Protospacers for the synthetic guide RNA (sgRNA) in the CRISPR/Cas9 experiments were designed to target each of the seven candidate genes (Supplementary Table 1). For screening purposes, protospacers with a restriction enzyme recognition site spanning the fourth base from the protospacer adjacent motif (PAM) were prioritized. A list of primers used to amplify the targeted region can be seen in Supplementary Table 2. Amplification was done using the Phusion^TM High-Fidelity DNA Polymerase (Thermo Scientific) as described by the manufacturer.

The protospacer sequences were cloned into the entry vector pJG85 under the control of the wheat U6 RNA polymerase III promoter (TaU6), flanked by the attL5 and attL2 Gateway cloning sites (Gil-Humanes et al., 2017). Another entry vector pJG80 (Gil-Humanes et al., 2017), with a promotor-less wheat optimized Cas9 (TaCas9) flanked by a attL1 and attR5 Gateway cloning site, was cloned, along with pJG85 using Gateway LR clonase enzyme mix (Invitrogen) into the destination vector pANIC6A described by Mann et al. (2012) (Supplementary Figure 1A).

For overexpression of candidate genes, the WBVec8 modified vector Ubi:USER:NOS was used with a maize ubiquitin constitutive promoter (Hebelstrup et al., 2010) (Supplementary Figure 1B). The vector was linearized using PacI leaving a 3′-AT overhang. The genes were either amplified from cDNA or genomic DNA with primers that had a 15 nt overhang homologous to the vector insertion site and AT overhang nucleotides (CAGGCTGAGGTCTTAAT) (Supplementary Table 3). The Phusion^TM High-Fidelity DNA Polymerase (Thermo Scientific) was used for amplification as instructed by the manufacturer. The fragments were cloned into the vector using In-Fusion^® HD kit (Clontech) as described by the manufacturer. The gene sequences amplified for overexpression were obtained from either genomic DNA or cDNA of barley cultivar Golden Promise (GP). The sequences are listed in the Supplementary Data 1 with the primers used for amplification marked on the sequences. The annotations used from either the old or new genome assembly are stated.

The vectors were cloned into Stellar^TM competent cells (Clontech) using heat shock transformation. Colonies were grown at 37°C on selective LB medium with 50 μg/ml spectinomycin or kanamycin for overexpression or CRISPR/Cas9 constructs, respectively.

All vectors were subsequently transformed into Agrobacterium strain AGL0 using the freeze/thaw method. Colonies were grown at 28°C on selective medium with rifampicin (25 μg/ml) and 50 μg/ml spectinomycin or kanamycin for overexpression or CRISPR/Cas9 constructs, respectively.

Crude DNA extraction from a small piece of leaf was performed using the Extract-N-Amp^TM Plant PCR kit (Sigma Aldrich), as described by the manufacturer. The crude DNA was diluted 1:3 before used as template for the polymerase from the same kit or the KAPA3G Plant PCR kit (Kapa biosystems).

DNA for cloning and sequencing was extracted from 10 cm young leaf pieces immediately frozen in liquid nitrogen. The leaves were crushed in a FastPrep-24 5G homogenizer (MP Biomedicals) while still frozen and DNA extracted by the phenol/chloroform extraction method.

Plants obtained by CRISPR/Cas9 transformation and by overexpression of the candidate genes were screened by PCR amplification of the hygromycin resistance gene in the T-DNA. The same primers were used for amplification of hygromycin in the knockout and overexpressing transformants Hyg_fw: 5′-ACTCACCGCGACGTCTGTCG-3′ and Hyg_rv: 5′-GCGCGTCTGCTGCTCCATA-3′. The program was 95°C for 3 min, then 40 cycles with 95°C for 3 min, 61°C for 15 s, 72°C for 40 s, and a final extension at 72°C for 1 min.

The product sizes were different in the knockout and overexpression constructs due to an intron in the hygromycin gene in the overexpression vector. The PCR products were 727 bp for knockout mutants and 917 bp for overexpression transformants.

Transformed plants obtained by CRISPR/Cas9 transformation were analyzed for the presence of mutations by PCR/restriction enzyme (PCR/RE) analysis as an initial screen to identify mutations. Uncut PCR products from the PCR/RE analysis of the CRISPR/Cas9 transformed plants were cloned and sequenced. Cloning was done using the Zero Blunt^TM TOPO^TM PCR cloning kit (Invitrogen). In the T₁ generation, progenies homozygous for the knockout mutations and no T-DNA insert were identified and selected for further analysis and for crossing.

