SNP markers have been gaining popularity for genetic studies since the last decade due to their high frequency, ability to high throughput automated analysis, and the decrease of their costs (Rafalski 2002; Davey et al. 2011). Even though, microsatellite markers remain the markers of choice for many genetic studies due to their attractive features (high variability, codominance, transferability, and reproducibility). This is particularly true for studies that require the use of relatively small number of markers or samples. However, the use of large sets of SSR markers is limited due to the high cost and the amount of work required for their analysis, as well as the difficulties of automation. Those limitations could be partially overcome by adopting multiplexing strategies (Butler 2005a, b; Missiaggia and Grattapaglia 2006; Blacket et al. 2012) that would led to significant savings in time, efforts, and laboratory reagents (Elnifro et al. 2000; Raabová et al. 2010). Empirical results demonstrated that SSR-based multiplexing and multiloading reduced the cost of PCR reagents up to 50 % and the cost of electrophoresis up to 85 % when compared to genotyping based on single-locus PCR (Masi et al. 2003; Merdinoglu et al. 2005). Guichoux et al. (2011) estimated that, even for a moderate number of samples, a 12-plex multiplexing could be eight times cheaper than simplex PCR. Additionally, genotyping errors due to human factor are less likely to occur, as laboratory manipulations are considerably reduced. Two approaches could be used for multiplexing: (1) the amplification of only one or a few markers followed by pooling the products from the individual PCRs prior to electrophoresis (Hall et al. 1996); (2) the joint amplification of all the loci to be loaded in the same electrophoresis run. In this study, we have used the second approach to address the development of an efficient tool that would allow grapevine genome-wide scanning using as fewer as possible PCR runs and electrophoresis tracks. Multiplex PCRs were developed following two main steps: (1) markers selection and organization in multiplex sets; and (2) the optimization of PCR conditions for the proposed multiplexes. In addition, one of the important issues in this work was the establishment of up to seven control points to avoid possible mistakes that could be very difficult to detect in large-scale genotyping projects (see “Data checking” on “Materials and methods”). The approach adopted in this study should be applicable for the development of similar tools in any other species.
SSR markers were selected based on three criteria: genome-wide coverage, high polymorphism and diversity in the fragment size ranges. In spite of the numerous efforts devoted to the development of SSR markers for grapevine genotyping since the early nineties, the number of available SSRs remains limited when compared to other genome sequenced crop species. For instance, whereas 4,109, 8,192, and even 70,732 entries for SSR probes were found at the NCBI for soybean, wheat, and rice, respectively, only 1,079 entries were found for grapevine (http://www.ncbi.nlm.nih.gov/probe). In addition, for most of them (62.5 %) fragment size ranges are concentrated between 150 and 250 bp (Fig. 4c). This limits marker combination possibilities, compromising hence the development of multiplex PCRs with high multiplexing levels (Hill et al. 2009). Moreover, the available information is usually scarce or incomplete, since many of these markers have been developed and/or tested using only a few individuals. For example, in our PCR conditions, five UDV (Di Gaspero et al. 2005), four Vitis Microsatellite Consortium (VMC), and one FAM (Huang et al. 2011) SSRs amplified in several accessions more than two fragments in a narrow size interval, resulting in unreliable electrophoretic profiles. This could be related to the presence of multiple binding sites for their respective primers in the grapevine genome. For instance, the design of alternative primers allowed recovering monolocus segregation for four of them. In the same way, three loci described as monomorphic in V. vinifera were selected because they were studied in only four (VVIB54; Merdinoglu et al. 2005) or even two accessions of this species (FAM18 and FAM104; Huang et al. 2011), suggesting that more alleles could be detected in a larger sample. Our results revealed that it is unlikely to detect more than one allele at these loci in V. vinifera. Nevertheless, those markers, as well as other 21 that detected only two or three alleles (Online Resource 1: Table S2), could be more informative in genetic backgrounds from other Vitis species. For example, the above-mentioned monomorphic markers detected more than one allele when other Vitis species were analyzed (Merdinoglu et al. 2005; Huang et al. 2011). On the other side, we found chromosomal regions spanning up to 10 Mbp for which SSR markers had never been described (Fig. 2). Attempting to develop new markers, we identified many microsatellite motifs (up to 153/Mbp) in the investigated regions, but successful primers could be designed for only a few of them. Moreover, PCR failures and/or irregularities in allele inheritance related to the presence of null alleles have been found for nearly half of the 22 new markers. Difficulties with SSR markers development have been already reported in plants (Tero et al. 2006) as well as animal species (Zhang 2004; McInerney et al. 2011), and linked to the presence of repetitive DNA and/or transposable elements. This might be also the case in our study, since most of the SSR gaps were located in low gene density regions of the grapevine genome (Online Resource 2), which had been found substantially complementary to high density of repetitive/transposable elements (Jaillon et al. 2007).
