MS26/CYP704B is required for anther and pollen wall development in bread wheat (Triticum aestivum L.) and combining mutations in all three homeologs causes male sterility

Development of anthers and pollen represents an important aspect of the life cycle in flowering plants. Genes contributing to anther and pollen development have been widely studied in many plant species. Ms26/CYP704B genes play an important role in pollen development through biosynthesis of sporopollenin for pollen exine formation. To investigate the role of Ms26/CYP704B genes in anther and pollen development of bread wheat, mutations in the A-, B-, and D-homeologs of the putative Ms26/CYP704B gene were analyzed. Single and double homozygous mutants in any of the homeologs did not affect pollen development and male fertility. Triple homozygous mutants resulted in completely male sterile plants that were defective in pollen and anther development. Additionally, double homozygous-single heterozygous mutants were also male sterile although with varying levels of residual fertility. The fertility of these triple mutants was dependent upon the homeolog contributing the wild-type allele. Two heterologous Ms26/CYP704B genes, when transformed into a triple homozygous mutant background, completely restored male fertility, whereas a single gene was unable to restore fertility. Functional analysis of Ms26/CYP704B furthers the understanding of male fertility genes which can be utilized for the development of novel hybrid seed production systems in wheat.


Introduction
Flowering plants have developed specialized structures for the production of male and female gametes. Successful production of male gametes relies on proper formation of male reproductive organs. Pollen grains (microgametophytes) are formed by the anther, the male reproductive organ, and deliver male gametes to organs bearing female gametes. are often adverse environmental conditions. The intine is composed of cellulosic material whereas the major component of exine is sporopollenin (reviewed in Quilichini et al. [1]). Sporopollenin fortifies the exine as the foundation of a skeletal structure as well as through formation of a durable covering. The components of the exine are synthesized by the surrounding tapetum and deposited on the surface of developing microspores within the anther locule [2], whereas the components of the intine are believed to be generated by the microspore vegetative cell [3]. Pollen exine in rice consists of two layers, the tectum (sexine and baculae) and the nexine (foot-layer). Exine development starts with the formation of primexine at tetrad stage (reviewed in Li and Zhang [4]). Primexine, a microfibrillar matrix composed mainly of cellulose, serves as a template for deposition of sporopollenin precursors. Following the release of microspores from tetrad, the tectum is formed on the primexine through deposition of sporopollenin precursors. With the progression in development, the sporopollenin is gradually deposited to thicken and consolidate the exine. Although a detailed biochemical analysis of sporopollenin has proven difficult, it is known to contain phenolics and polyhydroxylated aliphatics, covalently coupled by ether and ester bonds [5][6][7][8].
Recently, significant advances have been made in understanding the genes involved in pollen exine formation, including sporopollenin biosynthesis, in Arabidopsis and rice (reviewed in Zhang et al. [9]; Gomez et al. [10]). Several genes involved in synthesis of fatty acid precursors of sporopollenin have been identified [11]. Subfamilies CYP703A and CYP704 of cytochrome P450s have an essential role in hydroxylating the fatty acid constituents of predicted sporopollenin precursors. CYP703As catalyze the in-chain hydroxylation of fatty acids and heterologous CYP703A2 protein from Arabidopsis can catalyze in-chain hydroxylase of fatty acids with chain length from C10 to C16 [12]. Conversely, rice CYP703A3 has been shown to hydroxylase only lauric acid, preferably generating 7-hydroxylated lauric acid [13]. CYP703A3 in rice is required for the development of anther cuticle and pollen exine. The Arabidopsis CYP704B1 catalyzes the in-chain and ω-hydroxylation of fatty acids and is essential for exine biosynthesis [14]. Pollen from cyp704b1 mutant plants lack normal exine, but remain viable and capable of fertilization [14]. The rice OsCYP704B2 gene encodes a long-chain fatty acid hydroxylase capable of metabolizing very similar substrates in vitro as CYP704B1. Pollen grains in cyp704B2 mutants also lack a detectable exine resulting in male sterility. Anthers in these mutants have a defective tapetal layer and undeveloped cuticle [15]. The maize Ms26 gene encodes a cytochrome P450 mono-oxygenase enzyme (CYP704B1) [16,17] and microspores in ms26 mutants have a defective exine characterized by lack of sporopollenin deposition [18]. Similarly, sorghum mutants lacking functional MS26/CYP704B are male sterile due to defects in microspore development [19].
