mutator tagging screen (supplementary material Fig. S1). By DNA gel blot hybridization this allele contained a 4.9 kb autonomous MuDR insertion in intron 2 (Fig. 2B; supplementary material Fig. S1). Allelism analysis confirmed that mac1-Y211 is allelic to the original allele, verifying that GRMZM2G027522 is the mac1 gene (supplementary material Fig. S1). We now designate the original mutant allele as mac1-1.

We cloned a 1405 bp full-length cDNA from the W23 inbred line (GenBank Accession Number JN247438). The mac1 gene has four exons and the encoded protein is predicted to be 214 amino acids (22 kDa). MAC1 shares 58.7% and 39% similarity with rice TDL1A and Arabidopsis TPD1, respectively. Using SignalIP 3.0 software, the first 30 amino acids are a predicted cleavable signal peptide (Fig. 2B), which may target the protein into the endoplasmic reticulum for cellular export.

The MAC1 protein sequence was used in TBLASTN queries, and several related genes were found in maize, rice and Arabidopsis (cutoff E-value: e-10). Phylogenic analysis indicates that maize MAC1 and rice TDL1A form a branch in a clade that also contains Arabidopsis TPD1, whereas rice TDL1B occupies a more distant branch (supplementary material Fig. S2). Additionally, the region of the TDL1A gene on rice chromosome 12 is syntenic to the maize chromosome 10 region where mac1 resides. Considered together with the phylogeny, we conclude that mac1 is the maize ortholog of the rice TDL1A gene.

Maize anther development is highly regular relative to anther length (Kelliher and Walbot, 2011), and this metric was used to compare fertile and sterile anthers. In transverse sections examined by light microscopy, anther lobes of very young anthers (~100 μm in length) contain a group of L2-derived cells (L2-d). At this stage, all L2-d cells appear similar in morphology, size and shape; subsequently, AR cells are found centrally as enlarged germinal cells. No significant difference was observed between fertile and mac1 sterile anthers at this stage (Fig. 3A,F). At a slightly later stage (200-300 μm), fertile anthers contain one or two AR cells in each transverse section, surrounded by numerous L2-d somatic cells (Fig. 3B); subsequently, these L2-d cells divide periclinally (arrowhead in Fig. 3B), yielding the SPC and EN cells. Concurrently, the AR cells proliferate into several columns within the locule (Fig. 3C). Later in anther development (~700 μm in length), the secondary parietal cells divide periclinally (arrowhead in Fig. 3D) to form the TA that encircles the AR cells and the ML beneath the EN. During development, all cell layers divide anticlinally, increasing cell numbers in both the length and girth of each locule; for example, the number of EN cells increases from 8-10 (Fig. 3C) to 15-16 cells (Fig. 3E), as viewed in transverse sections. A few days prior to meiosis, each locule cell type has a distinct morphology (Fig. 3E): EP cells are elongated parallel to the long axis of the anther, EN cells are elongated 90^o relative to the epidermal cells, the ML is flattened and the TA cells are cuboidal with a densely stained cytoplasm.

At the 300 μm stage the first clear mac1 defect was observed in transverse section as an excess of AR cells enclosed by a single layer of subepidermal L2-d cells (Fig. 3G,K). Periclinal division of L2-d subepidermal cells in mac1 is delayed (Fig. 3H-L). The mutant L2-d cells do divide anticlinally; in transverse section, the subepidermal cell number is about 15-16 in mac1 anthers (Fig. 3I,L), so only the periclinal division is suppressed. Later, just prior to meiosis, some mac1 L2-d sub-epidermal cells divide irregularly (supplementary material Fig. S4), to form several cell layers with unclear cell identity. At later stages in mac1, the subepidermal layer shares some characteristics with EN; however, cells with characteristics of the ML or TA layers are never observed (Fig. 3J,M; supplementary material Fig. S5). Because the two alleles showed similar phenotypes, the following analyses were carried out using the mac1-1 allele, unless otherwise specified.

Confocal microscopy was used to compare reconstructions of sterile and fertile sibling anthers. A longitudinal optical section of fertile anthers (425 μm stage) showed clearly that AR cells are enclosed by three somatic cell layers: EP, EN and SPC (Fig. 4A). By contrast, a 425 μm mac1-1 anther had an excess number of AR cells enclosed by only the EP and L2-d cells (Fig. 4B). To examine the cell proliferation rate during anther development, EdU was incorporated into replicating DNA during S phase to label actively dividing cells (Fig. 4C,D). In mac1, the AR cells proliferate at least twofold more frequently (Fig. 4E). Consequently, mac1 anthers have accumulated three times more AR cells in a locule compared with fertile anthers at the 600 μm stage (supplementary material Fig. S3).

