Fernandez Gomez, José Fernández and Wilson, Zoe A. (2014) A barley PHD finger transcription factor that confers male sterility by affecting tapetal development. Plant Biotechnology Journal, 12 (6). pp. 765-777.

Summary Controlling pollen development is of major commercial importance in generating hybrid crops and selective breeding, but characterized genes for male sterility in crops are rare, with no current examples in barley. However, translation of knowledge from model species is now providing opportunities to understand and manipulate such processes in economically important crops. We have used information from regulatory networks in Arabidopsis to identify and functionally characterize a barley PHD transcription factor MALE STERTILITY1 ( MS1 ), which expresses in the anther tapetum and plays a critical role during pollen development. Comparative analysis of Arabidopsis , rice and Brachypodium genomes was used to identify conserved regions in MS1 for primer design to amplify the barley MS1 gene; RACE-PCR was subsequently used to generate the full-length sequence. This gene shows anther-speciﬁc tapetal expression, between late tetrad stage and early microspore release. HvMS1 silencing and overexpression in barley resulted in male sterility. Additionally, HvMS1 cDNA, controlled by the native Arabidopsis MS1 promoter, successfully complemented the homozygous ms1 Arabidopsis mutant. These results conﬁrm the conservation of MS1 function in higher plants and in particular in temperate cereals. This has provided the ﬁrst example of a characterized male sterility gene in barley, which presents a valuable tool for the future control of male fertility in barley for hybrid development.


Introduction
Demand for food is increasing because of population growth, urbanisation and increasing affluence in the developing world. The world's population is projected to increase from 6 to 9 billion by 2050, and food needs are expected to increase by 50% by 2030 and 100% by 2050 (FAO, 2008). Providing efficient mechanisms to increase crop productivity is key to addressing these challenges, and the control of plant fertility is critical to this. Barley is of major economic importance with a total UK crop production of >5.5 million tonnes, which is used for animal and human consumption as well as in the malting industry. Yield may be reduced by stress, biotic or abiotic, as well as by inbreeding depression due to its autogamous nature.
Control of male reproduction helps facilitate plant breeding and in particular the generation of high-yielding hybrid varieties, for example, hybrid rice yields 15-20% more than most inbred varieties (Cheng et al., 2007;Zhong et al., 2004). There are currently limited examples of hybrid lines in barley, although Syngenta has recently released HYVIDO, a barley hybrid that guaranties 0.5 t/ha more than conventional varieties (Syngenta, 2013). One reason for this has been the lack of fundamental knowledge and genetic resources for pollen development in barley. Male sterile lines are valuable resources that greatly facilitate the production of hybrids via cross-pollination. However, there are very few examples of characterized genes involved in male fertility in cereals and a dearth of information on the molecular mechanism of pollen development in these crops.
The study of the genes involved in pollen development has principally focused on the Arabidopsis model (Ma, 2005) and more recently on rice ). The transfer of knowledge from these regulatory networks to temperate cereals, for example barley and wheat, has been slow. However, recently, the characterization of important genes in cereal species has been greatly facilitated by the sequencing and annotation of a number of grass genomes. The complete genome of five grass species is available (Oryza sativa, Sorghum bicolor, Brachypodium distachyon and Zea mays B73) (International Brachypodium Initiative, 2010;International Rice Genome Sequencing Project, 2005;Paterson et al., 2009;Schnable et al., 2009) and more recently the barley genome (Mayer et al., 2012). Although these grass genomes vary greatly in size, partly due to expansion of retroelement repeats, there is an underlying conserved gene order or synteny (Moore et al., 1995;Wicker and Keller, 2007). This synteny can help in the translation of gene information from models to economically important species.
Much is now known about the regulatory networks in Arabidopsis associated with pollen wall development and the role that the anther tapetum plays in this process . Pollen formation in rice appears to follow a similar developmental pathway to Arabidopsis (Chen et al., 2005;Itoh et al., 2005;Wan et al., 2011;Wilson and Zhang, 2009) with the development of a secretory tapetum. A number of rice male sterile mutants have now been identified and are revealing high conservation to Arabidopsis pollen regulatory gene networks. Key Arabidopsis transcription factors have been identified in the regulation of tapetal development, including DYSFUNCTIONAL TAPETUM 1 (DYT1) (Zhang et al., 2006), MALE STERILITY1 (MS1) (Wilson et al., 2001) and ABORTED MICROSPORE1 (AMS1) (Sorensen et al., 2003), and many of these have also been shown to be conserved in rice ). For instance, OsUTD1 (UNDEVELOPED TAPETUM) encodes a bHLH protein, which acts after tapetum initiation in an analogous manner to AtDYT1 (Jung et al., 2005). Furthermore, OsTDR (Li et al., 2006;Zhang et al., 2008) has been shown to play an important role during rice tapetal development and lipid transport and metabolism for pollen wall formation. Phylogenetic analysis suggests that OsTDR is an orthologue of the Arabidopsis AMS gene (Sorensen et al., 2002). This conservation in function and regulation provides opportunities to identify similar genes in lesswell-characterized species.
Despite the economic importance of barley, little has been established relating to the molecular regulation of its pollen development, and to date, no barley male sterile mutants have been characterized. This is partly due to the difficulty of nondestructive staging of floral material, which has now been overcome by the development of vegetative markers linked to key anther stages (Gomez and Wilson, 2012). However, additional problems of slow plant growth, difficulties in transformation and the lack, until recently (Mayer et al., 2012), of full-genome information have hampered progress. In this paper, we present the first characterized male sterility gene in barley (Hv MALE STERILITY1 (HvMS1)), which has been identified using genomic information from Arabidopsis, rice and the bridging genome of Brachypodium distachyon.
Arabidopsis MALE STERILITY1 (MS1) is a plant homeodomain (PHD) finger motif transcription factor, which is critical for pollen development (Wilson et al., 2001). The rice PERSISTENT TAPETAL CELL1 (OsPTC1) (OsPTC1) gene has recently been identified as the putative orthologue of AtMS1 , which appears to have a similar function. To date, no related gene(s) have been identified in barley. AtMS1 shows very specific tapetal expression during microspore development and is not represented in the EST libraries. We therefore capitalized upon the close genomic relationship between Brachypodium and barley (Opanowicz et al., 2008) to serve as a valuable link for the identification of conserved genes in temperate cereals. We used the Brachypodium genome alongside the rice PTC1 gene (AtMS1 orthologue) sequences, to identify conserved regions to amplify equivalent sequences in barley. This approach, combined with RACE-PCR and subsequent functional testing using RNAi lines in barley and complementation analysis in Arabidopsis, has led to characterization of the first male sterility gene in barley, HvMALE STERILITY1 (HvMS1). This has demonstrated the conservation of gene function in pollen development between Arabidopsis and barley and has provided a valuable tool for the future manipulation of male fertility in barley.