The conserved role of MS26/CYP704B in different species provided an opportunity to investigate its role in wheat. Bread wheat (Triticum aestivum L.) is an allohexaploid (2n = 6x = 42) combining ancestral genomes of Triticum urartu, Aegilops speltoides, and Aegilops taushii, contributing the A-, B-, and D-genomes, respectively [20,21]. Gene expression in bread wheat is often characterized by asymmetrical contribution of the three homeologous genomes as a result of genetic and epigenetic changes following polyploidization [22][23][24]. Functional analysis of the homologs of Ms26/CYP704B genes can be important for understanding the role of these genes in wheat reproductive development. Generation of multiple recessive mutant alleles in the wheat homeologs of Ms26/CYP704B gene through a custom-designed homing endonuclease was previously reported [19]. In this study, functional and cytological analyses of these mutants and their combinations were performed. Results indicate the importance of all three homeologs towards male fertility albeit with underlying differences. Apart from elucidating the function of male fertility genes, Ms26/CYP704B mutant alleles provide a novel supported the decision to publish. The specific roles of these authors are articulated in the 'author contributions' section.
Competing interests: Manjit Singh, Manish Kumar, Katherine Thilges, Myeong-Je Cho, and A. Mark Cigan are or were employed by DuPont Pioneer. MS and AMC are inventors on pending patent applications on this work and are current or former employees of DuPont Pioneer who owns the pending patent applications PCT/US2015/065768 and provisional patent application number, 62/ 092,604: Restoration of Male Fertility in Wheat". Materials reported in this paper may be subject to third party ownership and/or to governmental regulations. Availability of materials reported in this paper to academic investigators for noncommercial research purposes under an applicable material transfer agreement will be subject to the requisite permission from any third-party owners of all or parts of the materials and to governmental regulation considerations. Obtaining the applicable permission from such third-party owners will be the responsibility of the requestor. Transgenic materials reported in this paper would only be made available if in full accordance with all applicable governmental regulations. This does not alter our adherence to all the PLOS ONE policies on sharing data, and sharing materials (subject to the above), as detailed online in the guide for authors. source of nuclear male sterility in wheat that can be utilized for the development of hybrid seed production systems and exploitation of heterosis.

Results
Homeologs of TaMs26 are highly similar to grass Ms26/CYP704B genes and are specifically expressed in anthers To compare the wheat genes homologous to Ms26/CYP704B, genomic sequences corresponding to the three homeologs were obtained through BLAST (Basic Local Alignment Research Tool; [25]) of the TGACv1 wheat genome assembly (T. aestivum cv. Chinese Spring) (http:// plants.ensembl.org/Triticum_aestivum/Info/Annotation) with maize Ms26 mRNA sequence (accession NM_001137176). Sequence analysis showed that each of the wheat homeologs have four predicted exons and three introns with a combined exon-intron length of 1926 bp for Agenome and 1920 bp for B-and D-genomes (S1 Fig). These genes represent coding DNA sequences (CDS) of 1665 bp for A-genome homeolog, and 1656 bp for B-and D-genome homeologs, which are capable of encoding for proteins of 554 and 551 amino acids in length, respectively. Comparison of the predicted amino acid sequences showed that the three homeologs are highly similar with 98-99% amino acid identity (Fig 1). The wheat Ms26/CYP704B proteins also show high similarity (96-97% identity) to putative Ms26/CYP704B from barley (BAK08270), while identity with Ms26/CYP704B proteins from rice (XP_015629295.1), maize (NP_001130648.1), sorghum (XP_002465796), and Brachypodium (XP_003558727.1) was 88-89% (Fig 1). Given the high sequence identity of the wheat genes to other grass genes orthologous to the Ms26/CYP704B maize and rice genes, wheat Ms26/CYP704B from here onwards is denoted as TaMs26; and TaMs26 homeologs for A-, B-, and D-genome are denoted as TaMs26-A, TaMs26-B and TaMs26-D, respectively.
