Two telomere-to-telomere gapless genomes reveal insights into Capsicum evolution and capsaicinoid biosynthesis

Two telomere-to-telomere gapless genomes reveal insights into Capsicum evolution and capsaicinoid biosynthesis | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Two telomere-to-telomere gapless genomes reveal insights into Capsicum evolution and capsaicinoid biosynthesis Li Guo, Weikai Chen, Xiangfeng Wang, Jie Sun, Xinrui Wang, Zhangsheng Zhu, and 18 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-3062426/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 20 May, 2024 Read the published version in Nature Communications → Version 1 posted You are reading this latest preprint version Abstract Chili pepper ( Capsicum ), one of oldest domesticated crops in human history, is best known for its unique fruit pungency given by capsaicinoids. The evolutionary history of capsaicinoid biosynthesis and the mechanism of tissue-specificity remain obscured due to the lack of high-quality Capsicum genomes. Here, we assembled two telomere-to-telomere (T2T) gap-free genomes for C. annuum and its wild non-pungent relative C. rhomboideum to dissect the evolution of fruit pungency in chili peppers. We precisely delineated Capsicum centromeres which, unlike Arabidopsis and human centromeres, lacked high-copy centromeric tandem repeats but were extensively invaded by CRM LTR retrotransposons. Phylogenomics estimated the evolutionary timings of capsaicinoid biosynthesis, and revealed eroded coding and regulatory regions of key biosynthesis genes in non-pungent species such as C. rhomboideum . We also found conserved placenta-specific accessible chromatin regions likely allowing tissue-specific biosynthesis gene co-regulation. Finally, from detected copy number variants of capsaicinoid biosynthesis genes using the C. annuum T2T genome as reference, we trained a machine-learning model to predict pungency for genomic selection. The new T2T genomic resources will accelerate chili pepper breeding, metabolic engineering and enrich our understanding of Capsicum biology and genome evolution. Biological sciences/Plant sciences/Plant evolution Biological sciences/Genetics/Genomics/Comparative genomics Biological sciences/Plant sciences/Secondary metabolism Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Introduction Chili pepper ( Capsicum annuum ), a member of the nightshade family (Solanaceae), is a worldwide cultivated vegetable and spice crop, bearing fruits best known for their fruit pungency given by capsaicinoids. Capsaicinoids are alkaloids synthesized via the convergence of the phenylpropanoid pathway and branched-chain fatty acid pathway, finished with condensation by capsaicin synthase (CS) 1 , although the full biosynthesis pathway is yet elucidated.The birth and death of capsaicinoid biosynthesis in the nightshade family remains a poorly understood question for plant evolutionary biologists. The biosynthesis of capsaicinoids occurs in the fruit placenta tissue of chili pepper, a unique trait of Capsicum spp. and missing in other Solanaceae plants such as tomato 2 . However, fruit pungency is not universally present across Capsicum , as nonpungency is found in several cultivars and wild relatives (e.g. C. rhomboideum ) of chili peppers 3 . How fruit pungency arose and lost in the Capsicum spp. and how capsaicinoids are exclusively produced in fruits remains poorly understood, due to the lack of high-quality genome resources across the Capsicum genus, especially non-pungent members. The first draft genome assembly of C. annuum (cultivar CM334) was released in 2014 sized 3.06 Gb with contig N50 of 30 kb 2 . Afterward, several assemblies showed improvement with the development of single-molecule DNA sequencing technologies, for example, cultivar ‘59’ (3.07 Gb, contig N50: 41.27 Mb) 4 , ‘Takanotsume’ (3.05 Gb, contig N50: 99.05 Mb) 5 and CC-090 (3.06 Gb, contig N50: 187.09 Mb) 6 . To date, 23 genome assemblies of cultivated peppers including C. annuum, C. baccatum , C. chinense ,and C. pubescens are publicly available 7-11 , whereas genome sequences for wild peppers are very scarce 7 . Also, despite continuous improvement, the published assemblies still contain numerous gaps and assembly errors, missing complete centromeres and telomeres. Pepper genomes are repeat-rich (~80%), making genome assembly extremely challenging. Assembly gaps and errors often lead to mis-annotation of genes and false discovery of genetic variants, confounding the pepper functional genomic research. Therefore, generating telomere-to-telomere (T2T) gapless and accurately annotated genome sequences is paramount to precision genetic improvement of peppers and dissecting full biosynthetic pathways for capsaicinoids and other valuable natural products. Accurate assembly of complete genome sequences remains a daunting task for eukaryotes with large and complex genomes with extensive repeats, high heterozygosity, or polyploidy. A breakthrough was recently achieved by the human T2T consortium in assembling the complete human genome sequence, a milestone in human genomics 12 . It has revolutionized the analysis of human genomic variants, and epigenetic and transcriptional signatures in centromeres 13-15 . Nearly two decades after the first available plant genome sequence, T2T gap-free genome assemblies were recently reported for Arabidopsis 16-18 , rice 19-20 , potato 21 and soybean 22 . However, these are considered near-complete genomes with either minor gaps in difficult-to-assemble regions or the omission of a few telomeres or centromeres typically containing high copies of tandem repeats. Furthermore, these published plant T2T genomes are relatively small in size (Arabidopsis 134 Mb 16 , rice 385 Mb 20 , potato 773 Mb 21 , and soybean 1.01 Gb 22 ). Until recently, a complete genome assembly was reported for maize (2.10 Gb) 23 . However, complete genomes remain rare for large complex plant genomes which are notoriously more difficult to assemble. Here, we de novo assembled and annotated two T2T gapless Capsicum genome sequences including a pungent pepper C. annuum and its non-pungent wild relative C. rhomboideum , which marks a new milepost for pepper genome research. The in-depth analysis of two T2T genomes revealed distinctive structural, epigenetic, and transcriptional features in their centromeres. Evolutionary insight into capsaicin biosynthesis pathway and regulation were provided using phylogenomics and epigenomic data analysis. Our study provides timely genomic resources to accelerate pepper research and genetic improvement. Results T2T gapless Capsicum genome assemblies Toassemble T2T gap-free genomes for pungent C. annuum and nonpungent C. rhomboideum , we generated high-coverage PacBio HiFi reads, Oxford Nanopore Technology (ONT) ultra-long reads, Illumina paired-end (NGS) reads and high-throughput chromatin conformation capture (Hi-C) sequencing reads for C . annuum double haploid cultivar G1-36576 and C. rhomboideum wild accession PI 645680 (Supplementary Fig. 1 and Supplementary Table 1). The genome assembly was performed using an in-house pipeline integrating various computational tools to maximize the strength of various data types (Supplementary Fig. 2 and Supplementary Table 2). Briefly, HiFi and ONT reads were first separately assembled using hifiasm 24 and NextDenovo 25 , respectively. The C. annuum HiFi-based assembly sized 3.13 Gb with contig N50 of 262.4 Mb, and contained 18 telomeres with thousands of copies of telomeric repeat units (TRUs) identified at one or both ends of 12 contigs, of which six were T2T. Assembling ONT reads produced a 3.10 Gb draft assembly with contig N50 of 177.8 Mb , containing 22 telomeres with over 10,000 copies of TRUs, of which four were nearly T2T. ONT assembly was then used to fill the gaps (Supplementary Table 3) and patch the telomeres of the HiFi assembly, yielding a hybrid assembly including 12 gapless chromosome-level contigs with 22 telomeres plus contigs containing 45S rDNA arrays. The nucleolus organizer regions (NOR) were separately assembled using 45S rDNA-containing HiFi reads, and the contigs were assembled to a single sequence based on the specific k -mer (Supplementary Fig. 3). To ensure accuracy, the sequences of ONT origin were replaced with their corresponding HiFi assembled contigs, followed by Hi-C scaffolding to 12 chromosomes ( Fig. 1A ) and manual correction for mis-assemblies in Juicebox 26 . After addition of the rDNA arrays and telomere patching, the final T2T gapless assembly of C. annuum genome (CaT2T) was 3.1Gb with a contig N50 of 262.6Mb ( Table 1 ), closing all 503 gaps of Ca59 ( Fig. 1B) and representing the first complete Capsicum genome assembly ( Fig. 1C ) and largest complete genome sequence. Using the same approach, we assembled a 1.70 Gb T2T gap-free genome sequence (CrT2T) for C. rhomboideum containing 13 chromosomes with a contig N50 of 146.0 Mb ( Fig. 1A ; Table 1 ), the first gap-free genome for non-domesticated Capsicum ( Fig. 1C ). We identified the complete set (24/24) of telomeres in C. annuum (Supplementary Fig. 4) and majority of (17/26) telomeres in C. rhomboideum (Supplementary Fig. 5). Synteny analysis revealed that 45.07% of C. rhomboideum genes are syntenic to C. annuum (Supplementary Table 4), but their whole-genome alignment showed low sequence identity, suggesting substantial divergence. Comparing the two genomes, we demonstrated that it required at least 10 fissions and 11 fusions of chromosomes to evolve the C. annuum karyotype from that of C. rhomboideum (Supplementary Fig. 6 ; Fig.1D ). Genome validation and annotation We performed extensive validations of the two T2T genome assemblies. Firstly, we examined their Hi-C chromatin interaction maps showing no obvious misplacement of contigs within the CaT2T and CrT2T assemblies (Supplementary Fig. 7AB). Then we mapped all HiFi, ONT, and NGS reads separately against the assemblies, yielding a mapping rate of over 99.96% for all three data types (Supplementary Table 5). Mapped HiFi or ONT reads showed uniform coverage across all whole genome with a few exceptions in CrT2T due to the presence of high-copy satellite repeats (Supplementary Fig. 7CD). The CaT2T and CrT2T had a QV (quality value) of 56.60 and 77.18, and a BUSCO score of 98.62% and 97.12%, respectively, demonstrating high accuracy and completeness of both assemblies ( Table 1 ). Furthermore, aligning a recently published genome assembly of C. annuum cultivar ‘59’ (hereafter Ca59) 4 against CaT2T showed strong collinearity between the two assemblies (Supplementary Fig. 2D). The high-quality assembly of CaT2T was well supported by a high-coverage of HiFi and ONT read mapping spanning these gap regions (Supplementary Fig. 8; Supplementary Table 6). Interestingly, we observed sporadic high coverage read mapping against CaT2T and CrT2T corresponding to intact mitochondrial or chloroplast genomes, which were validated by high coverage of ONT ultra-long read mapping spanning the entire integration sites (Supplementary Fig. 9), suggesting recent plastid genome integration in the nuclear genome. Repeat annotation showed that 79.5%(2.45 Gb) and 74.6% (1.28 Gb) of C. annuum and C. rhomboideum genomes were repetitive sequences, primarily composed of transposable elements (TEs), especially long terminal repeat (LTR) retrotransposons (Supplementary Table 7). While LTR insertion in C. rhomboideum occurred relatively recently, C. annuum had two bursts of insertion dated around 0.1 Mya (million years ago) and 3.9 Mya (Supplementary Fig. 10), consistent with previous report in Ca59 assembly 4 . Both pepper genomes had low contents of satellite repeats (< 0.01%), much fewer than human (4.5%), the model plant Arabidopsis (0.37%), and its Solanaceae relative tobacco (1.75%). A total of 34,428 and 33,512 protein-coding genes were predicted for CaT2T and CrT2T by using a combination of ab initio prediction, homologous proteins, and transcriptomic data. Filled gaps by CaT2T encoded 614 genes, of which 110 were newly annotated (nonsyntenic to Ca59) (Supplementary Fig. 11). Both Capsicum genomes were gene-dense towards the end of chromosomes but gene-sparse towards centromeres. CaT2T encoded 117 putative capsaicin biosynthesis genes (CBGs), including previously reported genes such as capsaicin synthase ( CS ), aminomethytransferase (AMT), ketoacyl-ACP synthase (Kas) and acyl carrier protein (ACL) 27 (Supplementary Table 8). These putative CBGs were expressed in at least one tissue, and 26 of them showed much higher expression levels in the fruit placenta compared to other tissues (Supplementary Fig. 12), therefore serving as strong candidates for full elucidation of capsaicin biosynthetic enzymes. Capsicum centromeres are extensively invaded by CRM retrotransposons Essential to faithful chromosomal segregations during cell division, centromeres are typically heterochromatic regions containing megabase-arrays of tandem repeats where the kinetochore protein complex binds 28 . We first identified centromeres of CaT2T by generating CENH3 ChIP-seq data of C. annuum , which clearly delineated the location and boundary of 12 centromeres in CaT2T ( Fig. 2A ). Then, we observed that the inter-chromosome interactions were always positively correlated with the ChIP-seq peak (Supplementary Fig. 13A), especially in CrT2T (Supplementary Fig. 13B). Unlike Arabidopsis 16 and human 13 centromeres, Capsicum centromeres lacked high-copy tandem satellite repeats and higher order repeats (HORs), suggestive of most newly-formed centromeres 29 . Rather, we found Capsicum centromeres were extensively invaded by Gypsy- LTRs taking up ~71% of total centromeric sequences ( Fig. 2B ), with only 47.3%~49.7% for whole genomes (Supplementary Table 7). This pattern was also reported in einkorn 30 and cotton 31 , where more than 80% of their functional centromeres are Gypsy- LTRs. We also found that Capsicum LTRs had a later burst of insertion in centromeres than in the whole genome, indicating recent evolution of centromeres shaped by LTR insertion ( Fig. 2B ). LTRs are typically subjected to reshuffle and rearrangement due to unequal homologous recombination events generating fragmented or solo LTRs 32 . A number of solo LTRs and intact LTRs were identified in two Capsicum genomes (Supplementary Table 9). We observed that Capsicum centromeres had a weaker ability than non-centromeres to remove LTRs, indicated by their significantly lower solo-to-intact LTR ratios compared to the whole genome ( Fig. 2C ). The synteny analysis showed that the LTR retrotransposon CRMs (centromeric retrotransposon of maize) were enriched in centromeres of several Capsicum genomes (Supplementary Fig. 13CD), suggesting the distribution of CRMs was a marker that could be used to identify Capsicum centromeres without relying on ChIP-seq data . CRMs possess chromodomain or CR motifs that potentially interact with the centromeric histones and play important roles in centromere evolution and function 33 . Furthermore , phylogenetic analysis of Gypsy -LTR retrotransposons in two Capsicum and potato genomes identified six subfamilies with two major ones: Athila and Tekay . Athila LTRs are the major LTRs in Arabidopsis centromeres 16 , unlike Capsicum centromeres which are enriched with CRM Gypsy- LTRs ( Fig. 2D; Supplementary Table 9; Supplementary Table 10). The lack of satellite repeats and enrichment of CRM LTRs ( Fig. 2E ) distinguished Capsicum centromeres from centromeres reported in other published plant T2T genomes 16,20-22 . Strikingly, CrT2T had a higher content of CRMs than CaT2T, contributing to a distinctive signature of sequence identity within their centromeres (Supplementary Fig. 14) . For both T2T genomes, we observed low inter-species and inter-chromosomal centromere sequence identity, suggesting a rapid divergence of Capsicum centromeres within and between species, consistent with the findings in a recent study of Arabidopsis centromeres 34 . Centromeres and telomeres are transcriptionally and epigenetically active Centromeres and telomeres are overall poorly understood genomic regions in terms of protein-coding genes, transcriptional and epigenetic control. Genome annotation predicted 60 genes in CaT2T centromeres, enriched in functions such as response to freezing, DNA topological change and meiotic chromosome separation (Supplementary Fig. 15). By contrast, CrT2T centromeres encoded 94 genes enriched in response to UV-B, photosynthesis and regulation of circadian rhythm. Interestingly, only six or seven centromere genes were homologous (Supplementary Table 11). The low homology of centromere genes reflected a high divergence of centromeres between the species, consistent with their poor whole-genome sequence alignment. We found that CaT2T centromeres ( Fig. 3A; Supplementary Fig. 16) and telomeres ( Fig. 3B ) had active transcriptions as suggested by RNA-seq analysis, both from transposons and protein-coding genes. For example, about 42 (70.0%) centromere-encoded genes were expressed in at least one tissue (TPM > 1) in C. annuum , including CaT2T07g00954 encoding a telomere maintenance protein that protects the ends of telomeres from attack and CaT2T01g02835 encoding a flowering time control protein. The expression of CrT2T centromere-encoded genes (92.5%) was more active, containing three tandem copies of RCC1 (Regulator of chromosome condensation) with an average TPM > 200, potentially playing key roles in the regulation of chromatin condensation in mitosis. Capsicum genomes contain rich epigenomic signatures such as DNA methylation, histone modifications, TADs (topological associated domains) and A/B compartments 4 . However, little is known about these epigenetic marks in centromeres and telomeres. Therefore, we mapped our own generated (Hi-C and bisulfite sequencing) and public epigenomic (histone ChIP-seq) data to CaT2T assembly. Hi-C data analysis detected A/B compartment, TADs, and small chromatin loops within centromeres ( Fig. 3A ). C. annuum centromeres primarily belonged to "B" compartment typically associated with low transcription, except that "A" compartments were detected in a few chromosomes such as Chr08, Chr09 and Chr10 (Supplementary Fig. 16). Consistently, high TE density, low gene density ( Fig. 3C ), and low histone H3K9me2 ChIP-seq peaks were detected for heterochromatic centromeres and pericentromeres with enriched CENH3 ChIP-seq signals ( Fig. 3D ). Interestingly, although DNA methylation levels in centromere and non-centromere regions were overall comparable ( Fig. 3C ), we found centromere-located CRMs showed lower gene-body CHG methylation compared to gene flanking regions ( Fig. 3E ; Supplementary Fig. 17), suggesting transcriptional activity of CRMs. Despite heterochromatin nature, transcription in centromeres was not silent as we identified some highly expressed genes and TEs, such as one on Chr07 encoding peroxiredoxin Q protein involved in cell redox homeostasis ( Fig. 3A ). Compared to centromeres, most telomeres had low Hi-C mapping signals (Fig. 3B ) perhaps due to the high density of tandem telomeric repeats. Subtelomere regions were relatively gene-rich mostly associated with A compartments and marked with low H3K9me3 ChIP-seq signals except in gene-sparse regions ( Fig. 3B ). These results provided unprecedented insights into the organization and functions of pepper complex genomic regions. Revisiting evolutionary history of capsaicinoid biosynthesis pathway in Capsicum The birth and death of capsaicinoid biosynthesis in plants remains poorly understood. Kim et al. first approached this question by comparing a fragmented pepper genome with tomato genome, revealing the mechanisms of pungency 2 . However, a better understanding of how the pathway emerged and evolved required an investigation in a broader phylogenetic context. Therefore, we revisited this question by preforming phylogenomics using two T2T Capsicum genomes and 14 other angiosperm genomes (Supplementary Table 12) including three pungent and 13 nonpungent species. We found that Capsicum was more related to Physalis (ground cherry) than to Solanum ( e.g. tomato), diverging from the two taxa at ~17 Mya and ~19 Mya, respectively ( Fig. 4A ). The fact that capsaicinoid biosynthesis is limited to Capsicum indicated that the specialized metabolite must have arisen in Capsicum after their divergence from Physalis at ~17 Mya. Furthermore, C. baccatum diverged from C. annuum and C. chinense at ~ 5Mya, which together diverged from the non-pungent C. rhomboideum at ~13.4 Mya, suggesting that the capsaicinoid pathway could have formed between 13.4 Mya and 5 Mya ( Fig. 4A ). To understand how the pathway arose in Capsicum plants, we identified genes from the 16 angiosperms with homologs to known CBGs using OrthoFinder 35 (Supplementary Table 13). All species, pungent or not, contained homologs of CBGs, except that CS , the most critical gene 1 , showed up as tandem duplicates only in Solanaceae at around 71 Mya ( Fig. 4B; Supplementary Fig. 18). Particularly, Capsicum species had the most copies of CS with seven in C. annuum , six in C. chinense , and four in C. pubescens , C. baccatum and C. rhomboideum . The C. pubescens and C. baccatum had fewer copies than C. annuum , due to either gene loss in the former two, or tandem duplications in the latter. In fact, CS duplication was widely detected in Solanaceae ( Fig. 4B ), as previously reported in tomato 2 , but tandem duplication mainly occurred in Solanum , Physalis , and Capsicum spp. between 30 Mya to 40 Mya ( Fig. 4A ). In addition, microsynteny analysis showed that CS tandem duplicates were syntenic in Capsicum, Physalis , and Solanum ( Fig. 4C ), so were other CBGs (Supplementary Fig. 19). This suggested that expression of CBGs in non-pungent species may have been disrupted. Indeed, we found many CBGs were highly expressed in fruits of pungent Capsicum species ( Fig. 4D ), whereas C. rhomboideum and Physalis CS and KasI genes were hardly expressed ( Fig. 4D ). Sequence alignment revealed that these syntenic CS copies ( CS-1 / CS-2 ) had conserved coding sequences (CDS) and upstream and downstream regulatory regions among pungent species, while non-pungent species had structural variations (SVs) within both CDS and flanking regions ( Fig. 4E; Supplementary Fig. 20). Besides, sequence variations were also observed for several other CBGs including ACL , BCAT, CCoAMT, FatA, and KasI (Supplementary Fig. 19). The highly conserved CS copies were only present in cultivated Capsicum suggesting they were relatively recent (~ 5 Mya) tandem duplicates from older CS genes. Taken together, although non-pungent species contained homologs of functional CBGs, their coding and regulatory regions had diverged significantly from those of C. annuum . Chromatin accessibility regulates tissue-specific capsaicinoid biosynthesis Capsaicinoid biosynthesis is highly tissue-specific, only accumulating in fruits particularly placenta beginning at ~16 days post anthesis. To understand how this tissue-specificity is achieved, we performed multi-omic co-profiling of C. annuum fruits and leaves including ATAC-seq (Assay for Transposase-Accessible Chromatin sequencing), whole-genome bisulfite sequencing and RNA-seq, and analyzed them using CaT2T as reference. RNA-seq analysis showed that CS and its transcriptional regulators MYB31 36 and MYB48 37 were specifically expressed in placenta ( Fig. 4F ). The placenta-specific open chromatin regions (OCRs) with low methylation levels were detected within 2 kb upstream of CS-2 , MYB31 and MYB48 , while CS-1 also showed OCRs in both pulp and seeds, suggesting that CS-2 is likely the primary functional gene that contributes to placenta-specific synthesis of capsaicinoids ( Fig. 4F and Supplementary Fig. 21). Other 26 putative CBGs encoding phenylpropanoid and valine pathways were expressed in both fruits and leaves but lacking tissue-specific OCRs, suggesting their functions beyond capsaicinoid biosynthesis (Supplementary Fig. 12). Nonetheless, their higher expression in placenta than in other tissues (Supplementary Fig. 12) indicated strong co-regulation for capsaicinoid biosynthesis. How co-regulation of genomically dispersed CBGs is achieved to confer this tempo-spatial specificity of capsaicinoid production remains unknown. One would expect there should be common regulatory elements for the CBGs. To validate this hypothesis, we extracted the placenta-associated OCRs identified by ATAC-seq and 2 kb upstream sequence of the CBGs to perform sequence motif enrichment analysis, yielding a total of 38 enriched motifs ( p < 0.01). Overlapping enriched motifs obtained from two enrichment analyses identified five transcription factor binding motifs (TFBS): MYB, G-box, Box-4, ABRE and MYC, present in all seven CBGs including CS, ACL, KasI , PAL , CCoAOMT and BCAT ( Fig. 4G ). It suggested these significantly enriched TFBS within common OCRs were likely recognized by certain TFs such as MYB31 co-regulating these CBGs in specific tissues 36 . In non-pungent C. annuum cultivars, the OCRs of CS-2 were lost due to a 2.4 kb deletion conferring the lack of capsaicinoids in these accessions ( Fig. 4F ). In short, multi-omic data analysis illustrated a likely epigenetic mechanism for the tissue-specific co-regulation of capsaicinoid biosynthesis genes in chili pepper. Machine-learning model for predicting fruit pungency from copy number variants Copy number variants (CNVs) have great impact on plant gene functions and traits such as grain size 39 and biosynthesis of natural products 40 . However, whether and how CNVs impact capsaicinoid biosynthesis is unknown. Fruit pungency levels in pepper cultivars are highly variable owing to artificial selection 41 . Using CaT2T as reference genome, we detected genomewide CNVs from resequencing data of 311 C. annuum accessions with capsaicinoid quantification 38 using AMYCNE (Supplementary