Background & Summary

Cultivated grapevines (Vitis vinifera ssp. vinifera) are usually grafted onto rootstocks derived from North American Vitis species (Fig. 1a). This practice was established during the 19th century in response to the near devastation of European vineyards by the grape root aphid phylloxera (Daktulosphaira vitifoliae Fitch)1. Grape phylloxera was introduced into Europe in the 1850s through the movement of plant material from North America2. Most North American Vitis species are resistant to phylloxera, likely as a result of co-evolution with the insect in their native environment. Vitis riparia and Vitis rupestris were the first wild grape species used as rootstock because they root easily from hardwood cuttings and have good grafting compatibility with the berry-producing scions3. However, these two species were not suitable for calcareous soils, which are common in Europe. Vitis berlandieri, another North American grape species, was then found to be resistant to phylloxera and lime-tolerant, although it poorly roots from dormant cuttings4. To introduce the lime-tolerance of V. berlandieri and improve its rootability, new rootstocks were bred crossing V. berlandieri with either V. riparia or V. rupestris. Today, commercialized rootstocks are mainly hybrids of these three grape species5. Among these, Richter 110 (110R; V. berlandieri x V. rupestris), Kober 5BB (V. berlandieri x V. riparia), and 101–14 Millardet et de Grasset (Mgt; V. riparia x V. rupestris) are the most commonly used worldwide (Fig. 1b). In addition to their resistance to phylloxera, grape rootstocks are chosen based on tolerance to biotic (e.g. nematodes) and abiotic stresses (e.g. drought), preference of soil physicochemical properties, and the vigor level they confer to the scion6. For instance, 101–14 Mgt generally triggers the precocity of the vegetative growth despite a moderate vigor, whereas 110R and Kober 5BB confer high vigor and delay plant maturity7. 110R is known for its drought tolerance and excess soil moisture has negative impacts on its development6. In contrast, 101–14 Mgt and Kober 5BB are not considered drought-tolerant and grow well in moist soils6. The three rootstocks also have different levels of tolerance to nematodes depending on the nematode species6,8.

Fig. 1
figure 1

Description of the three grape rootstocks 101–14 Mgt, 110R, and Kober 5BB. (a) Wine grapevine scion (Vitis vinifera spp. vinifera) grafted onto a rootstock from another Vitis species. (b) Schematic representation of haplotype composition of 101–14 Mgt, 110R, and Kober 5BB. Each pair of rootstocks shares a set of chromosomes from the same parental Vitis species. Shared haplotypes are represented with the same color.

Full size image

In addition to their commercial importance, rootstocks are valuable to study the genetic bases of grape adaptation to biotic and abiotic stresses9. However, to date only two genomes of V. riparia have been published10,11 and no reference genome is available for any of the commonly used rootstocks. This article describes the chromosome-scale assemblies of 110R, Kober 5BB, and 101–14 Mgt. Genomes were sequenced using highly accurate long-read sequencing (HiFi, Pacific Biosciences) and assembled with Hifiasm12. Each diploid draft genome was then scaffolded into two sets of pseudomolecules using the tool suite HaploSync13, and haplotypes were assigned to each Vitis parent based on sequence similarity between the haplotypes derived from the same species. These genomes represent an important resource for investigating the genetic basis of resistance to environmental factors and designing markers to accelerate rootstock breeding programs.

