Regulation of the Arabidopsis anther transcriptome by DYT1 for pollen development

Authors

  • Baomin Feng,

    1. Department of Biology, Huck Institutes of the Life Sciences, Pennsylvania State University, University Park, PA 16802, USA
    Search for more papers by this author
    • These authors contributed equally to this work.

    • Present address: Department of Plant and Microbial Biology, University of California, Berkeley, CA 94720, USA.

  • Dihong Lu,

    1. Intercollege Graduate Degree Program in Plant Biology, Huck Institutes of the Life Sciences, Pennsylvania State University, University Park, PA 16802, USA
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Xuan Ma,

    1. Intercollege Graduate Program in Cell and Developmental Biology, Huck Institutes of the Life Sciences, Pennsylvania State University, University Park, PA 16802, USA
    Search for more papers by this author
  • Yiben Peng,

    1. Peking–Yale Joint Center for Plant Molecular Genetics and Agrobiotechnology, National Laboratory for Protein Engineering and Plant Genetic Engineering, College of Life Sciences, Peking University, Beijing 100871, China
    Search for more papers by this author
  • Yujin Sun,

    1. Department of Biology, Huck Institutes of the Life Sciences, Pennsylvania State University, University Park, PA 16802, USA
    Search for more papers by this author
    • Present address: Department of Biology, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599, USA.

  • Gang Ning,

    1. Department of Biology, Huck Institutes of the Life Sciences, Pennsylvania State University, University Park, PA 16802, USA
    2. Microscopy and Cytometry Facility, Huck Institutes of the Life Sciences, Pennsylvania State University, University Park, PA 16802, USA
    Search for more papers by this author
  • Hong Ma

    Corresponding author
    1. State Key Laboratory of Genetic Engineering and Institute of Plant Biology, Center for Evolutionary Biology, School of Life Sciences, Fudan University, Shanghai 200433, China
    2. Institutes of Biomedical Sciences, Fudan University, Shanghai 200032, China
    Search for more papers by this author

(e-mail hongma@fudan.edu.cn).

Summary

Several genes encoding transcription factors have been shown to be essential for male fertility in plants, suggesting that transcriptional regulation is a major mechanism controlling anther development in Arabidopsis. DYSFUNCTIONAL TAPETUM 1 (DYT1), a putative bHLH transcription factor, plays a critical role in regulating tapetum function and pollen development. Here, we compare the transcriptomes of young anthers of wild-type and the dyt1 mutant, demonstrating that DYT1 is upstream of at least 22 genes encoding transcription factors and regulates the expression of a large number of genes, including genes involved in specific metabolic pathways. We also show that DYT1 can bind to DNA in a sequence-specific manner in vitro, and induction of DYT1 activity in vivo activated the expression of the downstream transcription factor genes MYB35 and MS1. We generated DYT1–SRDX transgenic plants whose fertility was dramatically reduced, implying that DYT1 probably acts as a transcriptional activator. Furthermore, we used yeast two-hybrid assays to show that DYT1 forms homodimers and heterodimers with other bHLH transcription factors. Our results demonstrate the important role of DYT1 in regulating anther transcriptome and function, and supporting normal pollen development.

Introduction

Somatic tissues in reproductive organs provide germ (reproductive) cells with nutrients and signal molecules. In typical flowering plants, the male reproductive organ (the anther) contains four similarly structured lobes, each with a few somatic cell layers surrounding the germ cells, which undergo meiosis to produce microspores. The tapetum layer immediately adjacent to the germ cells is highly active metabolically and nurtures pollen development from microspores. In Arabidopsis thaliana, the tapetal cells are formed at anther stage 5, function at anther stages 5–9, and degenerate via programmed cell death starting from anther stage 10 to release their contents for pollen wall formation (Sanders et al., 1999). Defects in the tapetum usually cause abnormal pollen development, resulting in reduced fertility.

Previous results support the idea that tapetum development is controlled by a complex transcriptional regulatory network including several key genes (Yang et al., 1999, 2007; Ito et al., 2007; Zhang et al., 2007; Thorstensen et al., 2008; Zhu et al., 2008). In particular, SPOROCYTELESS/NOZZLE (SPL/NZZ) encodes a putative transcription factor and is required for specification of both germ cells and somatic cell layers (including the tapetum) (Schiefthaler et al., 1999; Yang et al., 1999). ABORTED MICROSPORES (AMS) plays an important role in the tapetum for biosynthesis and secretion of pollen wall components (Sorensen et al., 2003; Thorstensen et al., 2008; Ma et al., 2012). Other transcription factor genes important for the tapetum and pollen development include MALE STERILITY 1 (MS1) (Wilson et al., 2001; Ito and Shinozaki, 2002; Ito et al., 2007; Yang et al., 2007) and MYB family members (Preston et al., 2004; Millar and Gubler, 2005; Zhang et al., 2007; Zhu et al., 2008). Analysis of genes for these transcription factors and other genes suggests that tapetum development is controlled by a sophisticated transcriptional regulatory network (Zhang et al., 2006; Wijeratne et al., 2007; Zhu et al., 2008).

Cell–cell signaling is also required to establish the tapetum identity. Previous studies revealed important roles for the mitogen-activated protein kinases MPK3 and MPK6, and the leucine-rich repeat receptor-like kinases ERECTA and ERECTA-LIKE1 and 2, in early anther development, including tapetal cell differentiation (Hord et al., 2008; Ge et al., 2010). A putative ligand–receptor pair encoded by TPD1 and EMS1 is essential for specification of tapetal cell fate (Zhao et al., 2002; Yang et al., 2003, 2005; Jia et al., 2008). Mutations in either gene cause formation of extra microsporocytes and failure to form tapetal cells. Changes in the ems1 transcriptome suggest that cell–cell signaling dramatically affects components of a transcriptional regulatory network in the tapetum (Wijeratne et al., 2007). Another study showed that brassinosteroids regulate key genes for anther development via the transcription factor BRI1–EMS SUPPRESSOR 1 (BES1), providing a link between phytohormone signaling and transcriptional regulation (Ye et al., 2010).