Table 14), finding CNVs for 121 CBGs including seven MYB TFs ( Fig. 5A and Supplementary Table 15). Expectedly, a strong association between CS CNV and capsaicinoid production was observed, where non-pungent cultivars had low copy of the putative CS gene ( Fig. 5A ) due to a 2.4 kb deletion spanning functional CDS 1,38 . For cultivars with intact CS gene, capsaicinoid levels might be affected by CNVs of other CBGs. To understand the relationship between CNV and capsaicinoid levels, we trained several machine-learning models including Random Forest, K-nearest neighbor, Gradient Boosting, and SVM (support vector machine), taking CNVs and capsaicinoid levels as input. After performing recursive feature selection and grid search, a Random Forest model achieved the highest average prediction accuracy of approximately 0.72 using 10-fold cross-validation on a dataset composed of CNV detection from newly generated resequencing data with approximately 50×coverage (Table S14). The ROC curve of the Random Forest model also displayed the highest AUC value of 0.85 on this same dataset (Fig. 5C), demonstrating the potential of the model to accurately classify capsaicin levels from CNVs for genomic prediction at an early seedling stage. Furthermore, the Random Forest model ranked 121 CBGs by their contribution to prediction performance. The top-ranked genes included well known CBGs such as CS-2, KasI, C3H, PAL, MYB31 and putative ones as DH , SAMSyn and IPMS etc. ( Fig. 5D ), majority of which were highly expressed in fruits (Supplementary Table 16). Remarkably, the model also correctly ranked the functional copy of CS-2 ( CaT2T02g02791 ) above the other six tandem copies, consistent with current knowledge about this gene. Overall, our machine-learning model demonstrated its prediction power and will facilitate elucidation of the full capsaicin biosynthesis pathway by prioritizing top-ranked CBGs. Discussion The two T2T gap-free Capsicum genome assemblies generated in this study represent key milestones in plant genome research. Firstly, almost ten years after the first sequenced pepper genome 2 , we mapped the complete genome sequence of C. annuum , representing the largest complete genome sequence reported so far. The genomics of chili peppers enter the era of T2T complete genomes, following the footsteps of human, model plant Arabidopsis, rice and maize. Secondly, to investigate the mechanisms of pungency evolution, we assembled a T2T gapless genome for a non-pungent wild pepper C. rhomboideum , the first gapless genome for a wild pepper species. Through phylogenomics involving five Capsicum (four pungent, one nonpungent) and ten non- Capsicum (all nonpungent) genomes, we have traced the evolutionary history of capsaicinoid biosynthesis pathways among closely-related Solanaceae plants by estimating the times for the emergence of pungency, tandem duplications of key genes, and a functional biosynthesis pathway along the evolutionary tree ( Fig. 6 ). However, due to lacking sufficient high quality Capsicum genomes in addition to the two assembled in this study, it is challenging to answer some key interesting evolutionary questions. For example, how did pungent C. baccatum lose all but one tandem duplicates of CS genes, whereas nonpungent C. rhombiodeum managed to keep all four copies? Is it possible the functional CS gene and cis -regulatory elements were already in the Solanaceae plants but later lost in nonpungent species? Alternatively, the five cultivated Capsicum species could gain functional CS genes or cis -regulatory element probably through natural mutations or TE transpositions during domestication. Capsicum genus has over 40 species with diverse genetic backgrounds and traits 41 , but reference genomes are only available for four species so far. With more genomes of Capsicum spp. with diverse capsaicin profiles available in the future, it is possible to better infer when and how the birth and death of the pathway occurred. Many plant secondary metabolites accumulate in multiple tissues and organs, but tissue-specific production of natural product molecules is not uncommon, with examples such as morphines in opium poppy capsules 42 and capsaicins in chili pepper fruits. The expression of the biosynthetic genes in specific tissues is required for the specificity, for which the regulatory mechanisms remain elusive. Typically, metabolic gene clusters can facilitate the co-regulation of biosynthetic genes as seen in the examples of morphine (poppy) 42 and thalianol (Arabidopsis) 43 . However, biosynthesis genes of many plant metabolites such as capsaicin and colchicine (lily) 44 are unclustered and dispersed genomically. The spatial-tempo regulatory mechanisms for biosynthetic genes are poorly understood overall. Through comparative ATAC-seq and RNA-seq analysis of fruit and leaf tissues, we detected fruit-specific open chromatin regions for several key biosynthetic genes and regulators encoding conserved TF binding sites ( Fig. 6 ). This concerted gene regulation likely enabled efficient production of biosynthetic enzymes at specific time and space. How these genes evolved such common cis-regulatory elements requires further investigation. Two T2T gapless genomes allowed us to gain insights into the complex genomic regions such as centromeres, telomeres and filled gaps. Centromeres were extensively studied in T2T genomes of human 13 , Arabidopsis 16 and rice 20 , all containing abundant high-copy tandem repeats. However, centromeres of Capsicum lacked such satellites but were enriched with Gypsy -LTR especially CRM retrotransposons. We also found this pattern was common in C. baccatum , C. chinense and C. pubescens based on the recently reported high-quality genomes 11 , although the composition of CRM retrotransposons varied among different Capsicum species. The potato genome also showed enrichment of CRM in centromeres 21 , but no CRM was identified in whole genome of P. pruinosa (1.38 Gb, contig N50: 82.2 Mb) 45 , suggesting that Solanaceae centromeres evolved rapidly and were diverse among species. Plant centromeres are rapidly diversifying due to cycles of transposon invasions even within species as shown by a recent study on 346 Arabidopsis centromeres 34 , demonstrating the so-called 'centromere paradox'. With more complete Capsicum genomes available, it will be interesting to compare centromeres among different Capsicum accessions or species to understand the evolution of centromeres during speciation and domestication. From application point of view, we first demonstrated two complete genome sequence of peppers, which can serve as a gold standard reference genome for improving variant analysis, and pangenome studies in the future. Secondly, one of the greatest challenges faced by crop breeding is the long breeding cycle and waiting time for trait evaluation. For example, fruit phenotyping of chili peppers could wait 120-150 days, begging for a method to accelerate germplasm assessment. Using capsaicinoid as an example, we performed CNV detection using T2T genome in a pepper population with known capsaicin levels, and trained a robust machine-learning model to predict fruit pungency from CNVs that can be quickly determined in pepper accessions of interest by genome resequencing analysis, or digital PCR of seedlings. To our knowledge, this is the first model that leverages biosynthesis gene copy numbers to infer levels of secondary metabolites, potentially turbo-charging natural product breeding. Admittedly, there is room for improvement as only a small dataset being used in training, but new datasets are expected to further train and strengthen the model. In summary, this study produced so far the largest complete plant genome assembly and the first two T2T gap-free genomes for Capsicum . Phylogenomics and multiomics unraveled the evolutionary mechanisms of Capsicum-unique and tissue-specific accumulation of capsaicinoids in pepper fruits. The new T2T genomic resources represent an important milestone in crop genome research, and will speed up pepper research and precise improvement. Methods Plant materials and sequencing Capsicum annuum double haploid line ‘G1-36576’ and C apsicum rhomboi deum wild accession ‘PI 645680’ plants were grown in regular azalea pots filled with a combination of potting mix, clay and vermiculite in greenhouse of Peking University Institute of Advanced Agricultural Sciences, Weifang (36°42'N and 119°10'E), Shandong province, China in the summer and autumn of 2022. The fresh leaves of four weeks old C apsicum annuum and C apsicum rhomboideum were harvested and used for DNA extraction and sequencing. Tissues of leaf, root, stem and flower at two days post anthesis, and fruits (separated into pericarp, placenta and seeds) at 21 days post anthesis for C. annuum and C. rhomboideum were collected for RNA extraction and sequencing. Tissues of leaf, placenta, pericarp and seeds at 21 days post anthesis were collected from C. annuum and used for ATAC-seq. Tissues of leaf and whole fruit at 21 days post anthesis were collected from C. annuum and used for bisulfite sequencing. Additionally, fresh leaves of nine pepperaccessions were harvested and used for re-sequencing. DNA and RNA isolation Isolation of high molecular weight (HMW) genomic DNA was conducted using CTAB (cetyltrimethylammonium bromide) method. Briefly, 10 μg clean and fresh leaves were ground in liquid nitrogen and then subjected to DNA extraction. The quality of DNA was checked using Qubit (Thermo Fisher Inc.) and a pulse field gel electrophoresis apparatus (BioRad) following manufacturer's recommendation. The isolation of total RNA was conducted using Trizol RNA extraction reagent (Thermo Fisher Inc.) following recommended protocol. The extracted RNA was assessed using RNA Nano 6000 Assay Kit of the Bioanalyzer 2100 system (Agilent Technologies, CA) following manufacturer’s recommendation. RNA samples with a RIN (RNA integrity number) value > 6.0 were proceeded to downstream library construction for RNA sequencing. Genome sequencing Illumina paired-end sequencing library was prepared using NEB Next® Ultra™ DNA Library Prep Kit for Illumina (NEB, USA) following its standard protocol. Briefly, 5 µg HMW DNA sample was fragmented by sonication to a size of 350 bp. DNA fragments were then end-polished, A-tailed, and ligated with the full-length Illumina sequencing adapters. A total of 300.8 Gb (~100 × genome coverage) 150 bp paired-end reads were produced using Illumina Novoseq 6000 platform by Novogene Biotechnologies Inc. (Tianjin, China). The clean data were used in genome survey, genome assembly polishing and assembly evaluation. To generate PacBio HiFi long reads, a total of 15 µg HMW DNA was sheared by gTUBEs (Covaris, MA, USA), and was used to construct a standard PacBio SMRTbell library using PacBio SMRTbell Express Template Prep Kit 2.0 (PacBio, CA, USA). The resultant library was separated on BluePippin (Sage Science, MA, USA) with 15 kb cutoff to remove short DNA fragments. Then a total of 356.3 Gb HiFi consensus reads with N50 length of 18.3 kb were generated using a PacBio Sequel II system at Novogene Biotechnologies Inc. (Tianjin, China). To generate Oxford Nanopore ultra-long reads, the long DNA fragments were size-selected and processed following the Ligation Sequencing SQK-LSK109 Kit (Oxford Nanopore Technologies, Oxford, UK) protocol. Briefly, DNA ends were FFPE (Formalin-Fixed and Parrffin Embedded) repaired and end-prepped/dA-tailed using the NEBNext End Repair/dA-tailing module (New England Biolabs, UK). Then sequencing adapters were ligated onto the prepared ends using the NEBNext Quick Ligation module (New England Biolabs, UK). The final DNA library was sequenced using the GridION X5/PromethION sequencer (Oxford Nanopore Technologies, Oxford, UK) by Single-Molecule Sequencing Platform at Peking University Institute of Advanced Agricultural Sciences (Weifang, China). A total of 261.5 Gb ultra-long reads were generated, with read length N50s of 100.3 kb and reads nearly up to 3 Mb. The Hi-C library was prepared from cross-linked chromatins of pepper leaves using a standard Hi-C protocol 46 . Then the library was sequenced using Illumina NovaSeq 6000 to obtain 2×150 bp paired-end reads at Novogene Biotechnologies Inc. (Tianjin, China). A total of 348.2 Gb of Hi-C data with ~112 × coverage was generated, and classified as valid or invalid interactions using HiC-Pro v3.1.0 47 , and only valid interaction pairs were retained for subsequent analysis. Genome assembly Step1: Genome size and heterozygosity of DH line ‘G1-36576’ were estimated using the Illumina data by Jellyfish v2.3.0 ( k -mer size = 19) 48 and Genomescope v1.0 (max k -mer coverage = 1,000,000) 49 . The estimated genome size is 3.19 Gb and the heterozygous rate is 0.207%. Step2: For the PacBio assembly, HiFi reads were assembled using Hifiasm (v0.16.1) 24 with the default parameters. The ONT assembly was conducted using NextDenovo (v2.5.0) 25 and polished using NextPolish (v1.4.0) 50 with parameters of ‘hifi_options = -min_read_len 1k -max_depth 100’ and ‘sgs_options = -max_depth 100 -bwa’. Then we aligned the contigs to the reference genomes of C. annuum chloroplast (GenBank accession NC_018552.1) and mitochondria (GenBank accession NC_024624.1) with Minimap2 (v2.24) 51 . Contigs with at least 50% of their bases covered by chloroplast or mitochondria genome sequences were removed from the assembly. Step3: Quickmerge 52 was used to joint contigs in the HiFi assembly (as the query) using the contigs from the ONT assembly (as the reference), which created a HiFi and ONT hybrid assembly. As ONT long reads are usually error-prone, we aligned the HiFi contigs to the merged assembly, and replaced the ONT originated sequences with corresponding HiFi contigs. After this step, ten nearly complete chromosome-level contigs were obtained, while two contigs corresponding to Chr04 and Chr08 only possessed one single telomere. Then Hi-C sequencing data were used to anchor all contigs using the pipeline of Juicer (v1.5) 53 , 3D-DNA (v180419) 54 and Juicebox (v1.11.08) 26 . For assembly validation, the contigs were manually checked, orientation tuned and any mis-assembly was adjusted within Juicebox 26 . Step4: The rDNA arrays on the acrocentric Chr08 are long tandem repeats of 45S unit (18S-5.8S-25S rDNA). To assemble the 45S rDNA arrays, we first estimated the rDNA copies. We used Barrnap v0.9 (https://github.com/tseemann/barrnap) to predict the location of rDNA in HiFi reads and extracted the 45S rDNA-containing HiFi reads. The copy number was estimated to be ~60,000/42 = 1,428 based on the 19-mers of 45S rDNA-containing HiFi reads (>20 kb, 42×depth). Two main types of repeat unit with different length were identified, Type A: 8,351-8,377 bp and Type B: 8,498-8,506 bp, each accounting for 70% and 30% of the total 45S rDNA arrays. To assemble the rDNA tandem arrays, we utilized the centroFlye HOR pipeline 55 as a reference. Due to the high similarities between rDNA units and the error-prone property of ONT long reads, we failed to assemble the NOR regions using ONT data. The 45S rDNA containing ONT long reads were used to extract prefix reads that containing telomeric repeats, internal reads that containing two types of rDNA at both ends and suffix reads that containing non-rDNA sequences. Then we assembled the 45S rDNA containing HiFi reads using Hifiasm 24 , generating a draft rDNA assembly (78 contigs, N50 of 445.3 kb, and sum of 15.5 Mb). Combined the extracted ONT reads and the assembled HiFi contigs, we identified rare 19-mers and connected the sequences with same unique 19-mers. We then used Hi-C data to anchor these sequences and filled the gaps by mapping the HiFi reads to the rDNA assembly using Winnowmap2 (v2.03, k=19, -x asm5) 56 . We finally obtained 12.66 Mb rDNA arrays with 1,506 rDNA copies, and added this sequence to the contig of Chr08. Step5: We extracted ultra long ONT reads (> 200kb) with at least ten copies of telomeric repeat motif ‘TTTAGGG’ or ‘TTCAGGG’ variant, and aligned these reads to above genome assembly using Winnowmap2 (v2.03, k=19, -ax map-ont) 56 . Using these alignment coordinates, the overhang sequences of telomere-containing reads were manually patched to each telomere. Telomeres were then manually confirmed to be structurally valid. Finally, we obtained a T2T genome assembly of C. annuum accession G1-36576, and named CaT2T. Similarly, the C. rhomboideum genome was assembled using the same strategy described above. The final assembly was named CrT2T. Genome quality assessment To assess the quality of the genome assembly, we firstly compared the genomic alignment dot plot between CaT2T and Ca59 assemblies using Minimap2 51 and D-GENIES 57 . For mapping statistics, the NGS short reads were mapped using BWA (v0.7.17) 58 , and the HiFi and ONT long reads were mapped using Minimap2 51 . Then Samtools (v1.10) 59 was used to count the mapping rates and coverage depth. The Ca59 gap regions that are resolved in CaT2T genome were manually checked in IGV (v2.12.3) 60 windows. To assess genome completeness, we applied BUSCO (v5.4.3) 61 for ortholog detection using solanales_odb10 database (n = 5,950). Quality value (QV) was estimated using Merqury (v1.3) 62 from HiFi reads. The telomere sequences were identified using Tandem Repeat Finder (TRF, v4.09.1) 63 with the parameters of ‘2 7 7 80 10 80 2000 -d -l 16’. The resulting ‘.dat file’ was transformed into a GFF3 file, which was then used to identify seven base telomeric repeats. Repeat annotation and TE analysis We used the universal Repbase database and a species specific de novo repeat library constructed by RepeatModeler (https://github.com/Dfam-consortium/RepeatModeler) to annotate the DNA sequences in two Capsicum species. The repetitive elements in the genome were then annotated and masked by RepeatMasker (v4.1.2) 64 using the following parameters of ‘-xsmall -s -no_id -cutoff 255 -frag 20000 -e ncbi’. To achieve large-scale accurate discovery of LTR retrotransposons, we applied LTR_Finder (v1.2) 65 , LTRharvest (v1.6.2) 66 , and LTR_retriever (v2.9.0) 67 to identify LTR elements. We identified 7,383 and 9,579 intact LTR-RTs candidates in CaT2T and CrT2T, respectively, which were used as input to the TEtranscripts analysis 68 . Then TEsorter (v1.3) 69 was implemented using HMM profiles obtained from the TE protein domain databases REXd-plant. The TE sequences were first translated in all six frames and the translated sequences were then searched against the database. Hits with coverage lower than 20% or E-value higher than 1e-3 were discarded. For the classifications of LTR-RTs, intact elements were identified and classified based on the presence and order of five conserved domains, including capsid protein (GAG), aspartic proteinase (AP), integrase (INT), reverse transcriptase (RT), and RNase H (RH). Then the number of LTR-RTs decreased to 5,202 and 6,834 in CaT2T and ChT2T after filtered by the conserved domains. Using TEsorter (v1.3) 69 , the Ty1- Copia elements was classified into several clades, including Ale, Alesia, Angela, Bianca, Ikeros, Ivana, SIRE, TAR, and Tork ; while the Ty3- Gypsy elements was classified into clades of Athila, CRM, Galadriel, Ogre, Reina, and Tekay . The ratios of solo LTRs to intact LTRs in each LTR family were calculated using the script of ‘solo_intact_ratio.pl’ in the LTR_retriever software. The insertion time of intact LTR retrotransposons were calculated using LTR_retriever according to the formula: T= K/2r, where K was the divergence between the two LTRs and r was the rate of nucleotide substitution. We employed an average substitution rate of (r) 7 × 10 -9 to estimate insertion times of LTR-RTs. Genome annotation Gene model prediction combined with the following three aspects of evidence: (a) ab initio prediction, (b) homologous protein, and (c) RNA-seq evidence, was conducted using MAKER (v2.31.11) 70 pipeline in two successive rounds. At the first round, short read and full-length RNA-seq and homology proteins were provided. The protein sequences used for homology-based prediction were from A. thaliana 16 , C. annuum 11 , S. tuberosum 21 , and universal Swiss-Prot proteins. To implement the MAKER pipeline, short read RNA-Seq data was assembled into a transcriptome using StringTie (v2.2.1) 71 . The PacBio long read transcriptome data were processed using the SMRT Analysis software Isoseq3 (https://github.com/PacificBiosciences/IsoSeq). MAKERemployed BLAST to align transcripts and proteins to the soft-masked genome, then Exonerate (v2.2.0) 72 were used to polish the BLAST hits and thereby accurately annotate the coding regions. The parameters of est2genome and protein2genome were set to 1, so that MAKERpredicted gene models only based on the provided transcripts and proteins. Then a subset of MAKERgene models with AED 73 scores < 0.25 were used to train SNAP 74 for three rounds. The GeneMark-ET and Augustus model were trained using BRAKER (v2.1.6) 75 pipeline. Briefly, the same data were aligned to the soft-masked genome using Exonerate 72 and HISAT2 (v2.1.0) 76 , then GeneMark-ET 77 was trained by the predicted gene structures, and the resulting ~6,000 good gene models were used for training AUGUSTUS (v3.2.3) 78 . In the second round, each set of gene predictions in round one was passed to MAKER through model_gff option and the evidence alignment options were turned off. The trained SNAP, GeneMark-ET and AUGUSTUS models were also integrated into MAKER to predict more credible genes. Finally, the unsupported gene models were filtered (keep_preds=0) and the highest-ranking gene sets with AED scores < 0.5 were retained. To compare the previously published genome annotation of C. annuum 4,11 with our CaT2T genome annotation, we also performed Liftoff (v1.6.3) 79 to annotate protein-coding genes of CaT2T assembly based on the reference with parameters of “-flank 0.1 -sc 0.99 --copies”. Then Gffread (v0.12.7) 80 was used to filter transcripts without normal open reading frames accordingly. Gene models were finally manually checked and corrected in IGV-GSAman (https://gitee.com/CJchen/IGV-sRNA) with the support of mapped RNA-seq reads and previous annotations 4,11 . Synteny and phylogenome analysis Non-redundant protein sequences from 16 species were prepared for ortholog analyses (Supplementary Table 12). Orthologues and orthogroups were then inferred using OrthoFinder (v2.5.4) 35 with default values setting and ‘-M msa’ activated. The longest predicted protein of each individual gene was used as the representative input for the OrthoFinder analysis. TrimAl (v1.4.12) 81 was used to remove poorly aligned regions of protein multiple sequence alignments. RAxML (v8.2.12) 82 was used to build Maximum Likelihood phylogenetic trees using the GAMMAJTT model, with rice as an out-group. TimeTree (www.timetree.org) is a public database containing divergence time estimates from various publications along with their own estimation. These estimates, ignoring the outliers, were used for selecting the range of lower and upper uniform calibration priors. The calibration values were chosen as 1.1-1.6, 109.2-123.5, and < 200 for the most common ancestor of the 13 species belonging to Solanum, dicotyledons, and all plants, respectively. The CodeML and MCMCTree programs in the PAML (v4.9) 83 were used to analyze amino acid substitution models and estimate divergence times. The CAFE5 84 was then applied to infer gene gain and loss rates in each genome. The orthogroups generated by OrthoFinder were regarded as distinct gene families and provided as inputs for CAFE5 analysis. The identified genes were subjected to GO and KEGG enrichment analysis, with the p-value of significant enrichment was set as 0.05. The syntenic analysis was performed by JCVI (v1.1.19) 85 . We identified synteny blocks by performing an all-against-all LAST search and chaining the hits with a distance cutoff of 20 genes. Additionally, we required each synteny block to have at least five gene pairs. The dotplot of major CBGs were shown by Gepard 86 . The Ks values of C. annuum syntenic block genes were calculated using ParaAT (v2.0) 87 . Machine-learning model development AMYCNE 88 was employed to investigate copy number variations (CNVs) of putative capsaicin biosynthesis genes from whole genome re-sequencing reads, including 311 publicly available accessions of C. annuum (BioProject accession: PRJCA004361) (for model training) and nine additional accessions sequenced in this study (for model testing). To train machine-learning models for capsaicin classification, we employed a balanced dataset created by dividing samples based on the median capsaicinoid content. The input features comprised CNVs of putative capsaicinoid biosynthetic genes, while the output was categorized into high and low capsaicinoid content based on the median value. The primary reason for choosing binary classification of capsaicinoid content stems from the specific goals of our study and the characteristics of our dataset. Given the limited size of our dataset, a binary classification model proved to be more robust and less sensitive to noise. Furthermore, the binary classification framework offers a clearer and more direct interpretation for our objectives. It simplifies the understanding of gene influence on capsaicinoid presence, which is essential for practical applications such as selective breeding in agriculture. This approach aligns closely with our aim to discern the genetic factors contributing to capsaicinoid content rather than quantifying it. Before proceeding with model training, we implemented a feature selection step to refine our input features and enhance model performance. Using a Random Forest Classifier as the base estimator, we applied Recursive Feature Elimination with Cross-Validation (RFECV) to identify the most predictive features. This method involved iteratively removing features and evaluating model accuracy using Stratified 5-fold cross-validation to ensure a balanced representation of each class. The RFECV process identified an optimal subset of features, which were then used for model training. Four primary algorithms—Random Forest, Support Vector Machines (SVM), Gradient Boosting, and K Nearest Neighbors (KNN)—were utilized to model the relationship between input features and classification outcomes. A 10-fold cross-validation approach was adopted for model evaluation, using implementations from the scikit-learn library. We applied a Grid Search approach to meticulously explore an extensive range of hyperparameters for each algorithm, aiming to identify the most effective model parameters. In the selection of kernel types for the SVM models, we focused on polynomial (poly) and linear kernels to enhance the interpretability of input features in subsequent analyses. Supplementary Table 17 provides a detailed overview of the models and their respective hyperparameter ranges. The random forest model with following parameters “max_depth=20, max_features=’sqrt’ and n_estimators=150” exhibited the highest accuracy, approximately 72%. This configuration achieved superior performance as evidenced by the ROC curve, attaining the highest AUC value of 0.85. This underscores its proficiency in classifying capsaicinoid content based on input CNVs of significant genes. Then the importance of input feature genes for capsaicinoid content were also evaluated with this classification model. CENH3 ChIP-seq An antigen with full peptide sequence corresponding to C. annuum CENH3 was used to produce the C. annuum anti-CENH3 antibodies in rabbit. For ChIP experiment, pepper seedlings were fixed with 1% formaldehyde solution in MS buffer (10 mM potassium phosphate, pH 7.0; 50 mM NaCl) at room temperature for 15 min in a vacuum. After fixation, the seedlings were incubated at room temperature for 5 min under vacuum with 0.15 M glycine. Approximately 1 g fixed tissue was homogenized with liquid nitrogen and purifying nuclei, and resuspended in 1 ml of cell lysis buffer & incubate for 10 min on ice & spin at 1500 rpm (RC-3B, 600 × g) for 5 min (Cell lysis buffer: 10 mM Tris, 10 mM NaCl, 0.2% NP-40 [pH 8.0], 1×protease inhibitors). The cell lysis was further resuspended in 1 ml of nuclei lysis buffer for 10 min on ice (Nuclei lysis buffer: 50 mM Tris, 10 mM EDTA, 1% SDS, 1×protease inhibitors) to isolate nuclei. The resuspended chromatin solution was sonicated five times for 15 s each at ∼10% power (setting 2.5 on the sonicator, Sanyo Soniprep 150). The volume of the chromatin sample was measured and then ChIP dilution buffer was added to 1 ml of chromatin with 2.5 µg of anti-H3K4me3 and incubated for 12 h at 4℃. 