Introduction

Autoresuscitation (AR) is a mechanism that allows mammals to survive transient periods of hypoxia. When animals are deprived of oxygen, they will stop breathing and then exhibit bradycardia and gasping. Normal breathing resumes if air is restored during the gasping period (AR). Evidence suggests that infants who die for no apparent reason during the first year of life (Sudden Infant Death Syndrome; SIDS) often have a defective AR mechanism (Poets et al., 1999; Sridhar et al., 2003). The SWR/J infant mouse model is believed to be a particularly good animal model for SIDS in that they fail to autoresuscitate after brief exposure to severe hypoxia. It is unclear to what extent variation in resuscitation ability is due to variation in genes or environment.

Mice, like most mammals, can autoresuscitate after exposure to brief severe hypoxia. Most strains of mice can successfully autoresuscitate if re-exposed to hypoxia immediately after the first trial. Repeated hypoxic episodes eventually result in failure to autoresuscitate after several successful trials. An inbred strain of Swiss mice (SWR/J) has a developmental window (19–22-days old) during which they fail to autoresuscitate. Before and after this, AR is effective in producing recovery. As most human SIDS deaths occur during a narrow developmental window, the genetic defects of the SWR mouse model may be relevant to SIDS in humans (Jacobi and Thach, 1989).

SWR/J mice exhibit AR failure whereas other unrelated strains tested autoresuscitate as expected, suggesting a genetic basis to variation in AR. Our hypotheses were (1) Mouse strains closely related to SWR/J will have a similar AR phenotype; (2) that F1 and F2 intercross animals between SWR/J and a normal strain (BALB/cJ) will vary in their AR response and that this variation will segregate with discrete regions of the genome (quantitative trait loci; QTLs); (3) BALB/c mice that are consomic for the SWR/J chromosome 10 QTL region, will fail to autoresuscitate after less trials than the recipient BALB/cJ strain.

Materials and methods

The mice were placed in the barometric plethymograph at the start of the experiment. Humidified 97% N2−3% CO2 was then flushed in at the onset of hypoxic seizures as indicated by rapid twitching of the hind limbs (Jacobi and Thach, 1989). A measure of 30 ml of O2 was injected into the chamber to give a concentration of 21% for adults and 35 ml for weanlings, with the volumes having been determined earlier to give an O2 concentration in the chamber of ∼21 % (Jacobi and Thach, 1989). All animals were tested between 19 and 22 days of age.

Animals from six different inbred strains of both sexes, NOR/LtJ (N=13), MA/MyJ (N=11), BALB/cJ (Brunton et al., 1979), SJL/J (N=8), FVB/NJ (N=16), and SWR/J (N=15), were tested for the number of ARs survived based on the protocol described above. Three of these strains, SJL/J, FVB/NJ, and SWR/J are closely related to one another whereas the others represent strains spread throughout the remainder of the inbred mouse strain geneology (Festing, 1998). The number of AR trials survived was also determined in a total of 16 F1 and 60 F2 animals, 28 from the intercross SWR/J × BALB/cJ and 32 from the reciprocal intercross BALB/cJ × SWR/J, the female partner listed first.

DNA was extracted from the tail clips of the 60 F2 animals and processed with Qiagen kits (QIAamp DNA kit, Valencia, CA, USA). Ninety-one microsatellite loci, polymorphic between BALB/cJ and SWR/J, were scored. Loci were spaced evenly across the entire genome, at an average interval of 19.3 cM (see Appendix 1). Only 2% of the microsatellite genotypes remained unresolved.

Differences among mouse strains, among the parental, F1 and F2 hybrids, and between the chromosome 10 congenic strain and the BALB/cJ recipient strain were evaluated using ANOVA (Sokal and Rohlf, 1995) with the number of AR trials survived as the dependent variable and strain as the independent variable. Overall genetic variation among strains is measured as the ratio of between-strain variance to total phenotypic variance. This is referred to as the broad-sense heritability (H2; 14). Pairwise strain differences were tested using Tukey's Honestly-Significant-Difference Test (Sokal and Rohlf, 1995). Means and standard errors reported here are the least squares means and associated standard errors produced by the ANOVA model.

