A structured mutant population for forward and reverse genetics in Barley (Hordeum vulgare L.)

Authors


(fax (44) 1382 568587; e-mail rwaugh@scri.sari.ac.uk).

Summary

Two large-scale ethylmethanesulfonate (EMS) mutant populations from barley (Hordeum vulgare L.) cv. Optic have been developed to promote both forward and reverse genetics in this crop. Leaf material and seed from approximately 20 000 M2 plants were individually harvested, freeze-dried and archived. DNA was isolated from 9216 plants from the 20 and 30 mm EMS treatments and assembled into 1152 eight-plant pools. To facilitate PCR-based mutation scanning an approach has been employed that combines cleavage of heteroduplexes using the Cel nuclease (Cel I), post-cleavage intercalating dye labeling and the subsequent detection of cleaved products on a Transgenomic WAVE-HS. The populations were evaluated by screening for induced mutations in two genes of interest and the induced mutations were validated by sequence analysis. To enhance the screening process, 12–16 M3 progeny from each of the M2 plants were assessed for visible phenotypes and the data entered into a web accessible database (http://bioinf.scri.sari.ac.uk/distilling/distilling.html).

Introduction

The ability to experimentally validate the function of gene sequences derived from large-scale expressed sequence tag (EST) programs has not kept pace with the rate at which they are discovered. As a result, there is an increasing abundance of sequences which, by electronic annotation, show no homology to any previously characterized genes or proteins and so no putative function can be assigned. While the use of Arabidopsis has greatly assisted in closing this gap from gene identification to gene function (Anonymous, 2000), as more research is conducted, its limited utility as a global model organism for determining gene function is becoming apparent. For example, the inability of Arabidopsis to form root nodules has prompted the development of functional genomic resources in legumes (Perry et al., 2003). Its limitations are further illustrated by the lack of conservation of the vernalization requirement pathway between Arabidopsis and wheat (Triticum monococcum) as well as barley (Yan et al., 2003, 2004). The fact that many traits, such as grain texture (Gautier et al., 2000) or inflorescence architecture (Komatsu et al., 2001), can only be studied within the species of interest has stressed the need for functional genomics resources in a variety of crop plants.

Barley is a large-genome monocotyledon (5300 Mbp) and a true diploid member of the Triticeae, which includes both wheat and rye. A long history of genetic studies has resulted in the generation of multiple genetic maps and an extensive assortment of molecular markers (Franckowiak, 1997; Kleinhofs and Graner, 2001; Kleinhofs and Han, 2001; Ramsay et al., 2000). An array of genomics-based resources has now also been developed and is expediting the identification of candidate genes for a number of traits (Close et al., 2004; Yu et al., 2000). Consequently, rapid and high-throughput methods for the confirmation and validation of gene function by targeted gene inactivation are currently established priorities. However, the lack of facile and high-throughput transformation systems for the Triticeae species coupled with their large genome size is hindering the development of resources, such as transposon or T-DNA tagging populations, which have greatly assisted functional genomics in model species (Feldmann et al., 1989; Hirochika, 2001; Jeon et al., 2000; Marsch-Martinez et al., 2002; Nakagawa et al., 2000; Weigel et al., 2000).

A long-recognized large-scale approach for achieving random gene inactivation is to induce mutations in a population of plants using chemical or physical mutagens. This approach is of particular interest because it is amenable to most plant species regardless of their transformability. When coupled with sensitive methods for the detection of ‘aberrant’ DNA fragments in complex PCR-derived mixtures, this previously random approach becomes amenable to the identification of mutations in targeted genes by reverse genetics. Recently, strategies deploying chemical mutagens to induce point mutations in DNA have been described in Arabidopsis and the acronym TILLING (Targeted Induced Local Lesions IN Genomes) has been coined (Colbert et al., 2001; McCallum et al., 2000a,b). TILLING is based on a simple PCR-screen coupled with variants of heteroduplex analysis or mismatch cleavage to detect mutations in a specified target region. Chemical mutagenesis has a number of inherent attractions such as the ability to use different mutagens, change mutagen doses and to easily scale the size of the mutagenesis procedure. Importantly, chemical mutagens generate an allelic series at any target locus, resulting in a change of function, reduced activity or specificity or a knockout mutation (Henikoff and Comai, 2003; Koornneef et al., 1982), that can be very valuable when attempting to assign function to a given gene.

