Main

Agronomy innovations and crop breeding have been instrumental in achieving sustained growth in global food production over the past century. However, ensuring food security for a growing population remains a major challenge, especially due to climate change and limited genetic resources1,2,3. Breeding efforts to develop genetically uniform cultivars have focused on a narrow pool of species and varieties, overlooking a wealth of locally adapted and genetically diverse traditional crops. While modern breeding, through prolonged domestication, has enhanced traits that boost productivity, it has also led to genetic bottlenecks, reducing genetic diversity and stress resistance in modern crop gene pools4. Crop wild relatives (CWRs) are reservoirs of valuable traits, including diverse forms of resistance to biotic and abiotic stresses, which are critical for adapting modern cultivars to future climate conditions5,6. Harnessing the genetic diversity of CWRs to improve stress resistance is paramount for future crop breeding efforts7.

Plants in the Triticeae tribe include some of the world’s most important cereal crops, such as wheat and barley, providing a substantial portion of global dietary intake. Among Triticeae CWRs, Hordeum I-genome species have garnered attention due to their exceptional stress tolerance and genetic diversity8,9,10. Species such as Hordeum brevisubulatum show superior adaptability to saline and alkaline soils11,12. With over one billion hectares of arable land worldwide affected by salt toxicity, approximately 60% of which is sodic13, transferring adaptive traits from Hordeum I-genome species into crops holds considerable promise for improving stress resilience in modern barley and wheat varieties. While considerable progress has been made in understanding the mechanisms behind plant salt tolerance, knowledge of alkaline salinity (alkali) tolerance remains limited. Alkali stress is more detrimental to plant growth than neutral salinity stress, particularly by reducing the uptake of essential nutrients and exacerbating oxidative stress14. Proteins with the evolutionarily conserved EF-hand motif—such as calmodulin and calmodulin-like—sense stress-triggered Ca2+ signals, thereby coordinating stress response pathways15,16. However, it is unknown whether novel Ca2+-binding proteins (CaBPs) specifically respond to alkali stress and enhance nutrient uptake for plant survival. In addition, horizontal gene transfer (HGT) events have been shown to contribute to plant adaptation17,18, yet the role of HGT-derived genes in the adaptation of Triticeae CWRs to saline–alkaline stress remains unclear.

To address these gaps, we de novo assembled a chromosome-scale genome for H. brevisubulatum (I genome) using PacBio high-fidelity (HiFi) long-read sequencing and high-throughput chromatin conformation capture (Hi-C). We performed whole-genome resequencing of 38 accessions from seven I-genome species, followed by comparative and population-based structural variations analyses with Hordeum marinum and Hordeum vulgare. Our findings reveal remarkable genomic rearrangements that have shaped the evolution and environmental adaptation of the I-genome species. We identified and functionally verified key candidate genes contributing to the superior alkali and salinity stress resistance of H. brevisubulatum. Furthermore, we evaluated the genetic potential of the I genome for crop improvement via a newly synthesized hexaploid species, Tritordeum (AABBII, 2n = 6x = 42), which features a substituted I-subgenome and shows enhanced alkaline–saline tolerance. Our study lays the foundation for Hordeum I-genome research, enhancing the potential for transferring elite I-genome alleles into future wheat and barley breeding programs to address global food security challenges under climate change.

Results

Genome assembly and composition

A superior alkaline–saline tolerant H. brevisubulatum line (accession PI 531775, 2n = 2x = 14) was selected for reference genome sequencing (Extended Data Fig. 1a,b). The genome size and ploidy were determined through flow cytometry analysis (Extended Data Fig. 1c and Supplementary Fig. 1). This I-genome accession exhibits a perennial growth habit, early spring regrowth and increased tiller numbers compared to the previous year’s growth (Extended Data Fig. 1d–f). To generate a high-quality reference genome, we used 45× PacBio HiFi sequencing and 158× Hi-C reads (Extended Data Fig. 1g). The assembly yielded 3.47 Gb of genome sequence, with a contig N50 size of 82.35 Mb and a scaffold N50 size of 476.19 Mb (Supplementary Table 1), accounting for 96.66% of the estimated genome size (3.59 Gb by K-mer; Extended Data Fig. 1h). Of the total assembly, 3.39 Gb (94.35%) were anchored and oriented into seven pseudo-chromosomes (1I–7I). The long terminal repeat (LTR) assembly index score of 22.98 for this assembly surpasses that of H. marinum (12.94) and H. vulgare (15.91) (Supplementary Table 2), indicating superior genome contiguity. In total, 73,672 protein-coding genes were annotated, with 99.4% mapped to the seven pseudo-chromosomes. Benchmarking Universal Single-Copy Orthologs (BUSCO) analysis identified 94.64% of complete Poales genes in the H. brevisubulatum assembly, further confirming its completeness (Supplementary Tables 1 and 2). Gene distribution analysis revealed a concentration of annotated genes in the distal chromosomal regions, where transposable elements were less abundant (Fig. 1a).

Fig. 1: Evolution, feature, and genomic diversity of the I genome.
figure 1

a, Genomic features of H. brevisubulatum. The proportion of LTR-RTs (including Gypsy and Copia, two major families of LTR-RTs), other repeats, genes, unannotated regions and centromeric repeats are represented. b, Phylogenetic relationship of 19 genomes, including I genome and other available Triticeae genomes. The divergence times among different species are labelled on the branch point. The numbers on each apex represent gene families’ expansion (dark blue) and contraction (orange). The star indicates H. brevisublatum assembled in this study. Divergence time points among Hordeum species are marked in red. The left lower part represents the geographical distribution of H. brevisubulatum accessions with different ploidies (Supplementary Data 10). Pink dots, 2n = 14; blue dots, 2n = 28; green dots, 2n = 42; brown dots, ploidies unknown. c,d, The occurrence of LTR-RTs (c) and DNA transposons (d) in the upstream and downstream (~5 kb) gene regions in the genomes of H. brevisubulatum, H. marinum and H. vulgare. e, Percentage of tandem duplication (TD), proximal duplication (PD) and transposed duplication (TRD) with (w/) and without (w/o) transposable elements (TE) in the genomes of H. brevisubulatum, H. marinum and H. vulgare. f, Synteny analysis among H. brevisubulatum (representing the I genome), H. vulgare (representing the H genome) and H. marinum (representing the Xa genome) by chromosomal positions. Inversions are highlighted in green. g, The phylogenetic tree of 62 accessions was constructed using SNPs of whole-genome single-copy genes. The 61 Hordeum accessions are clustered into five groups: H. brevisubulatum, H. roshevitzii, H. bogdanii, the New Word group and H. vulgare. T. elongatum was set as the out-group. The sizes of the dots on the nodes are proportional to bootstrap support values. h, PCA of 61 accessions (including 12 H. brevisubulatum, 15 H. bogdanii, 3 H. roshevitzii, 9 accessions from the New World and 22 barley accessions). The classifications of species are represented by shapes or colours.