The means and variances of the parental, F1 and F2 populations were used to estimate the effective number of factors (ne) responsible for differences between the parental strains (Lynch and Walsh, 1998). The effective number of factors is the number of freely segregating loci, all with the same additive effect, that would result in the same differences in parental means and segregation variance (difference between F1 and F2 trait variances) as observed here. It is usually an underestimate of the true number of loci responsible for parental strain differences.

QTL analysis was performed using the software R/qtl 1.08–56 (Broman et al., 2003). The Haldane mapping function was used and genotypes imputed every 1 cM. Separate significance thresholds for the autosomes and the X chromsome were generated using the ‘scanone’ and ‘scantwo’ functions with 4000 permutations (Broman et al., 2006). Mapping was performed using the Haley–Knott regression method (Haley and Knott, 1992; Broman et al., 2003). We tested for a single QTL on a chromosome, a sex-specific single QTL on a chromosome, and two interacting QTL on the same chromosome or on separate chromosomes. Finally, a test for epistatic interactions between loci identified as having a significant effect on the phenotype and the rest of the genome were performed (Cheverud, 2000).

The QTL identified on chromosome 10 was used to create a partial consomic strain by moving SWR/J's chromosome 10 into the BALB/cJ background (BALB/cJ-Chr10SWR/J). Standard speed congenic techniques were used selecting positively on SWR/J genotypes for chromosome 10 and negatively on SWR/J genotypes across the rest of the genome. Genotypes were obtained from a fixed whole genome panel of 768 mouse SNPs. The 768 SNP panel, which uses the Illumina genotyping platform, was provided by the Mutation Mapping and Developmental Analysis Project (MMDAP; Brigham and Women's Hospital, Harvard Medical School). It has an average SNP density of 3 Mb across autosomes and 7 Mb across chromosome X. Two-hundred and seventy-two of these SNPs were polymorphic between BALB/cJ and SWR/J and they were distributed, on average, every 9 Mb across the genome. After six generations of backcrossing, pups were tested for AR. The mother of these pups was consomic for chromosome 10 but still carried about 13% donor genome on other chromosomes. None of these donor regions had shown evidence for QTL effects in the F2 intercross.

Results

The average number of immediately repeated transient hypoxic episodes survived by the six inbred strains of mice is provided in Figure 1 and Table 1. The broad-sense heritability is 58% indicating a strong, highly significant (Prob.=2 × 10−11) genetic effect on strain differences. SWR/J mice failed to autoresuscitate from a single hypoxic period when oxygen was restored. SJL/J, a close relative of SWR/J, does not autoresuscitate significantly more than SWR/J (Prob.=0.46) surviving an average of 2.2 trials. FVB/NJ, another close relative of SWR/J, survives 4.5 trials but is only significantly different from BALB/cJ and SWR/J, the two most extreme strains. NOR/LtJ is quite similar to FVB/NJ surviving an average of 4.8 trials and only being significantly different from SWR/J. BALB/cJ and MA/MyJ survive 7.2 and 6.8 trials, respectively. BALB/cJ survives significantly more trials than FVB/NJ, SJL/J and SWR/J whereas MA/MyJ only survives significantly more trials than the last two strains.

Figure 1
figure 1

Mean number of autoresuscitation trials survived by various inbred mouse strains. FVB and SJL are close relatives of SWR. Horizontal bars over the columns indicate sets of strains that are not significantly different from one another using pairwise tests and correcting for multiple comparisons. Whiskers indicate one standard error above the mean.

Full size image
Table 1 Least squares mean, standard error, and sample size for the different mouse strains and hybrid populations
Full size table

SWR/J was crossed with BALB/cJ to examine the genetic basis of differences in AR ability (see Figure 2; Table 1). Autoresuscitation failure is dominant in this cross as the F1 hybrids (average 6.25 trials) are not significantly different from BALB/cJ (Prob.=0.74) but are highly significantly different from SWR/J (Prob.=4 × 10−6). The F2 hybrids survived significantly fewer trials (4.0 trials) than either the BALB/cJ parent strain (Prob.=0.001) or the F1 hybrids (Prob.=0.03) and significantly more trials than SWR/J (Prob.=1 × 10−5). The effective number of factors separating the parental strains is 3.73 so there are probably at least 3–4 loci responsible for the strain differences reported here.