To study the RNA expression of the wheat TaMs26 homeologs, RT-PCR was performed with primers located in Exon 4 sequence homologous to all three genomes. qRT-PCR and semi-quantitative RT-PCR revealed that the wheat TaMs26 homeologs are expressed in the anthers from tetrad (late meiosis II) to early uninucleate microspore stages (Fig 2 and S2 Fig) with maximum expression at the uninucleate microspore stage. No expression was detected in the anthers at pre-meiotic and late uninucleate microspore stages or in the ovary, leaf, and root tissues. To estimate the distribution of RNA transcripts from each homeolocus, Illumina deep sequencing was performed on the RT-PCR product from the early uninucleate stage. The number of sequence reads obtained corresponding to TaMs26-A, TaMs26-B, and TaMs26-D was 256,228 (33%), 282,829 (37%), and 231,026 (30%), respectively, out of 770,083 sequence reads; thus indicating that all three homeologs were expressed in the anthers at comparable levels.
Two wild-type alleles of TaMs26 are required for maintenance of male fertility To analyze the functional contribution of TaMs26 to pollen development in wheat, previously generated single recessive mutant alleles in the three homeologs [19] were utilized. For simplicity, homozygous mutations in A-, B-and D-homeologs are denoted as TaMs26-a, TaMs26b and TaMs26-d, while heterozygous mutations are denoted as TaMs26-Aa, TaMs26-Bb and TaMs26-Dd, respectively. Plants containing five non-identical mutations, i) 4 bp Δ in TaMs26-A, ii) 23 bp Δ in TaMs26-B, iii) 81 bp Δ in TaMs26-B, iv) 90 bp Δ TaMs26-D, and v) 96 bp Δ + 2 bp I in TaMs26-D, were used for crossing experiments and further analysis. To track the different alleles, PCR assays with genome-specific primers were designed to genotype mutations in TaMs26-A, TaMs26-B, and TaMs26-D (see Materials and methods). Previous analysis of single mutants showed that plants homozygous for mutations in any one of the TaMs26-a, TaMs26-b, or TaMs26-d homeologs were male fertile and capable of generating self-pollinated seed similar to wild-type plants [19]. These results suggested that no single TaMs26 homeolog was functionally dominant and that a combination of two or all three homeologs contributed towards male fertility. To examine the impact of homozygous mutations in two genomes on male fertility, TaMs26-a, TaMs26-b, or TaMs26-d mutations were combined through conventional crossing. As shown in Table 1, double homozygous mutant pairs were generated in combinations that retained a wild-type genome for TaMs26-A, TaMs26-B or TaMs26-D. These double mutants showed normal vegetative and floral growth and were male fertile, generating viable pollen as indicated by seed set on these plants that was comparable to wild-type plants (Table 1). This data demonstrated that plants wild-type for TaMs26 in any one genome were competent to maintain male fertility. Moreover, plants that contained a homozygous mutation in one genome and a heterozygous wild-type allele in each of the other two genomes were also male fertile; for example, plants that contained homozygous 4 bp Δ, 81 bp Δ/WT and 90 bp Δ/WT in TaMs26-A, TaMs26-B and TaMs26-D, respectively, were male fertile and set seed similar to wild-type plants. This observation demonstrated that two wild-type Ms26/CYP704B alleles derived either from a single genome or from two different genomes were sufficient for male fertility in wheat.