Source data

Full size image

Genomic evolution and features

The evolutionary relationship of the H. brevisubulatum I genome with other sequenced genomes within Triticeae (Secale cereale, Thinopyrum elongatum, Aegilops tauschii, Triticum urartu, Aegilops speltoides, Triticum aestivum, H. vulgare, Hordeum spontaneum and H. marinum), Aveneae (Avena sativa), and Poeae (Puccinellia tenuiflora) was resolved using a maximum likelihood phylogenetic tree based on 268 single-copy genes (Fig. 1b). H. brevisubulatum (I genome) was more closely related to H. marinum (Xa genome) than to H. vulgare (H genome) within the Hordeum clade. Divergence time estimation suggests that H. brevisubulatum split from H. marinum (sea barleygrass) approximately 5.34 Ma during the latest Miocene. Synonymous substitution distribution and fourfold degenerate site transversion (4dTv) confirmed a whole-genome duplication event in the Pooideae ancestor and identified a recent burst of gene duplication specific to Triticeae species19 (Extended Data Fig. 1i, j).

The H. brevisubulatum genome has a lower transposable element content (83.60%, 2.90 Gb) than H. marinum (86.74%, 3.31 Gb) and H. vulgare (88.44%, 3.71 Gb) (Supplementary Table 3), corresponding to a reduction of approximately 410 Mb and 810 Mb, respectively. As transposable element content substantially influences genome size in Triticeae20,21, this reduction likely makes H. brevisubulatum the smallest among sequenced diploid Hordeum species. Despite the lower transposable element content, the LTR retrotransposon (LTR-RT) subfamilies Fatima, Inga, Nusif, and Angala expanded substantially in H. brevisubulatum compared to H. marinum and H. vulgare (Extended Data Fig. 1k). Comparative analysis of transposable element insertion sites among the three Hordeum reference genomes revealed distinct patterns in H. brevisubulatum. LTR-RTs occurred most frequently adjacent to gene bodies (Fig. 1c), while DNA transposons were enriched in the upstream and downstream 5 kb regions but occurred less frequently near gene bodies (Fig. 1d). In addition, H. brevisubulatum exhibited higher proportions of tandem duplication, proximal duplication and transposed duplication (Supplementary Table 4), correlating with increased transposable element occurrence (Fig. 1e). Orthologous gene analysis across Triticeae, Aveneae and Poeae identified 3,602 genes unique to H. brevisubulatum (Supplementary Table 5 and Extended Data Fig. 1m). These genes were enriched in Gene Ontology (GO) categories related to cellular nitrogen metabolism and cellulose catabolism (Supplementary Data 1), suggesting their role in the adaptive evolution of H. brevisubulatum. It is worth noting that the FAR1 transcription factor gene family, associated with environmental adaptation22, exhibited exceptional expansion in H. brevisubulatum (Extended Data Fig. 1l and Supplementary Data 2), with several genes showing differential expression under salt or alkali stress (Supplementary Data 2).

Structural variation analysis between H. brevisubulatum, H. marinum and H. vulgare genomes revealed that megabase-scale inversions accounted for 78.6% of all inversions (Fig. 1f). Notable gene expansion events associated with large structural variations were observed on chromosomes 1I (8.27–46.16 Mb) and 7I (29.59–65.33 Mb; 504.96–524.83 Mb) of the H. brevisubulatum genome compared to H. marinum and H. vulgare. GO enrichment analysis showed that these expanded genes are related to stress, stimulus and defence responses and disease resistance (Supplementary Data 3). Furthermore, genes within 300 kb of inversion breakpoints in H. brevisubulatum were enriched in defence-related functions (Supplementary Data 4), such as those in the NB-ARC (nucleotide-binding adaptor shared by APAF-1, R proteins and CED-4) family (Supplementary Data 5) associated with disease resistance23. This enrichment was absent in the corresponding regions in H. marinum and H. vulgare (Supplementary Data 4). Comparisons with the 20-barley pan-genome revealed additional structural-variation-related gene expansions in H. brevisubulatum (Supplementary Fig. 2 and Supplementary Data 6). It is worth noting that a gene (HorBre03H01G096680) orthologous to Cul4 in barley and tin3 in wheat, both associated with tillering24,25,26, was identified near the breakpoint of a 674 kb inversion on chromosome 3IL.

Population structure and genomic diversity

To explore the genetic diversity and evolutionary patterns of the I genome, we resequenced 38 diploid accessions, including 12 H. brevisubulatum, 15 Hordeum bogdanii, three Hordeum roshevitzii and eight from four New World species (Hordeum chilense, Hordeum muticum, Hordeum pusillum and Hordeum stenostachys) (Supplementary Data 7). In addition, we incorporated previously published resequencing data for 22 H-genome accessions and one I-genome accession27,28,29,30 (Supplementary Data 7), identifying 2,209,666 high-quality single-nucleotide polymorphisms (SNPs) across all accessions. Phylogenetic analyses (Fig. 1g), principal component analysis (PCA; Fig. 1h and Extended Data Fig. 2b) and population structure assessments (Extended Data Fig. 2a) revealed three major clades among the I-genome species, reflecting species-level differentiation and evolutionary relationships. Clades I and II comprised H. brevisubulatum, H. bogdanii and H. roshevitzii from the Old World, while the more divergent Clade III contained New World species. It is worth noting that three accessions (IDs 27, 201-1 and 19-2) previously classified as H. brevisubulatum in the US National Plant Germplasm System (NPGS) were reassigned to H. bogdanii based on phylogenetic and field-based phenotypic data (Supplementary Data 7). Similarly, two highly heterozygous accessions (16-2 and 66-2) previously recorded as H. brevisubulatum were found to have predominantly H. bogdanii genomic background in the population structures analyses (Extended Data Fig. 2a). Discrepancies in clustering for some accessions (for example, 407-1, 1781-1 and 331-1) highlight the need for further investigation, including morphological examination and chloroplast genome sequencing.

The outcrossing H. brevisubulatum exhibited the highest nucleotide diversity (π = 3.43 × 10−4) among the analysed I- and H-genome species (Supplementary Table 6). Evidence of introgression from H. brevisubulatum to H. bogdanii (Extended Data Fig. 2c) was supported by shared derived alleles (Extended Data Fig. 2d). Selective sweep analyses (composite likelihood ratio and Tajima’s D) identified 98.92 Mb of selective sweep regions in the H. brevisubulatum genome, encompassing 3,162 protein-coding genes enriched in stress response pathways (Extended Data Fig. 2e,f and Supplementary Data 8). One notable gene, SOS1, part of the highly conserved salt resistance signalling pathway31,32, exhibited pronounced selective sweep signals in I-genome species (five copies in most Old World species and four copies in New World species) and the Xa reference genome (four copies) but remained a single copy in the 20 H. vulgare pan-genome accessions (Extended Data Fig. 2g and Supplementary Data 9). These findings underscore the adaptive evolution of H. brevisubulatum to stress environments.

Duplication and HGT facilitate saline–alkaline adaptation

In stress screening experiments, H. brevisubulatum showed superior tolerance to alkaline salinity compared to other Hordeum species (Extended Data Fig. 3a), exhibiting higher survival rates and biomass (Extended Data Fig. 3b,c) under both alkaline (pH ~9.2) and neutral (pH ~7.0) salinity. It also maintained nitrate uptake more effectively than H. vulgare under these stress conditions (Extended Data Fig. 3d). Genomic analyses identified several features that may have contributed to its enhanced saline–alkaline stress tolerance.

