6 Jan 2014:
Understanding relationships among species is a fundamental goal of evolutionary biology. Single nucleotide polymorphisms (SNPs) identified through next generation sequencing and related technologies enable phylogeny reconstruction by providing unprecedented numbers of characters for analysis. One approach to SNP-based phylogeny reconstruction is to identify SNPs in a subset of individuals, and then to compile SNPs on an array that can be used to genotype additional samples at hundreds or thousands of sites simultaneously. Although powerful and efficient, this method is subject to ascertainment bias because applying variation discovered in a representative subset to a larger sample favors identification of SNPs with high minor allele frequencies and introduces bias against rare alleles. Here, we demonstrate that the use of hybridization intensity data, rather than genotype calls, reduces the effects of ascertainment bias. Whereas traditional SNP calls assess known variants based on diversity housed in the discovery panel, hybridization intensity data survey variation in the broader sample pool, regardless of whether those variants are present in the initial SNP discovery process. We apply SNP genotype and hybridization intensity data derived from the Vitis9kSNP array developed for grape to show the effects of ascertainment bias and to reconstruct evolutionary relationships among Vitis species. We demonstrate that phylogenies constructed using hybridization intensities suffer less from the distorting effects of ascertainment bias, and are thus more accurate than phylogenies based on genotype calls. Moreover, we reconstruct the phylogeny of the genus Vitis using hybridization data, show that North American subgenus Vitis species are monophyletic, and resolve several previously poorly known relationships among North American species. This study builds on earlier work that applied the Vitis9kSNP array to evolutionary questions within Vitis vinifera and has general implications for addressing ascertainment bias in array-enabled phylogeny reconstruction.
Citation: Miller AJ, Matasci N, Schwaninger H, Aradhya MK, Prins B, Zhong G-Y, et al. (2013) Vitis Phylogenomics: Hybridization Intensities from a SNP Array Outperform Genotype Calls. PLoS ONE 8(11): e78680. doi:10.1371/journal.pone.0078680
Editor: Ting Wang, Wuhan Botanical Garden, Chinese Academy of Sciences, Wuhan, China
Received: June 27, 2013; Accepted: September 15, 2013; Published: November 13, 2013
This is an open-access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication.
Funding: This article was written, in part, thanks to funding from the Canada Research Chairs program and the National Sciences and Engineering Research Council of Canada. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Understanding relationships among species is the basis for modern classification schemes and provides the requisite framework for ecological and evolutionary analyses of diversity patterns and diversification processes , . Large-scale coordinated research programs, together with technical and analytical advances, have facilitated significant progress in current understanding of organismal phylogeny. Despite this, uncertainty regarding evolutionary relationships among species remains in many groups, including several that include economically important species such as apples , , grapes –, potatoes , and wheat .
Over the past five years, nearly all sub-disciplines within biology have been revolutionized in the wake of the genomics era , . Widespread adoption of next-generation sequencing (NGS) technologies have reduced the cost of DNA sequencing by orders of magnitude providing unprecedented access to the genome of an organism (www.genome.gov/sequencingcosts). One application of NGS is single nucleotide polymorphism (SNP) discovery through whole-genome sequencing or comparative sequence analysis of expressed sequence tags (ESTs) or reduced-representation libraries (RRLs) –. Resulting SNPs can be used to construct a SNP array, a compilation of hundreds, thousands, or even millions of polymorphic sites that enables genotyping of an individual at multiple loci simultaneously (e.g., ). To date, SNP arrays have been developed primarily in systems for which large amounts of genomic data are already available, including model organisms with sequenced genomes or domesticated species with significant EST libraries –. In combination with phenotypic data, SNP arrays have been used extensively in linkage mapping (e.g., ), association genetics (e.g., ), and genome-wide association studies ,  and have been particularly useful in screening variation in crop species . High-throughput genotyping via SNP arrays has contributed to current understanding of the genetic basis of agriculturally important traits and is supporting crop improvement efforts by accelerating marker-assisted selection and genomic selection .
In addition to crop improvement, SNP microarray technology holds great promise for studying evolutionary processes that shape variation in natural populations –. For example, SNP arrays have been used to characterize the genetic basis of local adaptation in Arabidopsis  , Douglas fir , loblolly pine , poplar , and Sitka Spruce , among others. The convenience of genotyping thousands of sites at the same time, together with the economy of scale, has propelled the use of array-generated genotypic data in a variety of evolutionary questions.
Phylogeny reconstruction based on genome-wide data (“phylogenomics”) is an exciting and important development in evolutionary biology , . SNP arrays present a potentially valuable source of data for this purpose and have already been used to genotype large numbers of individuals across multiple species. For example, evolutionary relationships among higher ruminants (e.g., cattle, sheep, goats, antelopes, deer, giraffes, pronghorn) were estimated using the Bovine SNP50 BeadChip, an array developed from variation detected among six cattle breeds and from heterozygous sites in the sequenced cattle genome , . Phylogenomic analyses based on 678 animals representing 61 species genotyped at more than 40,000 SNP sites yielded support for established clades and identified several new relationships. Similar studies have been completed in humans , horses and their wild relatives , and old world monkeys .
Utilization of SNP arrays involves applying variation discovered in one or a few individuals to a large range of accessions , . The number and diversity of individuals used in the SNP discovery process (the discovery panel) almost always leads to some degree of ascertainment bias because the discovery panel consists of only a small subset of the individuals to be genotyped on the array , . Frequently, the discovery panel favors identification of SNPs with high minor allele frequencies, introducing bias against rare alleles . Ascertainment bias becomes particularly acute when SNPs identified for one level of analysis (e.g., within species comparisons) are used at different scales (e.g., among species comparisons, as in phylogeny reconstruction) , . Indeed, it has been shown that the application of SNPs identified in a discovery panel to a broad set of samples is accompanied by losses in utility, particularly as genotyping is attempted for individuals that are increasingly evolutionarily divergent from the panel accessions –. We expect ascertainment bias to be particularly severe when assaying variation across a highly diverse genus like Vitis, where common ancestry between species is expected to date back tens of millions of years , .
Several approaches to reduce the effects of ascertainment bias have been proposed (reviewed in , , one of which involves the use of hybridization intensity data rather than genotype calls. Hybridization intensity data capture otherwise undetectable variation in SNP array data known as “off-target variants”, variation in genomic DNA that differs from the expected variant targeted by the array design . Characterizing site variation without directly querying alternative alleles at a locus has been used to identify polymorphisms between maize inbred lines , , in association mapping in Arabidopsis , and in phylogeny reconstruction . Summary statistics of fluorescence intensity values have been shown to outperform bi-allelic genotype calls for the purposes of linkage mapping in grape (Myles et al. unpublished data). Whereas traditional SNP calls assess known variants based on diversity housed in the discovery panel, hybridization intensity data characterize variation in the broader sample pool, regardless of whether or not those variants are present in the individuals used in the initial SNP discovery process.