Triple recessive mutations in TaMs26 result in male sterility
Triple homozygous mutants (Tams26-abd) were generated through conventional crossing to examine male fertility in the absence of a functional copy of TaMs26. Individual plants containing double homozygous TaMs26 mutants were crossed with plants containing a single homozygous TaMs26 mutation. Utilizing the above mentioned five deletions in A-, B-and Dgenomes, three sets of triple mutants were generated i) A-4 bp Δ, B-23 bp Δ, D-90 bp Δ; ii) A-4 bp Δ, B-81 bp Δ, D-90 bp Δ; iii) A-4 bp Δ, B-23 bp Δ, D-96 bp Δ + 2 bp I. Plants heterozygous for triple mutations (TaMs26-AaBbDd) were allowed to self-pollinate and progeny plants homozygous, heterozygous, and wild-type for each of the three mutations were identified. These plants were evaluated for vegetative and reproductive phenotypes. All triple homozygous mutants (Tams26-abd) were similar to wild-type plants with no obvious differences in vegetative growth characteristics or flowering characteristics (Fig 3A and 3B). Flowers from the Tams26-abd plants were identical to flowers from wild-type plants with the exception that anthers from the Tams26-abd plants were visibly shriveled in appearance ( Fig 3C). The presence of microspores was examined in the triple mutants and wild-type plants through confocal microscopy of anthers at the vacuolate stage of microspore development. In other grasses, such as maize, rice, and sorghum, mutations in the ortholog of the Ms26/CYP704B resulted in the breakdown of microspores shortly after tetrad release [15,18,19]. As shown in Fig 3D and  3E, anthers from a wild-type plant contained a high density of vacuolate microspores, while anthers from a Tams26-abd plant contained only a few vestigial microspores. These vestigial microspores were shrunken and appeared to have a deformed pollen wall compared to pollen from wild-type plants when examined by scanning electron microscopy (SEM) (Fig 4).
As has been shown in rice, CYP704B2 also has a role in the formation of anther cuticle. Unlike wild-type anthers, cyp704b2 mutant anthers exhibit abnormal epidermal cuticle formation which is smooth and glossy [15]. To investigate whether TaMs26 is similarly involved in anther cuticle formation, the anther epidermis at vacuolate microspore stage in the Tams26 Conserved role of Ms26/CYP704B in anther and pollen development and male fertility in wheat mutants was analyzed by SEM. In contrast to the epidermis of anthers from wild-type plants, which have well-developed cuticular and epicuticular structures (Fig 4A), the epidermal wall of anthers from Tams26-abd mutants was smooth and devoid of these cuticular structures ( Fig  4C). These observations demonstrated that, like rice, TaMs26 has a central role both in anther epidermis and pollen wall development in wheat.
One wild-type TaMs26 allele is insufficient for maintaining male fertility Interestingly, in addition to the triple homozygous mutants, anther structure and pollen viability was also affected in double homozygous-single heterozygous plants. Despite the presence of a single functional allele, seed number was severely reduced in these plants when compared to wild-type plants ( Table 2). Seed set in these mutant plants ranged from 0-32 seeds per plant and was dependent upon which genome contributed the wild-type allele. The lowest seed set was observed in Tams26-aBbd mutants and highest in Tams26-abDd mutants ( Table 2). These observed differences in genome contribution were more readily apparent when microspore development was analyzed in these plants. The most severe pollen phenotype was observed in Tams26-aBbd plants, which contained anthers that had severely collapsed and translucent microspores ( Fig 3G). The least severe phenotype was observed in anthers from Tams26-abDd plants, which contained numerous normal-looking microspores and a nearly equal number of collapsed and shrunken microspores (Fig 3H). The pollen phenotype in Tams26  Conserved role of Ms26/CYP704B in anther and pollen development and male fertility in wheat is insufficient to maintain male fertility in wheat and the alleles from three genomes are functionally non-equivalent.

Complementation of triple homozygous mutants requires two Ms26/ CYP704B genes
To further ascertain that the male sterility in Tams26  self-pollinate. F2 seed from this generation were planted and progeny genotyped for the introduced ZmMs26 gene and TaMs26 mutant alleles. Seed from these individual plants was counted as a qualitative measure of male fertility. As shown in Tables 2 and 3, in contrast to the low or no seed set observed with double homozygous-single heterozygous mutants, seed set was restored to wild-type levels when these plants contained a transformed copy of ZmMs26. However, ZmMs26 was unable to restore male fertility to Tams26-abd mutants when present as a single copy, and no seed set was observed from plants containing either of the two independent ZmMs26 insertions (Table 3). When coupled with the observation of male fertility in the double homozygous mutants and the single homozygous-double heterozygous mutants, observations from single gene complementation strongly suggested that two copies of the Ms26/CYP704B gene would be required to complement triple homozygous mutants. To test this hypothesis, the rice CYP704B2 gene (also referred as OsMs26) was added to a T-DNA expression cassette containing the ZmMs26 gene. Two Ms26 orthologs from maize and rice were selected based on the demonstration of function of these genes in previous studies [15,19]. This T-DNA expression cassette was transformed into double homozygous mutants (TaMs26-Abd), which were then crossed with A-genome homozygous mutants (TaMs26-aBbDd) to generate a population of triple  (Table 3). It was observed that both the anther epidermis and pollen wall phenotypes were restored to the wild-type phenotype in the plants carrying the ZmMs26-OsMs26 expression cassette (Fig 4E and 4F). Taken together, these results indicated that two functional alleles of Ms26/CYP704B gene are required for anther and pollen development in wheat.