Results
Amplification of the barley putative MS1 orthologous gene BLAST analysis using AtMS1 and its rice orthologue, OsPTC1, did not identify any related genes in the barley database. AtMS1 shows very specific, low-level expression only in the tapetum, from the tetrad breakdown stage to pollen mitosis I, and has not been identified in the Arabidopsis EST libraries. It was therefore expected that it would not be represented in barley cDNA libraries. Comparative analysis between the rice and Arabidopsis (PTC1 and MS1, respectively) sequences showed moderate similarity (53% , Table 1); however, due to limitations of barley genome sequence availability and the relatively low levels of homology between rice, Arabidopsis and barley, a bridging genome of the temperate grass Brachypodium was used to identify regions of high conservation for primer design ( Figure  S1).
BLAST analysis using the AtMS1 and OsPTC1 sequences against the Brachypodium genome identified a putative orthologue, Bradi4g31760, with high sequence similarity to AtMS1 and OsPTC1 (Table 1). Alignments between these sequences were used to identify conserved regions for primer design ( Figure S2; Table S1), which were used to amplify barley cDNA fragments (Figure 1a-d); sequencing of these showed high similarity to the putative Brachypodium orthologue and at a lower level to the AtMS1 and OsPTC1 sequences (Table S2). RACE-PCR, using primers based on these partial barley sequences, was used to amplify the full-length sequence (Figure 1e-g). The full-length genomic and cDNA sequences had very high similarities to the rice and Brachypodium orthologous nucleotide and protein sequences (Nucleotide: 87 and 91%; Protein: 83 and 83%, respectively, Table 1; Table S2). The barley gene comprised of three exons and two introns; the intron/exon borders in all four species were conserved ( Figure 1h). The barley sequence comprised a CDS of 2,016 bps, encoding a predicted protein of 672 amino acids, which included a PHD finger domain at the C-terminal region ( Figure 2). This conserved domain is also observed in AtMS1 and OsPTC1; however, the zinc finger domain, which is seen in the Arabidopsis MS1 protein, was not present in the barley, rice or Brachypodium MS1 sequences ( Figure 2). No analysis of the HvMS1 promoter region was possible due to a lack of sequence availability.
The putative HvMS1 gene shows expression in the anther tapetum during late meiosis and early microspore development Wild-type barley plants were analysed by RT-PCR and qRT-PCR to determine the expression of the putative HvMS1 gene, which was restricted to reproductive tissues ( Figure 3) and was seen only in floret samples containing anthers from the tetrad to free microspore stages (Figure 3a samples E and F; staging is shown in Figure S3). No expression was observed before or after these stages, reflecting the strict temporal regulation of expression of the gene. Ears of 3-4 cm, equating to the late tetrad stage, Table 1 Sequence similarity between MS1 genes (nucleotide and translated protein product) of Arabidopsis (AtMS1), rice (OsPTC1), Brachypodium (Bradi4g31760) and barley (HvMS1) To define the cell-specific expression pattern of HvMS1, RNA in situ hybridization was performed on transverse sections of wildtype inflorescences containing various stages of developing florets. The HvMS1 transcript was observed at a low level in the tapetal layer of the anther and only during a short period of time during late tetraspore/microspore release, (spike 3-4 cm) (Figure 3c), corresponding to that observed by RT-PCR analysis. No signal was observed in the sense control samples or in any other anther stages (Figure 3).
Altered expression of the putative HvMS1 gene by silencing, or overexpression, causes reductions in fertility due to impaired pollen and anther development HvMS1 function was analysed by RNA-mediated gene silencing in wild-type barley, variety Golden Promise. RNAi constructs were prepared using a 500-bp fragment of the HvMS1 cDNA sequence ( Figure S2), which lacked conserved motifs and did not show significant homology to other barley or Brachypodium sequences. Constructs were prepared by Gateway cloning into pBract207, driven by the maize Ubi1 promoter (Rooke et al., 2000) and used to transform immature barley embryos via Agrobacterium-med-iated transformation. These RNAi plantlets were transferred to soil and grown until anthesis. Regenerated T0 plantlets were confirmed as transgenic by PCR analysis using vector-specific and HvMS1-specific primers ( Figure S4; Table S1). Phenotypic analysis showed that several of the confirmed transgenic lines had poorly developed anthers that contained pollen with reduced viability (five lines; Figures 4 and S6). However, despite this reduction in viability, there was no overall decrease in seed set, presumably as sufficient pollen was still available for full fertilization.
Three to six plants generated from each T0 line (five independent lines) were grown until anthesis; these were further confirmed as transgenic by PCR analysis, and for three lines (Lines 14,19 & 20), the transgene copy number was determined (IDna Genetics Ltd., JIC Norwich; Figure S5a); qRT-PCR analysis indicated that levels of HvMS1 expression were reduced but varied slightly between tillers ( Figure S5b). This was not unexpected as different levels of fertility, and presumably RNAi silencing, were observed between tillers and also individual plants. All of these T1 RNAi lines showed complete sterility in some of their spikes ( (a) Transverse sections of anthers from sterile spikes of the T1 lines showed premature tapetal degeneration, just after tetrad release (Figure 4g,h). In contrast, the wild-type tapetum remained intact ( Figure 4i) and started degenerating significantly later during microspore mitosis stages. Florets collected from the RNAi lines at late heading stage (immediately prior to anthesis) were either empty or contained inviable pollen (Figure 4j-k), in contrast to the wild type that contained abundant viable pollen ( Figure 4l).
The effect of ectopic overexpression of the putative HvMS1 cDNA was also investigated by expressing the full-length HvMS1 cDNA regulated by the Ubi1 promoter. This was introduced into barley by Agrobacterium-mediated transformation; ten independent transgenic lines were generated and confirmed as expressing HvMS1 in leaf tissue ( Figure 5a); no expression was observed in leaves from nontransgenic control plants ( Figure 5a); high levels of HvMS1 expression was also confirmed by qRT-PCR using floral tissue from these transgenic lines ( Figure 5c). These overexpression lines did not display any apparent vegetative changes; however, at anthesis, all 10 lines were male sterile (Figure 5j-l). Sterility did not affect female organs, with full seed set obtained when crossed with wild-type pollen. The pollen from these overexpression lines appeared fully viable (Figure 5m,n); however, a failure of anther dehiscence and pollen release was observed in these lines ( Figure 5l). The anther surface showed deposits on the outer anther wall surface of the transgenic lines, which were reduced or absent in the wild type ( Figure 5i,o). Abnormal material deposition was observed on the surface of the pollen grains from the overexpression lines ( Figure 5u; black arrows), although the nature of this material is unknown. The Arabidopsis MS1 gene has been previously linked to biosynthesis and secretion of the pollen wall; it is likely that the deposits may be associated with alterations in the production and release of pollen wall materials from the tapetum. The overexpression lines also appeared to have a thicker anther endothecium (Figure 5t (black arrow)) than wild type (Figure 5q (black arrow)), and the anthers remained intact rather than opening (Figure 5s). In addition, traces of the tapetum still remained, indicating that complete tapetum degeneration did not occur (Figure 5t, white arrow). Moreover, pollen inside these anthers was sticky and failed to spread when dissected.

