A BLAST search with the rice ALS protein sequence (GI: 189031230) as query identified 13 entries (HB27A17r, HF08O07r, HF22F02r, HH04G02u, HO28K09S, HO28K09w, HQ01F18w, HS06N21r, HS17M10u, HS18N04r, HS18N04u, HT06N21r, and RUS50B01w) in the barley CR-EST database (IPK Gatersleben) and five in the EMBL-EBI ENA database (AF059600, HQ661102, HQ661103, AK361384, and AK368472). The sequences from both databases assembled into one contig. A PCR product obtained with primers ALSF (CAT GTC TCC ATT TGT GCA G) and ALSR (CTG CCA TCA CCC TCC ATG) and EST clone HQ01F18w as template was used to probe a Southern blot prepared from Golden Promise genomic DNA digested with the enzymes indicated in Figure 1A. The 5 kb region in the BglII digest visible in Figure 1A was excised from an agarose gel run in parallel, and a sub-genomic library was prepared (Seidman, 2001) by ligation into BamHI digested pBSK- vector (Stratagene). pBSK-ALS, the plasmid carrying the 5 kb fragment with the Golden Promise ALS gene was identified by PCR screening in pools of transformants. The insert was sequenced in both strands by Sanger sequencing. The DNA sequence is available at the European Nucleotide Archive (ENA) under accession number LT601589.

Genomic DNA was prepared as described (Dellaporta et al., 1983) or by the Qiagen Plant DNA easy kit (Qiagen, Hilden, Germany) as described by the manufacturer. PolyA⁺ RNA was extracted from leaf tissue using the Dynabeads mRNA DIRECT Kit (Invitrogen). Southern and Northern blotting was as described (Markmann-Mulisch et al., 2007). For Western blots, samples were prepared by crushing leaves and boiling the extract directly in sample buffer. The blots were prepared as described (Sambrook et al., 1989). Rabbit anti PpRAD51B antibody was obtained by commercial immunization (BioGenes, Gesellschaft für Biopolymere, Berlin) with purified protein overexpressed in E. coli (Ayora et al., 2002). The PCR analysis of gene targeting was done with Taq Polymerase (Ex Taq, Takara/ClonTech Europe) as described by the manufacturer using 35 cycles, denaturing temperature 98°C, annealing temperature 64°C, extension temperature 72°C. Primers were: PCR1, m567 (CCA TCA CCA AGC ACA ACT ACC TGG), m564 (GGT CAG CCG ACA ACT CTG AGG; PCR2, m567, m566 (GAG TGT CGT GCT CCA CCA TGT TG); PCR3, p35Sfwd (ACG CAC AAT CCC ACT ATC CTT C), m570 (CCG GAT CGG ACG ATT GCG TC).

For transformation, binary vectors pPEX-RAD51 and PEX-RAD54 were transformed into Agrobacterium strain AGL0, all others into AGL1 (Wang and Waterhouse, 2000). The corresponding strains were: p35S-ALS^S629N, A29; pGT-1ALS^mS629N, A27; pGT-2ALS^mS629N, A33; pINA-ALS^mS629N, A28; pWBVec10, Vec10. The barley cultivar Golden Promise was the genotype transformed throughout. Plant growth and transformation was essentially as described (Jacobsen et al., 2006; Watanabe et al., 2016). Selection conditions were: hygromycin: 50 mg/l; PPT: 8 mg/l; and Pursuit (application ready formulated herbicide solution, BASF): 400 nM. All tissue culture media were as described (Jacobsen et al., 2006) except that casein hydrolysate (Hydrolysate N-Z-Amine-A) was omitted. The leaf painting assay (Wu et al., 2003) was used for segregation analyses of greenhouse grown plants as described (Jones and Sparks, 2008).

For the analysis of transient expression assays, immature embryos were co-cultivated with Vec10 for 2 days and stained for β-glucuronidase (GUS) expression as described (Suter-Crazzolara et al., 1995).

