Whole-genome sequencing revealed the genomic origin of Solanum demissum Lindl., a Mexican hexaploid wild potato species Authors: Awie J Hosaka, Kazuyoshi Hosaka Published: 2026 ABSTRACT AbstractBackgroundSolanum demissum Lindl. (2n = 6x = 72) is the most popular wild species in potato breeding because of its resistance to late blight, caused by Phytophthora infestans (Mont.) de Bary. Regular meiosis with 36 bivalents indicates that S. demissum has three distinct genomes, but their origin remains unknown.ResultsWe constructed a chromosome-scale assembly of S. demissum using PacBio long-read sequencing and Hi-C scaffolding technologies. The final assembly comprised 2.035 Gb anchored to 36 chromosome-level scaffolds, which were phased into three subgenomes (SG1, SG2, and SG3, with sizes of 696.51 Mb, 685.32 Mb, and 653.29 Mb, respectively). Evidence-based annotation, using RNA-seq datasets, and deep learning-based prediction identified 32,888 genes in SG1, 33,028 in SG2, and 31,999 in SG3. Comparing gene sequences among the three subgenomes and 47 other genomes from diverse collections distinctly placed these subgenomes in the phylogenetic tree; the most closely related were SG1 with the maternal and A-genome ancestor S. verrucosum, SG2 with SG2 of S. acaule, and SG3 with SG1 of S. acaule. Syntenic ratios, indicating genome structure similarity, further supported these relationships. The three subgenomes of S. demissum exhibited mostly balanced gene expression. Additionally, the composition and genome occupancy of transposable elements, as well as the number of resistance genes, are similar across the three subgenomes and more closely resemble those of the ancestral genomes in S. acaule and S. verrucosum.ConclusionsSolanum demissum likely originated between a Mexican diploid S. verrucosum as the female parent and a South American tetraploid S. acaule as the male parent.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12870-026-08972-2.Keywords: Potato, Solanum demissum, Hexaploid, Mexican wild potatoes, Chromosome-scale assembly, Crop wild relatives, Solanum acaule, Solanum verrucosum Introduction The common potato (Solanum tuberosum L., 2n = 4x = 48) is the fourth most important food crop grown worldwide [26]. It has an enormous genetic reservoir, comprising over 100 closely related wild species from section Petota of the genus Solanum [38, 123]. These wild species vary in ploidy from diploid to hexaploid and are distributed from the central United States through Central America to Chile, with the Mexican highlands and the central Andes as centers of diversity [43]. Hawkes [38] recognized over 200 wild species and separated non-tuber-bearing and tuber-bearing species in section Petota into the subsections Estolonifera and Potatoe, respectively, and the species in subsection Potatoe into 19 taxonomic series. Recent molecular phylogenies have supported these two subsections as the Etuberosum and Petota clades, respectively, and together with Tomato and some other minor clades, they form the large Potato clade [29]. According to chloroplast DNA restriction site data, section Petota consists of four clades: Clade 1 contains the North and Central American diploid species, exclusive of S. bulbocastanum Dun., S. cardiophyllum Lindl., and S. verrucosum Schltdl.; Clade 2 contains S. bulbocastanum and S. cardiophyllum; Clade 3 contains all examined members of the South American series Piurana and some South American species placed in other series; Clade 4 contains all remaining South American species, S. verrucosum, and the North and Central American polyploid species, including all cultivated potato species [117]. Later, Clade 1 and Clade 2 were combined, and the three-clade phylogeny has been supported by full chloroplast DNA sequence data [53, 138], DNA sequences of granule-bound starch synthase I [122], nitrate reductase [108], and conserved orthologs [3], genome-wide single-nucleotide polymorphism (SNP) analysis [76], and whole-genome sequencing [1, 131, 148]. Among the 19 taxonomic series, series Acaulia, Longipedicellata, and Demissa consist exclusively of polyploid species. Well-known species in each series are S. acaule Bitter (2n = 4x = 48), S. stoloniferum Schltdl. (2n = 4x = 48), and S. demissum Lindl. (2n = 6x = 72), respectively. Compared to diploid species, which are usually adapted to narrow niches, S. stoloniferum and S. demissum are widely distributed in the Mexican highlands and adjacent regions. Solanum acaule is the most widely distributed species, extending from Ecuador southwards into northern Argentina [38, 124]. Genome similarity and differentiation are essential for understanding and utilizing genetic diversity as a genetic resource in breeding. Traditionally, genome similarity has been evaluated based on chromosome-pairing behaviors during meiosis. Diploid hybrids between South American species usually exhibit regular meiosis with 12 bivalents, indicating that the common A genome is shared among them [51]. Solanum tuberosum forms 2–4 multivalents (trivalents and quadrivalents); thus, it is cytogenetically a segmental allotetraploid (genome formula AAAtAt; [85]), although it is generally referred to as an autotetraploid (genomic formula AAAA). In contrast, S. acaule, S. stoloniferum, and S. demissum show regular meiosis with complete formation of bivalents [85]. Triploid hybrids, generated from a diploid A-genome species with S. acaule exhibit a high frequency of trivalents, indicating that the same A genome is shared [130]. A dihaploid clone induced through anther culture from S. acaule presented 10.3 to 10.6 bivalents and 2.8 to 3.4 univalents [13, 137], indicating considerable similarity between the two genomes of S. acaule. Thus, the genome formula AAAaAa has been given to S. acaule [83]. In contrast, triploid hybrids between S. stoloniferum and a diploid A-genome species presented 12 bivalents and 12 univalents, indicating that S. stoloniferum has an A and another genome [80, 81]. A dihaploid clone of S. stoloniferum predominantly exhibited univalents at meiosis, supporting its allotetraploid nature [56]. The second genome of S. stoloniferum has been referred to as the B genome [85]. Whole-genome sequencing of S. stoloniferum revealed its allopolyploid nature, and its B genome is highly homologous to the genomes of the Mexican diploid species belonging to Clade 1+2 [48]. The other genome so far identified in the diploid species is the E genome in the non-tuber-bearing species in the series Etuberosa [104]. Solanum demissum is the most popular wild potato species since the early 1900s because of its resistance to late blight, caused by Phytophthora infestans (Mont.) de Bary [101, 109]. Solanum demissum is highly self-fertile and sets abundant berries in nature. When S. demissum is used as a female parent, pentaploid hybrids (2n = 60) can be easily obtained from a cross with S. tuberosum [20, 111]. By continued backcrossing, many late blight resistant cultivars were released [101, 109]. To date, 11 race-specific resistance genes R1 − R11 have been identified from S. demissum, of which R1, R2, R3a, R3b, and R8 have been cloned [7, 55, 72, 78, 134]. However, the efficient reproduction of P. infestans in both asexual and sexual forms, combined with fast-evolving effector R genes necessary for successful infection [34], makes it a rapidly evolving pathogen that can readily generate new virulent strains. Consequently, none of the single or stacked resistance genes, even when combined in a single cultivar, conferred durable resistance [28]. The rapid adaptation of P. infestans makes it very difficult to breed durably resistant potato varieties [28, 87]. Solanum demissum regularly forms 36 bivalents at meiosis [79, 130]. Different clones of its triploid haploid formed 4.74 to 9.78 bivalents per cell in addition to univalents [5, 22, 52, 56, 79]. These observations suggested that S. demissum has two somewhat similar genomes and a third genome that is rather different. Mean bivalents frequencies in the tetraploid hybrids of S. demissum with A-genome diploid species have been reported as follows: 17.64 and 21.21 with S. verrucosum ( [102] and [65], respectively), 16.25−-16.46 with S. chacoense, 19.78 with S. phureja Juz. et Buk., and 20.47 with S. stenotomum Juz. et Buk. [82]. Based on these observations, the third genome of S. demissum