Haplotype‐Resolved Genome Assembly and Population Genomics Reveal Evolutionary History and Agronomic Traits of Mulberry

ABSTRACT

Mulberry is a representative economic tree species valued for both poverty alleviation and medicinal use. To advance the understanding of mulberry genomics and demography, we assembled high‐quality haploid genomes of two widely cultivated mulberry varieties NS14 and QS1, and analysed 376 accessions from 12 countries, including 39 ancient trees to investigate their origin and spreading. Population genetic analyses revealed that mulberry originated in the Yunnan‐Guizhou Plateau (YGP) and subsequently spread northward from South China to North China. This migration resulted in significant genetic differentiation between northern and southern populations, with the southern populations exhibiting higher genetic diversity. A total of 37 traits related to development and immunity were analysed in 203 accessions, and a genome‐wide association study (GWAS) was used to identify 204 associated loci. Five causal gene haplotypes were pinpointed for key production traits of mulberry trees, including branch pitch, branch length, budburst timing, leaf thickness, and leaf size (leaf area, leaf width, and leaf length). To further explore loci related to disease resistance, we examined the resistance of 538 F₁ hybrids derived from NS14 (resistant) and QS1 (susceptible). Through bulked segregant analysis and GWAS, we identified a G‐type RLK (receptor‐like kinase) tandem gene cluster. Transcriptomic analyses revealed opposite expression trends of these RLK genes in NS14 and QS1, further supporting their role in mulberry blight resistance. Our findings provide valuable genomic and demographic insights for future multi‐purpose breeding efforts in mulberry.

1. Introduction

Mulberry trees (Morus spp.) are crucial to sericulture, a low‐capital, high‐yield economic activity that has significantly contributed to poverty alleviation. More than 2000 years ago, sericulture catalysed trans‐Eurasian trade and was central to the Silk Road. This legacy continued into modern times, as in the 1970s, sericulture ranked as China's second‐largest export earner (Department of Agriculture and Rural Affairs of Guangxi Zhuang Autonomous Region n.d.). More recently, from 2013 to 2020, sericulture helped lift approximately 100 million people out of poverty across numerous countries in Asia (National Bureau of Statistics n.d.; Altman and Farrell 2022). In 2021, the “East Silk, West Shift” policy alone generated over 15 billion CNY in income for Guangxi province in China (Ministry of Agriculture and Rural Affairs of the People's Republic of China n.d.). Beyond silk production, mulberry's economic value is also derived from its medicinal properties and environmental adaptability (Liu et al. 2023; Jiang et al. 2017). Although with a remarkable cultivation history that spans over 5000 years, diverse and conflicting conclusions have been drawn on the origin and demographic history of mulberry trees, mainly because of the insufficient molecular evidence (Jiao et al. 2020; Dai et al. 2023). Resolving these questions requires more comprehensive molecular data, including valuable resources like ancient trees that serve as living records of historical genetic diversity (Kong et al. 2025).

Annual crops have entered the era of molecular‐assisted breeding through large‐scale trait‐gene mapping; the inherent complexities of woody species, including long generation cycles, low genetic polymorphism, and perennial growth characteristics, continue to impede functional gene discovery and molecular breeding applications. Mulberry trees are a promising exception. With a short breeding cycle (2–3 years), a compact genome (~330 Mb), and exceptional germplasm diversity (over 3000 accessions), mulberry enables rapid genetic studies and bypasses the logistical and temporal bottlenecks inherent in perennial species (Jiao et al. 2020; Dai et al. 2023; Dhanyalakshmi and Nataraja 2018; He et al. 2013; Jain et al. 2022; Ma et al. 2023; Xia et al. 2024). Despite this immense potential, mulberry research has not kept pace with annual crops, primarily because of three core challenges. First, a major challenge has been the lack of haplotype‐resolved genomes for key cultivated varieties. Although seven mulberry genome assemblies have been published, the absence of high‐quality haplotype information has limited our ability to fully capture gene information from both parents and has hindered the dissection of complex traits. Second, the evolutionary trajectory of mulberry remains unclear because of insufficient molecular evidence from prior studies, particularly the lack of ancient tree genomic data. Finally, trait dissection has been strikingly limited, with only two studies collectively analysing just four traits. This striking disparity between genetic potential and research focus leaves critical improvement targets for mulberry largely unaddressed. For instance, although disease resistance is paramount for crop sustainability, the genetic basis of mulberry blight resistance, the most damaging bacterial disease affecting mulberry cultivation, remains largely unexplored. This is the case despite the availability of two excellent materials with contrasting phenotypes for this critical trait: the widely cultivated variety NS14 (which historically covered 90% of China's cultivation area and exhibits robust resistance) and high‐yielding but susceptible successor QS1 (gradually supplanted NS14 after 2009).

In this study, we generated a high‐quality haplotype‐resolved genome assembly for Morus alba NS14 and QS1. A total of 242 mulberry accessions from 12 countries were sequenced, covering geographic areas that were previously underrepresented in earlier studies. These accessions included 39 ancient mulberry trees, among which 9 are over 1000 years old (China Agricultural Museum 2020), providing living records for evolutionary analyses. By integrating an additional 134 previously sequenced accessions, we established a comprehensive dataset of 376 accessions to reconstruct the domestication history and evolutionary trajectory of mulberry. Through large‐scale genome‐wide association studies (GWAS) of 37 agronomic traits, the vast majority of which have not been reported previously, we identified genetic determinants underlying critical traits including disease resistance, environmental adaptation, and yield components. These findings not only advance our understanding of perennial crop evolution but also provide potential genetic targets for future breeding programs.

2. Results

2.1. Haplotype‐Resolved Assembly and Annotation of Mulberry Genome

We conducted a crossbreeding experiment between NS14 and QS1 to investigate their genomic characteristics. A single F₁ progeny, QN52 (2n = 2x = 28), exhibiting strong disease resistance, was selected for genome sequencing using HiFi technology (Figure 1a,b). A total of approximately 34.3 Gb of sequencing data was generated, providing 100‐fold genome coverage with a read length N50 of 16 173 bp (Table S1). Two haplotypes of the heterozygous QN52 genome (heterozygosity rate = 1.17%) were assembled using hifiasm (Figure 1c, Figure S1), and subsequently phased using trio binning with short‐read sequencing data (~100×) from its two parents (heterozygosity: NS14, 1.05%; QS1, 1.11%). The final assembly contained two haplotype‐resolved genomes with sizes of 313.8 (NS14) and 316.3 Mb (QS1), consisting of 106 (NS14) and 142 (QS1) contigs, with the N50 lengths of 5 785 094 (NS14) and 6 239 259 (QS1) bp (Tables S2 and S3). Analysis using the BUSCO database revealed that 96.6% and 96.8% of conserved core single‐copy genes were present in the contig assemblies of NS14 and QS1, respectively, indicating high assembly completeness (Tables S4 and S5). Further scaffolding using Hi‐C data resulted in 98.31% (NS14) and 98.28% (QS1) haplotypes being assembled into 14 pseudo‐chromosomes. The scaffold N50 values reached 21.6 Mb (NS14) and 21.0 Mb (QS1), respectively (Tables S2 and S3). Repeat content analysis revealed that the NS14 and QS1 genomes consisted of 46.13% and 45.22% repetitive elements, with LTR Assembly Index (LAI) values of 23 (NS14) and 21.74 (QS1) (Table S6). These LAI values exceed the criterion for a high‐quality genome (LAI ≥ 20), demonstrating the high quality of the assembled genomes for NS14 and QS1.

FIGURE 1.

High‐quality genome assembly of NS14 and QS1 and phylogenetic analysis of Mulberry. (a) Growth phenotype (above) and Mulberry blight resistance (below) of NS14 and QS1. (b) QN52, a F₁‐resistant plant derived from hybridization of NS14 and QS1, was used for whole genome sequencing. (c) Genome features of NS14 and QS1. (d) Syntenic analysis between NS14/QS1 and Malus prunifolia . (e) Phylogenetic trees of 13 plant species in angiosperms.

