Regulation of Arabidopsis tapetum development and function by DYSFUNCTIONAL TAPETUM1 (DYT1) encoding a putative bHLH transcription factor

In flowering plants, pollen grains are formed within the anther region of the male reproductive organ, the stamen(Ma, 2005). Molecular genetic studies have revealed that the B and C functions of the wellknown ABC model together determine the stamen identity, whereas the C function alone specifies the carpel identity (Coen and Meyerowitz,1991; Ma, 2005). In Arabidopsis, APETALA3 (AP3) and PISTILLATA(PI) are two essential B function genes, and AGAMOUS(AG) is required for C function(Coen and Meyerowitz, 1991; Ma, 2005). The Arabidopsis anther is a bilaterally symmetrical structure with four lobes. Each lobe comprises four distinct somatic cell layers from outer to inner: they are the epidermis, endothecium, middle layer and tapetum(Goldberg et al., 1993; Sanders et al., 1999). The tapetum surrounds meiocytes and generates many proteins, lipids,polysaccharides and other molecules necessary for pollen development(Goldberg et al., 1993). Accordingly, the tapetum is characterized by active protein synthesis and secretion, a high rate of energy metabolism, and expression of many tapetum-preferential genes (Dickinson and Bell, 1976; Hernould et al.,1998; Liu and Dickinson,1989; Rubinelli et al.,1998; Scott et al.,2004; Taylor et al.,1998; Zheng et al.,2003).

Several genes have been identified that are required for normal anther development (Ma, 2005). For example, the SPOROCYTELESS/NOZZLE (SPL/NZZ) gene is required for cell type specification in both male and female reproductive organs(Schiefthaler et al., 1999; Yang et al., 1999). The spl/nzz mutant anthers lack the endothecium, middle layer, tapetum and meiocytes. Recently, Ito et al. (Ito et al., 2004) found that SPL/NZZ is a direct target of AG, and that ectopic expression of SPL/NZZ, independent of AG,induces the formation of anther locule with pollen grains(Ito et al., 2004). SPL/NZZ encodes a putative transcription factor(Schiefthaler et al., 1999; Yang et al., 1999), suggesting that SPL/NZZ regulates anther cell differentiation by activating downstream genes.

Furthermore, EXCESS MICROSPOROCYTES1/EXTRA SPOROGENOUS CELLS(EMS1/EXS) and TAPETUM DETERMINANT1 (TPD1) are early (pre-meiosis) genes that encode a receptor-like protein kinase and a small protein, respectively, and act in the same genetic pathway to control the tapetal cell identity (Canales et al.,2002; Sorensen et al.,2003; Yang et al.,2003; Yang et al.,2005; Zhao et al.,2002). Recently, the SERK1 and SERK2 genes encoding closely related receptor-like protein kinases were shown to have redundant functions in controlling tapetum formation(Albrecht et al., 2005; Colcombet et al., 2005). MALE STERILITY1 (MS1) and ABORTED MICROSPORES(AMS) act after meiosis and encode a PHD domain-containing protein and a bHLH transcription factor, respectively; they are essential for late stage functions of the tapetum (Ito and Shinozaki, 2002; Sorensen et al., 2003; Wilson et al.,2001). Remarkably, the `early' gene EMS1/EXS is expressed in both stamen and gynoecium, whereas the `late' genes MS1 and AMS are anther specific. However, it is not known how the`male-specific' MS1 and AMS expression is regulated.

Here, we report the isolation of a new Arabidopsis mutant, which is male sterile and defective in tapetum differentiation and function. Because the mutant has an abnormal tapetum, we named the gene DYSFUNCTIONAL TAPETUM1 (DYT1). DYT1 encodes a putative bHLH transcription factor and is preferentially expressed in tapetal cells as early as anther stage 5, spatially similar to, but temporally earlier than, MS1 and AMS. Furthermore, our results suggest that DYT1 probably acts downstream of SPL/NZZ and EMS1/EXS, and is required for normal expression of AMS, MS1and other tapetumpreferential genes. We propose that DYT1 is a component of a genetic network for tapetum differentiation and function.