is considered to be an A genome [56, 79, 82]. The currently proposed genome formula is AADDDdDd [85]. The most likely maternal and A-genome ancestor is S. verrucosum, as suggested by chloroplast DNA similarity and the shared nature of a mitochondrial molecule called Band 1 [49, 110, 119]. However, genomic in situ hybridization (GISH) analysis revealed that S. demissum has all three chromosome sets related to the basic A genome, similar to the GISH results for S. acaule [98]. Granule-bound starch synthase and nitrate reductase gene phylogenies indicated that the two distinct alleles identified in S. demissum clustered within the same clade as A-genome species [108, 122]. Molecular marker analyses revealed that the species most closely related to S. demissum are S. acaule and S. albicans (Ochoa) Ochoa of the series Acaulia [19, 63, 91]. Solanum albicans (2n = 6x = 72) was likely derived from S. acaule and a diploid species from another series [37], which possesses a clearly differentiated genome from A or Aa genomes [44, 85]. Many authors have recognized morphological similarities among S. demissum, S. acaule, and S. albicans [17, 39, 40, 62, 94, 120]. Along with molecular similarities, Spooner et al. [121] grouped S. demissum, S. acaule, and S. albicans into a single group, the Acaulia Group. Recently, Achakkagari et al. [1] conducted whole-genome sequencing for S. acaule and revealed that it originated from a maternal progenitor S. megistacrolobum Bitt. (combined into S. boliviense Dun. by Spooner et al. [124], an A-genome species from Clade 4, and an unknown or extinct species from Clade 3. Clade 3 is a large clade containing approximately one-third of the tuber-bearing Solanum species and is relatively unexploited [3, 123]. Thus, the second genome of S. acaule likely originated from an unexamined species of Clade 3. The presence and source of the distinct genome D or Dd in S. demissum have remained unknown. In this study, we sequenced the three subgenomes of S. demissum and revealed that the most closely related genomes were the A genome of S. verrucosum and the two S. acaule subgenomes. Therefore, we suggest that S. demissum originated from S. verrucosum as the female and S. acaule as the male. Potential disease- and pest-resistance genes were identified in the S. demissum genome, and their significance and aspects for utilization are also discussed. Materials and methods Plant material Solanum demissum clone 24H65-12 was used for sequencing, which was derived by self-crossing from 5H109-5. 5H109-5 was initially grown as a seedling (accession PI 186551), obtained from the US Potato Genebank at Sturgeon Bay, Wisconsin, USA, and used for genetic map construction [95] and QTL analysis of crossability [45]. Sequencing Young leaves were collected from the mature plant and frozen until use. Large-sized DNA was extracted from frozen leaves using a NucleoBond HMW DNA kit (Macherey-Nagel, Düren, Germany) according to the manufacturer’s protocol, with modifications as previously described [48]. PacBio HiFi reads were commercially obtained from the S. demissum genomic DNA at Macrogen Japan Corp. (Tokyo, Japan) using the PacBio Revio (PacBio). The Hi-C library was prepared as previously described [48]. The NextSeq 1000 (Illumina, Inc.) was used for sequencing the Hi-C library according to the manufacturer’s protocol. Genome assembly Hifiasm 0.25.0-r726 [15] was used for assembly with HiFi reads, employing the “–hom-cov 93” option based on the sequence coverage. To remove organellar genome-derived sequences from the assembled contigs, contigs shorter than 1 Mb were first extracted. Homologous regions to the mitochondrial and chloroplast genomes of S. phureja (DM v 8.1; Yang et al. [141] were subsequently searched using blastn 2.12.0 [2]. Contigs containing homologous regions longer than 10 kb were considered organellar-derived and excluded from downstream analyses. For chromosome-level assembly, a Hi-C library was prepared and sequenced on the NextSeq 1000. Hi-C reads were aligned to the contigs using BWA 0.7.17-r1188 [74]. PCR duplicates were removed using SAMBLASTER [27], and secondary and supplementary alignments were filtered out using the SAMtools view command with the “samtools view -S -h -b -F 3340” option [75]. The resulting BAM file was used in the HapHiC 1.0.7 pipeline [145] to scaffold contigs onto chromosomes. During this step, the option “-nchr 36” was specified to indicate the expected number of chromosomes. The scaffolds were reviewed and manually corrected using JuiceBox 1.11.08 (https://github.com/aidenlab/Juicebox). To phase the chromosomes into three subgenomes, SubPhaser 1.2.6. [58] was used with default parameters. RNA-seq RNA was extracted from leaves, floral buds, flowers, and stolons of the mature S. demissum plant using a Plant/Fungi Total RNA Purification Kit (Norgen Biotek Corp., Ontario, Canada). After fresh samples were frozen in liquid nitrogen, ground in a mortar with a pestle, and lysed with 600 µl of Lysis Buffer C at 55 ℃ for 5 min, the lysate was centrifuged at 12,000 rpm for 2 min to precipitate large debris and applied to the Filter Column. Subsequent steps were followed according to the manufacturer’s protocol, including DNase treatment with RNase-Free DNase I Kit (Norgen Biotek Corp.). Three biological replicates were taken from each tissue. Library preparation and sequencing were performed commercially by Novogene (Nippon Genetics Co., Ltd., Tokyo, Japan). The raw reads were qualified and trimmed with fastp 0.23.2 [16]. Annotation of protein-coding regions For annotation of the S. demissum genome, RNA-seq datasets were initially utilized for evidence-based annotation, as previously described [48]. Next, deep-learning-based annotation was performed by using the Helixer v0.3.4 web tool [127]. To merge the evidence-based annotation and deep-learning-based annotation, GffCompare v0.12.10 was used with the Helixer GFF as the query and the StringTie/TransDecoder-derived GFF3 as the reference (-r), with outputs written under the prefix dem_gffcmp (-o) [68, 99]. From the resulting .tmap file, entries assigned the class code u (transcripts unknown with respect to the reference, i.e., intergenic/novel) were extracted, and the corresponding query transcript IDs were collected. The original Helixer GFF was then filtered to retain only these u-class features, ensuring that only putative novel Helixer models were preserved while overlapping predictions were excluded. The retained Helixer features were concatenated with the StringTie/TransDecoder-derived GFF3, and the combined records were lexicographically sorted by sequence ID to generate a unified GFF file. To standardize feature identifiers, AGAT v1.3.3 (https://zenodo.org/records/18230424) was applied with a project-specific prefix (–prefix dem) and TAIR-style naming (–tair), resulting in a final annotation dataset. Evaluation of assembly and annotation quality Statistics on contig count and length distribution for both the draft and final assemblies were generated using the “stat” function in Seqkit v0.15.0 [113]. The completeness of genome assembly and gene annotation was further evaluated using Benchmarking Universal Single-Copy Orthologs (BUSCO; [114]), employing the Solanales odb10 dataset, which provides genomic and protein sequence references [69]. TE annotation Annotation of transposable elements (TEs) and phylogenetic analyses within S. demissum and related species were conducted as previously described [48]. Briefly, TEs were identified using EDTA v2.0 [96]. The two long terminal repeats (LTRs) at both ends of a retrotransposon are identical at the beginning of the insertion, but their differences increase over time. Thus, we obtained LTR identity information from intact TE annotations and examined LTR identity distributions as a proxy for insertion age to assess recent TE activity. To infer transposition histories of TEs, sequence divergence of TEs relative to their consensus sequences was estimated using RepeatMasker v4.1.1 (Smit AFA, Hubley R, and Green P, at http://repeatmasker.org) with EDTA-generated TE libraries, followed by calculating Kimura two-parameter (K2P) divergence metrics [67] using the accompanying RepeatMasker utility scripts. For Gypsy retrotransposon analyses, intact elements were extracted and classified based on conserved protein domains using