Our previous QTL analysis of a large F2
population identified five putative QTLs affecting overall survival time to HALI 
. Two of these QTLs, Shali1
, were validated as major effectors in a separate large backcross analysis 
of these recombinant populations generated from the B and X1 progenitor strains suggested that Shali1
increased HALI resistance and MST when carried in homozygous X1 strain alleles, but Shali2
increased resistance when homozygous for B strain alleles. The opposite scenario was also predicted, with homozygous B strain alleles for Shali1
and homozygous X1 strain alleles for Shali2
leading to increased sensitivity and earlier mortality in continuous >95% O2
. Thus, the Shali1
effect on HALI survival time directly correlated with the demonstrated outcome of the progenitors to a continuous hyperoxic exposure. To the contrary, the Shali2
effect was masked within the progenitor strain background, as clearly exemplified by the fact that B strain mice carried the putative resistance alleles for Shali2
, yet were HALI sensitive.
These large-effect QTLs with opposing allelic effects on HALI survival presented us with an intriguing opportunity to generate single congenic lines of mice that would synchronize these sensitivity or resistance QTLs in the same background strain and allow us to directly test in vivo our earlier in silico predictions. Because we were unsure whether either or both genetic backgrounds were important to support the predicted interactions, we generated reciprocal congenic lines for Shali1 and for Shali2 to coordinate the same effect QTLs in both the B and X1 backgrounds. This decision turned out to be critically important, since congenic strains for Shali1 and Shali2 generated on the B strain background showed much less effect compared to congenic lines on the recipient X1 strain. Since epistasis and other background effects cannot adequately be predicted for complex traits, these findings underscore the importance of generating reciprocal consomic and/or congenic strains to improve your chances of capturing the largest QTL effect(s).
This report presents the successful construction and subsequent testing of reciprocal congenic lines for the two major QTLs affecting HALI survival time. Most notably, we highlighted the dramatic and unpredicted phenotypic extremes demonstrated by the X1.B-1A (high penetrance of the sensitivity trait) and X1.B-4BB (high penetrance of the resistance trait) congenic strains. These lines were constructed to combine—within the same inbred strain—the reciprocal QTL effects of the two opposite-acting alleles for Shali1
, which originally derived from different chromosomes of the inbred X1 and B progenitor strains. In particular, the sensitive B alleles for Shali1
were combined with the existing sensitive X1 alleles for Shali2
in the X1 strain background, yielding the hypersensitive X1.B-1A congenic line. Based on our previous estimations for the Shali1
, X1.B-1A mice were expected to have about a 15-hr increase in sensitivity from resistant X1 mice, predicting a MST of ~115–120 hrs. Surprising, although on the resistant X1 strain background, the substitution of Shali1
with B alleles made the X1.B-1A congenic line significantly more sensitive than the sensitive B strain, indicating that Shali1
contained one or more key loci for overall HALI survival time. These results also strongly suggested that synchronizing the sensitivity alleles for Shali1
allowed direct interactions (either additively or synergistically) to further increase HALI sensitivity, and may signify important members of a similar pathway or process. As other sensitivity genes likely exist on the X1 strain background as well, the X1.B-1A line combines all resident sensitivity loci in the X1 strain with the Shali1
sensitivity region transferred in from the B strain. In this case, the Shali1
locus must contain at least one major-effect gene, which has the capacity to fully reverse the HALI phenotype of the resistant X1 strain. Therefore, the X1.B-1A data clearly validated the predicted sensitivity effect from Shali1
-derived B strain alleles.
In the complementary strategy, resistant B alleles for Shali2 were introgressed onto the X1 strain background, which already contained the resistant X1 alleles for Shali1. Together, the resistant allelic combination led to a large increase in HALI MST. In general, with the exception of a single female mouse with a survival time of 301 hrs (), individual mice of X1.B-4BB did not live longer than the most resistant X1 strain mice (i.e., 10–15% lived >200 hrs). Rather, because more than 90% of X1.B-4BB mice survived to the resistance threshold (and most survived significantly longer), the large increase in penetrance of the resistance trait can directly explain the 66-hr increase in MST of X1.B-4BB mice compared to the X1 strain. Again, effects larger than those predicted by the QTL analysis suggested an additive or potentially synergistic action for these two loci and/or other resistant loci present in the X1 strain background. In total, the resistance alleles for Shali2 (B alleles) combined with alleles for Shali1 and other existing resistance X1 alleles in the X1 strain background, almost totally eliminated the reduced penetrance of the resistance trait seen in the inbred X1 strain. Therefore, although the Shali2 effect was masked in the inbred B and X1 strain genomes, data from the X1.B-4BB congenic line confirmed the predicted resistance effect of B-derived Shali2 alleles.