Arabidopsis thaliana plants in this study are in the Landsberg erecta (Ler) background for all experiments with the exception of the mapping of the DYT1 gene, for which we crossed dyt1 with Columbia. The spl(Yang et al., 1999), ems1 (Zhao et al.,2002) and tpd1 (Yang et al., 2003) mutants were kindly provided by Drs W. Yang, D. Zhao and D. Ye, respectively. The dyt1 mutant was isolated from the progeny of a Ds insertional line (ET4262). The ems1 dyt1double mutant was obtained by pollinating the dyt1 pistil with pollen from an ems1/+ plant. Plants were grown under long-day conditions (16 hours light/8 hours dark) in a ∼22°C growth room.

Plants were photographed with a Sony digital camera, DSC-F707 (Sony Corp.,Tokyo, Japan). Flower pictures were taken using a Nikon dissecting microscope(Nikon Corp., Tokyo, Japan) with an Optronics digital camera (Optronics,Goleta, CA, USA). To determine pollen viability, mature anthers were stained with the Alexander solution (Alexander,1969) and photographed under an Olympus BX-51microscope (Olympus,Tokyo, Japan) with a SPOT II RT camera (Diagnostic Instruments, Sterling Heights, MI, USA). Chromosome spreading and DAPI staining were performed as described (Ross et al., 1996). Wild-type and mutant inflorescences were collected and fixed as described(Zhao et al., 2002). Floral buds were embedded in Spurr's resin; semi-thin (0.5 μm) sections were made by using an Ultracut E ultramicrotome (Leica Microsystems, Nussloch, Germany),stained with 0.05% of Toluidine Blue O for 40 to 60 seconds, and photographed under the Olympus BX-51 microscope with the SPOT II RT camera.

The dyt1 mutation was found to be distinct from the Dsinsertional locus (not shown). The dyt1 mutant (in the Lerbackground) was crossed with Columbia to obtain F1 and F2 seeds. About 500 F2 plants with mutant phenotypes were genotyped by using the SSLP and dCAPS markers (Li et al., 2001). Several predicted genes were found in the mapped region flanked by recombination events; these genes were amplified from wild-type and dyt1 plants and sequenced. TAIL PCR(Liu and Whittier, 1995) was used to determine the sequences of a retrotransposon insertion in the dyt1 mutant. To verify the DYT1-coding region, we used the primers oMC1872/oMC1873 (see Table 1 for all primer sequences), designed according to the annotated At4g21330 locus, to amplify wild-type Ler cDNA.

To generate a complementation construct, a 2.7 kb wild-type genomic fragment was amplified by PCR using primers oMC2241 and oMC2242, and then cloned into a modified pCAMBIA1300, yielding pMC2969. The Agrobacterium strain GV1301 was transformed with the plasmid pMC2969,then used to transform dyt1/+ plants. The seeds of transformed plants were screened for hygromycin-resistant seedlings, which were transferred into the soil. The T1 plants were genotyped to identify homozygous dyt1plants with the T-DNA, using a primer (oMC1945) for the retrotransposon insertion at the DYT1 locus and DYT1-gene-specific primer oMC1834. The oMC1823/oMC1873 and oMC2194/oMC1834 primers were used to identify the dyt1 allele and the transgene, respectively.

A set of genes were selected for expression analysis by RT-PCR because they are known to be important for male reproduction: SPL/NZZ(Schiefthaler et al., 1999; Yang et al., 1999), EMS1/EXS (Canales et al.,2002; Zhao et al.,2002), TPD1 (Yang et al., 2003; Yang et al.,2005), AMS (Sorensen et al., 2003), MS1(Ito and Shinozaki, 2002; Wilson et al., 2001), MALE STERILITY2 (Aarts et al.,1997), MALE STERILITY5(Glover et al., 1998), AtMYB32 (Preston et al.,2004), AtMYB103(Higginson et al., 2003), A6 and A9 (Sorensen et al.,2003), AtMYB33 and AtMYB65(Millar and Gubler, 2005), SOLO DANCERS (SDS) (Azumi et al., 2002), and ROCK-N-ROLLERS/AtMER3(RCK/AtMER3) (Chen et al.,2005; Mercier et al.,2005). A second group of 16 genes was identified using microarray data of wild-type and ems1 mutant anthers we had obtained in our laboratory (W.Z., Y.S., A. Wijeratne and H.M., unpublished). The genes that were expressed in the ems1 anther at levels that were one half or less of those in the wildtype anthers were regarded as candidate tapetum-preferential genes because the ems1 anthers lack tapetum. In addition, UBQ1 and AtMYB4(Vannini et al., 2004) were used as constitutive expression controls. The primers for RTPCR and relevant microarray data are provided in the Tables 1 and 2, respectively. The primers for real-time PCR are listed in Tables 1 and 3. The PCR and data treatment were carried out as described previously(Ni et al., 2004). Plant tissue was collected and quickly frozen in liquid nitrogen. The anthers at approximately anther stages 4 to 7 were collected under a dissection microscope. Total RNA was extracted using the RNeasy Plant Kit (Qiagen,Valencia, CA) from young inflorescences (approximately floral stage 1-10). 1-2μg total RNA was used for reverse transcription according to the manufacturer's instruction to synthesize cDNA, which was used directly as PCR templates (Invitrogen, Carlsbad, CA).