TEsorter [146] with the REXdb-plant database [92]. Phylogenetic relationships were inferred from concatenated reverse transcriptase (RT), RNase H (RH), and integrase (INT) domains using VeryFastTree 4.0.3 [100], and trees were visualized with GGTREE 3.10.1 [143]. HiFi read mapping, variant calling, and window-based heterozygosity analysis PacBio HiFi reads were aligned to the S. demissum reference genome (subgenome-resolved scaffolds) using minimap2 [73] with the map-hifi preset. The resulting alignments were converted to BAM format, sorted, and indexed using SAMtools 1.61.1 [75]. Single-nucleotide variants were called from the mapped reads using BCFtools 1.19 [18]. Briefly, pileup files were generated with the “bcftools mpileup” command, and variants were called with the “bcftools call” command under a diploid model. Variant calls were output in VCF format for downstream analyses. To assess genome-wide patterns of read depth and heterozygosity, the reference genome was divided into 1-Mb non-overlapping windows using BEDTools [103]. Read depth for each window was calculated using Mosdepth 0.3.3 [97]. Variant calls were filtered based on read depth to retain high-confidence sites. Only variants with total read depth values between 50 and 200, as estimated from both the DP and DP4 fields in the VCF, were retained. From these filtered variants, heterozygous sites (0/1 genotypes) were extracted. The number of heterozygous variants within each 1-Mb window was then counted using BEDTools, generating a chromosome-scale profile of heterozygosity across the genome. Expression analysis RNA-seq reads were aligned to the S. demissum genome using HISAT2 v2.2.1 [66]. Then, read counts for each gene were calculated using featureCounts v2.0.6 [77] and normalized to transcripts per million (TPM). Spearman’s rank correlation of single-copy orthologs was calculated from the TPM-normalized expression matrix, and a distance matrix was used for hierarchical clustering with the Ward.D2 method [90]. The clustering results were visualized as a dendrogram. To assess expression balance among subgenomes, only ortholog sets in which each gene was assigned to all three subgenomes (SG1, SG2, and SG3) were used. For each ortholog set, a proportion of the TPM value in each homoeolog to the sum of TPM values of the three homoeologs was calculated. The resulting expression balances were visualized using the R package ggtern [35]. For descriptive classification of homoeolog expression bias, ortholog sets were categorized as SG1-, SG2-, or SG3-dominant when one subgenome accounted for more than 60% of the total expression, and each of the other two accounted for less than 20%. Ortholog sets were categorized as SG1-, SG2-, or SG3-suppressed when one subgenome accounted for less than 20% of the total expression, and each of the other two accounted for more than 20%. All remaining ortholog sets were classified as balanced. Phylogenetic inference The phylogenetic relationships of the subgenomes of S. demissum among tuber-bearing Solanum species were inferred using OrthoFinder [25], as previously described [48]. The dataset comprised 50 genomes listed in Table 1, including a broad representation of diploid tuber-bearing Solanum species, the S. demissum subgenomes (SG1, SG2, and SG3), the subgenomes of the allotetraploids S. stoloniferum (A and B) and S. acaule (SG1 and SG2), and eggplant (S. melongena) as an outgroup (Table 1). Most of the whole-genome sequence data were obtained from publicly available databases. Briefly, single-copy ortholog groups were detected and aligned using MAFFT [64] by enabling the “-M msa” and “-A mafft” options. The tree was inferred using IQ-TREE 2 [89] with the “-m MFP” and “-bb 1000” options. The tree was visualized using FigTree v1.4.4 (http://tree.bio.ed.ac.uk/software/figtree/). Table 1Solanum genomes compared in this studyHawkes’ taxonomic series and speciesa)AccessionSpooner’s taxonomyGenome designation in this studyb)ReferenceS. melongena (2x)67/3S. melongenaS. melongena v4.0*Barchi et al. [8]S. lycopersicum (2x)cv. Heinz 1706S. lycopersicumS. lycopersicum v5.0*Zhou et al. [149]Series Etuberosa S. brevidens (2x; A, Ch)PG0009 (PI 558254)S. palustreS. palustreTang et al. [131] S. etuberosum (2x; Ch)PG0019 (PI 558302)S. etuberosumS. etuberosum*Tang et al. [131]Series Morelliformia S. morelliforme (2x; G, M, H)PG1011 (PI 243357)S. morelliformeS. morelliformeTang et al. [131]Series Bulbocastana S. bulbocastanum (2x; G, H, M)11H21 (PI 666967)S. bulbocastanumS. bulbocastanum*Hosaka et al. [47]Series Pinnatisecta S. brachistotrichum (2x; M)PG1015 (PI 498216)S. stenophyllidiumS. stenophyllidium*Zhang et al. [148] S. cardiophyllum ssp. cardiophyllum (2x; M)23H17-1 (PI 283062)S. cardiophyllumS. cardiophyllumHosaka et al. [48] S. cardiophyllum ssp. ehrenbergii (2x; M)PG1006 (PI 275212)S. ehrenbergiiS. ehrenbergii*Zhang et al. [148] S. jamesii (2x; M, USA)PG1008 (PI 458423)S. jamesiiS. jamesiiTang et al. [131] S. × michoacanum (2x; M)PG1023 (PI 653799)S. michoacanumS. michoacanum*Zhang et al. [148] S. pinnatisectum (2x; M)PG1013 (PI 275232)S. pinnatisectumS. pinnatisectumTang et al. [131] S. tarnii (2x; M)PG1018 (PI 570641)S. tarniiS. tarnii*Zhang et al. [148]Series Commersoniana S. commersonii ssp. commersonii (2x; A, Br, U)PG4049(PI 473412)S. commersoniiS. commersoniiTang et al. [131] S. commersonii ssp. malmeanum (2x; A, Br, Pr, U)PG4019 (PI 472841)S. malmeanumS. malmeanum*Zhang et al. [148]Series Circaeifolia S. circaeifolium (2x; B)PG4033 (PI 498117)S. stipuloideumS. stipuloideum*Zhang et al. [148]Series Lignicaulia S. lignicaule (2x; P)PG4017 (PI 473351)S. lignicauleS. lignicauleTang et al. [131]Series Yungasensa S. chacoense (2x; A, B, Br, Pr, P, U)M6S. chacoenseS. chacoense M6 v5*https://spuddb.uga.edu/M6_v5_0_download.shtml S. huancabambense (2x; P)PG3008 (PI 498244)S. huancabambenseS. huancabambense*Zhang et al. [148]Series Megistacroloba S. megistacrolobum (2x; A, B, P)PG5076 (PI 473122)S. bolivienseS. bolivienseTang et al. [131] S. sogarandinum (2x; P)PG4032 (PI 365360)S. sogarandinumS. sogarandinumTang et al. [131]Series Cuneoalata S. infundibuliforme (2x; A, B)PG4013 (PI 597699)S. infundibuliformeS. infundibuliforme*Zhang et al. [148]Series Conicibaccata S. buesii (2x; P)PG4041 (PI 607889)S. buesiiS. buesiiTang et al. [131] S. chomatophilum (2x; E, P)PG3005 (PI 310943)S. chomatophilumS. chomatophilumTang et al. [131] S. irosinum (2x; P)PG4005 (PI 568985)S. burkartiiS. burkartiiTang et al. [131]Series Piurana S. acroglossum (2x; P)PG3001 (PI 498204)S. acroglossumS. acroglossum*Zhang et al. [148] S. albornozii (2x; E)PG3002 (PI 561636)S. albornoziiS. albornozii*Zhang et al. [148] S. paucissectum (2x; P)PG3022(PI 365340)S. paucissectumS. paucissectumTang et al. [131] S. piurae (2x; P)PG3023 (PI 365365)S. piuraeS. piuraeTang et al. [131]Series Tuberosa S. acroscopicum (2x; P)PG4001 (PI 365314)S. acroscopicumS. acroscopicum*Zhang et al. [148] S. andreanum (2x; C, E)PG3003 (PI 498148)S. andreanumS. andreanumTang et al. [131] S. bukasovii (2x; P)PG5068 (PI 458379)S. candolleanumS. bukasovii*Tang et al. [131] S. cajamarquense (2x; P)PG6242 (CIP 762608)S. cajamarquenseS. cajamarquenseTang et al. [131] S. gourlayi (2x; A, B)PG5032(PI 545977)S. brevicauleS. gourlayiTang et al. [131] S. immite (2x; P)PG3009 (PI 498245)S. immiteS. immite*Zhang et al. [148] S. mochiquense (2x; P)PG3010 (PI 498411)S. mochiquenseS. mochiquense*Zhang et al. [148] S. multiinterruptum (2x; P)PG4060 (PI 498265)S. multiinterruptumS. multiinterruptumTang et al. [131] S. neorossii (2x; A)PG6243(CIP 764170)S. neorossiiS. neorossiiTang et al. [131] S. vernei (2x; A)PG4036(PI 473304)S. verneiS. verneiTang et al. [131] S. verrucosum (2x; M)11H23 (PI 666968)S. verrucosumS. verrucosum*Hosaka et al. [46]Series Tuberosa (cultivated) S. phureja (2x; B, C, E, P, V)DM1-3 516 R44S. tuberosum Andigenum Group (2x)S. phureja DM8.1*Yang et al. [141] S. stenotomum ssp. goniocalyx (2x; P)PG6359 (CIP 705468)S. tuberosum Andigenum Group (2x)S. stenotomum*Tang et al. [131]2x S. tuberosum (2x)RH89-039–16 2x S. tuberosumS. tuberosum RH*Tang et al. [131]Series Acaulia S. acaule (4x; A, B, E, P)AA9 (PI 498202)S. acauleS. acaule SG1*Achakkagari et al. [1]S. acaule SG2*Achakkagari et al. [1]Series Longipedicellata S. stoloniferum (4x; M, USA)11H31S. stoloniferumS. stoloniferum A*Hosaka et al. [48]S. stoloniferum B*Hosaka