50 µl protein A/G Beads were added and incubated for 4 h at 4℃. Beads were washed twice with each of the following buffers: wash buffer A (50 mM HEPES-KOH pH 7.5, 140 mM NaCl, 1 mM EDTA pH 8.0, 0.1% Na-Deoxycholate, 1% Triton X-100, 0.1% SDS), wash buffer B (50 mM HEPES-KOH pH 7.9, 500 mM NaCl, 1 mM EDTA pH 8.0, 0.1% Na-Deoxycholate, 1% Triton X-100, 0.1% SDS), wash buffer C (20 mM Tris-HCl pH8.0, 250 mM LiCl, 1 mM EDTA pH 8.0, 0.5% Na-Deoxycholate, 0.5% IGEPAL C-630, 0.1% SDS), wash buffer D (TE with 0.2% Triton X-100), and TE buffer. To purify eluted DNA, 200 μl TE was added and then RNA was degraded by the addition of 2.5 μl of 33 mg/mL RNase A (Sigma, R4642) and incubation at 37℃ for 2 hours. The DNA was then resuspended in 50 μl TE and amplified with the VAHTS® Universal DNA Library Prep Kit for Illumina V3 (Vazyme ND607). Amplified ChIP libraries were sequenced on the Illumina Novaseq 6000 platform. Epigenomic sequencing and data analysis Hi-C data was generated from leaf tissue as above mentioned, and processed using HiC-Pro (v3.1.0) 47 and Juicertools(v1.22.01) 53 , to generate 10 kb, 15 kb, 20 kb, 25 kb, 40 kb, 100 kb and 500 kb contact maps. A/B Compartment was identified by R (v4.2.0) script using HiTC (v1.42.0) 89 and Cworld-dekker (v0.0.1) (https://github.com/dekkerlab/cworld-dekker) in 100 kb iced contacts matrix. Chromatin accessibility of C. annuum was profiled using ATAC-seq construction kit (Vazyme Ltd. Nanjing China) following recommended protocol. The quality of the constructed libraries was assessed using a qubit followed by Agilent Bioanalyzer 2100 for fragment analysis. The libraries were sequenced using Illumina NovoSeq 6000 platform at Novogene Inc. (Tianjin, China). Three biological replicates were generated for each plant tissue and analyzed using the same computational methods. The ATAC-seq data were analyzed using an in-house computational pipeline. Basically, the clean reads of ATAC-seq were mapped to CaT2T reference genome using BWA-MEM (v2.2.1) 90 with default parameters. The alignment file (.bam) were used to call peaks by MACS2(v2.2.7.1) 91 . Histone modification ChIP-seq data was downloaded from public CNGBdb database with accession number CNP0001129. Whole genome bisulfite sequencing was conducted on tissues of leaf and whole fruit from C. annuum . ChIP-seq mapping and peaks calling were run with the commands, using bowtie2 (v2.5.1) 92 , Samtools(v1.10) 59 , and MACS2(v2.2.7.1) 91 . The visualization of centromeric repeats was accomplished using StainedGlass 93 . DNA methylation level was estimated using Bismark (v0.24.0) 94 , after mapping WGBS (whole-genome bisulfite sequencing) data to reference genome using BWA-MEM (v2.2.1) 90 . Transcriptome sequencing and analysis Total RNA was extracted from seven tissues, including leaf, flower, placenta, root, stem, seeds and pericarp. The mRNA was then subjected to transcriptome sequencing library construction using Illumina True-seq transcriptome kit (Illumina, CA). The libraries were then sequenced using Illumina Novaseq 6000 platform at Biomarker Technologies Corporation (QingDao, China) to generate 150 bp paired-end reads. For full-length transcriptome sequencing, about 5 µg mRNA was reverse-transcribed into full-length cDNA molecules with SMARTer™ PCR cDNA Synthesis Kit (Clontech, CA, USA) and the cDNA was further amplified by PCR. End repairing was conducted on amplified cDNAs, followed by SMRTbell adapter ligation. The ligation products are further treated by exonuclease to degrade failed ones before the Iso-seq library was sequenced using PacBio Sequal IIe instrument at Biomarker Technologies Corporation (QingDao, China). Full-length transcripts were assembled across tissues using the SMRTlink pipeline and used for guiding gene annotation. We quantified gene expression level using kallisto (v0.48.0) 95 . Counts for mapped reads were normalized by TPM (transcripts per million). Reads alignment was performed using HISAT2 (v2.1.0) 76 . To visualize the expression pattern of the focused genes among samples, heatmaps were generated using the R package. Quantification of capsaicinoids using HPLC-MS Capsaicinoids were extracted from mature fruits of chili peppers and subjected to HPCL-MS quantification according to the previous method 96 . All the detections were performed on a Vanquish UHPLC system combined with a TSQ Altis MS/MS system (Thermo Scientific, USA). ACQUITY UPLC@HSS T3 column (150 mm × 2.1 mm, 1.8 μm particle size, Waters) was used for the separation of samples with the column oven at 35℃. Isocratic elution was carried out at a flow rate of 0.5 mL/min using 50% Millipore water with 0.1% formic acid and 50% acetonitrile with 0.1% formic acid. The separation was finished at 35 min with an injection volume of 15 µL and the thermostat of 30℃ for the analysis. Declarations Data Availability The raw sequencing data (PacBio HiFi, ONT, Illumina paired-end and Hi-C) and genome assembly have been deposited in the National Center for Biotechnology Information (NCBI) under the Bioproject (PRJNA962192). The genome assembly and annotation are also available in Capsicum Genome Database (http://www.pepperbase.site/node/3) of Peking University Institute of Advanced Agricultural Sciences. Code Availability All codes were deposited at https://github.com/Weikai-47/Pepper_T2T. Any additional information required to reanalyze the data reported in this paper is available from the corresponding authors upon request. Acknowledgments We would like to thank the Bioinformatics Platform at Peking University Institute of Advanced Agricultural Sciences for providing the high-performance computing resources. We would like to thank East-West Seed Group for kindly providing the double haploid pepper seeds. This work was supported by Shandong Provincial Science and Technology Innovation Fund and the National Natural Science Foundation of China (31970317). LG is also supported by Taishan Scholars Program and Natural Science Foundation for Distinguished Young Scholars (ZR2023JQ010) of Shandong Province. Author Contributions L.G., H.H., X.W.D. and X.Z. conceived and designed the project. Y.M., J.B., B.L. and Z.Z. maintained the plant materials and coordinated sequencing data generation. W.C. and X.F.W. assembled the genomes and performed genome validations and annotations. X.F.W., X.R.W., W.C., J.S., S.Y., D.H.A., M.Y., K.W. and L.G. conducted bioinformatic analysis and prepared figures and tables. D.M. conducted epigenome sequencing. J.L. and L.W. performed quantification of capsaicinoids and gene copy numbers. S.C. constructed the genome database. T.M. developed the machine learning model. L.Z., R.C. and J.J. assisted in the result interpretation and discussion. L.G., W.C., L.Z. and D.H.A. wrote the manuscript. L.G., X.W.D., X.Z. and H.H. revised the manuscript. All authors read and approved the final version of the manuscript. Competing Interests The authors declare no competing interests. References Stewart, Jr.C. et al. The Pun1 gene for pungency in pepper encodes a putative acyltransferase. Plant J , 42 , 675-688 (2005). Kim, S.et al. Genome sequence of the hot pepper provides insights into the evolution of pungency in Capsicum species. Nat. Genet. 46 , 270-278 (2014). Stewart, Jr.C. et al. Genetic control of pungency in C. chinense via the Pun1 locus. J. Exp. Bot. 58 , 979-991 (2007). Liao, Y.et al. The 3D architecture of the pepper genome and its relationship to function and evolution. Nat. Commun. 13 , 3479 (2022). Shirasawa, K., Hosokawa, M., Yasui, Y., Toyoda, A., & Isobe, S. Chromosome-scale genome assembly of a Japanese chili pepper landrace, Capsicum annuum ‘Takanotsume’. DNA Res. 30 , dsac052 (2023). Lee, J.H. et al. High-quality chromosome-scale genomes facilitate effective identification of large structural variations in hot and sweet peppers. Hortic. Res. 9 , uhac210 (2022). Qin, C. et al. Whole-genome sequencing of cultivated and wild peppers provides insights into Capsicum domestication and specialization. Proc. Natl. Acad. Sci. USA 111 , 5135-5140 (2014). Kim, S. et al. New reference genome sequences of hot pepper reveal the massive evolution of plant disease-resistance genes by retroduplication. Genome Biol. 18 , 210 (2017). Hulse-Kemp, A.M. et al. Reference quality assembly of the 3.5-Gb genome of Capsicum annuum from a single linked-read library. Hortic. Res. 5 , 4 (2018). Kim, M.S. et al. Comparative analysis of de novo genomes reveals dynamic intra-species divergence of NLRs in pepper. BMC Plant Biol. 21 , 247 (2021). Liu, F. et al. Genomes of cultivated and wild Capsicum species provide insights into pepper domestication and population differentiation. Nat. Commun , 14 , 5487 (2023). Nurk, S.et al. The complete sequence of a human genome. Science 376 , 44-53 (2022). Altemose, N. et al. Complete genomic and epigenetic maps of human centromeres. Science 376 , eabl4178 (2022). Gershman, A.et al. Epigenetic patterns in a complete human genome. Science 376 , eabj5089 (2022). Aganezov, S. et al. A complete reference genome improves analysis of human genetic variation. Science 376 , eabl3533 (2022). Naish, M.et al. The genetic and epigenetic landscape of the Arabidopsis centromeres. Science 374 , eabi7489 (2021). Hou, X., Wang, D., Cheng, Z., Wang, Y., & Jiao, Y. A near-complete assembly of an Arabidopsis thaliana genome. Mol. Plant 15 , 1247-1250 (2022). Wang, B.et al. High-quality Arabidopsis thaliana genome assembly with Nanopore and HiFi long reads. Genom. Proteom. Bioinf. 20 , 4-13 (2022). Song, J.et al. Two gap-free reference genomes and a global view of the centromere architecture in rice. Mol. Plant 14 , 1757-1767 (2021). Shang, L. et al. A complete assembly of the rice Nipponbare reference genome. Mol. Plant 16 , 1232-1236 (2023). Yang, X. et al. The gap-free potato genome assembly reveals large tandem gene clusters of agronomical importance in highly repeated genomic regions. Mol. Plant 16 , 314-317. (2023). Wang, L. et al. A telomere-to-telomere gap-free assembly of soybean genome. Mol. Plant (2023). Chen, J. et al. A complete telomere-to-telomere assembly of the maize genome. Nature Genet , 1-11 (2023). Cheng, H., Concepcion, G.T., Feng, X., Zhang, H., & Li, H. Haplotype-resolved de novo assembly using phased assembly graphs with hifiasm. Nat. Methods 18 , 170-175 (2021). Hu, J.et al. An efficient error correction and accurate assembly tool for noisy long reads. Preprint at BioRxiv. (2023). Durand, N.C. et al. Juicebox provides a visualization system for Hi-C contact maps with unlimited zoom. Cell Syst. 3 , 99-101 (2016). Zhang, Z.X. et al. Discovery of putative capsaicin biosynthetic genes by RNA-Seq and digital gene expression analysis of pepper. Sci. Rep. 6 , 34121 (2016). Cleveland, D.W., Mao, Y., & Sullivan, K.F. Centromeres and kinetochores: from epigenetics to mitotic checkpoint signaling. Cell 112 , 407-421 (2003). Zhang, H. et al. Boom-bust turnovers of megabase-sized centromeric DNA in Solanum species: Rapid evolution of DNA sequences associated with centromeres. Plant Cell 26 , 1436-1447 (2014). Ahmed, H.I. et al. Einkorn genomics sheds light on history of the oldest domesticated wheat. Nature 620 , 830-838 (2023). Yang, Z. et al. Cotton D genome assemblies built with long-read data unveil mechanisms of centromere evolution and stress tolerance divergence. BMC Biol . 19 , 1-22 (2021). Vitte, C., & Panaud, O. Formation of solo-LTRs through unequal homologous recombination counterbalances amplifications of LTR retrotransposons in rice Oryza sativa L. Mol. Biol. Evol. 20 , 528-540 (2003). Neumann, P. et al. Plant centromeric retrotransposons: a structural and cytogenetic perspective. Mobile DNA 2 , 4 (2011). Wlodzimierz, P. et al. Cycles of satellite and transposon evolution in Arabidopsis centromeres. Nature 618 , 557-565 (2023). Emms, D.M., & Kelly, S. OrthoFinder: phylogenetic orthology inference for comparative genomics. Genome Biol. 20 , 238 (2019). Zhu, Z.et al. Natural variations in the MYB transcription factor MYB31 determine the evolution of extremely pungent peppers. New Phytol. 223 , 922-938 (2019). Sun, B. et al. Coexpression network analysis reveals an MYB transcriptional activator involved in capsaicinoid biosynthesis in hot peppers. Hortic Res 7 , 162 (2020). Cao, Y.et al. Pepper variome reveals the history and key loci associated with fruit domestication and diversification. Mol. Plant 15 , 1744-1758 (2022). Wang, S.et al. The OsSPL16-GW7 regulatory module determines grain shape and simultaneously improves rice yield and grain quality. Nat. Genet. 47 , 949-954 (2015). Shirai, K.et al. A highly specific genome-wide association study integrated with transcriptome data reveals the contribution of copy number variations to specialized metabolites in Arabidopsis thaliana accessions. Mol. Biol. Evol. 34 , 3111-3122 (2017). Carrizo, G.C. et al. Phylogenetic relationships, diversification and expansion of chili peppers ( Capsicum , Solanaceae). Ann. Bot. 118 , 35-51 (2016). Guo, L. et al. The opium poppy genome and morphinan production. Science 362 , 343-347 (2018). Huang, A.C. et al. A specialized metabolic network selectively modulates Arabidopsis root microbiota. Science 364 , eaau6389 (2019). Nett, R.S., Lau, W., & Sattely, E.S. Discovery and engineering of colchicine alkaloid biosynthesis. Nature 584 , 148-153 (2020). He, J. et al. Establishing Physalis as a Solanaceae model system enables genetic reevaluation of the inflated calyx syndrome. Plant Cell 35 , 351-368 (2023). Belton, J.M. et al. Hi-C: a comprehensive technique to capture the conformation of genomes. Methods 58 , 268-276 (2012). Servant, N. et al. HiC-Pro: an optimized and flexible pipeline for Hi-C data processing. Genome Biol. 16, 259 (2015). Marcais, G., & Kingsford, C. A fast, lock-free approach for efficient parallel counting of occurrences of k -mers. Bioinformatics 27 , 764-770 (2011). Vurture, G.W. et al. GenomeScope: fast reference-free genome profiling from short reads. Bioinformatics 33 , 2202-2204 (2017). Hu, J., Fan, J., Sun, Z., & Liu, S. NextPolish: a fast and efficient genome polishing tool for long-read assembly. Bioinformatics 36 , 2253-2255 (2020). Li, H. Minimap2: pairwise alignment for nucleotide sequences. Bioinformatics 34 , 3094-3100 (2018). Chakraborty, M., Baldwin-Brown, J.G., Long, A.D., & Emerson, J.J. Contiguous and accurate de novo assembly of metazoan genomes with modest long read coverage. Nucleic Acids Res. 44 , e147 (2016). Durand, N.C.et al. Juicer provides a one-click system for analyzing loop-resolution Hi-C experiments. Cell Syst 3 , 95-98 (2016). Dudchenko, O. et al. De novo assembly of the Aedes aegypti genome using Hi-C yields chromosome-length scaffolds. Science 356 , 92-95 (2017). Bzikadze, A.V., & Pevzner, PA. Automated assembly of centromeres from ultra-long error-prone reads. Nat. Biotechnol. 38 , 1309-1316 (2020). Jain, C., Rhie, A., Hansen, N.F., Koren, S., & Phillippy, A.M. Long-read mapping to repetitive reference sequences using Winnowmap2. Nat. Methods 19 , 705-710 (2022). Cabanettes, F., & Klopp, C. D-GENIES: dot plot large genomes in an interactive, efficient and simple way. PeerJ 6 , e4958 (2018). Li, H., & Durbin, R. Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics 25 , 1754-1760 (2009). Li, H. et al. The sequence alignment/map format and SAMtools. Bioinformatics 25 , 2078-2079 (2009). Robinson, J.T. et al. Integrative genomics viewer. Nat. Biotechnol. 29 , 24-26 (2011). Manni, M., Berkeley, M.R., Seppey, M., Simao, F.A., & Zdobnov, EM. BUSCO Update: novel and streamlined workflows along with broader and deeper phylogenetic coverage for scoring of eukaryotic, prokaryotic, and viral genomes. Mol. Biol. Evol. 38 , 4647-4654 (2021). Rhie, A., Walenz, B.P., Koren, S., & Phillippy, A.M. Merqury: reference-free quality, completeness, and phasing assessment for genome assemblies. Genome Biol. 21 , 245 (2020). Benson, G. Tandem repeats finder: a program to analyze DNA sequences. Nucleic Acids Res. 27 , 573-580 (1999). Tarailo-Graovac, M., & Chen, N. Using RepeatMasker to identify repetitive elements in genomic sequences. Curr. Protoc. Bioinformatics 5 , 4-10 (2009). Xu, Z., & Wang, H. LTR_FINDER: an efficient tool for the prediction of full-length LTR retrotransposons. Nucleic Acids Res. 35 , W265-W268 (2007). Ellinghaus, D., Kurtz, S., & Willhoeft, U. LTRharvest, an efficient and flexible software for de novo detection of LTR retrotransposons. BMC Bioinformatics 9 , 18 (2008). Ou, S., & Jiang, N. LTR_retriever: a highly accurate and sensitive program for identification of long terminal repeat retrotransposons. Plant Physiol. 176, 1410-1422 (2018). Jin, Y., Tam, O.H., Paniagua, E., & Hammell, M. TEtranscripts: a package for including transposable elements in differential expression analysis of RNA-seq datasets. Bioinformatics 31 , 3593-3599 (2015). Zhang, R.G. et al. TEsorter: an accurate and fast method to classify LTR-retrotransposons in plant genomes. Hortic. Res. 9 , uhac017 (2022). Cantarel, B.L. et al. MAKER: an easy-to-use annotation pipeline designed for emerging model organism genomes. Genome Res. 18 , 188-196 (2008). Pertea, M., et al. StringTie enables improved reconstruction of a transcriptome from RNA-seq reads. Nat. Biotechnol. 33 , 290-295 (2015). Slater, G.S., & Birney, E. Automated generation of heuristics for biological sequence comparison. BMC Bioinformatics 6 , 31 (2005). Eilbeck, K., Moore, B., Holt, C., & Yandell, M. Quantitative measures for the management and comparison of annotated genomes. BMC Bioinformatics 10 , 67 (2009). Korf, I. Gene finding in novel genomes. BMC Bioinformatics 5 , 59 (2004). Brůna, T., Hoff, K.J., Lomsadze, A., Stanke, M., & Borodovsky, M. BRAKER2: automatic eukaryotic genome annotation with GeneMark-EP+ and AUGUSTUS supported by a protein database. NAR Genom. Bioinform. 3 , lqaa108 (2021). Kim, D., Langmead, B., & Salzberg, S.L. HISAT: a fast spliced aligner with low memory requirements. Nat. Methods 12 , 357-360 (2015). Lomsadze, A., Ter-Hovhannisyan, V., Chernoff, Y.O., & Borodovsky, M. Gene identification in novel eukaryotic genomes by self-training algorithm. Nucleic Acids Res. 33 , 6494-6506 (2005). Stanke, M., Tzvetkova, A., & Morgenstern, B. AUGUSTUS at EGASP: using EST, protein and genomic alignments for improved gene prediction in the human genome. Genome Biol. 7 , S11 (2006). Shumate, A., & Salzberg, S.L. Liftoff: accurate mapping of gene annotations. Bioinformatics 37 , 1639-1643 (2021). Pertea, G., & Pertea, M. GFF Utilities: GffRead and GffCompare. F1000Res 9 . (2020). Capella-Gutiérrez, S., Silla-Martínez, J.M., & Gabaldón, T. trimAl: a tool for automated alignment trimming in large-scale phylogenetic analyses. Bioinformatics 25 , 1972-1973 (2009). Stamatakis, A. RAxML version 8: a tool for phylogenetic analysis and post-analysis of large phylogenies. Bioinformatics 30 , 1312-1313 (2014). Yang, Z. PAML 4: phylogenetic analysis by maximum likelihood. Mol. Biol. Evol. 24 , 1586-1591 (2007). Mendes, F.K., Vanderpool, D., Fulton, B., & Hahn, M.W. CAFE 5 models variation in evolutionary rates among gene families. Bioinformatics 36 , 5516-5518 (2021). Tang, H., Krishnakumar, V., & Li, J. jcvi: JCVI utility libraries. Zenodo. (2015). Krumsiek, J., Arnold, R., & Rattei, T. Gepard: a rapid and sensitive tool for creating dotplots on genome scale. Bioinformatics 23 , 1026-1028 (2007). Zhang, Z. et al. ParaAT: a parallel tool for constructing multiple protein-coding DNA alignments. Biochem. Biophys. Res. Commun. 419 , 779-781 (2012). Eisfeldt, J., Nilsson, D., Andersson-Assarsson, J. C., & Lindstrand, A. AMYCNE: Confident copy number assessment using whole genome sequencing data. PLoS One 13 , e0189710 (2018). Servant, N. et al. HiTC: exploration of high-throughput ‘C’ experiments. Bioinformatics 28 , 2843-2844 (2012). Li, H. Aligning sequence reads, clone sequences and assembly contigs with BWA-MEM. arXiv 1303 , 3997 (2013). Zhang, Y.et al. Model-based analysis of ChIP-Seq (MACS). Genome Biol. 9 , R137. (2008). Langmead, B., & Salzberg, S.L. Fast gapped-read alignment with Bowtie 2. Nat. Methods 9 , 357-359 (2012). Vollger, M.R. et al. StainedGlass: Interactive visualization of massive tandem repeat structures with identity heatmaps. Bioinformatics 38 , 2049-2051 (2022). Krueger, F., & Andrews, S.R. Bismark: a flexible aligner and methylation caller for Bisulfite-Seq applications. Bioinformatics 27 , 1571-1572 (2011). Bray, N.L., Pimentel, H., Melsted, P., & Pachter, L. Near-optimal probabilistic RNA-seq quantification. Nat. Biotechnol. 