Transcriptomes from six tissues (i.e., leaves, roots, male flowers, female flowers, twigs, and fruits) were used for genome annotation. A total of 26 498 genes and 27 228 proteins were predicted for haplotype NS14, with an estimated completeness of 93.5% (Table S2). Similarly, 27 781 genes and 29 765 proteins were annotated for haplotype QS1, with 94.0% completeness (Table S3). Such high completeness in NS14 and QS1 indicates relatively complete annotation.

We then examined genome synteny between Morus alba NS14/QS1 and another representative species Malus prunifolia . Both haplotypes exhibited a high degree of genome synteny with M. prunifolia , further validating the accuracy and completeness of the assemblies (Figure 1d). NS14 and QS1 haplotypes showed highly consistent gene order (Figure S2a). Nevertheless, a large number of variations were found between them, particularly in Chr1. Totally, we identified 1 453 962 SNPs, 371 526 insertion/deletion (InDels), and 13 128 structural variations between the two haplotype genomes (Table S7). The 45% of genes (8757 of 19 446) showed significant different allelic expression in the NS14/QS1, indicating both consistent and inconsistent allelic expression patterns (Figure S2b). Additionally, we analysed the evolutionary history of the genomes from 12 angiosperms, including four species of Morus Linn (including NS14/QS1) and eight other species, on the basis of their single‐copy gene families. The maximum‐likelihood phylogenetic tree revealed that the common ancestor of NS14/QS1 diverged from Artocarpus heterophyllus approximately 33.3 million years ago (MYA) during the Paleogene period and from Morus notabilis around 9.3 MYA during the Neogene period (Figure 1e).

2.2. Population Structure of Mulberry

We sequenced a total of 242 mulberry accessions from 26 provinces of China and 11 other countries (Afghanistan, Australia, USA, India, Ukraine, Republic of Serbia, Uzbekistan, Japan, Madagascar, Democratic People's Republic of Korea, and Sri Lanka) (Figure 2a, Table S8). Among these accessions, 39 ancient trees were specifically collected with ages ranging from more than 100 years to 1800 years (Table S8). Our sequencing generated approximately 1.2 Tb of sequences with an average depth of 15.5× per accession. Incorporating data from 134 accessions reported previously (Jiao et al. 2020), we identified 4 407 132 SNPs and 388 788 small indels (< 10 bp) (Table S9).

FIGURE 2.

Phylogenetic relationships and population structure of resequenced mulberry accessions. (a) Geographic distributions of 376 Mulberry accessions. The size and colour of each pie chart represent the sample size in a specific geographic location. (b) SNPs from 376 sequenced Mulberry accessions were used to construct a phylogenetic tree by neighbour‐joining (NJ). Artocarpus Heterophyllus served as an outgroup. The abbreviations in (a) and (b): Yunnan‐Guizhou Plateau (YGP), Guangdong and Guangxi Province (GGP), Yangtze River area (YaR), Taihu landrace group (TLG), Hubei landrace group (HLG), Taihu cultivar group (TCG), Northeast area (NE), Yellow River area (YeR), Xiajin County (XJ), Northwest area (NW), and Sichuan Province (SC). (c) Phylogenetic tree and model‐based clustering analysis with different number of clusters (K = 2–11) of the 376 sequenced Mulberry accessions. The red lines represent ancient accessions. (d) Genetic differentiation (Fst) analysis in each pair of Mulberry subpopulation. (e, f) Nucleotide diversity (π) (e) and linkage disequilibrium decay (f) analysis in Southern China and Northern China. The arrow with numerical values in the lines represented median values of the r ² and the corresponding physical distances in (f).

Using A. heterophyllus (Moraceae family) as an outgroup, we selected 1 342 173 unlinked/independent SNPs to analyse phylogenetic relationships and population structure. The neighbour‐joining phylogenetic tree revealed two groups of 376 mulberry accessions that could basically be categorised on the basis of their geographic locations within China: northern and southern regions (Figure 2b). The northern region consisted of five areas: Northeast area (NE), Yellow River area (YeR), Xiajin County (XJ), Northwest area (NW), and Sichuan Province (SC). Notably, the YeR region in XJ harbours an ancient mulberry cultivation community, recognised in 2018 by the FAO (Food and Agriculture Organization of the United Nations) as a “Globally Important Agricultural Heritage System” (Food and Agriculture Organization of the United Nations n.d.). The southern area contained three areas: Yunnan‐Guizhou Plateau (YGP), Guangdong and Guangxi Province (GGP), and Yangtze River area (YaR). The YaR group included three subgroups: Taihu landrace group (TLG), Hubei landrace group (HLG), and Taihu cultivar group (TCG) (Figure 2b). Because of their closer genetic distances among these three subgroups, we analysed them as a whole group (YaR) (Figure S3a). Accessions from abroad countries were mainly divided into two groups, suggesting a two‐stage expansion from their origin YGP to abroad. The mulberry population was investigated for genetic structure using clusters (K) ranging from 2 to 11 among the 376 mulberry accessions. When K = 2, the YGP population was separated from the whole mulberry population; when K = 11, clusters maximised the marginal likelihood, and clustering was basically consistent with the phylogenetic tree (Figure 2c, Figure S3b). The ancient trees clustered with their local cultivated/landrace species, which further supported the high consistency between our phylogenetic tree classification and geographical locations (Figure 2c).

We carried out genetic differentiation (Fst) analysis on each pair of mulberry subpopulations. The population in YGP exhibits significant genetic differentiation from the other populations (Figure 2d). The genetic diversity (π) analysis revealed that the population in the southern region had much higher nucleotide diversity than that in the northern region (Figure 2e, Figure S3c). The decay of LD with physical distance between SNPs revealed that half of the maximum values occurred at 0.26 kb in the southern region and 0.34 kb in the northern region (Figure 2f). These results suggest a higher degree of genetic recombination in the southern population, and this finding is consistent with the genetic diversity (π) analysis.

2.3. Domestication and Introgression of Mulberry

Principal component analysis (PCA) of 376 mulberry accessions revealed a clear geographic separation pattern (Figure 3a), which was consistent with the results of the phylogenetic tree and population structure (Figure 2b,c). These findings imply a potential evolutionary pathway for mulberry: the mulberry population originated in the YGP, subsequently spreading to the GGP, and from there, to abroad and the YaR. The population then expanded from the YaR to the NE, followed by dispersal to the YeR, and finally, to the NW and SC (Figure 3a,b). The TreeMix analysis supports our hypothesis, revealing significant gene flow among these mulberry populations, demonstrating that gene flow likely initiated with the YGP population and subsequently spread to foreign and northern populations (Figure 3c).

FIGURE 3.

Demographic history of mulberry. (a) The principal component analysis (PCA) of the 376 mulberry accessions. The arrow indicates the trajectory of germplasm transmission. (b) Putative spread routes of Mulberry. The arrows indicate the direction of germplasm transmission. (c) TreeMIX analysis of population splits and migrations among mulberry accessions from different areas. Red and orange lines between populations indicate gene flow. (d) Historical effective population sizes inferred from mulberry groups samples using PSMC, assuming a mutation rate of m = 7.0 × 10⁻⁹ and an average generation time of g = 2 years. (e) Divergence time for seven groups was estimated using PSMC.

