To identify regulatory sequences that direct preferential or specific expression in Arabidopsis microspores we analyzed normalized transcriptomic datasets. We compared the expression profiles of all gametophytically expressed genes with those expressed in the sporophyte. The male gametophyte microarray data set was obtained from our previous experiments involving analysis of four developmental stages; microspore, bicellular, tricellular and mature pollen 33. Sporophytic datasets were obtained from publicly available resources (NASC). We selected for genes exhibiting strict expression patterns at early stages of male gametophyte development with no or low signals in mature pollen. Genes with expression in inflorescences, flower buds 3134 and developing flowers 32 were retained as candidate genes, but genes with reliable signals in other sporophytic dataset(s) were excluded. Genes were further selected to retain those encoded as single copy genes and that showed a range of expression levels in microspores. This approach led to the identification of seven potential target genes (At5g40040, At5g59040, At5g46795, At4g26440, At2g03170, At3g14450 and At1g53650; Fig. 1).

The direct comparison of gene expression levels obtained from independent microarray datasets for male gametophytic cells and for sporophytic tissues is difficult 135. Therefore, the putative specificity of gene expression from microarray analyses was verified by RT-PCR analysis using RNA samples isolated from four stages of male gametophyte development, unicellular, bicellular, tricellular and mature pollen, and four sporophytic tissues and organs (flowers, leaves, stems and roots). Four genes showed expression in one or more sporophytic tissues and were eliminated from further characterization. The remaining three genes showed microspore-specific expression by RT-PCR analysis (Fig. 2). During selection we did not exclude genes from further analysis that also showed weak expression in open flowers since these contain mature pollen grains.

The putative microspore-specific promoter sequences of these genes, At5g59040, At5g46795, At4g26440, were termed MSP1, MSP2 and MSP3 respectively. Selected genes encoded proteins with quite distinct cellular functions. At5g59040 encodes COPT3, a putative pollen-specific member of a copper transporter family 36. At4g26440 encodes AtWRKY34, a group I WRKY family transcription factor 37. On the contrary, At5g46795 encodes an expressed protein of unknown function.

To test the activity of the selected MSP regulatory sequences in vivo, approximately 1 kb upstream genomic fragments were amplified for each of the genes and cloned into the GATEWAY-compatible destination vector, pKGWFS7 38. Three MSP-GFP::GUS constructs, MSP1 (At5g59040), MSP2 (At5g46795) and MSP3 (At4g26440), were introduced into wild type Arabidopsis plants and ~40 T1 transformants were analyzed histochemically for GUS activity in inflorescences, bud clusters and flowers. We found that 38/40 MSP1, 36/38 MSP2 and 18/34 MSP3 plants showed GUS expression in flower buds and/or in open flowers. GUS staining in MSP1 plants was clearly detectable in anthers of younger buds than those from MSP2 and MSP3, but gradually became weaker towards anthesis, while GUS expression from MSP2 and MSP3 remained high at later stages including in mature pollen grains.

Segregation analysis revealed that the majority of lines segregated 3:1 for kanamycin-resistant and kanamycin-sensitive plants. Plants that were hemizygous for MSP constructs showed 1:1 segregation of GUS staining and non-staining spores consistent with gametophytic expression (data not shown). We examined gametophytic expression in detail by staining isolated spores at different developmental stages and by sectioning stained flower buds. For all three promoters, GUS expression was first detectable in uninucleate microspores (Fig. 3P–R). Another common feature was that anther transverse sections clearly showed GUS staining in the tapetum (Fig 3D–F). Differences in expression profiles were noted between MSP1 and the other two promoters. While MSP1 showed earlier staining in microspores then a strong decline in mature flowers, MSP2 and MSP3 initiate expression in microspores, but GUS expression peaks later and accumulates in mature pollen.