Figure 2
figure 2

Mean number of autoresuscitation trials for parental strains and F1 and F2 hybrids. Horizontal bars over the columns indicate sets of strains that are not significantly different from one another using pairwise tests and correcting for multiple comparisons. Whiskers indicate one standard error above the mean.

Full size image

Linkage of AR was examined in the F2 generation. Sex differences in AR are borderline significant (P=0.055) in the F2 animals with males surviving an average of 3.17 trials and females 4.78 trials. Mapping is carried out both jointly, pooling the sexes, and in each sex separately. A locus 0.6 cM proximal to marker D10Mit113 was found to affect both males and females and account for 21% of the phenotypic variance (genome-wide corrected P=0.07; see Table 2). The dominance ratio (dominance genotypic value to additive genotypic value ratio; d/a) is 0.93, which indicates nearly complete dominance of the BALB/c allele. This locus is more strongly significant when just males are included in the analysis (genome-wide corrected P=0.02; 42% of variance; see Table 2). In the male-specific analysis the dominance effect is not significant (Prob.=0.41). The QTL identified on chromosome 10 is near D10Mit113, at 66.9 Mb. The confidence interval runs roughly from 51.7 to 82.1 Mb. We refer to this QTL as Auto10.1.

Table 2 QTL mapping results: ‘a’ is the additive genotypic value (half the difference between the two homozygotes), ‘d’ is the dominance genotypic value (deviation of the heterozygote from the midpoint of the two homozygotes), CI provides the genome coordinates of the confidence interval for the QTL
Full size table

Additionally, a locus with effects on males only was identified on chromosome 12 at marker D12Mit20 (genome-wide corrected P=0.04; 45% of variance; see Table 2). This locus has a significant genotype by sex interaction (Prob.=0.012) indicating sex differences in allelic effects. The male-specific locus on chromosome 12 is overdominant (da=5.21), the heterozygotes surviving more trials than either homozygote. This QTL is located at 115.6 Mb, which is the most distal marker on the chromosome. The confidence interval runs proximally to about 101.9 Mb. No significant effects were found in females alone. This locus is referred to as Auto12.1.

We also performed a scan for loci with epistatic interactions with Auto10.1 and Auto12.1. This scan was performed jointly for both sexes because there were not enough data to support separate two-locus scans in either sex alone. Results indicate epistatic interaction between the middle of chromosome 17 (Auto17.1e) with chromosome 10, and the distal end of chromosome 4 (Auto4.1e) with chromosome 12 (see Table 3 and Figure 3). The interaction between Auto10.1 and Auto17.1e includes negative additive by additive (aa=−1.55; P=0.016) and dominance by dominance (dd=−4.26; P=0.008) interactions and accounts for an additional 13.5% of the phenotypic variance relative to Auto10.1 alone. The interaction between Auto12.1 and Auto4.1e includes a negative additive by additive (aa=−1.52; P=0.022) and positive additive by dominance (ad=3.49; P=0.002) interaction and accounts for an additional 17% of the variance.

Table 3 Genotypic values, standard errors, and probabilities associated with the two-locus models for epistatically interacting QTLs: ‘a’ is the additive genotypic value, ‘d’ is the dominance genotypic value, ‘aa’ is the additive by additive epistasis, ‘ad’ is the additive by dominance epistasis, ‘da’ is the dominance by additive epistasis, ‘dd’ is the dominance by dominance epistasis
Full size table
Figure 3
figure 3

Epistatic interactions between Auto10.1 and Auto17.1e. The figures show how the effects of genotypes at the epistatic loci change depending on the genotype at the main effect locus, and vice versa. (a) Auto10.1 genotypes are along the horizontal axis and Auto17.1e genotypes are given to the right of the lines connecting Auto17.1e genotypes across Auto10.1 genotypes. (b) Epistatic interactions between Auto12.1 and Auto4.1e with Auto12.1 genotypes along the horizontal axis and Auto4.1e genotypes identified to the right of the lines connecting them across the Auto12.1 genotypes.

