Introduction

Preharvest sprouting (PHS) is the germination of mature grain on the mother plant before harvest under rainy or humid conditions (reviewed by Rodríguez et al. 2015). Lack of grain dormancy accounts for 60–80% of the variation for preharvest sprouting susceptibility in bread wheat (Triticum aestivum L.) (DePauw and McCaig 1991). Grain dormancy refers to the inability to germinate even under favorable environmental conditions (reviewed in Bewley et al. 2013). Grain is most dormant at physiological maturity and then lost through a period of dry storage called after-ripening (Paterson et al. 1989; Gerjets et al. 2010). Dormancy is also broken through imbibition in the cold (called cold stratification) or through scarification of the seed coat. Seed coat-imposed dormancy is relieved by seed coat scarification, whereas embryo dormancy is not. Wheat can have both forms of seed dormancy (Schramm et al. 2010).

A seed is considered germinated when the radicle or any other part of the embryo emerges from the seed coat (reviewed in Bewley et al. 2013; Mares and Mrva 2014). Hydrolytic enzymes like alpha-amylase are induced during germination to mobilize stored reserves for use in seedling growth. Alpha-amylase degrades starch during germination. Even before grain is visibly sprouted, alpha-amylase activity in dough can result in poor product end-use quality. While higher seed dormancy prevents PHS, too much dormancy at planting causes problems with uneven or poor seedling emergence (reviewed by Rodríguez et al. 2015). Thus, it is important to understand the genetic control of wheat seed dormancy and germination to develop varieties with an appropriate balance between dormancy and emergence.

Multiple environmental and genetic factors control the degree of seed dormancy. In wheat, seed dormancy is higher if the environment is cool and dry during embryo maturation (Nakamura et al. 2011; Mares and Mrva 2014). Plant hormone signaling controls seed dormancy and germination (reviewed in Finkelstein et al. 2008). The hormone abscisic acid (ABA) establishes seed dormancy during embryo maturation, maintains dormancy in mature seeds and inhibits the germination of mature seeds when exogenously applied. The hormone gibberellin A (GA) breaks dormancy and stimulates wheat seed germination (Tuttle et al. 2015). Thus, increased ABA signaling or decreased GA signaling is associated with higher seed dormancy. Dormancy is also higher in wheat varieties with red grain coat color than white (Himi et al. 2002, 2011). ABA and grain coat color can act independently and in concert to increase wheat grain dormancy (Himi et al. 2002). Mutant studies have shown that loss of ABA sensitivity can decrease seed dormancy in red wheat and that increased ABA sensitivity can increase seed dormancy in both red and white wheat varieties (Flintham 2000; Torada and Amano 2002; Schramm et al. 2010, 2012, 2013). Since white wheat varieties have lower seed dormancy and PHS tolerance, we have identified ethyl methanesulfonate (EMS)-induced genetic variants with increased ABA sensitivity as a method to increase grain dormancy and PHS tolerance in white wheat.

The EMS-induced semi-dominant ERA8 (Enhanced Response to ABA8) mutation was isolated in the soft white spring wheat cultivar ‘Zak’ (Kidwell et al. 2002; Schramm et al. 2013). The ERA mutants fail to germinate on ABA concentrations too low to prevent wild-type seed germination. The ‘ZakERA8′ line has increased seed dormancy, seed ABA sensitivity, and PHS tolerance than wild-type Zak (Schramm et al. 2013; Martinez et al. 2014, 2016). Since ERA8 dormancy is lost rapidly through 8–10 weeks of after-ripening, this mutation does not have an obvious adverse effect on emergence (Martinez et al. 2014, 2016). Thus, ERA8 is a good allele for improving PHS tolerance. Mapping would facilitate ERA8 selection as a source of PHS tolerance in white or red wheat. ABA signaling can be controlled both at the level of hormone accumulation and at the level of the ABA sensitivity. Endogenous embryo ABA measurements showed that ABA content decreased similarly with after-ripening of ERA8 and wild-type (WT) Zak, suggesting that the phenotype resulted from increased ABA signaling rather than decreased ABA turnover (Martinez et al. 2016). Thus, we hypothesize that ERA8 is a gain-of-function mutation in a positive regulator of ABA signaling.

Fine mapping and map-based cloning of wheat genes are challenging due to the large size (~ 17 Gb) and high redundancy of the allohexaploid bread wheat genome (Yan et al. 2003; Gupta et al. 2008; Shewry 2009). The large size is due to the fact that it contains over 80% noncoding DNA, and that it is derived from three diploid progenitors that each contributed seven homoeologous chromosomes referred to as the A, B, and D genomes (2n = 6x = 42, genomes AABBDD).

Quantitative trait loci (QTL) for PHS and dormancy have been identified throughout the wheat genome (Zanetti et al. 2000; Ogbonnaya et al. 2008; Munkvold et al. 2009; Kulwal et al. 2012; Jaiswal et al. 2012; Zhou et al. 2017; Martinez et al. 2018). While homologues of ABA signaling genes have been found on wheat chromosomes 3A, 4A, and 5A, only Viviparous-1/ABA-insensitive3 (VP1/ABI3) has been linked to differences in wheat PHS tolerance (Nakamura and Toyama 2001; McKibbin et al. 2002; Nakamura et al. 2007). To date, 18 dormancy and 13 PHS studies have identified a QTL on the long arm of chromosome 4A called the Phs1 locus (Liu et al. 2008, 2014; Mares and Mrva 2014; Shorinola et al. 2016; Martinez et al. 2018; Zuo et al. 2019). This 4A QTL has been cloned as the wheat Mitogen-Activated protein Kinase Kinase 3 gene (TaMKK3-A) (Nakamura et al. 2016; Torada et al. 2016; Shorinola et al. 2017). Another major PHS tolerance QTL on chromosome 3A was cloned from Japanese and US germplasm, the wheat gene MOTHER OF FLOWERING TIME AND TFL1 (TaMFT1) (Nakamura et al. 2011; Liu et al. 2013b).

The reference genome sequence of allohexaploid wheat landrace Chinese Spring has enabled the application of high-throughput sequencing methods to mapping and cloning in polyploid wheat (IWGSC et al. 2018). Even so, the large size and high redundancy of the wheat genome is still a challenge. Several methods facilitate the use of high-throughput sequence analysis for mapping in the complex wheat genome by enriching relevant regions. Genotyping-by-sequencing (GBS) entails restriction enzyme digestion of genomic DNA to enrich for DNA fragments that begin with common restriction sites for high-throughput genome resequencing (Poland et al. 2012; Truong et al. 2012). Exome sequencing (exome-seq), on the other hand, enriches specifically for the gene-containing regions of the genome through affinity capture and purification of the predicted wheat coding regions for high-throughput genome resequencing (Allen et al. 2013; Henry et al. 2014; Uauy et al. 2017; Krasileva et al. 2017; Mo et al. 2017).

