Rapid gene-based SNP and haplotype marker development in non-model eukaryotes using 3'UTR sequencing

Abstract

Background: Sweet cherry (Prunus avium L.), a non-model crop with narrow genetic diversity, is an important member of sub-family Amygdoloideae within Rosaceae. Compared to other important members like peach and apple, sweet cherry lacks in genetic and genomic information, impeding understanding of important biological processes and development of efficient breeding approaches. Availability of single nucleotide polymorphism (SNP)-based molecular markers can greatly benefit breeding efforts in such non-model species. RNA-seq approaches employing second generation sequencing platforms offer a unique avenue to rapidly identify gene-based SNPs. Additionally, haplotype markers can be rapidly generated from transcript-based SNPs since they have been found to be extremely utile in identification of genetic variants related to health, disease and response to environment as highlighted by the human HapMap project.

Results: RNA-seq was performed on two sweet cherry cultivars, Bing and Rainier using a 3' untranslated region (UTR) sequencing method yielding 43,396 assembled contigs. In order to test our approach of rapid identification of SNPs without any reference genome information, over 25% (10,100) of the contigs were screened for the SNPs. A total of 207 contigs from this set were identified to contain high quality SNPs. A set of 223 primer pairs were designed to amplify SNP containing regions from these contigs and high resolution melting (HRM) analysis was performed with eight important parental sweet cherry cultivars. Six of the parent cultivars were distantly related to Bing and Rainier, the cultivars used for initial SNP discovery. Further, HRM analysis was also performed on 13 seedlings derived from a cross between two of the parents. Our analysis resulted in the identification of 84 (38.7%) primer sets that demonstrated variation among the tested germplasm. Reassembly of the raw 3'UTR sequences using upgraded transcriptome assembly software yielded 34,620 contigs containing 2243 putative SNPs in 887 contigs after stringent filtering. Contigs with multiple SNPs were visually parsed to identify 685 putative haplotypes at 335 loci in 301 contigs.

Conclusions: This approach, which leverages the advantages of RNA-seq approaches, enabled rapid generation of gene-linked SNP and haplotype markers. The general approach presented in this study can be easily applied to other non-model eukaryotes irrespective of the ploidy level to identify gene-linked polymorphisms that are expected to facilitate efficient Gene Assisted Breeding (GAB), genotyping and population genetics studies. The identified SNP haplotypes reveal some of the allelic differences in the two sweet cherry cultivars analyzed. The identification of these SNP and haplotype markers is expected to significantly improve the genomic resources for sweet cherry and facilitate efficient GAB in this non-model crop.

Figure 1

General schema for rapid identification of SNPs. The method consists of four stages 1. Sample preparation, 2. Sequence data generation, 3. Data processing and 4. Variation screening or validation of polymorphism. Content in parentheses denotes the materials, software and methods used in this study. The variable polymorphic regions can facilitate efficient gene assisted breeding (GAB), genotyping and population genetics studies.

Figure 2

Analysis of variation of identified SNPs via high resolution melting (HRM) curves generated on 8 cultivars used in this study. HRM derivative plots, -(d/dT) fluorescence as a function of temperature, of several primer sets when analyzing 8 sweet cherry cultivars representing the common patterns observed during analysis. Comparisons outside one frame are not meaningful and the frames are not to scale with each other as the curve shape is the focus. A-C are from primers amplifying a region expected to contain 1 putative SNP while D contains 2, E has 3, and F contains 5. A. Primer set 121 produces a single curve pattern denoted by an arrow representing either a homozygous locus across all 8 cultivars tested or a heterozygous locus shared by all 8 tested cultivars. B. Primer set 100 has three distinct curve patterns highlighted as 1, 2 and 3 representing three allelic forms at the sampled locus. C. Primer set 115 has an indiscernible pattern. D-F. Each demonstrates variation in the population; however, the more SNPs present in the amplified region the smaller the differences among the melt curves.

Figure 3

Pedigree relationships of the 8 cultivars used in this study. Pedigree of the sweet cherry cultivars used for SNP development (blue box) and those used for HRM analysis of the SNPs (bold and all caps). The maternal parent is marked by a red line and the paternal parent by a blue line.

Figure 4

Four primer sets with the HRM curve for the two parents, Cowiche and Selah, on the left and the 13 seedlings on the right. HRM derivative plots, (-d/dT) fluorescence as a function of temperature. A-C contain 1 putative SNP while D. contains 2 putative SNPs. A. Primer set 131 shows no variation as expected for crossing two of the same homozygotes. Note the single curve profile in both the parents and seedlings. B. Primer set 189 shows a single curve profile in the parent panel which differentiates into a 1:2:1 (3:6:4) genotype ratio as expected for a heterozygous × heterozygous cross in the seedlings. It is represented by three different curve profiles where profile number 1 corresponds to the heterozygous parental profiles. C. Primer set 100 shows a 1:1 (8:5) genotype ratio as expected in a homozygous and heterozygous cross. The curve profiles are similar between the seedlings and the parents. D. Primer 92 shows 2 parental curve types. However, the seedlings show several distinct curve types which is not unexpected due to the presence of 2 high quality SNPs in this region.

Figure 5

Screenshot of the SeqMan (DNASTAR) visualization of contig456 showing 2 alleles at a single locus. Boxes 1 and 2 represent unique haplotypes obtained from the NGen 3.0 assembly of the 454 reads from Bing and Rainier according to the filtering parameters described in the methods. These haplotypes differ at each of the bases labeled in green on one of the haplotypes for a total of 10 SNPs between these haplotypes. Haplotype 1 consists of 11 reads from Bing and 1 of Rainier while haplotype 2 is entirely Rainier.