HvMS1 plays a conserved role in plant anther development
To establish the evolutionary relatedness of the putative HvMS1 and the AtMS1 gene, functional complementation of the Arabidopsis ms1 mutant was conducted. A binary plasmid (pGkGWY) carrying the Arabidopsis MS1 promoter and the HvMS1 coding region fused with the YFP (yellow fluorescent protein) was introduced into ms1MS1ttgTTG heterozygous Arabidopsis plants. The resultant segregating generation was analysed for complementation of the ms1ms1 mutation. Transgenic lines were initially screened to identify ttgttg homozygous plants; ttg is closely linked to the ms1 mutation enabling early selection at the seedling stage for lack of trichomes (ttg) and potential homozygosity at the ms1 locus. RT-PCR was subsequently used to genotype the segregating transgenic lines to confirm ms1 homozygosity and HvMS1 transgene expression. Two of ten lines were identified as ms1 homozygous mutants, as indicated by RT-PCR expression of a larger fragment due to loss of the first splice donor site in the ms1 mutant (Figure 6a, samples 1 and 6). All lines showed expression of the HvMS1 transgene ( Figure 6b) except for the Arabidopsis wild-type (Figure 6b: Wt1). No genomic DNA contamination was observed in the RT-PCR minus where RNA from samples 1 and 6 was used as template ( Figure 6c). In the ms1 mutant, no viable pollen was observed, and the siliques failed to set seeds (Figure 6g-h). However, ms1 homozygous lines expressing HvMS1 showed some viable pollen ( Figure 6e) and partial rescue of fertility with some siliques containing viable seed (Figure 6d and f). This confirmed that HvMS1 was able to functionally complement the Arabidopsis ms1 mutation and rescue fertility.