In this study, we apply SNP genotype and hybridization intensity data derived from the Vitis9kSNP array developed for grape ,  to characterize the effects of ascertainment bias and to reconstruct evolutionary relationships among Vitis species. A North Temperate genus comprising approximately 60 species, Vitis includes at least 14 species and three named hybrid taxa native to North America, one species complex in Europe (the cultivated grape V. vinifera ssp. vinifera (“vinifera”) and its wild progenitor V. vinifera ssp. sylvestris (“sylvestris”) –, and 37 species in China , . Previous phylogenetic analyses have demonstrated that Vitis is monophyletic and consists of two subgenera, subgenus Muscadinia (N = 2–3 North American species) and subgenus Vitis (N = ~60 species found in North America, Europe, and Asia) –, , , –. To date, chloroplast and nuclear sequence data, amplified fragment length polymorphism (AFLP), and microsatellites have been employed to describe the evolutionary relationships among subgenus Vitis species –, , ; these studies have generated support for some relationships within the genus, but several questions remain. Most notably, it is unclear if the North American subgenus Vitis species are monophyletic, and species-level relationships within the North American clades of subgenus Vitis remain largely unresolved.
Vitis presents an ideal system in which to explore the utility of SNP array data for phylogenetic analysis and to assess the effects of ascertainment bias on phylogeny reconstruction. This study system exhibits many attributes believed to exacerbate ascertainment bias: 1) Vitis is highly heterozygous; 2) common ancestry between species dates to at least 10 million years ago , ; and 3) the Vitis9kSNP array discovery panel was built using 17 individuals (eleven V. vinifera cultivars, one individual each of V. amurensis, V. cinerea, V. labrusca, V. palmata, V. rotundifolia, and V. vinifera ssp. sylvestris) but has been used to survey larger numbers of samples from a variety of taxa. In addition, previous phylogenetic analyses of Vitis have demonstrated consistent support for some relationships, for example, the progenitor-descendant relationship between sylvestris and the cultivated grape vinifera. Clades like this present an opportunity to evaluate whether genotype data or hybridization intensity data (or both) have the capacity to recover known relationships.
Here, we use the Vitis9kSNP array to characterize variation in approximately one third of Vitis species, genotyping over 1100 accessions at nearly 9000 sites . We demonstrate that phylogenies constructed using hybridization intensities suffer less from the distorting effects of ascertainment bias, and are thus more accurate, than phylogenies based on genotype calls. Moreover, we reconstruct the phylogeny of the genus Vitis using hybridization data, provide evidence to suggest that North American subgenus Vitis species are monophyletic, and identify several species-level relationships among North American Vitis species. This study builds on previous work that applied the Vitis9kSNP array to evolutionary questions within Vitis ; Myles et al. unpublished data), and has general implications for addressing ascertainment bias in array-enabled phylogeny reconstruction.
Leaves for DNA extraction were collected from the USDA grape germplasm collections in Davis, California, and Geneva, New York. Permission for tissue collection was obtained from the local USDA authorities. DNA was extracted using DNeasy Plant Mini Kits (Qiagen) and 1173 accessions representing 19 taxa (16 unique species, two hybrid taxa, one species with two intra-specific groups) were genotyped using the Vitis9kSNP array, which includes 8898 SNPs ,  (Table 1).
Genotype data curation
An initial principal components analysis (PCA) was conducted in R using the genotype calls from the Vitis9kSNP array in order to examine whether or not individuals clustered according to their assigned species. SNPs with low genotype quality scores (GenCall<0.2), low SNP quality scores (GenTrain score<0.3), MAF<0.05 and >20% missing data were excluded, which resulted in a data set of 4073 SNPs. For PCA, SNPs were pruned for linkage disequilibrium (LD) using PLINK  by considering a window of 10 SNPs, removing one of a pair of SNPs if LD>0.5, and then shifting the window by three SNPs and repeating the procedure (plink command: –indep-pairwise 10 3 0.5). After these filters, 3231 SNPs remained for PCA. PCA was run and individuals representing obvious curation errors (i.e. those carrying one species name but obviously clustering with individuals from another species) were removed from the remaining analyses. After these data curation steps, 1030 samples remained from 18 different taxa. Genotype and intensity data are available in the dryad digital repository.
Analyses of genotype data
To facilitate direct comparison between genotype data and hybridization intensity data, genotypes were used to calculate FST among species. Only SNPs with MAF>0.05 and <20% missing data were included, resulting in 4073 SNPs and 1030 samples. We calculated a weighted average FST between all pairs of species following equation 10 in . The resulting FST distance matrix was visualized with a multi-dimensional scaling (MDS) plot. The FST distance matrix was then used to construct phylogenetic trees using the “nj” function in the ape package in R . Neighbour-joining (NJ) trees rooted with V. rotundifolia, a representative of subgenus Muscadinia, were generated. To assess the impact of V. vinifera and V. sylvestris on the analysis, phylogenetic trees were constructed for the full dataset, as well as a reduced dataset with V. vinifera and V. sylvestris removed.
Analyses of hybridization intensity data
We investigated whether the effects of ascertainment bias on phylogenetic structure could be circumvented using normalized intensity data. Instead of forcing the intensity values generated from the probes on the array into categorical variables, i.e. genotype calls, we used the normalized intensity values as “quantitative genotypes” and calculated genetic distances between species using these scores. To explore the utility of hybridization intensity data in the reconstruction of evolutionary relationships, normalized intensity data from all 8898 SNPs assayed by the Vitis9kSNP array were used to calculate a genetic distance matrix between species. This matrix was generated using the same set of samples and has the same format as the FST distance matrix, facilitating comparison between relationships resolved using data from SNP genotype calls (previous section) and those resolved using the intensity data. For each SNP, the intensity data from the array consist of a normalized intensity for allele A (X) and a normalized intensity for allele B (Y) that captures information from an average of 30 probes querying that particular SNP. We investigated several summary statistics of these intensity values including X, Y, X+Y, X/(X+Y), ln(X/(X+Y)), ln(Y/(X+Y)), and ln(X/Y). To generate a single value for each SNP for each species, the median of the above summary statistics for each SNP was calculated for each species. Each of these matrices of summary statistics was converted into a distance matrix by calculating the Euclidean distances between each pair of species using the “dist” function in R. MDS plots were generated from these distance matrices to evaluate how well the intensity data captured relationships among samples. Distance matrices based on hybridization intensity were compared to one another and to the FST distance matrix generated from the genotype calls using mantel tests with 10000 permutations. For each summary statistic described above, rooted trees (with V. rotundifolia as the root) were generated. Topologies of pairs of trees were compared using the method of  where the “distance” between two trees is defined as twice the number of internal branches defining different bipartitions of the tips.