et al. [48]Series Demissa S. demissum (6x; G, M)24H65-12 (PI 186551)S. demissumS. demissum SG1*This studyS. demissum SG2*This studyS. demissum SG3*This studya)Ploidy and the countries distributed are given in parentheses. Country codes are A Argentina, B Bolivia, Br Brazil, C Colombia, Ch Chile, E Ecuador, G Guatemala, H Honduras, M Mexico, Pr Paraguay, P Peru, USA United States of America, U Uruguay, and V Venezuelab)Genomes and subgenomes (SGs) with chromosome-scale assemblies are shown with asterisks Solanum genomes compared in this study a)Ploidy and the countries distributed are given in parentheses. Country codes are A Argentina, B Bolivia, Br Brazil, C Colombia, Ch Chile, E Ecuador, G Guatemala, H Honduras, M Mexico, Pr Paraguay, P Peru, USA United States of America, U Uruguay, and V Venezuela b)Genomes and subgenomes (SGs) with chromosome-scale assemblies are shown with asterisks Comparison of genome structures Syntenic regions among S. demissum, S. verrucosum, and S. acaule were identified using SyRI v1.6.3 [33] as previously described [48]. Centromeric regions in S. verrucosum were previously determined using CENH3 ChIP-seq signals [46]. The results were visualized by Plotsr 0.5.4 [32]. To further investigate genome structure similarity among chromosome-scale assemblies of Solanum species (shown by asterisks in Table 1), pairwise ratios of syntenic regions were calculated by dividing the length of syntenic areas by the total chromosome length of each species. Based on the ratios, a dendrogram was constructed using the unweighted pair group method with arithmetic mean (UPGMA) method, and 100 bootstrap resamplings were performed using DendroUPGMA (http://genomes.urv.cat/UPGMA/). Divergence analysis To compare the sequence divergence among S. demissum, S. acaule, and S. verrucosum, coding DNA sequences (CDSs) in each single-copy ortholog group, detected using OrthoFinder, were translated with EMBOSS transeq v6.6.0.0 [9], aligned at the protein level using MAFFT with the “–localpair –maxiterate 1000” option, and back-translated to codon-preserving nucleotide alignments using PAL2NAL v14 [128] with the “-output paml -nogap” option. Pairwise synonymous substitution (dS) rates were estimated using codeml v4.10.7 [139], with a control file configured as follows: seqtype = 1 (codons), CodonFreq = 2 (F3 × 4), model = 2 (two or more branch ω ratios), NSsites = 0 (no among-site variation), cleandata = 1 (gap/ambiguous sites removed), and runmode = -2 (pairwise). To visualize chromosome-level patterns of synonymous divergence, dS rates were plotted along each chromosome using a sliding-window approach. For each chromosome, genes were ordered by genomic position, and the mean dS rate was calculated for consecutive windows of 10 genes. Detection of R-gene candidates Resistance (R) genes are often composed of specific domains: Drosophila Toll and mammalian interleukin-1 receptor (TIR), coiled-coil (CC), nucleotide-binding adaptor shared by the human apoptotic protease-activating factor 1 (NB-ARC), and leucine-rich repeat (LRR) domains [107]. Nucleotide-binding and leucine-rich repeat (NLR) genes were identified in S. demissum and other Solanum species using NLR-annotator v2.1 [126] with default parameters. The chromosomal distribution of NLR genes in S. demissum was visualized using R 4.3.3 (https://www.r-project.org/). Results Genome assembly The genome of the S. demissum clone 24H65-12 was assembled using PacBio HiFi reads. Sequencing on the PacBio Revio generated 11,553,623 HiFi reads, yielding 211.8 Gbp of sequence data, with a mean read length of 18,334 bp and an N50 length of 18,893 bp. The initial contig assembly yielded a total length of approximately 2.517 Gb with a contig N50 of 34.07 Mb (Table 2). The 944,446,258 Hi-C read pairs (95.4 Gbp) were mapped to the contigs, processed, and scaffolded (Fig. 1a). The final nuclear genome assembly spanned 2.314 Gb, with 2.035 Gb anchored to 36 chromosome-level scaffolds. The chromosome-level scaffolding significantly improved the contiguity, achieving a scaffold N50 of 56.05 Mb (Table 2). The 36 chromosome scaffolds were phased into three subgenomes (SG1, SG2, and SG3). The k-mer classification heatmap confirmed the successful differentiation of the three subgenomes (Fig. 1b). The resulting subgenomes (12 chromosomes each) showed slight size asymmetry. SG1 had the largest total length, 696.51 Mb, followed by SG2, 685.32 Mb, and SG3, 653.29 Mb (Table S1). Table 2Assembly statistics of the S. demissum genomeCategoryNumber of sequencesTotal length (bp)Average length (bp)Maximum length (bp)N50 (bp)Primary contig7,4562,516,682,040337,537.878,494,62334,073,135Final assembly Chromosome362,035,124,51856,531,236.686,927,26356,046,848 Unanchored4,120278,497,68167,596.52,447,82873,347 Organella3,250203,065,84162,481.8511,46562,834 Chloroplast3,060193,913,83563,370.5511,46564,396 Mitochondria1909,152,00648,168.5117,25648,538 Assembly statistics of the S. demissum genome Genome assembly and subgenome phasing of S. demissum. a Hi-C contact map of S. demissum chromosomes. Color intensity indicates contact frequency. b Subgenome-specific k-mer distribution, grouping to three sets of 12 chromosomes Gene annotation A total of 662.85 M read pairs were obtained for expression evidence, with individual samples ranging from 50.41 M to 72.67 M read pairs. The evidence-based annotation derived from RNA-seq datasets was merged with the deep learning prediction to produce a unified annotation dataset. The final gene features were standardized using AGAT 1.3.3, and a total of 97,915 protein-coding genes were identified in the final assembly, with an average protein length of 385.9 amino acids (Table S2). A slight asymmetry in gene counts was observed across the subgenomes: SG2 had the highest count (33,028), followed by SG1 (32,888), and SG3 had the lowest (31,999) (Table S2). Completeness of the assembly The completeness of the assembly was assessed using BUSCO scores for the genome sequences against the Solanales odb10 database (Table S3). The BUSCO score for overall assembly was 99.19%, indicating high completeness. This completeness is primarily driven by the duplicated fraction of genes (97.8%), reflecting the expected retention of core orthologs across the three subgenomes in the hexaploid species. The individual subgenomes (SG1, SG2, and SG3) each showed high levels of complete and single-copy BUSCOs (~ 95%) and low levels of complete and duplicated BUSCOs (~ 2.5%), confirming the successful separation of gene complements into the three near-complete subgenomes. The annotated protein set also showed high completeness (98.9%) for the overall assembly, validating the quality of the gene predictions (Table S3). Transposable element density, sequencing coverage, and heterozygosity Transposable elements (TEs) in S. demissum were annotated, and genome-wide TE densities were summarized for each subgenome (Table S4). A total of 68.5%, 68.1%, and 67.0% of the SG1, SG2, and SG3, respectively, were occupied by TEs. Gypsy elements were the most abundant TE family across all three S. demissum subgenomes (26.0–28.3%). The relative contributions of other major TE families, including Copia and DNA transposons, were also highly similar among subgenomes, and no pronounced subgenome-specific enrichment or depletion of individual TE families was detected. Chromosome-scale visualization of gene density, TE density, sequencing coverage, and heterozygosity revealed conserved structural features among the three subgenomes (Fig. S1). As shown for chromosome 3 as a representative example (Fig. 2), genes were preferentially distributed along chromosome arms, whereas LTR retrotransposons such as Gypsy and Copia (shown in blue) were enriched in gene-poor, pericentromeric regions. In contrast, DNA transposons (shown in purple) were more frequent in gene-rich subtelomeric regions. Distributions of genes, various transposable elements (TEs), HiFi-read coverages, and heterozygous portions (heterozygous SNPs/kb) on the S. demissum chromosome 3 across the three different subgenomes (SGs) Genome-wide coverage of PacBio HiFi reads was largely uniform across chromosomes and subgenomes, indicating that the assemblies do not suffer from major large-scale biases or structural artifacts. Solanum demissum is highly self-fertile, and accordingly, genome-wide