Ms26/CYP704B genes show high conservation of structure and function in wheat and other grasses
Sporopollenin, the main constituent of pollen exine is highly conserved both in structure and composition. Orthologs of the sporopollenin biosynthesis pathway, including Acos5, PksA, PksB, CYP703A and CYP704B, are conserved among flowering plants but are not present in green algae [10]. Several studies for expression and functional analysis of these genes support their conserved role. The Physcomitrella patens PpASCL is a functional ortholog of Arabidopsis PKSA as it is capable of hydroxyalkylpyrone synthase activity with in vitro preference for hydroxyl fatty acyl-CoA esters [26]. The expression profile of the CYP704B gene family in plant species from bryophytes to angiosperms is conserved to inflorescences, male organs, and spores highlighting their evolutionary protective role in spores [15,27,28]. Similar to Arabidopsis, rice and maize Ms26/CYP704B orthologs that exhibit male reproductive tissue preferential expression [14][15][16], the expression pattern of TaMs26 homeologs is anther-specific. Furthermore, all three TaMs26 homeologs were observed to be expressed at comparable levels implying that each allele contributes to CYP704B function during pollen development in wheat. The role of MS26/CYP704B orthologs in pollen exine formation has also been functionally demonstrated across plants. Arabidopsis, rice, and maize mutants of Ms26/CYP704B orthologs have been shown to have defects in pollen exine structure [14,15,18]. Similarly, the triple homozygous wheat (TaMs26-abd) mutants were male sterile due to defects in microspore morphology.
Interestingly, compared to wheat and other grass species, the Arabidopsis cyp704B1 mutants exhibit a milder phenotype. Pollen in cyp704B1 mutant plants lack a normal exine but remain viable and can set seed upon self-pollination [14]; whereas Ms26/CYP704B mutants in wheat, rice, maize, and sorghum are male sterile. In addition to microspore exine defects, wheat Tams26 and rice cyp704B2 mutants also exhibit defects in anther epidermis [15]. Although it is not known if anther epidermis is similarly affected in Arabidopsis cyp704B1 mutants, it is possible that the male sterility in wheat and rice (and possibly maize and sorghum) is due to a combination of anther epidermis and pollen exine defects. If so, then the CYP704B2 genes in wheat and rice (and other grasses) could have assumed a more central role by contributing not only to pollen exine formation but also to anther epidermis formation. In Arabidopsis, other members of the CYP86 clan such as those involved in cutin formation [29,30] may contribute towards anther epidermis development.
by genetic or epigenetic resetting to orchestrate gene expression from the complex allopolyploid genome [31]. Resetting can result in genomic asymmetry or expression of individual genes from specific genomes and can lead to partial or complete genome dominance (reviewed in Madlung and Wendel [32]). In wheat, such partial or complete genome dominance is evident for several traits including free threshing phenotype and brittle rachis which are related to crop domestication [20]. Analysis of TaMs26 double homozygous and triple homozygous mutants suggested functional equivalency of all three TaMs26 homeologs. Ms26/CYP704B genes have also been analyzed in Brassica napus, a tetraploid species, and it was observed that a double knockout in duplicate functionally redundant genes, BnMs1 (BnCYP704B1-1) and BnMs2 (BnCYP704B1-1), is required to obtain complete loss of function, which results in male sterility. However, in wheat, the analysis of the double homozygous-single heterozygous mutants revealed noticeable differences in the contribution of each genome. Pollen and anther phenotypes, and consequently seed set, of these mutants were dependent on the genome that contributed the heterozygous allele. The fertility of double homozygous-single heterozygous mutants was highest when the heterozygous allele was contributed by D-genome (Tams26-abDd). These observations suggested that the function of Ms26/CYP704B in Tams26-abDd plants was likely reduced but still significant. On the other hand, anther and pollen phenotypes as well as seed set were lowest when the heterozygous allele was contributed by the B-genome (Tams26-aBbd). When analyzed as a single wild-type allele in a background of five mutant alleles there are noticeable functional differences between the three TaMs26 homeologs. These differences suggest dominance, although subtle, of the D-genome over A-and B-genomes for this gene. In contrast, it was observed in Brassica napus that the genotypes BnMs1ms1ms2ms2 and Bnms1ms1Ms2ms2 exhibit wild-type male fertility phenotype [33,34]. Transcript counts for the three wheat homeologs in anthers showed comparable levels, suggesting that the differences observed in phenotype were not due to differential transcription of the homeologs. These can likely be attributed to post-transcriptional regulation. Differences in protein sequences can also lead variable enzymatic activity and although the predicted amino acid sequences of the three homeologs show 98-99% amino acid identity, several differences in amino acid sequence exist that can have a role in protein function.