Non-radioactive RNA in situ hybridization was performed as described(Li et al., 2004). A 624 bp DYT1 cDNA fragment was amplified using DYT1-specific primers with XbaI and BglII sites at the 5′ end: oMC2238 and oMC2239, respectively. The PCR product was confirmed by sequencing and cloned into the pGEM-T vector (Promega, Madison, WI) to yield plasmid pMC2949. Plasmid DNA was completely digested by XbaI or BglII and used as template for transcription with SP6 or T7 RNA polymerases,respectively (Roche, Mannheim, Germany). Images were obtained using the Olympus BX-51 microscope with the SPOT II RT camera and edited using PHOTOSHOP 5.0 (Adobe, San Jose, CA).

A DYT1 cDNA fragment was amplified using gene-specific primers oMC1872 and oMC1873 (Table 1),then cloned into pGEM-T vector to yield plasmid pMC2941. After verification by sequencing, the DYT1 cDNA fragment was subcloned into pCAMBIA1300 downstream of the CaMV 35S promoter to produce the plasmid pMC2942. An Agrobacterium strain C3581 carrying pMC2942 was used to transform wild-type, ems1/+ and spl/+ plants. The transgenic plants were selected by hygromycin resistance and verified using PCR with oMC1872 and oMC1873. EMS1 gene-specific primers oMC499 and oMC500, SPL/NZZ gene-specific primers oMC2044 and oMC2045, plus Ds5-specific primer oMC490 were used to identify ems1 and splheterozygous and homozygous plants, respectively. Paraffin sections were prepared as described for in situ experiments (above) and photographed as described for semi-thin sections.

The protein sequences of the nine genes from group 9 of the Arabidopsis bHLH family, including DYT1, were used to search for the closest homologs in both the rice genome(http://tigrblast.tigr.org/euk-blast/index.cgi?project=osa1)and Populus genome(http://www.floralgenome.org/cgi-bin/tribedb/tribe.cgi)using both BLAST and TBLASTN programs with a cutoff of 1E-15(Heim et al., 2003; Toledo-Ortiz et al., 2003). The multiple sequence alignment of full-length protein sequences was performed using ClustalX (Plate-Forme de Bio-Informatique, Illkirch Cedex, France) with a combination of GOP=4.0 and GEP=0.1. The bHLH domain region and additional conserved regions were aligned and used to perform neighbor joining (NJ)analyses with the `pairwise deletion' option, `P-distance' model and 1000 bootstrap replicates test using MEGA version 3.0(Kumar et al., 2004)(http://www.megasoftware.net/index.html).

To identify new Arabidopsis genes important for anther development, we screened for male sterile plants among Ds transposon insertional lines. One line was found to segregate normal and sterile plants with an approximate 3:1 ratio. Pollination of the mutant pistil with wild-type pollen indicated that the mutant is female fertile. The mutant was named dysfunctional tapetum1 (dyt1) because of its anther defects(see below). The vegetative growth of the dyt1 mutant appeared normal(Fig. 1A,B) and most mutant floral organs were also normal with the exception of shorter stamen filaments and smaller anthers (Fig. 1C,D). There were no pollen grains on the anther surface of opened flowers (Fig. 1G). Occasionally mutant siliques contained a few seeds (not shown), possibly owing to residual gene function (see below).