For the radiation resistance assays, homozygous GP-RAD51-1 and Golden Promise seeds were harvested and irradiated dry at the IAEA Terrestrial Environment Laboratory in Seibersdorf, Austria. Seeds were sown on soil, grown in an environmental chamber and evaluated after 4 weeks.

The barley ALS gene was identified in the CR-EST (IPK Gatersleben) and EMBL-EBI ENA databases by protein sequence homology searches with the rice gene as query. The corresponding DNA sequences were assembled into one contig and this information used to obtain a partial ALS sequence from EST clone HQ01F18w by PCR. A Southern blot was prepared from Golden Promise DNA and probed with this fragment. The results showed that ALS is a single copy gene in barley (Figure 1A). The genomic portion located on a BglII fragment was isolated, cloned and sequenced. The 5403 bp ALS-BglII fragment contained the complete ALS coding region including 277 bp upstream and 3185 bp downstream sequences. A comparison of this sequence to the published barley genome (International Barley Genome Sequencing Consortium et al., 2012) is shown in Figure 1Be. The sequence comparison revealed a remarkable number of differences between Golden Promise and the reference genome Morex, both in the coding (point mutations) and the non-coding regions (insertions/deletions) in this part of the genome.

A sequence comparison to the whole genome assembly available in the Gramene database¹ (Hordeum vulgare assembly ASM32608v1) revealed the location of ALS on the long arm of chromosome 6H (Figure 1Bc). A closer inspection showed that the ALS gene is located on contig_40275, except its 5′ region (Figure 1Bd). The 5′ part was also not found in the anchored neighboring contig_1558010 suggesting the existence of a large gap between the two contigs which is not bridged by our sequence. The Gramene database also provides data on the distribution of protein coding and repetitive sequences across all chromosomes (Figures 1Ba,b). A closer inspection of chromosome 6H in this respect showed an enrichment of protein coding genes in the region around the ALS gene, as expected from its position (International Barley Genome Sequencing Consortium et al., 2012; Higgins et al., 2014). However, this analysis also revealed the presence of repeats in this area which co-localize with genes. The presence of repetitive DNA directly next to the ALS gene was confirmed by an additional analysis which used the 3′ terminal 400 bp of the gene replacement fragment (described below) as a query in a BLASTN search against the entire barley genome. A high (1e^-50) threshold value (as defined in the Gramene database) setting identified sequence elements of 100 to 150 bp in length with almost perfect homology at eight other positions and on different chromosomes. A low threshold setting (1e^-5) showed that shorter and less conserved sequences exist thousands of times in the genome (Figure 1Bf). These data indicate that the terminal 400 bp of the gene targeting fragment are sequences with interspersed repetitive DNA.

Wild type ALS is inhibited by several different herbicides, one of which is the imidazolinone herbicide Imazethapyr, commercially available as Pursuit (BASF). A gene targeting assay system was developed for A. thaliana which is based on an ALS gene with a single point mutation at position 653 causing a serine to asparagine substitution and selection for Pursuit resistance (Badur and Reiss, 2004). The same mutation exists in imidazolinone herbicide resistant ALS genes of rice (Endo et al., 2006) and barley (Lee et al., 2011). This mutation was introduced into the Golden Promise HvALS gene by site directed mutagenesis. The homologous position in the barley protein sequence is S629 and conversion of the AGC codon at this position into AAC resulted in the desired S629N mutation. To express the Pursuit resistant ALS gene in barley, HvALS^S629N was placed between the CaMV 35S promoter (Odell et al., 1985) and polyadenylation signal (Sanfacon et al., 1991) and inserted into the barley transformation vector pMBVec8 (Wang et al., 1998). The resulting binary vector, p35S-ALS^S629N (Figure 2A) also harbors the intron-split hygromycin phosphotransferase (hpt) gene of pMBVec8 for selection on hygromycin. Vector p35S-ALS^S629N was transformed into Agrobacterium strain AGL1 to yield strain A29. For transformation control purposes, pMBVec10 (Wang and Waterhouse, 2000) with the same hpt gene and vector backbone as p35S-ALS^S629N was transformed into AGL1 to yield strain Vec10.