Discussion
Significant progress has been made towards understanding pollen development, and much of this has come from the study of male sterile mutants in the model system Arabidopsis.  (Mayer et al., 2012), is now making it possible to carry out these activities in crops. Nevertheless, extension of this knowledge to the temperate cereals, such as barley and wheat, has encountered difficulties, partly due to the lack of genome sequence and annotation of this sequence, but also to the scarcity of genetic resources for functional gene analysis. This is particularly apparent when trying to identify regulatory genes that may be expressed transiently, at a low level, and in a highly cell-specific manner, which are unlikely to be represented in cDNA libraries. We have shown this to be the case in the identification of the barley MS1 gene which like its  HvMS1OEx lines (s,t) at anthesis. At this stage, in the wild-type anther, the septum has already broken down, and the pollen is about to be released (p), whereas in the overexpression lines, the septum and therefore the locules remained intact (s), no anther opening and pollen release was observed.
In addition, at this stage, overexpression anthers showed a thicker anther endothecium (t: black arrow) and what appeared to be remains of the tapetum (t: white arrow). (r and u) SEM images of barley wild type (r) and T0 HvMS1OEx (u) pollen grains. HvMS1OEx samples showed abnormal deposits on pollen grains (u: black arrow) that might be related to the sticky appearance of the pollen in the HvMS1OEx anthers. Bars: d = 5 mm; e = 1 mm; f = 0.6 mm; g and m = 10 lm; h: 30 lm; n = 60 lm; i and o = 40 lm; j = 5 mm; k = 1 mm; l = 0.6 mm, p: 50 lm; q and t: 10 lm; r and u: 10 lm; s: 100 lm.  Moore et al., 1993), and this has been instrumental in enabling the characterization of the wheat gene Ph1 (Griffiths et al., 2006) and the barley gene Ppd-H1 (Turner et al., 2005). BLAST analysis using AtMS1 and OsPTC1 against the Brachypodium genome (Bd21 8x release, www.Brachypodium.org) identified Bradi4g31760, which has 54% and 85% homology to the Arabidopsis (MS1) and rice (PTC1) sequences, respectively. The high levels of homology observed imply that this may be the putative Brachypodium orthologue, and given the close relationship between Brachypodium and barley (Opanowicz et al., 2008) suggested that MS1 function would be conserved in barley and other monocots.

