Human‐associated genetic landscape of walnuts in the Himalaya: implications for conservation and utilization

A comprehensive understanding of the genetic variation of walnuts (Juglans regia and J. sigillata) in the Himalaya and its potential drivers are essential for the conservation and sustainable utilization of these plant genetic resources. In this study, we aimed to uncover the genetic landscape of walnuts and potential drivers in the Himalaya for better utilization, awareness, sustainable management and conservation of the extant genetic resources of walnuts in the vast Himalayan landscape.


| INTRODUC TI ON
The impact of human activities on Earth's plants varies across the biodiversity spectrum (Kress & Krupnick, 2022).Since the Agrarian Revolution around 12,000 years ago, humans have altered plants through domestication and ongoing crop improvement, shaping plant distribution, traits and genetic diversity to varying degrees (Doebley et al., 2006).Compared with their wild relatives, domesticated species have faced stronger evolutionary pressure from anthropogenic activities.Indeed, anthropogenic footprints in domesticated species are well-documented in numerous edible plants, such as rice (Huang et al., 2012;Khush, 1997;Zheng et al., 2022), tomatoes (Frary et al., 2000;Lin et al., 2014;Wang et al., 2023) and apples (Chen et al., 2021;Duan et al., 2017;Sun et al., 2020).However, previous studies on artificial manipulation of domesticated plants have mainly focussed on densely populated geographical regions, whereas the impact of human-mediated evolutionary processes on crop species growing in remote areas has rarely been explored.
Mountain regions, typically known to have high topographic and environmental heterogeneity, support about one-third of all terrestrial biodiversity (Spehn et al., 2011).The Himalaya, the most prominent mountain system on our planet, covers an area of 6.62 × 10 5 km 2 with a mean elevation of 3326 m (Liu, Milne, et al., 2022) and is one of the 36 global terrestrial biodiversity hotspots (Conservation International, 2023).It supports about 10,000 species of seed plants (Liu, Milne, et al., 2022;Rana et al., 2023), with approximately onethird being endemic to the region (Mittermeier et al., 2004;Wambulwa et al., 2021).Despite the apparent remoteness and inaccessibility of this region, human settlement in the Himalaya dates back thousands of years, and the region is now home to many unique and diverse ethnic groups (van Driem, 2014).However, the large spatial scale of the Himalaya has impeded systematic genetic diversity studies in the region (Wambulwa et al., 2021).To the best of our knowledge, the question of whether human activity plays a role in shaping the genetic structure of plant species in the Himalaya remains largely unexplored.
In wild plants, geographic isolation plays a critical role in shaping the genetic structure of populations (Smith, 1998).Natural geophysical barriers, such as rivers, shorelines, mountains and glaciers, can restrict gene flow and lead to genetic differentiation in isolated populations (Su et al., 2003), largely shaping the biogeographic landscape of many wild plants.For example, the high mountain barriers separating deep valleys in the Himalaya and adjacent regions have most likely promoted divergence in Cupressus (Xu et al., 2010), Taxus (Liu et al., 2013;Poudel, Moeller, et al., 2014), Hippophae tibetana (Qiong et al., 2017) and Koenigia forrestii (Rana et al., 2023).However, for domesticated species, these effects of geophysical barriers on the genetic landscape are made more complex by anthropogenic activity.
The domestication process of plants reduces genetic variation, thereby decreasing the adaptability and resilience of a species to environmental stresses (Krug et al., 2023;Miller & Schaal, 2006).
Conversely, the wild germplasm resources of domesticated species generally preserve superior alleles that confer the ability to survive in adverse environmental conditions.Therefore, wild material provides crucial genetic reservoirs for future breeding and genetic improvement.The description, exploration and preservation of biodiverse germplasm can provide breeders with a gene bank from which they can obtain desirable traits more accurately and rapidly (Khoury et al., 2019;Vahdati et al., 2019), as well as explore both wild and domesticated histories of a domesticated species (Helmstetter et al., 2020;Mabry et al., 2021).Thus, understanding the distribution and genetic diversity of wild relatives of domesticated species is critical for conservation and utilization in the face of emerging diseases and climate change (Bohra et al., 2022;Kovach & McCouch, 2008;Tanksley & McCouch, 1997;Zhang et al., 2017).Results: We detected three genetic groups of J. regia (JR1, JR2, JR3), one of J. sigillata (JS) and two hybrid types (JR1 × JS and JR3 × JS) of walnut in the Himalaya, with the western Himalaya identified as the genetic diversity hotspot of J. regia.The spatial genetic pattern of the J. regia was significantly influenced by geographic and climatic factors.Human-mediated dispersal probably promoted the hybridization and gene introgression between J. regia and J. sigillata, which reshaped the genetic landscape of walnut populations in the Himalaya.