It has been inferred that meiosis arrests in prophase I in mac1-1 (Golubovskaya et al., 1992a; Sheridan et al., 1999). Here, we examined RAD51 loading onto chromosomes. RAD51 is a RecA-like protein involved in homologous recombination, and it exists in foci on chromosomes during prophase I (Franklin et al., 1999). RAD51 loading is normal in the mac1-1 mutant (supplementary material Fig. S6). Furthermore, meiotic progression in mac1-Y211 is typically arrested before metaphase I; however, a few mutant meiocytes reach metaphase II (supplementary material Fig. S7). All anthers from both mutant alleles are completely sterile. We conclude that mac1 mutant meiocytes have the correct cell identity in both mutants but meiosis fails; because many mutants with tapetal defects exhibit ‘failure to thrive’ syndromes, we speculate that the absence of a functional tapetal layer results indirectly in meiotic arrest.

We further constructed double mutants with two meiotic genes, am1 (ameiotic1) and afd1 (absence of first division1). In am1-1 anthers and ovules, meiocytes conduct mitosis instead of meiosis (Golubovskaya et al., 1992b; Golubovskaya et al., 2006). In the afd1 mutant, the first division is absent and sister chromatids separate at metaphase I. The offspring of self-pollinated mac1/+; am1/+ and mac1/+; afd1/+ double heterozygous stocks were examined (supplementary material Tables S2, S3). In each family, the segregation of fertile and sterile plants was 9:7, as expected, and megaspore mother cells (MMC) isolated from single ovules on sterile plants were examined by light microscopy. In both mac1/mac1; am1/am1 and mac1/mac1; afd1/afd1 double mutant plants, multiple MMCs were observed, which is characteristic of mac1. For the mac1/mac1; afd1/afd1 double mutant, 1001 MMCs at various stages were observed in 183 ovules (supplementary material Table S4), showing an average of 5.5 MMCs in double mutants compared with one such cell per ovule in wild-type plants and nine MMCs in mac1 plants (Sheridan et al., 1996). By comparison, in anthers there was approximately a two- to threefold increase of PMCs in mac1 compared with fertile plants. In addition, as meiosis started, am1 or afd1 meiotic phenotypes appeared in both double mutants (supplementary material Figs S8, S9), suggesting that mac1 acts independently of both meiosis genes. As shown in supplementary material Tables S2 and S3, the segregation ratios were 9:3:3:1. The additive nature of phenotypes in both double mutants confirms that PMCs in mac1 mutants can enter the meiotic cell cycle appropriately.

mac1 mRNA levels in the anther, ear and other organs were examined by quantitative RT-PCR (Fig. 5A). The mac1 transcript level is low in anthers smaller than 500 μm and peaks during meiosis, then drops off sharply after meiosis is completed. mac1 transcripts are low in the ear but exhibit high expression in the root; no expression was detected in pollen, silk, embryos (16 days post fertilization), juvenile leaf or stem (10 days post germination). This pattern is similar to rice TDL1A expression, except that the rice transcript is also present in embryos (Zhao et al., 2008). By contrast, the RNA expression of the rice TDL1B gene is not consistent with its putative homolog in maize; there is no RNA expression in maize root, anther and ear tissues when assessed using quantitative RT-PCR.

To check whether mac1 roots have a scorable defect, we examined primary root growth and performed confocal imaging of maize roots; we found no obvious difference between wild-type sib and mac1 mutant plants (data not shown).

MAC1 contains a predicted cleavable signal peptide; the processed protein has a predicted mass of 18.6 kDa. We raised a polyclonal antibody against a partial sequence of MAC1 (amino acids 91-190). The 91-190 amino acid antigen affinity-purified antibody showed poor specificity (supplementary material Fig. S10). To improve it, we purified the antiserum with an affinity column coupled with a MAC1 peptide (amino acids 92-108). This affinity-purified antibody specifically recognizes an anther protein of ~18 kDa; this band is absent in anther protein extracts from both mutants (Fig. 5B). We conclude that the mature MAC1 protein is about 18 kDa and that both mutant alleles are null mutations. Although mac1 is required for normal embryo sac development in ears, the 18 kDa protein was not detected in the ear protein sample. Unlike the anther, which is rich in sporogenous tissue, immature ears are nearly all somatic cells; we propose that the level of protein in ears is too diluted to be detected.

