‘Chambourcin’ leaf material was obtained from a 12-year-old experimental vineyard located at the University of Missouri Southwest Research Station in Mount Vernon, Missouri, USA. For PacBio HiFi sequencing, high molecular weight (HMW) DNA was isolated using the Nucleobond Kit (Macherey-Nagel, Bethlehem, PA, USA) following the manufacturer’s protocol. Approximately 20 cg DNA was sheared to a center of mass of 10–20 kilobase (kb) in a Megaruptor 3 system. Next, a HiFi sequencing library was constructed following HiFi SMRTbell protocols for the Express Template Prep Kit 2.0 according to manufacturers’ recommendations (Pacific Biosciences, California). The library was sequenced using Sequel binding and sequencing chemistry v2.0 in circular consensus sequencing (CCS) mode in a Sequel II system with a movie collection (file format of HiFi data) time of 30 h. The HiFi reads were generated with the CCS mode of pbtools [12] using a minimum Predicted Accuracy of 0.990.

For the Bionano data, DNA was isolated from fresh young leaf tissue from a 12-year-old experimental vineyard located at the University of Missouri Southwest Research Station in Mount Vernon, Missouri, USA, using the Prep™ Plant DNA Isolation kit and labeled using the Bionano Prep™ DNA Labeling Kit Direct Label and Stain (DLS) (Bionano Genomics, San Diego, California). In total, 500 ng of ultra-high molecular weight (UHMW) DNA was used for the DLS reaction. DNA was incubated in the presence of DLE-1 Enzyme, DL-Green, and DLE-1 Buffer for 3:20 h at 37 °C. This was followed by proteinase K digestion at 50 °C for 30 min and double cleanup of the unincorporated DL-Green label. The resulting DLS sample was combined with the Flow Buffer, dithiothreitol (DTT) and DNA stain, mixed at slow speed in a rotator mixer for an hour, and then incubated overnight at 4 °C. The labeled sample was then loaded onto a Bionano flow cell in a Saphyr System for separation, imaging, and creation of digital molecules according to the manufacturer’s recommendations [13]. The raw molecule set was filtered to a molecule length of 250 kb and a minimum of nine CTTAAG labels per molecule. Bionano maps were assembled without pre-assembly using the non-haplotype parameters with no Complex Multi-Path Regions (CMPR) cut and without extend-split. Bionano software (Solve, Tools and Access, v1.5.1) [14] was used for data visualization, processing, and assembly of Bionano maps. The PacBio HiFi and Bionano sequencing were done at Corteva Agriscience, Johnston, Iowa, USA.

For the Illumina whole genome data, DNA was extracted from ‘Chambourcin’ leaf tissue collected from the USDA Grape Germplasm Collection located in Geneva, New York. DNA was extracted using Qiagen DNeasy Plant Mini Kits (Qiagen, Valencia, California, USA) and assessed for purity and concentration using a NanoDrop spectrophotometer and Qubit fluorometer. DNA was cleaned using the Qiagen Dneasy PowerClean Pro Cleanup Kit. DNA libraries were prepared and shotgun Illumina sequenced at Novogene (San Diego, California, USA) with paired-end 150 nt reads with 40X coverage. The raw Illumina reads were trimmed with Trimmomatic (v0.39; RRID:SCR_011848) [15] using HEADCROP:4 MINLEN:70 parameters.

The PacBio HiFi reads and 19 nt k-mers were used to estimate the genome heterozygosity using jellyfish (v2.3.0; RRID:SCR_005491) [16]. The resulting “.histo” file was visualized with GenomeScope (RRID:SCR_017014) [17].

