Synchronizing Allelic Effects of Opposing Quantitative Trait Loci Confirmed a Major Epistatic Interaction Affecting Acute Lung Injury Survival in Mice

Increased oxygen (O2) levels help manage severely injured patients, but too much for too long can cause acute lung injury (ALI), acute respiratory distress syndrome (ARDS) and even death. In fact, continuous hyperoxia has become a prototype in rodents to mimic salient clinical and pathological characteristics of ALI/ARDS. To identify genes affecting hyperoxia-induced ALI (HALI), we previously established a mouse model of differential susceptibility. Genetic analysis of backcross and F2 populations derived from sensitive (C57BL/6J; B) and resistant (129X1/SvJ; X1) inbred strains identified five quantitative trait loci (QTLs; Shali1-5) linked to HALI survival time. Interestingly, analysis of these recombinant populations supported opposite within-strain effects on survival for the two major-effect QTLs. Whereas Shali1 alleles imparted the expected survival time effects (i.e., X1 alleles increased HALI resistance and B alleles increased sensitivity), the allelic effects of Shali2 were reversed (i.e., X1 alleles increased HALI sensitivity and B alleles increased resistance). For in vivo validation of these inverse allelic effects, we constructed reciprocal congenic lines to synchronize the sensitivity or resistance alleles of Shali1 and Shali2 within the same strain. Specifically, B-derived Shali1 or Shali2 QTL regions were transferred to X1 mice and X1-derived QTL segments were transferred to B mice. Our previous QTL results predicted that substituting Shali1 B alleles onto the resistant X1 background would add sensitivity. Surprisingly, not only were these mice more sensitive than the resistant X1 strain, they were more sensitive than the sensitive B strain. In stark contrast, substituting the Shali2 interval from the sensitive B strain onto the X1 background markedly increased the survival time. Reciprocal congenic lines confirmed the opposing allelic effects of Shali1 and Shali2 on HALI survival time and provide unique models to identify their respective quantitative trait genes and to critically assess the apparent bidirectional epistatic interactions between these major-effect loci.