Detailed analyses were performed to understand the mutant developmental defects. Chromosome spread experiments were performed and revealed that normal meiotic features could be observed in the mutant meiocytes, demonstrating that meiotic nuclear divisions can proceed normally(Fig. 2). We also generated semi-thin anther sections to investigate the mutant anther development(Fig. 3). From anther stage 1 to 3, the dyt1 anthers appeared normal (data not shown). At stage 4,the mutant anther was similar to the wild-type anther in cell layer and cell number, but had a slightly different shape from the wild-type anther and was vacuolated in more cells than normal. In addition, the mutant sporogenous cells appeared more deeply stained than wild-type cells(Fig. 3A,B). At early stage 5,the dyt1 anther lobe had four cell types interior to the epidermis,similar to the wild type. The wild-type anther at late stage 5(Fig. 3E) was also vacuolated in more cells than earlier and had deeply stained meiocytes. Therefore, mutant anthers at stage 4 and early stage 5 exhibited these morphological features precociously (Fig. 3A-D).

In the wild-type anther at late stage 5(Fig. 3E), the tapetum had significantly larger cells than earlier. In the mutant(Fig. 3F), additional vacuoles were observed in tapetal cells, with a reduction of the cytoplasm. The vacuoles in the wild-type tapetum at this time are fewer and smaller than those in the mutant. In addition, the mutant middle layer maintained its thickness with vacuolation, unlike the reduced thickness of the wild-type middle layer (Fig. 3E,F). At stage 6, a thick callose wall forms around the meiocytes(Fig. 3G). By contrast, the mutant meiocytes had very thin callose cell walls(Fig. 3G,H). At this stage,most of the mutant meiocytes were undergoing meiosis(Fig. 2), but some of them had collapsed. At stage 7 and stage 8, the completion of wild-type meiosis results in the formation of tetrads and then microspores, but mutant tapetal and middle layer cells swelled with expanded vacuoles and filled the center of the locules where the meiocytes had collapsed and degraded(Fig. 3I-L).

To gain further insights into its function, we cloned the DYT1gene. The dyt1 mutant was from a Ds insertional line, but the dyt1 mutation was genetically separable from the Dselement (data not shown). To clone the gene, we mapped the dyt1 locus by analyzing ∼500 mutant F2 progenies from a cross between the dyt1 mutant (Ler) and Columbia wild-type plant. The mapping results indicated that the DYT1 gene was on chromosome 4, between At4g21220 and At4g21360 (data not shown). We sequenced candidate genes in this region from both wild-type and the dyt1 mutant, and found that only one locus, At4g21330, had an insertion mutation 109 bp upstream of the predicted translation initiation codon(Fig. 4A). We performed TAIL PCR to obtain the DNA at the 5′ and 3′ ends of the insertion and found that they each partially matched the same region of a putative retro-transposon At5g33382 in the Columbia genomic sequence. Further PCR and sequence analysis indicated that the insertion at the DYT1 locus matched exactly and completely to a seemingly intact retro-transposon in the Ler genome (W.Z., Y.S. and H.M., unpublished). To verify that At4g21330 is DYT1, we cloned an At4g21330 genomic fragment into a modified pCAMBIA1300 plasmid and used it to transform dyt1/+ plants. Fifty independent transgenic plants were analyzed; 12 lines were found to be homozygous for the dyt1 insertion. All 12 lines were fertile,including nine lines with normal fertility(Fig. 1E,H), confirming that At4g21330 is the DYT1 gene.

To determine the DYT1 cDNA sequence, we performed an RTPCR experiment with floral mRNA, and obtained a 624 bp product with an identical sequence to that indicated by the annotation. Additional RT-PCR experiments using primers that match sequences just beyond the annotated region yielded no product, confirming the annotated DYT1-coding region. The DYT1 gene encodes a putative transcription factor of 207 amino acid residues with a basic helix-loop-helix (bHLH) domain. According to the annotation(http://www.arabidopsis.org),the bHLH domain spans the region from Phe₂₉ to Gln₇₈(Fig. 4B)(Toledo-Ortiz et al., 2003). Interestingly, a preliminary search of public databases using the DYT1 protein sequence with the BLAST program showed that the DYT1 protein has the highest similarity to the Arabidopsis AMS protein(Sorensen et al., 2003).