Barley Golden Promise transformation comprises co-cultivation of immature embryos with Agrobacteria (Tingay et al., 1997) followed by callus, shoot, and root induction on different media and under different conditions. The callus induction phase is the most critical period for selection. Embryos regenerated on the published callus induction media (Jacobsen et al., 2006) were barely sensitive to Pursuit at concentrations of 300, 600, and 1500 nM. Barley tissue culture media contain high concentrations of amino acids in the form of casein hydrolysate which could compensate for ALS inhibition by Pursuit. Therefore casein hydrolysate was omitted from the media and the inhibitory effect of Pursuit tested at concentrations of 50, 150, 300, 400, 600, 900, 1500, and 3000 nM. Callus and shoot growth were still observed at concentrations up to 300 nM, but almost completely inhibited at 400 nM Pursuit. To exclude that the omission of casein hydrolysate had any effect on callus induction and transformation efficiencies, immature embryos were transformed with Vec10 and selected on hygromycin using media with and without casein hydrolysate. The direct comparison of callus induction and transformation efficiencies showed that omission of casein hydrolysate had no effect. To characterize Pursuit selection further, immature embryos were co-cultivated with A29 and Vec10 and one half of them selected on the modified media containing either 400 nM Pursuit or hygromycin. The comparison of growth on Pursuit and hygromycin showed that both procedures perform equally well. With A29, callus and shoot development was comparable on Pursuit and hygromycin while Vec10 transformed calli were severely inhibited in growth on Pursuit and gradually showed brownish discoloration (Figure 2B). The transformation efficiencies obtained with A29 on hygromycin and Pursuit were 44 and 34%, respectively. Those with Vec10 were 39 and 0%, respectively, on hygromycin and Pursuit (Table 1). The data show that the transformation efficiencies obtained with A29 on Pursuit and hygromycin are the same within the variability of the experiments. The same is true for hygromycin selection and strains Vec10 and A29 which shows that both vectors performed equally well.

The transformation efficiencies in these experiments were calculated as follows: A single, co-cultivated immature embryo usually yields a large, but variable number of shoots at the end of a transformation experiment. These can be of clonal origin or originate from different independent transformation events. To avoid ambiguities in the data, we counted only one shoot per embryo and defined an independent transformant as an embryo that produced at least one shoot forming a root in the presence of selective agent. Consequently, transformation efficiency is the number of immature embryos that formed shoots divided by total number of co-cultivated embryos.

The disadvantage of this definition is that it underestimates the true number of transformation events obtained in an experiment. To address this problem, we analyzed transient GUS expression in immature embryos co-cultivated with Vec10. These experiments (Figure 2C) confirm earlier observations (McCormac et al., 1998; Murray et al., 2004) showing that a large number of different cells is infected and at least transiently transformed by Agrobacterium in this procedure. Although clearly not all cells showing transient GUS expression give rise to stable transformants later, this assay is a reasonable proxy for transformation, and the data obtained with it suggest that a single immature embryo produces multiple independent transformation events.

A severe problem in barley transformation is overgrowth by Agrobacteria. Hygromycin selection was optimized to prevent overgrowth by Agrobacteria. The introduction of an intron-split hpt gene preventing its expression and the sensitivity of Agrobacteria to hygromycin largely solved the problem (Wang et al., 1998). However, there is no such system in Pursuit selection. Consequently, overgrowth was occasionally a larger problem with Pursuit selection that has caused loss of regenerating embryos. The problem was negligible with AGL1 provided the plant material was of good quality. However, the problem was quite severe with a different vector, pH001-ALS^S629N and strain AGL0. Many immature embryos were lost on Pursuit by overgrowth in this combination, suggesting that AGL1 is important for the performance and efficiency of Pursuit selection in barley.