Introduction
Acute lung injury (ALI) and acute respiratory distress syndrome (ARDS) continue to have high mortality [1][2][3][4], despite decades of research and numerous randomized clinical trials [5][6][7][8][9][10]. A significant reduction in mortality has been achieved with protective ventilation strategies [11], but pharmacological attempts remain disappointing [6,10,12]. Among the supportive measures available, use of supranormal oxygen (O 2 ) to correct the severe hypoxemia is integral to the management of ALI/ARDS patients. Oxygen therapy is also an essential treatment in many other acute (e.g., trauma) and chronic (e.g., COPD) cases. Nonetheless, the paradoxical detrimental pulmonary effects from high O 2 levels remain a complicating factor in critical care medicine. Basic research, on the other hand, has taken advantage of the damaging oxidative capacity of high-dose O 2 in laboratory animals to directly study ALI development and pathology [13][14][15][16][17][18], and to identify and critically evaluate potential protective and therapeutic strategies [19][20][21][22].
The biologic and pathologic features of ALI are well established [4,23]; yet, the identification of key genes affecting differential susceptibility and outcome has been elusive. With a long term goal to determine the critical genes controlling oxidant-induced ALI morbidity and mortality, we have established a susceptibility mouse model using hyperoxic acute lung injury (HALI) survival time for genetic analysis and gene discovery [24]. Two distinct large recombinant populations were generated for quantitative trait locus (QTL) analysis, using sensitive C57BL/6J (B) and the considerably more resistant 129X1/SvJ (X1) inbred mouse strains as the progenitors. Analysis of F 2 mice identified 5 QTLs (designated Shali for Survival time with hyperoxic acute lung injury) significantly linked to HALI survival time on Chr 1 (Shali1 and the male-specific QTL, Shali5), Chr 4 (Shali2), Chr 9 (Shali4), and Chr 15 (Shali3) [25]. The large backcross population also identified Shali1, Shali2, and Shali3 and, along with the F 2 data, consistently suggested that the Shali1 locus on Chr 1 and the Shali2 locus on Chr 4 had opposing allelic effects on overall HALI survival time within each inbred strain [26]. Specifically, QTL analysis of recombinants derived from the X1 and B progenitor strains determined that Shali1 directly correlated with the overall survival time trait of the parental strains, with resistant X1 strain alleles leading to an increased mean HALI survival time and sensitive B strain alleles yielding increased sensitivity. Conversely, phenotype data of backcross and F 2 recombinants supported that Shali2 had allelic effects in opposition to Shali1, with sensitive strain B alleles for Shali2 prolonging HALI survival time and resistant strain X1 alleles for Shali2 increasing sensitivity. The individual contributions of these QTLs to the overall survival time were estimated to change the survival time about 23 and 15 hrs in the corresponding direction for Shali1 and Shali2, respectively [25].
Besides the opposing allelic effects on overall HALI survival time for the two main QTLs, three additional factors have added significantly to the complexities of these genetic studies. First, the sensitive B strain showed a significant sex difference, with males more resistant than females. Interestingly, although no linkage was mapped to the X-chromosome, we did identify a male-specific QTL (Shali5) on distal Chr 1 [25]. Second, even though the X1 strain is considered a resistant strain (based on mean survival time; MST), the penetrance (i.e., the proportion of mice carrying the QTL and demonstrating the phenotype) of the resistance trait is low, with a total male-female average of ,33% penetrance. And, among the one-third of X1 mice that survived to the resistance threshold (.153 hrs), only about a third of these (i.e., ,10-15% of total X1 mice) reached a survival time of 200 hrs. This decreased penetrance can significantly increase the difficulties of identifying the resistance gene(s). Third, at least one additional locus (i.e., Shali5 and possibly Shali3) also showed an allelic effect on HALI survival opposite to that expected, with B alleles predicted to confer added HALI resistance and X1 alleles increased sensitivity. These complexities are likely inherent to most complex traits [27][28][29][30][31][32].
To verify the predicted QTL effects of the major loci in vivo, reciprocal congenic lines were constructed for Shali1 and for Shali2 on both parental strain backgrounds, which involved substituting the respective donor B strain segment onto the resistant X1 strain and substituting the donor X1 strain segment onto the sensitive B strain for both QTLs. Effects of this genetic restructuring on trait penetrance was also examined. Consistent with our earlier QTL analyses results, these congenic lines demonstrated significant changes in HALI survival time compared to their corresponding progenitor strain, thereby providing strong in vivo evidence for the existence and capture of these QTLs. By generating a congenic line on the resistant X1 strain background that was even more sensitive than the sensitive B strain, we further validated the bidirectional alleles of these QTLs. And, by significantly augmenting the overall effects on HALI survival time in both directions, these congenic lines revealed important epistatic interactions between the two Shali loci. These data support that the validated congenics represent an excellent model system to further delineate these loci and can be expanded to include congenics for additional putative Shali loci and other critical genes affecting this multigenic response.

Materials and Methods
Mice C57BL/6J (B) mice (i.e., the HALI-sensitive strain) and 129X1/ SvJ (X1) mice (i.e., the HALI-resistant strain) were purchased from the Jackson Laboratory (Bar Harbor, ME) and used in the construction of the reciprocal X1.B (i.e., X1 strain background with a B-derived QTL substitution) and B.X1 (i.e., B strain background with a X1-derived QTL substitution) congenic lines for Shali1 and Shali2, on chromosomes 1 and 4, respectively. Note: to avoid confusion, the ''X1'' abbreviation is used in this report (and all future communications) to denote the resistant inbred strain, which differs from our previous published work using ''S'' as the abbreviation for the resistant strain. B strain mice were brother-sister mated in-house to generate breeders and sensitivestrain controls for exposures. Inbred stocks were replaced at least every 10 generations to reduce strain drift. Due to their high rate of conjunctivitis, X1 mice were purchased (Jackson Laboratory) as needed for use as breeders and as resistant-strain controls for the hyperoxia exposures.