Recently, a bHLH transcription factor, DYSFUNCTIONAL TAPETUM 1 (DYT1), was discovered to be important for tapetum function (Zhang et al., 2006). In the dyt1 mutant, tapetal cells are formed but undergo premature vacuolation, with a reduction in the dense cytoplasm as observed in wild-type tapetal cells. Although dyt1 meiocytes undergo meiotic nuclear division to produce four haploid nuclei, the malfunction of the tapetum is linked to failure to produce microspores, resulting in the absence of any developing pollen grains. Previous studies suggested that DYT1 functions genetically downstream of SPL and TPD1/EMS1, but upstream of AMS and MS1 (Zhang et al., 2006; Sun et al., 2007). Therefore, DYT1 integrates early signaling and transcriptional pathways and controls the expression of late genes in a transcriptional regulatory network for tapetum development. However, the extent of the transcriptome regulated by DYT1 and the mechanisms of DYT1 function remain to be investigated.

In this study, a comparative transcriptomic analysis of dyt1 and wild-type anthers was performed to identify candidate genes regulated by DYT1. Several genes were shown to be activated after induction of DYT1 activity using a transgenic fusion. Because AMS is a paralog of DYT1 from a gene duplication shared by Arabidopsis and rice, and has important functions in anthers at post-meiotic stages, comparative transcriptomic analyses of ams, dyt1 and wild-type anthers were performed to identify genes possibly regulated by DYT1 and/or AMS. Several bHLH proteins are known to bind to a DNA sequence motif called an E-box (Martinez-Garcia et al., 2000; Toledo-Ortiz et al., 2003; Ito et al., 2012). Additional experiments described here indicated that DYT1 bound to a subset of E-boxes in vitro and interacted with other transcription factors in yeast two-hybrid assays, providing clues regarding the mechanisms of DYT1 action.

Results

Identification of differentially expressed genes in dyt1 anthers by microarray analysis

To identify genes regulated by DYT1 during anther development, microarray experiments were performed using Affymetrix ATH1 chips and cDNAs synthesized from total RNA of wild-type and dyt1 anthers at stages 4–7. A total of 3118 genes were found to be differentially expressed between wild-type and dyt1 anthers with a false-discovery rate (FDR) <0.05, including 435 that were down-regulated (Table S1) and 220 that were up-regulated (Table S2) by at least twofold in dyt1. However, the twofold difference is an arbitrary threshold, and some local or short-living interactions may have been retrieved by using a lower threshold. The microarray results were highly reproducible (Pearson’s coefficients 0.95–0.99), and were further verified by testing the expression of selected genes using RT-PCR and real-time PCR. The expression changes of nine genes encoding putative transcription factors and seven encoding enzymes (Table S3) as determined by real-time PCR were consistent with the microarray data (Figure S1).

Using agriGO (Du et al., 2010), over-represented GO categories in the down-regulated group according to molecular functions included lipid binding, transporter activity and hydrolase activity (hydrolyzing O-glycosyl compounds), and those according to biological processes included peptide transport, lipid transport, pollen exine formation and phenylpropanoid biosynthesis (hypergeometric P value <0.05) (Figure 1a,b). These results suggest that DYT1 probably controls tapetum function by positively regulating genes responsible for the biosynthesis, transport, assembly and modification of cell walls and pollen coat. In the up-regulated group, over-represented GO categories in biological processes included responses to heat, high light intensity or reactive oxygen species (hypergeometric P value <0.05) (Figure 1c). However, no categories on the basis of molecular function were over-represented in the up-regulated genes. These results suggest that DYT1 normally represses these stimulus-responsive genes; alternatively, the dyt1 mutation may cause cellular stresses that resulted in a general increase in expression of these genes.

Figure 1.

 Categories of genes with greater than twofold expression changes in dyt1 anthers.
(a) Molecular function categories enriched in genes down-regulated in dyt1 (< 0.05). Each bar represents the percentage of total genes with annotations (435 down-regulated genes in dyt1; 22 479 genes on the ATH1 chip).
(b) Biological process categories enriched in genes down-regulated in dyt1 (< 0.05).
(c) Biological process categories enriched in 204 up-regulated genes with annotations in dyt1 (< 0.05).

DYT1 positively regulates at least 22 genes encoding transcription factors

Among the 435 genes down-regulated by at least twofold in dyt1 anthers, 22 encode transcription factors (Table S4). MS1 was not among the 22 genes because there is no probe for it on the Affymetrix ATH1 chip. Among the 22 genes, AMS, MYB103/MYB80 and MYB35 are important for tapetum function and male fertility, indicating that DYT1 probably regulates at least some downstream genes indirectly through other transcription factors (Wilson et al., 2001; Sorensen et al., 2003; Zhu et al., 2008), in addition to possible direct regulation by DYT1. The remaining transcription factor genes in this group may be as yet undefined regulators downstream of DYT1, highlighting candidates for further functional studies. For example, FLOWERING LOCUS C (FLC) has been shown to be a repressor of flowering (Michaels and Amasino, 1999). However, its role in anther development remains to be investigated.