The vector constructed to express the P. patens PpRAD51B gene (Markmann-Mulisch et al., 2002), pPEX-RAD51 (Figure 3A) is described in Section “Materials and Methods.” Strain A12 was obtained by transformation into Agrobacterium AGL0. To produce the PpRAD51B overexpression lines, 60 immature embryos were co-cultivated with A12 and transformants selected on phosphinotricin (PPT) containing media. These experiments yielded 22 PPT resistant plants originating on eight different immature embryos. To identify the individuals with high PpRAD51B protein expression levels, primary transformants (T0) were analyzed by Western blotting and a PpRAD51B-specific antibody raised against purified protein overexpressed in E. coli (Ayora et al., 2002). This screen identified two transformants with high PpRAD51B expression levels, GP-RAD51-1 and GP-RAD51-2. These plants were self-pollinated and the progeny (T1) analyzed in more detail. To determine the number of independent transgene inserts in the genome, inheritance of PPT resistance in the segregating progeny was analyzed with the leaf painting assay (Wu et al., 2003). Sixteen out of 23 GP-RAD51-1 siblings were PPT resistant closely matching the expected ratio of 3:1 for Mendelian inheritance of a single gene. All 23 GP-RAD51-2 siblings were PPT resistant indicating the presence of multiple, independently segregating inserts. This generation was further analyzed by Southern blotting (Figure 3C). Genomic DNA was prepared, digested with EcoRV, an enzyme producing a single cut within the inserted transgene, blotted and the membrane probed with a PpRAD51B gene fragment. The results revealed the presence of a single transgene insert in GP-RAD51-1 and at least three independently segregating inserts in GP-RAD51-2. Western blotting of the same plants showed that all siblings with a transgene expressed PpRAD51B protein (Figure 3B). To identify homozygous GP-RAD51-1 individuals, PPT resistant T1 plants were self-pollinated and segregation of PPT resistance determined in the progeny. This analysis identified one homozygous line (line 2) with 100% PPT resistant progeny (T2). To obtain sufficient seeds for the gene targeting experiments, T2 plants were propagated by self-pollination and PpRAD51B expression confirmed by Northern blotting (Figure 3D) in the plants grown from them. For the Northern blot, Poly A⁺ RNA was prepared, separated by formaldehyde agarose gel electrophoresis, blotted and the membrane hybridized with the PpRAD51B gene probe. All plants expressed PpRAD51B confirming stable inheritance of an active gene in this generation. However, for unknown reasons mRNA as well as PpRAD51B protein expression seems to be variable.

For GP-RAD51-2, line 6 (T1) was chosen as a plant with an apparent single copy transgene integration. The plant was self-pollinated and siblings analyzed in more detail. Leaf painting showed that PPT resistance segregated in this population, but the transgene was present in all individuals, as shown by Southern blotting (Figure 3E), whether resistant or sensitive. To resolve this issue, a Southern blot with an enzyme (BclI) not cutting within the transgene was prepared. This blot showed the presence of two different insertions in line 6 (Figure 3E), one of which co-segregated with PPT resistance while the other one was present in all siblings. The corresponding Northern blot showed that all of them express the PpRAD51B gene (Figure 3F). These results indicate that GP-RAD51-2 line 6 is homozygous for a PpRAD51B expressing transgene and hemizygous for an additional transgene with a functional pat gene.

RAD51 overexpression in mammalian cells improved gene targeting and resistance to DNA damaging agents in some cases (Vispe et al., 1998; Yanez and Porter, 1999; Lundin et al., 2003) but not in others (Lambert and Lopez, 2000). To see whether RAD51 overexpression has an effect on radiation resistance in barley, homozygous GP-RAD51-1 seeds were gamma irradiated, and seed germination rates and leaf length at the seedling stage scored as parameters to judge resistance (Figure 3G). There was no significant difference between RAD51 overexpressing plants and wild type. Seeds germinated well up to a dose of 300 Gy and germination rates gradually dropped afterward. Leaf growth and development was affected by doses above 75 Gy with leaves becoming gradually shorter and growth stunted at higher doses. The data show that dry barley seeds are highly resistant to γ-rays and RAD51 expression does not seem to have an effect on resistance under these conditions. This result is different from that obtained with ScRAD54 in similar resistance assays (Shaked et al., 2005), but consistent with the results obtained with the Atrad51 mutant in A. thaliana which suggest that RAD51 is not involved in the repair of the type of DNA damage that γ-rays cause (Markmann-Mulisch et al., 2007).