Identification of the barley MALE STERILITY1 gene
Amplification using primers to conserved regions of the three MS1 orthologues, combined with 5 0 and 3 0 RACE-PCR, enabled identification of the barley putative HvMS1 gene. This sequence had very high similarity to the rice and Brachypodium nucleotide and protein sequences (Table 1) with a PHD finger domain (Cys4-His-Cys3 motif) at the C-terminal region (50-80 amino acids). The PHD finger domain is conserved in animals, yeast and higher plants (Halbach et al., 2000) and   (Table S1), the homozygous ms1 splice mutant generates a 506-bp band from the mutant ms1 gene, whereas the wild-type Ler MS1 fragment is smaller at 351 bp. The 506-bp mutant transcript was seen in lines 1 and 6, indicating that these are homozygous ms1 mutants, the other lines all showed the 351-bp WT MS1 band. WT 1 : Arabidopsis Ler inflorescence cDNA; WT 2 : Barley inflorescence cDNA; (b) HvMS1 RT-PCR expression (HvMS1-1F and HvMS1-3R primers, Table S1) was seen in cDNA from all of the putative transgenic inflorescences and the barley wild-type inflorescence material (WT 2 ); no expression was seen in Arabidopsis wild-type inflorescences (WT 1 , 2006). In plants, their biochemical function is unknown; however, they have been associated with chromatin modification (De Lucia et al., 2008). Phylogenetic analysis of the MS1 orthologues indicated that Arabidopsis MS1, poplar PtMS1 and rice PTC1 form a separate group within the entire family (Ito et al., 2007), suggesting a conserved role in plant reproductive development. The HvMS1 protein, and its orthologues in rice and Brachypodium, does not contain a leucine zipper motif (LZ), which is present within the AtMS1 protein, suggesting that this may have been associated with the monocot/dicot evolutionary divergence. The HvMS1 gene showed highly localized temporal and spatial expression in the anther tapetum from late tetrad to early microspore release (Figure 3). This localized expression was consistent with the tissue-specific expression of AtMS1 and OsPTC1 in Arabidopsis and Rice, respectively Yang et al., 2007), supporting the hypothesis that HvMS1 is the orthologue of AtMS1 in barley.

The HvMS1 gene plays a critical role in pollen development
The role of the putative HvMS1 gene was confirmed by reduced pollen viability in the T0 HvMS1RNAi transgenic lines ( Figure S6b, c), but the presence of some viable pollen was sufficient to maintain general fertility ( Figure S6b,c). However, an increase in sterility was seen in the subsequent T1 generation with complete sterility seen in some tillers (Figure 4). This effect was specific to pollen development, with no reductions in female fertility or changes in vegetative growth. The amount of sterile tillers varied between the different T1 transgenic lines but was consistent with lines showing reduced fertility in the T0 generation and correlated with presence of the RNAi construct and reduced expression of HvMS1 as indicated by qRT-PCR ( Figure S5). This complete sterility may be due to the increased transgene copy number in the T1 generation, increasing the silencing capacity beyond the required threshold level (Lindbo et al., 1993). This is a common occurrence, with RNAi lines frequently showing reductions in expression 'knocked down', rather than complete silencing 'knockouts' of the target gene (Yin et al., 2005). For example, (Moritoh et al., 2005) reported that in rice RNAi silencing of the OsGEN-like gene, a member of the RAD2/XPG nuclease family involved in microspore development, caused male sterility, but plants showed variable fertility depending on the expression level of the native gene. In the current work, RNAi silencing of the HvMS1 gene had a variable effect on different tillers within the same plant. The reason for this remains unknown, but may reflect variation in expression levels and stability between independently transformed plants and tillers (Jones et al., 1985;Peach and Velten, 1991;Walters et al., 1992). Alternative options for overcoming this variable silencing and obtaining full sterility could be achieved either by HvMS1 overexpression, the use of targeted mutagenesis, (TILLING), or via the use of transgene repressor constructs. Barley HvMS1 RNAi silencing lines (T1) showed premature tapetum degeneration at the free microspore stage (Figure 4gh), and by heading, anthers were empty or contained unviable pollen (Figure 4j,k). This was also observed in the Osptc1 mutant, which showed increased tapetal cell vacuolation at microspore release, and premature cytosolic secretion into the anther locule  and in Atms1 (Yang et al., 2007). This suggests that MS1 has a conserved function in both monocots and dicots, and acts in the regulation of the breakdown of the tapetum and in the secretion of materials into the locule for pollen wall formation.

