Powerful, efficient QTL mapping in Drosophila melanogaster using bulked phenotyping and pooled sequencing

Abstract

Despite the value of Recombinant Inbred Lines (RILs) for the dissection of complex traits, large panels can be difficult to maintain, distribute, and phenotype. An attractive alternative to RILs for many traits leverages selecting phenotypically extreme individuals from a segregating population, and subjecting pools of selected and control individuals to sequencing. Under a bulked or extreme segregant analysis paradigm, genomic regions contributing to trait variation are revealed as frequency differences between pools. Here we describe such an extreme quantitative trait locus, or X-QTL, mapping strategy that builds on an existing multiparental population, the DSPR (Drosophila Synthetic Population Resource), and involves phenotyping and genotyping a population derived by mixing hundreds of DSPR RILs. Simulations demonstrate that challenging, yet experimentally tractable X-QTL designs ( > =4 replicates, > =5000 individuals/replicate, and selecting the 5-10% most extreme animals) yield at least the same power as traditional RIL-based QTL mapping and can localize variants with sub-centimorgan resolution. We empirically demonstrate the effectiveness of the approach using a 4-fold replicated X-QTL experiment that identifies 7 QTL for caffeine resistance. Two mapped X-QTL factors replicate loci previously identified in RILs, 6/7 are associated with excellent candidate genes, and RNAi knock-downs support the involvement of 4 genes in the genetic control of trait variation. For many traits of interest to drosophilists, a bulked phenotyping/genotyping X-QTL design has considerable advantages.

Discovery of malathion resistance QTL inDrosophila melanogasterusing a bulked phenotyping approach

“…In replicate 1, 65 of 1,303 males (5.0%) and 43 of 1,338 females (3.2%) survived, while in replicate 2, 403 of 2,881 males (14.0%) and 693 of 2,817 females (24.6%) survived. Given simulation data presented in Macdonald et al (2022), the scale of the experiment was lower than optimal in replicate 1, while the intensity of the selection was sub-optimal for replicate 2.…”

“…The X-QTL base population used here is a derivation of the one created by Macdonald et al (2022), and consists of the outbred descendants of a mix of RILs from the DSPR collection (King et al 2012a; King et al 2012b).…”

“…This resulted in 50X coverage for the replicate 1 female pools, and 32-39X coverage for the remaining pools. Previous work indicates coverage at this level enables robust haplotype frequency estimation (Macdonald et al 2022). Next, the bcftools mpileup, call and query commands (Li 2011) were employed to generate a file of REF and ALT counts at all SNPs for each sample (founders plus X-QTL samples), and this was converted to REF allele frequencies per sample per SNP (often 0/1 for the inbred founders).…”

“…We call haplotypes for each X-QTL sample in windows of 1.5-cM, stepping through the genome in 0.05-cM increments, using a procedure described in more detail previously (Linder et al 2020; Macdonald et al 2022). Briefly, for each X-QTL pooled sample, and within each window, we use the R/limSolve package (Soetaert et al 2009) to find the set of 8 proportions (summing to 1) that minimizes the sum of the weighted squared differences between the known founder haplotypes, and the observed frequency of each SNP in the window in that sample.…”

“…We employ bulked phenotyping and pooled sequencing of malathion-selected and control individuals derived from a population constructed by mixing several hundred DSPR RILs. Our use of this design in a previous study of caffeine resistance recapitulated QTL isolated via traditional RIL-by-RIL phenotyping (Macdonald et al 2022; Najarro et al 2015), while being – in our opinion – more efficient to execute. Since variation in malathion resistance has been examined previously using the DGRP – Battlay et al (2018) found very strong associations at Ace and a more modest association at Cyp6g1 – we could execute a direct comparison of the outcomes of two different approaches to genetic analysis in flies.…”