Assessment of curation error
Using PCA, we identified and removed 54 samples that clearly did not cluster according to their species membership, and likely represented curation errors in the collection. These samples represent approximately 5% of the genotyped samples from the USDA grape germplasm collection. Using PCA, we demonstrate that, after excluding these curation errors, the samples used in the present study indeed cluster according to their taxonomic identity (Fig. 1). The removal of V. rotundifolia, V. sylvestris, and V. vinifera (Figs. 1b–d) shows that, even for the North American and Eurasian species, sample mix up or curation errors are unlikely to contribute to false phylogenetic inferences.
The proportion of the variance explained is found within parentheses on each axis. A) PCA with all samples included. PC1 clearly separates vinifera and sylvestris from the other wild Vitis species while PC2 separates rotundifolia from all others. B–D) PCA with rotundifolia, sylvestris and vinifera removed. Examining the distances between individual samples in PC space confirms that the curated sample set used in the present study does not likely suffer from mislabeling or curation error that would lead to false phylogenetic inference.
Assessment of ascertainment bias
The Vitis9kSNP array was constructed primarily to assay polymorphism within V. vinifera, with only a few probes designed specifically to query fixed differences among various Vitis species . While the SNP data clearly group individuals according to their taxonomic identity (Fig. 1), we find pervasive evidence of ascertainment bias. For example, the minor allele frequency (MAF) distribution in vinifera and its closely related ancestor sylvestris shows a large excess of intermediate frequency alleles relative to other wild Vitis species examined (Fig. 2). This pattern of MAF distributions across species is expected as most of the SNPs selected for the array were chosen specifically because they segregate within vinifera. This observed pattern of MAF distributions across species also means that pairs of wild species are fixed for identical or alternative alleles at many SNPs across the genome, while comparisons between vinifera or sylvestris and any other wild species will tend to involve an intermediate frequency allele compared to an allele found at a frequency of either 0 or 1. We demonstrate this by showing that species pairwise comparisons involving vinifera or sylvestris exhibit many fewer SNPs that are fixed for the same allele compared to species pairwise comparisons not involving vinifera or sylvestris (Fig. 3a). One result of this is that FST values from comparisons involving V. vinifera or V. sylvestris tend to generate intermediate FST values since many SNPs are fixed within a wild Vitis species but segregate within vinifera or sylvestris (Fig. 3b). Moreover, the biased FST values result in false phylogenetic inferences involving vinifera and sylvestris (described below).
MAF allele frequencies for other species look similar to V. coignetiae and V. riparia with a severe deficit of intermediate frequency alleles compared to vinifera and sylvestris.
A) Between-species comparisons with vinifera or sylvestris involve far fewer monomorphic SNPs than other comparisons. The number of monomorphic SNPs was calculated for every pairwise comparison between species. Because vinifera and sylvestris show an excess of intermediate frequency alleles compared to other Vitis species using the Vitis9KSNP array, comparisons involving vinifera or sylvestris display fewer monomorphic sites relative to comparisons involving other species pairs. B) The dotted lines indicated by “min” and “max” are the minimum and maximum FST values from comparisons between vinifera or sylvestris and other species. The ascertainment bias results in intermediate FST estimates with relatively little variation for pairwise comparisons between species involving vinifera or sylvestris.
Genetic distances based on SNP genotypes
Genetic distance among each pair of species was estimated using the FST statistic and MDS plots were used to visualize the resulting genetic distances among all species. NJ trees were rooted with V. rotundifolia and completed for the full filtered dataset of 1030 samples. As was the case using PCA (Fig. 1), V. rotundifolia is clearly distantly related to other Vitis species based on FST values (Figs. 4a, 5). However, vinifera and sylvestris appear misplaced in the MDS plot as they cluster more closely to North American Vitis than to Eurasian Vitis (Figs. 4a, b), which is neither in agreement with their geographic distribution nor with previous work –. Even more striking, phylogenetic analyses of the FST distance matrix of SNP genotypes fail to group vinifera with sylvestris, a well-known progenitor-descendent pair (Fig. 5a). Phylogenetic analysis of the genotype data places sylvestris with other Eurasian species V. amurensis, V. coignetieae, and V. piasezkii while vinifera falls outside of a large clade of North American and Eurasian subgenus Vitis, alongside the sole representative of subgenus Muscadinia, V. rotundifolia (Fig. 5a). This placement of vinifera renders subgenus Vitis non-monophyletic based on SNP genotype calls and is inconsistent with all known evidence and previous work.
A) Neighbour-joining (NJ) tree from FST estimates derived from SNP genotype calls from the Vitis9KSNP array. B) NJ tree from a distance measure derived from hybridization intensities from the Vitis9KSNP array.
Despite the effect of ascertainment bias on inferring relationships to vinifera, the MDS (Figs. 4a–b) and phylogenetic analyses (Fig. 5a) of SNP genotype data resolve some relationships identified in previous Vitis analyses , : 1) a Eurasian cluster in which sylvestris is basal to a group that includes V. piasezkii and V. amurensis+V. coignetiae; and 2) a clade of North American subgenus Vitis species in which V. palmata occupies the basal position; 2a) V. aestivalis+V. labrusca group together with V. cinerea+V. vulpina; and 2b) V. champinii+V. mustangensis form a clade that is sister to a clade of (V. monticola (V. girdiana (V. rupestris (V. riparia+V. acerifolia)))). With respect to Moore's classification scheme , these results support the monophyly of Moore's Series Ripariae (V. acerifolia, V. riparia, and V. rupestris), but do not support the monophyly of Series Cordifoliae (V. monticola Buckley, V. palmata Vahl, V. vulpina Linneaus), or Series Labruscae (V. labrusca Linneaus, V. mustangensis Buckley, V. shuttleworthii House). There is insufficient sampling/taxon identification (subspecific classification is not known for many accessions) to evaluate the monophyly of Moore's Series Aestivales (V. aestivalis Michaux, V. aestivalis var. aestivalis, V. aestivalis var. bicolor Dean, V. aestivalis var. lincecumii (Buckley) Munson) or Series Cinerescentes (V. cinerea Engelmann ex Millardet, V. cinerea var. baileyana (Munson) Comeaux, V. cinerea var cinerea, V. cinerea var. floridana Munson, V. berlandieri Planchon). SNP genotype data presented here corroborate several relationships identified in previous studies –.
Genetic distances based on hybridization intensities
The distance matrices generated from the various intensity data summary statistics (see Methods) were all highly correlated with one another (Mantel test; all pairwise comparisons p<1×10−4). This suggests that, regardless of the summary statistic used, the resulting genetic distance measures among species remain similar. Moreover, we compared phylogenetic tree topologies constructed from distance matrices derived from the various intensity data summary statistics and found that tree topology remains almost identical regardless of the summary statistic employed (Table S1). We therefore chose arbitrarily from among the summary statistics of the hybridization intensities and present results from the use of ln(X/Y). The genetic distance matrix generated from ln(X/Y) values was correlated with the FST distance matrix (Mantel test, p = 0.021). However, the genetic distances derived from intensity values recover a more accurate phylogeny of the genus Vitis than the genetic distances calculated from SNP genotypes (Figs. 4c, d; 5b). Most notably, the intensity data-based phylogenetic analyses resolve vinifera and sylvestris as sister taxa which share a most recent common ancestor with the Eurasian clade of V. piasezkii and V. amurensis+V. coignetiae. This is consistent with other phylogenetic analyses of Vitis that have suggested a close relationship between the cultivated grape and Eurasian Vitis species –.