Duplication of stress-sensing module enhances nitrogen use efficiency for alkali tolerance

Using H. brevisubulatum as a reference for the I genome and H. vulgare cv. Morex33 for the H genome, we examined gene copy number variations (gCNVs) across 39 I-genome and 22 H-genome accessions (Supplementary Data 7). Genes expanded in the I genome were enriched notably for GO terms associated with salt and alkali stress adaptation (Supplementary Data 11). Among these, orthologues to wheat aluminum-activated malate transporter134 and rice H+-ATPase35 (TaALMT1 and OsAHA3; Supplementary Table 7), typically single-copy genes in H. vulgare and H. marinum, were duplicated in H. brevisubulatum. Calcium-binding EF-hand protein-encoding genes (CaBPs) exhibited extensive expansion from two copies in H. vulgare to five in H. marinum and eight in H. brevisubulatum (Fig. 2a and Supplementary Table 7). These eight copies formed tandem repeats within a 0.3 Mb region (Extended Data Fig. 3e). All CaBP copies, except HbreCaBP.0750, were markedly upregulated early in the salt and alkali stress treatments, particularly under alkali stress, while HbreCaBP.0750 expression was downregulated under both stresses (Fig. 2b). To verify their function, we cloned one CaBP (HbreCaBP.0710) and overexpressed it in H. vulgare cv. Golden Promise, which markedly improved alkali tolerance (Fig. 2c), doubling biomass compared to the wild type (WT) under alkali stress (Extended Data Fig. 3f). In addition to increasing biomass under stress conditions, the transgenic barley exhibited an 87% increase in nitrogen uptake and a 68% enhancement in root nitrogen content (Fig. 2d,e). Moreover, soil-cultivated HbreCaBP.0710 transgenic barley produced up to 24.6% more tillers (Fig. 2f,g) and achieved a 28.7% higher grain yield (Fig. 2h) compared to the WT. These results confirm the positive role of CaBPs in alkali stress tolerance. The extensive expansion of CaBPs in H. brevisubulatum may have contributed to adaptation to alkali environments.

Fig. 2: Novel CaBP-NRT2 module duplicates enhance alkali tolerance via improvement of NUE.
figure 2

a, The presence of CaBP duplicates in the I-genome species, H. marinum and H. vulgare accessions. b, Heat map showing the expression patterns of HbreCaBPs in response to alkali and salt stresses in the roots of H. brevisubulatum. c, Phenotype of HbreCaBP transgenic plants under alkali stress. CK, control plants. Scale bars, 3 cm. d,e, Nitrate uptake rate (d) and nitrate nitrogen content (e) of HbreCaBP transgenic plants under alkali stress for 2 days. f, Phenotype of soil-grown HbreCaBP transgenic plants under alkali stress for 21 days. Scale bars, 5 cm. g,h, Comparison of tiller number (g) and grain weight per plant (h) between HbreCaBP transgenic plants and WT under alkali stress. i, The presence of NRT2 duplicates in the I-genome species, H marinum and H. vulgare accessions. j, The phenotype of HbreNRT2 transgenic plants under alkali stress. Scale bars, 3 cm. k,l, Nitrate uptake rate (k) and nitrate nitrogen content (l) of HbreNRT2 transgenic plants under alkali stress. m, A thematic model of stress sensor-responder module CaBP-NRT2. Boxplots show the median, 25th−75th interquartile range (IQR) and minima and maxima (whiskers) with n = 5 biological replicates for d, k and l; n = 6 for e; n = 8 for g; and n = 10 for h. Significant differences were calculated using a two-tailed Student’s t-test (*P < 0.05; **P < 0.01; ***P < 0.001; NS, non-significant).

Source data

Full size image

Given HbreCaBPs’ function in alkali tolerance (Fig. 2c–h), we further screened for gene interactions with HbreCaBP.0710 in H. brevisubulatum through yeast two-hybrid (Y2H) assays (Supplementary Table 8), which identified a high-affinity nitrate transporter NRT2 (Extended Data Fig. 3j and Supplementary Note 1). NRT2 is also present in the gCNV gene sets and has expanded extensively in H. brevisubulatum (25 copies) compared to other Poaceae members (H. vulgare, 11 copies; H. marinum, 19 copies; T. aestivum, 16 copies in A-subgenome, 13 in B-subgenome and 16 in D-subgenome; Zea mays, 7 copies; Oryza sativa, 4 copies), Asteraceae (Lactuca sativa, 13 copies; Cynara cardunculus, 8 copies; Artemisia annua, 11 copies) and Fabaceae (Medicago truncatula, 3 copies; Glycine max, 7 copies; Pisum sativum, 4 copies) (Supplementary Data 12). In Triticeae, NRT2 duplicates were predominantly located on chromosome 6, spanning approximately 1 Mb at the distal regions of chromosome 6S (Fig. 2i and Extended Data Fig. 3g). It is worth noting that NRT2 genes were induced exclusively under alkali stress (Extended Data Fig. 3h). Overexpression of one NRT2 (HbreNRT2.6420) in H. vulgare cv. Golden Promise enhanced alkali tolerance in transgenic lines (Fig. 2j and Extended Data Fig. 3i), boosting nitrate uptake rate by 24% and enhancing nitrate-nitrogen content by 63% (Fig. 2k,l), consistent with the higher nitrate uptake phenotype of HbreCaBP.0710 transgenic lines (Fig. 2d,e). These findings highlight the critical role of the CaBP-NRT2 complex in maintaining nitrate uptake and nitrogen content under alkali stress (Fig. 2m), potentially contributing to the adaptation of H. brevisubulatum to alkali stress.

Horizontally transferred gene from fungus enhances alkaline–saline tolerance via reactive oxygen species modulation

Reactive oxygen species (ROS) regulation is crucial for plant saline–alkaline tolerance. In H. brevisubulatum, ROS levels stabilized after alkali or salt stress, whereas cultivated barley (H. vulgare) exhibited excessive ROS accumulation (Extended Data Fig. 4a,b and Supplementary Note 2). Comparative analysis of the three Hordeum reference genomes (I, H and Xa) revealed extensive chromosomal inversions and translocations at the distal regions of chromosome 7L (Fig. 3a), including an inversion breakpoint containing four gene duplicates (HbreFhb7.1–HbreFhb7.4), homologous (>97.8%) to Fhb7 (ref. 18) (Fig. 3b, Supplementary Fig. 3 and Extended Data Fig. 4c–e). Originally acquired via HGT from the endophytic fungus Epichloë to T. elongatum and involved in Fusarium head blight resistance18, Fhb7 belongs to the GST superfamily and has no homologues outside Triticeae (Supplementary Fig. 3, Supplementary Data 13 and Supplementary Note 3). We isolated Epichloë strains from H. bogdanii seeds and shoots (Extended Data Fig. 4c,d), supporting the proposed HGT event18. Moreover, Fhb7 exhibited clear copy number variations (CNVs) across the Hordeum species, preferentially retained in Old World H. brevisubulatum and H. bogdanii (2–4 copies), with only one copy in New World H. pusillum and none in H. vulgare or H. marinum (Fig. 3c). We found that HbreFhb7s had various cis-elements involved in abiotic stress response in their promoter regions (Extended Data Fig. 4f), induced by alkali and salt stresses (Fig. 3d). Co-expression analysis linked HbreFhb7 with key transcription factors (MYB, WKRY, ERF and bHLH) and several genes encoding protein kinases (Extended Data Fig. 4g).

