Strategies for the design and analysis of introgression libraries and near-isogenic lines

Abstract

Introgression libraries (ILs) are used to make the genetic variation of unadapted germplasm or wild relatives of crops available for plant breeding. The libraries consist of near-isogenic lines (NILs) that carry small donor chromosome segments (DCS) from an exotic donor in the genetic background of an elite line. ILs can be used to make favorable genetic variation of exotic donor genotypes available in the genetic background of elite breeding material. In this way, ILs can help bridge the gap between elite and wild varieties, making further genetic resources available for breeding purposes. In many cases, the number of DCS exceeds the number of lines, and more specialized statistical methods are required in this case. The objective of this research was therefore to compare tests for QTL detection in ILs regarding detection power and false positive rate, as well as propose best practice for their use regarding known aspects of trait architecture and heritability. Additionally, statistical tests were adapted and tested for use in ILs that have more DCS than lines. To accomplish these objectives, ILs with different configurations of DCS (overlapping, nonoverlapping, and stepped aligned inbred recombinant straits (STAIRS) libraries) were simulated and experimental data from rapeseed and rye were used.Computer simulations with maize genome models were employed to investigate nonoverlapping, overlapping, and STAIRS libraries for traits with oligogenic inheritance. QTL detection power of the linear model (LM) and Dunnett (DT) tests were similar for nonoverlapping and STAIRS ILs; for overlapping ILs the DT was slightly superior. False positives were greatest for the t-test and lowest for the LM test. False positive sums with the DT were generally higher than for the linear model test if the heritability was 0.9 or lower. The linear model test outperformed the DT in nonoverlapping ILs and for overlapping ILs where trait heritabilities are below 0.9. Analysis of a rapeseed library additionally revealed that QTL localization ability using the LM test has a higher potential.A dataset of two rye introgression libraries was reanalyzed. After identifying ILs that had a significantly better per se or testcross performance than the recipient line with the DT test, the LM analysis was in most instances able to clearly identify the donor regions that were responsible for the superior performance. The precise localization of the favorable DCS allowed a detailed analysis of pleiotropic effects and the study of the consistency of effects for per se and testcross performance. These analyses also highlighted the potential power of the LMtest to localize QTL beyond what is possible using only the DT test.Extending the LM test to overparameterized ILs required adapted genome-wide prediction (GWP) methods in order to achieve DCS effects estimates when there are more DCS than lines. Computer simulations showed that GWP employing heteroscedastic marker variances had a greater power and a lower false positive rate compared with homoscedastic marker variances when the phenotypic difference between the donor and recipient lines was controlled by few genes. The simulations helped to interpret the analyses of glucosinolate and linolenic acid content in a rapeseed IL and plant height in a rye IL. These analyses support the superiority of genome-wide prediction approaches that use heteroscedastic marker variances. When coupled with permutation tests, GWP methods can be usefully applied to ILs.In a more detailed analysis of rye ILs, GWP was used to attempt QTL detection. For several traits, GWP enabled the detection of positive and negative donor effects in individual NILs. Two DCS effects were detected that significantly increased thousand-kernel weight. We found four beneficial DCS effects for protein, pentosan and starch content. Three DCS effects for protein content were observed for improving feeding purposes and one DCS effect for starch content to improve ethanol production. The effects were localized to small genomic regions.It can be concluded that the LM test has many advantages over the DT test, and can enhance the results of QTL analysis in ILs in many ways. The LM test has the potential to provide comparatively high levels of correct detection with low false positive rates in ILs with various configurations and allows a more accurate localization of the QTL effects compared to pairwise tests. However, it can only be used in cases where the number of lines exceeds the number of DCS effects to estimate. In cases where there are more DCS than lines, GWP methods are able to detect QTL when employed with permutation tests. The highest performance was achieved with GWP methods using heteroscedastic marker variances. These methods, as a whole, compose a toolbox for researchers to extract the most usefulness from ILs while providing guidelines to the expected outcomes depending on the structure of the library.