Main Conclusions:
The extant genetic landscape of walnuts in the Himalaya was driven by natural and anthropogenic forces.Regarding conservation, the western and eastern Himalaya are the genetic reservoir of J. regia and J. sigillata, and hence, pure individuals should be urgently protected from frequent hybridization and introgression.In addition, we propose the utilization of natural hybrid resources coupled with new breeding techniques that combine genomic and phenotypic data.The Himalaya are home to two walnut species: Juglans regia L. and J. sigillata Dode; both are wind-pollinated, monoecious, dichogamous, long-lived, outcrossing trees (Lu et al., 1999;Maity et al., 2019).Juglans regia (common, Persian, or English walnut) is naturally distributed from the Balkans through the Iranian Plateau and the Himalaya to southern China (Laufer, 1919;Zohary et al., 2012).It is widely cultivated throughout the temperate regions of the world for its high-quality timber, highly nutritious nuts and medicinal purposes (Lu et al., 1999).The population genetics of common walnut has therefore been well-investigated at regional, national and continental scales over the past two decades (Guney et al., 2021;Karimi et al., 2010;Pollegioni et al., 2020;Shahi Shavvon et al., 2023).Juglans sigillata (iron walnut) bears thick-shelled, edible nuts and is distributed in China, India, Myanmar, Bhutan and Nepal (Lu et al., 1999;Maity et al., 2019).Its genetic composition has also been explored in its native and cultivated areas with limited sampling, including eastern Tibet and northwestern Yunnan (Wambulwa et al., 2022;Wang et al., 2015;Yuan et al., 2018).Genetic barriers between J. regia and J. sigillata are weak, and hybridization has been reported in areas of sympatric distribution, in southwestern China (Gunn et al., 2010;Liu et al., 2023;Wambulwa et al., 2022).However, despite all this work, there has not yet been a thorough genetic analysis of the diversity of wild walnuts in remote areas, such as the Himalaya, largely due to geological, logistical and geopolitical challenges.Flora of China (Lu et al., 1999), 47 populations were identified as putative J. regia, and 18 as J. sigillata (Table S1; Figure 1).For each population, 6-30 adult trees (wild or feral) were sampled, with a distance of at least 100 m between any two sampled individuals.
Healthy and fresh leaves from each sampled tree were collected and preserved in silica gel pending DNA extraction.Voucher specimens were deposited at the herbarium of Kunming Institute of Botany, Chinese Academy of Sciences (KUN), or Calcutta University Herbarium (CUH).

| Population structure analysis
To examine the population structure of Himalayan Juglans, we performed individual-based Bayesian clustering analysis in STRUCTURE v2.3.4 (Pritchard et al., 2000) for the whole data set (N = 1410, hereafter referred to as Data set 1).We ran STRUCTURE from K = 2 to K = 14 under the admixture and correlated allele frequencies models, assuming no prior grouping information.For each K, 20 independent runs with 100,000 burn-in steps followed by 1,200,000 MCMC iterations were carried out, and the optimal number of clusters (K) was further determined by STRUCTURE HARVESTER v0.6.94 (Earl & vonHoldt, 2012).The outputs for different K values were processed with CLUMPP v1.1.2(Jakobsson & Rosenberg, 2007) and graphically displayed using DISTRUCT v1.1 (Rosenberg, 2004).
We assigned an individual to a certain cluster when its admixture coefficient (q) in STRUCTURE at K = 2 was ≥0.8.According to Vähä and Primmer (2006), when q = 0.2, STRUCTURE provided more accurate estimates for hybrids with different loci under various levels of divergence.In our case, individuals were assigned to three clusters: pure J. regia, pure J. sigillata and putative hybrids.
To minimize the effect of genetic introgression in subsequent analyses, we also created a second data set, Data set 2, by removing all putative hybrids inferred from STRUCTURE at K = 2.
Principal coordinates analysis (PCoA) was used to check the genetic clustering for Data set 1 and Data set 2 based on the covariance standardized method of pairwise Nei's genetic distance generated in GenAlEx v6.5 (Peakall & Smouse, 2012).In addition, to infer the genetic relationships of all Juglans individuals, we generated a neighbour-net tree based on Data set 1 using SplitsTree v4.18.3 (Huson & Bryant, 2006).To further explore the genetic differentiation and relationship among different walnut individuals, we built an unweighted neighbour-joining (NJ) tree in POPULATIONS v1.2.31 (Langella, 1999) using Data set 2. Both the neighbour-net and the NJ tree were built using Nei's standard genetic distance (Nei, 1987) computed among individuals and populations using MSA v4.05 (Dieringer & Schlötterer, 2003), respectively.The tree topologies were visualized and adjusted in the R package ggtree (Yu et al., 2017).