Methods

Library preparation and sequencing

Young leaves (1–2 cm-wide) were collected from 110R (FPS 01), Kober 5BB (FPS 06), and 101–14 Mgt (FPS 01) at Foundation Plant Services (University of California Davis, Davis, CA) and immediately frozen and ground to powder in liquid nitrogen. High molecular weight genomic DNA was extracted from 1 g of ground leaf tissue as described in Chin et al.14, and 12 µg of high molecular weight gDNA was sheared to a size distribution between 15 and 20 kbp using the Megaruptor® 2 (Diagenode, Denville, NJ, USA). For each accession, one HiFi sequencing library was prepared using the SMRTbell® Express Template Prep Kit 2.0 followed by immediate treatment with the Enzyme Clean Up Kit (Pacific Biosciences, Menlo Park, CA, USA). Libraries were size-selected using a BluePippin (Sage Sciences, Beverly, MA, USA) and HiFi SMRTbell® templates longer than 15 kbp were collected. Size-selected library fractions were cleaned using AMPure PB beads (Pacific Biosciences, Menlo Park, CA, USA). Concentration and final size distribution of the libraries were evaluated using a Qubit™ 1X dsDNA HS Assay Kit (Thermo Fisher, Waltham, MA, USA) and Femto Pulse System (Agilent, Santa Clara, CA, USA), respectively. HiFi libraries of 110R and Kober 5BB were sequenced using a PacBio Sequel II system (Pacific Biosciences, CA, USA) at the DNA Technology Core Facility, University of California, Davis (Davis, CA, USA). For 101–14 Mgt, sequencing was performed by Corteva Agriscience (Johnston, IA, USA) as an award from Pacific Biosciences to Dr. Noé Cochetel. An average of 26.5 ± 3.8 Gbp sequences were generated for each genome, corresponding to 53.1 ± 7.7 X-fold coverage of a 500 Mbp haploid genome (Table 1).

Table 1 Genome assembly statistics of the three rootstocks.
Full size table

Total RNA from V. berlandieri 9031, V. rupestris B38, and V. riparia HP-1 (PI588271) leaves was isolated using a Cetyltrimethyl Ammonium Bromide (CTAB)-based extraction protocol as described in Blanco-Ulate et al.15. RNA purity was evaluated with a Nanodrop 2000 spectrophotometer (Thermo Scientific, Hanover Park, IL, USA), and RNA integrity by electrophoresis and an Agilent 2100 Bioanalyzer (Agilent Technologies, CA, USA). RNA quantity was assessed with a Qubit 2.0 Fluorometer and a broad range RNA kit (Life Technologies, Carlsbad, CA, USA). Total RNA (300 ng, RNA Integrity Number >8.0) were used for library construction. Short-read cDNA libraries were prepared using the Illumina TruSeq RNA sample preparation kit v.2 (Illumina, CA, USA) following Illumina™ low-throughput protocol. Libraries were evaluated for quantity and quality with the High Sensitivity chip and an Agilent 2100 Bioanalyzer (Agilent Technologies, CA, USA). One library per species was sequenced using an Illumina HiSeq 4000 sequencer with a 2x100bp protocol (DNA Technology Core Facility, University of California, Davis, CA, USA). Long-read cDNA SMRTbell libraries were prepared for V. berlandieri and V. riparia. First-strand synthesis and cDNA amplification were accomplished using the NEB Next Single Cell/Low Input cDNA Synthesis & Amplification Module (New England, Ipswich, MA, USA). The cDNAs were subsequently purified with ProNex magnetic beads (Promega, WI, USA) following the instructions in the Iso-Seq Express Template Preparation for Sequel and Sequel II Systems protocol (Pacific Biosciences, Menlo Park, CA, USA). ProNex magnetic beads (86 µL) were used to select amplified cDNA (≥2 kbp). At least 80 ng of the size-selected amplified cDNA were used to prepare the cDNA SMRTbell library. DNA damage repair and SMRTbell ligation was performed with SMRTbell Express Template Prep Kit 2.0 (Pacific Biosciences, Menlo Park, CA, USA) following the manufacturer’s protocol. One SMRT cell was sequenced for each species on the PacBio Sequel I platform (DNA Technology Core Facility, University of California, Davis, CA, USA).