The second critical step for the development of the multiplex PCRs consisted in the establishment of suitable thermocycling programs and the adjustment of primers concentrations to obtain a balanced amplification for all the markers included in the same reaction. Touchdown PCR (Don et al. 1991) proved to be highly suitable for multiplex amplification. In fact, only two touchdown-based programs (Tp-A and Tp-B) were enough to run 90 % of the designed multiplex PCRs, which included primers with a wide range of annealing temperatures. As an example, Mx05 includes primers with annealing temperatures ranging from 48 to 67 °C. Touchdown-based thermocycling programs combined with the use of commercial kits optimized for multiplex PCR contribute to save time and labor in the optimization steps when compared to conventional protocols (Masi et al. 2003). In this study, they allowed the successful amplification of up to 15 primer-pairs in the same PCR reaction. As far as we know, this is the highest multiplexing level reached for microsatellite-based grapevine genotyping. Merdinoglu et al. (2005) considered the preferential amplification of small fragments over long fragments for multiplex design, which limited the multiplexing level that could be reached. Our results demonstrated that it is possible to amplify fragments with size differences up to 329 bp in the same reaction; as an example, Mx19 includes 10 loci with fragment sizes ranging from 76 to 401 bp. Nevertheless, an average of 3.8 (between two and eight) optimization assays were needed to achieve balanced amplifications.
The 45 developed multiplex PCRs were tested and optimized using seven grapevine accessions. In order to verify their validity for genotyping large diversity panels, we used them to genotype a set of 207 accessions representing practically most of the cultivated grapevine genetic diversity. Indeed, this collection was selected from the 1,852 V. vinifera accessions maintained at the germplasm bank of El Encín (Alcalá de Henares, Madrid, Spain), using the genotypic dataset of the 26 control loci through the genetic diversity maximization strategy (Le Cunff et al. 2008). Moreover, the main progenitors of grapevine cultivars (Lacombe et al. 2013) are present in the validation collection. The representativeness of the this collection was also verified by comparison with two large grapevine germplasm collections: (1) an Italian collection of 745 accessions (Cipriani et al. 2010)—a set of 22 “Vchr” markers identified an average of 9.8 (3 to 21) alleles/locus in that collection, almost the same as in the validation collection (between 3 and 20 with an average of 9.7). (2) A collection of 2,323 V. vinifera subsp. sativa cultivars conserved at INRA grape repository at Vassal (France) (Laucou et al. 2011)—twenty out of the 26 control loci detected a lower number of alleles (between 5 and 19, with an average of 11.95) in the validation collection than in the Vassal collection, where an average of 16.9 (6 to 36) alleles/locus were identified. However, these differences are mainly related to the presence of uncommon genetic material carrying rare alleles in the larger French collection (2,323 vs 207). For instance, 9 and 12 alleles not detected in our study by the markers VMC4F3 and VVIV67, respectively, had been detected in Laucou et al. (2011), but with frequencies lower than 0.005. These findings point out that, although additional alleles may be detected when studying more distant genetic material, the allelic ranges obtained using the validation collection should represent a good estimation of the actual diversity that can be found in cultivated grapevine. Online Resource 1: Table S5 shows the complete genotypes obtained for 25 reference cultivars at the 264 studied loci that can be used for inter-laboratory comparisons and protocols setting-up.
The presence of null alleles at microsatellite loci can introduce important biases into genetic studies (Callen et al. 1993; Pompanon et al. 2005). Their existence in grapevine have been already demonstrated by sequence analysis (Sefc et al. 1999). Another reliable approach for null allele detection is the analysis of allelic inheritance in family groups (Dakin and Avise 2004). Using this approach, we identified the probable existence of null alleles in 49 out of the 284 analyzed primer pairs. Most of them (69.39 %) showed estimated null allele frequencies (F) >0.10. However, the absence of null alleles in the 14 studied trio pedigrees does not discard their presence in the rest of the genotypes. In fact, 22 % of the loci with a moderate to high frequency of null alleles (0.10 < F < 0.34) did not show any incompatible genotype among the studied pedigrees. On the other hand, incompatible genotypes were detected for 67.5 % of primer pairs that, after PCR repetition, failed to amplify at least one sample. These results point out that most of the finally recorded PCR failures could be related to the presence of null alleles at homozygote state. Out of the 32 markers showing null alleles for which alternative primers were designed, 18 recovered the normal segregation of alleles, decreasing the estimated null allele frequency in most of the cases (from 0.170 to 0.002 on average), and no amplification failures were noticed for any redesigned primer pair.