To gain additional insights into the phylogenetic relationship between DYT1, AMS and other close homologs, we performed phylogenetic analysis of bHLH genes, including DYT1, AMS and a recently reported rice gene, UNDEVELOPED TAPETUM (OsUDT1), which is required for normal tapetum development (Jung et al., 2005). The phylogenetic analyses with the bHLH domain region alone, with both bHLH domain and conserved regions(Fig. 4C), or with the full-length sequences yielded trees with very similar topologies (others not shown). Our result indicates that DYT1, AMS, OsUDT1, Os02g02820 and PtDYT1-Like (Populus trichocarpa) form a separate clade within a group of related members of the bHLH gene family. Among them, AMS and Os02g02820 are supported as an orthologous pair, as are DYT1 and PtDYT1-like. In Arabidopsis, DYT1 and AMS are most closely related to each other, in agreement with the preliminary BLAST results. The rice OsUDT1 gene could be the ortholog of the DYT1 and PtDYT1-like genes.

To determine the DYT1 expression pattern, we performed a realtime PCR experiment (Fig. 5A). DYT1 expression was detected at a low level in young inflorescences with meiotic cells and in siliques, and at a high level in the wild-type anthers from stage 4 to 7, but not in vegetative tissues or in the post-meiotic stage-12 flower. In addition, RNA in situ hybridization experiments showed that DYT1 expression was detectable in the floral meristem and early anther primordia (Fig. 5C), and in archesporial cells at stage 2 (not shown). From stage 4 to early stage 5, a weak signal was detected in precursors of the middle layer, tapetum and meiocytes (Fig. 5D). From late stage 5 (Fig. 5E,H) to early stage 6, the DYT1 expression reached its highest level in the tapetum and is at a low level in meiocytes. However, at late stage 6, the DYT1 expression signal was drastically reduced to background levels (Fig. 5F). The DYT1 expression pattern is consistent to the observed tapetal defects in the dyt1 mutant anther. In the gynoecium, a weak signal was detected (data not shown). In the dyt1 anther, there was a low level expression, which was not specific to the tapetum(Fig. 5I). Therefore, the insertion upstream of the DYT1 ATG codon caused a great reduction of the level of DYT1 expression in the tapetum. The weak expression suggests that the dyt1 allele may have some residual function, which might account for the occasional seeds that were observed.

To test whether the DYT1 expression was affected by any known early anther development genes, we performed real-time PCR experiments with RNA from spl and ems1 mutant floral buds. Our results showed that the DYT1 expression was barely detectable in the splinflorescences and that the expression level in the ems1inflorescences was only ∼17% of the wild-type level(Fig. 5B). These results suggest that DYT1 might be downstream of SPL/NZZ and EMS1/EXS. We performed RNA in situ experiments to verify this possibility. In the spl anther, little DYT1 expression could be detected (Fig. 5J). A weak signal could be detected in the meiocytes in the ems1 anther at late stage 5 (Fig. 5K) and early stage 6, but, unlike the wild-type anther, the strong tapetal signal was not found in the ems1 anther (compare Fig. 5H with 5K). Therefore,the strong DYT1 expression in the tapetum requires EMS1/EXS. Both the real-time PCR and the RNA in situ hybridization results indicate that SPL/NZZ is essential for DYT1 expression and that EMS1/EXS promotes the high-level DYT1 expression specific to the tapetum.

The finding that DYT1 encodes a bHLH-type putative transcription factor suggests that DYT1 controls gene expression required for normal anther development. To test this hypothesis, we performed RT-PCR using primers for anther genes. We obtained results for a total of 32 genes(Fig. 6); among these, SPL/NZZ, EMS1/EXS and TPD1 are known early anther development genes (Canales et al.,2002; Schiefthaler et al.,1999; Yang et al.,1999; Yang et al.,2003; Yang et al.,2005; Zhao et al.,2002). The other genes were chosen for their tapetum-preferential expression according to either previous reports or to gene expression data obtained in our laboratory (Table 2). We found that for 21 genes out of 32, the expression was significantly reduced in the dyt1 mutant compared with the wild type(Fig. 6), indicating that indeed the normal expression of a large number of genes depends on the DYT1 gene function. In particular, two regulatory genes, MS1and AMS, which are important for tapetum development(Ito and Shinozaki, 2002; Sorensen et al., 2003; Wilson et al., 2001),exhibited greatly reduced levels of expression, suggesting that MS1and AMS act downstream of DYT1. By contrast, some tapetum preferential genes, such as A6 and A9, were still expressed in the dyt1 mutant at slightly reduced levels. In addition, the expression of SPL, TPD1, EMS1/EXS, AtMYB33 and AtMYB65 was not dramatically different in the dyt1 mutant, indicating that their expression does not require DYT1. To verify the RT-PCR results,selected genes were further analyzed using real-time PCR and the results(Fig. 6D) support the conclusion that AMS and MS1 expression requires DYT1 function.