Here, the generation of two structured chemically mutagenized populations of the barley cv. Optic is described, which will allow both forward genetics and genome-wide reverse genetics in a member of the Triticeae. In addition, an alternative mutation scanning method not relying upon fluorescent primers has been implemented. This is based on the double-stranded cleavage of heteroduplex molecules with Cel nuclease (Cel I) (Kulinski et al., 2000; Oleykowski et al., 1998; Qiu et al., 2004) with subsequent fragment detection on a Transgenomic WAVE High Sensitivity denaturing High Performance Liquid Chromatography (WAVE-HS dHPLC) system. As an illustration of the use of this resource and detection system, we describe the discovery of induced mutations in two genes of interest in our research program. Furthermore, to aid forward genetic screening, M3 visible mutant phenotypes have been documented in a web accessible database. The general utility and attributes of the population are discussed.

Results

Assembly of reverse genetics populations

Approximately 45 000 M1 barley plants (cv. Optic) were mutagenized with 20 and 30 mm doses of ethylmethanesulfonate (EMS) giving an M2 lethality of approximately 30 and 50%, respectively. Both doses of EMS yielded an acceptable level of M2 sterility (<10%). A 40 mm dose of EMS was investigated, but resulted in excessive M2 lethality and sterility (70 and 80%, respectively). From the seed setting M1, a single progeny was taken forward to the M2 generation. Leaf tissue was collected from approximately 20 000 surviving M2 plants at the seedling stage and grain harvested from each individual plant. Approximately 4600 random M2 individuals from each of the 20 and 30 mm EMS treatments were selected to construct two separate reverse genetics populations. The M2 genomic DNA was isolated and arrayed into twelve 96-well microtiter plates, with each well containing DNA pooled from eight individual plants (6 × 96-well plates for each of the populations). An overview of the process is given in Figure 1.

Figure 1.

Creation of a structured mutant population.
M0 seed was mutated, propagated, and a single M2 seed was taken forward from each chimeric M1 plant. Genomic DNA was isolated from each M2 plant and its M3 seed was archived for sibling analysis and phenotyping. ‘Reverse genetics’ screening was performed on the M2 DNA pools and ‘Forward genetics’ screens on the M3 families.

Cel nuclease/Transgenomic WAVE-HS mutation scanning method

Cel nuclease naturally possesses exo-nucleolytic activity that not only attacks the ends of double- stranded DNA, but also cleaves the fluorophores from the ends of fluorescently labeled amplicons (Henikoff and Comai, 2003). This is a major limiting factor in the detection of unknown signal nucleotide polymorphisms (SNPs) in deep pooling strategies. Thus, although the enzyme may cleave more heteroduplexes with a longer incubation or increased enzyme concentration, the cleaved products will not necessarily be visualized because of the simultaneous removal of the fluorophores. To circumvent this exo-nulceolytic issue, we employed a post-digestion intercalating dye labeling system coupled with detection by dHPLC attached to a high sensitivity fluorescent detector (Bahrami et al., 2002; Qiu et al., 2004). A series of test pools of barley genomic DNA of varying depths, each containing a single reference SNP from different genes (Hin-a, Mlo) were amplified with gene-specific primers and heteroduplexes induced. The heteroduplexes were double-strand digested with Cel I and the fragments were analyzed on a Transgenomic WAVE-HS. The detection of mutant fragments at a pool depth of 16–32 alleles was routinely obtained (Figure 2).

Figure 2.

Chromatograms illustrating the detection of a single mutation (mismatch) in a pool of 16 alleles (a) and 32 alleles (b). Cleaved products of 100 bp (x) and 300 bp (y), are easily visualized when overlaid with wild type uncut fragment, 400 bp (z).