The mulberry tree may have undergone strong selection during its long period of domestication, which could affect the effective population size (N _e) of existing genetic clusters. To address this issue, we used pairwise sequential Markovian coalescence analysis to study the changes in the effective population size (N _e) throughout the evolutionary history of mulberry. We found that seven groups showed similar demographic trajectories (Figure 3d). YGP, as the first diverged ancestral mulberry population, underwent its first expansion event around 180 000 years ago. Subsequently, around 150 000 years ago, other subpopulations began expanding simultaneously. The first mulberry expansion event came to an end around 20 000 years ago, a bottleneck that corresponds to the Last Glacial Maximum (LGM). The bottleneck events observed across most populations during the LGM align with the cold and arid conditions that would have contracted suitable habitats (Jackson et al. 2019). Meanwhile, around 15 000 years ago, the early mulberry (YGP) began its second expansion. As the early mulberry expanded, the population in NE began its second expansion around 10 000 years ago, and the mulberry in the YeR and YaR also experienced certain expansion events between 7500 and 3500 years ago (Figure 3d). The population expansion inferred in the lineage around 10 000 years ago coincides with the warm and stable climatic conditions of the Early Holocene, which likely provided favourable environments for population growth (Nielsen et al. 2018). This analysis provides evidence supporting the origin of mulberry and allows us to construct a model of its expansion and contraction in the main cultivation areas (Figure 3e).

2.4. Genetic Loci Related to Important Agronomic Traits

To explore the genomic basis of domestication and selection, two selective sweep analyses were performed: (1) YaR vs. YGP, representing modern cultivars vs. wild‐type accessions; (2) YaR vs. YeR, representing southern vs. northern cultivars. The top 5% Fst value for the YaR vs. YGP comparison was 0.55, significantly higher than the 0.14 observed for YaR vs. YeR (Figure S4). This higher Fst value between YaR and YGP indicates greater genetic differentiation between modern cultivars and wild‐type accessions. In contrast, the relatively lower Fst value between YaR and YeR suggests that southern and northern cultivars have experienced less genetic differentiation. Notably, the selective sweep analysis between YaR and YeR revealed specific genomic regions and genes related to regional adaptation (Figure S4). We identified strong selection signals in key genes governing traits essential for survival in distinct regional climates, such as the R gene cluster associated with disease resistance (Figure 6), the MaFAD2 gene involved in budburst timing (Zhao et al. 2024), and the MaUVR8 gene associated with environmental response to UV radiation (Yang et al. 2018). In total, 515 selected regions covering 5231 genes were identified across the two comparisons (Table S10). To further investigate the functional significance of these loci, 37 agronomic traits were analysed by GWAS in 203 mulberry samples, excluding ancient trees (Figure 4a, Table S11). A total of 204 loci linked to these traits were identified (Table S12).

FIGURE 6.

Identification of genes associated with bacterial disease resistance. (a) The Manhattan plot on the basis of GWAS analysis of resistance to bacterial mulberry blight. The dashed line represents the significant threshold (p < 1.452 × 10⁻⁶, Bonferroni‐corrected). (b) Bulked segregant analysis (BSA) of F₁ population of NS14 (resistant) × QS1 (susceptible). The pink threshold line indicates a significant difference, whereas the green threshold line indicates an extremely significant difference. (c) Genetic differentiation (Fst) and nucleotide diversity (π) analysis on Chr1. The blue background indicates the region of G‐type RLK enrichment associated with resistance in (a) and (b). (d) the LD block of this G‐type RLK enrichment region on Chr1. (e) The heatmap shows the variation of SNP loci in this G‐Type RLK region on Chr1 in the Yellow River and Yangtze River populations. Yellow represents the reference allele (Ref), red represents the alternative allele (Alt), and orange represents heterozygous (Het) sites. On the basis of the heatmap, the Yellow River and Yangtze River populations are divided into three haplotypes. (f, g) The proportion of germplasms with different resistance levels in the Yellow River and Yangtze River (f), and three haplotypes (g). In f and g, 1 represents hypersensitivity, 2 represents sensitiveness, 3 represents moderate resistance, and 4 represents high resistance. (h) Sequence synteny analysis of Chr1 between NS14 and QS1 genome. The red box line marks the location of G‐Type RLK, which was identified in a and b. The pink lines represent inversions (i) Schematic diagram of differential regulation in the G‐Type RLK region on Chr1 in NS14 and QS1 transcriptomic before and after inoculation with mulberry blight. The oval shapes represent genes. The ovals filled with red and blue, respectively, represent the genes up‐regulated and down‐regulated in NS14 and QS1 before and after pathogen inoculation (FDR < 0.05, Fold Change > 2). The curves connect homologous genes. The curves indicate that the transcriptional regulatory trends of homologous genes are the same (black) or opposite (green) of NS14 and QS1. (j) The number of differentially expressed resistance (R) genes in NS14 and QS1 transcriptomic before and after inoculation with mulberry blight. (k, l) The number of differentially expressed resistance (R) genes with different R gene types in NS14 (k) and QS1 (l) before and after inoculation with mulberry blight.

FIGURE 4.

GWAS analysis of 37 phenotypic traits in mulberry. (a) Distribution of chromosomal loci associated with 37 traits. (b, g) Phenotypic statistics of branch pitch (b) and branch length (g). (c, h) Manhattan plot (top) and LD block (bottom) on the basis of GWAS analysis surrounding the peak includes MaXylanase5 associated with branch pitch (c) and MaYUCCA8 associated with branch length (h). The blue box line marks the corresponding location of these genes. The red dashed line represents the significant threshold (p < 1.452 × 10⁻⁶, Bonferroni‐corrected) in (c) and (h). (d, i) Gene structure and mutation information of MaXylanase5 (d) and MaYUCCA8 (i). (e, j) Proportion of different haplotypes in the population related to branch pitch (e) and branch length (j). (f, k) Phenotype statistics of mulberry germplasm associated with various haplotypes of branch pitch (f) and branch length (k). Significant of phenotypic statistical was tested using Mann–Whitney U tests.

Shorter branch pitch is a key trait for selecting superior mulberry cultivars, as it increases leaf yield per unit area. Thus, genes involved in internode elongation are critical for the selection. The branch pitch of cultivated varieties (YeR and YaR) was significantly shorter than that of wild‐type germplasm (YGP) (Figure 4b). A significant GWAS signal on Chr7 was identified, containing the NS14_07G001134 gene, which encodes endo‐1,4‐β‐xylanase (MaXylanase5) (Figure 4c; Figure S5a). Three nonsynonymous variants of MaXylanase5 form three distinct haplotypes, with haplotype 3 predominantly found in cultivated varieties (YaR and YeR), and associated with shorter internode length (Figure 4d–f). MaXylanase5 promotes the hydrolysis of 1,4‐β‐D‐xylosidic linkages in xylans, a major component of hemicellulose in plant cell walls, and may be involved in cell wall remodelling during cell elongation (Abramson et al. 2010). The homologue of MaXylanase5 in maize was known to regulate stem internode length (Hu et al. 2020).

Branch length is a distinguishing feature between cultivated and wild mulberry varieties. Cultivated varieties (YeR and YaR) have significantly shorter branch lengths than wild‐type germplasm (YGP), which is advantageous for cultivation and harvesting (Figure 4g). A significant GWAS signal on Chr7 associated with branch length includes the NS14_07T000587 gene, which encodes an auxin synthase (MaYUCCA8) (Figure 4h; Figure S5b). The homologues of MaYUCCA8 have been reported to regulate hypocotyl length in A. thaliana (Cai et al. 2023; Gao et al. 2022). A nonsynonymous variant site (Chr7:4416417) in MaYUCCA8 leads to a substitution from Ala to Pro (Figure 4i). Haplotype 2, predominantly found in cultivated varieties (YeR and YaR), results in shorter branch lengths (Figure 4j,k).