Ectopic overexpression of HvMS1 results in abnormal anther development and functional male sterility in barley
Barley HvMS1 overexpression lines showed some similar phenotypes to the Arabidopsis MS1 overexpression plants, which showed occasional male sterility and abnormal growth (Yang et al., 2007), although unlike the Arabidopsis lines, no stunted plants or vegetative abnormalities were seen in the barley lines. Closer analysis of these barley lines revealed mature, apparently viable pollen but a failure of anther dehiscence. The reason for the lack of dehiscence and resultant sterility is unclear, but may be due to altered anther endothecium development and sticky pollen. Arabidopsis MS1 OEx lines had previously been reported to have increased indentation on the anther epidermal surface and to contain sticky pollen that suggested abnormalities in the pollen wall (Yang et al., 2007). A similar sticky appearance due to excess pollen wall materials (Figure 5u), which appeared to limit pollen release, was seen in the OExHvMS1 lines and may be equivalent to the enhanced lipid biosynthesis observed in Arabidopsis MS1 overexpression lines (Yang et al., 2007). The barley lines also showed abnormal endothecium expansion and development, with a lack of associated secondary thickening and an increased deposition of cellular materials ( Figure 5). The increased expression of HvMS1 appeared to be causing alteration in cell wall biosynthesis in these anther cell layers (Figure 5q,t). We have previously shown that the structure and thickening in the endothecium are critical for anther opening (Nelson et al., 2012;Yang et al., 2007); therefore, these abnormalities are likely to impair anther dehiscence. This phenotype has not been previously reported in Arabidopsis or rice; however, the severity of this effect may reflect increased HvMS1 expression using the ubiquitin promoter, as compared to those induced in Arabidopsis and rice using the CaMV35S promoter. Previous evidence with translational fusions in Arabidopsis supports very tight regulation of MS1 expression. It maybe that in these barley MS1 overexpression lines, the level of ectopic MS1 transcript is sufficiently high in the endothecium and other anther cell layers that normal development fails to occur.
HvMS1 is a functional orthologue of AtMS1 and can rescue fertility in Arabidopsis ms1 mutants Precise control of MS1/PTC1 expression appears critical for pollen development, as the MS1 rice orthologue, OsPTC1, was able to restore fertility in ms1ms1 homozygous lines when regulated using the native Arabidopsis promoter, but not under control of the CaMV35S promoter . Successful complementation of the Arabidopsis ms1 mutant was observed using the native Arabidopsis MS1 promoter to regulate the HvMS1 gene (Figure 6), confirming that MS1 function is maintained in barley. This work has demonstrated the conservation of the regulatory network of gene expression during pollen development from dicots to monocots, and more specifically within the temperate cereals. It has also demonstrated the value of using bridging genomic information to aid in gene identification. Such approaches are vital in species where full-genome data are not available, either due to the complexities of their genomes or because the species fall into the 'orphan crop' status. This regulatory network conservation will contribute to the identification of barley genes involved in anther and pollen development, and this in turn will provide targets that have application for switchable control of fertility as part of hybrid breeding programmes in cereals. We have therefore shown that the HvMS1 gene is critical to pollen development and that when HvMS1 expression is reduced a male sterile phenotype results. To our knowledge, this is the first example of a functionally characterized gene in barley that has been specifically linked to pollen development. Controlling crop fertility for hybrid development is a key breeding goal, as increased yield is frequently associated with hybrids. Understanding the molecular process of pollen development is critical to achieving this, and the characterization of such regulatory networks provides the first step in this process.