The PacBio HiFi assembly was generated using the Hifiasm assembler (RRID:SCR_021069; v0.13-r308) [18] with default parameters. To reduce the number of small, low-coverage artifactual contigs often generated by Hifiasm [18], the assembly was filtered to exclude less than 70,000 bp contigs. The resulting HiFi contigs were merged to the DLS Bionano maps with Bionano Solve (v3.5.1) [14] using the hybridscaffold.pl script of Bionano Solve (v3.5.1) [14] to get a hybrid assembly. Each scaffold of the hybrid assembly was then checked, and small overlapping contigs were curated and removed to make a contiguous sequence. This curated diploid assembly was examined to identify alternative contigs using Purge_Haplotigs (v1.1.1; RRID:SCR_017616) [19], and the primary assembly and haplotig assemblies were created. We mapped trimmed Illumina whole genome sequences to both assemblies separately with bowtie2 (v2.3.4; RRID:SCR_016368) [20] and samtools (v1.9; RRID:SCR_002105) [21]. The resulting .bam files were used for polishing both assemblies using Pilon (v1.23; RRID:SCR_014731) [22] with one round, and the final assembly (primary assembly) and haplotig assemblies were prepared. In this study, we used only the primary assembly for all downstream analysis, but the haplotigs are maintained to cover the total heterozygous genome. Scaffolds were aligned to the V. vinifera ‘PN40024’ 12X.v2 [23] reference genome using minimap2 (v2.17; RRID:SCR_018550) [24] and renamed based on the longest alignment with the reference genome V. vinifera ‘PN40024’ 12X.v2 chromosomes. We mapped two thousand ‘Chambourcin’ rhAmpSeq marker sequences [25] to the ‘Chambourcin’ genome assembly using the BWA aligner (v0.7.17; RRID:SCR_010910) [26]. The rhAmpSeq markers were designed to target the core Vitis genome and were developed from gene-rich collinear regions of 10 Vitis genomes [25]. These markers aid in mapping contigs on chromosomes and checking their orientation.

All assemblies generated by PacBio HiFi and Bionano data were assessed by Benchmarking Universal Single-Copy Orthologs (BUSCO) (v5.4.2; RRID:SCR_015008) [27] with genome mode and the embryophyta_odb10 dataset. The alignment of the two genomes was obtained using minimap2 (v2.17) [24] with default parameters, where the ‘Chambourcin’ primary assembly was considered as query while V. vinifera ‘PN40024’ 12X.v2, Shine Muscat [28], and V. riparia Gloire [29] were considered as the reference genome. A dot plot was obtained using the R (RRID:SCR_001905) script pafCoordsDotPlotly.R [30].

The trimmed Illumina whole genome sequences were mapped separately to diploid, primary, and haplotig ‘Chambourcin’ assemblies using KAT (v2.4.2; RRID:SCR_016741) [31]; specifically, we used kat comp and kat plot commands.

De novo repeats were identified with RepeatModeler2 (v2.0.2a) [32], and repeats were masked by RepeatMasker (v4.1.1; RRID:SCR_012954) [33]. ‘Chambourcin’ RNA-seq data were downloaded from a previously published study [7] and trimmed using Trimmomatic (v0.39) [15] with HEADCROP:15 LEADING:30 TRAILING:30 MINLEN:20 parameters. The trimmed ‘Chambourcin’ RNA-seq reads were then mapped to the masked ‘Chambourcin’ primary genome assembly using HISAT2 (v2.1.0; RRID:SCR_015530) [34] and samtools (v1.9) [21] with default parameters. The resulting alignments (.bam files) and protein sequences of the V. vinifera ‘PN40024’ 12X.v2, VCost.v3 were used for gene prediction using BRAKER2 (v2.1.6; RRID:SCR_018964) [35] with –prg=gth –gth2traingenes –gff3 parameters. The resulting gene predictions (proteins, coding sequences, and annotations) were completed separately for the ‘Chambourcin’ primary assembly and the ‘Chambourcin’ haplotig assembly. The quality of the predicted proteins was assessed using BUSCO (v5.4.2) [27] with protein mode and the embryophyta_odb10 dataset. The predicted proteins of the Vitis ‘Chambourcin’ primary assembly were then functionally annotated using eggNOG-mapper (v2) (RRID:SCR_021165) [36] and related to Gene Ontology (GO), KEGG pathway, and other functional information. The GO plot was developed using the WEGO tool [37]. For the analysis of orthologous gene models, the sequences of ‘Chambourcin’ primary gene models, V. vinifera PN40024 12X.v2, VCost.v3, Shine Muscat, and V. riparia Gloire were analyzed using OrthoVenn2 [38] with default settings, E-value: 1 × 10−5, and inflation value: 1.5.