DYT1 positively regulates genes for lipid metabolism, cell-wall modification and secondary metabolism that are important for pollen development

The GO annotations and MapMan analyses suggested that DYT1 positively regulates genes involved in lipid metabolism, cell-wall modification, and secondary metabolism (Thimm et al., 2004). Twenty-eight genes that are down-regulated in the dyt1 mutant are involved in lipid metabolism (Table S5), including lipid transfer proteins, acyl transferases, lipases, etc. Among them, MALE STERILE 2 (MS2) encodes a fatty acid reductase that is required for pollen wall formation (Aarts et al., 1997). Its expression level was reduced by 16-fold in dyt1 compared with the wild-type. Expression of nine pollen coat-related genes was also affected in dyt1, including genes encoding glycine-rich proteins/oleosins, arabinogalactan proteins and extracellular lipases (Table S6).

The expression levels of 33 genes involved in cell-wall modification were decreased in dyt1 anthers (Table S7). Most of these genes encode enzymes that catalyze breakdown of polysaccharides, including galactosidases, glucosidases, xylosidases, polygalacturonases, pectatelyases, etc. Four genes encode glycosyltransferases that may be involved in polysaccharide synthesis or glycosylation. Three genes encode invertase/pectin methylesterase inhibitors that may antagonize enzyme activities (Hothorn et al., 2004). These enzymes and inhibitors together may modify tapetum or meiocyte cell walls in a dynamic and balanced manner.

The microarray data also suggested that secondary metabolism was affected in the dyt1 mutant. Ten genes involved in lignin and flavonoid biosynthetic pathways showed reduced expression in dyt1 (Table S8). Among these genes, ACYL COA SYNTHETASE 5 (ACOS5) was reported to be expressed in the tapetum and important for pollen development (de Azevedo Souza et al., 2009). Twenty-seven transporter genes also showed lower expression in dyt1 (Table S9). These transporters may be important for transport of organic molecules (sugars, oligopeptides, amino acids, etc.) and inorganic molecules (Zn, Cu, nitrate, phosphate, etc.) to support active metabolism in the tapetum, or may export materials from the tapetum to anther locules to facilitate meiosis and pollen development.

Some defects in dyt1 anthers may be explained by the reduced expression of genes involved in the afore-mentioned pathways. Ultra-thin sections of dyt1 and wild-type anthers were analyzed by transmission electron microscopy (TEM) (Figure 2). At stage 5, dyt1 (Figure 2b,f) and wild-type (Figure 2a,e) anthers had similar tapetal cell-wall structures, but dyt1 tapetal cells were more vacuolated than wild-type. At early stage 6, dyt1 tapetal cells had much larger vacuoles and meiocytes had much thinner cell walls (callose walls) (Figure 2d,h) than those of wild-type (Figure 2c,g). As DYT1 is expressed in the tapetum and malfunction of the dyt1 tapetal cells precedes the abnormal cell-wall phenotype, it is possible that DYT1 regulates genes that are essential for active metabolism in the tapetal cells and for export of materials to meiocytes, such as those for callose wall formation.

Figure 2.

 TEM images of wild-type and dyt1 meiocytes and tapetal cells.
(a) Wild-type stage 5 anther section.
(b) dyt1 stage 5 anther section.
(c) Wild-type stage 6 anther section.
(d) dyt1 stage 6 anther section.
(e–h) Enlargements of boxed areas in (a–d), respectively.
Arrows indicate cell walls. Ms, meiocytes; T, tapetal cells; V, vacuoles. Scale bars = 10 μm (a–d) and 5 μm (e–h).

DYT1 binds to an E-box variant (TCACGTGA) in vitro

Because DYT1 contains a bHLH domain, we tested whether it bound to E-boxes found in the promoters of several downstream genes. We found that recombinant DYT1 proteins bound to an E-box from the MS1 promoter (Figure 3a–c). The binding specificity was validated by the findings that assays including unlabeled competitor probes with the same sequence or probes with point mutations (CACGTG →AACGTG or CACGTG → CGCGTG) showed no detectable DYT1 binding (Figure 3a,b). In addition, an increase in the DYT1 protein amount enhanced the level of DYT1–probe complexes (Figure 3c). These results strongly support the hypothesis that DYT1 specifically binds to E-boxes.

Figure 3.

 Determination of DYT1 in vitro binding sites.
(a) DNA probes used in gel-shift assays to show that DYT1 specifically binds to an E-box in the MS1 promoter (pMS1). pMS1 contains 20 bp MS1 promoter with an E-box and 20 bp region for oMC2746 annealing; M1 and M2 contain point mutations (red letters) in the E-box. The wild-type and mutated E-boxes are underlined.
(b) Gel-shift assay showing DYT1–E-box interaction. The presence of competitor probes (Comp) with the pMS1 sequence but unlabeled with biotin, and mutations in the M1 and M2 probes abolished DYT1–E-box binding. L, lysate of E. coli without DYT1 proteins.
(c) Gel-shift assay showing that higher amounts of DYT1 enhanced shifted bands.
(d) Consensus binding sequence of DYT1 determined by SELEX.
(e) Representative DNA probes enriched by SELEX (oMC6327–oMC6338). Possible binding sites are underlined. Red nucleotides are different from the consensus sequence.
(f) Gel-shift assay to confirm the interaction between probes listed in (e) with DYT1.
(g, h) Gel-shift assay to determine DYT1 binding specificity. A lysate of E. coli containing only the vector (v) did not bind probes. A 100-fold excess of unlabeled competitor probes and point mutations in half of the palindromic consensus sequence (probes p2–p5) abolished DYT1 binding. Probe (p6) with a mutation in the −1 position interacted with DYT1.

