In this study, we conducted a genome-wide deficiency screen for QTL for heat resistance. The correlation between the number of deleted genes and the degree of heat sensitivity was not significant, suggesting that the deletion of a number of genes in general did not influence heat sensitivity. As a result, we found 19 QTL for heat resistance on the 2nd and 3rd chromosomes. The result of the present study corresponded well to QTL mapping study conducted by Norry et al.  because 16 QTL found in this study were encompassed by QTL for KRHT found by Norry et al. . Norry et al.  localized 5 QTL for KRHT, and 3 of them encompassed 4 of the QTL for heat resistance found in this study. Similar distribution of QTL for heat resistance, i.e., resistance to mild heat stress during the pre-adult period, and those for KRHT suggest that the underlying mechanisms for those traits may partially overlap with each other, although they contribute to different aspects of heat resistance. In addition, colocalization of 4 QTL found in this study with 2 QTL found by Morgan & Mackay  suggests that resistance to mild heat stress during the pre-adult period and short-term heat shock at the adult stage share genetic architecture. Three of the heat-responsive genes found by Leemans et al.  were located in QTL found in this study, while the 16 non-overlapping heat-responsive genes found by Sorensen et al.  were encompassed by 11 QTL found in this study. To confirm whether QTL found in the previous studies and the ones found in the current study consist of the same genes, complementation test using both the deficiency strains and the strains from the previous QTL mapping studies would be necessary in future studies. Part of the reason for the difference could be the method of heat stress application. Leemans et al.  applied a short heat shock at the embryonic stage, while we provided mild heat stress (28°C) during the entire pre-adult stages. In addition, the previous QTL mapping studies utilized natural genetic variation to map candidate QTL while we used deficiency screening targeting arbitrary genome regions by design. When natural genetic variation is utilized to map QTL, the effect of the loci would not be detected if there is no genetic variation in the focal loci between the strains used for the mapping. Because QTL mapping is usually done with two representative strains, variable regions on the genome is limited. On the other hand, when well-designed deficiency collections such as DrosDel isogenic deficiency collections were used to map QTL, genetic variation between the deficiency and the control strains can be located even where no natural genetic variation exists in wild populations. The difference in the sources of genetic variation for the mappings could cause the difference of the QTL found in the previous and current studies.
Among the heat-responsive genes from Leemans et al.  and Sorensen et al. , 4 Hsp genes, Hsp70Ab, Hsp70Bb, Hsp70Bbb, and Hsp70Bc, were located in the QTL found in this study. In addition, 2 Hsp genes, Hsp70Aa and Hsp70Ba, were included in the QTL of which locations corresponded to the deficiency region of Df(3R)ED5577. These 6 copies of Hsp70 are nearly identical in sequence and are closely linked together on 87A and 87C of the right arm of the 3rd chromosome . Rapid induction of Hsp70 expression upon heat shock and the suppression of its expression are strictly regulated in a cell [24, 25]. The effect of Hsp70s on heat resistance has also been confirmed at the individual level based on the increased heat resistance in organisms with a high copy number of Hsp70 genes  and on the reduced heat resistance in Hsp70-null flies . We detected the effect of QTL, including those genes in our screening, and the findings support that Hsp70 genes are strong candidates for heat resistance in D. melanogaster. It also indicates that our screening has sufficient detection power to confirm the effect of Hsp70 genes and will be able to detect QTL with equivalent contribution to heat resistance as Hsp70 genes.
Six of the non-Hsp heat-responsive genes from Sorensen et al.  that were located inside QTL found in this study, CG17124, CG17108, CG3270, GstD2, GstD5, CG3301, have been reported to be involved in various stress responses, such as starvation stress, aging, and oxidative stress, and pesticide resistance [28–32]. There were 2 genes that were suggested to be involved in immune response: CG16749 and GstD5 [33, 34]. A general responsiveness to environmental stress of these genes suggests that they are strong candidates for heat resistance. Testing the individual effect of those genes on heat resistance will be necessary to understand how Hsp and non-Hsp genes act together to contribute to heat resistance. In addition, this result may suggest that the resistance for mild heat stress (28°C) during the entire pre-adult stage consists of multiple stress-resistance processes that are required for comprehensive homeostasis of development. Compared to a short-term heat shock at embryonic or adult stages, examination of the response to a long-term heat stress during the pre-adult stage may reveal ecologically important mechanisms for heat resistance in natural conditions.