The plant transcription factors for the gene models of the ‘Chambourcin’ primary assembly, V. vinifera PN40024 12X.v2, and VCost.v3 were identified using the Plant Transcription Factor Database (PlantTFDB v5.0; RRID:SCR_003362) [39]. The identified transcription factors were divided into subfamilies according to their sequence relationship with V. vinifera. For the circular phylogenetic tree and WRKY sequences of ‘Chambourcin’ primary gene models and V. vinifera PN40024 12X.v2, VCost.v3 gene models retrieved from PlantTFDB (5.0) [39] and aligned using ClustalW method in MEGA7 [40]. A phylogenetic analysis was carried out using the neighbor-joining method with 1,000 bootstrap replications, and the evolutionary distances were computed using the Poisson correction method with the Pairwise Deletion option. The WRKY classification of ‘Chambourcin’ primary gene models was carried out using the same method described in a previous study [41].

The ‘Chambourcin’ masked primary genome assembly and gene annotations were aligned to V. vinifera ‘PN40024’ 12X.v2, Shine Muscat, and V. riparia ‘Gloire’ genomes and gene annotations separately using the ‘promer’ option of the MUMmer program in SyMAP (v4.2) [42]. We used MIcroSAtellite [43] to find SSRs in the unmasked ‘Chambourcin’ primary genome assembly. The Circos plots were generated using circos (v0.69.6; RRID:SCR_011798) [44] with the ‘Chambourcin’ primary genome assembly, SSRs, and the ‘Chambourcin’ primary gene annotations.

The trimmed Illumina whole genome sequences were mapped to the diploid, primary, and haplotig ‘Chambourcin’ assemblies separately with bowtie2 (v2.3.4) [20] and samtools (v1.9) [21]. The resulting .bam files were used to obtain mapping results using the samtools flagstat [21] command. We also mapped the trimmed ‘Chambourcin’ RNA-seq reads [7] to diploid, primary, and haplotig ‘Chambourcin’ assemblies separately using HISAT2 (v2.1.0) [34] and samtools (v1.9) [21]. The resulting .bam files were used to obtain mapping results using the samtools flagstat [21] command.

We generated a high-quality and contiguous genome sequence of ‘Chambourcin’ using PacBio HiFi Sequencing, Bionano third-generation DNA sequencing, and Illumina short-read sequencing. A total of 1,634,814 PacBio HiFi filtered reads was produced with an average length of 16,148 bp and genome coverage of 28X. The filtered Bionano data resulted in a subset of 1,243,428 molecules with a total length of 429,808.857 Mbp and coverage of 188.70X. In total, 124 Bionano maps, with a total length of 962.964 Mbp and an N50 of 13,725 bp, were assembled, corresponding to the diploid complement. A total of 154,152,068 filtered Illumina short reads and genome coverage of 40X were generated for genome polishing. We estimated heterozygosity to be 2.28% in the ‘Chambourcin’ genome (Figure 1), which is higher than estimates for heterozygosity in any of the other Vitis genomes sequenced to date [23, 28, 29]. Relatively higher levels of heterozygosity in the ‘Chambourcin’ genome compared to other Vitis species are expected, given the complex interspecific pedigree of this cultivar. A GenomeScope plot of clean reads demonstrated two peaks of coverage; the first peak located at 25X coverage corresponds to the heterozygous portion of the genome, and the second peak at 52X coverage corresponds to the homozygous portion of the genome (Figure 1).

A de novo ‘Chambourcin’ genome was assembled using HiFi, Bionano, and Illumina data. First, a contig assembly of the PacBio HiFi reads resolved the reads into 196 contigs with an N50 of 12,215,205 bp and a total length of 949,347,381 bp (Table 1). The PacBio HiFi contig assembly was then merged with the Bionano maps to get an initial hybrid assembly comprising 67 scaffolds with an N50 length of 16,400,326 bp, a maximum scaffold length of 39,458,994 bp, and a total scaffold length of 903,810,753 bp (Table 1). After manual curation, the hybrid assembly included 64 scaffolds with an N50 length of 16,278,793 bp, a maximum length of 39,458,994 bp, and a total length of 869,222,201 bp (Table 1). The hybrid assembly was partitioned into a final primary assembly (493,554,689 bp) and a haplotig assembly (375,458,233 bp) (Table 1). Pilon (v1.23) corrected 19,771 single nucleotide polymorphisms (SNPs), 636 ambiguous bases, 16,469 small insertions totaling 121,315 bases, and 21,889 small deletions totaling 129,590 bases in the primary assembly; Pilon (v1.23) also corrected 21,394 SNPs, 606 ambiguous bases, 14,999 small insertions totaling 102,126 bases, and 18,929 small deletions totaling 105,957 bases in the haplotig assembly. After polishing for ‘Chambourcin’, the final primary assembly contained 26 scaffolds with an N50 length of 23,325,629 bp, and the longest scaffold measured 39,456,434 bp (Table 1). The secondary haplotig assembly after polishing contained 38 haplotig scaffolds with an N50 length of 1,2462,019 bp, and the longest scaffold measured 28,439,729 bp (Table 1). We identified 97.9% complete BUSCOs for primary genome assembly and 73.1% complete BUSCOs for haplotig genome assembly (Table 1).