Two functional alleles of CYP704B/Ms26 are required for male fertility in wheat Analyses of double and triple homozygous TaMs26 mutants suggested that two wild-type TaMs26 alleles are absolutely required for male fertility. These two wild-type alleles can be contributed by a single genome or two different genomes. The complementation experiments with one and two heterologous copies of Ms26/CYP704B genes further augmented this conclusion. When one copy of ZmMs26 gene was introduced into a background of triple homozygous mutants (Tams26-abd), the heterologous gene was unable to restore male fertility. However, when one copy of ZmMs26 gene was introduced into a background of double homozygous-single heterozygous mutants, male fertility was completely restored irrespective of which genome contributed the heterozygous alleles. This observation was further borne out by the requirement of two exogenously supplied Ms26 genes, ZmMs26 and OsMs26, to restore fertility to Tams26-abd mutant plants. Under the conditions tested in this report, it is evident that two functional alleles of an Ms26/CYP704B gene are required for male fertility in wheat whether derived from two endogenous wild-type alleles or orthologous genes from different species.
Previous studies have shown that a single copy of ZmMs26 gene can complement ms26/ ms26 mutants in maize and sorghum, both diploid species [16,19]. While a single wild-type allele of Ms26/CYP704B gene is sufficient for normal function in diploid species, data presented here indicates that hexaploid wheat has evolved to require at least two functional alleles. It appears that due to the presence of six functional alleles, the polyploid wheat has evolved to require a higher dosage of MS26/CYP704B as compared to diploid species. Interestingly, in the case of Ms1 gene, another male fertility gene in wheat, the existing data suggests that only the B-genome homeolog has functional relevance. In certain genetic backgrounds, homozygous recessive mutations in the B-genome homeolog resulted in complete male sterility [35]. While Ms1 functions as a gene in diploid species and heterozygotes (Ms1/ms1) are male fertile, TaMs26 likely represents the other end of the spectrum where all three genomes contribute towards a phenotype, but the minimum requirement for male fertility is two wild-type alleles.

Conclusion
Knowledge of genes contributing to reproductive development in wheat has lagged other plant species as polyploidy in wheat makes it less amenable for functional genomics using traditional approaches. However, the development of new biological tools has facilitated the genesis of functional analysis of genes in wheat [36,37]. In addition to new advances in DNA sequencing, DNA double-strand break (DSB) technologies have provided efficient methods to obtain multiple mutant alleles in the different wheat genomes. Using a custom designed homing endonuclease, mutations were previously generated in each of the three homeologs of the wheat TaMs26 gene [19]. These mutants allowed for the characterization of their gene products to further the understanding of anther and pollen development in wheat. Moreover, the generation of mutant alleles and the ability to restore fertility using orthologous Ms26 genes sets the foundation for the development of a TaMs26-based male fertility control system for hybrid seed production and exploitation of heterosis to increase wheat yields.

Plant material and growth conditions
The TaMs26 mutant stocks were derived from the single-genome mutations generated by Cigan et al. [19]. Plants were grown in the greenhouse at DuPont Pioneer research facilities in Johnston, Iowa. The greenhouse conditions were set to 22˚C and 20˚C average day and night temperatures, respectively, with 30% humidity and 16 hr day length period. Standard plant care practices were followed including fertilizer and pesticide applications.