34 , 525-527 (2016). Daood, H.G. et al. HPLC determination of capsaicinoids with cross-linked C18 column and buffer-free eluent. J. Chromatogr. Sci. 53 , 135-143 (2015). Table Table 1 | Statistics for genome assembly and annotation of two pepper species Genomic feature Capsicum annuum Capsicum rhomboideum Number of contigs 12 13 Total length (bp) 3,103,116,129 1,707,653,203 Contig N50 (bp) 262,573,928 145,987,823 Number of gaps 0 0 Number of telomeres 24 17 Number of centromeres 12 13 Number of gene models 34,428 33,512 GC content (%) 35.00 36.31 Repeat content (%) 79.50 74.64 Assembly BUSCOs (%) 98.62 97.12 Annotation BUSCOs (%) 97.04 93.23 QV 56.60 77.18 Completeness (%) 96.49 98.11 Additional Declarations There is NO Competing Interest. Supplementary Files PepperT2TSupplementaryTables.xlsx Supplementary Tables Supplementaryfigures.docx Cite Share Download PDF Status: Published Journal Publication published 20 May, 2024 Read the published version in Nature Communications → Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-3062426","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":273731429,"identity":"fff18b50-f379-4484-af29-38817ada15f5","order_by":0,"name":"Li Guo","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA7ElEQVRIiWNgGAWjYBACPmYwZcMgASQPQMQS8Gthg2hJI0ULhDoM1sJAnBZ2BubXPBXn8yRnJDAe+FFTx8DPnmPA8HMHXoexWfOcuV0sLZHAcLDn2GEGyZ43Boy9Z/BrMeZtu504TyL/wwEetgMMBjdyDJgZ2whqOQfUArTlz786BnsitDA/5m07kDgbqOUwbxszg4EEQS2MbYxzziQXS/Y8YDgs23eYR+LMs4KDvXi08PMfPvzhTYVdnsTxBOaPb77VyfG3J2988BOPFgYGxjZQlCTAuDwg4gA+DUDA/IGBYPSNglEwCkbBiAYAKvZIU/h6KrcAAAAASUVORK5CYII=","orcid":"https://orcid.org/0000-0001-6100-3481","institution":"Peking University","correspondingAuthor":true,"prefix":"","firstName":"Li","middleName":"","lastName":"Guo","suffix":""},{"id":273731430,"identity":"88febaac-b4c9-4d9f-a861-0faaf9980b84","order_by":1,"name":"Weikai Chen","email":"","orcid":"","institution":"Peking University Institute of Advanced Agricultural Sciences","correspondingAuthor":false,"prefix":"","firstName":"Weikai","middleName":"","lastName":"Chen","suffix":""},{"id":273731431,"identity":"771da0b0-289b-4cc0-8ec5-65bfb1b6b4d2","order_by":2,"name":"Xiangfeng Wang","email":"","orcid":"","institution":"Peking University Institute of Advanced Agricultural Sciences","correspondingAuthor":false,"prefix":"","firstName":"Xiangfeng","middleName":"","lastName":"Wang","suffix":""},{"id":273731432,"identity":"83093b89-4728-48bc-8675-b3db6e6c1903","order_by":3,"name":"Jie Sun","email":"","orcid":"","institution":"Peking University Institute of Advanced Agricultural Sciences","correspondingAuthor":false,"prefix":"","firstName":"Jie","middleName":"","lastName":"Sun","suffix":""},{"id":273731433,"identity":"9ce936ef-cfb9-459f-b73c-92c2bc79c53b","order_by":4,"name":"Xinrui Wang","email":"","orcid":"","institution":"Peking University Institute of Advanced Agricultural Sciences","correspondingAuthor":false,"prefix":"","firstName":"Xinrui","middleName":"","lastName":"Wang","suffix":""},{"id":273731434,"identity":"ffe11d68-a79e-4d7b-87bd-354eaa07ec92","order_by":5,"name":"Zhangsheng Zhu","email":"","orcid":"","institution":"South China Agricultural University","correspondingAuthor":false,"prefix":"","firstName":"Zhangsheng","middleName":"","lastName":"Zhu","suffix":""},{"id":273731435,"identity":"792efff7-6794-4898-862a-4ee7922b5f33","order_by":6,"name":"Dilay Hazal Ayhan","email":"","orcid":"https://orcid.org/0000-0002-0497-3294","institution":"Peking University Institute of Advanced Agricultural Sciences","correspondingAuthor":false,"prefix":"","firstName":"Dilay","middleName":"Hazal","lastName":"Ayhan","suffix":""},{"id":273731436,"identity":"33a16d8b-cd66-4ef2-b14b-f73573657398","order_by":7,"name":"Shu Yi","email":"","orcid":"","institution":"Peking University Institute of Advanced Agricultural Sciences","correspondingAuthor":false,"prefix":"","firstName":"Shu","middleName":"","lastName":"Yi","suffix":""},{"id":273731437,"identity":"750b4f91-cb5b-44f1-a220-840cffbb2fd6","order_by":8,"name":"Ming Yan","email":"","orcid":"","institution":"Peking University Institute of Advanced Agricultural Sciences","correspondingAuthor":false,"prefix":"","firstName":"Ming","middleName":"","lastName":"Yan","suffix":""},{"id":273731438,"identity":"8d658053-df91-4ff3-8aab-b22e4b5ec766","order_by":9,"name":"Lili Zhang","email":"","orcid":"","institution":"Weifang Institute of Technology","correspondingAuthor":false,"prefix":"","firstName":"Lili","middleName":"","lastName":"Zhang","suffix":""},{"id":273731439,"identity":"4c34875e-c577-4fd6-b0ed-859812e87bf7","order_by":10,"name":"Tan Meng","email":"","orcid":"","institution":"Peking University Institute of Advanced Agricultural Sciences","correspondingAuthor":false,"prefix":"","firstName":"Tan","middleName":"","lastName":"Meng","suffix":""},{"id":273731440,"identity":"e41e5b3f-8499-43e3-8511-6e3a2fe53d85","order_by":11,"name":"Yu Mu","email":"","orcid":"","institution":"Peking University Institute of Advanced Agricultural Sciences","correspondingAuthor":false,"prefix":"","firstName":"Yu","middleName":"","lastName":"Mu","suffix":""},{"id":273731441,"identity":"37dd54f3-b107-40fc-8c2b-ebeee3d88f7f","order_by":12,"name":"Jun Li","email":"","orcid":"","institution":"Peking University Institute of Advanced Agricultural Sciences","correspondingAuthor":false,"prefix":"","firstName":"Jun","middleName":"","lastName":"Li","suffix":""},{"id":273731442,"identity":"06e8a637-333f-4009-8862-9cbbd84b6858","order_by":13,"name":"Dian Meng","email":"","orcid":"","institution":"Peking University Institute of Advanced Agricultural Sciences","correspondingAuthor":false,"prefix":"","firstName":"Dian","middleName":"","lastName":"Meng","suffix":""},{"id":273731443,"identity":"8c8fe3d8-4b2d-4598-8d3c-141604e30f6e","order_by":14,"name":"Jianxin Bian","email":"","orcid":"","institution":"Peking University Institute of Advanced Agricultural Sciences","correspondingAuthor":false,"prefix":"","firstName":"Jianxin","middleName":"","lastName":"Bian","suffix":""},{"id":273731444,"identity":"b01c10a6-e414-4d29-82f7-e918306b5719","order_by":15,"name":"Ke Wang","email":"","orcid":"","institution":"Shandong Agriculture University","correspondingAuthor":false,"prefix":"","firstName":"Ke","middleName":"","lastName":"Wang","suffix":""},{"id":273731445,"identity":"82755c9f-1b0a-41d8-9496-2ec249aee1e9","order_by":16,"name":"Lu Wang","email":"","orcid":"","institution":"Peking University Institute of Advanced Agricultural Sciences","correspondingAuthor":false,"prefix":"","firstName":"Lu","middleName":"","lastName":"Wang","suffix":""},{"id":273731446,"identity":"f0dff3e3-6c09-40c4-b7f8-84386e6341bd","order_by":17,"name":"Shaoying Chen","email":"","orcid":"","institution":"Peking University Institute of Advanced Agricultural Sciences","correspondingAuthor":false,"prefix":"","firstName":"Shaoying","middleName":"","lastName":"Chen","suffix":""},{"id":273731447,"identity":"acc78d6c-6608-4180-a087-0fb09e54b2fa","order_by":18,"name":"Ruidong Chen","email":"","orcid":"","institution":"Peking University Institute of Advanced Agricultural Sciences","correspondingAuthor":false,"prefix":"","firstName":"Ruidong","middleName":"","lastName":"Chen","suffix":""},{"id":273731448,"identity":"78e9f4e5-3d72-4770-93a1-173f2a9c1214","order_by":19,"name":"Jingyun Jin","email":"","orcid":"","institution":"Peking University Institute of Advanced Agricultural Sciences","correspondingAuthor":false,"prefix":"","firstName":"Jingyun","middleName":"","lastName":"Jin","suffix":""},{"id":273731449,"identity":"d62f6f51-c89a-43af-88a9-73ac800d8ffe","order_by":20,"name":"Bosheng Li","email":"","orcid":"","institution":"Peking university Institute of Advanced Agricultural Science","correspondingAuthor":false,"prefix":"","firstName":"Bosheng","middleName":"","lastName":"Li","suffix":""},{"id":273731450,"identity":"5853d6fb-802c-401e-b204-48bd2a001ab5","order_by":21,"name":"Xingping Zhang","email":"","orcid":"","institution":"Institute of Advanced Agriculture Sciences, Peking University, Weifang, China","correspondingAuthor":false,"prefix":"","firstName":"Xingping","middleName":"","lastName":"Zhang","suffix":""},{"id":273731451,"identity":"fee636fa-f984-43cd-b2a6-12c93fefe2f8","order_by":22,"name":"Xingwang Deng","email":"","orcid":"https://orcid.org/0000-0001-8709-1467","institution":"Peking University","correspondingAuthor":false,"prefix":"","firstName":"Xingwang","middleName":"","lastName":"Deng","suffix":""},{"id":273731452,"identity":"7ca01354-8764-42f5-99f8-69c8dba0a051","order_by":23,"name":"Hang He","email":"","orcid":"https://orcid.org/0000-0003-3165-283X","institution":"Peking University","correspondingAuthor":false,"prefix":"","firstName":"Hang","middleName":"","lastName":"He","suffix":""}],"badges":[],"createdAt":"2023-06-14 10:31:19","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-3062426/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-3062426/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41467-024-48643-0","type":"published","date":"2024-05-20T04:00:00+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":51933722,"identity":"7262c847-b78d-4133-bb73-c112774f6a5b","added_by":"auto","created_at":"2024-03-04 06:22:00","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":2967758,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eT2T gap-free assembly of two \u003c/strong\u003e\u003cem\u003e\u003cstrong\u003eCapsicum \u003c/strong\u003e\u003c/em\u003e\u003cstrong\u003egenomes\u003c/strong\u003e. \u003cstrong\u003eA.\u003c/strong\u003eCircos plot showing the T2T genome assemblies for \u003cem\u003eC. annuum\u003c/em\u003e (CaT2T) and \u003cem\u003eC\u003c/em\u003e. \u003cem\u003erhomboideum \u003c/em\u003e(CrT2T)\u003cem\u003e. \u003c/em\u003eTrack a to g: chromosomes (Red: Ca, Blue: Cr), GC content, gene density, TE density, LTR/\u003cem\u003eGypsy\u003c/em\u003edensity, LTR/\u003cem\u003eCopia\u003c/em\u003e density and color ribbons representing genomewide syntenic blocks. Centromere regions (black) were denoted on chromosome track. \u003cstrong\u003eB.\u003c/strong\u003eA bubble plot highlights the key statistics of \u003cem\u003eCapsicum \u003c/em\u003egenome assemblies in this study (CaT2T and CrT2T) and reported previously. \u003cstrong\u003eC.\u003c/strong\u003e A chromosome ideogram of CaT2T genome assembly showing the mapping of filled gaps, candidate capsaicin biosynthesis genes (CBG), centromeres and telomeres, with a heatmap of gene density overlaid on each chromosome. \u003cstrong\u003eD.\u003c/strong\u003eReconstruction of the rearrangement (fissions and fusions) events between CaT2T and CrT2T. Synteny blocks were colored by MCScanX. The red rectangles denote the centromere regions.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-3062426/v1/ef0bbb7ee704deb36c7279cd.png"},{"id":51933723,"identity":"46c6a314-1a61-46ee-bf01-7f99c2c50f22","added_by":"auto","created_at":"2024-03-04 06:22:00","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":1194269,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eGenome structure of the centromeric regions in \u003c/strong\u003e\u003cem\u003e\u003cstrong\u003eC. annuum\u003c/strong\u003e\u003c/em\u003e\u003cstrong\u003e genome. A. \u003c/strong\u003eSchematic representation showing the distribution of different transposable elements across 12 centromeres in CaT2T. The CENH3 ChIP-seq signals (average of two biological replicates) were represented by the enrichment level in 30 kb windows and the ChIP-identified centromeres were marked by grey boxes. The red lines for the track of CRM indicated the intact CRM located in the centromeres. \u003cstrong\u003eB.\u003c/strong\u003e Specific LTR insertion time distributions of the whole genomes and centromeres. The center white line and black line in the plot indicate the median, the upper and lower quartiles for insertion times respectively. Significant difference between groups was assessed using the two-sided Wilcoxon rank-sum test. \u003cstrong\u003eC.\u003c/strong\u003eSpecific ratio of solo LTRs to intact LTR-RTs distributions of the whole genomes and centromeres. Significant difference between groups was assessed using the two-sided Wilcoxon rank-sum test. \u003cstrong\u003eD.\u003c/strong\u003e Neighbor-joining trees constructed from 9,712 intact Ty3-\u003cem\u003eGypsy \u003c/em\u003eelements from \u003cem\u003eC. annuum\u003c/em\u003e, \u003cem\u003eC. rhomboideum\u003c/em\u003e and \u003cem\u003eS. tuberosum\u003c/em\u003e. The red, green and yellow solid circle in the branches represent the elements located in centromere regions. \u003cstrong\u003eE.\u003c/strong\u003eAn example of pairwise sequence identity of non-overlapping 5 kb centromeric regions in \u003cem\u003eC. annuum \u003c/em\u003eand \u003cem\u003eC. rhomboideum\u003c/em\u003e.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-3062426/v1/af6967679b84cc7f2bc02d01.png"},{"id":51933725,"identity":"822b286d-b9f5-4270-a471-a7be21be34b2","added_by":"auto","created_at":"2024-03-04 06:22:00","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":809701,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eEpigenomic and transcriptional features of difficult-to-access regions in \u003c/strong\u003e\u003cem\u003e\u003cstrong\u003eCapsicum annuum \u003c/strong\u003e\u003c/em\u003e\u003cstrong\u003eT2T genome. A. \u003c/strong\u003eCharacteristics of the centromeres on \u003cem\u003eC. annuum \u003c/em\u003eChr07 as an example. The distributions of TAD (topological associated domains) and chromatin loops, A/B compartment, ChIP-seq signal (H3K27me3, red; H3K9me2, green; H3K4me3, purple), methylation level (CG, red; CHG, green; CHH, purple), ATAC-seq signal, TE elements (Athila, bule; CRM, yellow; Tekay, red; others, grey), genes (green), TE transcript abundances (blue) and gene transcript abundances (red) were plotted from the top to bottom successively. \u003cstrong\u003eB. \u003c/strong\u003eEpigenetic and transcriptional landscape of telomeres on \u003cem\u003eC. annuum \u003c/em\u003eChr01 as an example. \u003cstrong\u003eC. \u003c/strong\u003eEpigenetic signals detected at or in proximity to centromeres with TE, GC and gene density shown. \u003cstrong\u003eD.\u003c/strong\u003e CENH3 and Histone modification ChIP-seq signals at or in proximity to centromeres. \u003cstrong\u003eE.\u003c/strong\u003e The distribution of CG, CHG, and CHH methylation in Athila, CRM and Tekay regions including the upstream and downstream 10 kb of \u003cem\u003eC. annuum\u003c/em\u003e.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-3062426/v1/82cafc6e5ec12a1eeb53b096.png"},{"id":51933729,"identity":"3f09a869-74ad-4f38-b214-42ebf05be8e2","added_by":"auto","created_at":"2024-03-04 06:22:00","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":352335,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eEvolution of capsaicin biosynthesis genes and their tissue-specificity. (A). \u003c/strong\u003ePhylogenomic analysis of \u003cem\u003eC. annuum\u003c/em\u003e and related angiosperm species. Whole genome duplication (WGD) or triplication (WGT) events, and gene family expansion/contraction statistics were marked on the MCMC phylogenetic tree constructed using single-copy orthologs. MRCA represents most recent common ancestor. (\u003cstrong\u003eB\u003c/strong\u003e). A pannel alongside the species is a summary of abundance in gene family members related to biosynthesis of capsaicin under a phylogenomic context. (\u003cstrong\u003eC\u003c/strong\u003e) Microsynteny relationships of capsaisin synthase (CS) gene and its tandem copies present in a syntenic block are conserved among \u003cem\u003eSolanum lycopersicum \u003c/em\u003e(Sl), \u003cem\u003ePhysalis pubescens\u003c/em\u003e(Pp), \u003cem\u003eC. annuum\u003c/em\u003e (Ca), \u003cem\u003eC. rhomboideum\u003c/em\u003e (Cr), \u003cem\u003eC. baccatum\u003c/em\u003e(Cb), \u003cem\u003eC. chinense\u003c/em\u003e (Cc) and \u003cem\u003eC. pubescens \u003c/em\u003e(Cp). The red lines indicate the closest homologs of key \u003cem\u003eCS\u003c/em\u003e gene. (\u003cstrong\u003eD\u003c/strong\u003e) Left: diagram of capsaicinoid biosynthesis pathways and key genes. PAL, phenylalanine ammonia-lyase; C4H, cinnamate 4-hydroxylase; 4CL, 4-coumarate: CoA ligase; HCT, hydroxycinnamoyl transferase; CCoAOMT, caffeoyl-CoA 3-\u003cem\u003eO\u003c/em\u003e-methyltransferase; C3H, coumarate 3-hydroxylase; HCHL, hydroxyl cinnamyl-CoA hydrase/lyase; AMT, aminotransferase; BCAT, branched chain amino acid aminotransferase; BCKDH, branched-chain α-ketoacid dehydrogenase; Kas, β-ketoacyl-ACP synthase; ACL, acyl carrier protein; FatA, acyl-ACP-thiesterase; ACS, acetyl-CoA synthetase; and CS, capsaicin synthase. Right: transcriptional expression heatmap of capsaicin biosynthesis genes and their homologous genes in different tissues of five Solanaceae plants. (\u003cstrong\u003eE\u003c/strong\u003e) Dotplot of \u003cem\u003eCS\u003c/em\u003e nucleotide sequence from seven Solanaceae plants. The plotted sequence includes full coding sequences of \u003cem\u003eCS\u003c/em\u003eand tandem repeats as well as their 2 kb flanking sequences. (\u003cstrong\u003eF\u003c/strong\u003e) Epigenomic and transcriptomic patterns of two \u003cem\u003eCS\u003c/em\u003e genes and flanking regions. (\u003cstrong\u003eG\u003c/strong\u003e) Common transcription factor binding motifs enriched in both open chromatin regions (ATAC-seq) and upstream (2 kb) sequence of seven capsaicin biosynthesis genes.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-3062426/v1/f7a4ff104fa73b1563b5912e.png"},{"id":51933986,"identity":"5e215e79-ec01-4103-a53e-314242c8931c","added_by":"auto","created_at":"2024-03-04 06:30:00","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":1497877,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eMachine-learning model for fruit pungency prediction from capsaicin biosynthesis gene copy number variants. \u003c/strong\u003e(\u003cstrong\u003eA\u003c/strong\u003e) Copy number variations of candidate CBGs in 320 \u003cem\u003eC. annuum\u003c/em\u003e accessions including nine new resequencing samples. Black: absent, white: low-copy, red: high-copy. (B) Machine-learning models to capture the quantitative relationship between CNVs and capsaicin levels. (C) Evaluation of machine-learning models using ROC curves, with random forest model showing the highest AUC (Area Under Curve) for predicting fruit pungency from gene copy number variants. (D) Top-ranked CBGs predicted by random forest model with gene names shown in a word cloud where word sizes (small to large) are proportional to gene ranking (low to high).\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-3062426/v1/f62e43299714f15c7c921cd3.png"},{"id":51933726,"identity":"c294604e-0fc8-49ea-ad8f-dfe294fc4535","added_by":"auto","created_at":"2024-03-04 06:22:00","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":463865,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eThe model for the evolution of capsaicin synthase (CS) in pungent and non-pungent Solanaceae plants.\u003c/strong\u003e \u003cstrong\u003eA\u003c/strong\u003e. Pungent species \u003cem\u003eCapsicum annuum\u003c/em\u003e (CaT2T), \u003cem\u003eC. baccatum\u003c/em\u003e and \u003cem\u003eC. chinense \u003c/em\u003eall have at least one functional \u003cem\u003eCS\u003c/em\u003e gene. Besides, \u003cem\u003eC. annuum \u003c/em\u003eand \u003cem\u003eC. chinense\u003c/em\u003e have several tandem duplicates of CS genes (formed between 13.38 Mya and 4.8 Mya), most of which are partial gene paralogs and unfunctional. Mya: million years ago. CRE: \u003cem\u003ecis\u003c/em\u003e-regulatory elements. \u003cstrong\u003eB.\u003c/strong\u003eBy contrast, non-pungent species tomato (\u003cem\u003eS. lycopersicum\u003c/em\u003e) and ground berry (\u003cem\u003eP. pubescens\u003c/em\u003e) and \u003cem\u003eC. rhomboideum\u003c/em\u003e (CrT2T) all lack a functional \u003cem\u003eCS\u003c/em\u003e gene, but have various numbers of \u003cem\u003eCS\u003c/em\u003e tandem paralogs that include both partial \u003cem\u003eCS\u003c/em\u003e or non-expressed \u003cem\u003eCS\u003c/em\u003e due to deletion in CREs (formed between 19.28Mya and 13.38Mya). \u003cstrong\u003eC.\u003c/strong\u003e Pungent pepper achieves tissue-specific regulation of capsaicin production through placenta-specific opening of chromatin region around \u003cem\u003eCS\u003c/em\u003e gene (\u003cem\u003eCS2\u003c/em\u003e) identified by ATAC-seq, allowing its transcription (RNA-seq). By contrast, nonpungent species lack capsaicin biosynthesis due to missing either functional copy of \u003cem\u003eCS\u003c/em\u003e genes or corresponding CREs.\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-3062426/v1/904b29dfd7be838e66247bae.png"},{"id":56842111,"identity":"cfe237eb-cda6-4190-8ba5-ac8879999f1e","added_by":"auto","created_at":"2024-05-21 07:07:48","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":8521238,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-3062426/v1/9d4061fe-899e-4625-8174-14e4b5614b67.pdf"},{"id":51933724,"identity":"7ef11776-8eba-4019-b225-66c3ffa1bba4","added_by":"auto","created_at":"2024-03-04 06:22:00","extension":"xlsx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":291923,"visible":true,"origin":"","legend":"Supplementary Tables","description":"","filename":"PepperT2TSupplementaryTables.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-3062426/v1/244872c5e88a2c887982bbe6.xlsx"},{"id":51933731,"identity":"d9e70160-51fc-43fb-a3aa-a5b8b4ee5506","added_by":"auto","created_at":"2024-03-04 06:22:01","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":21990317,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cbr\u003e\u003c/p\u003e","description":"","filename":"Supplementaryfigures.docx","url":"https://assets-eu.researchsquare.com/files/rs-3062426/v1/77c40e8fc3691a06ed6a5fe2.docx"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"Two telomere-to-telomere gapless genomes reveal insights into \u003ci\u003eCapsicum\u003c/i\u003e evolution and capsaicinoid biosynthesis","fulltext":[{"header":"Introduction","content":"\u003cp\u003eChili pepper (\u003cem\u003eCapsicum annuum\u003c/em\u003e), a member of the nightshade family (Solanaceae), is a worldwide cultivated vegetable and spice crop, bearing fruits best known for their fruit pungency given by capsaicinoids. Capsaicinoids are alkaloids synthesized via the convergence of the phenylpropanoid pathway and branched-chain fatty acid pathway, finished with condensation by capsaicin synthase (CS)\u003csup\u003e1\u003c/sup\u003e, although the full biosynthesis pathway is yet elucidated.The birth and death of capsaicinoid biosynthesis in the nightshade family remains a poorly understood question for plant evolutionary biologists. The biosynthesis of capsaicinoids occurs in the fruit placenta tissue of chili pepper, a unique trait of \u003cem\u003eCapsicum\u003c/em\u003e spp. and missing in other Solanaceae plants such as tomato\u003csup\u003e2\u003c/sup\u003e. However, fruit pungency is not universally present across \u003cem\u003eCapsicum\u003c/em\u003e, as nonpungency is found in several cultivars and wild relatives (e.g. \u003cem\u003eC. rhomboideum\u003c/em\u003e) of chili peppers\u003csup\u003e3\u003c/sup\u003e. How fruit pungency arose and lost in the \u003cem\u003eCapsicum\u0026nbsp;\u003c/em\u003espp. and how capsaicinoids are exclusively produced in fruits remains poorly understood, due to the lack of high-quality genome resources across the \u003cem\u003eCapsicum\u0026nbsp;\u003c/em\u003egenus, especially non-pungent members.\u003c/p\u003e\n\u003cp\u003eThe first draft genome assembly of \u003cem\u003eC. annuum\u003c/em\u003e (cultivar CM334) was released in 2014 sized 3.06 Gb with contig N50 of 30 kb\u003csup\u003e2\u003c/sup\u003e. Afterward, several assemblies showed improvement with the development of\u0026nbsp;single-molecule DNA sequencing technologies, for example, cultivar ‘59’ (3.07 Gb, contig N50: 41.27 Mb)\u003csup\u003e4\u003c/sup\u003e, ‘Takanotsume’ (3.05 Gb, contig N50: 99.05 Mb)\u003csup\u003e5\u003c/sup\u003e and CC-090 (3.06 Gb,\u0026nbsp;contig N50: 187.09 Mb)\u003csup\u003e6\u003c/sup\u003e. To date, 23 genome assemblies of cultivated peppers including \u003cem\u003eC. annuum, C. baccatum\u003c/em\u003e, \u003cem\u003eC. chinense\u003c/em\u003e,and \u003cem\u003eC. pubescens\u003c/em\u003e are publicly available\u003csup\u003e7-11\u003c/sup\u003e, whereas genome sequences for wild peppers are very scarce\u003csup\u003e7\u003c/sup\u003e. Also, despite continuous improvement, the published assemblies still contain numerous gaps and assembly errors, missing complete centromeres and telomeres. Pepper genomes are repeat-rich (~80%), making genome assembly extremely challenging. Assembly gaps and errors often lead to mis-annotation of genes and false discovery of genetic variants, confounding the pepper functional genomic research. Therefore, generating telomere-to-telomere (T2T) gapless and accurately annotated genome sequences is paramount to precision genetic improvement of peppers and dissecting full biosynthetic pathways for capsaicinoids and other valuable natural products.\u003c/p\u003e\n\u003cp\u003eAccurate assembly of complete genome sequences remains a daunting task for eukaryotes with large and complex genomes with extensive repeats, high heterozygosity, or polyploidy. A breakthrough was recently achieved by the human T2T consortium in assembling the complete human genome sequence, a milestone in human genomics\u003csup\u003e12\u003c/sup\u003e. It has revolutionized the analysis of human genomic variants, and epigenetic and transcriptional signatures in centromeres\u003csup\u003e13-15\u003c/sup\u003e. Nearly two decades after the first available plant genome sequence, T2T gap-free genome assemblies were recently reported for Arabidopsis\u003csup\u003e16-18\u003c/sup\u003e, rice\u003csup\u003e19-20\u003c/sup\u003e, potato\u003csup\u003e21\u003c/sup\u003e and soybean\u003csup\u003e22\u003c/sup\u003e. However, these are considered near-complete genomes with either minor gaps in difficult-to-assemble regions or the omission of a few telomeres or centromeres typically containing high copies of tandem repeats. Furthermore, these published plant T2T genomes are relatively small in size (Arabidopsis 134 Mb\u003csup\u003e16\u003c/sup\u003e,\u0026nbsp;rice 385 Mb\u003csup\u003e20\u003c/sup\u003e, potato 773 Mb\u003csup\u003e21\u003c/sup\u003e, and soybean 1.01 Gb\u003csup\u003e22\u003c/sup\u003e). Until recently, a complete genome assembly was reported for maize (2.10 Gb)\u003csup\u003e23\u003c/sup\u003e. However, complete genomes remain rare for large complex plant genomes which are notoriously more difficult to assemble.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eHere, we \u003cem\u003ede novo\u003c/em\u003e assembled and annotated two T2T gapless \u003cem\u003eCapsicum\u0026nbsp;\u003c/em\u003egenome sequences including a pungent pepper \u003cem\u003eC. annuum\u0026nbsp;\u003c/em\u003eand its non-pungent wild relative \u003cem\u003eC. rhomboideum\u003c/em\u003e, which marks a new milepost for pepper genome research. The in-depth analysis of two T2T genomes revealed distinctive structural, epigenetic, and transcriptional features in their centromeres. Evolutionary insight into capsaicin biosynthesis pathway and regulation were provided using phylogenomics and epigenomic data analysis. Our study provides timely genomic resources to accelerate pepper research and genetic improvement.\u003c/p\u003e"},{"header":"Results","content":"\u003ch2\u003eT2T gapless \u003cem\u003eCapsicum\u0026nbsp;\u003c/em\u003egenome assemblies\u003c/h2\u003e\n\u003cp\u003eToassemble T2T gap-free genomes for pungent \u003cem\u003eC. annuum\u0026nbsp;\u003c/em\u003eand nonpungent \u003cem\u003eC.\u003c/em\u003e \u003cem\u003erhomboideum\u003c/em\u003e, we generated high-coverage PacBio\u0026nbsp;HiFi reads, Oxford Nanopore Technology (ONT) ultra-long reads, Illumina paired-end (NGS) reads and high-throughput chromatin conformation capture (Hi-C) sequencing reads for\u0026nbsp;\u003cem\u003eC\u003c/em\u003e\u003cem\u003e. annuum\u003c/em\u003e double haploid cultivar G1-36576 and \u003cem\u003eC.