Bulked segregant analysis (BSA) of genome resequencing data is a powerful approach for mapping and cloning. In this approach, pooled-DNA samples from plants with contrasting phenotypes are sequenced to identify polymorphisms co-segregating with the trait of interest (Michelmore et al. 1991). In organisms with smaller genomes, it is now common to clone mutagen-induced polymorphisms using BSA by contrasting whole genome DNA sequence of bulked wild-type and mutant backcross populations (reviewed in Thole and Strader 2015). However, the wheat genome is too large for this approach. Wheat genes have been fine-mapped by BSA using RNA-seq instead of DNA-seq to enrich for the expressed portion of the genome (Trick et al. 2012; Liu et al. 2012; Ramirez-Gonzalez et al. 2015a). The RNA-seq approach has the drawback that relative gene expression levels vary in different tissues and times in development, leading to the possibility that the gene of interest might be underrepresented in the sampled tissue. Exome capture has enabled affinity-purification of the predicted wheat coding regions before performing whole genome DNA resequencing (Allen et al. 2013; Henry et al. 2014; Uauy et al. 2017; Krasileva et al. 2017; Mo et al. 2017). This approach has been used to characterize the genome sequence of EMS-mutagenized tetraploid wheat TILLING populations (Harrington et al. 2019). Since this method was effective in identifying EMS-induced mutations responsible for a phenotype, this study used exome capture in a BSA-exome sequencing approach to map EMS-induced mutations linked to the ERA8 allele.

In this study, ERA8 was localized to chromosome 4A in a backcross population using bulked segregant analysis to map relative to EMS-generated polymorphisms characterized by high-throughput DNA sequencing of exome genomic DNA (BSA-exome-seq). Fine mapping with a larger backcross population was used to co-localize ERA8 with mutations in and near the TaMKK3-A gene that was previously cloned as a gene regulating wheat preharvest sprouting tolerance. QTL analysis of RIL populations was used to confirm the gene location. This information is being used to introgress ERA8 into soft white winter wheat to improve PHS tolerance by increasing ABA sensitivity.

Materials and methods

Plant material

Allohexaploid white spring wheat cultivars (Triticum aestivum L.) were used in this study. ‘ZakERA8’ (PI 669443) is derived from the soft white spring cultivar ‘Zak’ (PI 612956) via an EMS-induced mutation resulting in enhanced response to ABA during seed germination (Kidwell et al. 2002; Schramm et al. 2013; Martinez et al. 2014). Soft white spring ‘Louise’ (PI 634865) was selected from the cross ‘Wakanz’/‘Wawawai’ (Kidwell et al. 2006b). Hard white ‘Otis’ (PI 634866) is a selection from the cross ‘Idaho 377s’/3/‘Tanager S’/‘Torim 73’//‘Spillman’(Kidwell et al. 2006a). Mapping populations derived from these lines were grown in a controlled greenhouse environment with supplemental lighting to obtain a 16 h day/8 h night photoperiod with 15–17 °C night and 21–23 °C day temperatures.

Backcross (BC) populations were developed by three successive crosses of the ABA-hypersensitive ZakERA8 to the mutagenesis parent Zak (Fig. S1; Schramm et al. 2013). After the third consecutive cross, a single BC3F1 line was advanced to BC3F2 by self pollination. Because ERA8 is semi-dominant, homozygous ERA8/ERA8 and +/+BC3F2 were identified by progeny testing in the F3 generation based on a non-segregating ABA-hypersensitive ERA8-like and normal Zak-like germination phenotype. ERA8, and Zak genomic DNA was prepared from leaf tissue of the BC3F3 generation and used for bulked segregant analysis. Additional BC3F2:3 seeds (X5.1 and X5.2) were phenotyped and used for fine mapping.

Two populations were constructed to confirm the ERA8 map position, a Louise/ZakERA8 F2:5 recombinant inbred line (RIL) population and an Otis/ZakERA8 F2:3 population (Fig. S1). The Louise/ZakERA8 RIL population was derived from F2 seeds from ten independent F1 plants from the cross of Louise to ZakERA8 (Louise as the female parent). The population was developed without selection through single-seed descent of 225 RILs from the F2 generation to the F5 generation. Genomic DNA was extracted from single F5 generation plants grown in 3 L pots in the Plant Growth Facility at Washington State University, Pullman, WA. F6 seeds were harvested at physiological maturity (environment 1 or E1). For each RIL, 30 F6 seeds from E1 were advanced in headrows at the Washington State University Spillman Research Farm, Pullman, WA in 2014 and 2015 to produce F7 grain from the E2 and E3 environments, respectively. To construct an Otis/ZakERA8 F2:3 mapping population, Otis was pollinated with ZakERA8 (Kidwell et al. 2006a). F1 seeds were advanced in the greenhouse and F2 harvested at physiological maturity. Genotyping was performed using genomic DNA prepared from F2 leaf tissue, whereas F3 seeds were phenotyped by ABA germination assay. Genomic DNA was extracted from Louise/ZakERA8 RILs and Otis/ZakERA8 F2 plants on the Qiagen BioSprint-96. DNA concentrations were determined using the BioTek Gen5 plate reader, and DNA was diluted to 50 ng/µL.

Seed germination assays

Plating assays were used to compare ABA sensitivity during seed germination using experimentally determined conditions that best differentiated between wild type (WT) and ERA8/ERA8 parents for each population or after-ripening time point (Fig. S2). The wheat grain is a caryopsis because it includes a pericarp, but is referred to here as a “grain” or “seed” for the sake of simplicity (reviewed in Bewley et al. 2013). Seeds were harvested at physiological maturity, hand threshed, then dry after-ripened at ambient temperature (20–23 °C) and humidity (15–30%) for 5–7 weeks (as in Martinez et al. 2016). Seeds (30–90) were sterilized in a 50 mL tube with 20 mL of 10% bleach/0.01% SDS for 15 min and then rinsed 3x with 40 mL of sterile deionized water. Thirty seeds were plated on Petri dishes containing a 9-cm blue germination disk (Anchor Paper Co) saturated with 6 mL of 5 mM MES buffer, pH 5.5 (2-(N-morpholino)ethanesulfonic acid, Sigma-Aldrich) containing 2 µM, 5 µM, or 10 µM (±)-ABA (PhytoTechnology Laboratories). Stock solutions of 0.1 mM (±)-ABA were prepared in ethanol and diluted as indicated.

Population parents were plated on varying ABA concentrations weekly from 3 to 8 weeks of after-ripening to determine the after-ripening time point and ABA concentration with the optimal difference between parents (Fig. S2). Seeds were stored at − 20 °C for no more than 4 months to preserve dormancy until the optimal dry after-ripening time point was selected (Paterson et al. 1989). However, the larger X5 BC3F3 populations used for fine mapping were stored for 3 to 17 months at − 20 °C, and then the after-ripened seeds were plated on the optimal ABA concentration as indicated (Table S1). Because different wheat seed lots have different degrees of initial dormancy, and because grain after-ripens slowly in the freezer, different ABA concentrations were needed to score the phenotype in seed lots stored for different lengths of time at − 20 °C (Paterson et al. 1989; Schramm et al. 2013). Germination was scored daily over 5 days of incubation at 30 °C. Seeds were considered germinated upon embryonic root emergence. Germination index (GI), a value weighted for speed of germination, was calculated over 5 days of imbibition as (5 × gday1 + 4 × gday2 … + 1 × gday5)/(5 × n) where g is the number of additional seeds germinated on each day and n is the total number of seeds (Walker-Simmons 1987; Schramm et al. 2013). GI ranges from 0 to 1, where a GI of 1 indicates that all of the seeds germinated on day 1 of imbibition. Raw means and standard deviations were calculated from three technical replicates of ten seeds. Chi-squared analysis was used to determine goodness-of-fit to Mendelian segregation models for F2:3 seeds as in Schramm et al. (2013; Table S2). The χ2 statistic was calculated as Ʃ [(O − E)2/(E)] where O was the observed number of seeds germinated or ungerminated, and E is the expected number of seeds germinated or ungerminated based on Mendelian segregation. A χ2 distribution table was used to determine the p values based on the χ2 statistic and the degrees of freedom = 1. The model fits the observed values when p > 0.05 and does not fit when p < 0.05.