We also used an unbiased approach, systematic evolution of ligands by exponential enrichment (SELEX), to determine DYT1 binding sites in vitro. A consensus sequence TCACGTGA was found to be the strongest binding motif for DYT1 (Figure 3d). The palindromic nature of the consensus sequence suggested that DYT1 binds to DNA as homodimers. Twelve DNA probes enriched in the SELEX experiments were tested to determine whether they can be bound by DYT1 (Figure 3e,f). Directed mutagenesis analysis of half of the palindromic sequence indicated that all eight nucleotides are necessary for DYT1 binding under in vitro conditions. Changing a flanking nucleotide (p6, oMC6291) did not abolish DYT1 binding (Figure 3g,h). Sometimes two or more closely spaced sequences differing from the binding consensus may bind to a transcription factor in vitro or in vivo. We also found that DYT1 bound to probes oMC6328, oMC6330 and oMC6338 containing two E-boxes that differ from the consensus sequence at the first position (Figure 3e,f), suggesting that DYT1 may bind to distinct sequences with different affinity. Weak DYT1–DNA interactions that evade detection may be greatly enhanced by cooperative binding of DYT1 to multiple sites.

We found that 12 genes differentially expressed in dyt1, in addition to MS1, contain the consensus DYT1 binding sites within 1 kb from the transcription start sites (Table S10), indicating that they may be DYT1 target genes. However, many other genes affected in the dyt1 mutant do not contain the consensus DYT1 binding site, implying that they may be indirectly regulated by DYT1. If these genes are directly regulated by DYT1, then DYT1 may also bind to other sequences in vivo, perhaps by interacting with other transcription factors, as supported by results described below.

DYT1 activates the expression of downstream transcription factors

The observation that the dyt1 mutation caused a reduction in the expression of some genes while increasing the expression of others suggested that DYT1 may act either as an activator or a repressor of gene expression. To distinguish these possibilities, we reasoned that, if DYT1 is an activator, fusion of DYT1 to a strong repressor domain would inhibit its function, resulting in a dyt1 mutant-like phenotype; in contrast, if DYT1 is a repressor, then fusion to a repressor domain would not affect male fertility. Therefore, we generated transgenic plants containing a fusion of the DYT1 promoter and coding region to sequences encoding SRDX, a modified strong repression domain from the SUPERMAN protein (Hiratsu et al., 2003) proDYT1:DYT1-SRDX/Col-0 plants produced mostly short siliques (Figure 4b) and sterile (Figure 4d) or partially fertile anthers (Figure 4e). The suppression of male fertility by DYT1–SRDX strongly suggests that DYT1 acts as an activator of gene expression.

Figure 4.

 Expression of the DYT1–SRDX fusion suppresses male fertility.
(a) Col-0 plant.
(b) DYT1–SRDX/Col-0 plant.
(c) Col-0 anther.
(d) Completely sterile DYT1-SRDX/Col-0 anther.
(e) Partial fertile DYT1-SRDX/Col-0 anther.
Scale bars = 50 μm in (c–e).

We then tested whether DYT1 can activate downstream gene expression using a fusion to the ligand-binding domain of the glucocorticoid receptor (GR), which was shown to regulate plant transcription factors in a ligand-dependent manner (Aoyama and Chua, 1997; Gatz, 1997; Wagner et al., 1999; Ito et al., 2004). The fusion construct proDYT1:DYT1-GR was introduced into the dyt1 background. proDYT1:DYT1-GR/dyt1 plants without dexamethasone (DEX) treatment produced mostly short siliques (Figure 5a) and mainly dead pollen grains (Figure 5c), with a few fertile siliques, probably due to leaky nuclear transport of DYT1–GR proteins. A single treatment with 10 μm DEX restored the male fertility of proDYT1:DYT1-GR/dyt1 (Figure 5b,d). To test regulation by DYT1, expression levels of MYB35 and MS1 were determined by real-time PCR at several time points after 10 μm DEX treatment. MYB35 gene expression rapidly increased by more than threefold after 1.5 h treatment with 10 μm DEX, and its elevated expression level was maintained for several hours. After 6 h, MYB35 expression started to decrease, with a transient increase at 12 h. The fluctuation of the MYB35 level may result from the transient strong DEX induction of DYT1–GR and a subsequent decline due to diminishing DEX concentration. MS1 was induced after 12 h treatment with 10 μm DEX. These results suggest that DYT1 may regulate the expression of different genes in temporally distinctive patterns. The expression level of MYB35 in dyt1 was reduced by more than tenfold compared with wild-type. However, its expression did not change in ms1. Therefore, MYB35 probably functions downstream of DYT1 but upstream of MS1, as also suggested by a previous study (Zhu et al., 2008).

Figure 5.

 DYT1–GR induction restores male fertility in dyt1.
(a) DYT1–GR/dyt1 plant without DEX treatment. Most siliques are short.
(b) DYT1–GR/dyt1 plant 10 days after 10 μm DEX treatment. Most siliques are long.
(c) A DYT1–GR/dyt1 anther without DEX treatment contains mostly dead pollen grains.
(d) A DYT1–GR/dyt1 anther 10 days after 10 μm DEX treatment contains mostly mature pollen grains.
(e) Real-time PCR result for MYB35 and MS1 expression in 10 μm DEX-treated flower buds. Three biological replicates were performed for each time point. The gene expression levels in mock-treated samples were set as 1. Additional mock treatments at 1.5 and 6 h did not cause significant changes in gene expression. Scale bars = 50 μm (c, d).