Test screening of the barley reverse genetics population

To demonstrate the utility of the population for reverse genetics, we screened for mutations in Hordoindoline-a (Hin-a) and the Hordeum vulgare Floral Organ Regulator-1 (HvFor1) ortholog. Primers were designed to yield a 420 bp amplicon and a 913 bp amplicon for Hin-a and HvFor1, respectively. For ease of analysis of this proof of principle study we maintained a pool depth of eight plants. The 576 pools from both populations were PCR-amplified with the Hin-a and HvFor1 primers, heteroduplexes induced, cleaved with Cel I and the products analyzed on the Transgenomic WAVE-HS. Aberrant chromatograms were identified using the integrated analysis package of the Transgenomic Navigator Software. This initial screen yielded four positive pools for Hin-a and six positive pools for HvFor1. A single plant was identified from each of the positive pools, for both genes, that carried a putatively induced mutation. Subsequently, the Hin-a and HvFor1 genes were re-amplified four independent times and sequenced to 8× coverage in each of their corresponding putative mutant plants. The six induced alleles (all transition events) of the HvFor1 gene were confirmed and the precise mutations are illustrated in Table 1 along with the four induced alleles (two transition and two transversion events) of the Hin-a gene. To confirm the legitimacy of the two transversion events in Hin-a, the induced alleles were compared with data from an exhaustive SNP discovery project of the Hordoindoline gene family (74 cultivars, 15 landraces and 34 wild accessions) that yielded 28 distinct haplotypes for Hordonindoline-a (K.S. Caldwell, SCRI, Dundee, UK, personal communication). The induced alleles did not match any previously discovered SNPs within the Hin-a gene and no other SNPs were identified within a 1.4 kbp region surrounding the Hin-a coding region in the putative mutants. We therefore considered these SNPs bona fide induced mutations. Transversion events are generally considered a rare consequence of EMS mutagenesis. Indeed all other mutations identified within the population (including data not shown from two further genes) have been transition events.

Table 1.  Induced mutations in the coding sequences of the Hin-a and HvFor1 genes
GeneMutant lineNucleotideNucleotide changeAA locationAA change
  1. The nucleotide location is given from the first nucleotide of the cv. Optic coding sequence. All mutations were confirmed in the M3 generation. AA indicates amino acid, and the amino acid location is given from the Met in the pre-processed cv. Optic protein.

Hin-aEMS_R0010_28364G to A122Val to Met
Hin-aEMS_R0012_04182A to T61Lys to Met
Hin-aEMS_R0173_3145C to A15Ser to Arg
Hin-aEMS_R0184_18333C to T111Silent
HvFor1EMS_R0015_53278C to T93Ala to Val
HvFor1EMS_R0015_82619C to T207Pro to Ser
HvFor1EMS_R0038_32466A to G156Lys to Glu
HvFor1EMS_R0184_23315C to T105Silent
HvFor1EMS_R0186_55576G to A192Silent
HvFor1EMS_R0192_86615G to A205Silent

Forward genetics screening of M3 families and web accessible database

These structured mutant populations are also a valuable resource for forward genetics screens. To demonstrate their utility, we propagated 12–16 M3 individuals for each M2 parent that was used in the reverse genetics populations. Visible phenotypes were scored regularly during the growing season including key developmental time points. All information was entered directly into a portable web accessible database. A summary of this phenotypic screen is illustrated in Table 2. A visible phenotype was recorded for over 20% of the M3 families. As expected, the lines treated with 30 mm EMS, yielded a slightly higher percentage of the mutant phenotypes scored. Some examples of mutant phenotypes visualized during the screening are given in Figure 3. All information, including details of the phenotypes, is accessible via the following URL: http://bioinf.scri.sari.ac.uk/distilling/distilling.html and seed is available on request. The description of phenotypic categories is as follows: the leaf color phenotypic category included the following descriptors: dark green, pale green, yellow, albino, necrotic, glossy, matte, variegated and sectoring. Leaf appearance phenotypes included: corrugated, curly, twisted, serrated, auricle-less, displaced ligule, liguleless, smooth, hairy, more erect, and less erect. Leaf size phenotypes included: width, length, and thickness. Shoot appearance phenotypes included: fewer tillers, more tillers, more erect, more prostrate, node quantity, internode length, bowed, branched, fragile, and thickness. Floret appearance phenotypes included any alteration to the normal floret structure. Spike color descriptors are same as leaf color descriptors. Spike appearance phenotypes include the following: filled laterals (six rows), filled laterals, missing laterals, dense spike, less dense spike, short spike, long spike, brittle rachis, sterility, and floret quantity. Stature was divided into three classifications: taller, shorter, and extremely short. Plant development phenotypes included: seedling emergence time, flowering time, and time of senescence. Grain appearance phenotypes included overall size and shape.

Table 2.  Summary of phenotypic screening
PhenotypeNo. of mutants% Population
20 mm30 mm20 mm30 mm
  1. Phenotypes are grouped together based on their major classification. Statistics are represented as a percentage of plants from each respective reverse genetics population. All phenotypes were scored in reference to the parent cultivar (cv. Optic). Details of the phenotypes contributing to each major classification can be seen in the results section and further information can be obtained from the following URL: http://bioinf.scri.sari.ac.uk/distilling/distilling.html.