Exome capture and sequencing

High-quality genomic DNA from the Zak/ZakERA8 BC3F3 bulked populations was extracted using a phenol/chloroform-based DNA extraction method followed by RNAse A digestion (as in Yu et al. 2017). A 2.5-cm-long leaf sample was collected from BC3F3 plants and ground together with lines that comprised each bulk. For each individual parent, a 10–15-cm-long leaf sample was ground for DNA extraction.

Whole genome DNA resequencing of the large wheat genome (16 Mb) for bulked segregant analysis is prohibitively costly (Uauy et al. 2017). Exome capture was used to restrict genomic DNA sequencing to those regions encoding predicted genes (Krasileva et al. 2017). Library preparation and sequencing were carried out at the Earlham Institute, UK as described by Harrington et al. (2019). Separate barcoded libraries were prepared from the ERA8 parent (ERA8/ERA8), wild-type Zak parent (+/+), the bulked ERA8-like, and bulked Zak-like DNA with average insert sizes of 327, 345, 348, and 340 bp, respectively. The four libraries were multiplexed and captured on one exome-capture probeset using the wheat NimbleGen target capture following the SeqCapEZ protocol v5.0 (Krasileva et al. 2017). Sequencing was conducted on a single Illumina HiSeq 2500 lane producing 125-bp paired-end reads.

Read quality was analyzed using FastQC, and the reads with a base quality score ranging from 20 to 40 were retained (Table S3; Babraham Bioinformatics 2012). Adapter sequences were removed using Trimmomatic with the following parameters: ILLUMINACLIP (Truseq3-PE), CROP (125 bp), and MINIMUMLENGTH (50 bp) options (Bolger et al. 2014). The short read alignment tool, Bowtie2, was used to align trimmed reads to the wheat genome assembly, IWGSC RefSeq v1.0 from Chinese Spring, with some modifications (Langmead et al. 2009; Li and Durbin 2009; IWGSC et al. 2018; http://www.wheatgenome.org). First, we removed the unassigned chromosome (ChrUn) containing unanchored scaffolds in this assembly as these scaffolds lack positional information useful for mapping. Also, TaMKK3-A, an important gene on wheat chromosome 4A that controls PHS and dormancy in wheat is unanchored in the RefSeq v1.0 assembly (IWGSC et al. 2018), so we complemented the modified RefSeq v1.0 assembly with a 30.4 kb scaffold from the Synthetic W7984 wheat genome assembly (Chapman et al. 2015) spanning the TaMKK3-A region. The Bowtie2 end-to-end alignment algorithm was used with the sensitivity of the alignments increased by adjusting the minimum score threshold parameter (-score-min) to L, 0, 0.15 to allow only 2–3 mismatches per alignment depending on the base quality score. The raw sequence alignment map (SAM) files were processed (converted, sorted, and indexed) using SAMtools (Li et al. 2009a). To improve the accuracy of read depth values, duplicate reads resulting from PCR amplification during library preparation were removed using the Picard Markduplicate tool (http://broadinstitute.github.io/picard). Only properly paired reads, defined as correctly oriented paired reads with mapping distance less than average insert size, were retained.

Bulked segregant analysis

Exome variants including single-nucleotide polymorphism (SNP) and insertion or deletion (InDel) sites were first identified between the Zak WT and ERA8 parents using a Samtools (mpileup)-bcftools (view) variant-calling pipeline (Li et al. 2009a). Snpsift (snpEFF) was used for the analysis, manipulation, and interactive filtering of the VCF files (Cingolani et al. 2012). First, SNP variant sites with the non-reference allele in Zak were excluded as these most likely represent intervarietal SNPs between Chinese Spring and Zak. Second, ERA8-specific SNP sites with alternate (non-reference) alleles in ERA8 and reference allele in Zak WT were selected. Only SNPs with a variant quality score (“QUAL”) above 20 were selected to improve accuracy. Subsequent analysis concentrated on the 1854 transition variants (C to T or G to A) as these were likely caused by EMS mutagenesis. Variants in the 1854 sites between the bulked ERA8-like and WT-like sequence were called using the same variant-calling pipeline but with an additional flag to specify the position of the target sites in the mpileup analysis (–position). Bulked segregant analysis was performed to identify sites enriched in the ERA8-like versus WT-like bulked DNA. To do this, the ERA8-specific alternate allele frequencies at the selected sites were first calculated using the bulked allele frequency formula, BF = DV/DP, where DV is the number of reads with alternate allele and DP is the total read depth. To improve accuracy, analysis was restricted to sites with DP >=5. The bulked allele frequency difference (BFD) was then calculated between the two bulks as BFD = BFERA8_b − BFZWT_b. The analysis was based on the rationale that ERA8 plants should show an enriched frequency of the ERA8-like sequence variants in regions linked to the ERA8 gene, but should show random segregation of unlinked sequence variants. The BF and BFD across the whole genome was visualized with Circos to highlight the ERA8-linked region with the highest BFD in the genome (Krzywinski et al. 2009).

Fine mapping

ERA8 was fine-mapped relative to EMS-induced SNPs genotyped using Kompetitive Allele Specific PCR (KASP) in Zak/ZakERA8 BC3F2:3 lines (Smith and Maughan 2015). KASP markers for SNPs identified through whole genome exome resequencing of Zak and ZakERA8 were designed using PolyMarker with default settings (Table S4; Ramirez-Gonzalez et al. 2015b). The KASP assay for the naturally occurring preharvest sprouting susceptible (A660) versus resistant (C660) alleles of TaMKK3-A was performed as described in (Shorinola et al. 2017). Genomic DNA was extracted from BC3F2 leaf tissue on the Qiagen BioSprint-96, and DNA concentrations were determined using the BioTek Gen5 plate reader. A standard 10 µL KASP reaction volume contained 15 ng/µL gDNA, 1x KASP Master Mix (LGC Genomics; V4.0 2x Mastermix 96/384; Low Rox), 0.42 µM reverse primer, and 0.168 µM each of allele 1 and 2 forward primers (Smith and Maughan 2015). KASP assays were performed on a thermocycler as follows: 1 cycle of 94 °C for 15 min, 10 cycles of 94 °C for 20 s and 61–55 °C for 60 s (starting at 61 °C, with 0.6 °C decrease in temperature per cycle), then 26 cycles of 94 °C for 20 s and 57 °C for 60 s (Biometra Tadvanced, Analytik Jena). FAM and HEX fluorescence was measured on the LightCycler 480 System (Roche).

In a preliminary experiment, 2381 Zak/ZakERA8 BC3F2 lines were scored for WT versus ABA-hypersensitive ERA8 phenotype by plating on 2, 5, or 10 μM (±)-ABA. KASP assays were used to identify those lines containing recombination events within the ERA8-linked region. Of these, 424 were advanced to the BC3F3 and seeds were progeny tested for ABA sensitivity (Table S1). Based on Chi-squared analysis, BC3F2:3 from X5.1 (122 lines) and X5.2 (302 lines) showed Mendelian segregation for a single semi-dominant (or additive) gene (Fig. S1; Table S5). A linkage map of the X5.1 and X5.2 Zak/ZakERA8 BC3F2:3 populations was constructed using the regression algorithm and Kosambi mapping function in JoinMap v4.0 (Kosambi 1943; Van Ooijen 2006).

