In this study, we used 14 barley spring cultivars (Supplementary Table S2) and eight historic ert-k mutant lines (Supplementary Table S3) obtained from the Nordic Genetic Resource Center (NordGen), Alnarp, Sweden (www.nordgen.org). These lines and F₁ plants obtained from allelism crosses were grown in greenhouses at the Department of Biology, Lund University, Sweden. Plants were grown in soil with article number 744704 from SW Horto (www.swhorto.se) in 5-L pots. Greenhouse conditions were set to 16-h light/8-h dark cycles and temperature 20°C during the day and 16°C during the night. All statistical data presented for phenotypic traits are based on measurement of 6–11 plants of each line.

Genomic DNA for PCR reactions was extracted from fresh leaves using the REDExtract-N-AmpTM Plant PCR Kit (Sigma-Aldrich, St. Louis, MO, USA) according to the manufacturer’s instructions. Sanger sequencing were performed by Eurofins Genomics, Germany. Used primers are shown in Supplementary Table S4.

PCR were performed by initial denaturation at 94°C/3 min, followed by 35 cycles of 94°C/45 s, 56°C–62°C/45 s and 72°C/60–90 s, with a final extension step of 72°C/10 min.

Purification of PCR products was done by using Illustra ExoProStar 1-Step (Cytiva, Marlborough, MA, USA), following manufacturer’s protocol.

Found mutations were confirmed by Sanger sequencing of at least two independent reactions.

In order to analyze whether the eight ert-k mutants are transcription deficient for the Ert-k candidate gene, RT-PCR (not qRT-PCR) was performed. Plants were grown in soil and 100 mg material from a single leaf was harvested and immediately frozen in liquid nitrogen. Porcelain mortar was used to homogenize leaf material under liquid nitrogen. Total RNA was isolated using Trizol reagent (Invitrogen Carlsbad, CA, USA) according to the manufacturer’s instructions. Residual DNA was removed by DNAse I treatment (Thermo Fisher Scientific, USA). A 500 ng of total RNA was converted to single-stranded cDNA by using RevertAid First Strand cDNA synthesis kit (Thermo Fisher Scientific, USA) primed with (dT)15 in a 20 µL reaction volume (Stuart et al., 2021). Primers were designed to amplify a 1006 bp cDNA fragment. (Supplementary Table S4).

The ert-k.32 mutant was used as male in each cross. Mutants ert-k.76, ert-k.93, ert-k.302, ert-k.309, ert-k.435, ert-k.459, ert-k.477 and their mother cultivars Bonus and Foma were used as female. Pollination was performed 3 days after emasculation. F₁ plants were phenotyped and genotyped. For genotyping, specific primers were designed that could amplify either the wild type allele or the mutant allele (Supplementary Table S4). 6–11 F₁ plants of each cross were evaluated. The statistical analyses were based on comparisons between a mutant and its mother cultivar, or comparisons between a mutant crossed to ert-k.32 and its mother cultivar crossed to ert-k.32.

Genomic DNA for whole genome sequencing was extracted by a modified CTAB protocol (Doyle, 1991) as described in (Stuart et al., 2021). Genomic DNA was sent to the Earlham Institute (Norwich, UK, www.earlham.ac.uk) where DNA libraries were prepared and where DNA sequencing was performed on an Illumina HiSeq4000.150 bp paired-end reads were obtained.

BWA mem (version 0.7.17) was used to align reads to the barley MorexV3 reference genome (Li and Durbin, 2010; Mascher et al., 2021). Samtools fixmate (samtools version 1.10) was used to fill in mate coordinates. Aligned reads were further processed to remove PCR duplicates by running Samtools markdup (Li et al., 2009). To get vcf files, bcftools mpileup (bcftools version 1.17) was used with the following filtering options: q 60 -Q 30 -D (where -q is mapping quality for the alignment, -Q is minimum base quality, -D is instruction to run the BAQ algorithm not only in problematic regions, but on all reads) (Li, 2011). To identify alleles observed in a mutant sample and not found in the control (mother cultivars Bonus or Foma), bcftools in combination with the “contrast” plugin was used. Bcftools view was used to exclude sites with heterozygous genotypes (-g ^het), and to exclude sites with the reference genotype (-e ‘GT = "ref"'). Functional annotation of variants was performed with Ensembl Variant Effect Predictor (McLaren et al., 2016).

