Intraspecific diversity and the phenotype

Genomic sequencing of diverse genotypes in several plant species demonstrated that in addition to SNPs and SSR polymorphisms, extensive intraspecific differences include large insertions/deletions frequently composed of highly repetitive sequences such as retrotransposons and DNA transposons (Wang and Dooner, Reference Wang and Dooner2006), and in some cases also genes (Beló et al., Reference Beló, Beatty, Hondred, Fengler, Li and Rafalski2009; Springer et al., Reference Springer, Ying, Fu, Ji, Yeh, Jia, Wu, Richmond, Kitzman, Rosenbaum, Iniguez, Barbazuk, Jeddeloh, Nettleton and Schnable2009). For example, the complement of disease resistance genes may differ between accessions (Chin et al., Reference Chin, Arroyo-Garcia, Ochoa, Kesseli, Lavelle and Michelmore2001; Yahiaoui et al., Reference Yahiaoui, Kaur and Keller2009). Sequences that do not code for proteins may nevertheless affect the phenotype, by supplying enhancers or promoters to nearby genes, or code for small RNAs, which affect expression of other genes by a variety of mechanisms (Chen, Reference Chen2009). Pseudogenes, which in maize are frequently generated by Helitron transposons, are sometimes transcribed in sense or antisense direction, also affecting gene expression phenotype (Yang and Bennetzen, Reference Yang and Bennetzen2009).

If these types of polymorphisms are in linkage disequilibrium with genetic markers used for germplasm characterization (predominantly SNPs and SSRs), then no additional information other than marker genotype is needed to reflect correctly the underlying genetic relationships of accessions. However, if linkage disequilibrium (LD) between markers for germplasm fingerprinting and genic or non-genic large indel polymorphisms breaks down rapidly, direct genotyping of these differences may be necessary by DNA sequencing or other methods such as array comparative genomic hybridization (Beló et al., Reference Beló, Beatty, Hondred, Fengler, Li and Rafalski2009; Springer et al., Reference Springer, Ying, Fu, Ji, Yeh, Jia, Wu, Richmond, Kitzman, Rosenbaum, Iniguez, Barbazuk, Jeddeloh, Nettleton and Schnable2009). This is likely to occur in the case of variants, which occurred recently on the background of pre-existing haplotype pattern.

An important issue not always appreciated in the germplasm analysis context is the prevalence of ascertainment bias, which occurs when polymorphic loci are identified (ascertained) in one collection of germplasm, but used to evaluate diversity in another set (Clark et al., Reference Clark, Hubisz, Bustamante, Williamson and Nielsen2005). For example, a collection of SNP loci identified in a set of cultivated lines will not correctly represent polymorphic loci present in unadapted accessions, leading to incorrect estimates of genetic distances in the latter set of germplasm. Many polymorphic loci in the non-adapted accessions will not be represented in the SNP collection developed from adapted germplasm, and, in turn, some alleles common in adapted material may be rare in non-elite accessions. As a result, genetic distances determined in the ascertainment population may be lengthened in comparison with those in the non-ascertained population (Fig. 1). It is difficult to identify a priori an appropriate collection of germplasm for ascertainment (marker discovery), given unbalanced representation of different types of germplasm in many collections. Perhaps, the most appropriate unbiased methodology for germplasm fingerprinting is genotyping by genomic sequencing.

Fig. 1

An example of ascertainment bias. SNP markers ascertained in elite maize inbred collection were used to fingerprint a set of maize and teosinte lines. Genetic distances between maize lines appear much lengthened with respect to those between teosinte accessions, which are foreshortened. Using unbiased genotyping method eliminates this disparity. Data courtesy of Stan Luck (Pioneer Hi-Bred).

The sequencing technology is rapidly approaching the stage where it will become a cost-effective tool for genotyping (Edwards and Batley, Reference Edwards and Batley2009; Varshney et al., Reference Varshney, Nayak, May and Jackson2009). A number of accessions will be simultaneously sequenced in each lane of the instrument, after appropriate encoding. Depending on the size of the genome, some form of reduced representation analysis (Yuan et al., Reference Yuan, SanMiguel and Bennetzen2003) will probably be necessary to focus the effort on non-repetitive fraction of the genome.

Perspective on epigenotyping of germplasm

It is well established that epigenetic variation encoded by DNA base modifications such as 5-methylcytidine affects phenotype in animals and plants (Peaston and Whitelaw, Reference Peaston and Whitelaw2006; Henderson and Jacobsen, Reference Henderson and Jacobsen2007; Chandler and Alleman, Reference Chandler and Alleman2008). Some of the epialleles in plants are remarkably stable and affect important plant characteristics (Cubas et al., Reference Cubas, Vincent and Coen1999). It is therefore reasonable to propose that a complete characterization of a germplasm accession or a breeding stock should involve not only the description of the genotype but also of the epigenotype. It has recently been demonstrated that recursive selection for a yield component in canola results in plants that are genetically identical but can be distinguished by DNA methylation differences and exhibit significant differences in yield (Hauben et al., Reference Hauben, Haesendonckx, Standaert, Van Der Kelen, Azmi, Akpo, Van Breusegem, Guisez, Bots, Lambert, Laga and De Block2009). The tools for comprehensive epigenotyping are available and involve chemical deamination of m⁵C to U followed by DNA sequencing, enabling single base resolution across the whole genome, albeit at considerable expense (Lister and Ecker, Reference Lister and Ecker2009; Lister et al., Reference Lister, Pelizzola, Dowen, Hawkins, Hon, Tonti-Filippini, Nery, Lee, Ye, Ngo, Edsall, Antosiewicz-Bourget, Stewart, Ruotti, Millar, Thomson, Ren and Ecker2009; Wang et al., Reference Wang, Elling, Li, Li, Peng, He, Sun, Qi, Liu and Deng2009). The high throughput sequencing technology, especially rapidly developing single molecule sequencing (Edwards and Batley, Reference Edwards and Batley2009), promises to enable comprehensive epigenotyping of germplasm collections in the coming years. Currently, several options exist for epigenotyping of a subset of the genome, for example by excluding repetitive fraction of the genome (Peterson et al., Reference Peterson, Wessler and Paterson2002) or capturing specific sequences of interest (Hodges et al., Reference Hodges, Smith, Kendall, Xuan, Ravi, Rooks, Zhang, Ye, Bhattacharjee, Brizuela, McCombie, Wigler, Hannon and Hicks2009).