Ethics Statement
Mice were handled in accordance with the protocols approved by the IACUC of Cincinnati Children's Hospital Medical Center.

QTL genotype analysis
To demonstrate the reciprocity of the different allele combinations at the representative QTL peak loci and to help quantify the opposing effects of Shali1 and Shali2 in the model, our previous QTL dataset of 840 F 2 recombinants [25] was re-sorted and replotted for direct allelic comparisons. All F 2 mice (n = 840) were typed for D1Mit303 (the peak marker of Shali1) and sorted into three genotypes groups: those homozygous for X1-derived alleles, heterozygous, or homozygous for B-derived alleles; the MST was calculated for each allelic group of mice. Similarly, the 837 F 2 mice successfully typed for D4Mit308 (the peak marker of Shali2) were segregated into the three genotypic groups and their MSTs calculated. Next, the combined genetic contribution for Shali1 and Shali2 was estimated by first sorting the 837 mice typed for both D1Mit303 and D4Mit308 markers into the nine different possible genotypes, and then calculating and plotting the MSTs for these nine groups of mice.

Construction of consomic and congenic lines
The consomic and congenic lines described in these studies were generated over the past 3-4 years. Because Chr 1 contained two distinct QTLs separated by a large distance (i.e., Shali1 and the male-specific QTL, Shali5), and Chr 4 maintained a high LOD score for a large portion of the chromosome [25], we initiated construction of consomic lines for these two chromosomes. Reciprocal lines for Shali1 and Shali2 are designated herein as B.X1-1 and X1.B-1 and B.X1-4 and X1.B-4, respectively; background strain is listed first, followed by the substituted chromosome strain and chromosome number. A general backcross strategy and marker-assisted selection protocol [33,34] was employed, using donor males and recipient X1 strain or B strain females, where indicated. Male donors for breeding in the next generation were selected to be heterozygous for the entire donor chromosome of interest, and carrying the most recipient strain homozygosity for the remainder of the genome. Directional backcrosses were performed to ensure the corresponding recipient strain-derived mitochondria and Y-Chr in the final lines. A genomewide SNP analysis (Mouse MD Linkage Panel; Illumina, San Diego, CA) of lines at N 3-4 was performed at Cincinnati Children's Genetic Variation and Gene Discovery Core (http:// dna.chmcc.org/), and additional microsatellite markers were typed to directly target removal of the remaining heterozygous regions.
Near the end stages of the multi-generation screening process of constructing the four reciprocal consomic strains, select recombinants were identified that contained or did not contain a conservative Shali1 or Shali2 interval. After reproducing each recombination in the opposite sex by backcrossing to the appropriate parental strain, the congenic regions were fixed to homozygosity to 1) confirm the QTL effect and determine the level of overall phenotype represented by the captured region, 2) further resolve the congenic intervals, and 3) maintain the lines. The Mouse Universal Genotyping Array (MUGA; GeneSeek, Lincoln, NE), which includes .7,800 SNPs at an average spacing of 325 Kb, was used to verify the genetic makeup of the most relevant final consomic and congenic lines. Small regions of heterozygosity still remaining at this stage were eliminated with additional backcrosses using targeted microsatellite or SNP markers and the final lines reestablished to homozygosity.

DNA preparation
Genomic DNA was isolated from 2-3 mm tail clips using a commercial DNA extraction kit (Wizard Genomic DNA, Promega; Madison, WI). A Nanodrop 8000 (Thermo Fisher Scientific; Waltham MA) was used to analyze samples for purity (A 260 /A 280 ) and DNA content (A 260 ). DNAs were diluted to ,20 ng/ml for PCR requiring agarose gel separation or to ,5 ng/ml for PCR using fluorescence genotyping.