We also examined GUS expression in seedlings using ~20 lines of T2 generation plants for each construct and found no expression from all three MSP promoters, apart from 5–10% of anomalous lines that showed patchy expression in leaves and roots, or weak expression in stamen vascular tissues. None of the lines examined showed GUS activity in 5 day old seedlings. However interestingly, we consistently observed GUS activity at the distal tip of cotyledons and leaves in 10 day old MSP1-GUS seedlings (Fig. 3V). In summary all 3 MSP promoters were found to be specifically or highly preferentially expressed in microspores, developing pollen and the tapetum (Fig. 3)

To evaluate MSP promoters as tools for the manipulation of microspore gene expression in planta, we examined whether gene expression driven by MSP1 and MSP2 could complement a gametophytic mutant phenotype caused by defects that are known to depend upon gametophytic expression in developing microspores. Mutations in the Arabidopsis TWO-IN-ONE (TIO) protein kinase result in binucleate pollen grains due to the failure of cytokinesis in microspores at pollen mitosis I. Plants that are heterozygous for the T-DNA insertion allele, tio-3, show 50 % mutant pollen that results in a 2:2 segregation of wild type and mutant pollen in tetrads (Fig. 4A). Moreover cytokinesis defects in mutant tio-3 pollen completely block genetic transmission of tio-3 through pollen 3.

Test vectors were built in which MSP1 or MSP2 promoters drive the expression of full length TIO cDNA. pMSP1-TIO and pMSP2-TIO. Both vectors were transformed into tio-3 plants by floral dipping. Double selection for tio-3 (ppt) and pMSP-TIO (kanamycin) constructs led to the isolation of 19 nineteen transformants containing pMSP1-TIO and 10 containing pMSP2-TIO. Plants were screened for the frequency of wild type and mutant spores in mature pollen tetrads after DAPI staining. 18/19 lines from pMSP1-TIO and 4/10 from pMSP2-TIO showed an increase in the frequency of wild type pollen compared to heterozygous tio-3 plants (data not shown). This resulted in the frequent appearance of mature tetrads with three wild type spores and one mutant member, compared with heterozygous tio-3 plants that always showed a 2:2 segregation (Fig. 4).

We further tested genetic transmission of the tio-3 mutant allele in 15 complementing pMSP1-TIO lines and all of four complementing lines from pMSP2-TIO. In the F1 progenies generated from test crosses in which the pollen donor carried tio-3 and pMSP-TIO we observed approximately 30 to 50 % pptR progeny for pMSP1-TIO and 8 to 33 % for pMSP2-TIO. Table 1 shows the results for four representative lines for each construct. These results clearly demonstrate that gametophytic expression of the full-length TIO cDNA under the control of MSP1 and MSP2 promoters is sufficient to complement the tio pollen phenotype. Moreover, our results indicate that MSP1 is more effective than MSP2 in this complementation assay.

Arabidopsis plants were grown in controlled-environment cabinets at 21°C under illumination of 150 mmol m^-2 s^-1 with a 16-h photoperiod. For spore isolation, Arabidopsis ecotype Landsberg erecta (Ler) plants were used. Isolation of spores and pollen was described 33. Information on the purity of isolated fractions determined by light microscopy and 4,6-diamino-phenylindole staining and vital staining of isolated spore populations assessed by fluorescein-3,6-diacetate treatment were also described in 33. Roots were grown from plants in liquid cultures as previously described 33. Wild-type and transgenic seeds were sterilized according to published procedures 4344. Plants were transformed by floral dipping 45.

RNA isolation and hybridization of Affymetrix ATH1 genome arrays was described in 33. Twelve mixed and seventy-five sporophytic datasets used for comparison with the pollen transcriptome were obtained from the NASCArray database 46 through AffyWatch service 31. Sporophytic datasets represented seven vegetative tissues (seedlings, leaves, petioles, stems, roots, root hair zone and suspension cell cultures; 33. Mixed datasets originated from three experiments covering whole inflorescences and flower buds 3134 and whole flower development 32.

All gametophytic and sporophytic datasets were normalized using DNA-Chip Analyzer 1.3 (dChip) 47 as described previously 33. All raw and dChip-normalized transcriptomic datasets can be accessed and downloaded through the Arabidopsis Gene Family Profiler (aGFP) database 48.