\u003c/em\u003e \u003cem\u003erhomboideum\u003c/em\u003e wild accession PI\u0026nbsp;645680 (Supplementary Fig. 1\u0026nbsp;and Supplementary Table 1). The genome assembly was performed using an in-house pipeline integrating various computational tools to maximize the strength of various data types (Supplementary Fig. 2\u0026nbsp;and Supplementary Table 2). Briefly, HiFi and ONT reads were first separately assembled using \u003cem\u003ehifiasm\u003c/em\u003e\u003csup\u003e24\u003c/sup\u003eand \u003cem\u003eNextDenovo\u003c/em\u003e\u003csup\u003e25\u003c/sup\u003e, respectively. The\u0026nbsp;\u003cem\u003eC. annuum\u0026nbsp;\u003c/em\u003eHiFi-based assembly sized 3.13 Gb with contig N50 of 262.4 Mb, and contained 18 telomeres with thousands of copies of telomeric repeat units (TRUs) identified at one or both ends of 12 contigs, of which six were T2T. Assembling ONT reads produced a 3.10 Gb draft assembly with contig N50 of 177.8 Mb\u003cem\u003e,\u0026nbsp;\u003c/em\u003econtaining\u0026nbsp;22 telomeres with over 10,000 copies of TRUs, of which four were nearly T2T. ONT assembly was then used to fill the gaps (Supplementary Table 3) and patch the telomeres of the HiFi assembly, yielding a hybrid assembly including 12 gapless chromosome-level contigs with 22 telomeres plus contigs containing 45S rDNA arrays. The\u0026nbsp;nucleolus organizer regions (NOR)\u0026nbsp;were separately assembled using 45S rDNA-containing HiFi reads, and the contigs were assembled to a single sequence based on the specific \u003cem\u003ek\u003c/em\u003e-mer (Supplementary Fig. 3). To ensure accuracy, the sequences of ONT origin were replaced with their corresponding HiFi assembled contigs, followed by Hi-C scaffolding to 12 chromosomes (\u003cstrong\u003eFig.\u003c/strong\u003e\u003cstrong\u003e1A\u003c/strong\u003e\u003cstrong\u003e)\u003c/strong\u003e and manual correction for mis-assemblies in Juicebox\u003csup\u003e26\u003c/sup\u003e. After addition of the rDNA arrays and telomere patching, the final T2T gapless assembly of \u003cem\u003eC. annuum\u003c/em\u003e genome (CaT2T) was 3.1Gb with a contig N50 of 262.6Mb\u0026nbsp;(\u003cstrong\u003eTable\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e1\u003c/strong\u003e),\u0026nbsp;closing all 503 gaps of Ca59 (\u003cstrong\u003eFig.\u003c/strong\u003e\u003cstrong\u003e1B)\u003c/strong\u003e and representing the first complete \u003cem\u003eCapsicum\u0026nbsp;\u003c/em\u003egenome assembly\u0026nbsp;(\u003cstrong\u003eFig.\u003c/strong\u003e\u003cstrong\u003e1C\u003c/strong\u003e) and largest complete genome sequence. Using the same approach, we assembled a\u0026nbsp;1.70 Gb T2T\u0026nbsp;gap-free genome sequence (CrT2T) for \u003cem\u003eC.\u0026nbsp;\u003c/em\u003e\u003cem\u003erhomboideum\u0026nbsp;\u003c/em\u003econtaining 13 chromosomes with a contig N50 of 146.0 Mb (\u003cstrong\u003eFig.\u003c/strong\u003e\u003cstrong\u003e1A\u003c/strong\u003e\u003cstrong\u003e; Table\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e1\u003c/strong\u003e), the first\u0026nbsp;gap-free\u0026nbsp;genome for non-domesticated \u003cem\u003eCapsicum\u003c/em\u003e (\u003cstrong\u003eFig.\u003c/strong\u003e\u003cstrong\u003e1C\u003c/strong\u003e). We identified the complete set (24/24) of telomeres in \u003cem\u003eC. annuum\u003c/em\u003e (Supplementary Fig. 4) and majority of (17/26) telomeres in \u003cem\u003eC. rhomboideum\u003c/em\u003e (Supplementary Fig. 5). Synteny analysis revealed that 45.07% of \u003cem\u003eC. rhomboideum\u0026nbsp;\u003c/em\u003egenes are syntenic to \u003cem\u003eC. annuum\u0026nbsp;\u003c/em\u003e(Supplementary Table 4), but their whole-genome alignment showed low sequence identity, suggesting substantial divergence.\u0026nbsp;Comparing the two genomes, we demonstrated that it required at least 10 fissions and 11 fusions of chromosomes to evolve the \u003cem\u003eC. annuum\u003c/em\u003e karyotype from that of \u003cem\u003eC. rhomboideum\u003c/em\u003e (Supplementary Fig. 6\u003cstrong\u003e; Fig.1D\u003c/strong\u003e).\u003c/p\u003e\n\u003ch2\u003eGenome validation and annotation\u003c/h2\u003e\n\u003cp\u003eWe performed extensive validations of the two T2T genome assemblies. Firstly, we examined their Hi-C chromatin interaction maps showing no obvious misplacement of contigs within the CaT2T and CrT2T assemblies (Supplementary Fig. 7AB). Then we mapped all HiFi, ONT, and NGS reads separately against the assemblies, yielding a mapping rate of over 99.96% for all three data types (Supplementary Table 5). Mapped HiFi or ONT reads showed uniform coverage across all whole genome\u0026nbsp;with a few exceptions in CrT2T due to the presence of high-copy satellite repeats\u0026nbsp;(Supplementary Fig. 7CD). The CaT2T and CrT2T had a QV (quality value) of 56.60 and 77.18, and a BUSCO score of 98.62% and 97.12%, respectively, demonstrating high accuracy and completeness of both assemblies (\u003cstrong\u003eTable\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e1\u003c/strong\u003e).\u0026nbsp;Furthermore, aligning a recently published genome assembly of \u003cem\u003eC. annuum\u003c/em\u003e cultivar \u0026lsquo;59\u0026rsquo; (hereafter Ca59)\u003csup\u003e4\u003c/sup\u003e against CaT2T showed strong collinearity between the two assemblies (Supplementary Fig. 2D). The high-quality assembly of CaT2T was well supported by a high-coverage of HiFi and ONT read mapping spanning these gap regions (Supplementary Fig. 8; Supplementary Table 6). Interestingly, we observed sporadic high coverage read mapping against CaT2T and CrT2T corresponding to intact mitochondrial or chloroplast genomes, which were validated by high coverage of ONT ultra-long read mapping spanning the entire integration sites (Supplementary Fig. 9), suggesting recent plastid genome integration in the nuclear genome.\u003c/p\u003e\n\u003cp\u003eRepeat annotation showed that 79.5%(2.45 Gb) and 74.6% (1.28 Gb) of \u003cem\u003eC. annuum\u003c/em\u003e and \u003cem\u003eC. rhomboideum\u0026nbsp;\u003c/em\u003egenomes were repetitive sequences, primarily composed of transposable elements (TEs), especially long terminal repeat (LTR) retrotransposons (Supplementary Table 7). While LTR insertion in \u003cem\u003eC. rhomboideum\u003c/em\u003e occurred relatively recently, \u003cem\u003eC. annuum\u003c/em\u003e had two bursts of insertion dated around 0.1 Mya (million years ago) and 3.9 Mya (Supplementary Fig. 10), consistent with previous report in Ca59 assembly\u003csup\u003e4\u003c/sup\u003e. Both pepper genomes had low contents of satellite repeats (\u0026lt; 0.01%), much fewer than human (4.5%), the model plant Arabidopsis (0.37%), and its Solanaceae relative tobacco (1.75%). A total of 34,428 and 33,512 protein-coding genes were predicted for CaT2T and CrT2T by using a combination of \u003cem\u003eab initio\u0026nbsp;\u003c/em\u003eprediction, homologous proteins, and transcriptomic data. Filled gaps by CaT2T encoded 614 genes, of which 110 were newly annotated (nonsyntenic to Ca59) (Supplementary Fig. 11). Both \u003cem\u003eCapsicum\u0026nbsp;\u003c/em\u003egenomes were gene-dense towards the end of chromosomes but gene-sparse towards centromeres. CaT2T encoded 117 putative capsaicin biosynthesis genes (CBGs), including previously reported genes such as capsaicin synthase (\u003cem\u003eCS\u003c/em\u003e),\u0026nbsp;aminomethytransferase (AMT), ketoacyl-ACP synthase (Kas) and acyl carrier protein (ACL)\u003csup\u003e27\u003c/sup\u003e(Supplementary Table 8). These putative CBGs were expressed in at least one tissue, and 26 of them showed much higher expression levels in the fruit placenta compared to other tissues (Supplementary Fig. 12), therefore serving as strong candidates for full elucidation of capsaicin biosynthetic enzymes.\u003c/p\u003e\n\u003ch2\u003eCapsicum centromeres\u0026nbsp;are extensively invaded by CRM retrotransposons\u003c/h2\u003e\n\u003cp\u003eEssential to faithful chromosomal segregations during cell division, centromeres are typically heterochromatic regions containing megabase-arrays of tandem repeats where the kinetochore protein complex binds\u003csup\u003e28\u003c/sup\u003e. We first identified centromeres of CaT2T by generating CENH3 ChIP-seq data of \u003cem\u003eC. annuum\u003c/em\u003e, which clearly delineated the location and boundary of 12 centromeres in CaT2T (\u003cstrong\u003eFig.\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e2A\u003c/strong\u003e). Then, we observed that the inter-chromosome interactions were always positively correlated with the ChIP-seq peak (Supplementary Fig. 13A), especially in CrT2T (Supplementary Fig. 13B). Unlike Arabidopsis\u003csup\u003e16\u003c/sup\u003e and human\u003csup\u003e13\u003c/sup\u003e centromeres, \u003cem\u003eCapsicum\u0026nbsp;\u003c/em\u003ecentromeres lacked high-copy tandem satellite repeats and higher order repeats (HORs), suggestive of most newly-formed centromeres\u003csup\u003e29\u003c/sup\u003e. Rather, we found \u003cem\u003eCapsicum\u0026nbsp;\u003c/em\u003ecentromeres were extensively invaded by \u003cem\u003eGypsy-\u003c/em\u003eLTRs taking up ~71% of total centromeric sequences (\u003cstrong\u003eFig.\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;2B\u003c/strong\u003e), with only 47.3%~49.7% for whole genomes (Supplementary Table 7).\u0026nbsp;This pattern was also reported in einkorn\u003csup\u003e30\u003c/sup\u003e and cotton\u003csup\u003e31\u003c/sup\u003e, where more than 80% of their functional centromeres are \u003cem\u003eGypsy-\u003c/em\u003eLTRs. We also found that \u003cem\u003eCapsicum\u0026nbsp;\u003c/em\u003eLTRs had a later burst of insertion in centromeres than in the whole genome, indicating recent evolution of centromeres shaped by LTR insertion (\u003cstrong\u003eFig.\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e2B\u003c/strong\u003e). LTRs are typically subjected to reshuffle and rearrangement due to unequal homologous recombination events generating fragmented or solo LTRs\u003csup\u003e32\u003c/sup\u003e. A number of solo LTRs and intact LTRs were identified in two \u003cem\u003eCapsicum\u0026nbsp;\u003c/em\u003egenomes (Supplementary Table 9). We observed that \u003cem\u003eCapsicum\u0026nbsp;\u003c/em\u003ecentromeres had a weaker ability than non-centromeres to remove LTRs, indicated by their significantly lower solo-to-intact LTR ratios compared to the whole genome (\u003cstrong\u003eFig.\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e2C\u003c/strong\u003e). The synteny analysis showed that the LTR retrotransposon CRMs (centromeric retrotransposon of maize) were enriched in centromeres of several \u003cem\u003eCapsicum\u0026nbsp;\u003c/em\u003egenomes (Supplementary Fig. 13CD), suggesting the distribution of CRMs was a marker that could be used to identify \u003cem\u003eCapsicum\u0026nbsp;\u003c/em\u003ecentromeres without relying on ChIP-seq data\u003cem\u003e.\u0026nbsp;\u003c/em\u003eCRMs possess chromodomain or CR motifs that potentially interact with the centromeric histones and play important roles in centromere evolution and function\u003csup\u003e33\u003c/sup\u003e. Furthermore\u003cem\u003e,\u0026nbsp;\u003c/em\u003ephylogenetic analysis of \u003cem\u003eGypsy\u003c/em\u003e-LTR retrotransposons in two \u003cem\u003eCapsicum\u0026nbsp;\u003c/em\u003eand potato genomes identified six subfamilies with two major ones: \u003cem\u003eAthila\u0026nbsp;\u003c/em\u003eand \u003cem\u003eTekay\u003c/em\u003e. \u003cem\u003eAthila\u0026nbsp;\u003c/em\u003eLTRs are\u003cem\u003e\u0026nbsp;\u003c/em\u003ethe major LTRs in Arabidopsis centromeres\u003csup\u003e16\u003c/sup\u003e, unlike \u003cem\u003eCapsicum\u0026nbsp;\u003c/em\u003ecentromeres which are enriched with CRM \u003cem\u003eGypsy-\u003c/em\u003eLTRs (\u003cstrong\u003eFig.\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e2D;\u0026nbsp;\u003c/strong\u003eSupplementary Table 9; Supplementary Table 10). The lack of satellite repeats and enrichment of CRM LTRs (\u003cstrong\u003eFig.\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e2E\u003c/strong\u003e) distinguished \u003cem\u003eCapsicum\u0026nbsp;\u003c/em\u003ecentromeres from centromeres reported in other published plant T2T genomes\u003csup\u003e16,20-22\u003c/sup\u003e. Strikingly, CrT2T had a higher content of CRMs than CaT2T, contributing to a distinctive signature of sequence identity within their centromeres (Supplementary Fig. 14)\u003cem\u003e.\u0026nbsp;\u003c/em\u003eFor both T2T genomes, we observed\u003cem\u003e\u0026nbsp;\u003c/em\u003elow inter-species and inter-chromosomal centromere sequence identity, suggesting a rapid divergence of \u003cem\u003eCapsicum\u0026nbsp;\u003c/em\u003ecentromeres within and between species, consistent with the findings in a recent study of Arabidopsis centromeres\u003csup\u003e34\u003c/sup\u003e.\u0026nbsp;\u003c/p\u003e\n\u003ch2\u003eCentromeres and telomeres are transcriptionally and epigenetically active\u003c/h2\u003e\n\u003cp\u003eCentromeres and telomeres are overall poorly understood genomic regions in terms of protein-coding genes, transcriptional and epigenetic control. Genome annotation predicted 60 genes in CaT2T\u003cem\u003e\u0026nbsp;\u003c/em\u003ecentromeres, enriched in functions such as response to freezing, DNA topological change and meiotic chromosome separation (Supplementary Fig. 15). By contrast, CrT2T\u003cem\u003e\u0026nbsp;\u003c/em\u003ecentromeres encoded 94 genes enriched in response to UV-B, photosynthesis and regulation of circadian rhythm. Interestingly, only six or seven centromere genes were homologous (Supplementary Table 11). The\u0026nbsp;low homology of centromere genes reflected a high divergence of centromeres between the species, consistent with their poor whole-genome sequence alignment. We found that CaT2T centromeres (\u003cstrong\u003eFig.\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e3A;\u0026nbsp;\u003c/strong\u003eSupplementary Fig. 16) and telomeres (\u003cstrong\u003eFig.\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e3B\u003c/strong\u003e) had active transcriptions as suggested by RNA-seq analysis, both from transposons and protein-coding genes. For example, about 42 (70.0%) centromere-encoded genes were expressed in at least one tissue (TPM \u0026gt; 1) in \u003cem\u003eC. annuum\u003c/em\u003e, including \u003cem\u003eCaT2T07g00954\u003c/em\u003e encoding a telomere maintenance protein that protects the ends of telomeres from attack and \u003cem\u003eCaT2T01g02835\u003c/em\u003e encoding a flowering time control protein. The expression of CrT2T centromere-encoded genes (92.5%) was more active, containing three tandem copies of RCC1 (Regulator of chromosome condensation) with an average TPM \u0026gt; 200, potentially playing key roles in the regulation of chromatin condensation in mitosis.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eCapsicum\u0026nbsp;\u003c/em\u003egenomes contain rich epigenomic signatures such as DNA methylation, histone modifications, TADs (topological associated domains) and A/B compartments\u003csup\u003e4\u003c/sup\u003e. However, little is known about these epigenetic marks in centromeres and telomeres. Therefore, we mapped our own generated (Hi-C and bisulfite sequencing) and public epigenomic (histone ChIP-seq) data to CaT2T assembly. Hi-C data analysis detected A/B compartment, TADs, and small chromatin loops within centromeres (\u003cstrong\u003eFig.\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;3A\u003c/strong\u003e). \u003cem\u003eC. annuum\u003c/em\u003e centromeres primarily belonged to \u0026quot;B\u0026quot; compartment typically associated with low transcription, except that \u0026quot;A\u0026quot; compartments were detected in a few chromosomes such as Chr08, Chr09 and\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eChr10 (Supplementary Fig. 16). Consistently, high TE density, low gene density (\u003cstrong\u003eFig.\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e3C\u003c/strong\u003e), and low histone H3K9me2 ChIP-seq peaks were detected for heterochromatic centromeres and pericentromeres with enriched CENH3 ChIP-seq signals (\u003cstrong\u003eFig.\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e3D\u003c/strong\u003e). Interestingly,\u0026nbsp;although\u0026nbsp;DNA methylation levels in centromere and non-centromere regions were overall comparable (\u003cstrong\u003eFig.\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e3C\u003c/strong\u003e), we found\u0026nbsp;centromere-located CRMs showed lower gene-body CHG methylation compared to gene flanking regions\u0026nbsp;(\u003cstrong\u003eFig.\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e3E\u003c/strong\u003e\u003cstrong\u003e;\u0026nbsp;\u003c/strong\u003eSupplementary Fig. 17), suggesting transcriptional activity of CRMs. Despite heterochromatin nature, transcription in centromeres was not silent as\u0026nbsp;we identified some highly expressed genes and TEs, such as one on Chr07 encoding peroxiredoxin Q protein involved in cell redox homeostasis (\u003cstrong\u003eFig.\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e3A\u003c/strong\u003e). Compared to centromeres, most telomeres had low Hi-C mapping signals \u003cstrong\u003e(Fig.\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e3B\u003c/strong\u003e\u003cstrong\u003e)\u0026nbsp;\u003c/strong\u003eperhaps due to the high density of tandem telomeric repeats. Subtelomere regions were relatively gene-rich mostly associated with A compartments and marked with low H3K9me3 ChIP-seq signals except in gene-sparse regions (\u003cstrong\u003eFig.\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e3B\u003c/strong\u003e). These results provided unprecedented insights into the organization and functions of pepper complex genomic regions.\u003c/p\u003e\n\u003ch2\u003eRevisiting evolutionary history of capsaicinoid biosynthesis pathway in \u003cem\u003eCapsicum\u003c/em\u003e\u003c/h2\u003e\n\u003cp\u003eThe birth and death of capsaicinoid biosynthesis in plants remains poorly understood. Kim \u003cem\u003eet al.\u003c/em\u003e first approached this question by comparing a fragmented pepper genome with tomato genome, revealing the mechanisms of pungency\u003csup\u003e2\u003c/sup\u003e. However, a better understanding of how the pathway emerged and evolved required an investigation in a broader phylogenetic context. Therefore, we revisited this question by preforming phylogenomics using two T2T \u003cem\u003eCapsicum\u0026nbsp;\u003c/em\u003egenomes and 14 other angiosperm genomes (Supplementary Table 12) including three pungent and 13 nonpungent species. We found that \u003cem\u003eCapsicum\u0026nbsp;\u003c/em\u003ewas more related to \u003cem\u003ePhysalis\u0026nbsp;\u003c/em\u003e(ground cherry)\u003cem\u003e\u0026nbsp;\u003c/em\u003ethan to \u003cem\u003eSolanum\u0026nbsp;\u003c/em\u003e(\u003cem\u003ee.g.\u0026nbsp;\u003c/em\u003etomato), diverging from the two taxa at ~17 Mya and ~19 Mya, respectively (\u003cstrong\u003eFig.\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e4A\u003c/strong\u003e). The fact that capsaicinoid biosynthesis is limited to \u003cem\u003eCapsicum\u003c/em\u003e indicated that the specialized metabolite must have arisen in \u003cem\u003eCapsicum\u0026nbsp;\u003c/em\u003eafter their divergence from \u003cem\u003ePhysalis\u003c/em\u003e\u003cem\u003e\u0026nbsp;\u003c/em\u003eat ~17 Mya. Furthermore, \u003cem\u003eC.\u003c/em\u003e \u003cem\u003ebaccatum\u0026nbsp;\u003c/em\u003ediverged from \u003cem\u003eC. annuum\u003c/em\u003e and \u003cem\u003eC. chinense\u003c/em\u003e at ~ 5Mya, which together\u003cem\u003e\u0026nbsp;\u003c/em\u003ediverged from the non-pungent \u003cem\u003eC. rhomboideum\u003c/em\u003e at ~13.4 Mya, suggesting that the capsaicinoid pathway could have formed between 13.4 Mya and 5 Mya (\u003cstrong\u003eFig.\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e4A\u003c/strong\u003e). To understand how the pathway arose in \u003cem\u003eCapsicum\u0026nbsp;\u003c/em\u003eplants, we identified genes from the 16 angiosperms with homologs to known CBGs using \u003cem\u003eOrthoFinder\u003c/em\u003e\u003csup\u003e35\u003c/sup\u003e (Supplementary Table 13). All species, pungent or not, contained homologs of CBGs, except that \u003cem\u003eCS\u003c/em\u003e, the most critical gene\u003csup\u003e1\u003c/sup\u003e, showed up as tandem duplicates only in Solanaceae at around 71 Mya (\u003cstrong\u003eFig.\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e4B;\u0026nbsp;\u003c/strong\u003eSupplementary Fig. 18). Particularly, \u003cem\u003eCapsicum\u0026nbsp;\u003c/em\u003especies had the most copies of \u003cem\u003eCS\u0026nbsp;\u003c/em\u003ewith seven in \u003cem\u003eC. annuum\u003c/em\u003e, six in \u003cem\u003eC. chinense\u003c/em\u003e, and four in \u003cem\u003eC. pubescens\u003c/em\u003e, \u003cem\u003eC. baccatum\u0026nbsp;\u003c/em\u003eand\u003cem\u003e\u0026nbsp;C. rhomboideum\u003c/em\u003e. The \u003cem\u003eC. pubescens\u003c/em\u003e\u003cem\u003e\u0026nbsp;\u003c/em\u003eand\u003cem\u003e\u0026nbsp;C. baccatum\u0026nbsp;\u003c/em\u003ehad fewer copies than \u003cem\u003eC. annuum\u003c/em\u003e, due to either gene loss in the former two, or tandem duplications in the latter. In fact, \u003cem\u003eCS\u0026nbsp;\u003c/em\u003eduplication was widely detected in Solanaceae (\u003cstrong\u003eFig.\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e4B\u003c/strong\u003e), as previously reported in tomato\u003csup\u003e2\u003c/sup\u003e, but tandem duplication mainly occurred in \u003cem\u003eSolanum\u003c/em\u003e, \u003cem\u003ePhysalis\u003c/em\u003e,\u003cem\u003e\u0026nbsp;\u003c/em\u003eand \u003cem\u003eCapsicum\u0026nbsp;\u003c/em\u003espp. between 30 Mya to 40 Mya (\u003cstrong\u003eFig.\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e4A\u003c/strong\u003e). In addition, microsynteny analysis showed that \u003cem\u003eCS\u003c/em\u003e tandem duplicates were syntenic in \u003cem\u003eCapsicum, Physalis\u003c/em\u003e,\u003cem\u003e\u0026nbsp;\u003c/em\u003eand \u003cem\u003eSolanum\u0026nbsp;\u003c/em\u003e(\u003cstrong\u003eFig.\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e4C\u003c/strong\u003e), so were other CBGs (Supplementary Fig. 19). This suggested that expression of CBGs in non-pungent species may have been disrupted. Indeed, we found many CBGs were highly expressed in fruits of pungent \u003cem\u003eCapsicum\u003c/em\u003e species (\u003cstrong\u003eFig.\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e4D\u003c/strong\u003e), whereas \u003cem\u003eC. rhomboideum\u003c/em\u003e and \u003cem\u003ePhysalis CS\u0026nbsp;\u003c/em\u003eand \u003cem\u003eKasI\u0026nbsp;\u003c/em\u003egenes were hardly expressed (\u003cstrong\u003eFig.\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e4D\u003c/strong\u003e). Sequence alignment revealed that these syntenic \u003cem\u003eCS\u0026nbsp;\u003c/em\u003ecopies (\u003cem\u003eCS-1\u003c/em\u003e/\u003cem\u003eCS-2\u003c/em\u003e) had conserved coding sequences (CDS) and upstream and downstream regulatory regions among pungent species, while non-pungent species had structural variations (SVs) within both CDS and flanking regions (\u003cstrong\u003eFig. 4E;\u0026nbsp;\u003c/strong\u003eSupplementary Fig. 20). Besides, sequence variations were also observed for several other CBGs including \u003cem\u003eACL\u003c/em\u003e, \u003cem\u003eBCAT, CCoAMT, FatA,\u0026nbsp;\u003c/em\u003eand\u003cem\u003e\u0026nbsp;KasI\u0026nbsp;\u003c/em\u003e(Supplementary Fig. 19). The highly conserved CS copies were only present in cultivated \u003cem\u003eCapsicum\u0026nbsp;\u003c/em\u003esuggesting they were relatively recent (~ 5 Mya) tandem duplicates from older CS genes. Taken together, although non-pungent\u003cem\u003e\u0026nbsp;\u003c/em\u003especies contained homologs of functional CBGs, their coding and regulatory regions had diverged significantly from those of \u003cem\u003eC. annuum\u003c/em\u003e.\u003c/p\u003e\n\u003ch2\u003eChromatin accessibility regulates tissue-specific capsaicinoid biosynthesis\u003c/h2\u003e\n\u003cp\u003eCapsaicinoid\u0026nbsp;biosynthesis is highly tissue-specific, only accumulating in fruits particularly placenta beginning at ~16 days post anthesis. To understand how this tissue-specificity is achieved, we performed multi-omic co-profiling of \u003cem\u003eC. annuum\u003c/em\u003e fruits and leaves including ATAC-seq (Assay for Transposase-Accessible Chromatin sequencing), whole-genome bisulfite sequencing and RNA-seq, and analyzed them using CaT2T as reference. RNA-seq analysis showed that \u003cem\u003eCS\u0026nbsp;\u003c/em\u003eand its transcriptional regulators \u003cem\u003eMYB31\u003c/em\u003e\u003csup\u003e36\u003c/sup\u003e\u003cem\u003e\u0026nbsp;and MYB48\u003c/em\u003e\u003csup\u003e37\u003c/sup\u003e\u003cem\u003e\u0026nbsp;\u003c/em\u003ewere specifically expressed in placenta (\u003cstrong\u003eFig.\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e4F\u003c/strong\u003e). The placenta-specific open chromatin regions (OCRs) with low methylation levels were detected within 2 kb upstream of \u003cem\u003eCS-2\u003c/em\u003e,\u003cem\u003e\u0026nbsp;MYB31\u0026nbsp;\u003c/em\u003eand \u003cem\u003eMYB48\u003c/em\u003e, while \u003cem\u003eCS-1\u0026nbsp;\u003c/em\u003ealso showed OCRs in both pulp and seeds, suggesting that \u003cem\u003eCS-2\u003c/em\u003e is likely the primary functional gene that contributes to placenta-specific synthesis of capsaicinoids (\u003cstrong\u003eFig.\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e4F\u003c/strong\u003e and Supplementary Fig. 21).\u0026nbsp;Other 26 putative CBGs encoding phenylpropanoid and valine pathways were expressed in both fruits and leaves but lacking tissue-specific OCRs, suggesting their functions beyond capsaicinoid biosynthesis (Supplementary Fig. 12). Nonetheless, their higher expression in placenta than in other tissues (Supplementary Fig. 12) indicated strong co-regulation for capsaicinoid biosynthesis.