Immunolocalization assays using the peptide-affinity purified antibody were performed to elucidate MAC1 protein distribution. In 150 μm anthers, MAC1 protein was first visible in a few L2-d cells with weaker signals diffusely present in the four anther lobes, surprisingly including the EP and the connective tissue between the abaxial- and adaxial locules (Fig. 6A). Slightly later in development (300-400 μm), the MAC1 signal becomes more concentrated in the central cells of each locule, corresponding to the location of AR cells (Fig. 6B). In some sections, a weaker MAC1 signal was also observed in some epidermal cells at the outer ridge of anther locules (Fig. 6C). After all cell types have been established by 750 μm, the MAC1 signal was even stronger in the AR cells with a much weaker signal in the adjacent cells; it is possible that the low signals outside the AR cells here and at earlier stages represent diffusion of MAC1 protein after secretion by AR cells (Fig. 6D,E). No MAC1 signal was detected in the mac1-1 mutant (Fig. 6F), validating that the signals in fertile anthers represent detection of MAC1 protein.

Although MAC1 and its homologs are predicted to be secreted proteins, this feature has not been established experimentally. To determine the subcellular localization of MAC1, GFP was fused either to the C terminus of MAC1 or to the putative signal peptide (SP) (Fig. 7A) and expressed in onion epidermal cells. The control GFP vector exhibited both cytoplasmic and nuclear fluorescence (Fig. 7B), consistent with a previous report (Chiu et al., 1996). The SP-GFP fusion protein was secreted into the extracellular space between cells (Fig. 7C,D). We propose that the fusion protein is processed and secreted via the vesicular secretory pathway as SP-GFP signals were also observed in the endoplasmic reticulum around the nucleus (Fig. 7C). When the full-length MAC1 was fused to the GFP, the signal was observed only in the endoplasmic reticulum (Fig. 7E); no extracellular signal was apparent. It is likely that the large fusion protein (45 kDa) is unable to move through the cell wall (Baron-Epel et al., 1988). To discern whether the MAC1-GFP signal is secreted but hampered in movement by exclusion from the cell wall, onion epidermis was treated with hypertonic solution. After plasmolysis, it is evident that some MAC1-GFP fusion protein is associated with the cell wall (arrowhead in Fig. 7I), as well as the Hechtian strands (arrow in Fig. 7I). This was not observed with the GFP control (Fig. 7G,H). These observations indicate that MAC1 protein can be secreted via the endoplasmic reticulum-mediated secretory system.

Maize ocl4 (outer cell layer 4) encodes a transcription factor of the HD-ZIP class IV family, which is specifically expressed in the EP of numerous organs. In the ocl4 mutant, anthers begin their development normally but an additional periclinal cell division occurs after the L2-d cells divide periclinally to form the EN and SPC. The presumptive EN divides periclinally again in the locular bulge, leading to the formation of two cell layers with EN characteristics such as accumulation of starch granules (Vernoud et al., 2009) (Fig. 8B). At the same stage, in the mac1 mutant, one L2 layer is present subepidermally, but only late in premeiotic anther development does it exhibit any EN-like characteristics (Fig. 8C). In the mac1-1; ocl4 double mutant, excess AR cells are observed with disordered somatic cells around them as observed in mac1. Two EN-like layers with starch granules also develop (Fig. 8D). Taken together, these result indicate that the two mutant phenotypes are additive and that extra subepidermal cells generated by the ocl4 defect do not lead to differentiation of ML or TA types in the absence of MAC1.

To ascertain the timing of phenotypic deviations more precisely, mac1; ocl4 double mutant anthers were observed by confocal microscopy. In the ocl4, the extra periclinal division occurs after the initial L2-d periclinal division (Vernoud et al., 2009), a stage corresponding to ~400-500 μm anthers (Kelliher and Walbot, 2011). In fertile anthers at this stage, the EN and SPC have already developed (Fig. 8E), whereas, in mac1, only a single layer of L2-d subepidermal cells exists (Fig. 8F). In the double mutant, however, some cells in the L2-d subepidermal layer divide in 500 μm anthers (Fig. 8G, arrowhead). These divisions in the double mutant take place at about the same time as formation of the extra EN layer in the single ocl4 mutant; the number of AR cells resembles the mac1 mutant (Fig. 8G). Later, a burst of irregular cell division occurs in the double mutant to form unorganized cell layers with unclear cell identity, similar to mac1 (Fig. 8H). In summary, the double mutant shows an additive phenotype, indicating these two genes are probably acting independently to influence somatic cell division patterns in anther locules.