The barley lines expressing EGF and mLIF were kindly provided by ORF genetics (EGF^ORF and mLIF^ORF, respectively). They were generated by Agrobacterium-meditated transformation of Golden Promise. The T-DNA within the EGF^ORF line contained two human EGF copies driven by a seed-specific B-hordein promoter or an oat globulin seed-specific promoter, respectively. Both copies were terminated with a D-hordein terminator (Supplementary Figure 1C). The T-DNA within the mLIF^ORF line contained only a single copy of mLIF controlled by a seed specific D-hordein promoter and terminator (Supplementary Figure 1D). The EGF^ORF line was homozygous after double haploid formation whereas the mLIF^ORF was heterozygous.

Grain from transformed lines were counted and weighed for calculation of the 1,000-grain weight. Five g of grains from some of the lines were also sent to Eurofins for determination of total protein content.¹ This was done using the Dumas combustion method to determine the total nitrogen content in the samples.

The ER-stress assay was modified from a method described for Arabidopsis (McCormack et al., 2015). Two barley seeds per well in a 12-well plate were germinated in 800 μl 1/2 MS liquid media with 50 μg/ml ampicillin and incubated at 23°C, with 16/8 light/dark hours. After 4 days, the media was removed and 1.6 ml fresh media with either 0, 25, or 100 μg/ml TN dissolved in dimethyl sulfoxide (DMSO) was added. The plantlets were incubated for another 4 days, after which the media was removed. The wells were washed three times in sterile water before 1.6 ml media without TN was added to each well. After three more days of incubation, the plantlets were transferred to 2 ml Eppendorf tubes, frozen in liquid nitrogen and stored at −80°C until RNA extraction.

Frozen leaves were crushed using two glass beads in a FastPrep-24^TM 5G homogenizer (MP Biomedicals) for 6 s at speed 5.0. RNA extraction was done using the Spectrum^TM Plant Total RNA Kit (Sigma Aldrich) as described by the manufacturer. Potential residual DNA was removed by addition of RNase free DNase from Qiagen along with RNasin^® Ribonuclease inhibitor (Promega). Subsequently, the samples were purified using the NucleoSpin^® Gel and PCR Clean-up kit (Macherey-Nagel) with the NTC buffer. RNA concentration and quality were evaluated by NanoDrop 1000 (ThermoFisher Scientific) and on a 1% agarose gel to visualize the 18S and 28S ribosomal bands.

The extraction and analysis of recombinant EGF protein was performed as described in Panting et al. (2020). The Sandwhich ELISA assay of mLIF was done with a mouse LIF DuoSet ELISA kit from R&D Systems as described by the manufacturer. The samples were diluted 32,000 times before measuring the amount of mLIF protein. The mean μg/ml concentration of each biological replicate was calculated, and the standard error and a two-tailed Student’s t-test was calculated based on the mean concentrations. p < 0.05 was considered a significant change.

Knockout PDI mutant plants and GP controls were grown in the semi field and harvested at 20 days after pollination (DAP) for protein staining of the starchy endosperm. The method used for staining protein and sectioning was basically as described in a previous study (Wan et al., 2014). Samples were gradually dehydrated in ethanol dilution series going from 10 to 100% ethanol, increasing by 10% at each step with incubation for 1 h. Samples were kept overnight in 70% ethanol and dried two times in 100% ethanol. Sections were infiltrated with LR White resin using a mild vacuum for 1 h. The ethanol:LR White resin ratio went from 4:1 to 1:4 and the samples were kept at each concentration over night at 4°C before entering the next step. Finally, the samples were infiltrated with pure LR White resin. The pure LR White resin step was repeated three times and changed up to two times per day for 4 days or longer. Samples were incubated at room temperature on a rotator during the day and kept at 4°C at night. The embedded samples were polymerized at 55°C and cut into 1 μm slices with an ultramicrotome (Ultracut R, Leica, Germany). The sections were stained using 1% Napthol Blue Black in 7% acetic acid and visualized under a light microscope. Then, the stained sections were rinsed, dried, and photographed using a light microscope (DMLS, Leica, Germany) for protein body imaging. The experiment was done in triplicate.