DNA and protein sequence analysis of TaMs26 homeologs
To obtain complete genomic sequence for wheat A-, B-and D-homeologs of Ms26/CYP704B, TGACv1 (The Genome Analysis Centre) sequence database was BLAST (Basic Local Alignment Research Tool; Altschul et al. [25]) searched with the maize Ms26 mRNA sequence (accession NM_001137176). TaMs26 sequences were annotated for exons, introns, and CDS and translated into predicted amino acids based on maize sequence. Sequence comparisons and alignments were performed with DNA analysis software: Sequencher (Genecodes) and Vector NTI (Life Technologies Inc.). Sequences of sorghum (XP_002465796), rice (XP_015629295.1), barley (BAK08270), and Brachypodium (XP_003558727.1) orthologs of maize Ms26 were obtained from the NCBI database.

Genome-specific PCR assay and sequence analysis
For amplification of TaMs26 genomic fragments, DNA from T. aestivum (cv. Fielder) was extracted from four small leaf punches (0.5 cm in diameter) as described in Gao et al. [38].
PCR was performed using Phusion 1 High Fidelity PCR Master Mix (New England Biolabs Inc.) according to the manufacturer's recommendations. PCR amplified fragments were also subcloned for DNA sequence analysis using the pCR2.1-TOPO cloning vector (Life Technologies Inc.). A genomic fragment corresponding to TaMs26 Exon 4 was amplified using primer pair UNIMS26 5'-2 and UNIMS26 3'-1 (S1 Table), which are conserved across the three genomes. Genome-specific fragments were amplified using one genome-specific primer and a second UNIMS primer. A-genome fragment was amplified with primers TAMS26-A and UNIMS26 3'-1; B-genome fragment was amplified with primers TAMS26-B and UNIMS26 5'-2; and D-genome fragment was amplified with primers TAMS26-D and UNIMS26 5'-2 (S1 Table). For genotyping larger deletions (TaMs26-B 81 bp Δ, TaMs26-D 90 bp Δ, and TaMs26-D 96 bp Δ+ 2 bp I), corresponding fragments were amplified with genome-specific primers and size fractionated on 2% agarose gel. Smaller deletions (TaMs26-A 4 bp Δ and TaMs26-B 23 bp Δ) were amplified with corresponding genome-specific primers and part of the amplification product digested with BsiWI, the DNA restriction enzyme that recognizes the sequence 5'-CGTACG-3', as described in Djukanovic et al. [17]. Products of these reactions were electrophoresed on 1% agarose gels and screened for BsiWI digestion resistant PCR products indicative of mutations at the Ms26 target site.

RT-PCR analysis and transcript counts through deep sequencing
Anthers were collected for stages corresponding to pre-meiotic stage, meiosis II (tetrad) stage, early uninucleate microspores, and late uninucleate microspores. Ovaries were collected at stages corresponding to meiosis and gametogenesis. Leaf and root tissue were collected at seedling stage. RNA was extracted using TRIZOL reagent (Invitrogen Inc.) following manufacturer's protocol. A total of 1 mg RNA was treated with DNase I enzyme (Invitrogen Inc.) following the manufacturer's recommendations to remove DNA contamination. 500 ng of DNase-treated RNA was reverse transcribed with the iScript cDNA synthesis kit (Bio-Rad). The transcribed cDNA was diluted 5X in sterile water and used for PCR. Primer pair TAMS26-F and TAMS26-R was used to amplify TaMs26 transcript and primer pair TAAC-TIN-F and TAACTIN-R was used to amplify TaActin, and products were analyzed by gel electrophoresis. qRT-PCR was performed with PowerUp SYBR Green Master Mix (ThermoFisher Scientific) with above mentioned primer pairs on a ViiA™7 System (ThermoFisher Scientific). Relative quantification values were determined using the difference in Ct from the TaMs26 and the reference gene (TaActin).