Fig. 3: HGT and duplication of HbreFhb7 enhance stress adaptation in H. brevisubulatum.
figure 3

a, HbreFhb7 next to the inversion in the ends of chromosome 7L among three Hordeum genomes; this figure is shown by gene order. b, Phylogenetic relationship between Fhb7 from the reported Triticeae wild species including H. brevisubulatum and homologues in Epichloë species. The asterisk (*) represents the endophytic fungus Fhb7 isolated from Old World species H. bogdanii. The HGT-derived and duplicated Fhb7 in H. brevisubulatum was labelled with a dashed box. c, Distribution of the four copies of HbreFhb7 among Hordeum species. d, Radar map indicating expression pattern of HbreFhb7s in response to alkali and salt stresses in the roots of H. brevisubulatum. e, Phenotypes of HbreFhb7 transgenic plants under alkali and salt stresses. Scale bars, 5 cm. f,h, Comparison of ROS accumulation in roots of HbreFhb7 transgenic plants and the WT Golden Promise under salt (f) and alkali (h) stresses. Scale bars, 100 μm. g,i, Quantification of the relative fluorescence intensity of the roots stained with H2DCFDA under salt (g) and alkali (i) stresses. j, Work model indicating that HGT and duplication events of HbreFhb7 innovate saline–alkaline tolerance by suppressing ROS accumulation. The error bars represent mean ± s.d. with n = 7 biological replicates for g and i. Significant differences were calculated using a two-tailed Student’s t-test (****P < 0.0001; NS, non-significant).

Source data

Full size image

Functional validation of HbreFhb7 was conducted via overexpression of HbreFhb7.1 (preserved in most H. brevisubulatum accessions and highly induced upon alkali or salt stress) in barley cv. Golden Promise. Transgenic plants showed enhanced biomass under alkali and salt stress (Fig. 3e and Extended Data Fig. 4h) and reduced ROS accumulation (Fig. 3f–i), with no significant differences under control conditions, supporting HbreFhb7’s role in maintaining cellular redox homeostasis under alkali and salt stresses. The Y2H assays identified interactions between HbreFhb7 and several redox-related enzymes, including RBOH, GSTU6, GSTF11 and USP (Supplementary Table 9). We cloned the C-terminals of 10 RBOHs from H. brevisubulatum, confirming that HbreRBOHB1 interacted with HbreFhb7 (Extended Data Fig. 4i and Supplementary Note 1). A transient expression assay verified that ROS levels markedly declined when RBOHB1 was co-expressed with HbreFhb7 (Extended Data Fig. 4j and Supplementary Note 2). Figure 3j summarizes the genetic origin of HbreFhb7 and its biological mechanisms in conferring alkali and salt stress tolerance through regulating ROS production.

Potential of the Hordeum I genome for wheat improvement

The genetic potential of the Hordeum I genome for enhancing wheat stress tolerance was validated through Tritordeum8, a newly synthesized hexaploid species (AABBII) developed by replacing wheat’s D subgenome with the Hordeum I genome (Fig. 4a). Compared to hexaploid bread wheat ‘Chinese Spring’ (AABBDD), Tritordeum accession HT621 showed substantially improved tolerance to alkali and salt stresses (Fig. 4b and Extended Data Fig. 5a,b). Consistent with the Hordeum I genome’s inherent adaptability for maintaining nitrogen use efficiency under stress, HT621 exhibited a 47% increase in nitrate uptake rate and a 53% increase in nitrate nitrogen content compared to Chinese Spring under these conditions (Fig. 4c and Extended Data Fig. 5c).

Fig. 4: The Hordeum I genome holds great potential for improving stress tolerance.
figure 4

a, Graphic illustration of the creation of Tritordeum (AABBII). b, Comparison of phenotypic changes between Tritordeum HT621 and Chinese Spring under salt or alkali stress. Scale bars, 10 cm. c, Nitrate uptake rate of HT621 and Chinese Spring under salt or alkali stress. d, Ternary plot showing the relative expression levels of ancestral triads in Tritordeum HT621 roots. Each dot represents a gene triad with an A, B and I coordinate; balanced triads are represented by blue dots in the centre. e, Violin plots of the expression of ancestral triads in the A, B and I subgenomes of Tritordeum HT621 in response to alkali and salt stresses in roots. The significance of the difference is determined using a Kruskal–Wallis test (***P < 0.001; ****P < 0.0001; NS, non-significant). f, Heat map of the specific genes or duplicates in the I subgenome with markedly up-regulated expression but no homoeologues in the A and B subgenome of HT621. g, The nitrogen metabolic pathway and the expression of representative genes in the subgenomes of HT621 under alkali and salt stresses in roots. h, Thematic model illustrating the effect of substituting the D subgenome with the Hordeum I genome in improving Tritordeum alkali or salt stresses tolerance. Boxplots show the median, 25th−75th IQR, and minima and maxima (whiskers) with n = 5 biological replicates for c. Significant differences were calculated using a two-tailed Student’s t-test (**P < 0.01; NS, non-significant).

Source data

Full size image

Genomic analysis of HT621 revealed 4,248 homoeologous syntenic triads (1:1:1 ancestral triads) across its three subgenomes (Supplementary Data 14). Gene expression was balanced under control conditions (Fig. 4d and Extended Data Fig. 5d,f). However, under alkali and salt stresses, the I subgenome exhibited a more robust transcriptional response (Extended Data Fig. 5e), with higher expression of stress-responsive genes than the A and B subgenomes (Fig. 4e). Up to 300 differentially expressed genes were identified in the I subgenome, mainly 12 h and 24 h after treatment (Extended Data Fig. 5g), substantially enriching stress-responsive pathways (Extended Data Fig. 5h).

Stress-induced genes unique to the I subgenome (Supplementary Data 15) included those encoding Ca2+ signalling sensors (CaBP.0650 and CaBP.0710); transcription factors (WRKYs); Na+-exclusion protein (SOS1); transporters for phosphate, iron and sugar; and genes related to abscisic acid signalling, osmotic stress and ROS regulation, which were particularly responsive under alkali stress (Fig. 4f). Key nitrogen metabolic pathway genes, including NRT2 (high-affinity nitrate transporter), NR1.1 (nitrate reductase), GS1.1 (glutamine synthetase), NADH-GOGAT (NADH-dependent glutamate synthase) and Fd-GOGAT (ferredoxin-dependent glutamate synthase), exhibited higher expression in the I subgenome than in the A and B subgenomes (Fig. 4g). These genetic features likely underpin the enhanced nitrate uptake and nitrogen assimilation in Tritordeum, contributing to its improved salt and alkali tolerance. A model illustrating these pathways are presented in Fig. 4h.

Discussion

In this Article, we assembled a high-quality chromosome-scale H. brevisubulatum genome that can be used as a reference genome for I-genome species. Through comprehensive genomic, evolutionary and genetic analyses, we established a foundation for exploring the evolutionary relationships and adaptive traits of the Hordeum I-genome species. These species, recognized as important CWRs, harbour valuable genetic diversity that can enhance crop resilience and improve adaptive traits in breeding programs9. It is worth noting that H. brevisubulatum is a facultative halophyte with superior salt tolerance despite lacking specialized salt-exclusion structures such as salt glands12 (https://ehaloph.uc.pt/). Consistent with earlier research11, we found that H. brevisubulatum exhibited greater tolerance to neutral salts than Xa- and H-genome species (H. marinum and H. vulgare, respectively) and a unique ability to tolerate alkaline salts (alkali), underscoring its evolution of genetic strategies for both alkali and salt stress tolerance.