Genome assembly and pseudomolecule construction

HiFi reads were assembled using Hifiasm v.0.16.1-r37412. Multiple combinations of several assembly parameters were tested. A total of 1,939 assemblies were generated. The least fragmented assembly of each genotype was selected. The selected draft assemblies consisted of 406 ± 226 contigs with a N50 = 14.3 ± 0.6 Mbp (Table 1). Compared to other grape genomes previously generated with PacBio CLR technology, the PacBio HiFi reads greatly improves the contiguity of the draft assembly (PacBio CLR 1.2 ± 0.3 Mbp, Fig. 2a). Gene space completeness was assessed using BUSCO V.5.1 with the Viridiplantae and Embryophyta ODB10 datasets16 and by mapping PN40024 (V1 annotation17) single-copy genes using GMAP v.2019-09-12 (alignments with at least 80% coverage and 80% identity were considered). For each rootstock, the draft genome assembly underwent quality control and scaffolding into a diploid set of chromosome-scale pseudomolecules using HaploSync13 and the Vitis consensus genetic map developed by Zou et al.18. One cycle of HaploFill was used for each genotype. The use of PacBio HiFi reads reduced significantly the fragmentation of the draft assembly compared to recently published grape genomes sequenced using PacBio CLR technology (Fig. 2b)13,14,19. The lower fragmentation resulted in a 15 times smaller number of contigs necessary to scaffold a pseudomolecule (3.6 ± 2.0 HiFi contigs/pseudomolecule vs. 43.0 ± 20.6 CLR contigs/pseudomolecule) (Fig. 2b). Remarkably, in total across the three genomes, fifteen pseudomolecules were reconstructed from a single contig. Haplotype switches were identified based on sequence similarity of protein-coding sequences. Gene loci sequences of each rootstock were aligned against each others using minimap2 v.2.17-r94120 and the parameter “-x map-hifi”. Alignments with the highest coverage and identity were used to assign common species parentage and to detect haplotype switches along pseudomolecules (Fig. 3a). After manual correction of the haplotype switches, a second cycle of HaploFill13 was performed using the pseudomolecules derived from the same Vitis species as alternative haplotypes to help closing gaps with draft sequences.

Fig. 2
figure 2

PacBio HiFi sequencing technology substantially improves the contiguity of Vitis draft genome assembly. (a) Draft assembly fragmentation of 101–14 Mgt, 110R, Kober 5BB represented as distribution of contig NG(x) values. Muscadidia rotundifolia cv. Trayshed and V. vinifera cv. Cabernet Sauvignon, produced with CLR reads, were included as comparison. The NG(x) value is defined as the sequence length of the shortest contig necessary to achieve, cumulatively, a given fraction (x) of the expected diploid genome length (1 Gbp) when sequences are sorted from the longest to the shortest. Diploid assemblies produced with PacBio HiFi reads (101–14 Mgt, 110R, and Kober 5BB) resulted in a much more contiguous draft genome assembly compared to other grape genomes assembled with older long-read sequencing technologies despite a lower X-Fold coverage employed (PacBio Sequel CLR reads for Trayshed 140x X-Fold coverage;19,21 PacBio RSII CLR reads for Cabernet Sauvignon, 115X X-Fold coverage14). (b) Distribution of the number of contig scaffolded into complete pseudomolecules. The substantially lower fragmentation of the draft assemblies generated using PacBio HiFi reads (101–14 Mgt, 110R, and Kober 5BB) resulted on average in a 15x smaller number of contigs necessary to build a pseudomolecule.

Full size image
Fig. 3
figure 3

Haplotyping based on intraspecific sequence similarity. Shared parental species information was used to assign each haplotype to either V. riparia, V. rupestris, or V. berlandieri based on sequence similarity. This allowed to resolved assembly errors (i.e. haplotype switches). (a) Example of an haplotype switch found on chromosome 8 of Kober 5BB (left panel). After scaffolding of the pseudomolecules, an haplotype switch was observed at the end of chromosome 8 of Kober 5BB. The genes in the contig Kober5BB_h1tg000016l on haplotype 1 were highly similar to the genes located in 101–14 Mgt haplotype 2 (red), suggesting that Kober5BB_h1tg000016l derived from V. riparia, whereas the genes of Kober5BB_h1tg000038l corresponded to genes in haplotype 2 of chromosome 8 of 110R (light green), suggesting that Kober5BB_h1tg000038l derived from V. berlandieri. An opposite pattern was observed on haplotype 2, with the genes of the first 18.9 Mbp of the pseudomolecule similar to the genes of the haplotype 2 of 110R (light green) and the genes from the last 4.2 Mbp similar to the genes of 101–14 Mgt haplotype 2 (red). The haplotype switch was corrected by interchanging the contig Kober5BB_h1tg000038l with the corresponding region in the alternative haplotype, consisting of Kober5BB_h2tg000109l and 4.2 Mbp of Kober5BB_h2tg000017l (right panel). (b) Sequence similarity between haplotypes represented as the average percentage of the haploid chromosome set length not affected by structural variants (>50 bp), SNPs or InDels when compared with another haplotype. (c) Distribution of the percentage of sequence similarity (as defined in B) between haplotypes derived from the same species and haplotypes derived from different species (Statistical testing was performed with pairwise Wilcoxon rank sum test, density plot was produced with adjust = 1, n = 4096, kernel = “cosine” parameters).