MYB35 functions as a transcriptional activator downstream of DYT1

Among the 22 transcription factor genes whose expression decreased in dyt1 anthers, MYB35 expression showed the most severe reduction. To test whether MYB35 acts as a transcriptional activator, we generated proMYB35:MYB35-SRDX/Ler transgenic plants (MYB35-SRDX for short). Of 80 MYB35-SRDX plants, 11 were completely male sterile. Unlike wild-type anthers (Figure 6a), the anthers of these 11 plants did not produce pollen grains (Figure 6f). These plants produced seeds after being pollinated with wild-type pollen, indicating that female fertility was not impaired. Semi-thin sections showed that tapetal cells of MYB35-SRDX anthers had larger vacuoles than those of wild-type at stage 6 (Figure 6b,g). At stage 7, MYB35-SRDX anthers completed meiosis and produced tetrads, but the tapetal cells had even larger vacuoles than those in wild-type (Figure 6c,h). No obvious defects of chromosome behaviors were observed in MYB35-SRDX meiocytes by DAPI staining (Figure S3p–t), indicating that MYB35 probably does not affect meiosis. At stage 8, MYB35-SRDX microspores were not released from the tetrads, and the tapetum degenerated earlier than that of wild-type (Figure 6d,i). Finally, MYB35-SRDX anthers lacked microspores after stage 9 (Figure 6j), unlike the wild-type anther (Figure 6e). The phenotypes of MYB35-SRDX sterile plants were similar to those of MYB35 RNAi plants (Figure S3f–j) and those of the myb35 tdf1 mutant previously reported (Zhu et al., 2008). These results suggest that MYB35 most likely functions as a transcriptional activator.

Figure 6.

MYB35-SRDX transgenic plants have an abnormal tapetum and microspore development.
(a–e) Wild-type (WT) anthers.
(f–j) MYB35-SRDX transgenic plant anthers. (a) WT anther staining showing mature pollen grains. (b–e) WT anthers of stages 6–9. (f) MYB35-SRDX anther staining showing no pollen grains. (g–j) MYB35-SRDX anthers of stages 6–9. Scale bars = 50 μm (a, f) and 10 μm (b–e, g–j).

DYT1 interacts with itself, AMS and other bHLH proteins in yeast two-hybrid assays

Previous studies showed that bHLH transcription factors usually form dimers to regulate target gene expression (Toledo-Ortiz et al., 2003; Li et al., 2006). In Arabidopsis, five bHLH genes (DYT1, AMS, At1g06170, At2g31210 and At2g31220) were found to be preferentially expressed in the anther and may function downstream of SPL and EMS1 (Wijeratne et al., 2007). Our yeast two-hybrid assays showed that DYT1 interacts with itself, AMS and the three other bHLH proteins mentioned above (Figure 7). Some bHLH proteins interact with MYB proteins (Payne et al., 2000; Abe et al., 2003; Zimmermann et al., 2004; Zhao et al., 2008), but DYT1 showed no detectable interaction with MYB33 (Figure 7). These results suggest that DYT1 may form transcription factor complexes that have different DNA binding specificities from the binding consensus derived from in vitro studies.

Figure 7.

 DYT1 interacts with itself, AMS and three bHLH transcription factors preferentially expressed in the anther, but not with MYB33.
Interactions were tested using proteins fused to the DNA-binding domain (BD) and DYT1 fused to the activation domain (AD). Yeast transformants were spotted on SD plates lacking His and Ade to assess potential interactions.

DYT1 and AMS regulate partially overlapping groups of genes

Because AMS was previously reported to function downstream of DYT1 (Xu et al., 2010), and the AMS protein interacts with DYT1 as described above (Figure 7), we analyzed the ams anther transcriptome to further examine the DYT1-controlled regulatory network. The transcriptome of ams anthers at stages 4–7 was analyzed in a recent study using Affymetrix ATH1 chips (Ma et al., 2012). A total of 1368 genes were differentially expressed in ams compared with wild-type (FDR <0.05 and more than twofold change), with 519 and 849 genes down- and up-regulated in ams, respectively (Figure 8 and Table S11).

Figure 8.

 Venn diagram showing the overlap between genes affected in dyt1 and ams anthers.
The differentially expressed genes in dyt1 and ams anthers were divided into eight sets according to their overlap. The numbers in the circles represent the number of genes for individual sets. The numbers outside the circles represent the total numbers of genes that are differentially expressed in dyt1 or ams.

In totally, 1796 genes were differentially expressed (more than twofold change compared with wild-type) in dyt1 and/or ams (Figure 8), including 655 genes that were affected in dyt1 and 227 genes that were altered in both mutants, the latter group representing 12.6% of all differentially expressed genes (Figure 8, sets 1–4). Among the 227 genes affected in both mutants, 214 were affected in the same direction in both mutants, with more genes down-regulated (Figure 8, set 1) than up-regulated (Figure 8, set 2), implying that both DYT1 and AMS are likely to be activators. Among the 22 transcription factor genes down-regulated in dyt1, nine were also down-regulated in ams, including AGL40, FLC/FLF and MYB103/MYB80 (Table S12). Therefore, some of the genes affected in both dyt1 and ams mutants may be regulated by a common set of transcription factors downstream of both DYT1 and AMS. Because of the possible DYT1–AMS interaction suggested by the yeast two-hybrid experiment, we hypothesized that DYT1 and AMS may regulate expression of these genes by forming heterodimers. Another possibility is that DYT1 may regulate this set of genes indirectly through AMS.