| Plastid genome analysis
To infer the maternal origin of hybrids, 35 samples comprising five J. regia, five J. sigillata, and 25 hybrids were randomly selected to examine their evolutionary relationships based on the plastid genome.For each sample, genomic DNA was fragmented for the construction of a 350 bp short insert library, which was subsequently sequenced for 2× 150 bp paired-end reads on an Illumina HiSeq X-Ten (Illumina, San Diego, CA, USA) at BGI-Wuhan, China.
Approximately 2 GB of paired-end clean data were obtained for each sample.The plastid genome was assembled using the GetOrganelle v1.7.5 pipeline (Jin et al., 2020).The haplotypes of all individuals were identified in DnaSP v6 (Librado & Rozas, 2009), and a haplotype network was generated using PopArt v1.7 (Leigh & Bryant, 2015).Furthermore, to infer the maternal origin of hybrids, we built a rooted neighbour-joining (NJ) tree in MEGA v11 (Tamura et al., 2021) based on the whole plastid genome data set, with J. cathayensis as an outgroup.

| Genetic diversity and differentiation estimation
We estimated the genetic diversity for each locus and population based on the total number of alleles (N T ), private allelic richness (N P ), number of average alleles (N A ), number of effective alleles (N E ), observed heterozygosity (H O ), expected heterozygosity (H E ) and unbiased expected heterozygosity (uH E ) using GenAlEx.The allelic richness (A R ) was determined using the R package hierfstat (Goudet, 2005) after adjustment for sample size differences among populations.The inbreeding index (F IS ) was calculated in FSTAT v2.9.4 (Goudet, 2001).The above genetic diversity parameters were also calculated at the group level (inferred from STRUCTURE at K = 4).
In addition, we assessed the significance of pairwise genetic differentiation using F ST carried out in ARLEQUIN v3.5.2.2 (Excoffier & Lischer, 2010) and Nei's genetic distance estimated in GenAlEx v6.5 for all 65 populations.We also determined the degree of genetic differentiation (F ST ) among the six genetic groups using ARLEQUIN.Furthermore, we used GenAlEx to perform the analysis of molecular variance (AMOVA) separately on various levels: populations, species, groups, and populations within the same group.

| Spatial pattern of genetic diversity
We used the Inverse Distance Weighted (IDW) interpolation function in the Geographic Information System (GIS) software ArcGIS v10.7 (ESRI, Redlands, CA, USA) to infer the geographic patterns of allelic richness (A R ) and expected heterozygosity (H E ) for all 65 populations.Furthermore, we tested for correlation between genetic diversity (H E , expected heterozygosity) and elevation, altitude, longitude, latitude, annual mean temperature and annual mean precipitation using the R package vegan (Dixon, 2003) for the two species, that is, J. regia and J. sigillata.The climate data of annual mean temperature and annual mean precipitation were downloaded from WorldClim 2 (Fick & Hijmans, 2017), and subsequently extracted using GPS points for each population (Table S1) in ArcGIS.

| Gene flow and geographical barriers
We estimated the extent of gene flow for the different genetic groups of walnuts in the Himalaya based on STRUCTURE analysis using two approaches as was done for apples by Bina et al. (2022).Firstly, we tested whether there was a significant isolation-by-distance (IBD) pattern based on Data set 1. We estimated the correlation between and the geographic distance between geographical sites using the R package vegan.Second, we investigated the bidirectional gene flow for a pairwise group with a Bayesian assignment method implemented in ARLEQUIN and then visualized the results using the R package circlize (Gu et al., 2014).Based on the STRUCTURE clustering at K = 4, all individuals were divided into four groups (JR1, JR2, JR3, and JS) plus two types of hybrids.From this, the bidirectional pairwise gene flow among these six groupings was estimated.
We performed the geographical barrier test in Barrier v2.2 (Manni et al., 2004) to infer the geographical features corresponding to pronounced genetic discontinuity.Monmonier's maximum difference algorithm was used to identify the boundaries and barriers with 100% posterior density.