Leaf color2755146.4411.97
Leaf appearance14390.330.91
Leaf size44801.031.86
Leaf quantity230.050.07
Plant stature1905644.4513.14
Plant development3794198.889.76
Shoot appearance1202602.816.06
Shoot quantity13140.300.33
Floret appearance11350.260.82
Spike color2100.050.23
Spike appearance361060.842.47
Grain appearance290.050.21
Kernel color020.000.05
Figure 3.

Examples of phenotypes observed during M3 phenotypic screening.
Several spike phenotypes observed included; multiple florets (a) and branched spikes (b). Developmental mutants included phenotypes such as severe dwarfing with delayed flowering (c). Some plants illustrated very complex phenotypes such as: narrow curly matte leaves with thin prostrate tillers (d).

Discussion

When constructing a comprehensive reverse genetics population in any plant species, it ideally should be small enough to combat the high cost and labor-intensive screening process, yet of sufficient size to ensure a high success rate for the detection of mutations in the genes of interest. Originally the creation of a population based on the induction of deletions using chemical mutagens such as di-epoxy butane (DEB) (H. Leung, International Rice Research Institute, Los Baños, Laguna, Philippines, personal communication) was considered. However, the data from a series of pilot studies (data not shown) revealed that the number of phenotypically visible mutations induced by DEB in surviving M2 families was low in barley compared with EMS, and as a result the number of plants required for a population would be high. Similar conclusions were drawn from pilot studies using gamma irradiation and sodium azide (data not shown). The feasibility of creating a fast neutron deletion population (Delete-a-gene®) was not assessed, although, on the surface also appeared attractive (e.g. no transformation necessary, inexpensive to create the mutants, inexpensive to screen for induced mutations and non-transgenic). However, the low number of lesions per genome would require a population size estimated to be in excess of 100 000 plants for comprehensive coverage (>95% probability of detecting a mutation in a target gene) if the mutation rate is independent of genome size (Li et al., 2001, 2002). In addition, without the complete genome sequence of barley available and its large content of repetitive DNA, designing robust scanning windows for genes of interest could prove difficult.

The one reverse-genetics approach that appears to be relatively independent of genome size and requires a comparatively small amount of plants is EMS mutagenesis coupled with sensitive PCR-based mutation detection (McCallum et al., 2000a,b). Several attributes make this outline methodology attractive for large genome crop species. The first is that the mutation frequency of EMS appears to be high in all plant species and based on historical accounts appears to be independent of genome size (Greene et al., 2003; Henikoff and Comai, 2003). This translates into smaller population sizes. The second is the short development time required. Having the facilities, many medium-sized (∼10–15K plants) M1 populations of varying doses can be grown simultaneously and the most appropriate mutagen doses taken forward to the M2 generation. At least two generations a year can be achieved in most large genome crop species. In practical terms, this makes it possible to generate a working reverse genetics population in approximately one and a half years from the time of starting, without having to do any plant transformation, regeneration, or selection of transformants. As the EMS mutants are non-transgenic, subsequent generations can be grown under field conditions, without restrictions, for phenotypic analysis and advantageous alleles can be immediately incorporated into a fast-paced molecular breeding program using the characterized induced mutations as markers for selection.

Most analysis of induced or naturally occurring variation focuses on characterizing non-synonymous changes or insertion/deletions. While the majority of causal mutations will either be from amino acid substitutions or truncation events, there are instances where synonymous base changes or silent events can induce a significant phenotypic effect. For example, perturbing an intron splice recognition site or splicing enhancer, intronic or exonic, could impact upon spliceosome assembly and prevent correct or efficient splicing of the immature mRNA (Brown, 1996; Cartegni et al., 2002). The result may be a less abundant, dysfunctional or truncated protein. In addition, there has been a considerable volume of work published recently on the control of gene expression by microRNAs (miRNA). This mechanism has been shown in pathways controlling leaf morphogenesis (Palatnik et al., 2003) as well as flowering time and floral organ identity (Aukerman and Sakai, 2003). Base changes within the miRNA binding site, possibly even synonymous changes, could have dramatic effects on phenotype. While these instances may be infrequent, they should be considered during the accurate and exhaustive annotation of gene sequences prior to and after mutational analysis.