We also examined the non-overlapping sets of genes regulated by either DYT1 or AMS (Figure 8 and Table S13). Of the 1796 differentially expressed genes, 1569 were affected in only one mutant (Figure 8, sets 5–8), indicating that DYT1 and AMS have only partially overlapping functions in regulating the anther transcriptome. In contrast to the expectation that AMS may only regulate a subset of genes regulated by DYT1 because AMS expression is greatly reduced in dyt1, AMS actually regulates a larger number of genes (Figure 8). This contradiction may result from the residual expression of DYT1 (approximately one-third of the wild-type level) in the dyt1 mutant, in which the DYT1 promoter is partially disrupted but the coding region is intact (Zhang et al., 2006). The residual DYT1 function in dyt1 may be able to activate AMS expression to a low level, which may be sufficient to regulate a portion of downstream genes whose expression changes are not detected in the dyt1 mutant. However, ams is a null mutant, and AMS downstream genes are not buffered by residual AMS expression. Therefore, more genes were affected by the ams mutation than by the dyt1 mutation.

Discussion

For successful male reproductive development in plants, proper differentiation of meiocytes and surrounding somatic tissues including the tapetum is crucial. Transcriptional regulation is a major mechanism controlling cell differentiation. The dyt1 mutant is defective in tapetum function around the time of meiosis, providing an excellent system for studying regulation of the tapetum transcriptome. By investigating the transcriptome of anthers at stages 4–7, we were able to exclude noise from other stages or other floral organs. Our results on altered expression of genes in various biochemical pathways provide potential explanations for the morphological defects in the dyt1 tapetum. The tapetal cells in dyt1 are precociously vacuolated and degenerated near the time of meiosis, suggesting a role for DYT1 in promoting and maintaining active tapetum function to facilitate meiotic cytokinesis (Zhang et al., 2006). Our comparative transcriptome analysis of dyt1 and wild-type anthers indicates that DYT1 is required for normal expression of pathways including lipid metabolism and transport (Table S5), pollen coat formation (Table S6), cell-wall modification (Table S7), lignin and flavonoid biosynthesis (Table S8), and transport (Table S9).

The abnormally thin callose wall of dyt1 meiocytes (Figure 2) indicates that either callose synthesis in the meiocytes or callose precursor biosynthesis and transport processes in the tapetum are impaired. Our microarray data show that expression of known callose synthase genes was not reduced in dyt1, suggesting that meiocytes have callose synthesis machineries but lack building blocks (Ariizumi and Toriyama, 2011). Six sugar transporter genes are down-regulated in dyt1 (Table S9), suggesting that transport of callose precursors from the tapetum may be defective. Furthermore, 34 cell wall-related enzymes are down-regulated in dyt1 (Table S7). In normal anthers, these enzymes may be expressed in the tapetum during stages 4–7 and secreted for meiocyte callose wall dissolution after the tetrad stage. The enzymes may also be involved in dynamic tapetal cell-wall modifications to facilitate material transport from the tapetum to locules, as reported for the tomato anther (Polowick and Sawhney, 1993).

DYT1 also regulates the biosynthesis, storage and transport of pollen coat materials, including oleosins, arabinogalactan proteins, lipids and flavonoids (Table S5, S7, and S8). Oleosins, lipids and flavonoids are synthesized in tapetal cells and stored in organelles called tapetosomes until deposition onto the pollen coat after degeneration of tapetal cells (Hernandez-Pinzon et al., 1999; Hsieh and Huang, 2005, 2007). The glycine-rich protein GRP19 (oleosin-like protein), which is abundant on the pollen coat, functions as an emulsifier for lipid deposition, and is required for pollination (Mayfield et al., 2001). Defects in these pathways may affect male fertility (van der Meer et al., 1992; Aarts et al., 1997; Ariizumi et al., 2004). For example, the ms2 mutations cause pollen wall defects and male sterility (Aarts et al., 1997).

Previous studies suggested that tapetum function is tightly coupled with the progress of meiosis and pollen maturation (Ma, 2005; Zhu et al., 2008). Our transcriptome analysis suggested that DYT1 may function as a hub in the transcriptional regulatory network that precisely controls the tapetum function. We propose a putative regulatory network model (Figure 9) in which DYT1 regulates downstream genes through multiple mechanisms.

Figure 9.

 Model of gene regulatory network controlled by DYT1. DYT1, MYB35, AMS, MYB103/MYB80 and MS1 were reported to be essential for anther development.
Other genes are downstream of DYT1 but their functions in anther development are unknown. Arrows represent positive regulations. The double arrow represents a physical interaction.

In particular, DYT1 may regulate distinct temporal patterns of gene expression through feed-forward loops. A feed-forward loop is a minimally three-node network motif in which one transcription factor regulate a second transcription factor and together they regulate a third gene (Milo et al., 2002; Shen-Orr et al., 2002). DYT1 may regulate a target gene either by directly binding to its promoter or by activating an intermediate transcription factor that binds to the promoter of the same target gene (Figure 9). For example, DYT1 binds to the MS1 promoter, as shown by the gel-shift assay (Figure 3). DYT1 also positively regulates MYB99, which is also regulated by MS1 (Figure 9). Therefore, DYT1–MS1–MYB99 may form a feed-forward loop. In another transcription cascade, DYT1 regulates MYB35 and MYB103/MYB80, and MYB35 also regulates MYB103/MYB80 (Zhu et al., 2008). Thus, DYT1–MYB35–MYB103/MYB80 may also form a feed-forward loop (Figure 9). The feed-forward loops provide a mechanism for temporal separation of target genes of DYT1: those requiring only DYT1 may be activated as soon as the DYT1 level reaches a threshold, whereas those requiring both DYT1 and the product of another DYT1 target gene must wait for both regulators to be expressed at sufficient levels. However, concrete evidence for in vivo protein–promoter interactions is required to support the existence of these regulatory loops.

Another potential regulatory mechanism of DYT1 is direct regulation by DYT1 of a number of genes encoding enzymes and structural proteins, including cell-wall proteins; such a single input motif directly transfers a regulatory input signal to an effector for specific biochemical functions (Milo et al., 2002). Our microarray and in vitro binding site analyses suggest that DYT1 may directly regulate a group of effector genes to initiate tapetum function at early stages (Table S5–S10).