Genotype analysis
Primer pairs for polymorphic markers between the B and X1 strains were purchased from IDT (Coralville, IA). PCR was performed in 15-mL volumes in 96-well plates (USA Scientific; Ocala, FL) using a 4-block thermocycler (BioRad, Model PTC-225 or PTC-240) as described previously [25]. Markers with PCR product allele sizes of 5% or more were separated by 2.5-4% agarose (ISC BioExpress; Kaysville, UT) gels and stained with ethidium bromide. Microsatellite markers ,5% different in allele size were amplified from 20 ng total DNA using fluorescent primers synthesized by Applied Biosystems (ABI; Foster City, CA) and protocols provided. Fluorescent PCR products were separated using an ABI-3730xL sequencer located at the Cincinnati Children's Genetic Variation and Gene Discovery Core (http:// dna.chmcc.org/) and genotypes ascertained using GeneMapper software (V3.5, ABI). Besides using targeted microsatellite analysis whenever possible, real-time qPCR of individual SNPs (ABI) was used to remove any remaining small heterozygous regions identified by the MUGA SNP panel (GeneSeek, Inc.).

Hyperoxia Exposures
Mice in standard shoebox cages with food and water ad libitum were placed inside a 0.13-m 3 Plexiglas inhalation chamber (manufactured to specifications by Stellar Plastics, Detroit, MI) and exposed continuously to .95% O 2 until death. Each chamber housed up to 9 cages with up to 4 mice per cage; individual mice were clearly visible from above and through the sides of cages. O 2 level in the chamber was continuously controlled using a ProOx 110 portable O 2 monitor (Biospherix, Redfield, NY). Status of the exposures was closely monitored, such that the time between measures was within 5% error of the overall exposure to that point. The exposure time used for each mouse was the mean time between the last time check and the time the mouse was identified as dead. Mice were exposed between 6 and 12 weeks of age and, to prevent hypoxic seizure upon opening the chamber, each exposure was continued uninterrupted until all mice within the chamber succumbed. Because of the large number of control mice exposed over the years, and the decreased penetrance seen in the resistant X1 inbred strain, the B and X1 strains are presented as historic controls for group comparisons, and include data previously published [24] combined with new data for the total control mice.

Statistical analyses
Results are presented as mean survival times (MSTs) 6 SEM. Survival data related to the QTL genotype analyses and the strain comparisons were not normally distributed. For QTL genotype analyses, one-way ANOVA on ranks was used to compare MSTs of F 2 mice with the same single-marker (D1Mit303 or D4Mit308) or two-marker (D1Mit303+D4Mit308) genotypes (i.e., B-B, H, or X1-X1). Tukey posthoc adjustments for multiple comparisons were used to keep the overall family-wise type I error rate at 0.05. A Mann-Whitney-Wilcoxon (MWW) U-test was used for strain comparisons. Statistical differences were first assessed for sex in the control B and X1 strains. X1 strain males and females did not differ (p = 0.72), whereas B strain males were significantly more resistant than B strain females (p = 7.9610 25 ). Therefore, all data were stratified by sex. Sex-specific between group comparisons (B vs. each B.X1-1 congenic, B vs. each B.X1-4 congenic, X1 vs. each X1.B-1 congenic and X1 vs. each X1.B-4 congenic) were assessed using a MWW U-test and a Bonferroni correction (38 tests). A sexmatched comparison of the hypersensitive X1.B-1A line to the sensitive B strain control was also performed (MWW U-test). After accounting for multiple comparisons, differences were considered statistically significant when p,0.00132 (i.e., p,0.05/38).