In addition to developing and using this reverse genetics population in barley, we implemented a mutation scanning method based on the double-stranded cutting of mismatches with subsequent fluorescent labeling using an intercalating dye. This has been achieved by combining the Cel I cleavage assay (Kulinski et al., 2000; Oleykowski et al., 1998) from the Transgenomic Surveyor® kit (Transgenomic Inc., Omaha, NE, USA) with analysis on a Transgenomic WAVE-HS dHPLC fluorescent system (Qiu et al., 2004). It is recognized that the initial outlay required to purchase the dHPLC system may preclude many laboratories from adopting this approach. Therefore, gel or capillary electrophoresis-based methods continue to offer a practical alternative to the method described here. Furthermore, novel developments in mutation scanning technology will both increase the throughput and decrease the cost of rapid mutation discovery in mutagenized plant and other populations.

Based on the data obtained from screening Hin-a, HvFor1 and two other candidate genes (data not shown), the induced mutation frequency is approximately one mutation every one million base pairs or ∼5000 mutations per genome. This is a significantly lower frequency than observed in Arabidopsis. Our attempts to increase this frequency using a higher dose of EMS (40 mm) were unsuccessful because of high levels of lethality and sterility observed in the M2 lines. It is possible that an optimal dose of EMS that maximizes mutation frequency lies somewhere between 30 and 40 mm and we are creating additional M2 populations to investigate this. However, one other possibility for the large difference in mutation frequency between Arabidopsis and barley lies in their genome histories. Barley is currently regarded as a true diploid. In contrast, Arabidopsis possesses numerous segmental duplications that may provide significant potential for functional redundancy. It is therefore possible that the Arabidopsis genome can tolerate a higher mutation load than that of barley. A high level of functional redundancy has proven to be relevant even in Saccharomyces cerevisae where compensation of null mutations by duplicate genes has been clearly demonstrated (Gu et al., 2003).

In conclusion, we have created publicly available genome-wide reverse genetics populations of barley based on EMS mutagenesis and a mutation detection strategy exploiting Cel I cleavage of complex PCR amplicons followed by fragment detection using a Transgenomic WAVE-HS dHPLC. Although the estimated mutation frequency is significantly less than Arabidopsis, we have been able to identify several mutations in all of the four genes we have investigated to date. Here we have demonstrated the utility of the population for reverse genetics by identifying induced mutations in the Hin-a and HvFor1 genes. We propose that these populations will be a great asset for functional genomics research in the large genome cereals and will complement the insertional mutagenesis populations that are currently being developed elsewhere (Koprek et al., 2000, 2001). The addition of web accessible M3 phenotypic data will enhance the utility of the populations and will facilitate the identification of phenotypes targeted for gene isolation via forward genetics.

Experimental procedures

Plant materials

The barley (Hordeum vulgare) cv. Optic was used for all mutagenesis work and the development of the populations. Optic is a spring two-rowed Northern European malting variety with the sdw1 semi-dwarfing allele and the Mla12 allele. Plants were grown in the greenhouse under 20°C 16 h days and 15°C 8 h nights and were allowed to self-pollinate.

Ethylmethanesulfonate mutagenesis

Batches of 100 g of seed were imbibed in 125 ml of dH20 for 4 h at room temperature with gentle shaking (100 rpm). The 125 ml of dH20 was changed every hour for the duration of imbibition. The seeds were then incubated in 125 ml of varying concentrations (10, 20, 30 mm) of EMS for 16 h at room temperature with gentle shaking. The seeds were treated with two changes of 200 ml of 100 mm sodium thiosulfate for 10 min at room temperature then rinsed with 400 ml of dH20 for 30 min at room temperature with gentle shaking. The seeds were air-dried overnight on absorbent pads prior to planting.

Genomic DNA isolation

Genomic DNA was isolated using the following method and scaled appropriately: 50 mg of lyophilized barley leaf tissue was ground to a fine powder and 600 μl of urea extraction buffer [urea 42% (w/v), 0.35 m NaCl, 0.05 m Tris–HCl (pH 8.0), 0.02 m EDTA, Sarkosyl (n-lauroyl-sarcosine) 1% (w/v)] was added, incubated at 37°C for 30 min, 500 μl of phenol:chloroform:isoamyl alcohol (25:24:1) was added to each sample, mixed by vortexing, and spun for 15 min at 14 000 g. Subsequently, 400 μl of the supernatant was transferred to a fresh tube and treated with RNase-A (Amersham Biosciences, Piscataway, NJ, USA) prior to precipitation with an equal volume of isopropanol and 1/10 volume 3 m sodium acetate (pH 5.2). After pulse centrifugation, pelleted DNA was washed with 250 μl of 70% ethanol and dried at room temperature for 2 h prior to resuspension in TE (10 mm Tris–HCl, pH 8.0, 0.1 mm EDTA).