Our microarray results showed that expression of 166 genes was significantly reduced in both dyt1 and ams (Figure 8, set 1). These genes may be regulated by AMS alone according to the linear DYT1–MYB35–AMS cascade, or by DYT1–AMS heterodimers as suggested by the results of our yeast two-hybrid assay. The DYT1–AMS heterodimers may regulate a distinct group of genes that differ from those regulated by other complexes involving DYT1 or AMS separately. Expression of these genes requires a double regulatory input, and thus is limited to a defined time window when DYT1 and AMS are both present.

As shown in Figure 8, DYT1 and AMS also regulate distinct sets of genes. For example, 428 genes were regulated by DYT1 but not by AMS (Figure 8, sets 5 and 8). In other words, expression of these DYT1 downstream genes does not require AMS. Conversely, AMS regulates 1141 genes independently of DYT1 (Figure 8, sets 6 and 7). It is possible that the reduced AMS level in dyt1 (not a null mutant) allows residual AMS function, while the AMS function is abolished in ams (a null mutant). Formation of different transcriptional complexes may be another mechanism that differentiates target genes regulated by DYT1 and/or AMS. For example, a SET domain protein, ASH1-RELATED 3 (ASHR3), was shown to interact with AMS (Thorstensen et al., 2008). DYT1 may interact with various proteins, including three bHLH proteins (Figure 7). The different complexes may have different DNA binding specificities, and therefore regulate different genes.

In summary, this study highlights the function of DYT1 as a key transcriptional regulator of tapetum function and male fertility. Further investigations of the transcriptional regulatory network are necessary to provide more insights into the mechanisms controlling anther development in Arabidopsis.

Experimental Procedures

Plant materials and growth conditions

Plants were grown under long-day conditions (16 h light/8 h dark) in a 22°C growth room. The dyt1 mutant was genotyped as described previously (Zhang et al., 2006). The ams mutant (SALK_152147) was genotyped using primers oMC2325 and oMC1863 (Table S14).

Microarray

Anthers at stages 4–7 were dissected from 3 to 4-week-old plants. Total RNA was extracted using an RNeasy plant mini kit (Qiagen, www.qiagen.com). Three biological replicates for each genotype (Col-0, dyt1 and ams) were analyzed using Affymetrix Arabidopsis ATH1 chips (Affymetrix, www.affymetrix.com) at the Penn State Genomics Core Facility (University Park, PA). Raw data may be accessed through the National Center for Biotechnology Information Gene Expression Omnibus (accession number GSE18225). Microarray data were normalized using RAM (Bioconductor, www.bioconductor.org) in R. Genes determined to be absent by ‘MAS5 algorithm (Affymetrix) in wild-type anthers were considered as unreliable background signals (Zhang et al., 2005). The mean values of three replicates were transformed to log2 values. The LIMMA package (Bioconductor) was used to identify differentially expressed genes as described previously (Zhang et al., 2005). The differentially expressed genes were sorted in Excel with a twofold cutoff. Gene annotations were obtained from the Arabidopsis Information Resource website (http://www.arabidopsis.org).

Real-time PCR

cDNA was synthesized from 1 μg total RNA using SuperScript® II reverse transcriptase (Invitrogen, www.invitrogen.com) or M-MuLV reverse transcriptase (Fermentas, www.fermentas.com) and oligo(dT)16 according to the manufacturers’ instructions. The cDNA was diluted 50-fold and used as the template for real-time PCR using SYBR Advantage qPCR Premix (Clontech, www.clontech.com) according to the manufacturer’s instructions. The primers are listed in Table S14. Three biological replicates for each sample and three technical replicates for each gene were performed. UBQ1 or ACT2 was used as the internal control.

Gel-shift assay

The DYT1 coding sequence (CDS) was amplified by oMC2405 and oMC2406 using AccuPrimePfx DNA polymerase (Invitrogen), cloned into the pENTR-D/TOPO vector (Invitrogen) yielding pMC3329, and recombined into the pDEST17 vector (Invitrogen) via LR reaction using Gateway LR clonase II enzyme mix (Invitrogen). The resulting plasmid (pMC3250) was transformed into Escherichia coli BL21 (DE3) strain. Recombinant 6His–DYT1 proteins were purified as described previously (Huang et al., 1993).

Biotin-labeled dsDNA probes (pMS1, M1 and M2) were synthesized by annealing oligonucleotides oMC2781, oMC6286 or oMC6287, respectively, with oMC2746 and filling in using Klenow (NEB, www.neb.com). The in vitro binding reaction was performed as described previously (Huang et al., 1993). Free and bound probes were separated by 6% PAGE in 0.5× TBE at 100 V for 1 h, and transferred to Hybond N membrane (GE Healthcare, www.gelifesciences.com) using a Trans-Blot SD semi-dry transfer cell (Bio-Rad, www.bio-rad.com), and cross-linked to the membrane under UV light (120 mJ cm−2) for 20 sec. The signal was detected using a LightShift chemiluminescent EMSA kit (Pierce, www.piercenet.com) and X-ray film (Kodak, www.kodak.com).

Systematic evolution of ligands by exponential enrichment (SELEX)

The SELEX protocol was adapted from a previous study (Huang et al., 1993). A pool of ssDNA oMC2716 containing 30 nt random sequences flanked by 20 nt primer annealing regions was synthesized (Integrated DNA Technologies, www.idtdna.com). Biotin-labeled oMC2746 was annealed to oMC2716 to produce double-stranded DNAs after filling in using Klenow (NEB). The biotin-labeled dsDNA product was diluted 50-fold for gel-shift assays.