heterozygosity was generally very low, with heterozygous alleles rarely detected across most regions of the genome (Fig. S1). However, notable exceptions were observed. Specifically, striking increases in heterozygosity were detected in the distal regions of chromosomes 3 and 4 across all three subgenomes (Fig. 2, Fig. S1). Subgenome-specific increases were also detected in chromosome 1 of SG1, chromosome 5 of SG2, chromosome 7 of SG1 and SG2, and chromosome 9 of SG2 (Fig. S1). Subgenome expression bias To evaluate whether the three subgenomes of hexaploid S. demissum exhibit transcriptional bias, we first compared expression levels of genes encoded on each subgenome, along with the 15,737 single-copy orthologous gene pairs conserved among the three subgenomes, referred to as homoeologs hereafter, across four tissues (leaf, bud, flower, and stolon) (Table S5). Overall, the distributions of the number of transcripts per million (TPM) values were broadly similar among SG1, SG2, and SG3 for “All genes,” with no subgenome showing obviously higher or lower expression (Fig. 3a). Thus, the three subgenomes are transcriptionally active. For the homoeologs, TPM values were slightly higher than those of all genes across all tissues, consistent with these genes representing a conserved, relatively well-expressed gene set. Even within this subset, differences among SG1, SG2, and SG3 were minimal, suggesting the absence of any strong genome-wide dominance or suppression at the subgenome level. Expression balances among S. demissum subgenomes. a Boxplots indicate the means and ranges of transcripts per million (TPM) values of three biological replicates for all chromosomal genes (All) and single-copy orthologous gene pairs conserved among the three subgenomes (Homoeologs) in leaves, floral buds, flowers, and stolons, separately plotted for the subgenomes (SGs). b A dendrogram based on TPM values of homoeologous genes with the Ward.D2 method. c Expression balances of homoeologous genes. d The number of homoeologous genes in each category We next focused on the 15,737 single-copy orthologs shared among all three subgenomes to examine the expression balance of homoeologous triplets in greater detail (Fig. 3b–d). Hierarchical clustering, based on TPM-normalized expression profiles of these triplets, showed highly reproducible grouping of biological replicates (Fig. 3b). Within each tissue cluster, SG1 and SG2 consistently grouped more closely together, whereas SG3 tended to form a slightly more distant branch. Although this separation was subtle, it was stable across tissues. Ternary plots of relative expression among SG1, SG2, and SG3 (Fig. 3c) revealed that most homoeologous triplets were located near the center of the triangle in all tissues, indicating that the three homoeologs generally contributed similar proportions to the total expression of each triplet. In these plots, points near a vertex represent triplets in which one subgenome contributed disproportionately to total expression, whereas points near an edge indicate reduced contribution from one subgenome. Thus, the central enrichment observed across tissues indicates that strongly biased expression was uncommon and that expression was broadly balanced among the three subgenomes. Quantitative classification corroborated this pattern: across tissues, 8,600–9,800 triplets fell into the “Balanced” category, representing the largest category across all tissues (Table S6; Fig. 3d). Here, “Balanced” refers to triplets that did not meet the criteria for subgenome-dominant or subgenome-suppressed expression, indicating the absence of strong expression bias toward or against any one subgenome. Subgenome-dominant or -suppressed classes (SG1-, SG2-, or SG3-dominant/suppressed) each contained fewer than 1,300 triplets. The number of SG3-dominant triplets (297–366) was larger than that of SG1 and SG2 in all tissues (126–185), while the number of SG3-suppressed triplets (1019–1289) was also larger than that of SG1 and SG2 in all tissues (806–1050). Therefore, these comparisons indicate that SG3 has slightly different features in expression profiles than SG1 and SG2. Phylogenetic placement of the S. demissum subgenomes To explore the species relationships of S. demissum, we conducted a phylogenetic analysis based on gene sequence similarity using 50 genomes and subgenomes with sufficient sequence coverage (Table 1). OrthoFinder classified a total of 1,738,035 genes into 38,509 orthogroups, of which 4,038 were present in all the genomes. Among these common orthogroups, 956 were single-copy orthogroups (Table S7). A maximum-likelihood phylogeny using IQ-TREE 2 was constructed from these 956 single-copy orthogroups. The resulting tree clearly identified three clades among tuber-bearing Solanum species (Fig. 4a). The first group comprised Mexican diploid species with the B genome, corresponding to Clade 1+2. The second group comprised S. piurae and some other South American species, corresponding to Clade 3. The third group comprised South American A-genome species, including cultivated species and corresponding to Clade 4. Solanum stipuloideum Rusby (formerly, S. circaeifolium Bitter; [118] clustered as a sister to Clade 4 and to a group containing S. sogarandinum Ochoa and S. huancabambense Ochoa, supporting the findings by Zhang et al. [148] and the recognition of the’neocardenasii’ clade, which includes S. stipuloideum and S. neocardenasii Hawkes & Hjert. [125]. Bootstrap support was high for most internal nodes, indicating that the inferred topology is robust. Species relationships based on gene sequence similarity (a) and genome structure (b). a Phylogenetic tree generated using IQ-TREE 2, which was constructed using single-copy gene sequence similarity in 4,038 orthogroups commonly present in all 50 genomes. A bootstrap value is given at each node. b Dendrogram was constructed using the UPGMA method, based on pairwise ratios of syntenic regions among 23 chromosome-scale assemblies, showing genome structure similarity In the inferred phylogeny, the three S. demissum subgenomes were placed distinctly but were closely clustered with other species or genomes of interest. SG1 was closely clustered with the Mexican A-genome diploid species S. verrucosum within Clade 4. The S. verrucosum + S. demissum SG1 clade was clustered with the A genome of S. stoloniferum (depicted simply as S. stoloniferum A in Fig. 4). SG2 was also placed within Clade 4, but in a distinct clade from SG1 sister to S. acaule SG2 (Fig. 4a). This S. acaule SG2 + S. demissum SG2 clade was clustered with S. boliviense (= S. megistacrolobum). In contrast, SG3 was most closely clustered with SG1 of S. acaule as a lineage sister to Clade 3, forming a distinct branch that did not nest within Clade 3. Thus, S. demissum likely comprises genomes from S. acaule and S. verrucosum. Genomic divergence among S. demissum, S. acaule, and S. verrucosum To investigate genome structural relationships among S. demissum, S. verrucosum, and S. acaule, we performed a whole-chromosome synteny analysis using SyRI v1.6.3 (Fig. 5a). Large syntenic blocks were broadly conserved between S. verrucosum and S. demissum SG1, including partial conservation of centromeric regions. Although pericentromeric inversions were observed on chromosomes 3 and 8, the underlying sequences were largely retained. Although there are large pericentromeric inversions in chromosomes 1, 5, and 12 between S. demissum SG2 and S. acaule SG2, as well as in chromosomes 1, 4, and 10 between S. demissum SG3 and S. acaule SG1, extensively conserved synteny was observed. In contrast, comparisons between S. demissum SG1 and S. acaule SG2, and between S. demissum SG2 and S. acaule SG1, showed numerous duplicated and translocation connections (visualized as dense blue and green links) and, in addition, small and large inversions. Structural differences among S. acaule, S. demissum, and S. verrucosum genomes. a Structural and length variation generated by SyRI and visualized by Plotsr. b dS rates plotted along chromosome 2 among the S. demissum subgenomes (SGs, upper) and between each S. demissum subgenome and the subgenomes of S. acaule or S. verrucosum (lower). dS values were calculated from single-copy orthologs using a sliding-window approach of 10 consecutive genes ordered by genomic position. Because sliding-window