| Genetic clusters of walnuts in the Himalaya
We genotyped 1410 samples with 31 microsatellite markers, yielding a large data set with 0.47% missing data (Table S3).Most loci were in Hardy-Weinberg equilibrium except for a few loci that presented disequilibrium in a few populations (Table S4), and no pairs of loci were found to exhibit linkage disequilibrium (Table S5).Hence, all these data were used for downstream analysis.The STRUCTURE analysis of the whole data set indicated that K = 2 was the optimal K value, followed by K = 4 (Figure S1).At K = 2, all individuals from the 65 populations were divided into 866 J. regia, 291 J. sigillata, and 253 putative hybrids (0.2 < q < 0.8; Table S6; Figurea 2a and S2a).
Most populations showed substantial admixture from K = 2 to K = 4.
Interestingly, both JR1 and JR3 showed admixture with J. sigillata, but JR2 remained relatively pure (Figure 2a).Thus, at K = 4, all samples clustered into four pure groups (JR1, JR2, JR3, and JS) plus two hybrid types, namely Hybrid1 (JR1 × JS) and Hybrid2 (JR3 × JS; Table S6; Figure 2a).The two hybrid groups were distributed in the central Himalaya and northwestern Yunnan Province respectively, while the JS was distributed from the eastern Himalaya to the western Yunnan Province (Figure S2).Accordingly, STRUCTURE results at K = 6-10 also supported the further genetic separation of the two hybrid groups (Table S6; Figure S1a).
In the PCoA results for Data set 1 (with all 1410 samples included), the first and the second coordinates explained 12.64% and 8.84% of the variation, respectively (Figure 2b).Consistent with the STRUCTURE results, the four pure groups (JR1, JR2, JR3 and JS) clustered separately in the PCoA plot, with the putative hybrids dispersed between their respective parents (Figure 2b).Neighbour-net analysis results also showed a similar clustering pattern, with all JR1, Hybrid1, and some JS samples forming one cluster, all JR3, Hybrid2, and the remaining JS samples forming a second cluster, whereas JR2 formed a third cluster (Figure 2c).
When the 253 putative hybrids were removed, the four groups inferred from STRUCTURE at K = 4 formed clearly differentiated clusters in the PCoA based on Data set 2 (Figure 2d).The first and second coordinates explained 14.45% and 8.05% of the variation, respectively.The NJ tree with the putative hybrids excluded also confirmed the four genetic groups, with JR1, JR2, JR3 and JS, forming four distinct clades (Figure 2e).

| Maternal origin of hybrids
Haplotype network analysis based on the plastid genome indicated that J. regia and J. sigillata harbour different haplotypes, with the hybrids sharing the parental haplotypes (Figure 3a).Likewise, J. regia and J. sigillata formed two separate branches in the neighbourjoining (NJ) tree, with the hybrids embedded within them.Among the 25 tested hybrids, nine clustered with J. regia, and the other 16 hybrids with J. sigillata (Figure 3b).Each of the Hybrid1 and Hybrid2 groups contained haplotypes of both J. regia and J. sigillata, indicating bidirectional hybridization between the two species.