The Ert-k locus is represented by eight different alleles (Franckowiak and Lundqvist, 2012), which have been generated through treatment by X-rays, gamma-rays, neutrons and ethyl methanesulfonate of the cultivars Bonus and Foma (Supplementary Table S3). Early mapping experiments demonstrated that Ert-k is located on chromosome 6H (https://bgs.nordgen.org/). More recently, mutant ert-k.32 was backcrossed to the cultivar Bowman to create a near-isogenic line; BW314. By genotyping BW314 with 3,072 single nucleotide polymorphisms markers (SNPs), the introgression region was defined by 14 SNP markers spanning 22.3 cM (Figure 1) (Druka et al., 2011).

In our previous research, we further fine mapped the Ert-k locus by using three mapping populations; one double-haploid population derived from BW314 (ert-k.32) crossed to cultivar Quench, and F₂-mapping populations derived from crosses of BW314 to Bowman and Quench. The analysis confirmed the centromeric location of Ert-k on barley chromosome 6H and mapped Ert-k to a 15.7 cM region between markers 2_0291 and 1_0040 (Skov Kristensen et al., 2016). With help of the barley reference genome sequence (Mascher et al., 2021), it is now possible to translate the genetic map to the physical map. The distance between 2_0291 and 1_0040 is more than 400 Mbp comprising 73% of chromosome 6H (Figure 1). There are 1,986 high-confident genes between markers 2_0291 and 1_0040. Since recombination events are less frequent in centromeric regions, fine mapping is problematic. We therefore decided to search for the Ert-k gene by whole genome sequencing of allelic ert-k mutants. Among the eight ert-k mutants, we selected the five mutants which had been induced by X-rays, gamma rays and ethyl methanesulfonate (Supplementary Table S3). In total, severe homozygous mutations were found in seven genes on chromosome 6H of the five mutants compared to their respective mother cultivar (Table 1). These mutations are likely to be caused by the mutagenic treatment and cause frameshifts or missense mutations changing amino-acid residues in their respective protein.

Since the ert-k mutants are supposed to be allelic, we especially looked for genes where all five mutants would have a mutation. Such gene was not found. However, the gene HORVU.MOREX.r3.6HG0574880 got our attention since this gene was mutated in the coding region of three of the five analyzed mutants (Table 1). The gene encodes an alpha/beta hydrolase of 399 amino-acid residues, has 6 exons and contains a fermentation-respiration switch protein FrsA domain (Figure 2). Sanger sequencing of HORVU.MOREX.r3.6HG0574880 in all eight available ert-k mutants revealed an identical 4-bp deletion in exon 3 in both mutant ert-k.32 and ert-k.76, and a one-bp insertion in exon 5 in ert-k.309. In ert-k.32 and ert-k.76, the 4-bp deletion causes a truncated protein of 193 native amino-acid residues followed by an alanine residue. In ert-k.309, the one-bp insertion in exon 5 causes a frameshift. The truncated protein has 333 native amino-acid residues followed by glutamine and arginine (Figure 2). No mutations in HORVU.MOREX.r3.6HG0574880 were found in ert-k.93, ert-k.302, ert-k.435, ert-k.459 or ert-k.477. We therefore analyzed whether we could detect any mRNA of the candidate gene in the eight mutants. mRNA was detected by RT-PCR in all ert-k mutants, which suggested that the mutants are not transcript deficient (Figure 3). Since genomic DNA sequence information was available for ert-k.435 and ert-k.477, we looked for mutations in the intergenic regions in the vicinity of the candidate gene. In ert-k.477, we found a “C/T” SNP variation at position 6H:161619756. The mutation was confirmed by Sanger sequencing. The SNP is 133,213 bp upstream from the start codon of HORVU.MOREX.r3.6HG0574880. Three genes are located between the SNP and HORVU.MOREX.r3.6HG0574880, namely, HORVU.MOREX.r3.6HG0574850, HORVU.MOREX.r3.6HG0574860 and HORVU.MOREX.r3.6HG0574870. It is known that enhancers can be located hundreds of thousands base pairs away from the target gene and do not necessarily operate on the closest promoter (Pennacchio et al., 2013).