calculation requires a minimum of 10 genes, the plots do not start at the left chromosomal end (0 Mb); the first data point corresponds to the position of the fifth gene, and gene-poor regions near the chromosome end are not represented. Shaded regions indicate chromosomal segments with reduced dS values, consistent with post-origin recombination or homogenization between homoeologous chromosome 2 of SG1 and SG2 To further assess sequence-level divergence, we estimated pairwise synonymous substitution (dS) rates among S. demissum, S. verrucosum, and S. acaule using single-copy orthologs and visualized chromosome-scale patterns (Fig. S2; Fig. 5b). Across most chromosomes, dS distributions were consistent with the phylogenetic relationships inferred by OrthoFinder. For example, chromosome-wide dS rates among S. demissum subgenomes were generally higher between SG3 and SG1 or SG2 than between SG1 and SG2, indicating greater sequence divergence of SG3 relative to SG1 and SG2. In addition, pairwise comparisons between S. demissum SG1 and S. verrucosum or S. acaule genomes revealed clear differences in divergence levels: dS rates between SG1 and S. verrucosum were consistently low, whereas comparisons with S. acaule SG2 showed moderately higher dS rates, and those with S. acaule SG1 were higher. Likewise, dS rates between SG2 and S. acaule SG2, and between SG3 and S. acaule SG1, were consistently lower than those for the other comparisons. However, several notable local deviations were detected. Solanum demissum SG1 exhibited exceptionally low dS rates against S. acaule SG2 and high dS rates against S. verrucosum in the right terminal region of chromosome 2, whereas S. demissum SG2 exhibited the opposite pattern (Fig. 5b). This is likely caused by recombination between homoeologous chromosome 2 of SG1 and SG2, which occurred after S. demissum originated. On the right terminal region of chromosome 3, S. demissum SG1 against SG2 and SG2 against SG1 exhibited exceptionally low dS rates, indicating an unusually high sequence similarity between these regions (Fig. S2). In the same genomic interval, dS rates between SG1 and S. acaule SG2 were also markedly reduced. In addition, partial regions on chromosomes 1, 4, 6, 9, 10, and 11 of all three subgenomes displayed markedly elevated dS rates when compared against S. verrucosum (Fig. S2). These regions likely represent segments where the expected S. verrucosum-derived signal has been disrupted, indicating accelerated sequence divergence in S. verrucosum after S. demissum originated. We note that local gene-model disruption or partial CDS-prediction errors may also contribute to inflated dS estimates in some regions. Comparison of the genome structures of tuber-bearing Solanum species Genome structure similarity was investigated using pairwise ratios of syntenic regions obtained by SyRI. Although 30 chromosome-scale assemblies, out of 50 genomes used to evaluate gene sequence similarity, were loaded into SyRI, only 23 could be analyzed due to overload or other technical issues (Table S8). An average syntenic region with S. etuberosum was 33%, while those of the others ranged from 40s to 60s% (Table S8). The syntenic region between S. tuberosum RH and S. phureja DM8.1 was 72%. In contrast, syntenic regions between S. demissum SG1 and S. verrucisum (85%), between S. demissum SG2 and S. acaule SG2 (73%), and between S. demissum SG3 and S. acaule SG1 (76%) were the highest among all the others. Using the pairwise ratios, a dendrogram was constructed by the UPGMA method (Fig. 4b). Three large clusters were identified, corresponding to Clades 1+2, 3, and 4, similar to the results of gene sequence similarity (Fig. 4a). However, bootstrap support for higher-order topology was extremely low. The genome structures of S. demissum SG1, SG2, and SG3 were distinct from each other and most closely clustered with those of S. verrucosum, S. acaule SG2, and S. acaule SG1, respectively. The clusters of S. demissum SG2 and S. acaule SG2, and of S. demissum SG1 and S. verrucosum, were positioned within Clade 4, consistent with the gene-similarity-based phylogeny. However, the cluster comprising S. demissum SG3 and S. acaule SG1, which was most closely related to Clade 3 in the gene-similarity-based phylogeny, was grouped with Clade 1+2, although bootstrap support was extremely low. Solanum stipuloideum and S. huancabambense were distinct from the others and sister to the cluster of Clade 1+2 and Clade 3 with extremely low bootstrap support. Transposable element (TE) landscape in Solanum demissum To characterize the TE landscape of S. demissum, we compared its TE composition with those of tetraploid species, S. acaule and S. stoloniferum, and diploid species, S. bulbocastanum (B genome) and S. verrucosum (A genome). The subgenomes of polyploid species tended to show somewhat higher total TE occupancy (63.0–68.5%) than diploid species (56.4 and 59.7%) (Fig. 6a; Table S9). Dynamics of transposable elements (TEs) compared among genomes and subgenomes (SGs) of S. bulbocastanum, S. stoloniferum, S. acaule, S. demissum, and S. verrucosum. a Genome occupancy of the TEs. b Density plots of LTR identities in prominent LTR-type retrotransposon families, Gypsy, LTR, and Copia. Vertical lines indicate median values. c Kimura’s divergence values against consensus TE sequences. d Phylogenetic tree of Gypsy retrotransposons inferred by VeryFastTree from 6,894 RH-RT-INT concatenated sequences. The terminal nodes are colored based on the genomes from which their sequences are derived. Three rapidly diversifying clades were identified within the Gypsy families, Athila, Ogre, and CRM, as indicated by orange circles. e Copy numbers of TEs in each group of the rapidly diversifying clades Analyses of the LTR-type retrotransposon families, Gypsy, Copia, and unclassified LTR retrotransposons showed that, unlike the B genome of S. stoloniferum, which is characterized by an excess of very young Gypsy and LTR elements, S. demissum and S. acaule subgenomes did not exhibit pronounced enrichment of high-identity TE elements (Fig. 6b). Lower Kimura divergence values indicate young TEs, or recently inserted TEs, whereas higher values indicate old TEs (Fig. 6c). The B genome of S. stoloniferum has experienced rapid diversification or a burst of PIF Harbinger-type DNA transposons, with a sharp peak at approximately Kimura divergence value of 22. All three subgenomes of S. demissum exhibit two peaks of PIF Harbinger-type DNA transposons, though not as high as the B genome of S. stoloniferum, at approximately Kimura divergence values of 22 and 28, suggesting that there were TE bursts twice in their histories (Fig. 6c). The phylogenetic tree was inferred from 6,894 predicted protein sequences representing the RNase H (RH), reverse transcriptase (RT), and integrase (INT) domains encoded by Gypsy copies. Gypsy subfamily assignments were obtained using TEsorter, and the sequences were grouped into Athila, Ogre, Reina, CRM, Galadriel, and Tekay (Fig. 6d). Three highly diverse yet little differentiated Gypsy sequence groups with high copy numbers and short branch lengths were found in the Athila, Ogre, and CRM families (Fig. 6d). The copy numbers within these clades (Fig. 6e, Table S10) indicate that Gypsy copies from S. bulbocastanum (B genome) were rarely found in these rapidly diversifying clades. However, the B genome of S. stoloniferum showed copy numbers of CRM and Ogre families roughly comparable to those of the other genomes, but, like S. bulbocastanum, few copy numbers of the Athila family differed from the others. SG1, SG2, and SG3 of S. demissum showed approximately the same numbers as those of S. verrucosum, S. acaule SG2, and S. acaule SG1, respectively, for Gypsy copies of CRM, Ogre, and Athila families. Resistance gene repertoire and chromosomal distribution in S. demissum Several types of putative resistance (R) genes were identified among the 50 Solanum genomes and subgenomes (Fig. 7a). Total number of predicted R genes ranged from 281 (S. lycopersicum v5.0) to 1187 (S. vernei Bitter & Wittm.) (Table S11) and was significantly lower for diploid species in Clade 1+2 (mean of 459.9) than those of the species in Clade 3 and Clade 4 (means of 709.9 and 810.6, respectively) (t-test, P < 0.0001). The most common type was CC-NBARC-LRR, followed by NBARC-LRR, which were