Similar to the SNP genotype data, the intensity data resolve two clades within subgenus Vitis: 1) a Eurasian subgenus Vitis clade that includes (sylvestris+vinifera) and (V. piasezkii (V. amurensis+V. coignetiae)), and 2) a North American subgenus Vitis clade that includes V. palmata sister to 2a) (V. labrusca+V. aestivalis) and (V. cinerea+V. vulpina) and 2b) V. monticola (V. girdiana (V. rupestris (V. acerifolia+V. riparia))) and (V. champinii+V. mustangensis). Similar to the SNP genotype analysis (described above), the intensity data support the monophyly of Moore's Series Ripariae and fail to support the monophyly of Series Cordifoliae and Series Labruscae. The monophyly of Series Aestivales and Series Cinerscentes cannot be evaluated given present sampling and lack of sub-specific taxon identification). Although the intensity data resolve the two main clades within subgenus Vitis (a clade of North American subgenus Vitis species and a clade of Eurasian subgenus Vitis species), they fail to resolve a monophyletic subgenus Vitis.
This study offers a phylogenomic approach to elucidating relationships in the North Temperate genus Vitis, which includes the most economically important berry species in the world, the cultivated grapevine vinifera. Leveraging a SNP array designed primarily for the cultivated grapevine, polymorphic sites discovered in vinifera and a small group of wild Vitis individuals were screened in over 1100 accessions representing 19 Vitis taxa, and used to reconstruct evolutionary relationships within Vitis. These data suggest that the Vitis9KSNP array suffers from ascertainment bias: SNPs were discovered mainly in vinifera and these SNPs are thus more likely to segregate in vinifera and its closely related ancestor sylvestris than in more distantly related wild Vitis species . We investigated the effects of this ascertainment bias on phylogenetic inferences by analyzing relationships among diverse Vitis taxa using both SNP genotype calls and quantitative genotypes derived from hybridization intensity data. We demonstrate that ascertainment bias is pronounced when SNP genotypes are used to calculate genetic distances among taxa (Figs. 4a,b) and to construct phylogenies (Fig. 5a), leading to the failure to recover known clades. As an alternative to genotype calls plagued by ascertainment bias, summaries of hybridization intensity data provide a more accurate view of relationships among Vitis taxa (Figs. 4c,d; 5b). However, it is worth noting that even the hybridization intensity statistics are affected by ascertainment bias: genetic distance calculations based on intensity values involving vinifera or sylvestris are systematically upward biased (Fig. 6). This is unsurprising as we expect the probes on the Vitis9KSNP array, which were designed based on the Pinot Noir (vinifera cultivar) reference genome, to hybridize better to vinifera and sylvestris samples than to distantly related Vitis species whose sequences are not as complimentary to the probes on the array. Nevertheless, our analyses demonstrate that the severity of ascertainment bias when calling genotypes across diverse taxa results in incorrect phylogenetic inferences, while these obvious phylogenetic errors are not present when using intensity-based genetic distance measures. The data presented here confirm that SNP arrays developed for one taxon (e.g., vinifera) or one purpose (e.g., identifying gene regions associated with traits of agricultural importance) can be co-opted to study evolution and divergence at larger taxonomic scales, and that this work is enhanced significantly by the use of hybridization intensity data.
Each dot represents a pairwise comparison between two species. Pairwise comparisons involving V. vinifera or V. sylvestris are highlighted in red. Genetic distances for V. vinifera and V. sylvestris based on intensity values are systematically elevated compared to other pairwise comparisons.
Ascertainment bias in SNP arrays and the promise of hybridization intensity data in phylogenomics
SNP arrays have been developed for many crop plants  including apple , common bean ; citrus , corn , grape , peach , and rice ,  for the purposes of population genetics, gene discovery, and marker-assisted selection. However, the application of these arrays to broader phylogenomic questions has been limited. Transferability of SNP arrays seems plausible in long-lived perennials that are particularly heterozygous ; for example , used SNPs recovered in the clementine genome to examine evolutionary relationships among over 50 diverse accessions in the complex Citrus genus. Data presented here provide further support for this, suggesting that SNP arrays have tremendous potential for expanding current understanding of evolutionary relationships among crop species and their wild relatives.
Ascertainment bias is known to interfere with population genetic inferences . This study demonstrates that ascertainment bias is especially present in analyses above the species-level. The Vitis phylogeny built using SNP genotype data (Fig. 5a) failed to identify the close evolutionary link between the cultivated vinifera and its wild ancestor sylvestris, a well-known relationship that has been documented using molecular genetic data –. To address the problematic phylogeny that resulted from the ascertainment bias inherent in the genotype calls, we derived quantitative genotypes from the hybridization intensities and used these to estimate genetic distances among species. The resulting intensity-based phylogeny recovered most known clades and suggested other novel relationships not identified in previous analyses (described below).
Implications of phylogenomic analyses for understanding evolutionary relationships within Vitis
Phylogenomic analyses of Vitis based on the Vitis9kSNP array data resolve several clades identified in previous analyses –, – and suggest novel relationships not previously identified. On a broad phylogenetic scale, the hybridization intensity data support the distinction between subgenus Muscadinia (2n = 40) and two subgenus Vitis clades (2n = 38) ; however, neither the hybridization intensity analysis nor the SNP genotype analysis resolved a monophyletic subgenus Vitis (Fig. 5). This may be an example of ascertainment bias that is simply too strong to be overcome with hybridization intensity data. Of the 17 accessions used in the original discovery panel , only one came from subgenus Muscadinia. Perhaps any signal of differentiation between subgenus Muscadinia and subgenus Vitis may have been swamped by the sheer number of sites segregating within vinifera, and among vinifera and other subgenus Vitis taxa.
Subgenus Vitis exhibits a classic Eastern Asian-North American disjunct distribution with one species complex occurring in Eurasia. Although additional sampling representing both Eurasian and North American subgenus Vitis taxa is required to test the monophyly of the these groups, data presented here and in a previous study  indicate two evolutionarily distinct monophyletic groups within subgenus Vitis, one of which occupies Eurasia and the other which occupies North America. Some previous studies resolved a monophyletic Eurasian subgenus Vitis group, but did not support a monophyletic North American clade of subgenus Vitis , . These studies suggested that North American Vitis species are ancestral within subgenus Vitis, and that a Eurasian subgenus Vitis group evolved from within the North American Subgenus Vitis clade. A different group of analyses reported a clade of North American subgenus Vitis species nested within a paraphyletic Asian subgenus Vitis , , and/or various degrees of intermixing among Eurasian and North American subgenus Vitis taxa , , . The evolutionarily and geographically distinct Subgenus Vitis clades identified in this study could have resulted from a vicariant event (continental drift) leading to the geographic separation of Eurasian Vitis and North American Vitis, which was most likely associated by diversification of these groups on their respective continents . A well-documented aspect of the North Temperate disjunct pattern is that genera displaying this geographic distribution generally have more Eurasian species than North American species possibly due to greater net speciation and rates of molecular evolution , . This observation is corroborated in subgenus Vitis, where approximately 37 species have been recorded in Eurasia  and at least ~17 taxa in North America (Moore and Wen, unpublished data).