Full size image

A model including the two main effect QTLs (Auto10.1 and Auto12.1) and the main effects of the two epistatic QTLs (Auto4.1e and Auto17.1e) accounts for 26% of the phenotypic variance in AR. The model accounts for a total of 53% of the phenotypic variance when the epistatic interactions between Auto10.1 and Auto17.1e and between Auto12.1 and Auto4.1e are included. Hence, epistasis accounts for 27% of the phenotypic variance, the same as the main effects alone.

Animals from the BALB/cJ-chr10SWR/J partially consomic line survived significantly fewer trials, 2.28 trials, than the BALB/cJ recipient strain (7.2 trials; see Table 1; Prob.=2 × 10−6). This result shows that the AR phenotype has been transferred along with the selected region of chromosome 10 and confirms the mapping result obtained for chromosome 10. Although the consomic mean is not significantly different from the SWR/J parent (Prob.=0.50), it is significantly greater than zero (Prob.=0.004), suggesting that there are other factors responsible for the BALB/cJ—SWR/J strain differences in addition to chromosome 10.

Discussion

We found substantial genetic variation among inbred mouse strains for ability to repeatedly autoresuscitate, with a heritability of 58%. We confirm that mouse strain SWR/J fails to autoresuscitate even after a single episode of deprivation. Strains closely related to SWR/J, including SJL/J and FVB/NJ, only resuscitate a few times before expiry, whereas other strains, such as BALB/cJ and Ma/MyJ resuscitate repeatedly across trials. These results indicate that differences among strains in AR are likely to have a genetic basis that is shared, in part, among strains related to SWR/J. This relationship can be used along with haplotypes of the strains across QTL intervals to help fine-map the genes responsible for the QTL effect.

As BALB/cJ had the most robust AR phenotype, we intercrossed SWR/J and BALB/cJ to produce populations of F1 and F2 hybrids. The F1 hybrids autoresuscitated nearly as well as the parental BALB/cJ animals showing significant heterosis because the F1 hybrid mean was 2.6 trials higher than the midpoint of the two parental strains (3.6 trials), even though they autoresuscitated fewer times than the BALB/cJ strain. This indicates strong dominance in the BALB/cJ by SWR/J cross, which can be due to overdominance, where the heterozygote survives more trials than either homozygote, or directional dominance, where BALB/cJ alleles are regularly dominant to SWR/J alleles, at individual loci and/or dominance epistasis at pairs of loci. As expected under heterosis, the F2 population regressed towards lower levels of AR (Falconer and Mackay, 1996).

Mapping results identified two main effect QTLs significant at the genome-wide 5% level in males, Auto10.1 and Auto12.1. These loci jointly account for 27% of the phenotypic variance as indicated by the adjusted R2 value when both sexes are included and 59% in the males alone. Although the sample size is small, these two QTLs are very strongly supported and have less than a 5% chance of being false-positive results. Although the percentages of phenotypic variance accounted for by these QTLs are undoubtedly positively biased because of the small size of the F2 population (Beavis, 1994), their identity as genomic locations affecting AR is not in question because considerations of sample size were included in determining the appropriate significance thresholds. We also used the parental means and F1 and F2 variances to estimate the effective number of factors responsible for differences between the parental strains. This estimate was 3.73 indicating that there are probably at least 3–4 genes affecting AR survival segregating in the cross. Therefore, it is likely that there are additional loci responsible for the parental strain difference with effects too small to detect without a larger sample. Such loci remain to be discovered in a larger experiment.

Auto10.1 was further validated and replicated by the production of a partial consomic line with chromosome 10 transferred from SWR/J into a BALB/cJ background. However, the consmoic strain, unlike SWR/J does, on average, survive a couple of trials, suggesting that other genetic factors also contribute to the differences observed between the parental strains. However, it seems likely that Auto10.1 is a major reason for the strain difference as it recapitulates nearly 70% of the original observed difference.