broadly similar across the 50 genomes and subgenomes. SG1, SG2, and SG3 of S. demissum contained 589, 643, and 453 R genes, respectively, which were approximately comparable to those of S. verrucosum (564), S. acaule SG2 (662), and S. acaule SG1 (462), respectively (Table S11). Thus, the three subgenomes have largely retained their inherited R-gene complements, and no subgenome-specific expansion or contraction occurred following polyploidization. Within S. demissum, R genes were unevenly distributed along chromosomes, forming distinct clusters (Fig. 7b). Resistance (R) genes. a Comparison of nucleotide-binding and leucine-rich repeat (NLR) gene copy numbers across Solanum species and subgenomes (SGs). Bars indicate the total number of predicted NLR genes in each genome or subgenome. b Chromosomal distribution of NLR genes in S. demissum. Positions of previously characterized late blight resistance genes (R1, R3b, and R8) identified in this clone are indicated as red triangles The gene sequences of the five S. demissum-derived late blight resistance genes R1, R2, R3a, R3b, and R8 were searched using blastp [12] against the S. demissum assembly. R1, R3b, and R8 were found with > 99% homology, located on chromosome 5 of SG3, chromosome 11 of SG3, and chromosome 9 of SG1, respectively (Fig. 7b). R2 and R3a were not found in the present assembly with high homology. Discussion Subgenomes of S. demissum We sequenced three subgenomes (SG1, SG2, and SG3) of the hexaploid species S. demissum using PacBio long-read sequencing and Hi-C scaffolding technologies. These subgenomes range from 653 to 697 Mb in size and contain various TEs, along with chromosomal distributions of TEs and gene density, similar to previous reports in Solanum [4, 10, 30, 47, 48, 50, 144]. Higher TE contents are featured in subgenomes of polyploid species than in diploid species [48, 88]. Despite the predominantly homozygous background, common heterozygosity peaks across all three subgenomes and subgenome-specific heterozygosity peaks were found, suggesting that these regions may be subject to evolutionary constraints that favor the maintenance of allelic diversity, such as balancing selection, rather than being explained solely by recombination between homoeologous chromosomes or technical artifacts. Genomic or species origin of S. demissum The gene sequence-based phylogenetic tree clearly identified three clades among tuber-bearing Solanum species, consistent with the well-established three-clade structure of the Petota clade [3, 53, 76, 108, 122, 131, 148]. High similarity in both gene sequences and genome structure was observed between S. demissum SG1 and S. verrucosum, between S. demissum SG2 and S. acaule SG2, and between S. demissum SG3 and S. acaule SG1 (Fig. 4). Thus, we hypothesize that S. demissum originated from a cross between S. verrucosum as the female parent and S. acaule as the male parent. Solanum verrucosum is cross-compatible with most South American species as the female parent [24] and with some Mexican diploid species [42, 57]. Solanum acaule is also known as a bridging species between A- and B-genome species [21, 41]. The endosperm balance number (EBN) hypothesis [60] states that EBN is a species-specific genetic factor, independent of ploidy, that controls endosperm development in interspecific crosses. Solanum verrucosum and S. acaule have a balanced EBN value of 2 [36]. Therefore, the cross between S. verrucosum and S. acaule could be successful, yet results in sterile triploid hybrids. To overcome sterility, unreduced gametes or 2n gametes likely contributed to polyploidization because high rates of 2n pollen-producing plants were reported in S. acaule, S. verrucosum, and S. demissum [135]. Additivity of EBN [23] yields an EBN value of 4, which matches that of S. demissum [36]. A close relationship between S. demissum and S. acaule has long been recognized. However, S. verrucosum and S. demissum are distributed in Mexico, while S. acaule is distributed from Ecuador to northwest Argentina [38, 123]. There is no geographical overlap in their current distribution ranges. For this reason, no one considered S. acaule to be involved in the origin of S. demissum, except Spooner et al. [120], who proposed three hypotheses: convergence, a common ancestor, and a progenitor-derivative relationship to explain the similarities between S. demissum and S. albicans. The common ancestor hypothesis suggests that S. acaule is the common ancestor of S. demissum and S. albicans. The potential for long-distance dispersal of S. acaule is suggested, since the distance between the southern distribution of S. demissum in Guatemala and the northern distribution of S. acaule in Ecuador is approximately 1300 km [120], which is much shorter for the Mexican diploid species S. morelliforme Bitt. & Muench, which was discovered 4000 km apart in Bolivia [115]. Also, the wide distribution of S. acaule, from Ecuador to northern Argentina, demonstrates its dispersal ability [120] or suggests a previous expansion of its distribution area further north during glacial periods because S. acaule is adapted to very high altitudes over 3700 m [38]. An A-genome species or S. verrucosum migrated from South America to Mexico [38] and generated an allotetraploid species S. stoloniferum with a Mexican B-genome species [48]. Nevertheless, the A genome of S. verrucosum was most closely clustered to S. demissum SG1 rather than to S. stoloniferum A or other A-genome species (Fig. 4), suggesting that the hybridization between S. verrucosum and S. acaule occurred recently, not on the way to the migration of S. verrucosum to Mexico. Association between meiotic chromosome pairing and genomic homology The distinctiveness among the three subgenomes of S. demissum obviously results in regular meiosis with 36 bivalents [79, 85, 130]. SG1 and SG2 of S. demissum were placed within Clade 4, whereas SG3 was distinctly clustered either with Clade 3 or Clade 1+2 (Fig. 4). This relationship among the three subgenomes of S. demissum is consistent with previously observed chromosome behaviors at meiosis. Relatively high frequencies of bivalents (4.74 to 9.78) and univalents in the triploid haploid clones of S. demissum [5, 22, 52, 56, 79] indicate that S. demissum has two somewhat similar genomes, namely SG1 and SG2, and a third, rather different one, SG3. Solanum acaule shows regular meiosis with 24 bivalents, indicating that the two genomes, S. acaule SG1 and SG2, are distinct [56, 83, 129]. This supports the idea that S. demissum SG3 (or S. acaule SG1) differs from S. demissum SG2 (or S. acaule SG2) and SG1. As S. demissum SG1 is similar to the A genome of S. verrucosum, the A genome should be assigned to S. demissum SG1 and SG2, different from the previous interpretation that S. demissum SG3 is an A genome, different from SG1 and SG2, as indicated by Matsubayashi’s formula AADDDdDd [56, 65, 79, 85, 102]. Matsubayashi [84] obtained a heptaploid hybrid (2n = 84) from S. acaule × S. demissum, likely resulting from fertilization of a 2n egg of S. acaule with n pollen of S. demissum. He reported that the mean pairing frequency of the hybrid was 0.18 V + 1.11IV + 11.73III + 18.11II + 7.26I per cell, with 12III + 20II + 8I as the modal configuration. This pairing frequency was interpreted based on his genome formula [84], but can also be understood in light of our current understanding that S. demissum SG1 and SG2 are A genomes. Therefore, chromosome pairing behaviors clearly reflect similarities in genome sequence and structure. The genome structures of S. demissum SG3 and S. acaule SG1 are more similar to those of B-genome species (Fig. 4b). However, gene sequence similarity indicates that these two genomes are sister to Clade 3 species, as previously shown for S. acaule SG1 [1, 11]. GISH analysis also indicated that signals with the B-genome probe were absent in S. acaule and S. demissum [98]. Therefore, S. demissum SG3 and S. acaule SG1 are clearly not B genomes but are closely related to the genomes in Clade 3. The species in Clade 3 are sometimes represented as P-genome species [53, 85, 98, 108, 122]. However, there have been too few chromosome-pairing observations to assign a genome symbol to Clade 3 species and to S. demissum SG3. Subgenomic differentiation after hexaploidization Three subgenomes of S. demissum have not