North American subgenus Vitis species have been grouped by various authors, including M. O. Moore  who designated five series within subgenus Vitis in eastern North America based on morphological features: series Aestivales (includes V. aestivalis), series Cinerescentes (includes V. cinerea), series Cordifoliae (includes V. monticola, V. palmata, and V. vulpina), series Labruscae (includes V. labrusca, V. mustangensis, and V. shuttleworthii), and series Ripariae (includes V. acerifolia, V. riparia, and V. rupestris). Moore's (1991)  key to the series based on morphological features provides a framework of relationships among the series (Aestivales (Cinerscentes (Labruscae (Ripariae, Cordifoliae)))). Previous phylogenetic analyses have provided support for series Ripariae (V. acerifolia, V. riparia, and V. rupestris) , , . Zecca et al.  resolved a clade with V. riparia and V. rupestris, but V. acerifolia grouped with V. arizonica and V. girdiana, among others. All analyses performed here support a sister-taxon relationship between V. acerifolia and V. riparia, which together form a clade with V. rupestris. Although a close relationship between V. riparia and V. rupestris is widely supported, discrepancy in the placement of V. acerifolia may indicate that this species has a hybrid origin derived from a cross between V. riparia or V. rupestris and one of the southwestern species.
Expanding upon the V. acerifolia – riparia – rupestris group, the hybridization intensity data provide evidence for a clade of subgenus Vitis species found primarily in the central-southern-southeastern United States (V. riparia is an exception to this) by placing the V. acerifolia – riparia – rupestris clade with V. monticola, V. mustangensis, and their hybrid derivative V. x champinii (V. mustangensis x V. rupestris) (Fig. 5). Like V. acerifolia and V. rupestris, V. monticola and V. mustangensis are species whose primary distributions are in the central to central-southern United States. Vitis riparia is a widespread climbing vine found throughout the Midwest and the northeastern quarter of the United States. Previous authors grouped V. monticola with V. palmata and V. vulpina based on morphology , but some recent molecular analyses have suggested a relationship between V. monticola, V. mustangensis and the V. acerifolia – riparia – rupestris group  but see . The SNP genotype calls place the Californian species V. girdiana in this group as well, consistent with previous analyses –.
A second major clade within North American subgenus Vitis includes two species pairs: V. aestivalis+V. labrusca and V. cinerea+V. vulpina; V. palmata is basal among all North American subgenus Vitis species. Vitis aestivalis, V. cinerea, V. labrusca, and V. vulpina have largely overlapping distributions in the eastern half of the United States. These species clustered together in earlier studies ; most recently , identified a clade of (V aestivalis+V. labrusca)+V. vulpina, and a second clade of ([V. cinerea+V. palmata)+(V. mustangensis+V. shuttleworthii)]. While both this study and  find support for a close relationship between V. aestivalis and V. labrusca, the positions of V. monticola, V. palmata, and V. vulpina differ in the two analyses. The results of both studies conflict with Moore's ,  classification scheme. For example, Moore's series Cordifoliae includes V. monticola, V. palmata, and V. vulpina; analyses presented here suggest V. palmata is basal in North American subgenus Vitis and that V. vulpina forms a clade with V. cinerea. Similarly, Moore's  series Labrusceae posits a close relationship between V. labrusca, V. mustangensis, and V. shuttleworthii (not sampled in this study). However, data presented here suggest V. labrusca forms a clade with V. aestivalis, and that V. mustangensis groups with V. acerifolia, V. monticola, V. riparia, and V. rupestris.
Phylogenetic relationships of crop wild relatives can provide insights into the evolutionary history of a crop as well as a window into contemporary evolutionary processes such as hybridization between cultivated populations and wild progenitors or processes driving divergence among closely related species (e.g., ). In the case of grape, the wild progenitor and geographic origins of domesticated European grapevine are well known , , . However, lesser-known components of grapevine evolutionary biology include relationships among species that are used as parents in hybrid crosses (e.g., V. aestivalis, V. labrusca, V. vinifera) or those that are used as rootstocks (e.g., V. cinerea var. helleri, V. riparia, V. rupestris). For example, grafting vinifera scions to rootstocks of non-vinifera species dates back to the mid-1900's when the phylloxera invasion of France threatened to destroy the French grape crop . Rootstocks used to support vinifera come almost exclusively from North American species . Recently, hybrids between V. cinerea var. helleri and V. riparia or V. rupestris have been used to produce rootstock that is easy to propagate and that can withstand challenging abiotic conditions . Data presented here demonstrate that these important rootstock species occur in different clades within the North American subgenus Vitis: V. cinerea is most closely related to V. vulpina, while V. riparia and V. rupestris form a clade together with V. acerifolia. Building upon this phylogenetic framework, future work characterizing the diversity of abiotic and biotic pressures faced by natural populations and the genetic basis of abiotic and biotic stress response, will expand understanding of evolution and adaptation in Vitis, and may provide molecular tools to facilitate marker-assisted selection for rootstocks.
This study demonstrates that ascertainment bias presents a significant challenge for the application of SNP arrays in phylogenetic reconstruction; however, the effects of ascertainment bias can be minimized by using hybridization intensity rather than SNP genotype calls. We demonstrate that the Vitis9kSNP array, a panel developed based on variation discovered in 11 accessions of vinifera and single accessions of six other Vitis species, can be used to screen variation in a broad sample of over 1100 samples representing 18 taxa. Resulting data confirm relationships identified in previous studies (e.g., V. riparia+V. rupestris, vinifera+sylvestris) and suggest novel affinities among taxa (e.g., V. aestivalis+V. labrusca and V. cinerea+V. vulpina). This phylogenomic analysis of Vitis demonstrates the utility of SNP arrays in phylogeny reconstruction and expands current understanding of relationships among North American subgenus Vitis species.
Comparison of the performance of various genetic distance measures based on hybridization intensity. The values within each cell represent a measure of the difference between tree topologies generated from the summary statistics found in the respective row and column names. The distance measure between the pair of phylogenetic trees is defined as the twice the number of internal branches defining different bipartitions of the tips (Penny and Hendy 1985). All of the summary statistics generated phylogenetic trees with highly similar topologies.