In our deficiency screen, 3 QTL were located outside the QTL found in the previous QTL mapping studies [9–11]. As described above, the difference in heat stress application between the present study and others may result in this difference. The novel QTL may contain genes that specifically contribute to long-term mild heat stress response during the pre-adult period. One QTL corresponded to the deficiency region of Df(2R)ED4071 contained a heat-responsive gene, Ance-5, but the other two corresponded to the deficiency regions of Df(2R)ED2457 and Df(3R)ED6310 did not include any known heat-responsive genes. The 2 QTL contained 24 and 36 genes. A detailed examination of those genes in future research may reveal novel candidate genes for heat resistance.
The X-linked effect observed by Norry et al.  and Norry et al.  was not detected in this study. Part of the reason may be that most deficiencies on the X chromosome result in lethality in the male (see Additional file 1, Table S1). Thus, we evaluated heat sensitivity for the X chromosome in many cases for females only in this study, while Norry et al.  and Norry et al.  only used male flies for their QTL mapping. In addition, coverage of deficiencies was lowest on the X chromosome (54.1%) in this study, and it may lower the detection power on this chromosome. An effort to map with a higher coverage using both sexes is necessary for a more detailed analysis of the X chromosome.
In the current study, we searched for genomic deletions that increased the heat sensitivity to locate genomic regions with effect on heat tolerance, but we also found genomic deletions that decreased heat sensitivity. Such unexpected effect of the deletions could be due to the enhanced heat resistance at larval stage in exchange for fitness at adult stage. In this study, unfortunately, we did not measure their fitness at adult stage, and whether such trade-off exists or not is still unknown. Decreased heat sensitivity of the deficiency heterozygotes indicates that those QTL function to suppress heat tolerance in DSK001 homozygotes. No genes or no QTL have been reported to suppress heat resistance so far. Further study is necessary to understand the mechanism of the regulation of heat sensitivity thoroughly.
QTL mapping has been the most popular method to map candidate genomic regions for quantitative traits. Due to the recent development of the isogenic deficiency libraries such as DrosDel and Exelixis collections [14, 35], genome-wide deficiency screen became possible in D. melanogaster. In the current study, most of the deletions caused homozygous lethality, and it limits the experimental design to compare +/+ to Df/+. The large deletions suitable for an efficient screen tend to encompass recessive lethal alleles or alleles that cause recessive lethality when deleted together. In this experimental setting, QTL with dominant or recessive effect would not be detectable because +/+ and Df/+ are expected to show the same phenotype. In addition, when deficiencies with significant effect overlap with deficiencies without effect, it is possible to subtract the overlapping regions to reduce candidate regions down to smaller sizes. In the current study, however, there were cases where a deletion with significant effect was completely encompassed by a larger deletion without significant effect. Large deletions tend to encompass genes essential for survival, and it makes the subtraction approach lead to a false refinement of the candidate regions. To present the result in a conservative way, we avoided the subtraction approach in the current study. Even with this limitation, however, we detected 19 QTL for heat resistance, indicating that it is an effective approach when targeting QTL with additive effect. As for the resolution of the mapping, previously located QTL for heat resistance usually encompassed from several hundreds to thousands genes [9–11], while the QTL found in this study encompassed about 58 genes on average. The more than 10 times higher resolution of the present deficiency screen compared to the previous QTL mapping makes it possible to analyze the effect of individual genes in each QTL using mutation or RNAi approach in future studies. Although intensive effort is necessary to achieve high genome coverage, the current isogenic deficiency screen is a powerful approach to investigate genetic architecture of quantitative traits in D. melanogaster.
In the present study, we performed a novel screening for QTL for heat resistance and found 16 that overlapped with the previously known QTL. Combined with the results of gene expression profiling studies, we specified several putative candidate genes in the QTL. We also discovered 3 novel QTL for heat resistance. The high resolution mapping in this study compared to that of the previous QTL mapping studies has the advantage of a genome-wide screening using a newly available genetic tool, which is a collection of isogenic deficiency strains. Further deficiency screen with smaller deficiencies within the genomic regions with significant effect found in the current study would narrow down the candidate genes from the average 58 genes per candidate regions to smaller numbers. This makes the examination of the individual candidate genes using mutation analysis or RNAi knockdown more feasible in future studies. In addition to that, genome-wide deficiency screen of QTL for other aspects of heat resistance combined with more detailed gene expression profiling studies may provide a better understanding of the underlying mechanisms of heat resistance.