The ‘Chambourcin’ primary genome assembly was aligned to the reference genomes V. vinifera ‘PN40024’ 12X.v2 [23] (see Table 1 in GigaDB [45]), Shine Muscat [28], and V. riparia ‘Gloire’ [29]. A dot plot was generated to facilitate the comparisons among genomes. Collinearity between ‘Chambourcin’ and V. vinifera ‘PN40024’ 12X.v2, Shine Muscat, and V. riparia ‘Gloire’ was observed as a straight diagonal line without large gaps in the dot plot, confirming the high synteny of the ‘Chambourcin’ genome with V. vinifera ‘PN40024’ 12X.v2 (Figure 2A), Shine Muscat (Figure 2B), and V. riparia ‘Gloire’ (Figure 2C). To further validate the ‘Chambourcin’ genome assembly, we mapped 'Chambourcin' rhAmpSeq markers [25] to the ‘Chambourcin’ genome assembly. We found 99% of rhAmpSeq markers mapped to ‘Chambourcin’ scaffolds and mapped to the same chromosomes and positions the markers were derived from in the collinear Vitis core genome (see Table 2 in GigaDB [45]).

Synteny analyses of the ‘Chambourcin’ primary genome assembly with V. vinifera ‘PN40024’ 12X.v2, Shine Muscat, and V. riparia ‘Gloire’ genomes were used to identify syntenic blocks between species. The ‘Chambourcin’ primary assembly scaffolds aligned with larger syntenic blocks and covered the whole chromosomes of V. vinifera PN40024 12X.v2 (Figure 2D), Shine Muscat (Figure 2D), and V. riparia ‘Gloire’ (Figure 2E). This alignment of the primary genome with V. vinifera PN40024 12X.v2, Shine Muscat, andV. riparia ‘Gloire’ indicated highly contiguous ‘Chambourcin’ scaffolds useful for comparative genomic analyses.

The genome quality was assessed with Illumina whole genome reads separately for diploid, primary, and haplotig genome assemblies using KAT tool [31] and K-mer spectra plots were generated. A K-mer spectra is a graphical representation showing how many k-mers appear a certain number of times. The frequency of occurrence is plotted on the x-axis and the number of k-mers is plotted on the y-axis. All K-mer spectra plots for ‘Chambourcin’ diploid, primary and haplotig assemblies were identified with an error distribution under 10X, a heterozygous peak at 35X, and a homozygous peak at 67X (Figure 3A–C). The different colors in the K-mer spectra plot shows the different occurrences of k-mers. The black color represents read content occurs at zero time (0X), the red color represents unique content occurs at one time (1X), the purple color represents content occurs at two times (2X) and the green color represents content occurs at three times (3X). The K-mer spectra plot of the diploid genome assembly shows that the read content shown in black color is absent from the assembly, and red peak occurs once, showing most of the heterozygous content. At the same time, purple peak indicates more duplications on homozygous content (Figure 3A). The K-mer spectra plot of the primary genome is more collapsed, including mostly a single copy of the homozygous content and less of the heterozygous content (Figure 3B). The K-mer spectra plot for the haplotig genome assembly identified two black peaks representing read content in both the heterozygous and homozygous regions (Figure 3C). These K-mer spectra plots provides useful information for genome assembly assessment using whole genome short reads to identify duplicate regions in the assembly and visualize the genome assembly. This visualization is useful for genome assembly curation steps to identify accurate primary and haplotig assembly from a diploid genome assembly.