We thank the USDA staff that maintains the USDA grape germplasm collections. The authors acknowledge the contributions Kyle Gardner and Jeff Franklin, and members of the Miller Lab at Saint Louis University for helpful comments on previous versions of this manuscript.
Conceived and designed the experiments: GZ CS ESB MA SM. Performed the experiments: HS SM. Analyzed the data: NM AM SM. Contributed reagents/materials/analysis tools: GZ CS HS BP MA ESB. Wrote the paper: AM SM.
- 1. Donoghue MJ (2008) A phylogenetic perspective on the distribution of plant diversity. Proceedings of the National Academy of Sciences 105: 11549–11555. doi: 10.1073/pnas.0801962105
- 2. Webb C, Ackerly D, McPeek M, Donoghue M (2002) Phylogenies and community ecology. Annu Rev Ecol Syst 33: 475–505. doi: 10.1146/annurev.ecolsys.33.010802.150448
- 3. Coart E, Van Glabeke S, De Loose M, Larsen A, Roldán-Ruiz I (2006) Chloroplast diversity in the genus Malus: new insights into the relationship between the European wild apple (Malus sylvestris (L.) Mill.) and the domesticated apple (Malus domestica Borkh.). Mol Ecol 15: 2171–2182. doi: 10.1111/j.1365-294x.2006.02924.x
- 4. Lo EY, Donoghue M (2012) Expanded phylogenetic and dating analyses of the apples and their relatives (Pyreae, Rosaceae). Mol Phylogenet Evol 63: 230–243. doi: 10.1016/j.ympev.2011.10.005
- 5. Zecca G, Abbott JR, Sun W, Spada A, Sala F, et al. (2012) The timing and the mode of evolution of wild grapes (Vitis). Mol Phylogenet Evol 62: 736–747. doi: 10.1016/j.ympev.2011.11.015
- 6. Aradhya M, Wang Y, Walker M, Prins B, Koehmstedt A, et al. (2013) Genetic diversity, structure, and patterns of differentiation in the genus Vitis. Plant Syst Evol 299: 317–330. doi: 10.1007/s00606-012-0723-4
- 7. Wan Y, Schwaninger HR, Baldo AM, Labate JA, Zhong GY, et al. (2013) A Phylogenetic analysis of the grape genus (Vitis) reveals broad reticulation and concurrent diversification during Neogene and Quaternary Climate Change. BMC Evol Biol 13: 141. doi: 10.1186/1471-2148-13-141
- 8. Cai D, Rodríguez F, Teng Y, Ané C, Bonierbale M, et al. (2012) Single copy nuclear gene analysis of polyploidy in wild potatoes (Solanum section Petota). BMC Evol Biol 12. doi: 10.1186/1471-2148-12-70
- 9. Golovnina K, Glushkov S, Blinov A, Mayorov V, Adkison L, et al. (2007) Molecular phylogeny of the genus Triticum L. Plant Syst Evol 264: 195–216. doi: 10.1007/s00606-006-0478-x
- 10. Harrison N, Kidner CA (2011) Next-generation sequencing and systematics: What can a billion base pairs of DNA sequence data do for you? Taxon 60: 1552–1566.
- 11. Carstens B, Lemmon AR, Lemmon EM (2012) The promises and pitfalls of next-generation sequencing data in phylogeography. 6: 713–715.
- 12. Elshire RJ, Glaubitz JC, Sun Q, Poland J, Kawamoto K, et al. (2011) A Robust, Simple Genotyping-by-Sequencing (GBS) Approach for High Diversity Species. PLoS ONE 6: e19379. doi: 10.1371/journal.pone.0019379
- 13. McCue M, Bannasch D, Petersen J, Gurr J, Bailey E, et al. (2012) A High Density SNP Array for the Domestic Horse and Extant Perissodactyla: Utility for Association Mapping, Genetic Diversity, and Phylogeny Studies. Plos Genetics 8: e1002451. doi: 10.1371/journal.pgen.1002451
- 14. Myles S, Chia J, Hurwitz B, Simon C, Zhong GY, et al. (2010) Rapid Genomic Characterization of the Genus Vitis. Plos One 5: e8219. doi: 10.1371/journal.pone.0008219
- 15. Pavy N, Pelgas B, Beauseigle S, Blais S, Gagnon F, et al. (2008) Enhancing genetic mapping of complex genomes through the design of highly-multiplexed SNP arrays: application to the large and unsequenced genomes of white spruce and black spruce. BMC Genomics 9: 21. doi: 10.1186/1471-2164-9-21
- 16. Stoneking M, Krause J (2011) Learning about human population history from ancient and modern genomes. Nature Publishing Group 12: 603–614. doi: 10.1038/nrg3029
- 17. Antanaviciute L, Fernández-Fernández F, Jansen J, Banchi E, Evans KM, et al. (2012) Development of a dense SNP-based linkage map of an apple rootstock progeny using the Malus Infinium whole genome genotyping array. BMC Genomics 13: 203. doi: 10.1186/1471-2164-13-203
- 18. Jones F, Chan Y, Schmutz J, Grimwood J, Brady S, et al. (2012) A Genome-wide SNP Genotyping Array Reveals Patterns of Global and Repeated Species-Pair Divergence in Sticklebacks. Curr Biol 22: 83–90. doi: 10.1016/j.cub.2011.11.045
- 19. Verde I, Bassil N, Scalabrin S, Gilmore B, Lawley C, et al. (2012) Development and Evaluation of a 9K SNP Array for Peach by Internationally Coordinated SNP Detection and Validation in Breeding Germplasm. PLoS ONE 7: e35668. doi: 10.1371/journal.pone.0035668
- 20. Neves LG, Mamani EM, Alfenas AC, Kirst M, Grattapaglia D (2011) A high-density transcript linkage map with 1,845 expressed genes positioned by microarray-based Single Feature Polymorphisms (SFP) in Eucalyptus. BMC Genomics 12: 189. doi: 10.1186/1471-2164-12-189
- 21. Akey J, Ruhe A, Akey D, Wong A, Connelly C, et al. (2010) From the Cover: Tracking footprints of artificial selection in the dog genome. Proceedings of the National Academy of Sciences 107: 1160–1165. doi: 10.1073/pnas.0909918107
- 22. Huang X, Wei X, Sang T, Zhao Q, Feng Q, et al. (2010) Genome-wide association studies of 14 agronomic traits in rice landraces. Nat Genet 42: 961–967. doi: 10.1038/ng.695
- 23. Weng J, Xie C, Hao Z, Wang J, Liu C, et al. (2011) Genome-wide association study identifies candidate genes that affect plant height in Chinese elite maize (Zea mays L.) inbred lines. PLoS One 6: e29229. doi: 10.1371/journal.pone.0029229
- 24. Ganal M, Polley A, Graner E, Plieske J, Wieseke R, et al. (2012) Large SNP arrays for genotyping in crop plants. Journal of Biosciences 37: 821–828. doi: 10.1007/s12038-012-9225-3
- 25. Morrell PL, Buckler E, Ross-Ibarra J (2012) Crop genomics: advances and applications. Nature Publishing Group 13: 85–96. doi: 10.1038/nrg3097
- 26. Borevitz JO, Chory J (2004) Genomics tools for QTL analysis and gene discovery. Curr Opin Plant Biol 7: 132–136. doi: 10.1016/j.pbi.2004.01.011
- 27. Gilad Y, Borevitz J (2006) Using DNA microarrays to study natural variation. Curr Opin Genet Dev 16: 553–558. doi: 10.1016/j.gde.2006.09.005
- 28. Sacks B, Louie S (2008) Using the dog genome to find single nucleotide polymorphisms in red foxes and other distantly related members of the Canidae. Molecular Ecology Resources 8: 35–49. doi: 10.1111/j.1471-8286.2007.01830.x