Total RNA from 50 mg of leaves, stems, roots, inflorescences and isolated spores at each developmental stage was extracted using the RNeasy plant kit (Qiagen, Valencia, CA) according to the manufacturer's instructions. The yield and RNA purity were determined spectrophotometrically. Pollen, stem, leaf, and inflorescence samples were isolated from plants grown as described. Pollen RNA used for RT-PCR analyses was obtained from plants that were grown independently from those used to isolate RNA for microarray analysis. Samples of 1 μg total RNA were reverse transcribed in a 20-μL reaction using the ImProm-II Reverse Transcription System (Promega, Madison, WI) following the manufacturer's instructions with the exception that the oligo(dT)₁₅ primer was replaced with a custom-synthesized 3'-RACE primer (Tab. 2). The use of intron-spanning primer sets or nested reverse primers in cases where primers spanned no intron ensured that only cDNA was amplified. For PCR amplification, 1 μL of 50× diluted RT mix was used. The PCR reaction was carried out in 25 μL with 0.5 unit of Taq DNA polymerase (MBI Fermentas, Vilnius, Latvia), 1.2 mM MgCl₂, and 20 pmol of each primer. The PCR program was as follows: 2 min at 95°C, 33 cycles of 15 s at 94°C, 15 s at the optimal annealing temperature (63°C to 67°C), and 30 s at 72°C, followed by 10 min at 72°C. As a reverse primer, NESTED primer (Tab. 2) overlapping the 3'-RACE primer was used to eliminate genomic DNA amplification. The gene-specific forward primers were designed using Primer3 software 49 (Tab. 2).:

To examine the precise gene expression, each tested gene promoter region was fused with GUS to generate the MSP::GUS reporter using GATEWAY cloning system according to manufacturer's instructions (Invitrogen, Carlsbad, CA). Promoter fragments were amplified by two-step PCR from Col-0 genomic DNA isolated from 3-week-old seedlings using the KOD HiFi DNA Polymerase (Novagen, Darmstadt, Germany). For the first step, specific primers with appended adapters complementary to AttB1 and AttB2 sequences were used to generate 1000-bp promoter regions of tested genes (Tab. 2). Purified PCR products were cloned via pDONR201 donor vector (Invitrogen, Carlsbad, CA) into pKGWFS7 destination vector (VIB, Gent, Belgium; 38 to generate recombinant destination clones pMSP1, pMSP2 and pMSP3. All the recombinant plasmids were transformed into Agrobacterium tumefaciens GV3101 50. These strains were used to transform Arabidopsis thaliana ecotype Columbia using the floral dip method 45. Transgenic progenies were selected either on one-half strength Murashige and Skoog standard medium, supplemented with 50 mg/l kanamycin on 1-week-old seedlings.

Histochemical assays for GUS activity in T2 generation of Arabidopsis transgenic plants were performed according to the protocol described previously 4451. Seedlings and inflorescences incubated for 48 h at 37°C in GUS buffer (100 mM Na-phosphate, pH 7.2; 10 mM EDTA, pH 8.0; 0.1% Triton X-100; 2 mM K₃Fe [CN]₆) supplemented with 1 mM 5-bromo-4-chloro-3-indolyl b-D-glucuronide (X-gluc). GUS stained floral buds were fixed in 70% FAA (Formaldehyde/Acetic acid/Ethanol/water, 5/5/63/27, v/v/v/v) for at least 24 h. After washing with 70% ethanol, samples were dehydrated gradually in the ethanol-butanol series and infiltrated with paraffin. 10 μm thin cross sections were prepared with a microtome (Finesse ME, Shandon). GUS staining patterns were recorded using a Nikon Eclipse E600 microscope (Nikon Instruments, Melville, NY). Images were processed using Adobe Photoshop software (version CS2; Adobe Systems, San Jose, CA).

The full-length TIO cDNA clone, pKS-TIONC19, was modified to insert the ~1 kb MSP1 or MSP2 promoter fragments upstream of the TIO coding sequence using AscI and NotI. MSP promoter-TIO cDNA fusion fragments were subcloned into the binary vector pER10 52 using AscI and PacI to produce pMSP1-TIO and pMSP2-TIO. Before transformation into Agrobacterium tumefaciens strain GV3101, constructs were verified by restriction enzyme digestion and sequencing. Individual phosphinothricin (ppt) resistant heterozygous tio-3 mutants were transformed by floral dipping 45. Transformants containing both tio-3 and MSP promoter-TIO T-DNA constructs were selected on 15 cm MS agar plates supplemented with 10 mg/l ppt, 50 mg/l kanamycin, and 200 mg/l cefotaxime. Complementation was analyzed by scoring tio-3 pollen phenotype after 4'-6-Diamidino-2-phenylindole (DAPI) staining and also by scoring the genetic transmission of tio-3 allele through male after test crosses to Col-0 or ms1-1 as described by 253.