Allelic analysis of Shali1 and Shali2
Our previous QTL data suggested that the Shali1 and Shali2 allelic effects opposed each other in HALI survival time [24][25][26]. To quantitate and compare these QTL effects on survival time, allelic analysis of the major QTLs was performed in the F 2 dataset [25]. For this, the allelic effects of the peak markers for Shali1 and Shali2 loci were calculated, first for the separate QTLs and then for both QTLs. F 2 mice with homozygous X1 alleles for D1Mit303 were the most HALI resistant, with heterozygotes intermediate in MST and mice homozygous for B alleles at D1Mit303 being most sensitive. This directly correlated with X1 being the resistant strain. In contrast, F 2 mice that were homozygous for B alleles at D4Mit308 were the most HALI resistant, even though the inbred B strain is HALI sensitive. Similarly, F 2 mice that were homozygous for X1 alleles at D4Mit308 were most sensitive, even though inbred X1 mice are a resistant strain. Mice heterozygous at D4Mit308 were again intermediate in survival time. Because this data provided a key finding that formed the basis of the breeding strategies reported in this study, we have re-plotted these data to clearly demonstrate this inverse relationship. In this format, the opposing actions of the homozygous B and X1 alleles for Shali1 and Shali2 can be seen as near mirror images, showing a similar, but opposite effect of the two QTLs and HALI survival times ( Figure 1A). The overall survival time with homozygous X1 alleles at D1Mit303 (145 hrs) was similar to that for homozygous B alleles at D4Mit308 (140 hrs). Heterozygotes for either QTL marker also had similar MSTs to each other (both ,130 hrs) and to the control X1 strain (133 hrs).
The genotype and phenotype data for the F 2 population was further re-sorted into groups of mice that carried one of the 9 possible pairwise genotypes at D1Mit303 and D4Mit308; data for the nine groups was displayed according to increasing MST ( Figure 1B). A visual comparison of these 9 groups revealed a consistent pattern for the reciprocal allelic pairs. For example, bars 1 and 9, which include the reciprocal (B-B, X1-X1) and (X1-X1, B-B) genotype pair for the Shali1 and Shali2 markers, respectively, demonstrated the polar extremes of MSTs among the 9 genotype pairs. Moving inward from each end in Figure 1B, we see that bars 2 and 8, 3 and 7, and 4 and 6 directly matched reciprocal pairs at the corresponding increasing and decreasing MSTs. The group of mice heterozygous (i.e., B-X1 or X1-B; listed as H) for both Shali markers was in the exact center of the range, as would be predicted by the reciprocal effects. Again, the group of mice that were heterozygous at both loci had a MST no different than the control X1 strain.
To validate these in vitro allelic predictions in vivo, we generated a pair of reciprocal congenic lines for these two QTLs on both parental background strains. Figure 2 displays a genomic overview of the most important reciprocal congenic strains for Shali1 (X1.B-1A and B.X1-1A) and Shali2 (X1.B-4BB and B.X1-4A) compared to each other and to the X1 and B inbred progenitor strains. Specific details of these lines are described below. From this illustration, one can quickly note the genomic differences between these congenic lines and their respective progenitor (recipient) strains.