Plant organ development requires communication between cell layers to coordinate tissue differentiation (Fletcher, 2002; Feng and Dickinson, 2010b; Ma and Sundaresan, 2010). We demonstrate that early in anther ontogeny, MAC1 suppresses AR cell proliferation and that it is also required to promote periclinal division in the adjacent L2-d cells. The early roles of MAC1 suggest a function in regulating cell division rather than in guiding the SPC and TA differentiation to suppress trans-differentiation into PMC. In support of our conclusion, by immunolocalization, MAC1 is located primarily in the initial AR cells, with a diffuse gradient across the microsporangium. We demonstrate that MAC1 is a secreted protein and hypothesize that AR cells are a ‘signaling center’ from their inception. We further hypothesize that the dual early roles result from the secreted MAC1 protein via binding to a receptor kinase(s) in the AR cells to regulate their proliferation frequency and binding to a different receptor kinase(s) in the adjacent L2-d cells to regulate their periclinal division. These L2-d cells continue anticlinal cell divisions in mac1 mutants; therefore, the role of MAC1 is specific to promoting their periclinal division rather than being a general regulator of L2-d cell proliferation. Early steps in anther ontogeny have not yet been investigated in tpd1 Arabidopsis or tdl1a rice, whereas roles of MAC1 at later stages cannot be assessed in these mac1 mutants because cell differentiation is disrupted so early. Therefore, we cannot exclude the possibility that mac1 may induce the division of secondary parietal cells and then determine the fate of tapetal cells.

The ocl4 gene plays a role in suppressing additional periclinal division in EN precursors. In the absence of OCL4, these cells divide periclinally once to form an extra EN-like cell layer. These divisions occur in the outer locule arc distant from the central vascular column in the center of each anther (Vernoud et al., 2009). OCL4 is a transcription factor expressed exclusively in epidermal cells, suggesting that the signal(s) inhibiting periclinal division in EN cells is generated in the overlying layer. Double mutant analyses of mac1 and ocl4 demonstrated additive phenotypes, suggesting that these genes act independently in different pathways. The L2-d layer encircling excess AR cells in the mac1 mutant does not divide periclinally, a feature that we reasoned reflects the absence of MAC1. The absence of periclinal divisions may be reinforced by the normal OCL4-dependent signaling pathway generated by the overlying EP. In mac1; ocl4 double mutants, the L2-d cells do perform periclinal divisions starting at about the same stage as the extra periclinal divisions characteristic of ocl4. These observations suggest that the subepidermal L2-d cells in mac1 can respond to the signal from EP that inhibits additional division. Collectively, our analyses suggest that both EP and AR cells regulate the timing and location of periclinal divisions in L2-d cells (Fig. 9). The AR secretes a promoting factor required for the single periclinal divisions of L2-d subepidermal cells, and then the EP makes a suppressor acting on the outer layer after this division. Our observations provide strong evidence that cell-cell signaling is important in anther ontogeny and that signaling processes based on cell position rather than lineage are the major factor regulating differentiation of the L2-d cells into EN and SPC. Similar processes, perhaps involving some of the same proteins, may regulate the later step of secondary parietal periclinal division to generate the ML and TA.

We report here that mac1 encodes a small protein with an N-terminal signal peptide and that it shares similarity with rice TDL1A and Arabidopsis TPD1 genes. Both mac1 and rice tdla-RNAi mutants share the ovule development defect of excess MMCs (Sheridan et al., 1996; Zhao et al., 2008). tdla-RNAi rice exhibits normal anther development, a result interpreted to be consequence of failure to express sufficient RNAi in anthers (Zhao et al., 2008). Based on phenotypic similarity, syntenic chromosomal locations and phylogenic analysis, we conclude that maize mac1 is the ortholog of the rice TDL1A gene.

Both TDL1A and TPD1 were predicted to be secreted ligands, and they interact with the LRR-RLKs MSP1 and EMS1/EXS in rice and Arabidopsis, respectively (Canales et al., 2002; Zhao et al., 2002; Yang et al., 2003; Jia et al., 2008; Zhao et al., 2008). The rice msp1 mutant displayed an excessive number of both male and female sporogenous cells; anther wall layers were disordered and the TA layer was absent completely (Nonomura et al., 2003). These phenotypes are similar to maize mac1 in both anthers and ovules. By contrast, the Arabidopsis tpd1 or ems1/exs mutants show defects only in anthers (Zhao et al., 2002; Yang et al., 2003; Feng and Dickinson, 2010a). Therefore, there are differences in the signaling pathways for male and female reproductive organ development in grasses and dicots.