Earlier budburst in spring allows plants to access more light and soil resources, but also exposes them to the risk of late spring frosts, known as false spring, which can be lethal to plants (Chamberlain et al. 2019). Therefore, budburst timing is a critical trait in mulberry breeding because it directly influences the growth cycle and adaptability. The budburst time of the GGP group was significantly earlier than the YaR group and the YeR group (Figure 5a). A significant GWAS signal was identified on Chr1, which includes the NS14_01G001763 gene, encoding a B3‐domain transcription factor MaVAL1 (Figure 5b,c). Its homologous gene in A. thaliana, AtVAL1, has been reported to regulate flowering and vernalization processes (Tao et al. 2019; Jing et al. 2019). Three nonsynonymous variant sites in MaVAL1 form three haplotypes (Figure 5d). Haplotype 1, which leads to a late budburst time, is mainly found in the YeR population (the group with the latest budburst time) and also occurs significantly more frequently in the YaR population compared to the GGP group. In contrast, haplotype 2 results in early budburst and is predominantly present in the GGP group (Figure 5e,f).

FIGURE 5.

Identification of genes related to bud‐bursting time, leaf size and leaf thickness by GWAS analysis. (a, g, h, i, q) The phenotype statistics of bud‐bursting time (a), leaf width (g), leaf length (h), leaf area (i), and leaf thickness (q). (b, j) The Manhattan plot on the basis of GWAS analysis of bud‐bursting time (b) and leaf width (j). (c, k, r) Manhattan plot (top) and LD block (bottom) on the basis of GWAS analysis surrounding the peak includes MaVAL1 associated with bud‐bursting time (c), MaTEBICHI associated with leaf size (width, length and area) (k), and MaMSRB1 associated with leaf thickness (r). The blue box line marks the corresponding location of these genes. The dashed line in plot by GWAS analysis represents the significant threshold (p < 1.452 × 10⁻⁶, Bonferroni‐corrected). (d, l, s) Gene structure and mutation information of MaVAL1 (d), M MaTEBICHI (l), and MaMSRB1 (s). (e, m, t) Proportion of different haplotypes in the population related to bud‐bursting time (e), leaf size (m), and leaf thickness (t). (f, n, o, p, u) Phenotype statistics of mulberry germplasm associated with various haplotypes of bud‐bursting time (f), leaf width (n), leaf length (o), leaf area (p), and leaf thickness (u). Significant of phenotypic statistical was tested using Mann–Whitney U tests.

Mulberry leaves are the sole feed source for silkworms, and higher leaf biomass consumption directly correlates with increased silk yield (Jiao et al. 2020; He et al. 2013). Thus, leaf size and thickness are crucial indicators for improving biomass in mulberry breeding. Cultivated varieties (YaR) exhibited significantly larger leaf width and length compared to wild‐type germplasms (YGP) (Figure 5g,h). A significant GWAS signal on Chr5 was identified, associated with leaf width and leaf length (Figure 5j,k, Figure S6a,b). The signal includes a candidate gene, NS14_05G001588, which encodes a DNA polymerase (MaTEBICHI). Its homologue in Arabidopsis thaliana has been reported to influence leaf development (Inagaki et al. 2006). Nine nonsynonymous variants in the MaTEBICHI form three major haplotypes (Figure 5l), with haplotypes 2 and 3 resulting in increased leaf width and leaf length. Besides, these haplotypes are predominantly found in cultivated varieties (YaR and YeR) but not in wild‐type germplasms (YGP) (Figure 5m–o), highlighting a key target for domestication. The significantly higher transcript levels of MaTEBICHI shown in cultivated accessions (YaR/YeR) than in wild germplasm (YGP) were consistent with the haplotype distribution (Figure S6c), further supporting their positive roles in determining leaf dimensions. Although the leaf width and length were positively correlated with leaf area, the difference in leaf area between wild and cultivated types did not reach a significant level (Figure 5i,p; Figure S6d). This finding suggests that although key loci control size, the total projected leaf area is a highly complex and heterogeneous phenotype that is likely strongly influenced by additional genetic factors controlling leaf morphology, such as the degree of lobing and dissection, which varies widely in our panel (Table S11).

Chr9 harbours a significant GWAS signal associated with leaf thickness (Figure S6e). Trait surveys indicated that the leaf thickness in the YaR groups was significantly greater than in the YGP population (Figure 5q). The identified GWAS signal includes the gene NS14_09G001306, encoding a methionine sulfoxide reductase B (MaMSRB) that can reduce methionine sulfoxide into methionine, and this gene regulates rosette weight in A. thaliana (Figure 5r) (Laugier et al. 2009). Haplotype analysis revealed that a nonsynonymous SNP (Chr9:14110551) located in the exon region of MaMSRB results in four distinct haplotypes (Figure 5s). Haplotype 1, mainly present in the YGP group, represents a deletion at this site that causes a frameshift mutation in the MaMSRB gene, leading to reduced leaf thickness (Figure 5t,u). In contrast, haplotypes 3 and 4, predominantly found in cultivated varieties, are associated with increased leaf thickness (Figure 5t,u). Accordingly, MaMSRB1 transcript abundance was significantly higher in cultivated accessions (YaR/YeR) than in wild germplasm (YGP), further supporting its positive role in promoting leaf thickness (Figure S6f).

2.5. Genetic Loci Associated With Diverse Growth and Development Traits

We identified six loci associated with eight traits (leaf apex shape, leaf margin shape, leaf curvature, leaf angle, lenticel shape, bud differentiation, tender shoots colour, and branch colour) through GWAS, which are important evaluation indicators for distinguishing different varieties of mulberry trees and are also closely related to their ecological adaptability. Leaf morphology was reported to be highly correlated with plant adaptability (Nakayama et al. 2022). For example, leaf apex shape and leaf curvature both control rain drainage (Wang et al. 2020). The leaf apex and margin shape share an overlapping GWAS signal on Chr6, which contains the gene MaWOX2 (MaNS14_06G000802) (Figure S7a,b). The homologue of MaWOX2 in A. thaliana has been reported to play a role in the formation of young leaf boundaries (Lin et al. 2012). Another gene, MaLSG1 (NS14_06G001809), located on Chr6, was identified by GWAS as being significantly associated with leaf curvature (Figure S8c). MaLSG1 has been previously implicated in modulating leaf curvature in A. thaliana (Zhao et al. 2015).

Leaf angle regulates photosynthesis efficiency by affecting the spatial distribution and mutual shielding degree of leaves (Yang et al. 2023). A GWAS signal associated with leaf angle was identified, encompassing the NS14_14G000921 gene that encodes the auxin transport protein MaPIN1 (Figure S7d). This finding is consistent with GWAS analyses in Brassica napus , where leaf angle is also correlated with auxin transport family genes (Hu et al. 2021, 2022). Lenticels are channel‐like structures that promote gas exchange in secondary growth tissues, such as tree stems, and are closely related to pathogen invasion and drought stress (Lendzian 2006; Nemesio‐Gorriz et al. 2019; Zhong et al. 2024). Within the GWAS signal for lenticel shape, the MaCNGC4 gene (NS14_01G001786) was identified (Figure S7d). The homologue of MaCNGC4 has been previously documented to regulate stomatal size and density in A. thaliana (Kale et al. 2019). Given that lenticel formation occurs beneath the stomata and shares similarities with stomatal function (Zhong et al. 2024), MaCNGC4 may play a role in lenticel development in mulberry.

Bud differentiation (flowering and leafing order), branch colour, and tender shoots colour are important phenotypic characteristics. MaMED8 (NS14_02G000106), located on Chr2, has been found to be associated with bud differentiation (Figure S7f). Its homologue AtMED8 has been found to regulate floral transition through interaction with FPA to modulate FLC expression in A. thaliana (Yuan et al. 2023). In the traits of branch colour and tender shoots colour, an overlapping GWAS signal was detected on Chr1, including NS14_01G000512 gene (Figure S7g,h). This gene encodes a cinnamoyl‐CoA reductase (CCR) enzyme, which is recognised for its role in lignin biosynthesis (De Meester et al. 2018). The downregulation of CCR2 causes red discoloration in poplar wood (Van Acker et al. 2013).