Primers, PCR, and sequencing analysis

Primers were designed using Primer3 (http://frodo.wi.mit.edu//cgi-bin/primer3/primer3_www.cgi) based on the cv. Morex genomic sequence (obtained from K.S. Caldwell, personal communication) for Hin-a and an EST contig consensus sequence from the TIGR Gene Indices for HvFor1. The following primer sequences were used for PCR and sequencing of the population and putative mutants: Hin-a_L ggtctgcttgctttggtagc, Hin-a_R aatagtgctggggatgttgc, HvFor1_F aggcccgcactcaacacttc, HvFor1_R tgaggtccaggtaggtgagc. PCR was performed using Qiagen Hot Start Taq according to manufacturer's instructions (Qiagen, West Sussex, UK). Cycle sequencing was performed using ABI Big Dye V2 (Applied Biosystems, Foster City, CA, USA). Reactions were cleaned up according to manufacturer's directions and subsequently analyzed on either an ABI-377 or ABI-3700 (Applied Biosystems). Sequences were analyzed using SEQUENCHERTM 4.1 (Gene Codes, Ann Arbor, MI, USA)

Cel nuclease mismatch cleavage assay

Cel nuclease assay was performed according to the manufacturer's directions for the Surveyor® Mutation Detection Kit for WAVE HS systems (Transgenomic Inc, Omaha, NE, USA). Briefly, 10 μl of a PCR product were added to 42 μl of dH20, 6 μl of 10X Surveyor reaction buffer, 1 μl of Cel nuclease-W and 1 μl of Enhancer-W. The reaction was incubated at 42°C for 20 min and stopped by adding 6 μl of the Surveyor Stop solution. The entire 60 μl reaction was injected into a Transgenomic WAVE-HS dHPLC under non-denaturing conditions (50°C). The WAVE Optimized HS Staining Solution I (Transgenomic Inc.) was mixed automatically with samples (0.1 ml min−1) after elution from the DNASep column (Transgenomic Inc.) using the HSX-3500 dye pump. Subsequent chromatograms were analyzed in the analysis module of the Navigator software (Transgenomic Inc.).

Forward genetics screening

Twelve to 16 M3 progeny were planted under field conditions for each of the M2 lines of the 20 and 30 mm EMS reverse genetics populations. Plants were scored for visible phenotypes every 2–4 weeks from time of planting, awn emergence, time of heading, and final maturity. All information was collected directly into portable databases built on MySQL, Apache and Perl. All phenotypes were scored in reference to the parent cultivar (cv. Optic). All information from the phenotypic screen is accessible through a graphical interface at the following URL: http://bioinf.scri.sari.ac.uk/distilling/distilling.html.

Acknowledgements

The authors are indebted to a large number of individuals for their considerable assistance in harvesting, processing of plant material and for their general help with the work presented here. These include: Linda Cardle, Luke Ramsay, Arnis Druka, Hui Liu, Peter Hedley, Sharon Mudie, Sandie Williamson, Euan Caldwell, Nils Rostoks, Ingo Hein, Joanne Russell, Bill Thomas, Richard Keith, Ilza Druka, Irene Tierney, Derek Matthew, Graeme Dargie, Fiona Napier, John Marshall, Karen McLean and Malcolm Macaulay. We also thank Tony Yeung for his contribution of Cel I earlier in our research. We thank the following from Transgenomic for their assistance: Craig Parker, Ben Legendre, and Gary Gerard. We thank Katherine S. Caldwell and Wayne Powell for access to the Hordoindoline-a haplotype data prior to publication. Finally, we thank Gordon G. Simpson and Hajime Sakai for their critical reading of this manuscript. This work was supported by the Scottish Executive Environment and Rural Affairs Department through BBSRC grant no. IGD12397 awarded to Robbie Waugh and by the University of Minnesota Graduate School (McKnight Land Grant Professorship Program) for Gary J. Muehlbauer.

Ancillary