The vector constructed to express the budding yeast ScRAD54 gene (Shaked et al., 2005), pPEX-RAD54 (Figure 4A) is described in Section “Materials and Methods.” Strain A13 was obtained by transformation into Agrobacterium AGL0. To obtain ScRAD54 overexpressing plants, 60 immature embryos were co-cultivated with A13 and transformants selected for PPT resistance. This transformation yielded 21 shoots originating on seven different immature embryos, 18 of which were viable and set seeds. These transformants were screened for ScRAD54 expression by Northern blotting (Figure 4B) and the ScRAD54 gene as probe. GP-RAD54-9, the primary transformant (T0) with the highest ScRAD54 expression was self-pollinated and analyzed in the next generation by Southern blotting. Genomic DNA from plants in a segregating population was digested with HindIII cutting within the ScRAD54 gene and the blot hybridized with the ScRAD54 probe (Figure 4C). This blot indicated the presence of at least two independently segregating insertions in GP-RAD54- 9. Selected T1 plants were self-fertilized and the progeny tested for PPT resistance. The progeny of one plant (line 24) with 100% PPT resistant siblings was analyzed further. Northern blotting (Figure 4E) showed that all individuals expressed ScRAD54. Southern blotting showed that all had one transgene insertion in common (upper band in Figure 4D), while some of them carried an additional insertion (lower band in Figure 4D). An additional band appears on top of the upper band and is visible in most lanes and in both Southern blots (Figures 4C,D). This band is likely due to partial digestion, but could be another insertion. These results indicate that GP-RAD54-9/24 is homozygous for a transgene insertion harboring expressed pat and ScRAD54 genes, and hemizygous for another insertion with unknown composition.

To get an estimate of how many different transformation events can be generated with one co-cultivated immature embryo, the Southern blots produced to characterize the overexpression lines were evaluated quantitatively. In this set, eight immature embryos produced a total of 23 shoots. Eighteen of these shoots differed in their integration pattern and therefore were independent transformants. Consequently, one transformed immature embryo produced 2.25 different transformation events on average in this sample.

The ALS-based gene targeting assay system established for barley is shown in Figure 5A. The design follows the one used to analyze gene targeting in A. thaliana before (Badur and Reiss, 2004; Prinzenberg, 2006). The system comprises a gene replacement fragment with a non-functional, 5′ terminally truncated ALS gene carrying the S629N mutation and a second, silent diagnostic mutation which allows independent identification of the introduced gene. Then an intron-split hpt gene under 35S promoter control follows which was inserted between the putative polyadenylation/termination signal of ALS and the following 2 kb of sequences present in the genome. This design provides 1.2 kb of sequence homology to the genome at the 5′ and 2 kb at the 3′ end of the replacement fragment. Two different gene targeting vectors (Figure 5B) were obtained by insertion of the same gene targeting fragment into two different vector backbones. In one of them, pGT-1ALS^mS629N the backbone was pMBVec8. For the other one, pGT-2ALS^mS629N, a newly constructed vector with minimalised T-DNA border sequences was used to reduce heterology at the ends. Both plasmids were transformed into AGL1 resulting in strains A27 for pGT-1ALS^mS629N and A33 for pGT-2ALS^mS629N.

To analyze gene targeting, immature wild type embryos were co-cultivated with A27 or A33 and transformants selected on Pursuit. To control the quality of each single transformation experiment, an additional aliquot of 20 immature embryos was co-cultivated in parallel with the same A27 or A33 Agrobacterium culture and selected on hygromycin. This reference transformation served to assure quality of individual transformation experiments and to determine the transformation efficiency obtained in each of them. Transformation efficiencies were determined as before except that rooting assays were omitted for hygromycin. However, they were continued for Pursuit selection. For wild type, 887 embryos were transformed with A27 or A33 and selected on Pursuit (Table 2). The reference transformations on hygromycin predicted that this number would have generated 465 independent transformants. A total of 130 calli formed shoots on Pursuit in these experiments, but none of them rooted in the presence of Pursuit later. Consequently, no Pursuit resistant transformant was among them and these experiments did not yield a single gene targeting event.