2.6. Identification of a G‐Type RLK Gene Cluster Associated With Bacterial Disease Resistance

Mulberry blight is a significant bacterial disease that severely restricts the development of the mulberry industry. To investigate resistance to mulberry blight, we conducted a three‐year assessment of the 203 mulberry accessions to record the relevant traits (Figure S8 and Table S11). To identify gene loci associated with disease resistance, we performed GWAS analysis and conducted an F₁ population of NS14 (resistant) × QS1 (susceptible) followed by performing bulked segregant analysis (BSA). A significant peak signal of BSA overlapped with the GWAS signal on Chr1 (41.72–44.28 M), which was a G‐type RLK enriched region (Figure 6a,b and Figure S9). Linkage disequilibrium (LD) analysis revealed a clear linkage within this region, and selective sweep analysis indicated that this region underwent selection in the YeR population compared to the YaR population (Figure 6c,d). Consistent with these observations, the YeR population has a higher proportion of resistant accessions compared to the YaR population. Haplotype analysis further revealed three distinct haplotypes within this region, among which haplotype 1 contains a higher number of resistant accessions and is predominantly found in the YeR populations (Figure 6e–g).

To further elucidate the genetic basis of this resistance, we conducted additional analyses focusing on the G‐type RLK‐enriched region. Genome synteny analysis revealed that this region was highly divergent between NS14 and QS1 (Figure 6h). Analysis of the structural variation (SV) showed that there were a large number of SVs in this region (Figure S10). This region contains a large cluster of G‐type RLK family genes, which makes it difficult to identify the potential causal genes solely on the basis of functional annotation (Figure 6h, Table S13). Thus, we then conducted transcriptomic analyses of NS14 and QS1 before and after inoculation with the pathogen of mulberry blight to identify the RLK genes responding to the disease (Figure S11a). Within the region, we noticed several homologous G‐type RLK gene pairs that showed opposite expression trends in NS14 and QS1 (Figures 6i, S11b,c; Tables S14, S15). To verify the transcriptional dynamics of these candidates, including their early response timing, we performed qRT − PCR analyses across multiple early time points post‐inoculation (Figure S11d). The opposite expression trends of G‐type RLKs in resistant (NS14) and susceptible (QS1) varieties were confirmed by qRT‐PCR, which revealed significant and rapid responses as early as 3 h post‐inoculation and maintained expression differences consistent with the transcriptome data (Figure S11d). A genome‐wide analysis of resistance (R) genes showed that the overall number and classification of R‐genes were highly similar between the NS14 and QS1 genomes (Figure S11e–g). However, when we analysed the differentially expressed resistance genes across the entire genome, the QS1 genome had many more down‐regulated RLK family genes compared to the NS14 genome (Figure 6j–l, Tables S16, S17). Overall, the G‐type RLK gene cluster on Chr1 (41.72–44.28 M) is likely responsible for mulberry blight resistance. The structural differences in this gene cluster between resistant (NS14) and susceptible (QS1) varieties may lead to different expression trends of these RLK genes, which is likely one of the reasons for the observed resistance differences.

3. Discussion

Disease resistance genes (R genes) are often found in clusters in many species (Parniske et al. 1999; Tang et al. 2022; Shang et al. 2022; van Wersch and Li 2019). Previous studies have indicated a unique arrangement of R genes in woody plants compared to herbaceous plants (van Wersch and Li 2019), with additional divergent groups of G‐type and L‐type RLKs emerging in core woody eudicots. In addition to the arrangement, G‐type RLKs showed an overall expansion in woody plants, such as Populus and Eucalyptus (Ngou et al. 2022; Yang et al. 2016). RLKs clustered in forestry species could provide the needs for survival in the same location for an extended period of time (Yang et al. 2016). However, the dosage effect of clustered RLK genes on gene expression has not been reported yet. In this study, we identified G‐type RLK genes exhibiting opposite expression trends in resistant and susceptible mulberry trees, suggesting a dosage effect that might help defend against bacterial pathogens. Future investigations focusing on diverse woody species are warranted to validate whether this dosage‐dependent regulation of RLK clusters is conserved across long‐lived forest species and to unravel its broader role in perennial pathogen defence strategies.

Ancient trees provide critical evolutionary anchors for perennial crops, as their genetic profiles are less obscured by modern breeding practices. In the total of eight geographic regions of China reported in this study, we were able to collect ancient mulberry trees from six of these regions (Table S8). Phylogenetically, these ancient trees predominantly occupied basal positions within their respective regional clusters (Figure 2c), closely linking modern landraces to wild progenitors. Intriguingly, ancient specimens from the Yunnan‐Guizhou Plateau (YGP) deviated from this pattern; their non‐basal positions within the phylogeny align with YGP's extraordinary genetic diversity, reinforcing its dual role as both a biodiversity hotspot and the putative epicentre of mulberry domestication. These findings collectively demonstrate how geographic isolation and localised adaptation have sculpted the phylogeographic architecture of cultivated mulberries. Similar results were seen in tea plants. By studying 40 ancient tea trees from southwest China, researchers confirmed these as evolutionary links between wild and cultivated varieties, proving the region as tea's birthplace (Kong et al. 2025). Together, these studies show how ancient trees can help uncover crop domestication histories that are often hidden by modern breeding.

Our mulberry accessions were clustered strongly on the basis of geographic separation patterns. The northern accessions exhibit a distinct cluster but also appear to be the progeny of southern accessions, suggesting their adaptation to the northern environment. Among the whole northern accessions, NE shows the highest similarity to the southern accessions, especially YaR ones. The YeR area is located between NE and YaR, so the intuitive idea of species spreading would be YaR to YeR then NE. However, our results suggested the sequence as YaR to NE then YeR. This raises a question: How did mulberry trees skip over the YeR and spread to NE? Here, we proposed two hypotheses, respectively, on the basis of the evolutionary rate and migratory bird movement. (1) Evolutionary rate. Cold temperatures and reduced sunlight in northern areas lead to a prolonged development time for angiosperms, extending the period from juvenile to reproductive stages and resulting in much longer life cycles (Guo et al. 2018; Crisp and Cook 2011). We hypothesize that mulberry trees indeed spread through the intuitive route, YaR to YeR, then NE. Because of the colder climate and reduced sunlight in Northeast China, the extended seedling‐to‐flowering time results in longer life cycles, allowing these northern accessions to retain more ancestral traits (similar to those in southern regions) than those in the YeR region. This hypothesis is supported by a famous example in plant evolution: gymnosperms, which appeared later than angiosperms, possess longer life cycles, enabling them to retain more ancestral traits (van Wersch and Li 2019). In other words, a later appearance but more ancestral traits due to a longer life cycle. (2) Migratory bird movement. In this hypothesis, we proposed that the spreading to NE was directly from YaR by migratory birds. NE is the hotspot of migrating routes for at least 350 migratory bird species (Kirby et al. 2008). Mulberry fruits and berries are a major food source for at least 43 migratory bird species, with migratory birds contributing to more than half of the total consumption of mulberry fruits (Kratenko et al. 2020). Apart from the NE, the mulberry germplasm resources of SC are particularly intriguing. Located in southern China, SC's germplasm is evolutionarily close to that of Shaanxi, a province in northwest China. However, SC is situated right next to the origin of YGP, prompting an intuitive question: why were the mulberry trees not in SC directly derived from YGP? This unique evolutionary positioning can be attributed to SC's historical geographical context. Known as Ba Shu in ancient times, SC was an isolated basin area, with the Shu Road serving as the only passage connecting Shaanxi and SC, and the sole route for SC to interact with the outside world (Jupp 2007). The region's distinctive geographical features, surrounded by natural barriers and accessible only through limited pathways, likely shaped the migration and distribution of its mulberry germplasm.