In ems1 and tpd1 Arabidopsis mutants, excess sporogenous cells appeared at stage 5 anthers and the TA layer was missing (Zhao et al., 2002; Yang et al., 2003). The number of AR in these mutants is similar to the sum of the TA and the AR in fertile anthers (Zhao et al., 2002). These data were invoked to propose that the TPD1-EMS1 signaling pathway specifies TA cell fate after the periclinal division of the SPC: in effect, suppressing the germinal fate in the tapetal cells (Feng and Dickinson, 2010a; Ma and Sundaresan, 2010). Exploiting the large size of maize anthers during the initial steps of cell fate setting, we find that mac1 anthers contain excess AR from the earliest stages examined (<300 μm) and that these AR cells subsequently proliferate at more than twice the rate observed in fertile anthers throughout the period before SPCs in fertile anthers would divide to generate the tapetal layer. Therefore, at least in maize, MAC1 functions as a suppressor of AR proliferation in an autoregulatory circuit. Absence of this suppressive function is sufficient to explain the excess PMC phenotype without invoking trans-differentiation of a newly formed tapetal layer.

By confocal analysis, maize ARs are specified before 300 μm within dozens of L2-derived somatic cells in each lobe. There was no evidence for a single hypodermal cell dividing asymmetrically to form a founding primary parietal cell and a single primary sporogenous cell (Kelliher and Walbot, 2011). How AR cell fate is set in maize remains largely unknown. In Arabidopsis the transcription factor SPOROCYTELESS/NOZZLE (SPL/NZZ) promotes AR specification (Yang et al., 1999; Ito et al., 2004). SPL/NZZ is thought to interact with two LRR-RLKs, BARELY ANY MERISTEM 1 and 2 (BAM1 and BAM2) (Hord et al., 2006). Interestingly, both maize and rice lack a homolog of SPL/NZZ, which provides a second example in which grasses, and perhaps monocots in general, regulate the male component of reproduction differently from Arabidopsis.

There are some parallels between the role of MAC1 in suppressing AR proliferation and the role of CLAVATA3 (CLV3) in shoot apical and floral meristems (Willmann, 2000; Rojo et al., 2002). CLV3 acts as a negative regulator to control the number of stem cells (Brand et al., 2000). CLV3 is a secreted protein that functions as a ligand for the CLV1-CLV2 receptor kinase complex (Willmann, 2000; Sharma et al., 2003). CLV3 activates a signal transduction cascade to restrict expression of the WUSCHEL (WUS) transcription factor, a protein that promotes stem cell formation and also provides a negative-feedback regulation on CLV3 expression (Muller et al., 2006). The size of the apical meristem cell population is controlled by the balance of CLV3 and WUS expression. If this analogy holds in anther locules, then a second factor opposing MAC1 action is predicted to promote AR proliferation.

When excess AR cells are first observed in mac1, there is only one layer of encircling L2-d cells, a state that persists until close to the start of meiosis (Sheridan et al., 1999; Abramova et al., 2002). MAC1 suppresses AR proliferation, but it is also required to promote periclinal division in the L2-d cells. In mac1 mutant, L2-d cells divide regularly anticlinally to sustain anther growth. At the stage when all five cell types are present in normal locules, the mac1 L2-d cells begin to divide irregularly (not strictly in the periclinal plane) to form patches of multilayered somatic wall tissue. We attribute the absence of early periclinal L2-d cell divisions in mac1 anthers to the absence of MAC1 protein secretion by the central AR cells. The divisions near the onset of meiosis could arise from a different signal or mechanism.

When TPD1 is expressed ectopically in wild-type Arabidopsis, the number of carpel cells increased significantly, which supports the positive role that this type of protein can have on somatic cell proliferation (Yang et al., 2005). TPD1 activation of cell division in the transgenic carpel is dependent on its receptor, EMS1/EXS (Yang et al., 2005). Feng and Dickinson (Feng and Dickinson, 2010a) expressed EMS1 under the control of the A9 tapetal promoter in ems1 plants and found that locular peripheral cells (similar to our designation of L2-d cells) could respond late in anther ontogeny to differentiate as functional tapetal cells. If TPD1 is the ligand late in development, this supports a third role for the triad of MAC1/TDL1A/TPD1 proteins in regulating tapetal differentiation. We hypothesize one or more kinases may interact with MAC1 in maize to support its multiple roles in anther and ovule.

We thank Bill Sheridan (University of North Dakota) and David Duncan for mac1 mutants; David Duncan was supported by a VPUE summer research award from Stanford University to conduct the directed tagging and initial analysis of new mutants of mac1. We thank Ling Meng and Peggy G. Lemaux (UC Berkeley) for help with onion transient expression assay, and Jennifer Fletcher (Plant Gene Expression Center, USDA) for her gift of pAVA120. We thank Wann-Neng Jane and Plant Cell Biology Core Lab (IPMB, Academia Sinica) for help in anther sectioning. We thank Jihyun Moon (UC Berkeley) for her comments on the manuscript.