In addition to the tapetum-preferential genes, we also tested the expression of two meiosis-specific genes: SDS and RCK/AtMER3(Azumi et al., 2002; Chen et al., 2005; Mercier et al., 2005). Although the expression level of SDS did not change, the expression of RCK/AtMER3 was significantly reduced. SDS is known to act earlier than RCK/AtMER3 in prophase I, suggesting that the dyt1 mutation might affect the expression of late prophase I genes more than early prophase I genes.

Both ems1/exs and dyt1 mutations affect tapetum development. Previous reports and our results suggest that DYT1 acts downstream of EMS1/EXS. To test this further, we generated the ems1 dyt1 double mutant and examined its early anther development. We found that the double mutant resembles the ems1 mutant in that the double mutant anther also completely lacks the tapetum(Fig. 7), suggesting that indeed EMS1/EXS is upstream of DYT1 in the same pathway. In addition, to test whether DYT1 is sufficient to alter anther development, we generated transgenic plants carrying a 35S-DYT1fusion in wild-type, spl/nzz and ems1/exsbackgrounds. Transgenic lines with DYT1 overexpression were identified by RTPCR (Fig. 6E;data not shown) and analyzed for their anther morphology. In all cases, the 35S-DYT1 transgenic anthers had morphologies resembling those of the corresponding genotypes without the transgene (see Fig. S1 in the supplementary material). Therefore, the overexpression of DYT1 was not able to suppress the spl/nzz and ems1/exs mutant phenotypes, indicating that other genes acting downstream of SPL/NZZand EMS1/EXS are probably required for normal tapetum development. Because DYT1 was found to be required for normal expression of a number of tapetum genes, we tested selected genes by real-time PCR to determine whether the 35S-DYT1 transgene was able to stimulate the expression of these genes. Our results indicate that the 35S-DYT1transgene did not alter the expression of these genes substantially(Fig. 6E), consistent with the morphological results.

Our results and previous studies suggest that additional genes other than DYT1 probably function downstream of SPL/NZZ and EMS1/EXS to promote tapetum development. It is possible that some of these genes might encode regulatory proteins. For example, the AtMYB33 and AtMYB65 genes are known to be important for tapetum development, and might also function downstream of SPL/NZZand EMS1/EXS. To test these ideas, we examined the expression in wild-type and mutants of these genes and of other genes, which were identified as tapetum-preferential from our microarray data (W.Z., Y.S. and H.M.,unpublished) and encode bHLH or WD-40 proteins. Real-time PCR results(Fig. 6F) indicate that the expression of two bHLH genes (At2g31210 and At1g06170) was greatly reduced in the spl mutant, and one of them (At2g31210) was expressed at a lower level in the ems1 mutant. By contrast, the expression levels of AtMYB33, AtMYB65 and two other genes were either nearly normal or increased.

Our analysis of the dyt1 mutant phenotype indicates that DYT1 is required for normal tapetum development following its formation. Furthermore, DYT1 is expressed preferentially in the tapetum at late stage 5, consistent with its role in this layer. The EMS1/EXS, SERK1/SERK2 and TPD1 genes(Albrecht et al., 2005; Canales et al., 2002; Colcombet et al., 2005; Yang et al., 2003; Yang et al., 2005; Zhao et al., 2002) are important for the formation of the tapetal layer; therefore, these genes probably act at earlier stages than DYT1, as supported by the observed strong tapetal expression of EMS1/EXS and TPD1(Canales et al., 2002; Yang et al., 2003; Zhao et al., 2002) prior to the strong DYT1 expression in the tapetum. Our expression and double mutant analyses support the hypothesis that DYT1 acts downstream of SPL/NZZ and EMS1/EXS.