Full size image

Gene prediction and repeat annotation

Gene structural annotations were predicted using the procedures described in https://github.com/andreaminio/AnnotationPipeline-EVM_based-DClab21. For each rootstock, Iso-Seq data from the corresponding parental species were concatenated with the de novo assembled transcripts from RNA-seq reads before generating the gene models. Iso-Seq libraries underwent extraction, demultiplexing and error correction using IsoSeq3 v.3.3.0 protocol (https://github.com/PacificBiosciences/IsoSeq). Low-quality and single isoforms dataset were further polished using LSC v2.022. RNA-seq reads were quality-filtered and adapters were trimmed with Trimmomatic v.0.36 and the options “ILLUMINACLIP:2:30:10 LEADING:7 TRAILING:7 SLIDINGWINDOW:10:20 MINLEN:36”23. High-quality RNA-seq reads from each Vitis species were assembled with three different protocols: (i) Trinity v.2.6.524 with the “de novo” protocol, (ii) Trinity v.2.6.524 using the “On-genome” protocol, (iii) Stringtie v.1.3.4d25 using the reads found to align on the genome sequences with HISAT2 v.2.0.5 and the parameter “--very-sensitive”26. Transcript sequences common to the three assembly methods were then pooled with the Iso-Seq reads. Sequence redundancy was reduced using CD-HIT v4.627 with the parameters “cd-hit-est -c 0.99 -g 0 -r 0 -s 0.70 -aS 0.99”. Non-redundant transcripts were processed with PASA v.2.3.328 to obtain the final training model sets. Combined with data from public databases, the derived transcript and protein evidences were aligned on the genome assembly using a multi-aligner pipeline including Exonerate v.2.2.029 and Pasa v.2.3.328. To produce the final set of consensus gene models with EvidenceModeler v.1.1.130, ab initio predictions were also generated using Augustus v.3.0.331, BUSCO v.3.0.232, GeneMark v.3.4733, and SNAP v.2006-07-2834. For the repeat annotation, RepeatMasker v.open-4.0.635 was used. To assign a functional annotation to each of these gene models, results from diamond v2.0.13.15136,37 blastp matches on the Refseq plant protein database (https://ftp.ncbi.nlm.nih.gov/refseq/, retrieved January 17th, 2019) and from InterProScan v.5.28–67.038 were parsed through Blast2GO v.4.1.939. A total of 56,768 protein-coding gene loci were annotated in the genome assembly of 110R, 59,807 in Kober 5BB and 72,758 in 101–14 Mgt. On average, 124,991 ± 36,197 protein-coding alternative splicing variants were identified per haplotype. The unplaced sequences were composed of 2,747 ± 2,821 gene loci (Table 1).

Analysis of colinearity between haplotypes

Colinear gene loci were identified using MCScanX v.11.Nov.201340. Annotated protein-coding sequences of the three rootstocks were aligned against each other using GMAP v.2019-09-1241 with the parameters “-B 4 -x 30–split-output”. Alignments with both identity and coverage greater than 80% were retained. Alignments corresponding to annotated mRNA regions were identified using mapBed from Bedtools v2.29.242 with the parameters “-F 0.75 -f 0.5 -e”. Colinear blocks were then detected with MCScanx_h (MCScanX v.11.Nov.201340) tool using the following parameters “-s 10 -m 5 -w 5”.

Identification of sequence polymorphisms and structural variants between haplotypes

Pseudomolecule sequences were aligned against each other using nucmer tool from MUMmer4 v.4.0.0.beta543. SNPs and short indels between haplotypes were identified from alignments with show-snps tool (MUMmer4 v.4.0.0.beta543) with parameters “-Clr -x” and longer structural variants with show-diff tool (MUMmer4 v.4.0.0.beta543) with default parameters.