- 29. Shiu S, Borevitz J (2008) The next generation of microarray research: applications in evolutionary and ecological genomics. Heredity 100: 141–149. doi: 10.1038/sj.hdy.6800916
- 30. Garvin M, Saitoh K, Gharrett A (2010) Application of single nucleotide polymorphisms to non-model species: a technical review. Molecular Ecology Resources 10: 915–934. doi: 10.1111/j.1755-0998.2010.02891.x
- 31. Horton MW, Hancock AM, Huang YS, Toomajian C, Atwell S, et al. (2012) Genome-wide patterns of genetic variation in worldwide Arabidopsis thaliana accessions from the RegMap panel. Nature Publishing Group 44: 212–216. doi: 10.1038/ng.1042
- 32. Lasky J, Des Marais D, McKay J, Richards J, Juenger T, et al. (2012) Characterizing genomic variation of Arabidopsis thaliana: the roles of geography and climate. Mol Ecol 21: 5512–5529. doi: 10.1111/j.1365-294x.2012.05709.x
- 33. Eckert A, Bower A, Wegrzyn J, Pande B, Jermstad K, et al. (2009) Association Genetics of Coastal Douglas Fir (Pseudotsuga menziesii var. menziesii, Pinaceae). I. Cold-Hardiness Related Traits. Genetics 182: 1289–1302. doi: 10.1534/genetics.109.102350
- 34. Eckert AJ, Bower AD, Gonzalez-Martinez SC, Wegrzyn JL, Coop G, et al. (2010) Back to nature: ecological genomics of loblolly pine (Pinus taeda, Pinaceae). Mol Ecol 19: 3789–3805. doi: 10.1111/j.1365-294x.2010.04698.x
- 35. Keller S, Levsen N, Olson M, Tiffin P (2012) Local Adaptation in the Flowering-Time Gene Network of Balsam Poplar, Populus balsamifera L. Mol Biol Evol 29: 3143–3152. doi: 10.1093/molbev/mss121
- 36. Holliday J, Ritland K, Aitken S (2010) Widespread, ecologically relevant genetic markers developed from association mapping of climate-related traits in Sitka spruce (Picea sitchensis). New Phytol 188: 501–514. doi: 10.1111/j.1469-8137.2010.03380.x
- 37. Straub S, Parks M, Weitemier K, Fishbein M, Cronn R, et al. (2012) Navigating the tip of the genomic iceberg: Next-generation sequencing for plant systematics. Am J Bot 99: 349–364. doi: 10.3732/ajb.1100335
- 38. Decker J (2009) Resolving the evolution of extant and extinct ruminants with high-throughput phylogenomics. 1–7. doi: 10.1073/pnas.0904691106
- 39. Matukumalli L, Lawley C, Schnabel R, Taylor J, Allan M, et al. (2009) Development and Characterization of a High Density SNP Genotyping Assay for Cattle. PLoS ONE 4: e5350. doi: 10.1371/journal.pone.0005350
- 40. Malhi R, Trask J, Shattuck M, Johnson J, Chakraborty D, et al. (2011) Genotyping single nucleotide polymorphisms (SNPs) across species in Old World Monkeys. Am J Primatol 73: 1031–1040. doi: 10.1002/ajp.20969
- 41. Pindo M, Vezzulli S, Coppola G, Cartwright D, Zharkikh A, et al. (2008) SNP high-throughput screening in grapevine using the SNPlex™ genotyping system. BMC Plant Biol 8: 12. doi: 10.1186/1471-2229-8-12
- 42. Terol J, Naranjo MA, Ollitrault P, Talon M (2008) Development of genomic resources for Citrus clementina: characterization of three deep-coverage BAC libraries and analysis of 46,000 BAC end sequences. BMC Genomics 9: 423. doi: 10.1186/1471-2164-9-423
- 43. Nielsen R (2000) Estimation of population parameters and recombination rates from single nucleotide polymporhisms. Genetics 1–12.
- 44. Albrechtsen A, Nielsen F, Nielsen R (2010) Ascertainment Biases in SNP Chips Affect Measures of Population Divergence. Mol Biol Evol 27: 2534–2547. doi: 10.1093/molbev/msq148
- 45. Wang Y, Nielsen R (2011) Estimating population divergence time and phylogeny from single-nucleotide polymorphisms data with outgroup ascertainment bias. Mol Ecol 21: 974–986. doi: 10.1111/j.1365-294x.2011.05413.x
- 46. Helyar S, Hemmer-Hansen J, Bekkevold D, Taylor M, Ogden R, et al. (2011) Application of SNPs for population genetics of nonmodel organisms: new opportunities and challenges. Molecular Ecology Resources 11: 123–136. doi: 10.1111/j.1755-0998.2010.02943.x
- 47. Vezzulli S, Micheletti D, Riaz S, Pindo M, Viola R, et al. (2008) A SNP transferability survey within the genus Vitis. BMC Plant Biol 8: 128. doi: 10.1186/1471-2229-8-128
- 48. Bradbury I, Hubert S, Higgins B, Bowman S, Paterson I, et al. (2011) Evaluating SNP ascertainment bias and its impact on population assignment in Atlantic cod, Gadus morhua. Molecular Ecology Resources 11: 218–225. doi: 10.1111/j.1755-0998.2010.02949.x
- 49. Didion JP, Yang H, Sheppard K, Fu C, McMillan L, et al. (2012) Discovery of novel variants in genotyping arrays improves genotype retention and reduces ascertainment bias. BMC Genomics 13: 34. doi: 10.1186/1471-2164-13-34
- 50. Miller J, Kijas J, Heaton M, McEwan J, Coltman D (2012) Consistent divergence times and allele sharing measured from cross-species application of SNP chips developed for three domestic species. Molecular Ecology Resources 12: 1145–1150. doi: 10.1111/1755-0998.12017
- 51. Ollitrault P, Terol J, Garcia-Lor A, Bérard A, Chauveau A, et al. (2012) SNP mining in C. clemintina BAC end sequences; transferability in the Citrus genus (Rutaceae), phylogenetic inferences and perspectives for genetic mapping. BMC Genomics 13: 13. doi: 10.1186/1471-2164-13-13
- 52. Springer N, Ying K, Fu Y, Ji T, Yeh C, et al. (2009) Maize Inbreds Exhibit High Levels of Copy Number Variation (CNV) and Presence/Absence Variation (PAV) in Genome Content. PLoS Genetics 5: e1000734. doi: 10.1371/journal.pgen.1000734
- 53. Fu Y, Springer N, Ying K, Yeh C, Iniguez A, et al. (2010) High-Resolution Genotyping via Whole Genome Hybridizations to Microarrays Containing Long Oligonucleotide Probes. PLoS One 5: e14178. doi: 10.1371/journal.pone.0014178
- 54. Kim S, Zhao K, Jiang R, Molitor J, Borevitz J, et al. (2006) Association mapping with single-feature polymorphisms. Genetics 173: 1125–1133. doi: 10.1534/genetics.105.052720
- 55. Myles S, Boyko A, Owens C, Brown P, Grassi F, et al. (2011) From the Cover: Genetic structure and domestication history of the grape. Proceedings of the National Academy of Sciences 108: 3530–3535. doi: 10.1073/pnas.1009363108
- 56. Soejima A, Wen J (2006) Phylogenetic analysis of the grape family (Vitaceae) based on three chloroplast markers. Am J Bot 93: 278–287. doi: 10.3732/ajb.93.2.278
- 57. Arroyo-García R, Ruiz-García L, Bolling L, Ocete R, López M, et al. (2006) Multiple origins of cultivated grapevine (Vitis vinifera L. ssp. sativa) based on chloroplast DNA polymorphisms. Mol Ecol 15: 3707–3714. doi: 10.1111/j.1365-294x.2006.03049.x
- 58. Ren H, Wen J (2007) Vitis. Flora of China 12: 210–222.