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eHow co-regulation of genomically dispersed CBGs is achieved to confer this tempo-spatial specificity of capsaicinoid production remains unknown. One would expect there should be common regulatory elements for the CBGs. To validate this hypothesis, we extracted the placenta-associated OCRs identified by ATAC-seq and 2 kb upstream sequence of the CBGs to perform sequence motif enrichment analysis, yielding a total of 38 enriched motifs (\u003cem\u003ep\u003c/em\u003e \u0026lt; 0.01). Overlapping\u0026nbsp;enriched motifs obtained from two enrichment analyses identified five transcription factor binding motifs (TFBS): MYB, G-box, Box-4, ABRE and MYC, present in all seven CBGs including \u003cem\u003eCS, ACL, KasI\u003c/em\u003e, \u003cem\u003ePAL\u003c/em\u003e, \u003cem\u003eCCoAOMT\u0026nbsp;\u003c/em\u003eand \u003cem\u003eBCAT\u003c/em\u003e (\u003cstrong\u003eFig.\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e4G\u003c/strong\u003e).\u0026nbsp;It suggested these significantly enriched TFBS within common OCRs were likely recognized by certain TFs such as MYB31\u003cem\u003e\u0026nbsp;\u003c/em\u003eco-regulating these CBGs in specific tissues\u003csup\u003e36\u003c/sup\u003e. In non-pungent \u003cem\u003eC. annuum\u003c/em\u003e cultivars, the\u0026nbsp;OCRs of \u003cem\u003eCS-2\u0026nbsp;\u003c/em\u003ewere lost due to a 2.4 kb deletion conferring the lack of capsaicinoids in these accessions (\u003cstrong\u003eFig.\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e4F\u003c/strong\u003e). In short, multi-omic data analysis illustrated a likely epigenetic mechanism for the tissue-specific co-regulation of capsaicinoid biosynthesis genes in chili pepper.\u0026nbsp;\u003c/p\u003e\n\u003ch2\u003eMachine-learning model for predicting fruit pungency from copy number variants\u003c/h2\u003e\n\u003cp\u003eCopy number variants (CNVs) have great impact on plant gene functions and traits such as grain size\u003csup\u003e39\u003c/sup\u003e and biosynthesis of natural products\u003csup\u003e40\u003c/sup\u003e. However, whether and how CNVs impact capsaicinoid biosynthesis is unknown. Fruit pungency levels in pepper cultivars are highly variable owing to artificial selection\u003csup\u003e41\u003c/sup\u003e. Using CaT2T as reference genome, we detected genomewide CNVs from resequencing data of 311 \u003cem\u003eC. annuum\u003c/em\u003e accessions with capsaicinoid quantification\u003csup\u003e38\u0026nbsp;\u003c/sup\u003eusing AMYCNE (Supplementary Table 14), finding CNVs for 121 CBGs including seven \u003cem\u003eMYB\u003c/em\u003e TFs (\u003cstrong\u003eFig.\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e5A\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eand Supplementary Table 15). Expectedly, a strong association between \u003cem\u003eCS\u003c/em\u003e CNV and capsaicinoid production was observed, where non-pungent cultivars had low copy of the putative CS gene (\u003cstrong\u003eFig.\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;5A\u003c/strong\u003e) due to a 2.4 kb deletion spanning\u0026nbsp;functional\u0026nbsp;CDS\u003csup\u003e1,38\u003c/sup\u003e. For cultivars with intact \u003cem\u003eCS\u003c/em\u003e gene, capsaicinoid levels might be affected by CNVs of other CBGs. To understand the relationship between CNV and capsaicinoid levels, we trained several machine-learning models including Random Forest, K-nearest neighbor, Gradient Boosting, and SVM (support vector machine), taking CNVs and capsaicinoid levels as input. After performing recursive feature selection and grid search, a Random Forest model achieved the highest average prediction accuracy of approximately 0.72 using 10-fold cross-validation on a dataset composed of CNV detection from newly generated resequencing data with approximately 50\u0026times;coverage (Table S14). The ROC curve of the Random Forest model also displayed the highest AUC value of 0.85 on this same dataset (Fig. 5C), demonstrating the potential of the model to accurately classify capsaicin levels from CNVs for genomic prediction at an early seedling stage. Furthermore, the Random Forest model ranked 121 CBGs by their contribution to prediction performance. The top-ranked genes included well known CBGs such as \u003cem\u003eCS-2, KasI, C3H, PAL,\u0026nbsp;\u003c/em\u003e\u003cem\u003eMYB31\u0026nbsp;\u003c/em\u003eand putative ones as \u003cem\u003eDH\u003c/em\u003e,\u003cem\u003e\u0026nbsp;SAMSyn\u003c/em\u003e and \u003cem\u003eIPMS\u003c/em\u003e \u003cem\u003eetc.\u0026nbsp;\u003c/em\u003e(\u003cstrong\u003eFig.\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e5D\u003c/strong\u003e), majority of which were highly expressed in fruits (Supplementary Table 16). Remarkably, the model also correctly ranked the functional copy of \u003cem\u003eCS-2\u0026nbsp;\u003c/em\u003e(\u003cem\u003eCaT2T02g02791\u003c/em\u003e) above the other six tandem copies, consistent with current knowledge about this gene. Overall, our machine-learning model demonstrated its prediction power and will facilitate elucidation of the full capsaicin biosynthesis pathway by prioritizing top-ranked CBGs.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThe two T2T gap-free \u003cem\u003eCapsicum\u0026nbsp;\u003c/em\u003egenome assemblies generated in this study represent key milestones in plant genome research. Firstly, almost ten years after the first sequenced pepper genome\u003csup\u003e2\u003c/sup\u003e, we mapped the complete genome sequence of \u003cem\u003eC.\u003c/em\u003e \u003cem\u003eannuum\u003c/em\u003e, representing the largest complete genome sequence reported so far. The genomics of chili peppers enter the era of T2T complete genomes, following the footsteps of human, model plant Arabidopsis, rice and maize. Secondly, to investigate the mechanisms of pungency evolution, we assembled a T2T gapless genome for a non-pungent wild pepper \u003cem\u003eC. rhomboideum\u003c/em\u003e, the first gapless genome for a wild pepper species. Through phylogenomics involving five \u003cem\u003eCapsicum\u0026nbsp;\u003c/em\u003e(four pungent, one nonpungent) and ten non-\u003cem\u003eCapsicum\u0026nbsp;\u003c/em\u003e(all nonpungent) genomes, we have traced the evolutionary history of capsaicinoid biosynthesis pathways among closely-related Solanaceae plants by estimating the times for the emergence of pungency, tandem duplications of key genes, and a functional biosynthesis pathway along the evolutionary tree\u0026nbsp;(\u003cstrong\u003eFig.\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e6\u003c/strong\u003e). However, due to lacking sufficient high quality \u003cem\u003eCapsicum\u0026nbsp;\u003c/em\u003egenomes in addition to the two assembled in this study, it is challenging to answer some key interesting evolutionary questions. For example, how did pungent \u003cem\u003eC. baccatum\u003c/em\u003e lose all but one tandem duplicates of \u003cem\u003eCS\u003c/em\u003e genes, whereas nonpungent \u003cem\u003eC. rhombiodeum\u003c/em\u003e managed to keep all four copies? Is it possible the functional \u003cem\u003eCS\u003c/em\u003e gene and \u003cem\u003ecis\u003c/em\u003e-regulatory elements were already in the Solanaceae plants but later lost in nonpungent species? Alternatively, the five cultivated \u003cem\u003eCapsicum\u0026nbsp;\u003c/em\u003especies could gain functional \u003cem\u003eCS\u003c/em\u003e genes or \u003cem\u003ecis\u003c/em\u003e-regulatory element probably through natural mutations or TE transpositions during domestication. \u003cem\u003eCapsicum\u0026nbsp;\u003c/em\u003egenus has over 40 species with diverse genetic backgrounds and traits\u003csup\u003e41\u003c/sup\u003e, but reference genomes are only available for four species so far. With more genomes of \u003cem\u003eCapsicum\u0026nbsp;\u003c/em\u003espp. with diverse capsaicin profiles available in the future, it is possible to better infer when and how the birth and death of the pathway occurred.\u003c/p\u003e\n\u003cp\u003eMany plant secondary metabolites accumulate in multiple tissues and organs, but tissue-specific production of natural product molecules is not uncommon, with examples such as morphines in opium poppy capsules\u003csup\u003e42\u003c/sup\u003e and capsaicins in chili pepper fruits. The expression of the biosynthetic genes in specific tissues is required for the specificity, for which the regulatory mechanisms remain elusive. Typically, metabolic gene clusters can facilitate the co-regulation of biosynthetic genes as seen in the examples of morphine (poppy)\u003csup\u003e42\u003c/sup\u003e and thalianol (Arabidopsis)\u003csup\u003e43\u003c/sup\u003e. However, biosynthesis genes of many plant metabolites such as capsaicin and colchicine (lily)\u003csup\u003e44\u003c/sup\u003e are unclustered and dispersed genomically. The spatial-tempo regulatory mechanisms for biosynthetic genes are poorly understood overall. Through comparative ATAC-seq and RNA-seq analysis of fruit and leaf tissues, we detected fruit-specific open chromatin regions for several key biosynthetic genes and regulators encoding conserved TF binding sites (\u003cstrong\u003eFig.\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;6\u003c/strong\u003e). This concerted gene regulation likely enabled efficient production of biosynthetic enzymes at specific time and space. How these genes evolved such common cis-regulatory elements requires further investigation.\u003c/p\u003e\n\u003cp\u003eTwo T2T gapless genomes allowed us to gain insights into the complex genomic regions such as centromeres, telomeres and filled gaps. Centromeres were extensively studied in T2T genomes of human\u003csup\u003e13\u003c/sup\u003e, Arabidopsis\u003csup\u003e16\u003c/sup\u003e and rice\u003csup\u003e20\u003c/sup\u003e, all containing abundant high-copy tandem repeats. However, centromeres of Capsicum lacked such satellites but were enriched with \u003cem\u003eGypsy\u003c/em\u003e-LTR especially CRM retrotransposons. We also found this pattern was common in\u003cem\u003e\u0026nbsp;C. baccatum\u003c/em\u003e, \u003cem\u003eC. chinense\u003c/em\u003e and \u003cem\u003eC. pubescens\u0026nbsp;\u003c/em\u003ebased on the recently reported high-quality genomes\u003csup\u003e11\u003c/sup\u003e, although the composition of CRM retrotransposons varied among different \u003cem\u003eCapsicum\u003c/em\u003e species. The potato genome also showed enrichment of CRM in centromeres\u003csup\u003e21\u003c/sup\u003e, but no CRM was identified in whole genome of \u003cem\u003eP. pruinosa\u003c/em\u003e (1.38 Gb, contig N50: 82.2 Mb)\u003csup\u003e45\u003c/sup\u003e, suggesting that Solanaceae centromeres evolved rapidly and were diverse among species. Plant centromeres are rapidly diversifying due to cycles of transposon invasions even within species as shown by a recent study on 346 Arabidopsis centromeres\u003csup\u003e34\u003c/sup\u003e, demonstrating the so-called \u0026apos;centromere paradox\u0026apos;. With more complete \u003cem\u003eCapsicum\u0026nbsp;\u003c/em\u003egenomes available, it will be interesting to compare centromeres among different \u003cem\u003eCapsicum\u0026nbsp;\u003c/em\u003eaccessions or\u003cem\u003e\u0026nbsp;\u003c/em\u003especies to understand the evolution of centromeres during speciation and domestication.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFrom application point of view, we first demonstrated two complete genome sequence of peppers, which can serve as a gold standard reference genome for improving variant analysis, and pangenome studies in the future. Secondly, one of the greatest challenges faced by crop breeding is the long breeding cycle and waiting time for trait evaluation. For example, fruit phenotyping of chili peppers could wait 120-150 days, begging for a method to accelerate germplasm assessment. Using capsaicinoid as an example, we performed CNV detection using T2T genome in a pepper population with known capsaicin levels, and trained a robust machine-learning model to predict fruit pungency from CNVs that can be quickly determined in pepper accessions of interest by genome resequencing analysis, or digital PCR of seedlings. To our knowledge, this is the first model that leverages biosynthesis gene copy numbers to infer levels of secondary metabolites, potentially turbo-charging natural product breeding. Admittedly, there is room for improvement as only a small dataset being used in training, but new datasets are expected to further train and strengthen the model.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIn summary, this study produced so far the largest complete plant genome assembly and the first two T2T gap-free genomes for \u003cem\u003eCapsicum\u003c/em\u003e. Phylogenomics and multiomics unraveled the evolutionary mechanisms of Capsicum-unique and tissue-specific accumulation of capsaicinoids in pepper fruits. The new T2T genomic resources represent an important milestone in crop genome research, and will speed up pepper research and precise improvement.\u0026nbsp;\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003e\u003cstrong\u003e\u003cem\u003ePlant materials and sequencing\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eCapsicum annuum\u003c/em\u003edouble haploid line ‘G1-36576’ and \u003cem\u003eC\u003c/em\u003e\u003cem\u003eapsicum\u003c/em\u003e\u003cem\u003e\u0026nbsp;rhomboi\u003c/em\u003e\u003cem\u003edeum\u003c/em\u003e wild accession ‘PI 645680’\u0026nbsp;plants were grown in regular azalea pots filled with a combination of potting mix, clay and vermiculite in greenhouse of Peking University Institute of Advanced Agricultural Sciences, Weifang (36°42'N and 119°10'E), Shandong province, China in the summer and autumn of 2022.\u0026nbsp;The fresh leaves of four weeks old \u003cem\u003eC\u003c/em\u003e\u003cem\u003eapsicum\u003c/em\u003e\u003cem\u003e\u0026nbsp;annuum\u003c/em\u003e and \u003cem\u003eC\u003c/em\u003e\u003cem\u003eapsicum\u003c/em\u003e\u003cem\u003e\u0026nbsp;rhomboideum\u0026nbsp;\u003c/em\u003ewere harvested and used for DNA extraction and sequencing. Tissues of leaf, root, stem and flower at two days post anthesis, and fruits (separated into pericarp, placenta and seeds)\u0026nbsp;at 21 days post anthesis for \u003cem\u003eC. annuum\u003c/em\u003e and \u003cem\u003eC. rhomboideum\u003c/em\u003e were collected for RNA extraction and sequencing. Tissues of leaf, placenta,\u0026nbsp;pericarp\u0026nbsp;and seeds\u0026nbsp;at 21 days post anthesis\u0026nbsp;were collected from \u003cem\u003eC. annuum\u003c/em\u003e and used for ATAC-seq.\u0026nbsp;Tissues of leaf and whole\u0026nbsp;fruit\u0026nbsp;at 21 days post anthesis\u0026nbsp;were collected from \u003cem\u003eC. annuum\u003c/em\u003e and used for bisulfite sequencing.\u0026nbsp;Additionally, fresh leaves of nine pepperaccessions were harvested and used for re-sequencing.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eDNA and RNA isolation\u0026nbsp;\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIsolation of high molecular weight (HMW) genomic DNA was conducted using CTAB (cetyltrimethylammonium bromide) method. Briefly, 10 μg clean and fresh leaves were ground in liquid nitrogen and then subjected to DNA extraction. The quality of DNA was checked using Qubit (Thermo Fisher Inc.) and a pulse field gel electrophoresis apparatus (BioRad) following manufacturer's recommendation. The isolation of total RNA was conducted using Trizol RNA extraction reagent (Thermo Fisher Inc.) following recommended protocol. The extracted RNA was assessed using RNA Nano 6000 Assay Kit of the Bioanalyzer 2100 system (Agilent Technologies, CA) following manufacturer’s recommendation. RNA samples with a RIN (RNA integrity number) value \u0026gt; 6.0 were proceeded to downstream library construction for RNA sequencing.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eGenome sequencing\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIllumina\u0026nbsp;paired-end sequencing library was prepared using NEB Next® Ultra™ DNA Library Prep Kit for Illumina (NEB, USA) following its standard protocol. Briefly, 5 µg HMW DNA sample was fragmented by sonication to a size of 350 bp. DNA fragments were then end-polished, A-tailed, and ligated with the full-length Illumina sequencing adapters. A total of 300.8 Gb (~100 × genome coverage) 150 bp paired-end reads were produced using Illumina Novoseq 6000 platform by Novogene Biotechnologies Inc. (Tianjin, China). The clean data were used in genome survey, genome assembly polishing and assembly evaluation. To generate PacBio HiFi long reads, a total of 15 µg HMW DNA was sheared by gTUBEs (Covaris, MA, USA), and was used to construct a standard PacBio SMRTbell library using PacBio SMRTbell Express Template Prep Kit 2.0 (PacBio, CA, USA). The resultant library was separated on BluePippin (Sage Science, MA, USA) with 15 kb cutoff to remove short DNA fragments. Then a total of 356.3 Gb HiFi consensus reads with N50 length of 18.3 kb were generated using a PacBio Sequel II system at Novogene Biotechnologies Inc. (Tianjin, China). To generate Oxford Nanopore ultra-long reads, the long DNA fragments were size-selected and processed following the Ligation Sequencing SQK-LSK109 Kit (Oxford Nanopore Technologies, Oxford, UK) protocol. Briefly, DNA ends were FFPE (Formalin-Fixed and Parrffin Embedded) repaired and end-prepped/dA-tailed using the NEBNext End Repair/dA-tailing module (New England Biolabs, UK). Then sequencing adapters were ligated onto the prepared ends using the NEBNext Quick Ligation module (New England Biolabs, UK). The final DNA library was sequenced using the GridION X5/PromethION sequencer (Oxford Nanopore Technologies, Oxford, UK) by Single-Molecule Sequencing Platform at Peking University Institute of Advanced Agricultural Sciences (Weifang, China). A total of 261.5 Gb ultra-long reads were generated, with read length N50s of 100.3 kb and reads nearly up to 3 Mb. The Hi-C library was prepared from cross-linked chromatins of pepper leaves using a standard Hi-C protocol\u003csup\u003e46\u003c/sup\u003e. Then the library was sequenced using Illumina NovaSeq 6000 to obtain 2×150 bp paired-end reads at Novogene Biotechnologies Inc. (Tianjin, China). A total of 348.2 Gb of Hi-C data with ~112 × coverage was generated, and classified as valid or invalid interactions using HiC-Pro v3.1.0\u003csup\u003e47\u003c/sup\u003e, and only valid interaction pairs were retained for subsequent analysis.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eGenome assembly\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eStep1: Genome size and heterozygosity of DH line ‘G1-36576’ were estimated using the Illumina data by Jellyfish v2.3.0 (\u003cem\u003ek\u003c/em\u003e-mer size = 19)\u003csup\u003e48\u003c/sup\u003e and Genomescope v1.0 (max \u003cem\u003ek\u003c/em\u003e-mer coverage = 1,000,000)\u003csup\u003e49\u003c/sup\u003e. The estimated genome size is 3.19 Gb and the heterozygous rate is 0.207%.\u003c/p\u003e\n\u003cp\u003eStep2: For the PacBio assembly, HiFi reads were assembled using Hifiasm (v0.16.1)\u003csup\u003e24\u003c/sup\u003e with the default parameters. The ONT assembly was conducted using NextDenovo (v2.5.0)\u003csup\u003e25\u003c/sup\u003e and polished using NextPolish (v1.4.0)\u003csup\u003e50\u003c/sup\u003e with parameters of ‘hifi_options = -min_read_len 1k -max_depth 100’ and ‘sgs_options = -max_depth 100 -bwa’. Then we aligned the contigs to the reference genomes of \u003cem\u003eC. annuum\u003c/em\u003e chloroplast (GenBank accession NC_018552.1) and mitochondria (GenBank accession NC_024624.1) with Minimap2 (v2.24)\u003csup\u003e51\u003c/sup\u003e. Contigs with at least 50% of their bases covered by chloroplast or mitochondria genome sequences were removed from the assembly.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eStep3: Quickmerge\u003csup\u003e52\u003c/sup\u003e was used to joint contigs in the HiFi assembly (as the query) using the contigs from the ONT assembly (as the reference), which created a HiFi and ONT hybrid assembly. As ONT long reads are usually error-prone, we aligned the HiFi contigs to the merged assembly, and replaced the ONT originated sequences with corresponding HiFi contigs. After this step, ten nearly complete chromosome-level contigs were obtained, while two contigs corresponding to Chr04 and Chr08 only possessed one single telomere. Then Hi-C sequencing data were used to anchor all contigs using the pipeline of Juicer (v1.5)\u003csup\u003e53\u003c/sup\u003e, 3D-DNA (v180419)\u003csup\u003e54\u003c/sup\u003e and Juicebox (v1.11.08)\u003csup\u003e26\u003c/sup\u003e. For assembly validation, the contigs were manually checked, orientation tuned and any mis-assembly was adjusted within Juicebox\u003csup\u003e26\u003c/sup\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eStep4: The rDNA arrays on the\u0026nbsp;acrocentric Chr08 are long tandem repeats of\u0026nbsp;45S unit (18S-5.8S-25S rDNA). To assemble the 45S rDNA arrays, we first estimated the rDNA copies. We used Barrnap v0.9 (https://github.com/tseemann/barrnap) to predict the location of rDNA in HiFi reads and extracted the 45S rDNA-containing HiFi reads. The copy number was estimated to be ~60,000/42 = 1,428 based on the 19-mers of 45S rDNA-containing HiFi reads (\u0026gt;20 kb, 42×depth). Two main types of repeat unit with different length were identified, Type A: 8,351-8,377 bp and Type B: 8,498-8,506 bp, each accounting for 70% and 30% of the total 45S rDNA arrays. To assemble the rDNA tandem arrays, we utilized the centroFlye HOR pipeline\u003csup\u003e55\u003c/sup\u003e as a reference. Due to the high similarities between rDNA units and the error-prone property of ONT long reads, we failed to assemble the NOR regions using ONT data. The 45S rDNA containing ONT long reads were used to extract prefix reads that containing telomeric repeats, internal reads that containing two types of rDNA at both ends and suffix reads that containing non-rDNA sequences. Then we assembled the 45S rDNA containing HiFi reads using Hifiasm\u003csup\u003e24\u003c/sup\u003e, generating a draft rDNA assembly (78 contigs, N50 of 445.3 kb, and sum of 15.5 Mb). Combined the extracted ONT reads and the assembled HiFi contigs, we identified rare 19-mers and connected the sequences with same unique 19-mers. We then used Hi-C data to anchor these sequences and filled the gaps by mapping the HiFi reads to the rDNA assembly using Winnowmap2 (v2.03, k=19, -x asm5)\u003csup\u003e56\u003c/sup\u003e. We finally obtained 12.66 Mb rDNA arrays with 1,506 rDNA copies, and added this sequence to the contig of Chr08.\u003c/p\u003e\n\u003cp\u003eStep5: We extracted ultra long ONT reads (\u0026gt; 200kb) with at least ten copies of telomeric repeat motif ‘TTTAGGG’ or ‘TTCAGGG’ variant, and aligned these reads to above genome assembly using Winnowmap2 (v2.03, k=19, -ax map-ont)\u003csup\u003e56\u003c/sup\u003e. Using these alignment coordinates, the overhang sequences of telomere-containing reads were manually patched to each telomere. Telomeres were then manually confirmed to be structurally valid. Finally, we obtained a T2T genome assembly of \u003cem\u003eC. annuum\u003c/em\u003e accession G1-36576, and named CaT2T. Similarly, the \u003cem\u003eC. rhomboideum\u003c/em\u003e genome was assembled using the same strategy described above. The final assembly was named CrT2T.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eGenome quality assessment\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo assess the quality of the genome assembly, we firstly compared the genomic alignment dot plot between CaT2T and Ca59 assemblies using Minimap2\u003csup\u003e51\u003c/sup\u003e and D-GENIES\u003csup\u003e57\u003c/sup\u003e. For mapping statistics, the NGS short reads were mapped using BWA (v0.7.17)\u003csup\u003e58\u003c/sup\u003e, and the HiFi and ONT long reads were mapped using Minimap2\u003csup\u003e51\u003c/sup\u003e. Then Samtools (v1.10)\u003csup\u003e59\u003c/sup\u003e was used to count the mapping rates and coverage depth. The Ca59 gap regions that are resolved in CaT2T genome were manually checked in IGV (v2.12.3)\u003csup\u003e60\u003c/sup\u003e windows. To assess genome completeness, we applied BUSCO (v5.4.3)\u003csup\u003e61\u003c/sup\u003e for ortholog detection using solanales_odb10 database (n = 5,950). Quality value (QV) was estimated using Merqury (v1.3)\u003csup\u003e62\u003c/sup\u003e from HiFi reads. The telomere sequences were identified using Tandem Repeat Finder (TRF, v4.09.1)\u003csup\u003e63\u003c/sup\u003e with the parameters of ‘2 7 7 80 10 80 2000 -d -l 16’. The resulting ‘.dat file’ was transformed into a GFF3 file, which was then used to identify seven base telomeric repeats.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eRepeat annotation and TE analysis\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe used the universal Repbase database and a species specific \u003cem\u003ede novo\u003c/em\u003e repeat library constructed by RepeatModeler (https://github.com/Dfam-consortium/RepeatModeler) to annotate the DNA sequences in two Capsicum species. The repetitive elements in the genome were then annotated and masked by RepeatMasker (v4.1.2)\u003csup\u003e64\u003c/sup\u003e using the following parameters of ‘-xsmall -s -no_id -cutoff 255 -frag 20000 -e ncbi’.\u0026nbsp;To achieve large-scale accurate discovery of LTR retrotransposons,\u0026nbsp;we applied LTR_Finder (v1.2)\u003csup\u003e65\u003c/sup\u003e, LTRharvest (v1.6.2)\u003csup\u003e66\u003c/sup\u003e, and LTR_retriever (v2.9.0)\u003csup\u003e67\u003c/sup\u003e to identify LTR elements.\u0026nbsp;We identified 7,383 and 9,579 intact LTR-RTs candidates\u0026nbsp;in CaT2T and CrT2T, respectively, which were used as input to the TEtranscripts analysis\u003csup\u003e68\u003c/sup\u003e. Then TEsorter (v1.3)\u003csup\u003e69\u003c/sup\u003e was implemented using HMM profiles obtained from the TE protein domain databases REXd-plant. The TE sequences were first translated in all six frames and the translated sequences were then searched against the database. Hits with coverage lower than 20% or E-value higher than 1e-3 were discarded. For the classifications of LTR-RTs,\u0026nbsp;intact elements were identified and classified based on the presence and order of five conserved domains, including capsid protein (GAG), aspartic proteinase (AP), integrase (INT), reverse transcriptase (RT), and RNase H (RH). Then the number of LTR-RTs decreased to 5,202 and 6,834 in CaT2T and ChT2T after filtered by the\u0026nbsp;conserved domains. Using TEsorter\u0026nbsp;(v1.3)\u003csup\u003e69\u003c/sup\u003e, the Ty1-\u003cem\u003eCopia\u0026nbsp;\u003c/em\u003eelements was classified into several clades, including \u003cem\u003eAle, Alesia, Angela, Bianca, Ikeros, Ivana, SIRE, TAR,\u0026nbsp;\u003c/em\u003eand \u003cem\u003eTork\u003c/em\u003e; while the Ty3-\u003cem\u003eGypsy\u0026nbsp;\u003c/em\u003eelements was classified into clades of \u003cem\u003eAthila, CRM, Galadriel, Ogre, Reina,\u0026nbsp;\u003c/em\u003eand \u003cem\u003eTekay\u003c/em\u003e.