undergone subfunctionalization and are active, with no clear expression biases, as most homoeologous genes exhibit balanced expression, consistent with the results in an allotetraploid S. stoloniferum [48] and other allopolyploid crop species such as cotton [147], Brassica [14, 140], and peanut [142]. Achakkagari et al. [1] also reported a balanced expression pattern between the subgenomes of S. acaule. However, subtle expression bias toward more dominant and suppressed expression was observed in SG3, consistent with its phylogenetic position, which is more divergent from SG1 and SG2, reflecting underlying evolutionary history rather than a pronounced expression-level bias. TE compositions and genome occupancy frequencies are also similar across the three subgenomes, but they are more similar to those of the ancestral genomes in S. acaule and S. verrucosum. Subgenome-specific TE bursts, previously observed in the B genome of the allotetraploid species S. stoloniferum [48], were not observed in any subgenomes of S. demissum. Although recent TE bursts have been detected in the Gypsy families of S. demissum subgenomes, they are inherited from the ancestral genomes in S. acaule and S. verrucosum. Thus, TE activities in response to genome shock [86] at the origin of S. demissum were likely minimal. The synteny and dS analyses demonstrated that, while the overall genome architecture of S. demissum largely preserves the structural features of its progenitor genomes, its chromosomes have undergone extensive localized remodeling after polyploid formation (Fig. 5). For example, the dS analysis uncovered recombination that occurred after S. demissum originated, between homoeologous chromosomes, resulting in the exchange of terminal segments between chromosome 2 of SG1 and SG2, and in the duplication of terminal segments in chromosome 3 of SG1 and SG2. S. demissum armed by quantity rather than quality of R genes? Consistent with previous studies [6, 61, 136], the most common type of R genes was CC-NBARC-LRR, followed by NBARC-LRR, which were broadly similar across the 50 genomes and subgenomes. These putative R genes were unevenly distributed along chromosomes, forming distinct clusters, consistent with the locations of NBARC-LRR genes reported by Jupe et al. [61], which are known as resistance gene hotspots [31]. The Mexican highlands are the center of diversity for P. infestans [93], the causal pathogen of late blight, the most devastating disease in potatoes. Thus, Mexican wild species, including S. demissum, have coevolved and survived with P. infestans to date. Mexican diploid species have fewer R genes (mean of 459.9) than Clade 3 and Clade 4 species (mean of 709.9 and 810.6, respectively). Solanum demissum has a 3.7 times greater number of R genes (1685), equivalent to the sum of R genes in the parental genomes in S. verrucosum and S. acaule. The identified late blight resistance genes in the present S. demissum clone are located on chromosomes 5 and 11 of SG3 (R1 and R3b) and chromosome 9 of SG1 (R8) in similar locations to those originally reported [54, 59, 71]. Thus, not a specific subgenome supplied late blight resistance genes, but all subgenomes likely did. It is known that any single clone of S. demissum has at least eight resistance genes to late blight and protects itself to the level of coexistence with P. infestans in natural habitats [105, 106]. Thus, many R genes from different subgenomes act in a complementary manner to protect S. demissum against P. infestans. In turn, during the past decades, strong and broad-spectrum late blight resistance genes have been identified from Mexican diploid species: RB/Rpi-blb1, Rpi-blb2, Rpi-blb3, and Rpi-abpt from S. bulbocastanum [78, 116, 132, 133], and Rpi1 and Rpi-blb3 from S. pinnatisectum [70, 112]. R1 to R11 were identified from S. demissum, but none, as single genes, were durable; their resistance was broken. In contrast, resistance genes isolated from S. bulbocastanum or S. pinnatisectum, although fewer in number, conferred broad-spectrum resistance against multiple races. A similar pattern might be expected for other pathogens, indicating that Mexican diploid species are useful sources of disease resistance in potato breeding. Conclusions We provided a high-quality de novo chromosome-scale assembly of the allohexaploid species S. demissum, which is historically the most popular wild species in potato breeding, and revealed that its genome originated from S. verrucosum and S. acaule. During the origin of S. demissum, parental genomes did not undergo significant rearrangement or reorganization and were maintained almost intact, contrasting with the origin of a Mexican alloteraploid species, S. stoloniferum [48]. Due to balanced EBN values between S. demissum and S. tuberosum, the two species can be crossed easily, though the resulting hybrids are pentaploid, and backcrossed to S. tuberosum [111]. This study will provide important insights into polyploid evolution in tuber-bearing Solanum species and into the development of a strategy for resistance breeding. Supplementary Information Supplementary Material 1: Fig. S1 Distributions of genes, transposable elements, HiFi-read coverages, heterozygous portions (heterozygous SNPs/kb), and subgenome-specific k-mers on the S. demissum chromosomes. Fig. S2 Chromosome-by-chromosome dS ratios among the S. demissum subgenomes (upper) and between each subgenome of S. demissum and the S. acaule subgenomes or S. verrucosum (lower). Supplementary Material 2: Table S1 Chromosome lengths of S. demissum. Table S2 Annotated genes. Table S3 BUSCO statistics. Table S4 Transposable elements. Table S5 Statistics of the RNA-seq analysis. Table S6 The number of homoeologous genes in each category. Table S7 Statistics of the OrthoFinder analysis. Table S8 Pairwise ratios of syntenic regions. Table S9 Comparison of genome occupancy of the TEs. Table S10 The number of copies in Gypsy sequence groups with high copy numbers and short branch lengths. Table S11 R gene count. Acknowledgements We thank Rena Sanetomo of Obihiro University of Agriculture and Veterinary Medicine for constructive comments on the manuscript and Kazumi Tanaka for assistance with plant maintenance. The Human Genome Center at the University of Tokyo provided supercomputing resources. This research was supported by potato industries and growers: Calbee, Inc., Hokkaido Potato Growers Association, KENKO Mayonnaise Co., Ltd., KOIKE-YA Inc., Yamayoshi Seika Co., Ltd., Yamazaki-Biscuit Co., Ltd., KIKUSUI-Do Co., Ltd., and Fukagawa Oil & Fat Co., Ltd. Authors’ contributions KH designed the research. KH prepared plant materials and extracted DNA for sequencing. AH performed genome sequencing, assembly, annotation, and comparative genome analysis. AH and KH wrote the manuscript. Data availability The raw sequencing reads and the assembly of S. demissum have been deposited into the National Center for Biotechnology Information under BioProject Number PRJNA1405828. All the other data supporting the reported results are included in the article or the Supplementary Information. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1: Fig. S1 Distributions of genes, transposable elements, HiFi-read coverages, heterozygous portions (heterozygous SNPs/kb), and subgenome-specific k-mers on the S. demissum chromosomes. Fig. S2 Chromosome-by-chromosome dS ratios among the S. demissum subgenomes (upper) and between each subgenome of S. demissum and the S. acaule subgenomes or S. verrucosum (lower). Supplementary Material 2: Table S1 Chromosome lengths of S. demissum. Table S2 Annotated genes. Table S3 BUSCO statistics. Table S4 Transposable elements. Table S5 Statistics of the RNA-seq analysis. Table S6 The number of homoeologous genes in each category. Table S7 Statistics of the OrthoFinder analysis. Table S8 Pairwise ratios of syntenic regions. Table S9 Comparison of genome occupancy of the TEs. Table S10 The number of copies in Gypsy sequence groups with high copy numbers and short branch lengths. Table S11 R gene count. Data Availability Statement The raw sequencing reads and the assembly of S. demissum have been deposited into the National Center for Biotechnology Information under BioProject Number PRJNA1405828. All the other data supporting the reported results are included in the article or the Supplementary Information.