| Genetic diversity and differentiation at the population and group level
Overall, the levels of genetic diversity were similar across loci.
The total number of alleles per locus (N A ) ranged from 5 to 12, with a mean N A of 9 (Table S7).Across populations, N A varied from 2.10 (pop 46) to 4.52 (pop 10), H O from 0.29 (pop 64) to 0.64 (pop 1) and H E from 0.29 (pop 64) to 0.61 (pop 12; Table S1).In general, the populations from the western Himalaya showed relatively higher genetic diversity than those from the eastern side (Table S1; Figure S3).Specifically, populations 12, 16, 19 and 20   S1).
The inbreeding coefficient (F IS ) ranged from −0.11 (pop 23) to 0.15 (pop 18), indicating relatively low levels of inbreeding.Estimates of A R varied considerably between populations, ranging from 2.09 in population 46 to 4.19 in population 20.The number of private alleles was relatively low, varying from 0 to 3, with an average of 0.43 across the 65 populations (Table S1).Analysis of molecular variance (AMOVA) for all walnut populations in the Himalaya indicated that 62% of the diversity resided within the populations and 38% among them (Table S8).AMOVA at the species level indicated that 28% of the genetic diversity was partitioned between J. regia and J. sigillata and 72% within them.An AMOVA examining the six groups determined that 26% of the variation was among groups, and 74% within them.Similar patterns were also detected for AMOVA in different groups (Table S8).
Analysis of genetic differentiation based on all 31 loci suggested moderate-to-strong genetic structure among the 65 populations.
Patterns of pairwise unbiased Nei (uNei) distance were consistent with the F ST values.Pairwise F ST among populations varied from 0.005 to 0.535 (Table S9).According to Wright's (1978) criterion, about 50% of population pairs showed a significantly high level of genetic differentiation (F ST > 0.25), with most of the other half showing medium levels of genetic differentiation (0.05 < F ST < 0.25), and only a few (about 2% of population pairs) exhibiting low genetic differentiation (F ST < 0.05; Table S9; Figure S4) .Overall, J. sigillata showed lower genetic differentiation between intraspecific populations than J. regia (Figure S4).Genetic diversity and pairwise differentiation were also evaluated among the six groups inferred based on STRUCTURE results at K = 4 (Table 1; Figure 4a).Among the three groups of J. regia, JR1 had the highest number of total alleles (N T = 222), private alleles (N P = 19), average alleles (N A = 7.16), expected heterozygosity (H E = 0.63) and allelic richness (A R = 6.40).The pairwise F ST values among the six groups ranged from 0.04 to 0.28 (Figure 4a).The genetic differentiation between JR2 and JR3 (0.21) was higher than that between JR1 and JR2 (0.10), or between JR1 and JR3 (0.13).
We noted a relatively higher degree of differentiation between the two Juglans species, with F ST values of JR1-JS, JR2-JS, and JR3-JS being 0.16, 0.28, and 0.27, respectively (Figure 4a).As expected, the F ST estimate between a hybrid and its nonparental group was higher than that between the hybrid and its parental group.For instance, Hybrid1-JR3 (0.21) > Hybrid1-JR2 (0.15) > Hybrid1-JR1 (Figure 4a).

| Spatial pattern of genetic diversity
We detected a significant pattern of isolation by distance across the Himalaya for both J. regia (r = 0.209, p = 0.0001) and J. sigillata (r = 0.764, p = 0.0001; Figure 5a).For J. regia, H E had a significant negative correlation with both altitude and longitude but was positively correlated with latitude and mean annual temperature (Figure 5b-e) but had no significant correlation between H E and annual mean precipitation.None of these variables were significantly correlated with H E for J. sigillata.

| Gene flow and biogeographic barriers
We calculated the bidirectional pairwise contemporary gene flow among the four genetic groups and two hybrid classes.Results suggested strong gene flow between hybrids and their parents, with minimal gene flow between the two species (Figure 4b).Juglans populations in the Himalayan region with 100% posteriors (Figure 6).Barriers I, II and III separated JR2 from most of JR1, a single eastern outlier of JR1 (population 28, in which 22 individuals matched J. sigillata for morphology but four were hybrids), and JR3 plus J. sigillata, respectively.Barrier IV separated JR3 from an area occupied by only J. sigillata to its west, although this area of JR3 appeared to grade into J. sigillata to its south, with no barrier detected there (Figure 6).Overall, the biogeographical barriers supported the genetic grouping inferred from STRUCTURE at K = 4 (Figures 2a and 6).