Plant material and growth conditions
Two double-rowed spring barley varieties, Optic, (kindly provided by The James Hutton Institute), Golden Promise used for transformation studies and regenerated plants, were grown under controlled conditions (15°C/12°C; 16 h photoperiod; 80% RH, 500 lmol/m 2 /s metal halide lamps (HQI) supplemented with tungsten bulbs). Seeds were sown in 12-well pots using John Innes no. 3 compost and after 2-3 weeks transferred to 5 L pots containing Levington C2 compost (four plants each). Transgenic barley plants were generated using Agrobacterium-mediated transformation of immature embryos (Golden Promise) (Harwood et al., 2008).
Alignment between AtMS1, OsPTC1 and the Brachypodium putative orthologous gene, Bradi4g31760, was performed using the Needleman-Wunsch global sequence alignment tool (Needleman and Wunsch, 1970;NCBI). This global alignment was only used on sequences that were expected to share significant similarity over most of their length. These alignments were used to identify conserved regions for primer design (Figure 2, Table S1).

Isolation of HvMS1 full-length cDNA and cloning
Total RNA was isolated using the Qiagen RNeasy Mini Kit (Qiagen, Manchester, UK) following the manufacturer instructions from wild-type barley florets collected from 3.5 to 4.5 cm spikes (Gomez and Wilson, 2012). 1.5 lg of total RNA was used to synthesize cDNA using the SuperscriptIII cDNA synthesis kit (Invitrogen, Life Technologies, Paisley, UK). Primers designed using the most conserved regions between OsPTC1 and Bradi4g31760 (Table S1) were used to amplify the barley putative orthologue. RACE-PCR was conducted to obtain the full length, 5 0 and 3 0 ends, of the wild-type HvMS1 mRNA.

In situ hybridization of HvMS1
A 500-bp HvMS1 cDNA PCR product (amplified using primers HvMS1-3 F and HvMS1-3R; Table S1) was inserted into pCR4Blunt TOPO to act as a template for the hybridization probe. The fragment was amplified using M13F/R primers and purified. DIG-UTP-labelled sense and antisense probes were generated by run-off transcription with T3 and T7 RNA polymerases, respectively (Roche Applied Science).
Optic florets from various stages of development were fixed overnight with 4% (v/v) paraformaldehyde in 1xPBS at 4°C and dehydrated in an ethanol series (30-100% (v/v)), embedded in paraffin (Paramat extra pastille, BDH Laboratory) and cut into 0.8lm sections using a microtome HM315. Sections were hybridized with probes as previously described (Wilson, 2000) and signal detected using a DIG Nucleic Acid Detection Kit (Roche).

Microscopy
Alexander stain (Alexander, 1969) was used to check pollen viability. The stain contains malachite green, which stains the cellulose in pollen walls and acid fuchsin, which stains the pollen protoplasm; aborted pollen grains lack protoplasm and thus fail to stain with acid fuchsin. Anthers from flowers containing mature, nondehisced anthers, just before anthesis were placed into a droplet of stain and shaken gently to release the pollen grains. Viable grains stained dark blue or purple with a defined intact shape; dead grains stained pale turquoise-blue, often with a broken wall structure. Zhang, W., Sun, Y.L., Timofejeva, L., Chen, C., Grossniklaus, U. and Ma, H. (2006)

Supporting information
Additional Supporting information may be found in the online version of this article: Figure S1 Phylogenetic tree of putative MS1 orthologous cDNA sequences (the phylogenetic tree was generated using Geneious software).

Figure S2
AtMS1/HvMS1/OsPTC1/Bradi4g31760 transcript alignment. Boxes show the conserved regions where primers were designed to amplify the barley orthologue HvMS1. Figure S3 Toluidine blue stained transverse sections through wild-type barley anthers showing representative stages of pollen development equivalent to the RT-PCR samples (Figure 3). Figure S4 Genotyping of T0 Barley RNAi-HvMS1 lines. Figure S5 Analysis of RNAi lines. Figure S6 Phenotypic analysis of anther and pollen development in HvMS1RNAi T0 lines. Figure S7 Comparison between T1 RNAi ears (Line 14). This line showed complete sterility in some tillers (s), whilst others were completely fertile (f). Bar: 0.5 cm.