Genotyping-by-sequencing

Genotyping-by-sequencing (GBS) of Louise/ZakERA8 RILs was performed on a Thermo-Fisher Scientific Ion-Proton system (Poland et al. 2012; Truong et al. 2012). Briefly, to construct GBS libraries, DNA was digested with PstI and MspI and then ligated to barcode adapters using T4 DNA ligase. Samples were purified, pooled and sequenced. The USDA-ARS Western Regional Small Grains Genotyping Lab SNP marker pipeline was used to analyze the sequencing reads. This pipeline used built-in Linux commands and routines written in the Perl 5 programming language (http://www.perl.org) to remove extraneous bases 5’ to the PstI site, trim the sequence reads to uniform length “tags” of 70 or 100 bases and then identify tags unique to each of the parent lines. These tags were aligned to the RefSeq v1.0 wheat genome (IWGSC et al. 2018) using SOAP2 allowing 0–2 mismatches (Li et al. 2009b). Tags that aligned to two or more chromosomes were discarded, leaving only tags that aligned to 0 (unknown) or 1 chromosome. A Perl 5 routine was then used to identify pairs of these filtered tags that differed by one or two SNPs between the parents, Louise and ZakERA8, possibly representing alternate alleles. Linux built-in commands were then used to score the presence of those markers in the progeny population, and a Perl 5 routine was used to convert the presence/absence data to genotype calls. Markers that were scored as missing (neither allele present) in more than 75% of the progeny were removed from the data set. An iterative imputation of missing data was calculated using the ‘missforest’ package in R (Stekhoven and Bühlmann 2012). If a marker appeared to be heterozygous for 21 or more of the 207 RILs, then that marker was excluded from analysis to reduce false-positive results.

Genetic linkage map and QTL mapping

A linkage map of the Louise/ZakERA8 RIL population was constructed using SNP data from GBS using the maximum likelihood algorithm in JoinMap v4.0 (Van Ooijen 2006). Linkage groups of 5 or more markers were selected with a LOD above 7.0 (Table S6). The 92 of 2234 markers showing segregation distortion, p < 2e−05, may have resulted from genotyping errors and so were omitted from the analysis. Composite interval mapping (CIM) was performed with the R/qtl package in R v3.2.5 using the Haley–Knott regression algorithm, and the Kosambi mapping function with step = 0 (Haley and Knott 1992; Broman et al. 2003). Significant LOD scores were determined using the ‘scanone’ function in R/qtl with the Haley–Knott method and 1000 permutations for each trait. QTL graphs were constructed using the ‘ggplot2’ package v3.1.0 in R v3.2.5 (Wickham 2016).

Linkage analysis was performed in the Louise/ZakERA8 RIL, Zak/ZakERA8 BC3F2:3, and Otis/ZakERA8 F2:3 populations using genotype information from the ERA8 EMS-induced KASP markers (Table S4; Supplementary Material 3). QTL analysis for both populations was conducted as described above for the Louise/ZakERA8 GBS data. The QTL analysis R script can be found in the public repository ‘ERA8-Mapping-2019’ on GitHub (https://github.com/shantel-martinez/ERA8-Mapping-2019).

RNA sequencing

An RNA-seq experiment (2.5 million reads) was performed on the original BC3 parents Zak (Zak2.7.3.7.2.3.10.3) and ZakERA8 (BC2F3; 95.6.85.4). Seeds were imbibed on 5 mM MES, pH 5.5 for 8 h. Embryos were dissected and frozen in liquid nitrogen under a green safelight to avoid induction of photosynthesis genes. RNA was extracted from 20 dissected embryos per sample using a modified version of an RNA extraction technique which yielded an average of 2300 ng/μL of RNA with a RIN > 8.0 (Onate-Sánchez and Vicente-Carbajosa 2008; Nelson and Steber 2017). This was replicated three times for each parent sample. Sequencing was performed using the Illumina HiSeq 2500 (WSU Genomics Core). Quality of the RNA reads was checked using FastQC, and all samples had a quality score > 32 (Babraham Bioinformatics 2012). Pseudo-alignment and quantification of RNA-seq reads were conducted with kallisto-0.42.3 using the quant command with options −b 100 and −t 8 (Bray et al. 2016). The wheat RefSeq v1.0 transcriptome assembly (IWGSC et al. 2018) was used as the reference genome for this alignment. SAMtools was used to sort and index the BAM files v0.2.19 (Li et al. 2009a). Gene differential expression analysis was done using the ‘sleuth’ package in R (Pimentel et al. 2017).

Results

ERA8 mapping by bulked segregant analysis of a backcross population

ERA8 was mapped in the Zak/ZakERA8 backcross population relative to DNA polymorphisms induced by the EMS mutagenesis of Zak (Fig. 1; Fig. S1). ERA8 segregates as a single semi-dominant Mendelian trait in backcross populations when the phenotype is examined at optimized ABA concentrations and after-ripening time points (Schramm et al. 2013; Martinez et al. 2014). Consistent with this, the Zak/ZakERA8 BC3F2 population (X5.2) showed segregation as a single semi-dominant gene on 2 μM ABA at 5 weeks of after-ripening (χ2 = 1.62, p value = 0.20; Table S2). The germination of the 300 BC3F2:3 lines exhibited a normal distribution in the presence of 2 μM ABA, and the corresponding Zak WT and ERA8 controls showed an average of 66.9% (SD ± 16.3) and 16.8% (SD ± 8.8) germination after 5 days of imbibition, respectively (Fig. 1b). Thus, an F2:3 line was considered to be homozygous WT (+/+) if the percent germination was above 73% and to be homozygous ERA8 (ERA8/ERA8) if below 24%. These individuals were tested on ABA two more times to confirm the phenotype. The 15 BC3F2:3 lines that consistently showed high percent germination on ABA were defined as WT-like (WT_b:15), and the 17 individuals that consistently showed low percent germination on ABA were defined as ERA8-like (ERA8_b:17) (Fig. 1b).

Fig. 1
figure 1

The BSA-exome-seq method. a Genetics. The BC3F2:3 population was developed by crossing Zak WT (+/+) and BC2F3 ZakERA8 (ERA8/ERA8). BC3F2:3 genomic DNA was used for exome-seq. The phenotype was BC3F2:3 grain germination on ABA. b Selection of bulked lines. Compared to the parental ERA8 (purple arrow) and WT (green arrow) germination on ABA, 17 ERA8-like and 15 WT-like lines were chosen for bulked segregant analysis. c The number of 125 bp paired-end Illumina HiSeq 2000 reads generated for each parent, WT bulk (WT_b), and ERA8 bulk (ERA8_b) are shown. d Alignment of parental reads to the Chinese Spring IWGSC reference sequence (v 1.0), where varietal differences from Chinese Spring were shared by the Zak WT and ERA8 parents, and mutagen-induced (C/G to T/A) were ERA8-specific. e Unlinked SNPs showed random variation in WT_b versus ERA8_b, whereas ERA8-linked SNPs did not. An example of an exon alignment and bulk frequency ratios is shown. f KASP markers were designed to score SNPs. SNP_17 is shown (color figure online)

