Single nucleotide polymorphims (SNPs) have become a genomic commodity as they are becoming indispensable in various genome scans aimed at mapping genomes [1–6], finding associations with complex traits [7–10], and population genomics [11, 12]. They are distributed along the various regions of the genomes [13, 14] and are frequent in coding regions of angiosperms [15, 16] and conifers [17–19]. However, the efficiency of genome scans is not only dependant on a wide genomic distribution of SNPs. Indeed, it also relies on the ability to genotype large numbers of SNPs over large sets of individuals.
SNP genome scans in non model species usually involve two steps: the discovery of SNPs and genotyping. With no a priori knowledge of DNA polymorphisms, SNPs are usually discovered through various strategies of individual or pool DNA sequencing , or by using tilling techniques, a high-throughput strategy relying on the enzymatic cleavage of mismatches . For a number of crop species, current resequencing efforts have led to the development of SNP databases and generate a wealth of SNPs usable in genome scans. In conifers, large-scale EST sequencing projects have been initiated [22–25], providing a starting point to develop SNP resources in pine  and spruce .
Several SNP genotyping array approaches have been developed with variable success. The accuracy of innovative SNP genotyping technologies has been assessed mostly through the development of assays suitable for analysing variations in the human genome. Broadly speaking, four reaction principles govern SNP genotyping assays: hybridization with allele-specific oligonucleotide probes, oligonucleotide ligation, single nucleotide primer extension, and enzymatic cleavage reviewed in [26–28]. Among these approaches, the GoldenGate assay developed by Illumina and relying on the bead array technology has demonstrated high performance with high levels of call rate, reproducibility, and overall success rate for the analysis of the human genome [29–31].
High-throughput SNP assays have recently been applied to plants. Large datasets of SNP-based markers are being developed in barley through the development of genotyping assays relying on Illumina's technologies , leading to the undertaking of an international SNP project . The same genotyping approach has made it possible to map large datasets of SNPs even in complex and duplicated genomes such as soybean , and projects are underway in hexaploid wheat  and poplar .
In the present study, we are asking whether high-throughput SNP genotyping technologies developed for human population genomics applications, such as the Illumina GoldenGate SNP assay, are applicable to large and essentially unsequenced genomes as seen in conifers. Conifer genomes reach very large sizes, around 10,000–40,000 Mb , consisting mostly of repetitive sequences . For the two conifers considered herein, white spruce and black spruce, genome sizes are well in excess of 10e10 bp .
Moreover, the partial knowledge of the large and redundant genomes of conifers can be a limiting factor to design an efficient SNP genotyping assay. Indeed, sequences located upstream and downstream the SNP cannot be fully validated for locus specificity and the possible presence of repetitive elements [29, 30]. The possible effect of such a drawback remains to be verified for most crop and tree species which genomes are essentially not sequenced. Based on EST sequence data available for white spruce , we have designed primers and resequenced genomic DNA for hundreds of genes in white spruce and black spruce. The high quality SNP datasets developed were used to select SNPs amenable to the GoldenGate genotyping assay and test the technology for these two species. Then, we integrated these SNP data into linkage maps of expressed genes and illustrated the possibility to rapidly improve the density of existing genetic maps for spruce species.