| Population structure and genetic diversity of walnuts in the Himalaya
Genetic clustering analysis (STRUCTURE, PCoA, Neighbour net analyses and Neighbour-joining tree) produced concordant results, dividing all 1410 Juglans samples into six groups: one for J. sigillata (JS), three for J. regia (JR1, JR2, JR3), and two hybrid groups, Hybrid1 (JS × JR1) and Hybrid 2 (JS × JR3; Figures 2 and S1).The natural dispersal distance of walnuts is quite limited (Tamura et al., 1999), making geographical isolation the key factor in driving differentiation between and within species.Accordingly, biogeographical barriers were detected, which roughly fit with the ridge of the eastern Himalaya, Yarlung Zangbo Grand Canyon, and Baxoila Ling.
These barriers may have effectively hindered the natural dispersal of walnut pollen and seeds and also blocked historical germplasm exchange by human beings, hence possibly explaining the distinct and non-overlapping ranges of the three genetic groups of J. regia, and the genetic differentiation between them (JR1, JR2 and JR3).
Although we detected one geographical barrier (IV) within J. sigillata populations, this species comprised only one genetic group (JS in Figures 2a and S1) separated into two geographical areas (Figure 6), and the barrier might reflect separation between one of these areas and intervening material of J. regia group JR3, rather than separation within J. sigillata.
The Himalaya are considered as one of the multiple refugia and genetic reservoirs of J. regia during the Last Glacial Maximum (Beer et al., 2008;Pollegioni et al., 2014Pollegioni et al., , 2017)).This view is supported by the presence of J. regia pollen in the Gorkha Himalaya (Central Nepal) dated to ~18,000 years ago (Schlütz & Zech, 2004), and nutshells from Kashmir dated to ~4700-4000 years ago (Pokharia et al., 2018).Our study provides genetic evidence that supports and refines this hypothesis.Our Mantel test results and spatial interpolation of genetic diversity indicated that genetic   S1).
Similar patterns of higher genetic diversity in the west were also found in the similarly distributed Taxus contorta and Incarvillea arguta (Poudel et al., 2014a(Poudel et al., , 2014b;;Rana et al., 2021).Therefore, the western Himalaya appears to be a genetic diversity hotspot of common walnut and might have contained the main glacial refugium for this species.
The PCoA and NJ analysis for all populations indicated that within J. regia, the JR2 group is genetically closer to JR1 than to JR3 (Figure 2d,e).Indeed, the pairwise F ST value between JR2 and JR3 was higher (F ST = 0.21) than the value between JR1 and JR3 (F ST = 0.13) or JR1 and JR2 (F ST = 0.10), despite JR1 and JR3 being most distant geographically.Hence, the degree of genetic differentiation for different group pairs was not positively correlated with the geographical distance between them (Figure 4a).This indicates that the two barriers separating JR2 from JR3 (i.e., III, Yarlung Zangbo Grand Canyon and IV, Baxoila Ling; Figure 6) might together form a more powerful barrier to gene flow than barriers I or II.Consistent with this, bidirectional gene flow between JR1 and JR2 was much stronger than between the pairs JR1-JR3 and JR2-JR3 (Figure 4b).
However, this pattern might also result from J. regia dividing into separate western and eastern refugia during glacial maxima, with JR1 and JR2 deriving from the former, and JR3 the latter.Consistent with JR2 having been derived from westward expansion from a major refugium further west, JR1 had 19 private alleles compared with only one in JR3 and none in JR2 (Table 1).However, STRUCTURE at K = 3 resolved JR2 as a mixture of JR1 and JR3, indicating that the barrier between JR2 and JR3 might not be absolute.

| Drivers of the spatial genetic structure of walnuts in the Himalaya
A combination of biotic and abiotic factors, together with species traits, can shape the distribution of species in different ways at different spatial scales (Chau et al., 2019).This complex interplay of human and environmental factors probably also occurred in the Himalaya.The STRUCTURE, PCoA, Neighbour Joining and population divergence analyses indicated considerable genetic differentiation among the four walnut groups in the Himalaya (JR1, JR2, JR3 and JS; Figures 2 and 4).In addition, Mantel tests supported the isolation by distance model of expansion for J. regia across the Himalaya (Figure 5).Specifically, geographic and climatic factors including altitude, longitude, latitude, and annual mean temperature, plus the barriers noted above, had all significantly affected the genetic diversity of J. regia.Furthermore, the geographical barriers shaped and maintained the genetic clustering of J. regia in the Himalaya, the detected four barriers separated J. regia into three groups (JR1, JR2, JR3; Figure 6).Therefore, the abiotic factors might have affected the spatial genetic diversity and shaped the spatial genetic structure of walnuts in the Himalaya.
In addition to abiotic factors, human-mediated dispersal processes might have influenced the spatial genetic patterns of J. regia over the centuries (Pollegioni et al., 2020).Juglans regia is phylogenetically closest to J. sigillata (Aradhya et al., 2007), and both species have been cultivated in China for thousands of years (Xi & Zhang, 1996).It is reported that J. regia and J. sigillata had different natural distribution ranges (Laufer, 1919;Lu et al., 1999;Zohary et al., 2012), and were allopatric until humans expanded the range of J. regia.In our study, we detected frequent bi-directional hybridization and gene flow between J. regia and J. sigillata in the Himalaya (Figures 2-4), and the two hybrid types were detected in 35 of the 65 populations (Table S6; Figures 2 and 6).(Qi et al., 2013;Wang, 1994).Other recent evidence suggests that modern humans occupied the higher altitudes of the Tibetan Plateau and Himalaya at least 3000 years before the present (Chen et al., 2015;Liu, Witonsky, et al., 2022;Zhang et al., 2022).In addition, a fossil record of walnut nutshells, dating back to ~3200 years ago, has been confirmed in an ancient market site in Pakistan (located in the western Himalaya) (Spengler et al., 2020).These pieces of evidence point to a long history of human activity in the Himalaya, which probably drove the range expansion of J. regia and J. sigillata between two sides of Himalaya, ultimately causing interspecific genetic introgression in areas of secondary contact.
Our Mantel test showed a significant negative correlation between genetic diversity and altitude (Figure 5), indicating that higher altitude populations derive from a recent upward range expansion.Natural J. regia material generally occurs at 800 m to 2000 m a.s.l.(Eastwood et al., 2009), but the current study recorded populations as high as 3715 m (Figure 5b; Table S1), which might have been founded by movement of cultivated material by ancient Tibetan communities, who migrated seasonally between high and low altitudes.
Collectively, geographic and climatic factors are the key forces driving the spatial genetic structure of J. regia across the Himalaya, while human-mediated dispersal probably promoted dispersal and gene flow between these two species and reshaped the genetic landscape of walnuts in the Himalaya.