Generation of Shali1 lines
Reciprocal consomics. Previous QTL results supported an increased survival time for X1 alleles and increased sensitivity for B alleles occurring at the Shali1 locus. QTL results also suggested that X1 and B alleles for Shali5, a putative male-specific QTL on distal Chr 1, had reversed effects on HALI survival time in males when compared to Shali1 (i.e., X1 alleles caused increased sensitivity in males, whereas B alleles added resistance in males). Reciprocal Chr 1 consomic lines were constructed to capture both Shali1 and Shali5 QTLs in the same background strain and to serve as the source and backup lines for current and future Chr 1 congenics ( Figure 3). The B.X1 and X1.B reciprocal consomic lines for Shali1 were fixed to homozygosity using microsatellite markers and/or the Illumina medium-density SNP array. The final genetic makeup of the reciprocal Chr 1 consomics was determined using the newly available MUGA SNP panel (GeneSeek, Inc.). The X1 alleles for the distal ,5 Mbp of Chr 1 were not captured in the B.X1-1 consomic line.
Reciprocal congenics. The Shali1 congenic strains were conservatively constructed to contain more than the 95% confidence interval transferred from the B strain onto the resistant X1 strain background (designated herein as X1.B-1A), or the reciprocal congenic line that contained Shali1 from the X1 strain, transferred to the sensitive B strain background (designated herein as B.X1-1A) ( Figure 3). The X1.B-1A congenic strain originated from a recombinant male that was heterozygous for the proximal half of Chr 1, but homozygous for X1 alleles at the tested microsatellites throughout the remainder of the genome. A female with the same recombination was generated from, and then bred with, the original male to fix the X1.B-1A line. Final MUGA SNP analysis of the fixed X1.B-1A congenic line mapped the crossover between SNPs at 102.558 Mbp and 103.336 Mbp. Similarly, a recombinant male carrying the distal half of Chr1 (and Shali5), with a crossover between D1Mit440 and D1Mit389 was also identified and fixed to homozygosity to capture the putative malespecific QTL (Figure 3).

Generation of Shali2 lines
The B strain is sensitive to the detrimental effects of hyperoxia. However, our earlier QTL results supported that F 2 mice carrying homozygous B alleles for Shali2 had increased resistance and those carrying homozygous X1 alleles from the resistant X1 strain demonstrated increased HALI sensitivity. Three overlapping congenic lines (X1.B-4A, X1.B-4B, and X1.B-4C) for Chr 4 were successfully produced from a short-lived Chr 4 consomic line ( Figure 4). Collectively, these 3 congenics covered most of Chr 4; MUGA SNP analyses of these lines indicated that B-derived alleles for ,66-81 Mbp were not captured. Due to poor breeding performance and poor mothering skills, few mice were generated from the X1.B-4A and X1.B-4B congenic lines. Fortunately, the results from exposing small numbers of mice for these lines quickly steered the breeding towards the X1.B-4B line, leading to the generation of X1.B-4BA and X1.B-4BB subcongenic strains ( Figure 4). The X1.B-4BB line bred reasonably well, but the X1.B-4C line, which also contains the entire Shali1 region, bred considerably better than all other Chr 4 lines, allowing us to more quickly test a larger population of mice for this line. The reciprocal B.X1-4 consomic line was bred to completion and recombinants were identified to establish three B.X1-4 congenics (B.X1-4A, B.X1-4BA, and B.X1-4C). All consomic and congenic lines were fixed to apparent homozygosity and exposed continuously to .95% O 2 . MSTs were determined for both sexes of each line and compared to the opposite sex and their appropriate parental controls.