Full size image

Bulked segregant analysis was performed using bulked genomic DNA from the ERA8-like (ERA8_b) and WT-like (WT_b) BC3F2:3 lines. In order to maximize the number of informative sequences, genomic DNA was enriched for coding sequence using exome capture before performing high-throughput DNA sequencing on the Illumina HiSeq 2500 (48–67 million reads per sample; Fig. 1c; Table S3). Likely, EMS-induced C to T and G to A polymorphisms between Zak WT and ERA8 were identified by aligning to the RefSeq v1.0 wheat genome (Fig. 1d; IWGSC et al. 2018). SNPs identified between WT and ERA8 were then screened against the ERA8-like and WT-like bulked exome sequences. If a region of the chromosome was not associated with the ABA-hypersensitive phenotype, then a polymorphic SNP should segregate randomly, appearing in about 50% of the WT-bulk reads and 50% of the ERA8-bulk reads (Fig. 1e). Conversely, if a SNP is associated with the ABA-hypersensitive phenotype, then the SNP allele from the ERA8 parent should be found in nearly 100% of the reads from the ERA8 bulk and nearly 0% of the WT-bulk reads. A large region of chromosome 4A showed strong enrichment for 80 EMS-induced mutations in the ERA8-like bulk sequences (Fig. 2). These 80 SNPs were in 46 genes spanning chromosome 4A (33–613 Mb). Of these, 42 SNPs in the coding region of 27 genes were located on the long arm of chromosome 4A. To our knowledge, this is the first example where an EMS-generated allele has been successfully mapped in allohexaploid wheat using bulked segregant analysis of high-throughput DNA exome-seq data.

Fig. 2
figure 2

Bulk sequence frequencies aligned to the wheat genome. a The outer track is the physical size (million bp per tick mark) of each chromosome based on the RefSeq v1.0 wheat reference genome sequence. b Bulk frequencies (BF) of the ERA8_b (purple) and WT_b (green) were aligned to the reference genome and were calculated as a fraction of the alternate allele read depth over the total number of reads. c The BF differences (BFD) were calculated from the differences between the BF of ERA8_b and the BF of WT_b (color figure online)

Full size image

Fine mapping of ERA8 in backcross and RIL populations

The ERA8 map position was further refined using recombination events from a larger backcross-population mapped relative to mutagen-induced SNPs. SNPs identified in the BSA-exome-seq analysis were scored by KASP assay (SNP_1 through SNP_30; Table S4). Based on an additional 424 random Zak/ZakERA8 BC3F2:3 lines, the ERA8 phenotype was not associated with the centromere-proximal SNP_1 to SNP_6 region, but rather was strongly associated with a more distal region on the long arm of chromosome 4A spanning SNP_20, SNP_17, and SNP_29 (Fig. 3a). SNP_20 and SNP_29 flank a 4.6 Mb interval containing 70 high confidence predicted genes in the IWGSC RefSeq v1.0 annotation (Table S7). SNP_17, located within this interval had the highest LOD score, ranging from 5.2 to 16.5 (Fig. 3a). The X5.2 population likely segregated for a weaker ERA8 phenotype because it lost more dormancy due to longer storage in the freezer before plating (Table S1). Interestingly, SNP_17 is an ERA8-linked missense mutation in the cloned preharvest sprouting tolerance gene, TaMKK3-A (Nakamura et al. 2016; Torada et al. 2016; Shorinola et al. 2017). Because the TaMKK3-A gene was not present in the assigned chromosomes of the RefSeq v1.0 wheat genome reference assembly, its relative position was estimated based on neighboring genes in the BAC-based Chinese Spring assembly (IWGSC et al. 2018; Shorinola et al. 2017).

Fig. 3
figure 3

Fine mapping ERA8 on chromosome 4A. QTL analysis was performed using mutagen-induced SNPs and germination index on ABA in the: a Zak/ZakERA8 BC3F3 from crosses X5.1 and X5.2, b Louise/ZakERA8 F6:7, and c the Otis/ZakERA8 F2:3 populations. Population sizes were 424, 207, and 108 individuals, respectively. The Louise/ZakERA8 population was phenotyped over three environments (E1, E2, and E3). Marker distances (cM) of the 4A region of interest were calculated using JoinMap. SNP positions are marked above the graph

Full size image

Based on the fine mapping, ERA8 appeared to be within the 4,597,054 bp region between SNP_20 and SNP_29 (Figs. 3a, 4a–c). This region contains 70 genes of which 3 carry EMS-induced mutations including SNP_20 in a predicted GSK1 transcription factor (TraesCS4A01G312200), SNP_17 in the TaMKK3-A gene (TraesCSU01G167000) and SNP_29 in a gibberellin 20-oxidase gene (TraesCS4A01G319100). Two additional SNPs were found in low-confidence genes in the interval: SNP_33 in a suspected pseudogene, and SNP_34 in a low confidence predicted Ser/Thr phosphatase (Table S7). Unfortunately, we were unable to develop KASP markers to clearly distinguish between alleles of SNP_33 and SNP_34. The exome-seq analysis selected only C to T and G to A transitions for mapping in order to filter out sequence differences that were not caused by EMS mutagenesis. Additional analysis detected no insertion or deletion (in/del) polymorphisms within the SNP_20 to SNP_29 interval (Fig. S3). This in/del analysis detected a large deletion on the distal arm of chromosome 2DL and a rearrangement between chromosome arms 6AL and 6BS in ERA8. However, neither of these polymorphisms were linked with increased ABA sensitivity (Fig. 2).

Fig. 4
figure 4

The ERA8 interval. a Initial BSA based on exome-seq of 32 BC3F2:3 identified the ERA8 region (gray) on chromosome 4A. The centromeric region is in red. b QTL analysis relative to EMS-induced SNPs in 424 BC3F2:3 mapped a 4.6 cM region containing the ERA8-linked interval (purple bar) spanning c the 4 unique genes containing SNP_6 through SNP_29. The number of predicted genes between SNPs is in brackets. The TaMKK3-A position is estimated based on previous mapping studies. d Lines containing recombination events in the SNP_6 to SNP_29 region were placed into groups I to VII based on whether they carried the ERA8 (purple) or Zak WT (white) SNP allele. e The ABA germination phenotype of groups I-VII is shown as a box and whisker plot where dots show the percent germination of individual BC3F2:3-4 lines. The population parents ERA8 and Zak, and BC3F2:3 line where all four SNPs had the ERA8 genotype (ERA8-g) or Zak genotype (Zak-g) are shown as controls. ERA8-like (purple) and Zak-like (green) groups were determined based on Student’s t test (p ≤ 0.05) compared to the parents. n = the number of lines per group (color figure online)

Full size image

A comparison of phenotype to the location of recombination events was used to further define the ERA8 position relative to the four mutagen-induced SNPs in the ERA8-linked region (SNPs 6, 20, 17, and 29; Fig. 4; Table S4). Homozygous lines were placed into haplotype groups (I to VII) based on the location of recombination events (Fig. 4d-e). Then, the germination phenotype of each group was placed into the ERA8-like (purple) or WT-like (green) class based on whether or not they showed a statistically significant difference from the ERA8 or Zak WT parent based on Student’s t test. The location of recombination events (Fig. 4d) was compared to the germination phenotype on ABA (Fig. 4e). Based on the germination phenotypes of Group I and Group IV, SNP_29 does not appear to determine the ERA8 phenotype. Based on Groups III, IV, and VI, the ERA8-determining mutation is most tightly linked to the SNP_20 to SNP_17 region containing 25 predicted genes. Based on Groups II, V, and VI, the region from SNP_6 to SNP_20 is not sufficient to determine the germination phenotype, although this conclusion is supported by fewer lines. All haplotypes carrying SNP_17 have an ERA8 phenotype, suggesting that either SNP_17 or a tightly linked noncoding region mutation between SNP_20 and SNP_29 cause the ERA8 phenotype. Unfortunately, our ability to fine-map ERA8 is severely limited by having only the four mutagen-induced SNPs in the ERA8 region.