- 59. Wen J, Nie Z, Soejima A, Meng Y (2007) Phylogeny of Vitaceae based on the nuclear GAI1gene sequences. Can J Bot 85: 731–745. doi: 10.1139/b07-119
- 60. Trondle D, Schroder S, Kassemeyer H, Kiefer C, Koch M, et al. (2010) Molecular phylogeny of the genus Vitis (Vitaceae) based on plastid markers. Am J Bot 97: 1168–1178. doi: 10.3732/ajb.0900218
- 61. Wada (2008) Systematics and Evolution of Vitis. Davis: University of California. 102 p.
- 62. Ren H, Lu L-M, Soejima A, Luke Q, Zhang D-X, et al. (2012) Phylogenetic analysis of the grape family (Vitaceae) based on the noncoding plastid trnC-petN, trnH-psbA, and trnL-F sequences. Taxon 60: 629–637.
- 63. Péros J, Berger G, Portemont A, Boursiquot J, Lacombe T (2010) Genetic variation and biogeography of the disjunct Vitis subg. Vitis (Vitaceae). J Biogeogr 38: 471–486. doi: 10.1111/j.1365-2699.2010.02410.x
- 64. Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MA, et al. (2007) PLINK: a tool set for whole-genome association and population-based linkage analyses. The American Journal of Human Genetics 81: 559–575. doi: 10.1086/519795
- 65. Weir BS, Cockerham CC (1984) Estimating F-statistics for the analysis of population structure. Evolution 38: 1358–1370. doi: 10.2307/2408641
- 66. Paradis E, Claude J, Strimmer K (2004) APE: Analysis of Phylogenetics and Evolution in R language. Bioinformatics 20: 289–290. doi: 10.1093/bioinformatics/btg412
- 67. Penny D, Hendy MD (1985) The use of tree comparision metrics. Syst Zool 34: 75–82. doi: 10.2307/2413347
- 68. Moore M (1991) Classification and systematics of eastern North American Vitis L. (Vitaceae) North of Mexico. SIDA Contrib Bot 14: 339–367.
- 69. Galbraith D, Edwards J (2010) Applications of Microarrays for Crop Improvement: Here, There, and Everywhere. Bioscience 60: 337–348. doi: 10.1525/bio.2010.60.5.4
- 70. Chagné D, Crowhurst R, Troggio M, Davey M, Gilmore B, et al. (2012) Genome-Wide SNP Detection, Validation, and Development of an 8K SNP Array for Apple. PLoS ONE 7: e31745. doi: 10.1371/journal.pone.0031745
- 71. Hyten DL, Song Q, Fickus EW, Quigley CV, Lim J, et al. (2010) High-throughput SNP discovery and assay development in common bean. 1–8. doi: 10.1186/1471-2164-11-475
- 72. Ganal M, Durstewitz G, Polley A, Bérard A, Buckler E, et al. (2011) A Large Maize (Zea mays L.) SNP Genotyping Array: Development and Germplasm Genotyping, and Genetic Mapping to Compare with the B73 Reference Genome. PLoS ONE 6: e28334. doi: 10.1371/journal.pone.0028334
- 73. McCouch S, Zhao K, Wright M, Tung C, Ebana K, et al. (2010) Development of genome-wide SNP assays for rice. Breed Sci 60: 524–535. doi: 10.1270/jsbbs.60.524
- 74. Thomson M, Zhao K, Wright M, McNally K, Rey J, et al. (2012) High-throughput single nucleotide polymorphism genotyping for breeding applications in rice using the BeadXpress platform. Mol Breed 29: 875–886. doi: 10.1007/s11032-011-9663-x
- 75. Petit RJ, Hampe A (2006) Some evolutionary consequences of being a tree. Annual Review of Ecology, Evolution, and Systematics 187–214. doi: 10.1146/annurev.ecolsys.37.091305.110215
- 76. Clark AHM, Bustamante CD, Williamson SH, Nielsen R (2005) Ascertainment bias in studies of human genome-wide polymorphism. Genome Res 15: 1496–1502. doi: 10.1101/gr.4107905
- 77. Wen J (1999) Evolution of eastern Asian and eastern North American disjunct distributions in flowering plants. Annu Rev Ecol Syst 30: 421–455. doi: 10.1146/annurev.ecolsys.30.1.421
- 78. Xiang QJ, Zhang WH, Ricklefs RE, Qian H, Chen ZD, et al. (2004) Regional differences in rates of plant speciation and molecular evolution: a comparison between. Eastern Asia and Eastern North America Evolution 58: 2175–2184. doi: 10.1111/j.0014-3820.2004.tb01596.x
- 79. Aradhya M, Dangl G, Prins B, Boursiquot J, Walker M, et al. (2003) Genetic structure and differentiation in cultivated grape, Vitis vinifera L. Genet Res 81: S0016672303006177. doi: 10.1017/s0016672303006177
- 80. Galet P (1979) A Practical Ampelography. Translated and adapted by Lucie T. Morton. Morton LT, translator. Ithaca, NY: Cornell University Press.
- 81. Mullins MG, Bouquet A, Williams LE (1992) Biology of the Grapevine. Cambridge, United Kingdom: Cambridge University Press.