Adaptation to salt and alkali stress in the Hordeum I genome

Structural variations substantially contribute to speciation and genome evolution36. The H. brevisubulatum genome revealed notable chromosome inversions, translocation and gCNVs compared to the Xa-genome (H. marinum) and H-genome (H. vulgare) reference genomes. These structural variations, particularly the extensive expansion of stress-related genes, align with observations in other halophytic species, such as Achnatherum splendens37 and Spartina alterniflora38, and likely drive adaptation to harsh environments.

Among these structural variations, gCNVs have emerged as key contributors to plant evolution and adaptation19,39. Unique gCNVs in the H. brevisubulatum genome were enriched with GO terms related to stress adaptation, including several stress-responsive and transporter-encoding genes. The SOS1 gene, a well-characterized locus essential for salt tolerance and Na+ exclusion within the Salt Overly Sensitive (SOS) pathway40, has expanded to five copies in H. brevisubulatum, compared to one copy in H. vulgare and four copies in H. marinum. In the halophytic Arabidopsis thaliana relative Thellungiella salsuginea, SOS1 is a key determinant of halophytism41. Supporting its role in tolerance, studies have shown that mutations in SOS1 in the halophytic grass H. marinum lead to diminished salt tolerance42. The expansion of SOS1 in H. brevisubulatum likely contributes to its enhanced salt tolerance, reinforcing the critical role of this gene in plant adaptation to saline environments. Similarly, genes pivotal for alkali tolerance, such as AHA3 (regulating H+ flux and pH gradients in saline–alkaline stress35,43) and ALMT1 (a malate-GABA (gamma-aminobutyric acid) transporter conferring alkaline tolerance in wheat34), were duplicated in the H. brevisubulatum genome. These duplications, alongside the expansion of other stress-related genes, may help explain the superior salt and alkali tolerance observed in I-genome species compared to Xa- and H-genome species. Unlike other sequenced Hordeum species, which are annuals, H. brevisubulatum is perennial. This difference, coupled with the gCNV-mediated expansion of stress-tolerance genes such as SOS1, AHA3 and ALMT1, suggests broader adaptive traits beyond salt and alkali tolerance, including life history strategies. These findings underscore the importance of continued explorations of H. brevisubulatum to understand its unique genetic and adaptive strategies.

Salt and alkali tolerance mechanisms in H. brevisubulatum

While the molecular mechanisms of salt tolerance in plants are well studied44, the understanding of alkali tolerance mechanisms remains limited14. Calcium signalling plays a crucial role in salt tolerance; however, how alkali stress signals are detected is not fully characterized. One key player is CaBP, an EF-hand Ca2+-binding protein that senses stress signals by binding Ca2+ and transducing them to downstream genes15. Studies have shown that CaBPs can act as negative or positive regulators under saline–alkaline stress16,45. Leveraging the H. brevisubulatum reference genome, we identified an expansion of HbreCaBP. Experimental validation revealed that these proteins are induced under high alkali stress and contribute positively to alkali tolerance in transgenic barley, suggesting their role as Ca2+ sensors and regulators. This expansion may enhance stress responses, indicating a potential target for breeding alkali-tolerant crops. Further studies are needed to explore allelic variations and their contributions to this adaptive trait.

Our findings also revealed that H. brevisubulatum maintains a higher nitrate uptake rate under alkali or salt stress than H. vulgare, indicating this is a critical adaptive strategy. This capability is underpinned by a dramatic expansion of the high-affinity nitrate transporter-encoding gene NRT2, which has 25 copies in H. brevisubulatum—the highest among Poaceae species with sequenced genomes. NRT2 genes are known to facilitate nitrate uptake under low-nitrogen conditions46,47,48, but their interaction with the expanded CaBP suggests a sensor-responder module that contributes to alkali tolerance. This module likely helps mitigate the nutritional imbalance caused by alkali stress, strengthening the plant’s resilience. Further research is required to elucidate the signal transduction and activation between CaBPs and NRT2 under stress. Drawing parallels with studies on alkali-tolerant legumes such as Sesbania cannabina, which uses expanded phosphorus-transporter genes to maintain nutrient uptake49, we propose that H. brevisubulatum uses the expanded CaBP-NRT2 module similarly.

Another important finding is the retention and expansion of HGT-derived Fhb7 genes in H. brevisubulatum. Originally identified as a product of a fungus-to-plant HGT event in T. elongatum18, Fhb7 has since been observed in other Triticeae species50. Symbiotic associations between endophytic Epichloë fungi and Pooideae species likely serve as the basis for the transfer of Fhb7 and other HGT-derived genes51,52. We isolated the Epichloë bromicola strain from Old World H. bogdanii, supporting the occurrence of the HGT event. The presence of Fhb7 in multiple Triticeae species suggests either a shared ancestral HGT event or multiple independent occurrences. Remarkably, H. brevisubulatum has expanded Fhb7 from one to four copies, distinguishing it from other species. These genes are more abundant in Old World I-genome species, indicating dynamic changes in structurally active genomic regions, consistent with the positioning of Fhb7 near a structural inversion in our assembled genome. Beyond its known role in Fusarium head blight tolerance in wheat18, our transgenic studies revealed that HbreFhb7 potentially contributes to alkali and salt tolerance by reducing ROS accumulation. The conservation and expansion of HbreFhb7 likely play a crucial role in H. brevisubulatum’s adaptation to saline–alkaline environments.

Genetic potential of I-genome species for crop breeding

The superior abiotic stress resilience of I-genome species offers immense potential for enhancing barley and wheat breeding programs. Several I-genome species have already been used in traditional hybridization efforts, improving agronomic performance in modern wheat and barley varieties8,9. The H. brevisubulatum genome sequence, combined with insights into its adaptive strategies and alkaline–saline tolerance mechanisms, provides a valuable resource for optimizing the use of I-genome species in crop improvement.

Through phenotypic evaluations of hexaploid Tritordeum (AABBII) under stress conditions, we demonstrated that substituting the D-subgenome with the I-subgenome substantially enhanced alkaline–saline tolerance. It is worth noting that the hexaploid Tritordeum exhibited higher nitrate uptake rates and nitrogen content under alkali and salt stress, supporting the role of the expanded CaBP-NRT2 module as a critical molecular mechanism underlying I-genome species’ stress adaptation. Moreover, we found that I-subgenome genes in the synthesized hexaploid showed enhanced transcriptional responsiveness to stress conditions. This transcriptional regulation—observed under stress conditions but not in the control—highlights an additional evolutionary strategy, complementing genomic structural variations, used by I-genome species for stress adaptation.

Our findings highlight the immense potential of harnessing the genetic diversity of I-genome species into breeding programs through transgenic and crossbreeding approaches to develop stress-resilient crops. Alternatively, de novo domestication of I-genome species with agronomic potential presents a promising avenue for future crop development53. By decoding the Hordeum I genome, we have created a strong foundation for leveraging these genetic resources, which are crucial for advancing sustainable agriculture and ensuring food security, particularly for developing crops that can thrive in soils affected by salt or alkali toxicity. Our study paves the way for establishing an advanced crop breeding platform to improve wheat and barley resilience and productivity.