ERA8 mapping by QTL analysis of a recombinant inbred line population

A Louise/ZakERA8 RIL population of 225 lines was created to confirm the ERA8 map position by QTL analysis of the ABA-hypersensitive germination phenotype. QTL analysis was performed using germination data from three environments referred to as E1, E2, and E3 (Table S8). Because dormancy and wheat ABA sensitivity vary with environment, the RIL plating assays were performed using the after-ripening time point and ABA concentration resulting in the strongest germination difference between the Louise and ZakERA8 parents (Fig. 5a). It proved difficult to find conditions that differentiated between Louise and ZakERA8 because Louise had more ABA sensitivity than the original wild-type Zak parent. When whole-seed germination was examined, ERA8 was only slightly more ABA sensitive than Louise (5.6% vs. 25.6% germination at 6 weeks of after-ripening). A stronger difference between Louise and ZakERA8 was observed when grains were cut to break coat-imposed dormancy, and ABA sensitivity examined during embryo germination (Fig. 5b). For example, in E1 with 7 weeks of after-ripening, Louise and ERA8 showed 92% and 4% germination on 5 µM ABA, respectively. In E1, the strongest difference in Louise and ERA8 germination was observed at day one of imbibition (Table S8). Both percent germination and GI were used in the QTL analysis because the percent germination of the RILs had a normal distribution, whereas the GI distribution was skewed toward the nondormant parent Louise (Table S8; Fig. 5c). While there was a stronger difference between parents in E2 and E3 using germination index (GI) than using percent germination, all of the significant QTL were identified using percent germination data.

Fig. 5
figure 5

ABA sensitivity of Louise (blue), ERA8 (purple), Zak WT (green), and the Louise/ZakERA8 RIL(gray) population. The parents (Louise, ERA8, and Zak WT) percent germination after 5 days of imbibition over an after-ripening time course in environment 1 were assayed in: a whole seeds on 2 µM ABA and b cut seeds on 0, 2, 5, and 10 µM ABA. c The RIL population was assayed across three environments with the percent germination after 2, 3, and 1 day of imbibition on 5 µM, 2 µM, and 2 µM of ABA shown, respectively. Environments: E1) F6 seed from the greenhouse 2013; E2) F7 seed from the field 2014; and E3) F7 seed from the field 2015 (color figure online)

Full size image

Both Louise and ERA8 contributed multiple QTL impacting ABA sensitivity during germination. Linkage analysis of GBS data identified 2234 single-nucleotide polymorphisms (SNPs) between Louise and ERA8 that fell into 45 linkage groups spanning all 21 chromosomes (Table S6). Seven significant QTL mapped to chromosomes 2D, 4A, 4B, and 7B (Table 1). None of the QTL was significant in multiple environments. Louise contributed higher ABA sensitivity due to the QABA.wsu-2D and QABA.wsu-4A.3 loci. In fact, QABA.wsu-2D had the highest LOD score of 5.69. ERA8 contributed higher ABA sensitivity due to QABA.wsu-4A.1, QABA.wsu-4A.2, QABA.wsu-4B, QABA.wsu-7B.1, and QABA.wsu-7B.2 loci. The two major ERA8-associated loci were on chromosome 4A, with LOD scores of 5.32 and 5.05, respectively. None of these QTL was associated with a strong ABA-hypersensitive phenotype because they were significant with only one to three days of imbibition. This suggests that Louise and ZakERA8 did not differ greatly for ABA sensitivity. The QTL identified, however, had additive effects since RILs carrying an increasing number of ABA sensitivity loci had increasing ABA sensitivity (Fig. S4). None of the QTL identified co-localized with QTL for plant height mapped in the same RIL population (Fig. S5; Table S9). However, a Louise/ZakERA8 heading date QTL co-localized with QABA.wsu-4A.9 in the combined analysis using GBS and EMS-generated SNPs.

Table 1 Significant QTL found in the Louise/ZakERA8 RIL population
Full size table

To determine whether either of the ERA8 QTLs on chromosome 4A co-localized with the ERA8 region mapped in the backcross population, QTL analysis of the Louise/ZakERA8 RILs was performed using only KASP markers for 12 mutagen-induced SNPs. QTL mapping identified one significant QTL on chromosome 4A located at the 4A centromere (QABA.wsu-4A.2, LOD 3.5) (Fig. 3b). Thus, the ERA8 phenotype was not significantly associated with SNP_17 in the Louise/ZakERA8 RIL population. There are two possible explanations for this result, either there are other loci in the Louise background that suppress the ERA8 phenotype, or both Louise and ZakERA8 contributed independent ABA-sensitive alleles of TaMKK3-A to the RIL population.

Alleles of TaMKK3-A

Based on the hypothesis that the ERA8 phenotype may have resulted from a mutation in the TaMKK3-A coding region, we characterized the TaMKK3-A alleles of Zak, ZakERA8, and Louise using KASP assays (Shorinola et al. 2017). The naturally occurring PHS-tolerant allele of TaMKK3-A/Phs1 in wheat results from a C nucleotide at position + 660 associated with a recessive dormant phenotype, whereas the PHS susceptible allele is an A nucleotide at position + 660 (Torada et al. 2016). Interestingly, KASP assays showed that Louise has the dormant TaMKK3-A-C660R allele, whereas wild-type Zak and ZakERA8 carry the susceptible TaMKK3-A-A660S allele (Fig. 6; Fig. S6; Shorinola et al., 2017). Only the ZakERA8 line carries the EMS-induced G to A transition at position +1093 of TaMKK3-A. This missense mutation results in a change from glutamic acid (E) to the basic amino acid lysine (K) at position 365 within the conserved Nuclear Transport Factor 2 (NTF2) domain. Thus, ZakERA8 and Louise carry two different TaMKK3-A alleles associated with higher dormancy and ABA sensitivity. This suggests that the dormant ERA8 phenotype failed to segregate in the Louise/ZakERA8 RIL population because both parents contributed dormant alleles of TaMKK3-A. Thus, the QTL mapped were due to contributions from other loci affecting ABA sensitivity in the population. If this hypothesis is correct, then the ABA-hypersensitive phenotype should map to the ERA8 region identified in the backcross population if ZakERA8 is crossed to a line carrying the nondormant TaMKK3-A-A660S allele, such as hard white spring Otis. QTL analysis of 108 Otis/ZakERA8 F2:3 lines detected a significant QTL associated with ABA-hypersensitive germination at SNP_20, suggesting that ERA8 is in the SNP_20 to SNP_29 region (Fig. 3c). Taken together, these results support the hypothesis that the ERA8 phenotype is due to the EMS-induced TaMKK3-A-A1093R allele.