| Conservation and utilization implications
Due to anthropogenic habitat fragmentation caused by urbanization, commercialization of traditional orchards, use of wild trees for grafting (Liu et al., 2023), and felling trees for timber, a large proportion of walnut genetic resources is under threat (Vahdati et al., 2019).
Although the common walnut was listed under the 'Least Concern' category in the most recent IUCN Red List (Rivers & Allen, 2017), owing to its exceptionally high genetic diversity, wild J. regia in Kyrgyzstan, Kazakhstan, Uzbekistan, Turkmenistan, and Tajikistan is listed in 'The Red List of Trees of Central Asia' by the BGCI (Eastwood et al., 2009).In addition, J. regia was found to have the highest residual value in terms of plant communities and vegetation ecosystem services in the Naran Valley of western Himalaya, suggesting that it was overused by the local inhabitants and was threatened or vulnerable at the national level (Khan, 2012).Therefore, hotspots of walnut diversity need to be identified to serve as the basis for effective conservation strategies, such as the western Himalaya one identified here.
Frequent bidirectional hybridization with J. sigillata was detected in multiple common walnut populations of the western, central, and eastern Himalaya (Figures 2 and 5), indicating that some populations have been altered by introgression.Such hybridization and introgression may cause population extinction through genetic swamping (Rhymer & Simberloff, 1996;Todesco et al., 2016).Therefore, conservation efforts must also consider threats to the genetic purity of J. regia and J. sigillata populations.On the other hand, the natural walnut hybrids in the Himalaya harbour high levels of genetic and phenotypic diversity (Table 1) and could provide valuable genetic resources for walnut breeding.Generally, hybridization can produce new genotypes and introduce novel adaptive traits to resist exotic pests and diseases or promote adaptation to the changing climate (Fritz, 1999), and it has been a driving force in the evolution of cultivars, such as apples (Chen et al., 2022).In walnut breeding, artificial crossing is a conventional technique (Vahdati et al., 2019), but this approach is limited by the long generation time and larger growing space of walnut trees.The utilization of biotechnology and genomic information can help breeders bypass the long breeding cycles, and hence potentially aid de novo domestication (Huang et al., 2022), and the examination of useful traits in the natural hybrids detected here could aid this process.However, the links between commercial traits and the genome must be bridged before the approach is applied.Thus, to fully utilize these hybrid resources, genome-wide association studies (GWAS) on the phenotypes of different hybrid lineages are urgently needed.

| CON CLUS ION
In this study, we performed the first large-scale analysis of walnuts in the Himalaya, in which the genetic diversity, structure, and differentiation of J. regia and J. sigillata were thoroughly evaluated.
In this study, we explored the landscape of genetic variation of walnuts in the Himalaya by analysing 31 microsatellite markers and 35 representative plastid genomes of 1410 Juglans individuals collected from 65 populations across China, Nepal, India, Pakistan and Afghanistan in the Himalaya.By integrating multiple lines of evidence, we examined the potential influence of human activities on the genetic landscape of wild plants of domesticated walnut species in the Himalaya.In detail, we aimed to (1) determine the genetic diversity and spatial genetic structure of walnuts, (2) provide new insights into the possible drivers of the genetic landscape of walnuts and (3) offer conservation and management recommendations for wild walnut genetic resources.2 | MATERIAL S AND ME THODS 2.1 | Sample collection, DNA extraction and microsatellite genotyping A total of 65 populations comprising 1410 individuals from across the Himalayan region were used in this study.These included the 11 Pakistani populations comprising 187 individuals reported by Magige et al. (2022).Using morphological traits according to the F I G U R E 1 Spatial map of the 65 walnut populations (see Table S1 for detailed information).(a) Geographical location of all populations.(b) The eight populations distributed in the Pakistani Himalaya enclosed in the green rectangle in (a).(c) The 34 populations distributed in the eastern Himalaya enclosed in the orange rectangle in (a).The blue circles indicate Juglans regia, whereas the red one represents J. sigillata.