Methods

Plant materials and genome sequencing

H. brevisubulatum accession PI531775 was sourced from the US NPGS and cultivated in a greenhouse under a 16 h light/8 h dark cycle at 24 °C and 16 °C, respectively. Genome size (~3.65–3.77 Gb) and diploidy were confirmed via flow cytometry using Z. mays cv. B73 (~2.3 Gb) and H. vulgare cv. Morex (~5.04 Gb) as standards. DNA was extracted from 3-week-old leaves using a DNAsecure Plant Kit (TIANGEN). Seven ~15 kb DNA-insert libraries were sequenced on a PacBio SEQUEL II platform at Novogene, generating 162.12 Gb HiFi reads (~45× coverage). In addition, 350 bp DNA-insert libraries were sequenced on an Illumina HiSeq platform (150PE) at BioMarker, generating 579.78 Gb paired-end reads (~161× coverage). Five Hi-C libraries were constructed from cross-linked chromatins using a standard Hi-C protocol and sequenced on the Illumina HiSeq platform at BioMarker, yielding 568.61 Gb Hi-C reads (~158× coverage).

Genome assembly and annotation

Genome size estimation

Genome size was estimated using K-mer distribution analysis with Kmc v3.2.1 (ref. 54) and genomescope v1.0.0 (https://github.com/schatzlab/genomescope/). The K-mer value used for the genome survey was 21.

Genome assembly

Contig-level assembly was performed using hifiasm v0.14.0 (ref. 55) with HiFi reads as input, applying default parameters. Read error correction used a Falcon-like algorithm integrated into the hifiasm pipeline. Hi-C scaffolding was conducted by aligning Hi-C sequencing data to the assembled genome using Juice_tools v1.8.9 (ref. 56), followed by chromosome construction with 3D-DNA v180419 (ref. 57).

Gene annotation

Gene structures were predicted using the GETA v2.4.5 pipeline (https://github.com/chenlianfu/geta), integrating ab initio gene prediction, homologous proteins and transcriptome data using Augustus v2.5.5 (ref. 58), Trimmomatic v0.40 (ref. 59), HiSAT2 v2.2.1 (ref. 60) and genewise v2.4.1 (https://www.ebi.ac.uk/~birney/wise2/).

Repetitive sequences were identified and masked to reduce background noise from frequently duplicated transposable elements. The RNA-sequencing (RNA-seq) data underwent trimming with Trimmomatic before alignment to the reference genomes using HISAT2. Subsequently, the GETA pipeline calculated coverage thresholds specific to each alignment region based on the sequencing depth. Transcripts falling below the coverage cutoff were filtered out to prioritize reliable introns and optimize transcripts. The remaining high-quality transcripts were subjected to open reading frame (ORF) prediction in TransDecoder v 5.5.0 (https://github.com/TransDecoder/TransDecoder). Guided by earlier predicted intron and exon structures, the gene models underwent iterative training using Augustus v2.5.5 (ref. 58) to achieve the best score. In addition, homologous proteins from A. thaliana, O. sativa, Actinidia chinensis from Phytozome v12 (https://phytozome.jgi.doe.gov/), Brachypodium distachyon, Sorghum bicolor and Z. mays from EnsemblPlants release-50, and previously published barley27 and wheat61 proteins were used for further protein identification via Genewise v2.4.1.

Repeat annotation

Genome-wide transposable element annotation was conducted using EDTA v2.0.0 (ref. 62). Transposable elements, including LTR-RTs and DNA transposons, were identified in H. brevisubulatum, H. marinum and H. vulgare. RepeatMasker v4.1.5 (ref. 63) was used to align the genome assemblies of these species against the clariTeRep repeat library in CLARI-TE (https://github.com/jdaron/CLARI-TE), enabling LTR subfamily classification. The LTR assembly index was evaluated for H. brevisubulatum, H. marinum, H. vulgare, T. elongatum and Dasypyrum villosum by integrating results from LTR_Finder v1.07 (ref. 64) and LTR_harvest v1.6.1 (ref. 65) using LTR_retriever v2.9.0 (ref. 66). To infer centromere positions, Gypsy retrotransposons Cereba and Quinta, enriched in centromeric regions of Triticum species61,67, were analysed42,68. Putative centromeric regions in H. brevisubulatum were identified based on the enrichment of RLG_famc8.3 (Cereba), RLG_famc8.1 (Quinta) and RLG_famc8.2 (Quinta), following methodologies established in previous studies42,68.

BUSCO assessment

Genome completeness was assessed using BUSCO analysis. The completeness evaluation for H. brevisubulatum, H. marinum, H. vulgare, T. elongatum and D. villosum was performed using compleasm v0.2.5 (ref. 69) with the Poales_odb10 database.

Comparative genomics and genome evolution analysis

Phylogenetic tree construction

Orthologous gene families across 15 species—H. brevisubulatum, H. marinum42, H. vulgare cv. Morex33, H. vulgare B1K041227, T. aestivum61 (subgenomes A, B and D), A. speltoides70, T. urartu71, A. tauschii72, T. elongatum18, S. cereale73, A. sativa74 (subgenomes A, C and D), P. tenuiflora75, B. distachyon76, O. sativa77 and Ananas comosus78—were identified using OrthoFinder v2.5.4 (ref. 79) after excluding sequences shorter than 100 amino acids. All-vs-all Blastp v2.2.31 for the 268 identified single-copy genes identified only 10 gene pairs with minimal redundancy (4%) with high sequence identity (>50%) among 11 unique genes, which was deemed negligible for species phylogeny.

Three approaches were used to construct the species tree80,81. First, each coding sequences (CDS) of the 268 single-copy homologous genes was aligned using MUSCLE v3.8.31 (ref. 82) and trimmed with trimAL v1.2 (ref. 83) (‘-automated1’). IQTREE2 v2.1.4-beta (ref. 84) was used to construct 268 trees with self-estimated best substitution models, and a coalescent-based species tree was generated with Astral v 5.7.8 (ref. 85). Second, concatenated protein sequences from OrthoFinder were aligned using MUSCLE v3.8.31, converted to CDS alignments using PAL2NAL v14 (http://www.bork.embl.de/pal2nal/index.cgi#RunP2N) and trimmed using trimAL v1.2 (‘-automated1’). ModelTest-NG v 0.1.6 (ref. 86) selected the best-fit substitution model (GTR + I + G4) for maximum likelihood tree construction using PhyML v3.088 (ref. 87). Third, Bayesian phylogenies were constructed with MrBayes v3.2.7a using the general time-reversible (GTR) model with six substitution types (nst=6) and invgamma rate distribution, running two independent Markov chain Monte Carlo chains every 100 trees for 200,000 generations. Convergence was confirmed with standard deviations below 0.05 and effective sample size (ESS) above 200. The verified maximum likelihood phylogenetic tree was visualized using iTOL (https://itol.embl.de/).