Regenerated CRISPR/Cas9 transformed T₀ plants were screened for the presence of the T-DNA. Between 12 and 39 transformants were screened for the hygromycin gene and positive plants were screened for mutations using PCR/RE or directly sequenced. Plants with partial or no digestion of the PCR-product were selected for further characterization and the PCR-product from 5 to 13 plants were cloned and sequenced to identify the mutation (Table 1). The subsequent T₁ plants from selected mutants were again screened by PCR/RE and for the presence of the hygromycin resistance gene in order to select T₁-plants homozygous for the mutation but without the CRISPR/Cas9 T-DNA construct. The PCR-product containing the mutation from these T₁-plants was sequenced to verify the heredity of the mutation (Table 1). The knockout lines were named with GP followed by the knockout gene in superscript small letters (GP^hsp70, GP^hsp26, GP^hsp16.9, GP^gst, GP^crt, GP^ipi, and GP^pdi). Supplementary Table 5 lists all generated plants for this study, and which experiments they are included in. For the GP^crt mutant line, a homozygous mutant was first identified in the T₂ generation and GP^hsp26 was first identified in the T₁ generation. Six out of the seven mutants selected for further study had a +1 nt insert. The last mutant line was a −1 nt deletion of HSP16.9 in GP^hsp16.9.

Despite several transformation attempts on 450 and 175 embryos with the overexpression CRT and HSP16.9 constructs, respectively, it was not possible to obtain any transformed plants. Plants transformed with the other five target genes were regenerated and screened by PCR and for overexpression of the inserted gene in T₀ or T₁ plants (Supplementary Figure 2). From expression analysis of 6 to 12 overexpressing plant lines per construct, a high expression line from each construct was selected and homozygous lines generated for further analysis. These were named with GP followed by superscript large letters (GP^IPI, GP^HSP26, GP^PDI, GP^GST, and GP^HSP70).

The stress responses were evaluated in barley plants where the seven candidate genes one by one were either knocked out or overexpressed, except for the overexpression of CRT and HSP16.9. Expression of the known UPR responsive genes BiP, CRT, and PDI were monitored in control plantlets (Lu and Christopher, 2008; Kim et al., 2013). These genes have previously been correlated to the UPR. Plantlets from knockout mutants and overexpressing lines exposed to either 0, 25, or 100 μg/ml TN. In the GP control, an expected gradual increase in BiP and PDI gene expression was observed as the TN concentration increased. For CRT, the expression level was similar between 0 and 25 μg/ml TN whereas increased expression was observed at 100 μg/ml TN (Supplementary Figure 3).

The PDI knockout and overexpressing lines GP^pdi and GP^PDI showed an expected lower and higher expressions of PDI, respectively, compared to GP (Figure 1A). Moreover, GP^PDI had significantly increased expression of BiP at 0 and 100 μg/ml, and CRT expression was significantly higher at 0 and 25 μg/ml TN.

Both GP^hsp70 and GP^HSP70 showed an increased expression of all target genes at 25 μg/ml (Figure 1B). At 100 μg/ml TN, gene expression levels dropped again, with the lowest level seen in GP^hsp70.

For the HSP26 knockout and overexpression lines, there was a general increase of expression at 25 μg/ml TN, but only CRT expression in GP^HSP26 was significant. At 100 μg/ml TN, stress gene expression levels in GP^hsp26 dropped to wild type levels whereas expression levels in GP^HSP26 were decreased to below wild type levels (Figure 1C).

In GP^ipi, the BiP and PDI genes were upregulated at 0 μg/ml TN. Opposite, overexpression of IPI in GP^IPI caused overexpression of PDI and CRT genes (Figure 1D). At 25 μg/ml TN, both GP^ipi and GP^IPI had increased expression of all genes, particularly, CRT.

In the gst knockout line GP^gst the BiP, CRT, and PDI were increasingly expressed at 0 μg/ml TN (Figure 1E). The trend was less clear at 25 and 100 μg/ml although there were indications of increased PDI and CRT expression at 25 μg/ml and BIP at 100 μg/ml, respectively. For GST overexpressing plantlets, TN treatments at 25 and 100 μg/ml caused increased expression for the three studied genes, although only PDI and CRT at 25 μg/ml and PDI at 100 μg/ml had statistically significant.

As expected, the crt knockout in line GP^crt led to no expression of CRT at any TN concentration (Figure 1F). In contrast, increased expression was seen for BiP, statically significant at 0 and 100 μg/ml TN. PDI expression in the crt knockout was only significantly increased in expression at 0 μg/ml TN. The GP^hsp16.9 mutant line had increased expression of PDI and BiP at 0 μg/ml TN. All three genes were increasingly expressed at 25 μg/ml and only BiP at 100 μg/ml TN.