F
Genetic structure and relationships of Juglans in the Himalaya.Across all five diagrams, different colours represent the different genetic groups: blue for JR1, green for JR2, yellow for JR3, and red for JS.(a) STRUCTURE results for K = 2 to K = 4.The population numbers highlighted with grey and purple indicate populations that comprise at least 20% of Hybrid1 and Hybrid2, respectively.(b) Principal coordinates analysis (PCoA) plot for 1410 individuals based on SSR data.(c) Neighbour-net tree showing the genetic relationships among individuals, with coloured solid circles at branch terminals corresponding to the four genetic groups inferred in STRUCTURE at K = 4, plus Hybrid1 (grey), and Hybrid2 (purple).(d) Principal coordinates analysis (PCoA) of 1157 individuals based on SSR data with all putative hybrids (0.2 < q < 0.8 at K = 2 in STRUCTURE) removed.(e) Neighbour-joining tree for all 1157 individuals without putative hybrids.F I G U R E 3 Evolutionary relationships among Juglans regia (JR), J. sigillata (JS), and the two hybrid groups (Hybrid1, Hybrid2) inferred using the whole chloroplast genomes of 35 randomly selected samples with J. cathayensis as an outgroup.(a) The chloroplast haplotype network of the two Juglans species and their hybrids.(b) A rooted neighbour-joining (NJ) tree.

TA B L E 1
Genetic diversity estimates for six genetic groups identified based on SSR data.F I G U R E 4 Genetic diversity, divergence and gene flow among different groups of walnuts in the Himalaya.(a) Genetic diversity (H E ) within, and population divergence (F ST ) between, the six genetic groups (JR1, JR2, JR3, JS, Hybrid1 and Hybrid2).Values in circles indicate a genetic diversity estimate; values in black on each dashed line represent pairwise population divergence among groups.(b) Chord diagram of bidirectional gene flow among the six genetic groups.The gene flow out of a group to each of the other five groups was indicated by the same colour.For example, blue indicates gene flow out of JR1.Four potential biogeographic barriers were detected among 65

F
Correlation between genetic variation and environmental factors for the 59 walnut populations.(a) Mantel test between pairwise genetic differentiation (F ST /1-F ST ) and geographic distance; (b-f) Correlations between genetic diversity (H E ) and abiotic factors (i.e., altitude, longitude, latitude, annual mean temperature, annual mean precipitation).
in the west, progressively decreasing eastwards (Figures5c and S2), and the four populations with the highest genetic diversity were all in the western group JR1 (Table

F
Spatial genetic structure and geographical barriers for all 65 populations inferred from STRUCTURE at K = 4.The red lines represent the biogeographic barriers detected with 100% posteriors.The four barriers were identified based on Monmonier's maximumdifference algorithm.(a) Geographical location of all populations, and (b and c) enlarged views of population distributions within the two areas indicated by rectangles in (a).| 11 of 15 YAN et al.
Hybrid1 might have originated from the hybridization between remotely distributed groups JR1 and JS, while Hybrid2 likely originated from the hybridization between the geographically closer related JR3 and JS.Due to the strong biogeographic barriers (three barriers were detected with 100% posteriors) between JR1 and JS (Figure 6), plus the limited natural dispersal distance of walnut pollen and seeds, human-mediated dispersal is likely the main driving force of gene flow between JR1 and JS.Human activity in the Himalayan region can be traced back thousands of years.For example, Chinese archaeologists have proposed that the Di-Qiang people, or Proto-Tibeto-Burman populations, inhabited the Tibetan Plateau during the Yang-Shao epoch (about 8000 years ago) and brought agriculture to the Himalayan region Xingdian Talent Support Program of Yunnan Province, Grant/Award Number: YNWR-QNBJ-2018-146 and XDYC-QNRC-2022-0068; CAS "Light of West China" Program; Natural Science Foundation of Yunnan, Grant/Award Number: 202201AT070222; Science and Engineering Research Board, Govt. of Inida, New Delhi, Grant/Award Number: CRG/2021/000665; Ministry of Environment Forest and Climate Change, Govt. of India, New Delhi, Grant/Award Number: 22018/13/2015-RE(Tax) Note: N, sample size; N T , number of total alleles; N P , private alleles; N A , average number of alleles; N E , number of effective alleles; H O , observed heterozygosity; H E , expected heterozygosity; uH E , unbiased expected heterozygosity; A R , allelic richness; F IS , inbreeding coefficient.