Phylogenomic dating and branch length analysis were conducted using MCMCtree in PAML v4.9 (https://github.com/abacus-gene/paml), using an auto-correlated JC69 substitution model with uniform priors on relative node times. Calibration times were derived from known species divergence dat in Timetree (http://www.timetree.org/).

Gene family analysis

Gene family expansions and contractions were detected using CAFE v4.2.1 (ref. 88) with a significance threshold (P < 0.05). Orthologous families identified via OrthoFinder were filtered to exclude those with abnormal gene counts. Filtered families and the ultrametric tree were input into CAFE following the tutorial (https://hahnlab.github.io/CAFE/src_docs/html/index.html), optimizing the birth and death rate parameter (λ) across all families with the ‘-s’ and ‘-t’ parameter. Gene family identification was conducted using Hmmer v3.3.2 (ref. 89) and Blastp. Model HMM files from the Pfam database (v35.0) were used to detect family-specific protein sequences (e < 0.01). Homologous protein sequences were identified with Blastp (e < 1 × 10–10). Phylogenetic trees were constructed with RAxML-NG v1.2.2 (ref. 90), while candidate gene family proteins were validated against Pfam and National Center for Biotechnology Information Conserved Domains databases. Transcription factor families were identified using PlantTFDB (http://planttfdb.gao-lab.org/prediction.php). Predicted transcription factors were manually validated by confirming associated domains in the InterPro database (https://www.ebi.ac.uk/interpro/).

Genome collinearity and duplication analysis

Structural variations among the genomes of H. brevisubulatum, H. marinum and H. vulgare cv. Morex were analysed using GeneSpace v1.1.4 (ref. 91), which leverages orthogroup identification tools such as OrthoFinder, MCScanX and Diamond v 2.1.8 to define syntenic regions92. The H. vulgare cv. Morex genome was used as the reference, with inversions highlighted in green. Plots were generated based on both physical and gene rank order. Collinear gene pairs identified by MCScanX v1.0.0 were further processed to calculate whole-genome 4dtv and synonymous substitution values using wgd v1.1.2 (ref. 93). Micro-synteny analyses of HbreCaBP and HbreNRT2 genes between species were performed with MCScanX (default parameters) and visualized using jcvi v1.2.10 (https://github.com/tanghaibao/jcvi). Gene duplications in H. brevisubulatum, H. marinum and H. vulgare cv. Morex were classified into five types—whole-genome duplication (WGD), tandem duplication, proximal duplication, transposed duplication and dispersed duplication—using DupGen_finder v1 (ref. 94) with default settings.

Resequencing and population analysis

Resequencing reads mapping and variant calling

Thirty-eight I-genome accessions from populations across China (Nei Monggol, Xinjiang, Gansu, Qinghai and Tibet provinces), Kazakhstan, Iran, South Africa, the United States, Peru, Uruguay and Argentina were identified as diploids using flow cytometry, with H. brevisubulatum PI531775 or Z. mays cv. B73 as the reference standard (Supplementary Data 7). Leaves were collected, and DNA libraries were constructed using a DNAsecure Plant Kit (TIANGEN). Sequencing was performed on the Illumina HiSeq-4000 platform in 150 bp paired-end mode. In addition, raw resequencing data for 22 H-genome accessions and 1 I-genome species were obtained (Supplementary Data 7). Quality control of raw sequencing reads was carried out using FastQC v0.12.1 (ref. 95), followed by adapter and low-quality read trimming using Trimmomatic v0.40 (ref. 59). Clean reads, with an average genome coverage depth of 16.3× (Supplementary Data 7), were aligned to the H. brevisubulatum genome using a modified Burrows-Wheeler Aligner (BWA) program from the Sentieon package (v202112.02). The Binary Alignment/Map format (BAM) files underwent an accelerated Genome Analysis Toolkit (GATK)-compliant pipeline, adhering to the ‘Broad Institute Best Practices Pipeline’. Quality control metrics included MeanQualityByCycle (MQ > 30.0), QualDistribution (QD > 2.0), GCBias, AlignmentStat and InsertSizeMetricAlgo. Duplicates were identified and removed using LocusCollector and the Dedup algorithm. Uniquely mapped reads with mapping quality >30 were selected (via samtools) and processed with IndelRealigner for accurate small indel detection (1–10 bp). Base quality recalibration was performed using QualCal. Variants were called using the HaplotypeCaller algorithm to identify SNPs and indels, with joint calling performed using GVCFtyper. Variants with low depths, high repetition or more than 15% missing data were filtered from the raw SNP Variant Call Format (VCF) file using VCFtools v0.1.16 (ref. 96) with the following parameters: minDP = 4, minQ = 30, max-missing = 0.85, mac = 4, max-alleles = 2 and min-alleles = 2. Further filtering of SNPs was performed using BCFtools v 1.19 (ref. 97) with the parameters ‘-e GT == ‘het’ & DP < 15’. Linkage disequilibrium pruning was conducted with Plink v 1.9.0 using the parameters maf = 0.05 indep = 50 5 2. In total, 810,571 SNPs from single-copy genes were extracted to reconstruct phylogenetic relationships among the 61 accessions. In addition, 1,595,091 SNPs within the CDS region of genes from these accessions were selected for further analysis.

Phylogeny analysis

The SNP VCF file was converted into a DNA sequence alignment file using a custom Perl script. ModelTest-NG v 0.1.6 (ref. 86) was used to identify the best-fit nucleotide substitution model ‘GTRGAMMAX’. Maximum likelihood tree was constructed using RAxML-NG v1.2.2, and phylogenetic tree was visualized with iTOL v7.

PCA

VCFtools was used to convert the VCF file into PLINK format, and PCA was conducted using GCTA v1.94.1 (ref. 98). The top three principal components were visualized and retained for further population structure analysis.

Population genetic structure

Population structure (61 accessions) was inferred using Admixture v1.3.0 (ref. 99) with ancestral population sizes set from K = 2 to K = 9. The optimal structure was determined at K = 5, with the lowest cross-validation error (0.25).

Genetic diversity and introgression estimation

Genome-wide Tajima’s D and nucleotide diversity (θπ) were calculated using VCFtools with a 50 kb sliding window. Composite likelihood ratio values for 50 kb windows were computed using SweeD v3.3.1 (ref. 100). Gene flow was modeled using TreeMix v 1.13 (ref. 101) with the following parameters: -se, –bootstrap and -k set to 1,000. The number of migration edges (-m) ranged from 1 to 10, with three iterations of each. OptM v0.1.8 (ref. 102) was used to determine the optimal number of migration edges. Introgression was further analysed using Patterson’s D-statistic (ABBA-BABA test), implemented in Dsuite v 0.5 (ref. 103).

Population CNVs

Population CNVs were identified using popCNV v1.1.0 (https://github.com/sc-zhang/popCNV), which estimates gene copy number based on sequencing depth and guanine-cytosine (GC) content correction. Sequencing depth was calculated using Mosdepth v0.3.3 (ref. 104) with a 1 kb window size. Population CNVs were identified for the 39 I-genome accessions using H. brevisubulatum as the reference genome and for the 22 H-genome accessions using H. vulgare cv. Morex33 as the reference genome. Gene clustering into orthologous groups for H. brevisubulatum, H. vulgare cv. Morex33 and H. marinum42 was performed using OrthoFinder. The number of genes in each orthologous group from the I-, Xa- and H-genome accessions were counted to infer CNVs. Candidate gCNVs for 18 H-genome accessions were further validated using Blastn v2.2.31. In cases of discrepancies, Blastn results were prioritized. CNVs for candidate genes were visualized using heat maps generated in the pheatmap package v1.0.12 in R.