\u0026nbsp;The ratios of solo LTRs to intact LTRs in each LTR family were calculated using the script of ‘solo_intact_ratio.pl’ in the LTR_retriever software.\u0026nbsp;The insertion time of intact LTR retrotransposons were calculated using LTR_retriever according to the formula: T= K/2r, where K was the divergence between the two LTRs and r was the rate of nucleotide substitution. We employed an average substitution rate of (r) 7 × 10\u003csup\u003e-9\u003c/sup\u003e to estimate insertion times of LTR-RTs.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eGenome annotation\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eGene model prediction combined with the following three aspects of evidence: (a) \u003cem\u003eab initio\u0026nbsp;\u003c/em\u003eprediction, (b) homologous protein, and (c) RNA-seq evidence, was conducted using MAKER (v2.31.11)\u003csup\u003e70\u003c/sup\u003e pipeline in two successive rounds. At the first round, short read and full-length RNA-seq and homology proteins were provided. The protein sequences used for homology-based prediction were from \u003cem\u003eA. thaliana\u003c/em\u003e\u003csup\u003e16\u003c/sup\u003e, \u003cem\u003eC. annuum\u003c/em\u003e\u003csup\u003e11\u003c/sup\u003e, \u003cem\u003eS. tuberosum\u003c/em\u003e\u003csup\u003e21\u003c/sup\u003e, and universal Swiss-Prot proteins.\u0026nbsp;To implement the MAKER pipeline, short read RNA-Seq data was assembled into a transcriptome using StringTie (v2.2.1)\u003csup\u003e71\u003c/sup\u003e. The PacBio long read transcriptome data were processed using the SMRT Analysis software Isoseq3 (https://github.com/PacificBiosciences/IsoSeq). MAKERemployed BLAST to align transcripts and proteins to the soft-masked genome, then Exonerate (v2.2.0)\u003csup\u003e72\u003c/sup\u003e were used to polish the BLAST hits and thereby accurately annotate the coding regions. The parameters of est2genome and protein2genome were set to 1, so that MAKERpredicted gene models only based on the provided transcripts and proteins. Then a subset of MAKERgene models with AED\u003csup\u003e73\u003c/sup\u003e scores \u0026lt; 0.25 were used to train SNAP\u003csup\u003e74\u003c/sup\u003e for three rounds. The GeneMark-ET and Augustus model were trained using BRAKER (v2.1.6)\u003csup\u003e75\u003c/sup\u003e pipeline. Briefly, the same data were aligned to the soft-masked genome using Exonerate\u003csup\u003e72\u0026nbsp;\u003c/sup\u003eand HISAT2 (v2.1.0)\u003csup\u003e76\u003c/sup\u003e, then GeneMark-ET\u003csup\u003e77\u003c/sup\u003e was trained by the predicted gene structures, and the resulting ~6,000 good gene models were used for training AUGUSTUS (v3.2.3)\u003csup\u003e78\u003c/sup\u003e. In the second round, each set of gene predictions in round one was passed to MAKER through model_gff option and the evidence alignment options were turned off. The trained SNAP, GeneMark-ET and AUGUSTUS models were also integrated into MAKER to predict more credible genes. Finally, the unsupported gene models were filtered (keep_preds=0) and the highest-ranking gene sets with AED scores \u0026lt; 0.5 were retained.\u003c/p\u003e\n\u003cp\u003eTo compare the previously published genome annotation of \u003cem\u003eC. annuum\u003c/em\u003e\u003csup\u003e4,11\u003c/sup\u003e with our CaT2T genome annotation, we also performed Liftoff (v1.6.3)\u003csup\u003e79\u003c/sup\u003e to annotate protein-coding genes of CaT2T assembly based on the reference with parameters of “-flank 0.1 -sc 0.99 --copies”. Then Gffread (v0.12.7)\u003csup\u003e80\u003c/sup\u003e was used to filter transcripts without normal open reading frames accordingly. Gene models were finally manually checked and corrected in IGV-GSAman (https://gitee.com/CJchen/IGV-sRNA) with the support of mapped RNA-seq reads and previous annotations\u003csup\u003e4,11\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eSynteny and phylogenome analysis\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNon-redundant protein sequences from 16 species were prepared for ortholog analyses\u0026nbsp;(Supplementary Table 12).\u0026nbsp;Orthologues and orthogroups were then inferred using OrthoFinder (v2.5.4)\u003csup\u003e35\u003c/sup\u003e with default values setting and ‘-M msa’ activated. The longest predicted protein of each individual gene was used as the representative input for the OrthoFinder analysis. TrimAl (v1.4.12)\u003csup\u003e81\u003c/sup\u003e was used to remove poorly aligned regions of protein multiple sequence alignments. RAxML (v8.2.12)\u003csup\u003e82\u003c/sup\u003e was used to build Maximum Likelihood phylogenetic trees using the GAMMAJTT model, with rice as an out-group. TimeTree (www.timetree.org) is a public database containing divergence time estimates from various publications along with their own estimation. These estimates, ignoring the outliers, were used for selecting the range of lower and upper uniform calibration priors. The calibration values were chosen as 1.1-1.6, 109.2-123.5, and \u0026lt; 200 for the most common ancestor of the 13 species belonging to Solanum, dicotyledons, and all plants, respectively. The CodeML and MCMCTree programs in the PAML (v4.9)\u003csup\u003e83\u003c/sup\u003e were used to analyze amino acid substitution models and estimate divergence times. The CAFE5\u003csup\u003e84\u003c/sup\u003e was then applied to infer gene gain and loss rates in each genome. The orthogroups generated by OrthoFinder were regarded as distinct gene families and provided as inputs for CAFE5 analysis. The identified genes were subjected to GO and KEGG enrichment analysis, with the p-value of significant enrichment was set as 0.05. The syntenic analysis was performed by JCVI (v1.1.19)\u003csup\u003e85\u003c/sup\u003e. We identified synteny blocks by performing an all-against-all LAST search and chaining the hits with a distance cutoff of 20 genes. Additionally, we required each synteny block to have at least five gene pairs. The dotplot of major CBGs were shown by Gepard\u003csup\u003e86\u003c/sup\u003e. The \u003cem\u003eKs\u003c/em\u003e values of \u003cem\u003eC. annuum\u003c/em\u003e syntenic block genes were calculated using ParaAT (v2.0)\u003csup\u003e87\u003c/sup\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eMachine-learning model development\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAMYCNE\u003csup\u003e88\u003c/sup\u003e was employed to investigate copy number variations (CNVs) of putative capsaicin biosynthesis genes from whole genome re-sequencing reads, including 311 publicly available accessions of \u003cem\u003eC. annuum\u003c/em\u003e (BioProject accession: PRJCA004361) (for model training) and nine additional accessions sequenced in this study (for model testing).\u0026nbsp;To train machine-learning models for capsaicin classification, we employed a balanced dataset created by dividing samples based on the median capsaicinoid content. The input features comprised CNVs of putative capsaicinoid biosynthetic genes, while the output was categorized into high and low capsaicinoid content based on the median value. The primary reason for choosing binary classification of capsaicinoid content stems from the specific goals of our study and the characteristics of our dataset. Given the limited size of our dataset, a binary classification model proved to be more robust and less sensitive to noise. Furthermore, the binary classification framework offers a clearer and more direct interpretation for our objectives. It simplifies the understanding of gene influence on capsaicinoid presence, which is essential for practical applications such as selective breeding in agriculture. This approach aligns closely with our aim to discern the genetic factors contributing to capsaicinoid content rather than quantifying it. Before proceeding with model training, we implemented a feature selection step to refine our input features and enhance model performance. Using a Random Forest Classifier as the base estimator, we applied Recursive Feature Elimination with Cross-Validation (RFECV) to identify the most predictive features. This method involved iteratively removing features and evaluating model accuracy using Stratified 5-fold cross-validation to ensure a balanced representation of each class. The RFECV process identified an optimal subset of features, which were then used for model training.\u003c/p\u003e\n\u003cp\u003eFour primary algorithms—Random Forest, Support Vector Machines (SVM), Gradient Boosting, and K Nearest Neighbors (KNN)—were utilized to model the relationship between input features and classification outcomes. A 10-fold cross-validation approach was adopted for model evaluation, using implementations from the scikit-learn library. We applied a Grid Search approach to meticulously explore an extensive range of hyperparameters for each algorithm, aiming to identify the most effective model parameters. In the selection of kernel types for the SVM models, we focused on polynomial (poly) and linear kernels to enhance the interpretability of input features in subsequent analyses. Supplementary Table 17 provides a detailed overview of the models and their respective hyperparameter ranges. The random forest model with following parameters “max_depth=20, max_features=’sqrt’ and n_estimators=150”\u0026nbsp;exhibited the highest accuracy, approximately 72%. This configuration achieved superior performance as evidenced by the ROC curve, attaining the highest AUC value of 0.85. This underscores its proficiency in classifying capsaicinoid content based on input CNVs of significant genes. Then the importance of input feature genes for capsaicinoid content were also evaluated with this classification model.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eCENH3 ChIP-seq\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAn antigen with full peptide sequence corresponding to \u003cem\u003eC. annuum\u003c/em\u003e CENH3 was used to produce the \u003cem\u003eC. annuum\u0026nbsp;\u003c/em\u003eanti-CENH3 antibodies in rabbit. For ChIP experiment, pepper seedlings were fixed with 1% formaldehyde solution in MS buffer (10 mM potassium phosphate, pH 7.0; 50 mM NaCl) at room temperature for 15 min in a vacuum. After fixation, the seedlings were incubated at room temperature for 5 min under vacuum with 0.15 M glycine. Approximately 1 g fixed tissue was homogenized with liquid nitrogen and purifying nuclei, and resuspended in 1 ml of cell lysis buffer \u0026amp; incubate for 10 min on ice \u0026amp; spin at 1500 rpm (RC-3B, 600 × g) for 5 min (Cell lysis buffer: 10 mM Tris, 10 mM NaCl, 0.2% NP-40 [pH 8.0], 1×protease inhibitors). The cell lysis was further resuspended in 1 ml of nuclei lysis buffer for 10 min on ice (Nuclei lysis buffer: 50 mM Tris, 10 mM EDTA, 1% SDS, 1×protease inhibitors) to isolate nuclei. The resuspended chromatin solution was sonicated five times for 15 s each at\u0026nbsp;∼10% power (setting 2.5 on the sonicator, Sanyo Soniprep 150). The volume of the chromatin sample was measured and then ChIP dilution buffer was added to 1 ml of chromatin with 2.5 µg of anti-H3K4me3 and incubated for 12 h at 4℃. 50 µl protein A/G Beads were added and incubated for 4 h at 4℃. Beads were washed twice with each of the following buffers: wash buffer A (50 mM HEPES-KOH pH 7.5, 140 mM NaCl, 1 mM EDTA pH 8.0, 0.1% Na-Deoxycholate, 1% Triton X-100, 0.1% SDS), wash buffer B (50 mM HEPES-KOH pH 7.9, 500 mM NaCl, 1 mM EDTA pH 8.0, 0.1% Na-Deoxycholate, 1% Triton X-100, 0.1% SDS), wash buffer C (20 mM Tris-HCl pH8.0, 250 mM LiCl, 1 mM EDTA pH 8.0, 0.5% Na-Deoxycholate, 0.5% IGEPAL C-630, 0.1% SDS), wash buffer D (TE with 0.2% Triton X-100), and TE buffer. To purify eluted DNA, 200 μl TE was added and then RNA was degraded by the addition of 2.5 μl of 33 mg/mL RNase A (Sigma, R4642) and incubation at 37℃ for 2 hours. The DNA was then resuspended in 50 μl TE and amplified with the VAHTS® Universal DNA Library Prep Kit for Illumina V3 (Vazyme ND607). Amplified ChIP libraries were sequenced on the Illumina Novaseq 6000 platform.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eEpigenomic sequencing and data analysis\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eHi-C data was generated from leaf tissue as above mentioned,\u0026nbsp;and processed using HiC-Pro (v3.1.0)\u003csup\u003e47\u003c/sup\u003e and Juicertools(v1.22.01)\u003csup\u003e53\u003c/sup\u003e, to generate 10 kb, 15 kb, 20 kb, 25 kb, 40 kb, 100 kb and 500 kb contact maps. A/B Compartment was identified by R (v4.2.0) script using HiTC (v1.42.0)\u003csup\u003e89\u003c/sup\u003e and Cworld-dekker (v0.0.1) (https://github.com/dekkerlab/cworld-dekker) in 100 kb iced contacts matrix.\u0026nbsp;Chromatin accessibility of \u003cem\u003eC. annuum\u003c/em\u003e was profiled using ATAC-seq construction kit (Vazyme Ltd. Nanjing China) following recommended protocol. The quality of the constructed libraries was assessed using a qubit followed by Agilent Bioanalyzer 2100 for fragment analysis. The libraries were sequenced using Illumina NovoSeq 6000 platform at Novogene Inc. (Tianjin, China). Three biological replicates were generated for each plant tissue and analyzed using the same computational methods. The ATAC-seq data were analyzed using an in-house computational pipeline. Basically, the clean reads of ATAC-seq were mapped to CaT2T reference genome using BWA-MEM\u0026nbsp;(v2.2.1)\u003csup\u003e90\u003c/sup\u003e with default parameters. The alignment file (.bam) were used to call peaks by MACS2(v2.2.7.1)\u003csup\u003e91\u003c/sup\u003e. Histone modification\u0026nbsp;ChIP-seq data was downloaded from public CNGBdb database with accession number CNP0001129. Whole genome bisulfite sequencing was conducted on tissues of leaf and whole fruit from \u003cem\u003eC. annuum\u003c/em\u003e. ChIP-seq mapping and peaks calling were run with the commands, using\u0026nbsp;bowtie2\u0026nbsp;(v2.5.1)\u003csup\u003e92\u003c/sup\u003e, Samtools(v1.10)\u003csup\u003e59\u003c/sup\u003e, and\u0026nbsp;MACS2(v2.2.7.1)\u003csup\u003e91\u003c/sup\u003e. The visualization of centromeric repeats was accomplished using StainedGlass\u003csup\u003e93\u003c/sup\u003e. DNA methylation level was estimated using \u003cem\u003eBismark\u0026nbsp;\u003c/em\u003e(v0.24.0)\u003csup\u003e94\u003c/sup\u003e, after mapping WGBS (whole-genome bisulfite sequencing) data to reference genome using\u0026nbsp;BWA-MEM\u0026nbsp;(v2.2.1)\u003csup\u003e90\u003c/sup\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eTranscriptome sequencing and analysis\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTotal RNA was extracted from seven tissues, including leaf, flower, placenta, root, stem, seeds and pericarp. The mRNA was then subjected to transcriptome sequencing library construction using Illumina True-seq transcriptome kit (Illumina, CA). The libraries were then sequenced using Illumina Novaseq 6000 platform at Biomarker Technologies Corporation (QingDao, China) to generate 150 bp paired-end reads. For full-length transcriptome sequencing, about 5 µg mRNA was reverse-transcribed into full-length cDNA molecules with SMARTer™ PCR cDNA Synthesis Kit (Clontech, CA, USA) and the cDNA was further amplified by PCR. End repairing was conducted on amplified cDNAs, followed by SMRTbell adapter ligation. The ligation products are further treated by exonuclease to degrade failed ones before the Iso-seq library was sequenced using PacBio Sequal IIe instrument at Biomarker Technologies Corporation (QingDao, China). Full-length transcripts were assembled across tissues using the SMRTlink pipeline and used for guiding gene annotation. We quantified gene expression level using kallisto (v0.48.0)\u003csup\u003e95\u003c/sup\u003e. Counts for mapped reads were normalized by TPM (transcripts per million). Reads alignment was performed using HISAT2 (v2.1.0)\u003csup\u003e76\u003c/sup\u003e. To visualize the expression pattern of the focused genes among samples, heatmaps were generated using the R package.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eQuantification of capsaicinoids using HPLC-MS\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eCapsaicinoids were extracted from mature fruits of chili peppers and subjected to HPCL-MS quantification according to the previous method\u003csup\u003e96\u003c/sup\u003e. All the detections were performed on a Vanquish UHPLC system combined with a TSQ Altis MS/MS system (Thermo Scientific, USA). ACQUITY UPLC@HSS T3 column (150 mm × 2.1 mm, 1.8 μm particle size, Waters) was used for the separation of samples with the column oven at 35℃. Isocratic elution was carried out at a flow rate of 0.5 mL/min using 50% Millipore water with 0.1% formic acid and 50% acetonitrile with 0.1% formic acid. The separation was finished at 35 min with an injection volume of 15 µL and the thermostat of 30℃ for the analysis.\u0026nbsp;\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003eData Availability\u003c/p\u003e\n\u003cp\u003eThe raw sequencing data (PacBio HiFi, ONT, Illumina paired-end and Hi-C) and genome assembly have been deposited in the National Center for Biotechnology Information (NCBI) under the Bioproject (PRJNA962192). The genome assembly and annotation are also available in \u003cem\u003eCapsicum\u003c/em\u003e Genome Database (http://www.pepperbase.site/node/3) of Peking University Institute of Advanced Agricultural Sciences.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eCode Availability\u003c/p\u003e\n\u003cp\u003eAll codes were deposited at https://github.com/Weikai-47/Pepper_T2T. Any additional information required to reanalyze the data reported in this paper is available from the corresponding authors upon request.\u003c/p\u003e\n\u003cp\u003eAcknowledgments\u003c/p\u003e\n\u003cp\u003eWe would like to thank the Bioinformatics Platform at Peking University Institute of Advanced Agricultural Sciences for providing the high-performance computing resources. We would like to thank East-West Seed Group for kindly providing the double haploid pepper seeds.\u0026nbsp;This work was supported by Shandong Provincial Science and Technology Innovation Fund and the National Natural Science Foundation of China (31970317). LG is also supported by Taishan Scholars Program and Natural Science Foundation for Distinguished Young Scholars (ZR2023JQ010) of Shandong Province.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAuthor\u0026nbsp;Contributions\u003c/p\u003e\n\u003cp\u003eL.G., H.H., X.W.D. and X.Z. conceived and designed the project. Y.M., J.B., B.L. and Z.Z. maintained the plant materials and coordinated sequencing data generation. W.C. and X.F.W. assembled the genomes and performed genome validations and annotations. X.F.W., X.R.W., W.C., J.S., S.Y., D.H.A., M.Y., K.W. and L.G. conducted bioinformatic analysis and prepared figures and tables. D.M. conducted epigenome sequencing. J.L. and L.W. performed quantification of capsaicinoids and gene copy numbers. S.C. constructed the genome database. T.M. developed the machine learning model. L.Z., R.C. and J.J. assisted in the result interpretation and discussion. L.G., W.C., L.Z. and D.H.A. wrote the manuscript. L.G., X.W.D., X.Z. and H.H. revised the manuscript. All authors read and approved the final version of the manuscript.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eCompeting\u0026nbsp;Interests\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eStewart, Jr.C. et al. The Pun1 gene for pungency in pepper encodes a putative acyltransferase. \u003cem\u003ePlant J\u003c/em\u003e, \u003cstrong\u003e42\u003c/strong\u003e, 675-688 (2005).\u003c/li\u003e\n\u003cli\u003eKim, S.et al. Genome sequence of the hot pepper provides insights into the evolution of pungency in \u003cem\u003eCapsicum\u003c/em\u003e species. \u003cem\u003eNat. Genet. \u003c/em\u003e\u003cstrong\u003e46\u003c/strong\u003e, 270-278 (2014).\u003c/li\u003e\n\u003cli\u003eStewart, Jr.C. et al. Genetic control of pungency in \u003cem\u003eC. chinense \u003c/em\u003evia the \u003cem\u003ePun1\u003c/em\u003e locus. \u003cem\u003eJ. Exp. Bot. \u003c/em\u003e\u003cstrong\u003e58\u003c/strong\u003e, 979-991 (2007).\u003c/li\u003e\n\u003cli\u003eLiao, Y.et al. The 3D architecture of the pepper genome and its relationship to function and evolution. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 3479 (2022).\u003c/li\u003e\n\u003cli\u003eShirasawa, K., Hosokawa, M., Yasui, Y., Toyoda, A., \u0026amp; Isobe, S. Chromosome-scale genome assembly of a Japanese chili pepper landrace, \u003cem\u003eCapsicum annuum\u003c/em\u003e \u0026lsquo;Takanotsume\u0026rsquo;.\u003cem\u003e DNA Res. \u003c/em\u003e\u003cstrong\u003e30\u003c/strong\u003e, dsac052 (2023).\u003c/li\u003e\n\u003cli\u003eLee, J.H. et al. High-quality chromosome-scale genomes facilitate effective identification of large structural variations in hot and sweet peppers. \u003cem\u003eHortic. Res.\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, uhac210 (2022).\u003c/li\u003e\n\u003cli\u003eQin, C. et al. Whole-genome sequencing of cultivated and wild peppers provides insights into \u003cem\u003eCapsicum\u003c/em\u003e domestication and specialization. \u003cem\u003eProc. Natl. Acad. Sci. USA\u003c/em\u003e \u003cstrong\u003e111\u003c/strong\u003e, 5135-5140 (2014).\u003c/li\u003e\n\u003cli\u003eKim, S. et al. New reference genome sequences of hot pepper reveal the massive evolution of plant disease-resistance genes by retroduplication. \u003cem\u003eGenome Biol.\u003c/em\u003e \u003cstrong\u003e18\u003c/strong\u003e, 210 (2017).\u003c/li\u003e\n\u003cli\u003eHulse-Kemp, A.M. et al. Reference quality assembly of the 3.5-Gb genome of \u003cem\u003eCapsicum annuum\u003c/em\u003e from a single linked-read library. \u003cem\u003eHortic. Res.\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, 4 (2018).\u003c/li\u003e\n\u003cli\u003eKim, M.S. et al. Comparative analysis of \u003cem\u003ede novo\u003c/em\u003e genomes reveals dynamic intra-species divergence of NLRs in pepper.\u003cem\u003e BMC Plant Biol. \u003c/em\u003e\u003cstrong\u003e21\u003c/strong\u003e, 247 (2021).\u003c/li\u003e\n\u003cli\u003eLiu, F. et al. Genomes of cultivated and wild \u003cem\u003eCapsicum\u003c/em\u003e species provide insights into pepper domestication and population differentiation. \u003cem\u003eNat. Commun\u003c/em\u003e, \u003cstrong\u003e14\u003c/strong\u003e, 5487 (2023).\u003c/li\u003e\n\u003cli\u003eNurk, S.et al. The complete sequence of a human genome. \u003cem\u003eScience\u003c/em\u003e \u003cstrong\u003e376\u003c/strong\u003e, 44-53 (2022).\u003c/li\u003e\n\u003cli\u003eAltemose, N. et al. Complete genomic and epigenetic maps of human centromeres. \u003cem\u003eScience\u003c/em\u003e \u003cstrong\u003e376\u003c/strong\u003e, eabl4178 (2022).\u003c/li\u003e\n\u003cli\u003eGershman, A.et al. Epigenetic patterns in a complete human genome. \u003cem\u003eScience\u003c/em\u003e \u003cstrong\u003e376\u003c/strong\u003e, eabj5089 (2022).\u003c/li\u003e\n\u003cli\u003eAganezov, S. et al. A complete reference genome improves analysis of human genetic variation. \u003cem\u003eScience\u003c/em\u003e \u003cstrong\u003e376\u003c/strong\u003e, eabl3533 (2022).\u003c/li\u003e\n\u003cli\u003eNaish, M.et al. The genetic and epigenetic landscape of the \u003cem\u003eArabidopsis\u003c/em\u003e centromeres. \u003cem\u003eScience\u003c/em\u003e \u003cstrong\u003e374\u003c/strong\u003e, eabi7489 (2021).\u003c/li\u003e\n\u003cli\u003eHou, X., Wang, D., Cheng, Z., Wang, Y., \u0026amp; Jiao, Y. A near-complete assembly of an \u003cem\u003eArabidopsis thaliana\u003c/em\u003e genome. \u003cem\u003eMol. Plant \u003c/em\u003e\u003cstrong\u003e15\u003c/strong\u003e, 1247-1250 (2022).\u003c/li\u003e\n\u003cli\u003eWang, B.et al. High-quality \u003cem\u003eArabidopsis thaliana\u003c/em\u003e genome assembly with Nanopore and HiFi long reads. \u003cem\u003eGenom. Proteom. Bioinf. \u003c/em\u003e\u003cstrong\u003e20\u003c/strong\u003e, 4-13 (2022).\u003c/li\u003e\n\u003cli\u003eSong, J.et al. Two gap-free reference genomes and a global view of the centromere architecture in rice. \u003cem\u003eMol. Plant\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 1757-1767 (2021). \u003c/li\u003e\n\u003cli\u003eShang, L. et al. A complete assembly of the rice Nipponbare reference genome. \u003cem\u003eMol. Plant \u003c/em\u003e\u003cstrong\u003e16\u003c/strong\u003e, 1232-1236 (2023).\u003c/li\u003e\n\u003cli\u003eYang, X. et al. The gap-free potato genome assembly reveals large tandem gene clusters of agronomical importance in highly repeated genomic regions. \u003cem\u003eMol. Plant\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 314-317. (2023).\u003c/li\u003e\n\u003cli\u003eWang, L. et al. A telomere-to-telomere gap-free assembly of soybean genome. \u003cem\u003eMol. Plant\u003c/em\u003e (2023).\u003c/li\u003e\n\u003cli\u003eChen, J. et al. A complete telomere-to-telomere assembly of the maize genome. \u003cem\u003eNature Genet\u003c/em\u003e, 1-11 (2023).\u003c/li\u003e\n\u003cli\u003eCheng, H., Concepcion, G.T., Feng, X., Zhang, H., \u0026amp; Li, H. Haplotype-resolved \u003cem\u003ede novo\u003c/em\u003e assembly using phased assembly graphs with hifiasm. \u003cem\u003eNat. Methods \u003c/em\u003e\u003cstrong\u003e18\u003c/strong\u003e, 170-175 (2021).