4. Material Method

4.1. Plant Materials

Pollen from Nong Sang 14 (NS14) was used to fertilise the pistils of female flowers of Qiang Sang 1 (QS1), and the resulting F₁ generations were planted and evaluated for their resistance to mulberry bacterial blight. A single F₁ plant QN52, which exhibits strong disease resistance, was selected for genome assembly. For population analysis of mulberry, NS14, QS1, and a diverse collection panel of 201 mulberry accessions were planted in the mulberry resource nursery in Hangzhou, Zhejiang Province, China (113.44047 E, 23.388464 N). A total of 39 ancient mulberry accessions were also utilised in this study. Among these, 21 were sourced from the National Germplasm Repository for Mulberry (Zhenjiang), whereas the remaining 18 were collected from various other regions within China. Detailed information about the 242 accessions is provided in Table S7. Young leaves were collected and frozen quickly in liquid nitrogen for DNA extraction. A total of 51 traits were investigated in this study. The experiments were conducted in 2020 for all traits except for disease resistance, which was assessed over a 3‐year period from 2020 to 2022. Detailed phenotyping information is provided in Table S10.

4.2. Pathogen Inoculation and Disease Resistance Assay

The pathogen used in this study is a strain of Pectobacterium carotovorum subsp. mori 18#, which we have isolated and identified as the causal agent of bacterial blight of mulberry. After overnight incubation with shaking, the bacterial culture was collected and diluted in sterile water to an optical density (OD) of 0.2 at 600 nm. For the resistance assessment in the GWAS population (natural population), three plants per variety were selected, and three branches per plant were inoculated. For the F₁ population (BSA population), one plant with four branches was used. Sterile needles were used to create small puncture wounds at the fifth node of each branch, and then a sterile cotton ball was placed over the wound, followed by the application of 100 μL of bacterial suspension, and wrapped with plastic wrap. The plastic wrap was removed after 24 h, and resistance was scored at 5 days post‐inoculation. For transcriptome analysis, we used detached leaves. Small puncture wounds were created on the fifth fully expanded leaves from the top with sterile needles, and a 5 mm filter paper disc was placed onto each puncture site. A 10 μL droplet of the bacterial suspension was applied onto each filter paper disc. The inoculated leaves were then placed in the dark at 28°C for 24 h to facilitate pathogen colonisation. Mock inoculations were performed under the same conditions using sterile water instead of the bacterial suspension. Leaf disks (LD) near the inoculated sites were excised at 24 h post‐inoculation (hpi) using a 10 mm‐diameter cork borer for RNA‐sequencing analysis.

4.3. Resolving the NS14 and QS1 Haplotypes With Trio Binning

High‐quality genomic DNA was extracted from fresh leaves of QN52 and subjected to library construction following the standard protocol of PacBio (Pacific Biosciences). Sequencing was performed on the PacBio Sequel II HiFi platform, which generates high‐fidelity long reads with CCS (v.4.2.0), at Shanghai OE Biotech. Additionally, 100× Illumina PE 150 reads were generated for the parental genomic DNA of NS14 and QS1. We applied the hifiasm (Cheng et al. 2021) (v.0.14.2) trio mode with a 31‐mer database from parental short reads from yak (https://github.com/lh3/yak), to resolve the haplotypes from NS14 and QS1. Organelle DNA or rDNA fragments and other non‐plant contig fragments were identified by BLASTN to the nt database and organelle reference. Hi‐C reads were mapped to the assembly with the Juicer pipeline (v.1.5.7) (Durand, Shamim, et al. 2016) and scaffolded by 3D‐DNA (version: 180419) with the parameters ‘‐r 0 ‐m haploid’. False duplications and phase errors were manually curated on the basis of yak trioeval within Juicebox (v.1.11.08) (Durand, Robinson, et al. 2016). Finally, yak was used to assess the base accuracy of the genome assembly.

Transposon elements were annotated utilizing EDTA (version 1.9.5) (Ou et al. 2019), referencing the pan‐genome TE database available at the HuffordLab GitHub repository (NAM‐genomes/te‐annotation). Protein‐coding genes were inferred employing MAKER2, leveraging homologous evidence from RNA‐seq and protein databases. RNA samples were procured from three distinct tissues of QN52 (root, twig, and leaf), two from QS1 (female flower and fruit), and one from NS14 (male flower). Subsequently, the RNA was aligned to the reference genome using HISAT2 (version 2.10.2) (Li and Godzik 2006). Protein sequences for two plant species, including M. alba and Ficus erecta, were sourced from UniProt (Viridiplantae) (https://www.uniprot.org) and integrated with CD‐HIT (version 4.6) (Brůna et al. 2021) employing a sequence identity threshold of 99%. De novo gene prediction was executed with SNAP (version 2006‐07‐28), AUGUSTUS (version 3.3.3), and GeneMark (version 4.3.8). SNAP was calibrated on the basis of the initial MAKER2 annotations, whereas AUGUSTUS and GeneMark were trained using BRAKER2 (Kim et al. 2019) with RNA‐seq and protein database evidence. Gene models with annotation edit distances below 0.5 were selected for further analysis.

4.4. Whole‐Genome Collinearity Analysis

We conducted a whole‐genome collinearity analysis to detect regions of collinearity and genome variation between the NS14 (reference) and QS1 (query) haplotypes, utilising minimap2 “‐ax asm5 ‐eqx” (v2.17) (Li and Birol 2018) for the alignment. The alignment was provided to SyRI (Goel et al. 2019) (‐k ‐F ‐S, v1.3) to identify synteny, single‐nucleotide level differences, and large‐scale structural variations consisting of insertions, deletions, inversions, duplications, and translocations (≥ 50 bp). Only SVs with a quality score of 10 or above and supported by at least five reads were retained.

4.5. Phylogenetic Analysis and Molecular Dating

Orthogroups for the selected species were identified using OrthoFinder v2.5.4 (Emms and Kelly 2019) with default parameters. Single‐copy orthogroups were aligned with MAFFT (Katoh et al. 2002) using the L‐INS‐i model, and the resulting alignments were concatenated into a supermatrix. This supermatrix was used to construct the species phylogeny with IQ‐TREE2 (Minh et al. 2020) under the best‐fit model JTT + F + I + G4, with 1000 bootstrap replicates. Divergence times were estimated using the MCMCtree program in PAML v4.9 (Yang 2007), with parameters set to burnin = 200 000, sampfreq = 100, and nsample = 50 000. Fossil calibration points were set according to the records in the TimeTree database (http://timetree.org).

4.6. Population Sequencing

We utilised the TruSeq Nano DNA LT Sample Preparation Kit (Illumina, San Diego, CA, USA) to construct sequencing libraries from 1.5 μg of genomic DNA per sample, following the manufacturer's protocol and incorporating unique index codes for sample identification. The genomic DNA was fragmented to approximately 350 bp using the S220 Focused‐ultrasonicators (Covaris, USA). The DNA fragments were then end‐repaired, A‐tailed, and ligated with full‐length adapters for Illumina sequencing, followed by PCR amplification. The amplified products were purified using the AMPure XP bead system. Library size distribution was assessed with an Agilent 2100 Bioanalyzer, and library quantification was performed using real‐time PCR. Sequencing was performed on the Illumina Novaseq 6000 platform in OE Biotech to generate raw sequence data with a read length of 150 bp.