AMS and MS1 are also important for tapetum function, but they are required at post-meiotic steps(Ito and Shinozaki, 2002; Sorensen et al., 2003; Wilson et al., 2001). In ams anthers, meiosis is normal and microspores are formed; however,the newly formed microspores soon degenerate. In ms1 anthers, pollen development is abnormal and no normal mature pollen is produced. Compared with AMS and MS1, DYT1 acts at an earlier stage, before the completion of meiosis. Therefore, DYT1 is required for a key step in tapetum development. In other words, tapetum development requires the combined activities of the EMS1/EXS, SERK1/SERK2, TPD1, DYT1, AMS and MS1 genes: first EMS1/EXS, SERK1/SERK2 and TPD1 specify the tapetal cells as distinct from meiocytes at the time of the cell division that form the tapetal cells, then DYT1 is required to promote correct tapetal cell fate for proper function, and finally AMS and MS1 further regulate the tapetal cell function supporting normal microspore development.

As DYT1 encodes a bHLH putative transcription factor, it is likely that it regulates the expression of tapetal genes. We found that the expression of a majority of tapetum-preferential genes tested depends on DYT1. The greatly reduced expression in the dyt1 mutant of many tapetum-preferential genes, particularly those encoding transcription factors, supports the idea that DYT1 is a key component of a regulatory step in normal tapetum development. The Arabidopsis bHLH gene family has over 140 members, making this the second largest gene family of transcription factors (Toledo-Ortiz et al., 2003). Although DYT1 and AMS clearly have non-redundant and distinct functions, they are members of the same subfamily. Phylogenetic analysis performed here including the closest rice homologs of DYT1 indicates that the rice gene, OsUDT1, is also a member of this subfamily and a putative DYT1 ortholog. A mutation in the OsUDT1 gene results in a defective tapetum(Jung et al., 2005), similar to the dyt1 tapetum. In addition, the expression pattern of the OsUDT1 gene (Jung et al.,2005) is different from that of DYT1. Therefore, this bHLH subfamily contains phylogenetically and functionally distinct members.

Some tapetum marker genes, such as A6 and A9, were expressed in the dyt1 mutant at slightly reduced levels. It is possible that other genes are also important for the activation of some tapetum genes. In addition to DYT1, previous reports described mutants with similar phenotypes,such as fat tapetum, gne1 and gne4(Sanders et al., 1999; Sorensen et al., 2002). Although the molecular nature of the FAT TAPETUM, GNE1 and GNE4 genes are unknown at this time, it is likely that additional loci are involved in defining the tapetal cell fate. In other words, DYT1 is essential, but not sufficient, for the specification of the tapetum identity, as supported by our observation that overexpression of DYT1 did not alter anther phenotypes in wild-type or mutant backgrounds. Recently, it was reported that the AtMYB33 and AtMYB65 genes redundantly facilitate tapetum development, with the double mutant having tapetum defects before the completion of meiosis(Millar and Gubler, 2005). It is known that bHLH transcription factors can form homodimers or heterodimers with other bHLH proteins. In some cases, it has been shown that bHLH proteins can form complexes with MYB proteins and WD-40 proteins(Ramsay and Glover, 2005). It is possible that the AtMYB33 and AtMYB65 proteins may form heterodimers with DYT1 to regulate tapetum-preferential gene expression. This idea is consistent with our result that the expression of these two MYB genes is not altered in the dyt1 mutant.

Our analysis of gene expression in the spl mutant suggest that two other genes (At2g31210 and At1g06170) encoding bHLH proteins might also act downstream of SPL/NZZ and that these bHLH genes might in turn regulate other genes. Indeed, an examination of upstream sequences (500 bp from the ATG codon) of 163 putative tapetum-preferential genes (identified from microarray data, W.Z., A. Wijeratne, Y.S. and H.M., unpublished) revealed that 143 of them have potential binding sites for bHLH proteins, and 69 genes have putative MYB-binding sites, with 59 genes have both types of elements. These observations support the hypothesis that bHLH and MYB proteins regulate overlapping but non-identical sets of tapetum-preferential genes. In particular, 18 out of 21 genes that exhibit reduced expression in the dyt1 mutant have putative bHLH-binding sites; in addition, between 500-1000 bp upstream of the AMS ATG codon, there is a bHLH-binding consensus site. Further experiments are needed to test whether these genes are direct targets of DYT1 and/or other bHLH proteins.