Fig. 6
figure 6

Alleles of the TaMKK3-A gene. Increased dormancy and PHS tolerance may result either from the recessive Phs1 allele at + 660 bp or from the semi-dominant EMS-induced ERA8 allele at + 1093 bp of MKK3-A. Nucleotide and amino acid differences are shown for the S (susceptible) and R (resistant) alleles. The genotypes of ERA8, Zak, Otis, and Louise are shown at both positions

Full size image

Differentially expressed genes in ERA8

There are 70 genes in the SNP_20 to SNP_29 region. If any gene in this region resulted in an ABA-hypersensitive phenotype due to noncoding region mutations such as a promoter or intron, then we would expect to see a difference in expression levels in ERA8 versus wild-type Zak. The semi-dominant germination phenotype would then result from a change in the dosage of a regulatory gene. RNA-seq was used to identify additional ERA8 candidate genes in the ERA8-linked region based on differential expression in imbibing ZakERA8 versus wild-type seeds. Differences in seed dormancy should result in many transcriptome differences downstream of the ERA8 mutation (Barrero et al. 2009; Liu et al. 2013a; Nelson and Steber 2017). To focus on transcriptome differences possibly resulting from an ERA8 promoter or intron mutation, transcriptome analysis was performed comparing ERA8 and Zak WT seed after-ripened long enough to reduce differences in seed dormancy. ZakERA8 BC2F3 and Zak WT seeds after-ripened for 1.5 years both showed 100% germination within two days of imbibition. The mRNA was prepared from seeds imbibed for 8 h without ABA in the dark. Embryos were harvested into liquid nitrogen under a green safe light to enrich for transcripts expressed during seed germination while avoiding the induction of highly expressed photosynthesis-related transcripts that can represent as much as 50% of RNA-seq reads from photosynthetic tissue (Trick et al. 2012). The most significantly differentially regulated genes were located on chromosomes 2B (ERA8-up/WT-down) and 2D (WT-up/ERA8-down) (Table S10). On chromosome 4A, there were four WT-up/ERA8-down genes, and eight ERA8-up/WT-down genes (Table S10). Only two differentially expressed genes were within the SNP_6 to SNP_30 interval, both homologues of BED zinc finger family proteins believed to be involved in stress response (Joly-Lopez et al. 2017). However, both BED zinc finger homologues were located between SNP_29 and 30, indicating that they are not in the strongest ERA8-linked region between SNP_ 20 and SNP_29. TaMKK3-A did not show any significant change in expression in WT versus ERA8 under the conditions examined. In conclusion, the RNA-seq data did not reveal additional ERA8 candidate genes based on differential expression analysis from the time point examined.

Discussion

BSA-exome-seq as a method to map EMS-induced mutations

Here, we combined two previously published methods, exome capture and bulked segregant analysis of a backcross population, to map a mutagen-induced trait with variable expressivity in the large allohexaploid wheat genome. Exome capture was previously implemented to characterize all of the coding-region EMS-induced mutations in a wheat TILLING population (Krasileva et al. 2017). BSA was previously used to map genes relative to varietal DNA polymorphisms identified by RNA-seq (Trick et al. 2012). We combined these two approaches to map a mutation using bulked segregant analysis of exome sequence from pools of genomic DNA from WT-like and ERA8-like individuals from a BC3F2:3 population. A similar approach was used in the tetraploid Kronos TILLING population to map tall versus short EMS-induced alleles from a segregating M4 population and a leaf chlorosis phenotype in an F2 population (Mo et al. 2017; Harrington et al. 2019). In both of these studies, the phenotypic differences between wild-type and mutant individuals were relatively large compared to the more difficult-to-phenotype ERA8 mutation. BSA of a backcross population was particularly useful for mapping ERA8 since variation in dormancy is a quantitative trait in RIL populations. BSA allowed the use of a relatively small number of confidently phenotyped BC3F2:3 individuals. It was beneficial to map in an ERA8 BC3 population because the mutant plants became healthier and the ABA-hypersensitive phenotype more consistent with additional backcrosses. This may have occurred because background EMS mutations were removed by backcrossing (Uauy et al. 2017). Researchers interested in mapping mutant phenotypes with incomplete penetrance, variable expressivity, or requiring careful developmental staging may find it helpful to consider additional backcrosses before investing in BSA-exome-seq.

Comparison of ERA8 mapping by QTL analysis and BSA-exome-seq

There are both advantages and disadvantages to mapping in a backcross population as opposed to an RIL population. Creation of RIL populations may require more time than creating a population from a single backcross event since plants must be advanced by single-seed descent to at least the F5 generation. Moreover, mapping in a backcross population simplifies the genetics, decreasing the impact of other loci. For example, the backcross population had the advantage that ERA8 showed Mendelian segregation since it was the major locus contributing ABA-hypersensitive germination (Table S2; Table S5; Martinez et al. 2014). While Louise was selected as an RIL population parent based on having dormancy characteristics similar to Zak, we were blind to the precise alleles contributing to the dormancy or lack of dormancy in Louise (Fig. 5). Thus, the fact that Louise carried a dormant allele of the TaMKK3-A gene was a surprise, making Louise less than an ideal choice given the hypothesis that ERA8 is a new dormant allele of TaMKK3-A. Both the Louise and ZakERA8 parents contributed other QTL associated with higher ABA sensitivity during germination to the RIL population (Table 1; Table S9). These QTL are of potential use for PHS tolerance breeding in northwestern US wheat since pyramiding of multiple ABA-sensitive QTL provided increasing seed dormancy (Fig. S4). A drawback to the mapping in the backcross instead of RIL population was the paucity of polymorphisms in the ERA8 interval, such that ERA8 could not be mapped to a region smaller than the 4,597,054 bp between SNP_20 and SNP_29 (Table S7). A mapping population derived from a cross between near-isogenic lines containing varietal differences within one chromosome segment should have a higher density of polymorphisms (Yan et al. 2003; Trick et al. 2012). But such a population may have multiple sequence polymorphisms within a candidate gene(s), making it more challenging to identify the mutation causing the phenotype of interest (Shorinola et al. 2017). Future efforts to determine if the ERA8 phenotype is caused by the TaMKK3-A-G1093A mutation will need to identify additional recombination events within the 70 gene SNP_20 to SNP_29 interval to examine which genes are associated with the ABA-hypersensitive phenotype. This could be done by identifying additional polymorphisms by sequencing this interval in Zak and ZakERA8 (as in Harrington et al. 2019), or by using a population with a higher density of polymorphisms, such as an Otis/ZakERA8 population.

TaMKK3-A is the strongest ERA8 candidate gene

Several lines of evidence indicate that TaMKK3-A is the strongest candidate for the ERA8 gene. First, SNP_17/TaMKK3-A-G1093A is the strongest QTL mapped in the backcross population, with a LOD score as high as 16.5 (Fig. 3a). Based on examination of recombination events in the region, it does not appear that either SNP_20 or SNP_29 is required for the ERA8 phenotype (Fig. 4d,e). Second, G1093A was the only EMS-induced mutation identified in a high confidence gene within the ERA8-linked interval between SNP_20 and SNP_29 (Table S7). The only other mutations in this interval were SNP_33 in a pseudogene and SNP_34 in a low-confidence gene (TraesCS4A01G473200LC). Third, there were no genes in the SNP_20 to SNP_29 interval showing differential expression between Zak and ZakERA8 during seed imbibition, when ERA8 is expected to be expressed (Table S10). Thus, it appears unlikely that the ERA8 phenotype results from a promoter mutation in this interval leading to altered expression of a regulatory gene. Fourth, ERA8 could be mapped to the region flanked by SNP_20 and SNP_29 in the Otis/ZakERA8 population, but not in Louise/ZakERA8 RIL population (Fig. 3b,c). This appears to be an inadvertent complementation test where both Louise and ZakERA8 carry dormant alleles and Otis a nondormant allele of TaMKK3-A. Finally, MKK3 is a strong candidate for ERA8 because it is known to function in wheat and barley grain dormancy as well as in Arabidopsis ABA signaling (Danquah et al. 2015; Nakamura et al. 2016; Torada et al. 2016; Shorinola et al. 2017). Taken together, these data suggest that the G1093A mutation in TaMKK3-A is the most likely cause of the ERA8 phenotype, although we cannot discard the presence of other mutations in the 70-gene SNP_20 to SNP_29 interval. While SNP_29 is a mutation in a GA20 oxidase previously shown to impact seed dormancy in rice, this is a synonymous mutation to another codon for the amino acid cysteine (Ye et al. 2013). Future work will need to determine if the TaMKK3-A-G1093A mutation results in an ABA-hypersensitive germination phenotype when transformed into Zak or another variety carrying the PHS susceptible TaMKK3-A-A660S allele.