In order to analyze if the eight ert-k mutants are truly allelic, we crossed ert-k.32 with the other seven ert-k mutants or with cultivars Bonus and Foma. Mutant ert-k.32 was used as father in the crosses and the success of the crosses was verified by the heterozygous presence of the ert-k.32 allele in the resulting F₁ plants. If two recessive mutants are deficient in the same gene, i.e., truly allelic, the F₁ generation will show the mutant phenotype. In contrast, if the two recessive mutants are deficient in two different genes, a wild-type phenotype is expected in F₁ plants. F₁ seeds were planted in the greenhouse together with Bonus, Foma and each of the eight ert-k mutants. The plants were phenotyped by visual inspection (Figure 4) and spike length and the length between rachis internodes 5 and 15 were measured (Table 2). The allelism test suggested that mutants ert-k.93 and ert-k.302 might be non-allelic to ert-k.32, which would explain why no mutations in HORVU.MOREX.r3.6HG0574880 were found in these two mutants. Also ert-k.435 and ert-k.459 showed no mutations in this gene although the obtained F₁ plants displayed a mutant phenotype. Since we have no genome sequence data of ert-k.459, we could not analyze this mutant for intergenic SNP variation. In our previous work with barley mutants, we used to identify approximately 75% of the mutations in historic accessions that are supposed to be allelic (Zakhrabekova et al., 2012; Zakhrabekova et al., 2015; Matyszczak et al., 2020). Because the ert-k phenotype is similar to other ert phenotypes, we sequenced all ert-k mutants for the miRNA172 binding site of Ert-r (APETALA2) (Houston et al., 2013). The reason why we sequenced only the miRNA172 binding site of Ert-r is that it is known that only mutations within this domain are causing a dwarf phenotype (Houston et al., 2013). No mutations were found (not shown).

Mutants ert-k.32 and ert-k.76 were found to have the same 4-bp deletion, which was unexpected since the mutagenic process is random and they were induced with different mutagens and were isolated in different years (Supplementary Table S3). To rule out whether the two mutant lines have been mixed up over the years or if they represent two identical mutations generated independently of each other, we analyzed the SNP variation in a 40 Mbp region around the suggested Ert-k candidate gene HORVU.MOREX.r3.6HG0574880. Due to the tight linkage in the centromeric region around this gene, it is very likely that existing homozygous SNPs are inherited together. The presence of identical SNPs around the mutation would indicate a common origin of the mutants, i.e., seeds had been mixed up over the years. In contrast, the presence of unique SNPs would indicate that the two mutants are the results of two independent mutagenic events. In ert-k.32 and ert-k.76 we identified 17 and 12 SNPs, respectively, within the 40 Mbp region. Six of these were present in both mutants, whereas the remaining 11 and 6 SNPs were unique to each mutant (Figure 5). The six common SNPs probably reflect differences between the Bonus line used for mutagenesis in 1947 and 1955 that generated ert-k.32 and ert-k.76, respectively, and the Bonus accession we sequenced in this experiment. The unique SNPs were likely obtained through the mutagenic treatment. In order to investigate whether the 4-bp deletion was a result of the mutagenic treatment or was already present in an “ancient” Bonus line, we sequenced the site of the 4-bp deletion from all Bonus accessions available at the Nordic Genetic Stock Center (www.nordgen.org) and the cultivars Gull, Opal Abed, Maja Abed, Binder Abed, Guld Svalöf, Seger and Hanna, which are in the pedigree of Bonus. None of the listed cultivars showed the 4-bp deletion, which supports that the deletion is a result of the mutagenic treatment.

Since Pallas was used in plant breeding shortly after it was released as a cultivar, we analyzed cultivars with Pallas in their pedigree and four accessions of Pallas that are available at the Nordic Genetic Stock Center (www.nordgen.org) (Supplementary Table S5). The 4-bp deletion seen in ert-k.32 and ert-k.76 was found in all cultivars except Jenny, which in addition did not show the characteristic compact spike phenotype of Pallas (Figure 6). The fact that all four Pallas accessions and the cultivars Hellas, Senat and Visir contained the 4-bp deletion further supports that HORVU.MOREX.r3.6HG0574880 is the Ert-k gene.

Alpha/beta hydrolases are characterized by an active site composed of a catalytic triad of three involved amino-acid residues. Typically, a serine residue plays the role of a nucleophile in the reaction. The serine is activated by a catalytic base and a catalytic acid, which are typically a histidine and aspartate residue, respectively. To analyze if the ErtK candidate has a catalytic triad and to identify the involved residues, we performed an alignment of the ErtK candidate with several alpha/beta hydrolases with published protein structures (Mindrebo et al., 2016) (Figure 7). The alignment suggested that Ser-167, His-261 and D-232 are the essential amino-acid residues for the activity of the suggested ErtK hydrolase (Figure 7). The 4-bp deletion in ert-k.32 and ert-k.76 is located before the codon corresponding to His-261 and D-232, whereas the 1-bp insertion in ert-k.309 is located after.