In the F2 generation, females survived about 1.6 trials more than males, although this difference was not quite statistically significant. Furthermore, both of the main effect QTLs were stronger in males than in females, significantly so at Auto12.1. This indicates genetic variation in sexual dimorphism for the AR trait. These results are of particular relevance to SIDS in that this bias towards effects in males is consistent with the epidemiology of SIDS in humans, in which males comprise 60% of the SIDS cases.

Auto10.1 and Auto12.1 were found to epistatically interact with chromosomes 17 and 4, respectively. These epistatic interactions served to mask the effects of Auto17.1e and Auto4.1e in the F2 population. Inspection of Figure 3 indicates that there are no differences among genotypes at Auto17.1e when averaged over the genotypes at Auto10.1. However, Auto17.1e has an additive effect with the SWR/J allele promoting resuscitation in Auto10.1 BALB/cJ homozygotes, underdominance in Auto10.1 heterozygotes, and overdominance in Auto10.1 SWR/J homozygotes. Likewise, the effects of Auto4.1e are not apparent when averaged over the Auto12.1 genotypes (see Figure 3). Here, Auto4.1e has a strong additive effect with SWR/J dominant in the Auto12.1 SWR/J homozygotes, no effect in Auto12.1 heterozygotes, and an overdominant effect in Auto12.1 BALB/cJ homozygotes. Epistatic interactions explained an additional 27% of the phenotypic variance in AR, arguing for their importance in determining variation in this mouse cross. Further confirmation of these interactions and a search for more interacting loci will require larger samples than reported here.

The comparison of the parental and hybrid generation means indicated that there was heterosis in the cross, although the F1 hybrid mean was not higher than that of the BALB/cJ parental strain. The QTL results support this conclusion with the dominance of the BALB/cJ allele at Auto10.1 and the overdominance found at Auto12.1. We also noted dominance epistasis involving both these loci. Together, these effects are consistent with the level of heterosis seen in the cross.

Cardiac glycogen reaches a developmental nadir in 21-day-old mice. Significantly cardiac glycogen is very much lower in SWR mice compared with other strains (Despande et al., 1999). We have proposed that decreased cardiac glycogen is an important factor in the failure of AR in SWR mice. With onset of hypoxia, the capacity of the heart to function depends on anaerobic glycogenolysis, which depends, in turn, on availability of adequate glycogen stores. We searched for candidate genes within the confidence regions of the two main effect QTLs. There are 361 genes in the Auto10.1 interval. The first gene in the Auto10.1 region was at 51 742 237 (Vg ll) and the last is at 82 113 069 (Tdq).

We have found a candidate gene on chromosome 10. Serine threonine protein kinase encodes an enzyme that interacts with glycogen synthase kinase 3 in the regulation of glycogen synthesis in heart muscle (Biondi and Nebreda, 2003; Taegtmeyer, 2004).

There are 274 genes identified in the chromosome 12 QTL region. The initial gene in this region is at 101 787 988 (Kp56a5) and the last at 115 595 970 (Iqhv13–2). There are two candidate genes in this region. Hexokinase-1 is among the genes we have identified on chromosome 12. Hexokinase-1 is crucial for phosphorylation of glucose and therefore is important in glycogen synthesis and also in anaerobic glycolysis (Voet and Voet, 2004). Hence, a defect in this gene could have a critical role contributing to decreased cardiac muscle glycogen as well as the decrease in anaerobic glycolysis essential for heart function during hypoxia (Beck et al., 2000). Another candidate gene in this region is 6-phosphofructokinase. The positive effect of this enzyme on glycogenolysis in perfused rat hearts has been documented (Vercesi and Focesi, 1975).

In summary, we have found 4 genomic regions (Auto10.1, Auto12.1, Auto17.1e, Auto4.1e) with effects on ability to repeatedly resuscitate in mice in a F2 intercross between SWR/J and BALB/cJ. Mapping results were found to be stronger in males than in females, consistent with the male bias in SIDS cases. Auto10.1 was replicated in a partial consomic line and seems to have a large effect on AR variation. Further fine mapping of this region will provide a tighter support interval for the QTL effect and a consequent reduction in the number of positional candidate genes.