MKK3 protein family structure and function

MKK3 family predicted-proteins contain a conserved ATP-binding domain, a kinase domain, and an NTF2 (Nuclear Transport Factor 2)-like domain. Both of the natural dormant MKK3 variants of wheat and barley are amino acid polymorphisms in the kinase domain, lysine (K) 220 to asparagine (N) in wheat and asparagine (N) 260 instead of threonine (T) in barley (Nakamura et al. 2016; Torada et al. 2016). Interestingly, the dormant wheat MKK3-A-C660R allele is recessive, whereas the dormant barley allele is dominant. Thus, it appears that in the Triticeae a dormant phenotype can result from either gain or loss-of-function alleles of MKK3. If ERA8 is an allele of TaMKK3-A, then the fact that ERA8 is semi-dominant suggests that MKK3 is a positive regulator of ABA signaling in wheat. Previous work, however, suggests that the recessive dormant MKK3-A-C660R is the ancestral allele. In order to define whether TaMKK3-A is a positive or negative regulator, future work will need to determine whether a null allele results in increased or decreased seed dormancy and ABA sensitivity.

The semi-dominant ERA8 phenotype is linked to the G1093A mutation resulting in a change from glutamate (E) 365 to lysine (K) within the conserved NTF2-like domain (Danquah et al. 2015). Since E365K is a change from a negatively to positively charged amino acid, this change may alter protein function. The NTF2 domain is unique to the MKK3 family of map kinases (Jiang and Chu 2018) that appears to be involved in protein–protein interaction with downstream MKK3-activated group-C MAP-kinases, MPK7, 1, and 3 (Dóczi et al. 2007). A semi-dominant phenotype might result from increased affinity for one of the wheat MKK3-target proteins.

TaMKK3-A is expressed in many tissues where ABA signaling is expected to occur such as roots, leaves, seedlings, and grain (Torada et al. 2016; Ramírez-González et al. 2018). But the transcript levels are highest in maturing embryos, when dormancy and desiccation tolerance are established. The ERA8 mutation alters ABA sensitivity in seeds but not during leaf stomatal closure (Schramm 2010). In contrast, the ABA-hypersensitive Warm4 (Wheat ABA-responsive mutant 4) mutant of wheat exhibited altered ABA sensitivity both during seed germination and in vegetative tissues (Schramm et al. 2010). This observation suggests that an ERA8 mutation in the MKK3 NTF2 domain may increase binding to a target protein involved in seed but not leaf ABA signaling.

The potential role of MKK3 signaling in controlling wheat grain dormancy and germination

If the ERA8 phenotype truly results from the TaMKK3-A-G1093A mutation, then future work will need to further investigate its role in wheat ABA signaling and seed dormancy. The recovery of MKK3 alleles providing increased dormancy and PHS tolerance in wheat and barley based both on natural variation and chemical mutagenesis would suggest that MKK3 is a key regulator of these responses in cereals. This likely has broad relevance in plants since the MKK3 gene family is highly conserved in eukaryotes, including Arabidopsis (AtMKK3), barley (HvMKK3), rice (OsMKK3), and tobacco (NtNPK2) (Shibata et al. 1995; Xiong and Yang 2003; Hamel et al. 2006; Nakamura et al. 2016; Torada et al. 2016). The mitogen-activated kinase signaling cascade was initially characterized based on its role in controlling mitosis and cell growth in yeast and humans (reviewed by Rodriguez et al. 2010). Identification of ABA-activated mitogen-activated protein kinases (MAPKs) in the barley aleurone system first suggested a role for MAPKs in ABA regulation of cereal grain germination (Knetsch et al. 1996). MAP kinases phosphorylate proteins in a wide range of signaling pathways. The Arabidopsis MAP kinase MKK3 performs jasmonic acid (JA)-stimulated phosphorylation of the MAP kinase MPK6, which in turn negatively regulates the MYC2/JIN1 transcription factor, a positive regulator of root growth (Takahashi et al. 2007). This is interesting given that JA hormone appears to stimulate germination in barley and in wheat (Barrero et al. 2009; Tuttle et al. 2015; Xu et al. 2016; Martinez et al. 2016). MKK3 and MPK6 also have roles in ABA signaling and seed dormancy in Arabidopsis, together with other map kinases MPK3, MPK17, MPK18, MKK1, MKK9, MPK1, MPK2, MPK7, and MPK14 (Liu 2012; Danquah et al. 2015). Loss of AtMKK3 function in Arabidopsis led to ABA-hypersensitive seed germination (Danquah et al. 2015). Future work will need to identify TaMKK3-A target proteins, and to define domain functions through structure–function analyses. The ERA8 allele could be a useful tool for analysis of protein–protein interactions in wheat. Such analyses would need to examine MKK3 domain functions in the context of ABA responses including seed germination, root elongation, and stomatal closure.

Breeding for PHS tolerance with ERA8

Even if ERA8 is not an allele of TaMKK3, the identification of SNP_17 as a strong ERA8-linked marker will have utility for deploying ERA8 in breeding programs. Before crossing to ERA8, it will be important to determine if the starting MKK3 allele is the susceptible A660 or the tolerant C660, in order to know which dormancy alleles will be segregating in the cross. Deploying an EMS-induced mutation is advantageous since the marker associated with the ERA8 allele will be polymorphic in most outcrosses. The exome-seq data for ERA8 revealed a large deletion on chromosome arm 2DL and an insertion on 6BS, likely induced by mutagenesis. To avoid possible negative effects of these in-dels, it would be wise to select against them during the breeding process. Since ERA8 was selected in a soft white wheat background, this gene can be used to breed for PHS resistance regardless of seed coat color. Future breeding efforts will need to move ERA8 into a hard wheat background to produce bread products versus soft wheat products. Previous work established that ERA8 has sufficient dormancy to prevent preharvest sprouting at maturity, but loses dormancy rapidly enough through 8–10 weeks of after-ripening to allow fall planting of winter wheat without a strong negative impact on emergence or yield (Martinez et al. 2014, 2016). Farmers would need to be advised to wait 4–9 weeks before replanting ERA8-carrying winter wheat. ERA8 has also been shown to maintain higher seed dormancy than Zak at temperatures as low as 20 °C (Martinez et al. 2016). Maintaining dormancy at lower temperatures is beneficial since cooler conditions exacerbate PHS.