Microsatellite and SNP markers have become in the last years the markers of choice for genetic analyses in grapevine, including genetic mapping and QTL identification (Huang et al. 2012; Doligez et al. 2013; Battilana et al. 2013; Barba et al. 2014), linkage disequilibrium and association analyses (Emanuelli et al. 2010; Barnaud et al. 2010; Cardoso et al. 2012; Vargas et al. 2013), and varietal identification (Myles et al. 2010; Cabezas et al. 2011; Laucou et al. 2011; Migliaro et al. 2012). As in many other woody plant species, genetic mapping in grapevine has been carried out using mainly the double pseudo test-cross strategy (Grattapaglia et al. 1995), which is based on the study of allelic segregation of markers found in heterozygosis in one or both progenitors of an F1 progeny. The development of new genotyping platforms and technologies has increased the number of SNPs that can be genotyped to hundreds of thousands allowing the construction of highly saturated genetic maps. However, with the currently available information, these markers are not suitable for a direct comparison with the previously published genetic maps and QTLs in grapevine, which are predominantly based on microsatellite markers. The number of markers that can be genotyped using the panel of multiplex PCRs developed in this study is large enough to allow that comparisons. For example, the analysis of the segregation types inferred from the genotypes showed on Online Resource 1: Table S5 points out that in a supposed project involving a mapping progeny derived from a cross between Cabernet Sauvignon and Pinot Blanc would allow to map 235 out of the 264 studied loci, with an average of 12.37 per chromosome (from 7 in chromosome 10 to 17 in chromosome 5). On the other side, linkage disequilibrium in V. vinifera expands up to 16 cM when studied using microsatellite markers (Barnaud et al. 2006, 2010). This suggests that a low-density whole-genome genotyping, using the 45 developed multiplex PCRs as genotyping tool, should be useful for QTL detection through association mapping studies.
American and Asian Vitis species constitute a valuable source of resistance genes that can be introduced into V. vinifera varieties through interspecific breeding programs (Töpfer et al. 2011). The limiting step in this kind of programs is the recovery of the genetic background of the recurrent parent (vinifera) by backcrossing. The assistance of the selection process by genome-wide distributed SSRs organized in ready to use multiplex PCRs, such as the developed in this study, has the potential to make considerable savings in time and costs. The efficiency of such a tool could be improved by organizing SSRs from each chromosome in a distinct multiplex PCR as suggested in a previous simulation study (Herzog et al. 2013).
Multiplexes Mx01 and Mx02 allow the study of 26 unlinked markers, including at least one in each of the 19 V. vinifera chromosomes. They were designed re-organizing the multiplex PCRs S, A, and B used by Ibáñez et al. (2009) to genotype 376 table grape accessions. These 26 microsatellites have been used to genotype the complete germplasm bank of El Encín (Alcalá de Henares, Madrid, Spain). Twenty of them have been also used through eight multiplex PCRs and three sequencing runs to characterize most of the INRA (France) grape repository (Laucou et al. 2011). Additionally, smaller subsets of the control loci used in this study (including the six OIV and the nine GrapeGen06 SSRs) have been used to characterize grapevine genetic resources from most of the viticultural regions in the world (This et al. 2011). Given the large amount of information available for these markers, Mx01 and Mx02 can be used for a rapid and cost efficient characterization of unexplored grapevine germplasm and short range genetic studies, such as parentage analysis. In this work, we have used Mx01 and Mx02 to certify the identity of each of the extracted DNAs. Moreover, the inclusion of one of the loci amplified with Mx01 and Mx02 as control markers in each multiplex PCR allowed to verify assay reproducibility and sample traceability throughout the study, factors which should be considered by any researcher aware of the consequences of genotyping errors (Pompanon et al. 2005).
Although important cost and time savings might be obtained using the panel of multiplex PCRs presented in this study, a considerable investment in primers labeling will be still required. Economic labeling methods could be used to further decrease the costs. As an example, Blacket et al. (2012) proposed the use of four fluorescently labelled universal primers (one primer for each fluorescent dye) through the three primer PCR approach. This kind of techniques would offer an inexpensive alternative to the commercial synthesizing of custom labeled primers for multiplex genotyping. However, it should be noted that additional optimization steps may be needed because sometimes the three primers approach seem to decrease PCR efficiency (de Arruda et al. 2010).