Data Records

Sequencing data were deposited at NCBI under BioProject number PRJNA858084, SRA accessions SRR2081042144, SRR2081042245, SRR2081042346, SRR2081042447, SRR2081042548, SRR2081042649, and SRR2081042750. Genome assemblies are available at EMBL-EBI under BioProject number PRJEB5501351. Genome assemblies, gene annotation and repeat annotation files are at Zenodo under the https://doi.org/10.5281/zenodo.682432352, and at http://www.grapegenomics.com53. A genome browser and a blast tool are available for each rootstock at http://www.grapegenomics.com53.

Technical Validation

The genome assemblies were evaluated for completeness of the diploid sequence and gene content, and for correct haplotype phasing. The average size of each set of 19 pseudomolecules was 494.5 ± 5.5 Mbp (diploid genome size: 1,015.0 ± 7.9 Mbp, Supplemental figure 1), which is close to the length of the parental haploid genome size estimated by flow cytometry (499.3 ± 37.3 Mbp54) suggesting that the three genomes were entirely assembled. Only 36.1 Mbp (3.5%), 19.9 Mbp (2.0%), and 23.3 Mbp (2.3%) of the draft sequences could not be placed into any pseudomolecules of 101–14 Mgt, 110R, and Kober 5BB genomes, respectively. The unplaced sequences were mostly composed of repeats (68.0% ± 12.3%). These results are comparable with the latest release of the V. vinifera PN40024 reference haploid genome assembly, for which the location of 27.4 Mbp (5.6%) remains undetermined55.

Each set of 19 pseudomolecules was evaluated for gene space completeness using both conserved single-copy orthologs of plant genes (BUSCOs) and the single-copy gene content of V. vinifera PN40024. Complete copies of 98.1 ± 0.14% of the BUSCO models were found in each set of pseudomolecules (Supplemental Table 1). Similarly, almost all of the single-copy genes of PN40024 aligned to each set of pseudomolecules (95.01% ± 0.3%). The gene space present in the unplaced sequences was limited to 0.69 ± 0.8% of the BUSCO models and 1.79 ± 0.8% of the PN40024 genes. The completeness of the gene space is another strong evidence that the assemblies are a complete representation of the diploid genomes of the three rootstocks. On both haplotypes of 101–14 Mgt we found more gene loci (33,379 ± 328) than in 110R and Kober 5BB (28,584 ± 863). Further genome-wide gene expression analyses are required to determine if the larger number of gene loci identified in 101–14 Mgt corresponds to a larger number of expressed transcripts than in the other rootstocks.

Using the pedigree information of each rootstock (Fig. 1b), we assigned each pseudomolecule to its parental Vitis species, i.e. either V. riparia, V. rupestris, or V. berlandieri. For each pseudomolecule, we identified the three pairs of haplotypes having the highest gene sequence similarity and assigned them to the shared parental Vitis species. This allowed us to manually detect and correct the phasing errors (i.e. haplotype switches) introduced during the assembly of the draft sequences or the scaffolding of the pseudomolecules (Fig. 3a). Whole-sequence comparison of the six haplotypes of each pseudomolecule showed that the haplotypes assigned to the same Vitis species were more similar (80.5% ± 1.4% identity) than those that do not share the same species (74.0% ± 3.3% identity; p value = 0.0003, W = 142, n = 30 unpaired Wilcoxon rank sum test; Fig. 3b,c). These results suggest that the haplotypes of the three rootstock genomes were correctly phased. Despite the variable levels of sequence polymorphism, pseudomolecules of the three rootstock genomes were highly colinear regardless of their species of origin. When considering both gene sequence similarity, gene order, and physical location, 73.1% ± 3.5% of the protein-coding loci were found in at least one colinear block when comparing haplotypes with shared parental origin, and 71.5% ± 3.5% between haplotypes of different species (Supplemental figure 2). Overall, an average of 82.4% ± 2.6% of the genomic sequences are covered by colinear blocks (Supplemental figure 3), which reflects a remarkable conservation of chromosome structure among these Vitis species.