\u003c/li\u003e\n\u003cli\u003eHu, J.et al. An efficient error correction and accurate assembly tool for noisy long reads. Preprint at BioRxiv. (2023).\u003c/li\u003e\n\u003cli\u003eDurand, N.C. et al. Juicebox provides a visualization system for Hi-C contact maps with unlimited zoom. \u003cem\u003eCell Syst.\u003c/em\u003e \u003cstrong\u003e3\u003c/strong\u003e, 99-101 (2016).\u003c/li\u003e\n\u003cli\u003eZhang, Z.X. et al. Discovery of putative capsaicin biosynthetic genes by RNA-Seq and digital gene expression analysis of pepper. \u003cem\u003eSci. Rep.\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, 34121 (2016).\u003c/li\u003e\n\u003cli\u003eCleveland, D.W., Mao, Y., \u0026amp; Sullivan, K.F. Centromeres and kinetochores: from epigenetics to mitotic checkpoint signaling. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e112\u003c/strong\u003e, 407-421 (2003).\u003c/li\u003e\n\u003cli\u003eZhang, H. et al. Boom-bust turnovers of megabase-sized centromeric DNA in \u003cem\u003eSolanum\u003c/em\u003e species: Rapid evolution of DNA sequences associated with centromeres.\u003cem\u003e Plant Cell\u003c/em\u003e \u003cstrong\u003e26\u003c/strong\u003e, 1436-1447 (2014).\u003c/li\u003e\n\u003cli\u003eAhmed, H.I. et al. Einkorn genomics sheds light on history of the oldest domesticated wheat. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e620\u003c/strong\u003e, 830-838 (2023).\u003c/li\u003e\n\u003cli\u003eYang, Z. et al. Cotton D genome assemblies built with long-read data unveil mechanisms of centromere evolution and stress tolerance divergence. \u003cem\u003eBMC Biol\u003c/em\u003e. \u003cstrong\u003e19\u003c/strong\u003e, 1-22 (2021).\u003c/li\u003e\n\u003cli\u003eVitte, C., \u0026amp; Panaud, O. Formation of solo-LTRs through unequal homologous recombination counterbalances amplifications of LTR retrotransposons in rice \u003cem\u003eOryza sativa\u003c/em\u003e L. \u003cem\u003eMol. Biol. Evol.\u003c/em\u003e \u003cstrong\u003e20\u003c/strong\u003e, 528-540 (2003).\u003c/li\u003e\n\u003cli\u003eNeumann, P. et al. Plant centromeric retrotransposons: a structural and cytogenetic perspective. \u003cem\u003eMobile DNA\u003c/em\u003e \u003cstrong\u003e2\u003c/strong\u003e, 4 (2011).\u003c/li\u003e\n\u003cli\u003eWlodzimierz, P. et al. Cycles of satellite and transposon evolution in \u003cem\u003eArabidopsis\u003c/em\u003e centromeres. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e618\u003c/strong\u003e, 557-565 (2023).\u003c/li\u003e\n\u003cli\u003eEmms, D.M., \u0026amp; Kelly, S. OrthoFinder: phylogenetic orthology inference for comparative genomics.\u003cem\u003e Genome Biol.\u003c/em\u003e \u003cstrong\u003e20\u003c/strong\u003e, 238 (2019).\u003c/li\u003e\n\u003cli\u003eZhu, Z.et al. Natural variations in the MYB transcription factor MYB31 determine the evolution of extremely pungent peppers. \u003cem\u003eNew Phytol.\u003c/em\u003e \u003cstrong\u003e223\u003c/strong\u003e, 922-938 (2019).\u003c/li\u003e\n\u003cli\u003eSun, B. et al. Coexpression network analysis reveals an MYB transcriptional activator involved in capsaicinoid biosynthesis in hot peppers. \u003cem\u003eHortic Res\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, 162 (2020).\u003c/li\u003e\n\u003cli\u003eCao, Y.et al. Pepper variome reveals the history and key loci associated with fruit domestication and diversification. \u003cem\u003eMol. Plant \u003c/em\u003e\u003cstrong\u003e15\u003c/strong\u003e, 1744-1758 (2022).\u003c/li\u003e\n\u003cli\u003eWang, S.et al. The OsSPL16-GW7 regulatory module determines grain shape and simultaneously improves rice yield and grain quality. \u003cem\u003eNat. Genet.\u003c/em\u003e \u003cstrong\u003e47\u003c/strong\u003e, 949-954 (2015). \u003c/li\u003e\n\u003cli\u003eShirai, K.et al. A highly specific genome-wide association study integrated with transcriptome data reveals the contribution of copy number variations to specialized metabolites in \u003cem\u003eArabidopsis thaliana\u003c/em\u003e accessions. \u003cem\u003eMol. Biol. Evol. \u003c/em\u003e\u003cstrong\u003e34\u003c/strong\u003e, 3111-3122 (2017).\u003c/li\u003e\n\u003cli\u003eCarrizo, G.C. et al. Phylogenetic relationships, diversification and expansion of chili peppers (\u003cem\u003eCapsicum\u003c/em\u003e, Solanaceae). \u003cem\u003eAnn. Bot.\u003c/em\u003e \u003cstrong\u003e118\u003c/strong\u003e, 35-51 (2016).\u003c/li\u003e\n\u003cli\u003eGuo, L. et al. The opium poppy genome and morphinan production. \u003cem\u003eScience\u003c/em\u003e \u003cstrong\u003e362\u003c/strong\u003e, 343-347 (2018).\u003c/li\u003e\n\u003cli\u003eHuang, A.C. et al. A specialized metabolic network selectively modulates \u003cem\u003eArabidopsis\u003c/em\u003e root microbiota. \u003cem\u003eScience\u003c/em\u003e \u003cstrong\u003e364\u003c/strong\u003e, eaau6389 (2019).\u003c/li\u003e\n\u003cli\u003eNett, R.S., Lau, W., \u0026amp; Sattely, E.S. Discovery and engineering of colchicine alkaloid biosynthesis. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e584\u003c/strong\u003e, 148-153 (2020).\u003c/li\u003e\n\u003cli\u003eHe, J. et al. Establishing \u003cem\u003ePhysalis\u003c/em\u003e as a \u003cem\u003eSolanaceae\u003c/em\u003e model system enables genetic reevaluation of the inflated calyx syndrome. \u003cem\u003ePlant Cell\u003c/em\u003e \u003cstrong\u003e35\u003c/strong\u003e, 351-368 (2023).\u003c/li\u003e\n\u003cli\u003eBelton, J.M. et al. Hi-C: a comprehensive technique to capture the conformation of genomes. \u003cem\u003eMethods\u003c/em\u003e \u003cstrong\u003e58\u003c/strong\u003e, 268-276 (2012).\u003c/li\u003e\n\u003cli\u003eServant, N. et al. HiC-Pro: an optimized and flexible pipeline for Hi-C data processing. \u003cem\u003eGenome Biol.\u003c/em\u003e 16, 259 (2015).\u003c/li\u003e\n\u003cli\u003eMarcais, G., \u0026amp; Kingsford, C. A fast, lock-free approach for efficient parallel counting of occurrences of \u003cem\u003ek\u003c/em\u003e-mers. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e27\u003c/strong\u003e, 764-770 (2011).\u003c/li\u003e\n\u003cli\u003eVurture, G.W. et al. GenomeScope: fast reference-free genome profiling from short reads. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e33\u003c/strong\u003e, 2202-2204 (2017).\u003c/li\u003e\n\u003cli\u003eHu, J., Fan, J., Sun, Z., \u0026amp; Liu, S. NextPolish: a fast and efficient genome polishing tool for long-read assembly. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e36\u003c/strong\u003e, 2253-2255 (2020).\u003c/li\u003e\n\u003cli\u003eLi, H. Minimap2: pairwise alignment for nucleotide sequences. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e34\u003c/strong\u003e, 3094-3100 (2018).\u003c/li\u003e\n\u003cli\u003eChakraborty, M., Baldwin-Brown, J.G., Long, A.D., \u0026amp; Emerson, J.J. Contiguous and accurate de novo assembly of metazoan genomes with modest long read coverage. \u003cem\u003eNucleic Acids Res.\u003c/em\u003e \u003cstrong\u003e44\u003c/strong\u003e, e147 (2016).\u003c/li\u003e\n\u003cli\u003eDurand, N.C.et al. Juicer provides a one-click system for analyzing loop-resolution Hi-C experiments. \u003cem\u003eCell Syst\u003c/em\u003e \u003cstrong\u003e3\u003c/strong\u003e, 95-98 (2016).\u003c/li\u003e\n\u003cli\u003eDudchenko, O. et al. De novo assembly of the\u003cem\u003e Aedes aegypti\u003c/em\u003e genome using Hi-C yields chromosome-length scaffolds. \u003cem\u003eScience\u003c/em\u003e \u003cstrong\u003e356\u003c/strong\u003e, 92-95 (2017).\u003c/li\u003e\n\u003cli\u003eBzikadze, A.V., \u0026amp; Pevzner, PA. Automated assembly of centromeres from ultra-long error-prone reads. \u003cem\u003eNat. Biotechnol. \u003c/em\u003e\u003cstrong\u003e38\u003c/strong\u003e, 1309-1316 (2020).\u003c/li\u003e\n\u003cli\u003eJain, C., Rhie, A., Hansen, N.F., Koren, S., \u0026amp; Phillippy, A.M. Long-read mapping to repetitive reference sequences using Winnowmap2. \u003cem\u003eNat. Methods\u003c/em\u003e \u003cstrong\u003e19\u003c/strong\u003e, 705-710 (2022).\u003c/li\u003e\n\u003cli\u003eCabanettes, F., \u0026amp; Klopp, C. D-GENIES: dot plot large genomes in an interactive, efficient and simple way. \u003cem\u003ePeerJ\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, e4958 (2018).\u003c/li\u003e\n\u003cli\u003eLi, H., \u0026amp; Durbin, R. Fast and accurate short read alignment with Burrows-Wheeler transform. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e25\u003c/strong\u003e, 1754-1760 (2009).\u003c/li\u003e\n\u003cli\u003eLi, H. et al. The sequence alignment/map format and SAMtools. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e25\u003c/strong\u003e, 2078-2079 (2009).\u003c/li\u003e\n\u003cli\u003eRobinson, J.T. et al. Integrative genomics viewer. \u003cem\u003eNat. Biotechnol.\u003c/em\u003e \u003cstrong\u003e29\u003c/strong\u003e, 24-26 (2011).\u003c/li\u003e\n\u003cli\u003eManni, M., Berkeley, M.R., Seppey, M., Simao, F.A., \u0026amp; Zdobnov, EM. BUSCO Update: novel and streamlined workflows along with broader and deeper phylogenetic coverage for scoring of eukaryotic, prokaryotic, and viral genomes. \u003cem\u003eMol. Biol. Evol. \u003c/em\u003e\u003cstrong\u003e38\u003c/strong\u003e, 4647-4654 (2021).\u003c/li\u003e\n\u003cli\u003eRhie, A., Walenz, B.P., Koren, S., \u0026amp; Phillippy, A.M. Merqury: reference-free quality, completeness, and phasing assessment for genome assemblies.\u003cem\u003e Genome Biol.\u003c/em\u003e \u003cstrong\u003e21\u003c/strong\u003e, 245 (2020).\u003c/li\u003e\n\u003cli\u003eBenson, G. Tandem repeats finder: a program to analyze DNA sequences.\u003cem\u003e Nucleic Acids Res.\u003c/em\u003e \u003cstrong\u003e27\u003c/strong\u003e, 573-580 (1999).\u003c/li\u003e\n\u003cli\u003eTarailo-Graovac, M., \u0026amp; Chen, N. Using RepeatMasker to identify repetitive elements in genomic sequences. Curr. Protoc. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, 4-10 (2009).\u003c/li\u003e\n\u003cli\u003eXu, Z., \u0026amp; Wang, H. LTR_FINDER: an efficient tool for the prediction of full-length LTR retrotransposons. \u003cem\u003eNucleic Acids Res.\u003c/em\u003e \u003cstrong\u003e35\u003c/strong\u003e, W265-W268 (2007).\u003c/li\u003e\n\u003cli\u003eEllinghaus, D., Kurtz, S., \u0026amp; Willhoeft, U. LTRharvest, an efficient and flexible software for de novo detection of LTR retrotransposons.\u003cem\u003e BMC Bioinformatics \u003c/em\u003e\u003cstrong\u003e9\u003c/strong\u003e, 18 (2008).\u003c/li\u003e\n\u003cli\u003eOu, S., \u0026amp; Jiang, N. LTR_retriever: a highly accurate and sensitive program for identification of long terminal repeat retrotransposons. Plant Physiol. 176, 1410-1422 (2018).\u003c/li\u003e\n\u003cli\u003eJin, Y., Tam, O.H., Paniagua, E., \u0026amp; Hammell, M. TEtranscripts: a package for including transposable elements in differential expression analysis of RNA-seq datasets. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e31\u003c/strong\u003e, 3593-3599 (2015).\u003c/li\u003e\n\u003cli\u003eZhang, R.G. et al. TEsorter: an accurate and fast method to classify LTR-retrotransposons in plant genomes. \u003cem\u003eHortic. Res.\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, uhac017 (2022).\u003c/li\u003e\n\u003cli\u003eCantarel, B.L. et al. MAKER: an easy-to-use annotation pipeline designed for emerging model organism genomes. \u003cem\u003eGenome Res.\u003c/em\u003e \u003cstrong\u003e18\u003c/strong\u003e, 188-196 (2008).\u003c/li\u003e\n\u003cli\u003ePertea, M., et al. StringTie enables improved reconstruction of a transcriptome from RNA-seq reads. \u003cem\u003eNat. Biotechnol. \u003c/em\u003e\u003cstrong\u003e33\u003c/strong\u003e, 290-295 (2015).\u003c/li\u003e\n\u003cli\u003eSlater, G.S., \u0026amp; Birney, E. Automated generation of heuristics for biological sequence comparison.\u003cem\u003e BMC Bioinformatics\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, 31 (2005).\u003c/li\u003e\n\u003cli\u003eEilbeck, K., Moore, B., Holt, C., \u0026amp; Yandell, M. Quantitative measures for the management and comparison of annotated genomes. \u003cem\u003eBMC Bioinformatics \u003c/em\u003e\u003cstrong\u003e10\u003c/strong\u003e, 67 (2009).\u003c/li\u003e\n\u003cli\u003eKorf, I. Gene finding in novel genomes. \u003cem\u003eBMC Bioinformatics\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, 59 (2004).\u003c/li\u003e\n\u003cli\u003eBrůna, T., Hoff, K.J., Lomsadze, A., Stanke, M., \u0026amp; Borodovsky, M. BRAKER2: automatic eukaryotic genome annotation with GeneMark-EP+ and AUGUSTUS supported by a protein database. \u003cem\u003eNAR Genom. Bioinform. \u003c/em\u003e\u003cstrong\u003e3\u003c/strong\u003e, lqaa108 (2021).\u003c/li\u003e\n\u003cli\u003eKim, D., Langmead, B., \u0026amp; Salzberg, S.L. HISAT: a fast spliced aligner with low memory requirements. \u003cem\u003eNat. Methods \u003c/em\u003e\u003cstrong\u003e12\u003c/strong\u003e, 357-360 (2015).\u003c/li\u003e\n\u003cli\u003eLomsadze, A., Ter-Hovhannisyan, V., Chernoff, Y.O., \u0026amp; Borodovsky, M. Gene identification in novel eukaryotic genomes by self-training algorithm. \u003cem\u003eNucleic Acids Res. \u003c/em\u003e\u003cstrong\u003e33\u003c/strong\u003e, 6494-6506 (2005).\u003c/li\u003e\n\u003cli\u003eStanke, M., Tzvetkova, A., \u0026amp; Morgenstern, B. AUGUSTUS at EGASP: using EST, protein and genomic alignments for improved gene prediction in the human genome. \u003cem\u003eGenome Biol. \u003c/em\u003e\u003cstrong\u003e7\u003c/strong\u003e, S11 (2006).\u003c/li\u003e\n\u003cli\u003eShumate, A., \u0026amp; Salzberg, S.L. Liftoff: accurate mapping of gene annotations. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e37\u003c/strong\u003e, 1639-1643 (2021).\u003c/li\u003e\n\u003cli\u003ePertea, G., \u0026amp; Pertea, M. GFF Utilities: GffRead and GffCompare. \u003cem\u003eF1000Res\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e. (2020).\u003c/li\u003e\n\u003cli\u003eCapella-Guti\u0026eacute;rrez, S., Silla-Mart\u0026iacute;nez, J.M., \u0026amp; Gabald\u0026oacute;n, T. trimAl: a tool for automated alignment trimming in large-scale phylogenetic analyses. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e25\u003c/strong\u003e, 1972-1973 (2009).\u003c/li\u003e\n\u003cli\u003eStamatakis, A. RAxML version 8: a tool for phylogenetic analysis and post-analysis of large phylogenies. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e, 1312-1313 (2014).\u003c/li\u003e\n\u003cli\u003eYang, Z. PAML 4: phylogenetic analysis by maximum likelihood. \u003cem\u003eMol. Biol. Evol. \u003c/em\u003e\u003cstrong\u003e24\u003c/strong\u003e, 1586-1591 (2007).\u003c/li\u003e\n\u003cli\u003eMendes, F.K., Vanderpool, D., Fulton, B., \u0026amp; Hahn, M.W. CAFE 5 models variation in evolutionary rates among gene families. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e36\u003c/strong\u003e, 5516-5518 (2021).\u003c/li\u003e\n\u003cli\u003eTang, H., Krishnakumar, V., \u0026amp; Li, J. jcvi: JCVI utility libraries. Zenodo. (2015).\u003c/li\u003e\n\u003cli\u003eKrumsiek, J., Arnold, R., \u0026amp; Rattei, T. Gepard: a rapid and sensitive tool for creating dotplots on genome scale. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e23\u003c/strong\u003e, 1026-1028 (2007).\u003c/li\u003e\n\u003cli\u003eZhang, Z. et al. ParaAT: a parallel tool for constructing multiple protein-coding DNA alignments. \u003cem\u003eBiochem. Biophys. Res. Commun. \u003c/em\u003e\u003cstrong\u003e419\u003c/strong\u003e, 779-781 (2012).\u003c/li\u003e\n\u003cli\u003eEisfeldt, J., Nilsson, D., Andersson-Assarsson, J. C., \u0026amp; Lindstrand, A. AMYCNE: Confident copy number assessment using whole genome sequencing data. \u003cem\u003ePLoS One\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, e0189710 (2018).\u003c/li\u003e\n\u003cli\u003eServant, N. et al. HiTC: exploration of high-throughput \u0026lsquo;C\u0026rsquo; experiments. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e28\u003c/strong\u003e, 2843-2844 (2012).\u003c/li\u003e\n\u003cli\u003eLi, H. Aligning sequence reads, clone sequences and assembly contigs with BWA-MEM. \u003cem\u003earXiv\u003c/em\u003e \u003cstrong\u003e1303\u003c/strong\u003e, 3997 (2013).\u003c/li\u003e\n\u003cli\u003eZhang, Y.et al. Model-based analysis of ChIP-Seq (MACS). \u003cem\u003eGenome Biol.\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, R137. (2008).\u003c/li\u003e\n\u003cli\u003eLangmead, B., \u0026amp; Salzberg, S.L. Fast gapped-read alignment with Bowtie 2. \u003cem\u003eNat. Methods\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 357-359 (2012).\u003c/li\u003e\n\u003cli\u003eVollger, M.R. et al. StainedGlass: Interactive visualization of massive tandem repeat structures with identity heatmaps. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e38\u003c/strong\u003e, 2049-2051 (2022).\u003c/li\u003e\n\u003cli\u003eKrueger, F., \u0026amp; Andrews, S.R. Bismark: a flexible aligner and methylation caller for Bisulfite-Seq applications. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e27\u003c/strong\u003e, 1571-1572 (2011).\u003c/li\u003e\n\u003cli\u003eBray, N.L., Pimentel, H., Melsted, P., \u0026amp; Pachter, L. Near-optimal probabilistic RNA-seq quantification. \u003cem\u003eNat. Biotechnol. \u003c/em\u003e\u003cstrong\u003e34\u003c/strong\u003e, 525-527 (2016).\u003c/li\u003e\n\u003cli\u003eDaood, H.G. et al. HPLC determination of capsaicinoids with cross-linked C18 column and buffer-free eluent. \u003cem\u003eJ. Chromatogr. Sci.\u003c/em\u003e \u003cstrong\u003e53\u003c/strong\u003e, 135-143 (2015).\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Table","content":"\u003cp\u003e\u003cstrong\u003eTable 1 | Statistics for genome assembly and annotation of two pepper species\u003c/strong\u003e\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"609\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd width=\"38.423645320197046%\"\u003e\n \u003cp\u003eGenomic feature\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"31.362889983579638%\"\u003e\n \u003cp\u003e\u003cem\u003eCapsicum annuum\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"30.213464696223316%\"\u003e\n \u003cp\u003e\u003cem\u003eCapsicum rhomboideum\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"38.423645320197046%\"\u003e\n \u003cp\u003eNumber of contigs\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"31.362889983579638%\"\u003e\n \u003cp\u003e12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"30.213464696223316%\"\u003e\n \u003cp\u003e13\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"38.423645320197046%\"\u003e\n \u003cp\u003eTotal length (bp)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"31.362889983579638%\"\u003e\n \u003cp\u003e3,103,116,129\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"30.213464696223316%\"\u003e\n \u003cp\u003e1,707,653,203\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"38.423645320197046%\"\u003e\n \u003cp\u003eContig N50 (bp)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"31.362889983579638%\"\u003e\n \u003cp\u003e262,573,928\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"30.213464696223316%\"\u003e\n \u003cp\u003e145,987,823\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"38.423645320197046%\"\u003e\n \u003cp\u003eNumber of gaps\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"31.362889983579638%\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"30.213464696223316%\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"38.423645320197046%\"\u003e\n \u003cp\u003eNumber of telomeres\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"31.362889983579638%\"\u003e\n \u003cp\u003e24\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"30.213464696223316%\"\u003e\n \u003cp\u003e17\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"38.423645320197046%\"\u003e\n \u003cp\u003eNumber of centromeres\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"31.362889983579638%\"\u003e\n \u003cp\u003e12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"30.213464696223316%\"\u003e\n \u003cp\u003e13\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"38.423645320197046%\"\u003e\n \u003cp\u003eNumber of gene models\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"31.362889983579638%\"\u003e\n \u003cp\u003e34,428\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"30.213464696223316%\"\u003e\n \u003cp\u003e33,512\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"38.423645320197046%\"\u003e\n \u003cp\u003eGC content (%)\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"31.362889983579638%\"\u003e\n \u003cp\u003e35.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"30.213464696223316%\"\u003e\n \u003cp\u003e36.31\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"38.423645320197046%\"\u003e\n \u003cp\u003eRepeat content (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"31.362889983579638%\"\u003e\n \u003cp\u003e79.50\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"30.213464696223316%\"\u003e\n \u003cp\u003e74.64\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"38.423645320197046%\"\u003e\n \u003cp\u003eAssembly BUSCOs (%)\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"31.362889983579638%\"\u003e\n \u003cp\u003e98.62\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"30.213464696223316%\"\u003e\n \u003cp\u003e97.12\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"38.423645320197046%\"\u003e\n \u003cp\u003eAnnotation BUSCOs (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"31.362889983579638%\"\u003e\n \u003cp\u003e97.04\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"30.213464696223316%\"\u003e\n \u003cp\u003e93.23\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"38.423645320197046%\"\u003e\n \u003cp\u003eQV\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"31.362889983579638%\"\u003e\n \u003cp\u003e56.60\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"30.213464696223316%\"\u003e\n \u003cp\u003e77.18\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"38.423645320197046%\"\u003e\n \u003cp\u003eCompleteness (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"31.362889983579638%\"\u003e\n \u003cp\u003e96.49\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"30.213464696223316%\"\u003e\n \u003cp\u003e98.11\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"nature-portfolio","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"","title":"Nature Portfolio","twitterHandle":"","acdcEnabled":false,"dfaEnabled":false,"editorialSystem":"ejp","reportingPortfolio":"","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-3062426/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-3062426/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eChili pepper (\u003cem\u003eCapsicum\u003c/em\u003e), one of oldest domesticated crops in human history, is best known for its unique fruit pungency given by capsaicinoids. The evolutionary history of capsaicinoid biosynthesis and the mechanism of tissue-specificity remain obscured due to the lack of high-quality \u003cem\u003eCapsicum \u003c/em\u003egenomes. Here, we assembled two telomere-to-telomere (T2T) gap-free genomes for \u003cem\u003eC. annuum\u003c/em\u003e and its wild non-pungent relative \u003cem\u003eC. rhomboideum\u003c/em\u003e to dissect the evolution of fruit pungency in chili peppers. We precisely delineated \u003cem\u003eCapsicum \u003c/em\u003ecentromeres which, unlike Arabidopsis and human centromeres, lacked high-copy centromeric tandem repeats but were extensively invaded by CRM LTR retrotransposons. Phylogenomics estimated the evolutionary timings of capsaicinoid biosynthesis, and revealed eroded coding and regulatory regions of key biosynthesis genes in non-pungent species such as \u003cem\u003eC. rhomboideum\u003c/em\u003e. We also found conserved placenta-specific accessible chromatin regions likely allowing tissue-specific biosynthesis gene co-regulation. Finally, from detected copy number variants of capsaicinoid biosynthesis genes using the \u003cem\u003eC. annuum\u003c/em\u003e T2T genome as reference, we trained a machine-learning model to predict pungency for genomic selection. The new T2T genomic resources will accelerate chili pepper breeding, metabolic engineering and enrich our understanding of \u003cem\u003eCapsicum \u003c/em\u003ebiology and genome evolution.\u003c/p\u003e","manuscriptTitle":"Two telomere-to-telomere gapless genomes reveal insights into Capsicum evolution and capsaicinoid biosynthesis","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-03-04 06:21:55","doi":"10.21203/rs.3.rs-3062426/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"nature-communications","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"NCOMMS","sideBox":"Learn more about [Nature Communications](http://www.nature.com/ncomms/)","snPcode":"","submissionUrl":"https://mts-ncomms.nature.com/","title":"Nature Communications","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature Communications","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"2476c00b-1e5c-4e46-a2a1-84281ed5f6b1","owner":[],"postedDate":"March 4th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":28848229,"name":"Biological sciences/Plant sciences/Plant evolution"},{"id":28848230,"name":"Biological sciences/Genetics/Genomics/Comparative genomics"},{"id":28848231,"name":"Biological sciences/Plant sciences/Secondary metabolism"}],"tags":[],"updatedAt":"2024-05-21T07:07:39+00:00","versionOfRecord":{"articleIdentity":"rs-3062426","link":"https://doi.org/10.1038/s41467-024-48643-0","journal":{"identity":"nature-communications","isVorOnly":false,"title":"Nature Communications"},"publishedOn":"2024-05-20 04:00:00","publishedOnDateReadable":"May 20th, 2024"},"versionCreatedAt":"2024-03-04 06:21:55","video":"","vorDoi":"10.1038/s41467-024-48643-0","vorDoiUrl":"https://doi.org/10.1038/s41467-024-48643-0","workflowStages":[]},"version":"v1","identity":"rs-3062426","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-3062426","identity":"rs-3062426","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}