In summary, the TN assay was functional in barley for evaluating the expression of the stress responsive genes PDI, BPI, and CRT. Treatments of wild type plantlets with increasing TN levels induced expression of the PDI, BPI, and CRT genes. Moreover, the stress response of the PDI, BPI, and CRT genes were modulated by individual overexpression or knockout of the seven genes CRT, PDI, IPI, GST, HSP70, HSP26, and HSP16.9. All lines showed at least one significant change in expression of at least one UPR stress gene, except for GP^hsp26.

The effect of EGF and mLIF expression and stress gene overexpression or knockout on grain development was evaluated through 1,000-grain weight measurements (Table 2). First, expression of the two heterologous genes for EGF and mLIF caused no change in 1,000-grain weight. However, despite grown under the same growth conditions, only GP^ipi showed the same 1,000-grain weight (45.9 g) as wild type (46 g). Grains from all other lines with modulated genes had a reduced 1,000-grain weight compared to GP. The lowest 1,000-grain weight was seen for GP^pdi (29.5 g). Plants from the two overexpressing lines GP^HSP26 and GP^GST grown in the greenhouse had a reduced 1,000-grain weight when compared to wild-type seeds grown under the same conditions. A general observation was that greenhouse conditions resulted in a lower 1,000-grain weight compared to the semi field.

Visually, the grain from GP^pdi with the lowest 1,000-grain weight had a shrunken grain phenotype compared to wild type GP and GP^PDI (Figure 2).

GP^pdi grains with a strongly reduced 1,000-grain weight and shrunken grains were selected for microscopy and protein staining (Figure 3). In GP wild type control, distinct, dark blue protein bodies were clearly visible among starch granules across the entire endosperm, with a denser coloration in the outer cell layers (Figures 3D–F). The protein distribution in GP^pdi was considerably less distinct with fewer protein bodies (Figures 3A–C). The endosperm of GP^pdi displayed a light blue coloration, almost covering the entirety of the cells. The proteins in the starchy endosperm were not properly assembled into protein bodies, but instead diffused within the cells. From these images, we were not able to determine if there was a reduction of protein within the grain of GP^pdi or if it was only the distribution of protein that were affected. We therefore estimated the protein content in mature grain of GP^pdi, GP^PDI, and GP grown in the semi field. Both samples from GP^pdi and GP^PDI had a higher protein content than GP wild type with 14, 15.7, and 10.9%, respectively (Table 2). The amount of protein per 1,000 grains were calculated to be 5.01, 4.13, and 5.99 g for GP, GP^pdi, and GP^PDI, respectively. In conclusion, the knockout of PDI both reduces the amount of protein and disrupts the cellular organization of the protein in the grains of these lines.

The crossed offspring selected above was analyzed for their content of either EGF or mLIF by a sandwich ELISA assay. The mLIF screening was made on plants grown in either a growth chamber or in the greenhouse. The F₂ plants, mLIF^gst, and mLIF^wt were grown in a growth chamber. The mLIF concentration in mLIF^wt was significantly lower than that in mLIF^ORF. The mLIF concentration in mLIF^gst was higher than mLIF^wt but not to a significant level. The other crosses expressing mLIF were grown in the greenhouse, and all showed a significant reduction on mLIF concentration compared to mLIF^ORF (Figure 4). This was the case with both overexpression and knockout of IPI and knockout of CRT genes.

The accumulation of EGF in mature grains from crosses with EGF^ORF was also analyzed (Figure 5). A F₂ wild type from a cross with the GP^hsp16.9 knockout mutant was included as a control (EGF^wt). The knockout of PDI had a significant negative effect on EGF accumulation showing a decrease from 1.4 μg/mg in EGF^wt to.7 μg/mg in EGF^pdi (Figure 5). The EGF^PDI did, however, not show a positive effect on the accumulation of EGF. The two plants overexpressing IPI both had significantly higher EGF content than EGF^wt with 1.7 and 1.8 μg/mg for EGF^IPI–1 and EGF^IPI–3, respectively. Overexpression of the HSP70 gene had no effect on the accumulation of EGF in the offspring whereas EGF^hsp70 knockout significantly decreased the EGF level. The studied offspring from crossing with GP^hsp26 showed varying EGF concentration. EGF^hsp26–2 had a significantly less EGF than the EGF^wt control, while the EGF^hsp26–5 had an increase in EGF to 1.9 μg/mg (Figure 5). However, due to high standard error in the EGF^hsp26–5 the increased EGF was not significantly higher according to Student’s t-test.