4.7. Sequence Alignment and Variation Calling

The raw reads were subjected to a quality check and then filtered by fastp (Chen et al. 2018). The remaining high‐quality reads were mapped to the reference genome (NS14) using Burrows‐Wheeler Aligner (BWA) (Li and Durbin 2010). In order to reduce mismatches generated by PCR amplification before sequencing, Picard (http://broadinstitute.github.io/picard/) was employed to mark duplicate reads, and duplicated reads were removed using SAMtools (v0.1.1) (Li et al. 2009). After alignment, the genomic variants, including SNPs and InDels, were identified by Genome Analysis Toolkit (GATK) software (McKenna et al. 2010). Variants were filtered to retain those with genotype scores greater than 30 and read depth scores > 2. Heterozygous sites were also filtered to retain SNPs with minor allele frequency (MAF) greater than 5%. The identified SNPs and indels were further annotated with ANNOVAR (Wang et al. 2010) tool software (version 2013‐05‐20).

4.8. Population Structure and Genetic Diversity Analysis

The SNP data of 242 accessions, along with 134 accessions from the previous study, were genotyped using GATK (McKenna et al. 2010) (V.4.1.9.0). SNPs that passed the screening criteria were extracted and gathered as high‐confidence variants. A total of 4 407 132 SNPs were identified, among them 1 342 173 independent SNPs (‐‐indep‐pairwise 50 1 0.2) were randomly selected with minimum missing data for phylogenetic tree analysis, and an individual‐based neighbour‐joining (NJ) tree was constructed on the basis of the p‐distance using the software TreeBest (v1.9.2) (Vilella et al. 2009) with bootstrap 1000 replications. The population genetic structure was examined using the program ADMIXTURE (v1.23) (Alexander et al. 2009), with all SNPs specifying K ranging from 2 to 11. PCA was performed using the EIGENSOFT program (Price et al. 2006). To estimate and compare the pattern of linkage disequilibrium (LD), the squared correlation coefficient (r (National Bureau of Statistics n.d.)) between pairwise SNPs was computed using PopLDdecay (Zhang et al. 2019), with the parameters in the program set as ‘‐MaxDist 1000 kb’. The average r ² value was calculated for pairwise markers in a 1000 bp window, and values were averaged across the whole genome. To identify genomic regions under selection, population fixation statistics (Fst) and nucleotide diversity (π) for each sliding window (in 100 kb windows with 10 kb step size) were calculated using VCF tools (Danecek et al. 2011). The putative selection targets were designated as the top 5% of log‐odds ratios for both π and Fst. Candidate genes were annotated within functional categories on the basis of Gene Ontology and KEGG and databases. To trace potential historical fluctuations in population size, PSMC software (Li and Durbin 2011) was used with parameters ‘‐N 30 ‐t 5 ‐r 5 ‐p “4+30*2+4+6+10”’. Average generation time was set to 2 years and the mutation rate to 2.5 × 10⁻⁸ per site per generation. To detect gene flow between different populations, TreeMix56 was used to estimate a maximum likelihood (ML) tree (Tang et al. 2012).

4.9. Measurement Method of Leaf Thickness

Leaf thickness was phenotyped by measuring total leaf disc mass, which is the standard metric for measuring leaf thickness in mulberry breeding programs. To perform the measurement, three biological replicates were taken for each germplasm. For each replicate, 10 mature leaves were collected and stacked together. One 3.6 cm‐diameter disc was then sampled by punching through all 10 stacked leaves simultaneously at their center. The cumulative weight of this single stack of 10 discs was recorded as the trait value for that replicate. Finally, the average cumulative weight across the three replicates was used as the final quantitative trait value for the leaf thickness phenotype of the germplasm.

4.10. Genome‐Wide Association Study

A total of 4 407 132 SNPs with minor allele frequency (MAF) > 0.05 were used in the GWAS. The association analysis was done using the EMMAX package (v.2012–021‐0) (http://genetics.cs.ucla.edu/emmax/index.html) with default parameters (Kang et al. 2010). The first five principal components were used as covariates. An effective number of independent markers (SNP and SVs) was estimated to be 688 790 (‐‐indep‐pairwise 50 1 0.2), and we defined the significance threshold by Bonferroni‐corrected genome‐wide significance (α = 1) (p < 1.452 × 10⁻⁶) (He et al. 2023).

4.11. BSA of F₁ Population by Whole‐Genome Resequencing

On the basis of the double pseudo test cross theory proposed by Hemmat et al. (1994), the F₁ generation can be utilised to construct genetic linkage maps in species with highly heterozygous genomes. In 2019, an F₁ population consisting of 538 lines was developed in Hangzhou, China, from a cross between the strains NS14 and QS1. The resistance to mulberry bacterial blight of each accession was evaluated. Genomic DNA was extracted from fresh leaves using the CTAB method. For Bulked Segregant Analysis (BSA), DNA samples were pooled by combining equal amounts of DNA from 30 individuals that exhibited either hypersensitive or hyper‐resistant phenotypes. In total, 100× depth genome sequences for each of the parents and 100× data for the bulked sample were generated. Short reads underwent a quality check and filtering using fastp (Version 0.19.5) (Chen et al. 2018), and then clean reads were aligned to the reference genome using Burrows–Wheeler Aligner (BWA, Version 0.7.12) (Li and Durbin 2010) with default options. The mapped reads were sorted and indexed using SAMtools (Version 1.4) (Li et al. 2009), and PCR duplicates were removed with Picard (http://broadinstitute.github.io/picard/, Version 4.1.0.0). Variants, including SNPs and InDels, were called using GATK (Version 4.1.0.0) (McKenna et al. 2010), and annotated with SnpEff (Cingolani et al. 2014). To identify candidate regions associated with resistance to mulberry bacterial blight, the SNP‐index and Δ(SNP‐index) were calculated for all genomic positions. The SNP‐index (Abe et al. 2012) was estimated from the proportion of reads harbouring SNPs from the total number of reads compared to the reference genome sequence. The Δ(SNP‐index) (Takagi et al. 2013) was calculated by subtracting the SNP‐index of the High bulk‐pool from that of the Low bulk‐pool. SNP‐index and Δ(SNP‐index) values were plotted on the mulberry genome physical map in a 1‐Mb window size, sliding the window with 10‐kb increments. Statistical confidence intervals of Δ(SNP‐index) were calculated for all SNP positions with given read depths under the null hypothesis of the existence of no QTLs. The 95% confidence intervals of Δ(SNP‐index) were used as the screening threshold.

4.12. Structural Variations Analysis

To precisely analyse the complex structural variations in the RLK gene region, HiFi data were aligned with NS14 and QS1, respectively, using pbmm2 (v1.17.0) with the parameters “‐‐preset CCS ‐‐sort”. Subsequently, the default CCS data analysis pipeline in pbsv (v2.11.0) (first running pbsv discover and then pbsv call ‐‐ccs) was employed for structural variation analysis. This approach enables base‐resolution level variation detection and accurately identifies various types of variations such as deletions (DEL), duplications (DUP), and insertions (INS).

4.13. RNA Sequencing and Transcriptome Analysis

Total RNA was extracted from the leaves of mulberry varieties NS14 and QS1 following 24 h post‐inoculation (hpi) with the pathogen and mock treatments, utilising the RNeasy Plant Mini Kit (Qiagen) according to the manufacturer's protocol. Subsequent sequencing was conducted on the Illumina Novaseq 6000 platform, yielding approximately 6 Gb of 150‐bp paired‐end reads per sample. The clean reads obtained from these RNA‐sequencing experiments were aligned to the NS14 and QS1 reference genomes using TopHat2 (v.2.1.1) (Trapnell et al. 2012). Expression levels of each transcript were quantified and normalised to fragments per kilobase of transcript per million mapped reads (FPKM) using Cufflinks (v.2.1.1) (Trapnell et al. 2010). RT reactions were performed using PrimeScript RT Reverse Transcription Reagents (Vazyme) according to the manufacturer's protocol. RT‐qPCR was carried out with RT‐qPCR SYBR Green Mix (Yeasen) on a QuantStudio Real‐Time PCR System. Gene expression was normalised to the expression of the MaActin gene. The primers used for RT‐qPCR are listed in Table S18.