Transcriptome sequencing

H. brevisubulatum and Tritordeum seedlings were grown hydroponically in 1/4 Hoagland solution (pH 5.8) under controlled conditions (16 h light at 24 °C, 8 h dark at 16 °C) for 2 weeks. Stress treatments involved exposure to either 200 mM NaCl (pH 5.8) or 200 mM NaHCO3 (pH 9.0), with the solution refreshed every 2 days to maintain pH. Roots were collected at ZT6 (3 h after stress), ZT15 (12 h after stress) and ZT6 day 2 (24 h after stress). Total RNA was extracted using an RNAprep Pure kit (Tiangen) with three biological replicates per time point. RNA-seq libraries (350 bp inserts) were prepared and sequenced using Illumina NovaSeq in 150 nt paired-end mode, generating ~6 Gb of raw reads per sample. For Tritordeum (AABBII), reads were aligned to a concatenated genome combining the AABB genome of Triticum turgidum cv. Svevo105 and the II genome of H. brevisubulatum using STAR v2.7.10 (ref. 106). Low-quality reads were filtered out, and uniquely mapped reads were used to create count matrixes with featureCounts v2.0.4 (ref. 107), normalized to reads per kilobase per million mapped reads (RPKM) via a customized Perl script. Differential expression was conducted using edgeR v4.4.2 with >2-fold change and Benjamini–Hochberg false discovery rate < 0.01. Heat maps were created using the pheatmap package v1.0.12. GO enrichment analysis was performed with the topGO package v2.58.0 and visualized using ggplot2 v3.4.4. Correlation networks were constructed using the weighted correlation network analysis package v1.61 (ref. 108).

Expression bias of Tritordeum subgenomes

Homologous gene expression bias in hexaploid Tritordeum was analysed following previous methodologies109 with slight modifications. In total, 4,248 triads of homologous genes were identified using Blast v2.2.31 and MCSanX v1.0.0, representing 1:1:1 correspondences among the three subgenomes and referred to as ancestral triads. Gene expression levels were normalized by calculating each homologue’s contribution to total expression within its triad. Based on a threshold of 0.15, each homologous gene pair was classified into one of three expression types: dominant, suppressed or balanced. A ternary plot illustrating subgenomic contributions was generated using the ggtern package v3.4.2 in R.

Salt and alkali treatment

H. brevisubulatum, H. vulgare and barley transgenic lines were germinated on wet filter paper before salt and alkali treatment via hydroponic cultivation. After approximately 3 days, seedlings were transferred to 1/4 Hoagland solution (pH 5.8) and grown under controlled conditions (16 h light at 24 °C, 8 h dark at 16 °C). Seedlings of H. vulgare were grown for 2 weeks, while H. brevisubulatum were grown for 3 weeks, until they reached the three-leaf stage. Treatments involved exposing seedlings to 200 mM NaCl (pH 5.8) or 100 mM NaHCO3 (pH 9.0), with the solution refreshed every 2 days to maintain pH. After 7 days of treatment, plants were photographed, and shoot and root fresh weights were measured. For 15N uptake analysis, seedlings cultured for 2 weeks were subjected to stress treatments with 100 mM NaCl or 50 mM NaHCO3, as described above. Following this, seedlings were treated with K15NO3 for 30 min, rinsed with 0.1 mM CaSO4 solution and deionized water, and dried at 70 °C for 3 days. The dried samples were ground into powder, and the 15N content was measured using an isotope ratio mass spectrometer with five biological replicates. The nitrate uptake rate was defined as the amount of 15N absorbed per unit weight of seedling per unit time, calculated by dividing the total 15N content (μmol 15N) of the whole plant by the seedling’s dry weight (g DW) over a 3 h period, following established methods110,111.

Phenotypic evaluations were conducted on H. brevisubulatum, H. marinum (PI 200341 from NPGS), Hordeum murinum (PI 266198 from NPGS), H. vulgare (Golden Promise), Tritordeum (HT621), bread wheat (Chinese Spring) and HbreCaBP transgenic lines. Seeds were sterilized in 70% ethanol for 30 s, washed three times with distilled water, shaken in 10% bleach solution for 15 min and washed thoroughly with distilled water 8 to 10 times. Seedlings were grown in a controlled greenhouse under long-day conditions (16 h light at 24 °C, 8 h dark at 16 °C). Salt (NaCl) and mixed alkali (NaHCO3:Na2CO3 = 5:1, pH 9.2) treatments were applied to soil-grown plants at the third-leaf stage, gradually increasing from 100 mM to 300 mM. After 21 days of treatment, plants were photographed, and shoot and root fresh weights were measured. For transgenic and WT plants, 150 mM mixed alkali treatments were applied for 21 days. Control plants were grown under normal conditions without stress treatments. Phenotypic data, including tiller number and grain yield, were recorded.

Transgenic plant validation

The coding sequences of HbreCaBP, HbreNRT2 and HbreFhb7 (Supplementary Data 16) were cloned into the pCAMBIA1300 vector under their respective native promoters and terminators. Barley transformation was conducted following a previously published protocol112 with minor modifications. Briefly, immature embryos from the barley cv. Golden Promise were isolated and incubated with Agrobacterium for 10 min. Co-cultivation was performed for 2 days on CM medium (1/10 Murashige and Skoog (MS) medium plus 10 g l−1 glucose, 100 μM acetosyringone, and 3.5 g l−1 phytagel). After removing embryo axes, scutella were cultured on the first selection medium (MS medium plus myoinositol 0.35 g l−1, 0.69 g l−1 proline, 1 mg l−1 thiamine HCl, 2.5 mg l−1 dicamba, 1 g l−1 casein hydrolysate, 10 mg l−1 hygromycin B, 160 mg l−1 timentin, 1 g l−1 2-(N-morpholino) ethanesulfonic acid and 3.5 g l−1 phytagel). After 2 weeks, tissues were transferred to the second selection medium, identical to the first but containing 20 mg l−1 hygromycin B. Three weeks later, embryonic calli were cultured on DM medium (first selection medium supplemented with 2.5 mg l−1 CuSO4·5H2O, 1 mg l−1 kinetin, 0.5 mg l−1 6-benzylaminopurine and 0.05 mg l−1 1-naphthaleneacetic acid) at 25 °C under light conditions to induce differentiation. Regenerated shoots were transferred to RT medium (1/2 MS medium with 1 mg l−1 indole butyric acid) and subsequently transplanted into pots. Transgenic plants were confirmed via PCR analysis using gene-specific primers.

Y2H assay

To screen the H. brevisubulatum complementary DNA library, the coding regions of HbreCaBP or HbreFhb7 (Supplementary Data 16) were cloned into the pGBKT7 vector to create a bait construct, which was co-transformed with a prey cDNA library into the yeast strain AH109. The prey cDNA library was generated by fusing cDNAs from various H. brevisubulatum tissues into the pGADT7 vector. Transformants were selected using Quadruple DO medium (SD/-Leu/-Trp/-His/-Ade, -L-T-H-Ade).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.