For transcript count by deep sequencing, RT-PCR was performed on the cDNA synthesized above with Phusion 1 High Fidelity PCR Master Mix (New England Biolabs) adding on the sequences necessary for amplicon-specific barcodes and Illumina sequencing using "tailed" primers through two rounds of PCR. The primers used in the primary PCR reaction were Ms26F and Ms26R while the primers used in the secondary PCR reaction were F2 and R2. The product amplified by these primers contained two single nucleotide polymorphisms (SNPs) that could discriminate between the three homeologs. The resulting PCR products were concentrated using a Qiagen MinElute 1 PCR purification spin column (QIAGEN, Inc.). A portion of the amplification products was visualized on an agarose gel and the remainder was single read 100 nt-length deep sequenced on an Illumina Genome Analyzer IIx (ELIM Biopharmaceuticals, Inc.) with a 30-40% (v/v) spike of PhiX control v3 (Illumina, Inc.) to off-set sequence bias. Two independent reactions were performed. The reads were clustered based on their SNPs and the number of reads was counted using in-house designed sequence analysis software.

Ms26 complementation constructs and transformation
To generate the ms26 fertility complementation T-DNA vector, the Ms45 gene in PHP24490 [39] was replaced by a 3.9 kb DNA fragment [Chromosome 1, nt 4509689-14506399 Maize (B73) RefGen_v2 (MGSC)], which contained the wild-type maize Ms26 gene (also referred to as ZmMs26). The T-DNA expression cassette also carried an α-amylase gene under the control of pollen-specific PG47 promoter along with DsRed gene regulated by a seed-specific promoter as the transformation selectable marker. The α-amylase prevented the pollen-specific transmission of the ZmMs26 gene and resulted in 50:50 segregation of ZmMs26 in transformed plants. The final T-DNA expression cassette, pPG47::Bt1:ZmAA1//Ms26//35SEN-pEND2:: DsRed2, was introduced into Agrobacterium strain LBA4404 as described above. To make the construct with two Ms26 orthologous genes, a 3.2.0 kb genomic fragment of the rice Ms26/ CYP704B gene (LOC_Os03g07250.1; chromosome 3 nt 3701612-3704874, Os-Nipponbare ref IRGSP-1.0MSU Rice Genome Annotation Project) including approximately 1.0 kb fragment 5' of the ATG translation start site and 0.5 kb 3' of the translation stop site, was amplified from Orzya sativa L. ssp. Japonica, cv. Kitaake, with primers OSMS26-F and OSMS26-R. Amplification and sequencing was performed as described for the wheat TaMs26 genomic fragments. The rice Ms26 ortholog (also referred to as OsMs26) was then cloned into the above mentioned vector, which contained the ZmMs26 gene. The T-DNA expression cassette was transformed into T. aestivum (cv. Fielder) containing homozygous Tams26-a and b mutations as described in Cigan et al. [19]. Transformed plants were crossed with pollen from Tams26aBbd plants and progeny analyzed (see S4 Fig for crossing strategy). Plants transformed with ZmMs26 and ZmMs26-OsMs26 constructs were genotyped with the ZMMS26 TERM primer pair and OSMS26PRO primer pair, respectively (S1 Table).

Microscopy
For scanning electron microscopy (SEM), anthers and pollen were fixed in FAA (formalin, acetic acid, EtOH) and dehydrated in a graded EtOH series (50%, 75%, 95%, 100%). At this point, the samples were critical point dried using liquid carbon dioxide (Denton Vacuum). Anthers and pollen were mounted on aluminum stubs with double-sided sticky tabs, and surrounded with silver paint. The samples were then sputter-coated (Denton Vacuum) with palladium-gold (60:40) and observed using a JEOL 5800 scanning electron microscope (JEOL USA, Inc.) at 10 kV.
For confocal scanning laser microscopy (CSLM), flowers were collected and fixed in 2% paraformaldedye: 4% glutaraldehyde in PBS buffer and vacuum infiltrated at 10 psi to increase the penetration of fixative. Samples were rinsed in PBS and then cleared in graded series of 2,2'-Thiodiethanol (TDE) (10%, 25%, 50%, 75%, and 97%). Samples were mounted on slides in TDE and imaged using the Leica TCS SPE CSLM (Lecia Microsystems) with 405 nm, 488 nm and 532 nm laser lines. 20 anthers from each mutant class and wild type plants were analyzed.  Table. List of primers used in the study. (PDF)