Developing biparental mapping populations
The watermelon mapping populations were developed at North Carolina State University. Two parental lines, ‘Charleston Gray’(resistant, female parent, P1) developed by C. F. Andrus in 1954 (Andrus 1955), and ‘New Hampshire Midget’ (susceptible, male parent, P2) were used to generate F1, F2, BC1P1, and BC1P2 mapping populations. The mapping populations consisted of 228 F2 individuals as well as 60 individuals each in BC1P1 and BC1P2.
Inoculum preparation and pathogen inoculation
Colletotrichum orbiculare race 1, collected in North Carolina in 1998, was used to inoculate seedlings. The inoculum preparation and inoculation were conducted as described by Patel (2019). In brief, the fungus was grown on green bean agar (GBA) media for three-weeks. Spores were harvested by adding 10 to 15 mL distilled water to each agar plate, rubbing the surface of the agar with a sterile metal spreader, pouring the spore suspension into a sterile conical flask, and passing it through four layers of cheesecloth. Concentration of the inoculum was measured using a hemocytometer and adjusted to 100,000 spores mL− 1 prior to inoculation. One drop of Tween-20 was added to every 500 mL of the spore inoculum. The three-week-old watermelon seedlings grown in the greenhouse were inoculated with the spore inoculum. After inoculation, seedlings were kept in a humidity chamber, in the greenhouse, for 48 h in darkness at 80–100% relative humidity, and at a temperature of 22 to 24°C. Then, seedlings were moved to the natural light, and rated at 8, 11, and 14 days post inoculation (dpi).
Disease rating
The disease index was rated on a scale of 0 to 100%, with an interval of 5%, with weightage on different parts of the plants - true leaves (50% total: yellowing- 5%, complete necrotic leaf- 40%, petiole-10%), meristem (25% total: necrosis spots-10%, mostly necrotic- 20%, dead-25%), hypocotyl (20% total: 1–2 brown patches-5%, many brown patches-15%, completely brown-20%), cotyledons (5% total: little to complete necrosis: 5%). Individuals were designated as resistant and susceptible when the overall rating score was ≤ 40%, and ≥ 41%, respectively (Patel 2019).
DNA isolation, ddRADseq library construction, and genotyping by sequencing
A total of 360 watermelon leaf samples (three P1, three P2, six F1, 60 BC1P1, 60 BC1P2, and 228 F2 individuals) were collected from three-week-old seedlings. Samples were freeze-dried immediately, and genomic DNA was extracted from lyophilized samples using E.Z.N.A. Plant DNA Kit (Omega Bio-tek, GA, USA) following manufacturer’s protocol. The DNA were quantified using Quant-iT-PicoGreen (Invitrogen, Thermo Fisher Scientific, USA) following the manufacturer’s instructions. Due to some samples yielding low amounts of DNA, a total of 188 watermelon samples (three P1, two P2, six F1, 48 BC1P1 and 129 F2 individuals) were sent to Texas A&M AgriLife Genomics and Bioinformatics Service, College Station, TX (https://www.txgen.tamu.edu/) for double digest restriction-site associated DNA sequencing (ddRADseq) as described previously (Yang et al. 2020) with the following changes. The restriction enzymes EcoRI and NlaIII were used for library prep and inserts from 400 to 600 bp were selected on a Pippin prep (Sage Science, Boston, MA, USA). The ddRADseq libraries were sequenced using 40% of a NovaSeq S4 X lane (2 x150 bp paired-end run; Illumina, Inc., San Diego, CA, USA).