Given the sizes of the chromosomal regions remaining in these congenic lines, it is premature to identify and speculate on potential positional candidate genes. Most congenics in this report contain extremely large regions of transferred DNA. Even the smallest region, which is currently in the B.X1-1.303 congenic strain, is more than 10 Mbp and contains hundreds of genes. Since the ~10 Mbp region of B.X1-1.303 has not yet been captured on the more robust X1 strain background, we still consider this refined interval tentative. And, because it is not uncommon to see a single QTL dissect further into multiple, closely linked QTLs that often show opposing phenotypic effects 
, further resolution of these regions and additional complementary strategies are needed before a highly informative candidate gene list can be generated. Because of their high penetrance, RNA-Seq analysis of the X1.B-1A and X1.B-4BB congenic lines will be used to help identify putative positional candidate genes for further study.
Besides this report, major QTLs on different chromosomes with counteracting effects within the same background strain have been published for other complex traits, although these findings are very rarely reported for rodents 
. The difficulties to detect such gene-gene interactions are likely to be an inherent shortcoming of the QTL analysis mapping strategy, suggesting that this will remain a significant challenge in complex trait analysis. For example, when two or more major genes of a complex trait have opposing effects, their interaction can partially or fully hide the overall measured effect, thereby leading to an inability to detect the individual QTLs or to a reduced capacity to determine their separate contributions. As described previously for lung cancer susceptibility 
, such complicated gene-gene interactions are likely to be commonplace in complex traits and contribute greatly to the heritable portion of phenotypic variation 
. Interestingly, the MST difference between the sensitive B (105 hrs) and resistant X1 (133 hrs) inbred strains was only about 1.26-fold. Traditional recommendations of using two polar-responding inbred strains for QTL analysis argues against using our model for QTL analysis. Although adhering to this recommendation will improve the odds of tracking and identifying QTLs, our data demonstrate that a large differential trait response between progenitors is not an absolute requirement. Given that the complexity of a trait is usually not known, many competing factors such as multiple genes, opposing effects, sex effects, gene-gene interactions and incomplete penetrance, can distort the overall measure differently on different background strains. For this reason, our established reciprocal congenic mouse models, which have captured significant gene-gene interactions for HALI susceptibility, provide us with unique and potentially powerful capacity to identify the critical genes for differential HALI susceptibility and to further characterize the complex interactions of these, and potentially other relevant background strain loci.
Additional findings with these reciprocal lines suggest other potential opportunities for follow-on breeding studies, either by creating specific informative multi-QTL congenics or by using other available congenic resources. First, the X1.B-1B line, which carries the proximal half of Chr 1, including the male-specific QTL Shali5
, showed an increase in HALI resistance for males. As seen for Shali2
, the B alleles were associated with increased resistance, although only males in the case of Shali5
. These results suggest that a double congenic line containing the X1.B-4BB and X1.B-1B regions may show additional resistance for males by establishing a putative gene-gene interaction. Second, the X1.B-4A line carries the proximal region of Chr 4 derived from the B strain, an area not previously associated with a change in HALI survival time. However, the X1.B-4A mice demonstrated increased HALI sensitivity in females, suggesting a possible new sex-specific sensitivity QTL when homozygous for the B allele. Double congenics with the X1.B-1A and X1.B-4A lines could help validate this new QTL and their possible gene interactions for added sensitivity. Third, although the B.S congenics for both Shali1
often had little or no effects, select lines were statistically significant and, therefore, may provide potentially revealing information. For example, the substituted region that is associated with increased sensitivity in X1.B-1A mice is ~102 Mbp (per MUGA SNP analysis). However, when combined with data from the B.X1-1.303 congenic line, the Shali1
region can be tentatively refined to a 10.8-Mbp interval surrounding D1Mit303
, from 59 to 69.8 Mbp (). MUGA SNP analysis of B.X1-1C, which potentially carries an informative recombination in the Shali1
interval, could further narrow the critical region. Because the X1 background shows more effect, this refined region can now be targeted for recombinants in the X1.B-1A line to validate this prediction, narrow the region of interest and allow selection of positional candidate genes. Thus, although the B strain background appears to be less informative for determining the Shali
effects, their data in combination with the reciprocal lines can help narrow the important intervals. Fourth, because most knockout mice are often generated from 129-derived stem cells placed on the B strain background, our mouse model may allow us to take advantage of the knockout lines available for the genes that map to these QTL intervals to refine Shali1
, a strategy first proposed by Bolivar et al.
. This strategy would be most relevant for the B.X1-1.303 congenic line, which has demonstrated a significant effect on the B strain background.
In summary, these studies present the first description of congenic lines that validated two major QTLs on different chromosomes with opposite allelic effects for a complex trait in the same background inbred strain. These congenic lines further demonstrated large changes in the respective control strain phenotypes, which are indicative of these lines capturing and/or creating significant gene-gene interactions. The reciprocal congenic lines for Shali1 and Shali2 represent unique allelic mouse models to identify the quantitative trait genes for these QTLs and to critically assess the bidirectional epistatic interactions between these major-effect loci.