Repeated regions were binned into seven different classes: long interspersed nuclear elements (LINEs) (4.43%), long terminal repeats (LTRs) (15.66%), DNA transposons (2.03%), rolling-circles (0.58%), low complexity repeats (0.37%), simple repeats (1.21%), and unclassified repeats (31.95%) (Figure 4; Table 2). The repetitive sequence content in the ‘Chambourcin’ primary genome assembly (56.23%) was higher than previously reported for V. riparia ‘Manitoba 37’ (46%) [41], V. vinifera ‘PN40024’ 12X.v2 (35.12%) [23], Shine Muscat (48%) [28], and V. riparia ‘Gloire’ (33.94%) [29]. SSRs are tandem repeats of DNA that have been used to develop robust genetic markers. We identified 304,571 SSRs, repeating units of 1–6 base pairs in length, in the ‘Chambourcin’ primary genome assembly (Figure 4; and see GigaDB Table 3 [45]).

A total of 33,791 gene models were predicted for the ‘Chambourcin’ primary genome assembly (Figure 4). We identified 94.6% complete BUSCOs (C); of these, 86.9% were designated single-copy BUSCOs (S), and 7.7% were designated duplicated BUSCOs (D) (Table 3). As evidenced by the high number of complete single-copy genes identified, the BUSCO results indicate that the ‘Chambourcin’ primary genome assembly offers comprehensive coverage of the expected gene space. Functional annotation of the ‘Chambourcin’ primary gene models (33,791) was done using the EggNOG database (see GigaDB Table 4 [45]). A total of 27,075 ‘Chambourcin’ primary proteins were annotated, and 86% (22,977) of these proteins were annotated with V. vinifera V1 gene models (see GigaDB Table 4 [45]). Out of the 27,075 ‘Chambourcin’ annotated primary proteins, 13,311 gene models were identified with GO accessions and further classified into three sub-ontologies: biological process (11,399), cellular component (11,472), and molecular function (9,977) (Figure 5F) (GigaDB Table 4 [45]). A total of 8,460 ‘Chambourcin’ primary proteins were annotated with KEGG pathways (GigaDB Table 4 [45]). Using OrthoVenn2, we identified 16,056 common orthologs between ‘Chambourcin’ primary gene models, V. vinifera PN40024 12X.v2 annotation, Shine Muscat, and V. riparia ‘Gloire’ (Figure 5A). In total, 16,476 orthologous gene models were found between the ‘Chambourcin’, Shine Muscat, and V. riparia ‘Gloire’ (Figure 5B). Finally, 19,477 gene models were orthologous with V. vinifera PN40024 12X.v2 VCost.v3 proteins (Figure 5C), 18,669 gene models were orthologous with Shine Muscat (Figure 5D), and 18,183 gene models were orthologous with V. riparia ‘Gloire’ (Figure 5E).

Using PlantTFDB 5.0, 1,606 plant transcription factors representing 58 gene families were identified from ‘Chambourcin’ primary proteins (see GigaDB Table 5 [45]). A similar number of transcription factors was identified for the AP2, NAC, RAV, and WRKY gene families, as found in V. vinifera ‘PN40024’ 12X.v2, VCost.v3. We identified 65 WRKY sequences in ‘Chambourcin’ and 62 in V. vinifera PN40024 12X.v2, VCost.v3 (Figure 6) (Table 4; GigaDB Table 5 [45]). WRKY transcription factors regulate many processes in plants and algae, such as the responses to biotic and abiotic stresses and seed dormancy. The 'Chambourcin' WRKY subfamily classification was similar to V. vinifera ‘PN40024’ 12X.v2 and V. riparia ‘Manitoba 37’ (Table 4). These results show the high coverage of ‘Chambourcin’ primary proteins.

We aligned trimmed Illumina whole genome reads to 'Chambourcin' diploid, primary, and haplotig assemblies separately and obtained an average of 95.08%, 90.17%, and 75.88% mapping results, respectively (Table 5). We also separately mapped trimmed RNA-seq reads to 'Chambourcin' diploid, primary, and haplotig assemblies and obtained 87%, 80.81%, and 60.77% mapping results, respectively. The mapping results of both trimmed Illumina whole genome and trimmed RNA-seq reads to genome assemblies show that most reads were mapped to the diploid assembly, followed by primary and haplotig assemblies. The mapping results for the primary genome assembly retained most of the genome from the diploid assembly, while the smallest number of mapped reads belonged to the haplotig assembly. These results suggest that the smallest genome portion is missing in the haplotig assembly.