It is known that a functional tapetum is required for normal pollen development following meiosis, as shown by molecular ablation studies and the characterization of mutants such as ams(Aarts et al., 1997; Ito and Shinozaki, 2002; Sorensen et al., 2003; Wilson et al., 2001). In the ems1/exs, serk1 serk2 and tpd1 mutants, the tapetum is missing and excess meiocytes occupy the position of the tapetum(Albrecht et al., 2005; Colcombet et al., 2005; Yang et al., 2003; Zhao et al., 2002). Nevertheless, meiotic nuclear divisions still occur in the ems1mutant, indicating that the tapetum is not required for meiotic nuclear events. However, the ems1 meiocytes do not undergo cytokinesis,suggesting that tapetum might be needed for the completion of meiosis. The dyt1 mutant phenotypes provide further support for this idea. In addition to a morphologically abnormal tapetum, the dyt1 mutant meiocytes were found to have thinner callose cell walls than normal,suggesting that normal tapetum function is needed for the formation of the callose wall. Nevertheless, DYT1 expression was detected at a low level in the meiocytes and expression of some meiotic genes was reduced;therefore, it is possible that DYT1 might also function in meiocytes.

This and previous studies support a model for the genetic control of tapetum development and function (Fig. 8), although evidence for biochemical interactions is not yet available. SPL/NZZ is required for the formation of sporogenous cells and surrounding somatic cell layers, including the tapetum(Yang et al., 1999). Recently,Ito et al. showed that AG is a direct activator of SPL/NZZexpression (Ito et al., 2004). EMS1/EXS, TPD1 and SERK1/SERK2 are required for the formation of tapetum (Albrecht et al.,2005; Colcombet et al.,2005; Yang et al.,2003; Zhao et al.,2002). In addition, phenotypic changes caused by TPD1overexpression are dependent on EMS1/EXS(Yang et al., 2005). We show here that SPL/NZZ and EMS1/EXS positively regulate expression of DYT1. The AtMYB33 and AtMYB65 genes are expressed in the tapetum and their expression does not require the SPL/NZZ, EMS1/EXS or DYT1 gene.

In addition, it is likely that the genes that depend on DYT1 for normal expression support the tapetum function that produces the enzymatic activities and materials needed for pollen development(Dickinson and Bell, 1976; Hernould et al., 1998; Liu and Dickinson, 1989; Rubinelli et al., 1998; Scott et al., 2004; Taylor et al., 1998; Zheng et al., 2003). Among the genes regulated by DYT1 are AMS and MS1, which also encode transcription factors that probably regulate late tapetum genes. Analysis of DYT1 overexpression transgenic plants and expression studies in various mutants suggest strongly that other genes are needed for normal tapetum development. Phenotypic similarities suggest that AtMYB33 and AtMYB65, as well as potentially others (such as GNE1)(Sorensen et al., 2002), may act at approximately the same step as DYT1. Therefore, DYT1is a crucial component at a key step in the regulatory network responsible for tapetum development and function (Fig. 8).

We thank J.-P. Vielle Calzada for identifying the Ds line carrying the dyt1 mutation; W. Yang, D. Zhao and D. Ye for spl, ems1 and tpd1 mutants; A. Omeis and J. Wang for plant care;D. Grove and A. Price for DNA sequencing and real-time PCR; and G. Ning, R. Haldeman and M. Hazen for assistance in preparing anther sections and for use of the Olympus BX-51 microscope and camera. We thank A. Wijeratne for help in real-time PCR; and B. Feng, C. H. Hord, W. Hu, A. Surcel, G. Wang, A. Wijeratne and L. Zahn for helpful comments on the manuscript. This work was supported by a grant from the US Department of Energy (DE-FG02-02ER15332) to H.M. and used plant materials generated with support from a National Science Foundation grant (0215923). H.M. gratefully acknowledges the support of the John Simon Guggenheim Memorial Foundation.