To analyze gene targeting in PpRAD51B and ScRAD54 overexpressing lines, immature embryos obtained from plants grown from homozygous GP-RAD51-1/2 and GP-RAD54-9/24 seeds were co-cultivated with A27 and A33 and selected on Pursuit. The experiments included the same reference transformations as for wild type described above. In all transformation experiments together, 1861 RAD51 and 136 RAD54 transgenic embryos were transformed with A27 or A33 (Table 2). The reference transformation on hygromycin predicted that 1304 independent transformants were obtained on Pursuit altogether (Table 2). Although many calli produced shoots, none of them was a Pursuit resistant transformant.

The reference transformation also showed that the pGT-1ALS^mS629N (A27) and pGT-2ALS^mS629N (A33) targeting vectors performed equally well since both yield comparable transformation efficiencies on hygromycin, 62±23% and 72±14% (Table 2). Therefore, potential differences between the strains seem to be negligible and the data were pooled.

In P-N selection, the positive marker is used to select for the integration of the gene replacement fragment while the negative marker prevents random integration. A potent negative selection marker is the 35S-promoter-driven diphtheria toxin-A (DT-A) gene. Placed at the ends of the gene replacement fragment, these sequences are lost upon homology-mediated integration but maintained in random integration and cause cell death. To obtain a P-N selection system in barley, the ALS^mS629N gene replacement fragment was inserted between the two DT-A genes flanking each T-DNA border region of pINA134 (Terada et al., 2002). The resulting vector, pINA-ALS^mS629N (Figure 5C) was transformed into AGL1 and the resulting strain (A28) used for co-cultivation of wild type embryos. In total, 1645 immature embryos were transformed and selected on hygromycin (Table 3). Transformation efficiencies were monitored by transforming an aliquot of the same batch with Vec10, selection on hygromycin and scoring callus growth. Compared to Vec10, A28 transformed calli developed considerably slower and poorer on hygromycin, as expected for negative selection. Very few calli produced poorly growing shoots, only one of which formed roots on hygromycin. Molecular analysis revealed later that this one was not targeted. The reference transformation with Vec10 suggests that 1270 independent transformants were generated in the P-N transformation experiments.

The ALS-based gene targeting assays rely on the formation of a functional, Pursuit resistance conferring gene. To exclude that selection problems have prevented rooting on Pursuit, all shoots obtained in the above assays were tested in pools of 100 by PCR. All gene targeting vectors used the same replacement fragment (Figure 5) and would create the same 5′ recombination junction upon HR with the ALS gene in the genome. This junction is detected specifically in PCR2 (Figure 5A). Diploid barley nuclei contain 5.79 pg of nuclear DNA. This number corresponds to 20.000 diploid genomes in the amount of DNA used in a standard PCR reaction (120 ng), or 200 genomes that one individual contributes to a pool of 100 transformants. To optimize the PCR reaction, serial dilutions of a control plasmid which contained a 5′ recombination junction generated in vitro by cloning were mixed with 120 ng genomic Golden Promise DNA and conditions optimized until 200 recombination junction molecules were reliably detected. Additional PCRs were used to detect the genomic ALS gene (PCR1) and the hpt gene in the gene replacement fragment (PCR3).

Shoots obtained by Pursuit or P-N selection were collected in pools of 100, DNA prepared and amplified by PCR2. The corresponding product was not present in any of the pools, but it was easily amplified from the control DNA (Figure 5D). The same was true for the single hygromycin resistant shoot obtained by P-N selection. Since PCR1 showed that all DNA preparations were PCR-amplifiable (Figure 5D) the results show that the 5′ recombination junction was not present in any of the transformed plants. In addition, many of the pools did not even contain transformants since the hpt gene detected with PCR3 was not present in all them. By contrast, the same fragment was present in all 33 individual A27 transformants selected on hygromycin.