Genetic map construction and QTL analysis

A genetic linkage map was constructed using MapDisto 2.0 (Ref³¹). The command ‘Find linkage groups’ was used to construct linkage groups. The minimum log-of-odds (LOD) threshold of 3 and maximum recombination fraction (r) of 0.35 were used to search for linkage groups. The commands ‘Compare all orders’ and ‘Order a linkage group’ were used to compute best order of sequences respectively for linkage groups with short sequences (< 10 loci) and long sequences (> 10 loci). The ‘Sum of adjacent recombination frequencies (SARF)’ option was chosen for ordering sequences. The commands ‘Ripple order’ and ‘Check inversions’ were used to refine the order of sequences generated by ‘Order a linkage group’ command. Double recombinants were removed using ‘Show double recombinants’ command followed by ‘Replace error candidates with missing data’ command. Finally, ‘Bootstrap order’ command with 1000 permutations was used to evaluate the stability or robustness of a given order.

Because many SNPs were very closely linked and did not provide additional information about the QTL locations, we selected only 251 SNPs that were evenly distributed across nine linkage groups for final linkage map construction and QTL mapping. The final linkage map contained 251 SNPs and seven SSRs. Linkage groups were aligned with the pseudo-molecules in the lettuce genome (Genome ID 28333) by BLASTing sequences containing SSRs and SNPs against the lettuce whole genome sequence³². The sequences for SSRs were retrieved from NCBI database³³. The data were used to perform a QTL analysis using QTL Cartographer v2.5 (Ref³⁴). Composite interval mapping (CIM) was performed to build an initial model for multiple interval mapping (MIM) procedure. The CIM analysis was implemented using the following criteria: standard model, ‘forward and backward method’ for automatic cofactor selection, a 10-cM window size, automatic selection of 5 control markers, walking speed of 1 cM, and threshold LOD score estimated empirically with 1000 permutations. The MIM analysis was conducted using QTL peaks with empirical LOD threshold value of 2.0 and minimum 5-cM interval between QTL as the initial model. Because the QTL analysis is an iterative process, QTL were searched and refined in a cyclic stepwise fashion using ‘Searching for new QTL’, ‘Testing for existing QTL’, and ‘Optimizing QTL positions’ commands. QTL detected at large marker intervals were deleted as they could be ‘ghost QTL’. New QTL models were only accepted into the current model if they reduced the Bayesian Information Criterion (BIC) values. The epistatic interactions between each pair of QTL were tested using the option ‘QTL interaction’. Non-significant epistatic interactions between QTL were deleted from the model. The model with the least BIC was selected to report the statistically significant putative QTL. QTL effects were estimated using the ‘summary’ option. Sequences for markers in the vicinity of detected QTL were BLASTed³⁵ to identify the putative functions of the QTL. The QTL was visualized by MapChart 2.32 (Ref³⁶).