Exposures of the reciprocal consomics and congenics
Chr 1: Shali1 and Shali5 (Figure 3). Based on results from the early Illumina SNP panel and targeted microsatellite analysis, the X1.B-1 consomic was fixed; therefore we began collecting survival time data. Recent MUGA SNP analysis of the X1.B-1 consomic identified 3 small regions still segregating for B alleles; however, we found no difference in HALI survival times for X1.B-1 mice still segregating for any of these 3 regions. Exposure of X1.B-1 mice demonstrated that B-derived full Chr 1 substitution yielded a slight, but not significant, overall decrease in survival time in X1.B-1 males compared to X1 males (124 hrs vs. 142 hrs; p = 0.13). However, a significant increase in sensitivity was found for X1.B-1 females, as compared to X1 females (99 hrs vs. 138 hrs; p = 3.1610 25 ). In fact, the MST of the X1.B-1 females matched the MST of sensitive B strain females (99 hrs), and directly correlated with substitution of Chr 1 with the B strain.
A different picture emerged for the X1.B-1A congenic line, which has the resistant X1 strain background and substitution of the proximal half of Chr 1 from the sensitive B strain, including the Shali1 interval (Figures 2 and 3). Therefore, compared to the X1.B-1 consomic strain, X1.B-1A mice lack B alleles for the distal half of Chr 1, which contains the Shali5 region. As predicted, X1.B-1A mice were significantly more sensitive than X1 mice (males, p = 3.6610 212 ; females, p = 2.2610 216 ). Surprisingly, the X1.B-1A line also differed markedly in HALI sensitivity from the sensitive B strain control for males (p = 7.2610 28 ) and females (p = 9.9610 211 ). Similar to that found previously for the control B strain [24] and the X1.B-1 consomic line (described above), X1.B-1A males and females differed significantly (p = 5.4610 210 ) in HALI MSTs (Figures 3 and 5). X1.B-1A males (MST = 93 hrs) died an average of 49 hrs (35%) earlier than recipient X1 strain males (MST = 142 hrs). Even more dramatic, X1.B-1A females (MST = 76 hrs) died an average of 62 hrs (45%) earlier than females from its background X1 strain (MST = 138 hrs). In fact, not one X1.B-1A female (of 46 mice tested) survived to the MST of sensitive B strain females (99 hrs), with survival times ranging from 61-96 hrs ( Figure 5). Accordingly, we now consider the X1.B-1A congenic line as a hypersensitive strain. Subcongenic lines for the X1.B-1A interval are in progress to capture and narrow the region to at least that identified by the B.X1-1.303 line (i.e., 59-69.8 Mbp). For X1.B-1B, which carries Shali5 -the putative male-specific QTL, males trended towards being significantly more resistant than X1 males (164 hrs vs. 142 hrs; p,0.10), whereas female X1.B-1B mice did not differ from X1 females (131 hrs vs. 138 hrs; p = 0.59). These data correlated with earlier QTL analysis results and are consistent with Shali5 containing one or more male-specific effector genes; however, additional mice are needed to more reliably validate these early findings. Our attempts to increase the sample size and to refine the Shali5 region have been unsuccessful in multiple lines, due to extremely poor survival of pups containing B alleles for Shali5, without B alleles for Shali1.
The B.X1-1 consomic line carries homozygous X1 alleles for Chr 1, including Shali1 and probably Shali5 (MUGA SNP analysis showed a loss of the distal ,3 Mbp of Chr 1). Males did not differ from B strain control males for any of the B.X1-1 consomic and congenic lines. B.X1-1 consomic females, which carry both Shali1 and Shali5, also did not differ from B strain females (p = 0.0076; compared to the Bonferroni corrected p-value of p,0.00132), but certainly trended towards the expected resistance. However, females for B.X1-1A (p = 1.8610 24 ), and B.X1-1AA (p = 6.1610 24 ) congenic lines differed significantly from B strain females. Both of these congenic lines carry Shali1, but not Shali5, supporting the resistance effect of Shali1 X1 alleles. The B.X1-1.303 congenic line carries the segment around the microsatellite marker with the highest LOD score for Shali1. Again, B.X1-1.303 males did not differ from B strain controls. B.X1-1.303 females, however, were significantly more resistant than control B females (p = 5610 27 ), thereby potentially reducing the Shali1 interval to ,10.8 Mbp. As seen with the other B.X1 substituted Chr 1 lines, the MSTs for B.X1-1.303 males and females were similar. Since control B males were considerably more resistant than B females, these congenic lines suggest a possible female-specific resistance effect for Shali1 X1 alleles (Figure 3).
Chr 4: Shali2 (Figure 4). The X1.B-4A congenic line, which does not include any part of the Shali2 interval, showed an unpredicted increase in sensitivity for males (119 hrs vs. 142 hrs), and females (104 hrs vs. 138 hrs), although these did not reach significance due to high variability and low sample size. A similar result was found for the X1.B-4BA subcongenic line. For X1.B-4B, although just a dozen mice were generated and exposed to hyperoxia, results gave good indication that the X1.B-4B line contained one or more significant resistance loci (though only males were resistant in this small subset). crossovers within the Shali2 interval, their ''no effect'' outcome can potentially be combined with the crossovers in the X1.B-4 lines that did demonstrate an effect to help refine the Shali2 segment. Subcongenic lines for the Shali2 region are in progress.