4.14. Availability of Data and Materials

The data supporting the findings of this work are available within the paper and its Supporting Information. The datasets and plant materials generated and analysed during the current study can be obtained from the corresponding author upon reasonable request. The raw data of resequenced 242 mulberry accessions, raw data for QN52 genome assembly, and the newly assembled genomes of NS14 and QS1 can be accessed through the Genome Sequence Archive (GSA) (Chen et al. 2021) database, which is maintained by the National Genomics Data Center (NGDC) (Xue et al. 2022) and the China National Center for Bioinformation (CNCB) / Beijing Institute of Genomics, Chinese Academy of Sciences. These datasets are publicly accessible at https://ngdc.cncb.ac.cn under accession number PRJCA035398.

Author Contributions

J.W., Z.W., Y.‐C.J.L., X.S., and Z.L. conceived the project and designed the study. J.W., P.L., Z.X., T.L., Y.Z., Y.H., N.C., M.Z., Y.L., H.L., B.S., and F.Z. collected samples and performed experiments. J.W., Z.W., C.J., Y.Z., and X.J. performed data analyses. J.W., Z.W., and Y.‐C.J.L. wrote the manuscript, and X.S., C.J., Q.L., and Y.Z. revised the manuscript. All authors read and approved the manuscript.

Funding

This work was supported by the Science and Technology Program of Zhejiang Province, 2021C02072, LQN25C160001. Ministry of Agriculture and Rural Affairs of the People's Republic of China, CARS‐18. Science Technology Department of Xinjing Uygur Autonomous Region, 2023A02008–2.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: High‐quality genome assembly of NS14 and QS1. (a, b) Genome Hi‐C heat map for NS14 (a) and QS1 (b). (c) Genomic synteny between homologous chromosomes of the diploid assemblies.

Figure S2: Genetic variations and allelic imbalance in NS14 and QS1. (a) Identification of the syntenic and non‐syntenic region between NS14 and QS1. (b) Identification of allele‐specific expression in leaves of NS14 and QS1. Differentially expressed genes were defined as a threshold (false discovery rate < 0.05; Fold Change > 2). Coordinates are logarithmically scaled (log₁₀). Blue dots indicate allele‐specific expression genes, and grey dots represent non‐allele‐specific expression genes.

Figure S3: Genetic similarity and Nucleotide diversity analysis between Mulberry groups. (a) Identity by state (IBS) analysis of the genetic similarity between Mulberry groups in this study. (b) The dot plot was utilised for cross‐validation purposes, with the aim of identifying the most suitable K value that corresponds to the population's differentiation history. The optimal K value is pointed out by the arrow on the plot. (c) Nucleotide diversity (π) analysis in Mulberry groups in this study.

Figure S4: Selective sweeps through comparisons of YGP versus the Yangtze River and the Yangtze River versus Yellow River. (a, b) The distribution of population differentiation (Fst) and log₂ π ratio as (π_YaR/π_YGP) (a) or (π_YaR/π_YeR) (b) using a 50‐kb sliding window with 25‐kb steps. The upper dots, indicated by blue and green, represent the top 5% of selection regions analysed by θπ. The upper right dots, marked in yellow, represent the top 5% of selection regions analysed by F _ST. In a, the overlapping regions represent the selected in YaR (marked in blue) and YGP (marked in green). In b, the overlapping regions represent selected in YaR (marked in blue) and YeR (marked in green). (c, d) Selective sweeps analysis between YGP versus YaR (c) and YaR versus YeR (d) on the basis of population differentiation (Fst). The dashed line indicates the threshold for selection analysis.

Figure S5: GWAS analysis traits associated with branch pitch (a) and branch length (b). The dashed line in the plot by GWAS analysis represents the significant threshold (p < 1.452 × 10⁻⁶, Bonferroni‐corrected).

Figure S6: Trait correlation analysis. (a, b, d) Manhattan plot on the basis of GWAS analysis of leaf area (a), leaf length (b) and leaf thickness (e). The dashed line in the plot by GWAS analysis represents the significant threshold (p < 1.452 × 10⁻⁶, Bonferroni‐corrected). (c, f) Expression levels of MaTEBICHI and MaMSRB1 in YGP, YaR and YeR accessions. Significance was tested using two‐tailed t‐tests. (d) Each square in the heatmap represents the correlation coefficient between two variables, and the darker the colour, the stronger the correlation.

Figure S7: GWAS analysis of forest tree growth traits. (a–h), Phenotype and Manhattan plot on the basis of GWAS analysis associate with leaf apex shape (a), leaf margin shape (b), leaf curvature (c), leaf angle (d), lenticel shape (e), bud differentiation (f), tender shoots colour (g), branch colour (h). The dashed line in the plot by GWAS analysis represents the significant threshold (p < 1.452 × 10⁻⁶, Bonferroni‐corrected).

Figure S8: Schematic diagram of field inoculation of mulberry blight and determination of resistance level. (a) Schematic diagram of inoculation of mulberry blight in the field. (b) The disease resistance level was determined on the fifth day after inoculation.

Figure S9: Bulked segregant analysis (BSA) of F₁ population of NS14 (resistant) × QS1 (susceptible). (a) SNP density distribution for BSA analysis. (b) Delta SNP index plot showing the difference between the hyperresistant and hypersusceptible bulks. (c) Hyperresistance index distribution. (d) Hypersusceptible index distribution. The black threshold line indicates a significant difference, whereas the red threshold line indicates an extremely significant difference.

Figure S10: Number of structural variations in the G‐type RLK region of chromosome 1 in the NS14 and QS1 genomes.

Figure S11: RNA‐seq analysis of mulberry blight inoculation in NS14 and QS1. (a) Schematic diagram of mulberry blight inoculation in leaves. (b) The difference in disease resistance in mulberry blight inoculation between NS14 and QS1. (c) The Neighbour‐Joining (NJ) phylogenetic tree was used to analyse the evolutionary relationship of differentially expressed G‐type RLKs in the transcriptomes of mulberry blight inoculated with NS14 and QS1. G‐type RLK genes with protein sequence similarity over 70% respectively marked in red and blue font, are up‐ and down‐regulated. (d) Transcriptional regulation differences of homologous gene pairs before and after inoculation with mulberry blight were analysed by qRT‐PCR across different time points. Significance was tested using two‐tailed t‐tests. (e) The number of resistance (R) genes in the NS14 and QS1 genomes. (f) The number of resistance (R) genes in chromosomes of NS14 and QS1. (g) The number of resistance (R) genes in different types of R genes in the NS14 and QS1 genomes.

Table S1: Summary of data used for genome assembly.

Table S2: NS14 haplotype outcome quality statistics.

Table S3: QS1 haplotype outcome quality statistics.

Table S4: NS14 haplotype BUSCO statistical results.

Table S5: QS1 haplotype BUSCO statistical results.

Table S6: Genome quality assessed by the LAI index.

Table S7: Genome variations.

Table S8: Information of the 376 mulberry accessions.

Table S9: Summary of the SNPs and InDels.

Table S10: Selective sweep region.

Table S11: Summary of the phenotype data in this study.

Table S12: Summary of the GWAS results.

Table S13: Disease‐resistant region in NS14 and QS1.

Table S14: The differentially expressed genes in the disease‐resistant region in NS14 transcriptomic before and after inoculation with mulberry blight.

Table S15: The differentially expressed genes in the disease‐resistant region in QS1 transcriptomic before and after inoculation with mulberry blight.

Table S16: Resistance genes in NS14 transcriptomic before and after inoculation with mulberry blight.

Table S17: Resistance genes in QS1 transcriptomic before and after inoculation with mulberry blight.

Table S18: Primer in this study.

Data Availability Statement

The data that support the findings of this study are openly available in Genome Sequence Archive at https://ngdc.cncb.ac.cn/gsa/, reference number PRJCA035398.