DYT1 protein extract (5 μl) and DNA probes (10 μl) were incubated in binding buffer (10 mm Tris/HCI, pH 7.5, 50 mm NaCl, 1 mm DTT, 1 mm EDTA, 5% glycerol, 50 μg ml−1 poly(dI-dC)(Sigma, P4929), 100 μg ml−1 BSA) for 20 min at 25°C. Then the mixture was separated in a 5% polyacrylamide gel (acylamide:bis-acrylamide=29:1) in 0.5× TBE running buffer. The gel slices containing the DYT1–DNA complex were cut out (the position for the shifted band was determined by running a duplicate sample in a separate lane and visualizing using a LightShift chemiluminescent EMSA kit), and soaked in 1× TE buffer overnight at 4°C. The eluted DNA was amplified by PCR for 20–30 cycles (each cycle consisted of 95°C for 20 sec, 60°C for 20 sec, and 72°C for 20 sec) using oMC2718 and oMC2746, followed by DNA purification before the next round of the gel-shift assay. After six rounds, the DNA was amplified for 30 cycles and cloned into pGEM-T vector (Promega, www.promega.com). Plasmids were sequenced using the T7 primer to determine the insert sequences. The WebLogo tool (http://weblogo.berkeley.edu) was used to analyze the enrichment of binding sites among the cloned DNA probes.

Construction of the proDYT1:DYT1-GR and proDYT1:DYT1-SRDX constructs

A genomic region containing the DYT1 gene (without the stop codon) and 1444 bp of upstream sequence was amplified using oMC2469 and oMC2470. The GR domain was amplified from the pBI-GR plasmid (Lloyd et al., 1995) using oMC2471 and oMC2472. These two fragments were cloned into pCAMBIA1300 (Cambia, www.cambia.org/daisy/cambia/585) yielding proDYT1:DYT1-GR. proDYT1:DYT1-SRDX was amplified using oMC2801 and oMC2802 and cloned into pCAMBIA1300. The constructs were transformed into Agrobacterium tumefaciens strain C3581, which was used to transform plants by the floral-dip method (Clough and Bent, 1998). proDYT1:DYT1-GR/dyt1 plants were treated with 10 μm DEX (10 mm DEX stock solution in 100% ethanol diluted 1000-fold in 0.015% Silwet L-77) or mock-treated using 0.1% ethanol in 0.015% Silwet L-77. (LEHLE SEEDS, www.arabidopsis.com)

Construction of the proMYB35:MYB35-SRDX and MYB35 RNAi plasmids

A genomic region containing the MYB35 gene without the stop codon and 1290 bp of the upstream region was amplified using oMC2922 and oMC2923 and cloned into pCAMBIA1300. oMC3095 and oMC3096 were annealed to form dsDNA encoding the SRDX domain with PstI and HindIII overhangs, before being ligated to the 3′ end of proMYB35:MYB35. To generate the MYB35 RNAi construct, the MYB35 coding region was cloned into the pRR2222 vector as inverted repeats (Zhao et al., 2003). The native promoter and the inverted repeats of MYB35 were inserted into plant expression vector pMDC7 (Curtis and Grossniklaus, 2003). Plant transformation was performed using the floral-dip method (Clough and Bent, 1998).

Yeast two-hybrid assay

The DYT1 CDS was cloned into pDEST22 (N-GAL4 AD, Invitrogen) using the Gateway cloning system (Invitrogen).The DYT1, AMS, At1g06170, At2g31210, At2g31220 and MYB33 CDS were cloned separately into pDEST32 (N-GAL4 BD, Invitrogen). Plasmids were transformed into the PJ694A yeast strain using the LiAc/polyethylene glycol method. Transformants containing both the bait and prey constructs were selected on SC dropout mixture lacking Trp and Leu plates and streaked onto plates containing SC dropout mixture lacking His and Ade to test for positive interactions.

Transmission electron microscopy

TEM samples were prepared as previously described (Li et al., 2004). Ultra-thin sections were analyzed using a JEOL JEM 1200 EXII transmission electron microscope (JEOL, www.jeol.com) at the Microscopy and Cytometry Facility of Pennsylvania State University (University Park, PA).

Light microscopy

Arabidopsis flower buds were fixed and embedded in Spurr’s resin as described previously (Zhang et al., 2006). Semi-thin (5 μm) sections were prepared using a Reichert-Jung Ultracut E microtome (Leica, www.leica.com/) at the Pennsylvania State University Microscopy and Cytometry Facility (University Park, PA. Sections were stained with 0.1% toluidine blue O for 30 sec at 60°C. Images were taken using Nikon Eclipse E400 microscope (Nikon, www.nikon.com) connected to a DEI-750 CE camera head (Optronics, www.optronics.com).

Acknowledgements

We thank Dr H. Huang ((Institute of Plant Physiology and Ecology, Shanghai, China) for advice on gel-shift assays, Dr N. Altman (Department of Statistics, Pennsylvania State University, University Park, USA) for help with statistical analysis of microarray data, Dr W. Zhang (Shanghai Key Laboratory of Bio-energy Crops, Shanghai University, Shanghai, China) for helpful discussion, and Ms J. Wang for plant care. We thank Dr Z. Wilson (School of Biosciences, University of Nottingham, Sutton Bonington Campus, Loughborough, Leics, UK) for sharing unpublished results. This work was supported by a US Department of Energy grant (DE-FG02-02ER15332) to H.M. and by the Department of Biology Huck Institutes of the Life Sciences, Pennsylvania State University (University Park, PA) and Fudan University (Shanghai, China).

Ancillary