Penetrance of the high sensitivity and high resistance traits in consomic and congenic lines
Penetrance refers to the frequency in which individuals carrying a specific gene express the correlating trait. In this case, penetrance of the resistance trait was designated as a survival time .153 hrs, as defined previously [24]. Penetrance of high sensitivity was taken as the proportion of mice with a survival time less than the sex-specific MST of the sensitive B strain (i.e., females ,99 hrs and males ,114 hrs). Table 1 gives the percent penetrance (sensitivity or resistance) for the relevant X1.B-1 and X1.B-4 congenic lines showing notable changes in MST, along with the percent penetrance of increased sensitivity and resistance for the B and X1 progenitor strains.
Of all sensitive B strain mice tested, none of either sex reached the threshold survival time for designation as ''resistant''. However, 39/81 males and 32/68 females met the threshold for high sensitivity, resulting in 48% and 47% penetrance of the sensitivity trait in B strain males and females, respectively. For the X1 strain, 22/62 males and 23/74 females were resistant, giving 35 and 31% penetrance, respectively, for the resistance trait. Total penetrance for resistance (both sexes) was 33%.
Penetrance of the X1.B congenic lines differed strikingly from their background X1 strain in both directions, depending on the Shali substitution. Although X1 mice are considered a resistant strain (i.e., 33% resistant), 18% of X1 mice (10/62 males and 15/ 74 females) showed high sensitivity, with a survival time less than the sex-matched mean of the sensitive B strain (Table 1). Remarkably, even though the B substitution was on the resistant X1 strain background, the penetrance of HALI sensitivity for the X1.B-1A congenic line was 93% of total mice, with 100% of X1.B-1A females (46/46) and 88% of males (59/67) demonstrating high sensitivity. In fact, of all X1.B-1A mice tested (n = 113), only a single X1.B-1A male lived past the MST of the recipient X1 strain males (Figure 3). The X1.B-1B congenic line, which carries the male-specific Shali5 region, had 60% penetrance of HALI resistance in males (6/10), but none of 8 females were resistant. Unexpectedly, the X1.B-4A congenic line, which contains no part of Shali2, demonstrated increased HALI sensitivity compared to X1 mice, with 5/12 males and 6/10 females showing highly sensitive survival times; only one of these 22 mice had a resistant phenotype. Although these data did not significantly differ from controls (due to small sample sizes relative to the reduced penetrance), the trends suggest further studies are needed for this putative susceptibility locus. The X1.B-4BB and X1.B-4C congenic lines both carry the Shali2 interval. However, whereas the X1.B-4BB had 90% penetrance of resistance in males (19/21) and females (19/21), the X1.B-4C line showed increased penetrance of the resistance trait in males only (22/43 males vs. 8/33 females) when compared to the recipient X1 strain (Table 1).

Discussion
Our previous QTL analysis of a large F 2 population identified five putative QTLs affecting overall survival time to HALI [25]. Two of these QTLs, Shali1 and Shali2, were validated as major effectors in a separate large backcross analysis [26]. Results 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% O 2 . 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 and Shali2, 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 effect [25], 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 and Shali2 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 ( Figure 5), 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 [29,31,32,[35][36][37][38][39], 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 [40,41]. 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 [42], such complicated gene-gene interactions are likely to be commonplace in complex traits and contribute greatly to the heritable portion of phenotypic variation [43,44]. 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, genegene 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 and Shali2 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 (Figure 2). 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 and Shali2, a strategy first proposed by Bolivar et al. [45]. 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.