Raw sequences were demultiplexed using Illumina bcl2fastq, allowing for 1 base error in the barcode sequences. Sequences were first quality-filtered using the program FASTX-Toolkit (http://hannonlab.cshl.edu/fastx-toolkit). Raw sequencing reads were first trimmed to remove low quality bases with quality score less than 20 on the ends of reads and reads with 30% or more bases showing low quality score (Q < 15) were removed. The reference genome for watermelon was downloaded from NCBI website (GCA_000238415.2). Bowtie2 [http://bowtie-bio.sourceforge.net/bowtie2/index.shtml] was used to align quality-filtered reads to the reference with the default parameters. Aligned reads were then processed with SAMtools v1.19 to generate coordinate sorted binary SAM files (BAM). Reads with mapping quality (MQ) less than 5 were removed. The local re-alignment tool in the Genome Analysis Toolkit (GATK, https://software.broadinstitute.org/gatk/) was used to perform re-alignment in Insertion/Deletion regions as previously described. Finally, the processed alignment files were fed to the tool HaplotypeCaller, which is part of the GATK, to call variations and perform genotyping for each sample. Once the SNP calling process was completed, individual SNPs with more than 20% missing data and Minor Allele Frequency (MAF) less than 0.05 in each population group were removed.
QTL mapping
Genotypic data were assigned to A (P1 type, homozygous resistant), B (P2 type, homozygous susceptible), H (heterozygous), and X (missing) types. Since the phenotypic disease rating data for F2 population were found to be in a non-normal distribution, QTL analysis was done on ‘qtl’ package (Broman et al. 2003) with a non-parametric method (‘model = np’) on R software (R Core Team 2014; version 3.6.2) with RStudio GUI (RStudio-Team 2021). The logarithm of odds (LOD) threshold of 4.11 for QTL detection was estimated with 1,000 permutations. As genotyping-by-sequencing (GBS) genotypic data had higher missing values, QTL analysis was also performed after imputing missing genotypic data using a multiple imputation method (‘method = imp’) (Sen and Churchill 2001) on R ‘qtl’. A graphical display of allele effects was done using “Effect Plot” function on R ‘qtl’. A genetic linkage map was constructed using IciMapping V4.1 (Meng et al., 2014), whereas the linkage map was displayed using MapChart version 2.32 (Voorrips 2002).
PACE-based SNP genotyping and non-parametric analysis
Allele-specific primers were designed for 34 SNP markers in and around the QTL region using PrimerQuest™ Tool (IDT, Coralville, IA, USA) and confirmed manually. Polymerase chain reaction (PCR) allelic competitive extension (PACE) genotyping chemistry constituting FAM, HEX, and ROX fluorophores was used to analyze the SNPs (3CR Bioscience, Essex, UK). The polymorphic PACE SNP markers (Supplementary Table S1) were used to genotype the mapping populations (N = 360) as well as watermelon germplasm (N = 61). The PACE SNP marker was also designed for a previously reported high resolution melting SNP marker, CL14-27-9 (Supplementary Fig. S1), for CC-NBS-LRR gene (CNL; Cla001017 or ClCG08G002410) (Jang et al. 2019), and designated as S8_5149002 to match the physical coordinates of the Charleston Gray genome. The PACE PCR components included 4 µL of PACE Genotyping Master Mix (2X) (3CR Bioscience, Essex, UK), 0.11 µL primer assay mix (72X), 2 µL template DNA and 2 µL of molecular biology grade water. The PCR was carried out in the Eppendorf flexlid nexus gradient Mastercycler (Eppendorf, Hamburg, Germany). The PCR conditions included one cycle of enzyme activation (94℃, 15 min), followed by 10 cycles each of template denaturation (94℃, 20 s) and annealing/extension with drop of 0.8℃ per cycle (65 to 57℃, 60 s), and 27 cycles each of denaturation (94℃, 20 s), and annealing/extension (57℃, 60 s) (3crBioscience 2018). KlusterCaller software version 3.4.1.36 (LGC Genomics, Herts, UK) was used to cluster genotypes using BMG Labtech Omega machine (BMG Labtech, Ortenberg, Germany). An additional 3 to 9 cycles of final denaturation and annealing/extension was done to improve the amplification, as well as to obtain tight and well separated clusters. Several non-parametric analysis - Chi-Square, Mann-Whitney-Wilcoxon test (Wilcoxon 1945; Mann and Whitney 1947), Kruskal-Wallis test (Kruskal and Wallis 1952), and Dunn’s test (Dunn 1964) were conducted on data on R software (R Core Team 2014; version 3.6.2) with RStudio GUI (RStudio-Team 2021) using functions ‘chisq.test’, ‘wilcox.test’, ‘kruskal.test’, and ‘dunn.test’, respectively.