Identification of a Genetic Region in Mice that Specifies Sensitivity to Propofol 

ILS, ISS, and LSXSS RI mice were produced in the Specific Pathogen Free facility at the Institute for Behavioral Genetics, University of Colorado at Boulder. ILS and ISS were derived from LS and SS by 20 rounds of brother-sister mating in the absence of any behavioral selection. [15]The LSXSS RIs were similarly derived without selection by reciprocal intercrosses between LS and SS followed by > 20 generations of brother-sister mating. [15,20]Although the RI inbreeding began with 40 sibling pairs, only 25 strains were extant at the time of this study. Animals were weaned at about 25 days of age and housed one to four per cage with like-sex littermates on a 12-h light/dark cycle. Experimental protocols were approved by animal care and use committees at the University of Colorado at Boulder and University of Colorado Health Sciences Center at Denver. For the LSXSS RIs, 5 - 15 mice of each sex were tested per strain. Because mice of the same RI strain are genetically identical, the difference between testing 5 or 15 mice was minor and was accounted for by differences in the SEM. For the ILS by ISS F²s, 194 mice were phenotyped, but 30 mice were inadvertently discarded without being genotyped; therefore the F²variance of Table 1is based on 194 mice, but only 164 mice were analyzed for genetic mapping. C57BL/6J coisogenic mice having the spontaneous albino mutation c^2jwere obtained from Jackson Laboratory, Bar Harbor, Maine.

LS and SS have been shown to be differentially sensitive to propofol at all doses tested (ranging from 15–35 mg/kg), and all doses had about equal efficacy in bringing about differential sensitivity. [18]In this study, a dose of 20 mg/kg was selected for its reliability in producing LORR without causing death. Propofol (10 mg/ml in Intralipid, Stuart Chemicals, Wilmington, DE) injections were performed when mice were 60 to 94 days old. F²s had a mean age of 73.9 days (SD, 5) and a mean weight of 22.4 g (SD, 3). Dosages of 20 mg/kg (volumes ranging from 30–60 micro liter) were administered by injection into the retroorbital venous sinus using a 26-gauge, 3/8-inch needle and Hamilton syringe. Room temperature was 23–26 [degree sign] Celsius. Propofol sensitivity was measured as LORR. [18]At injection, mice immediately lost consciousness and were placed on their backs in V-shaped plexiglass troughs until recovery of the righting reflex. Animals were judged to have regained righting reflex when they turned over three times in less than 1 min. Full recovery after regaining consciousness was usually immediate. The RI and F²mice were naive to any previous injections and were injected only once with propofol. The LORRs of the (ISS c/c x C57BL/6 c^2j/C) F¹mice represent the mean of two injections spaced 7 days apart. The mean of the second injection was not significantly different from the mean of the first injection (P = 0.36; analysis of variance, two-tailed significance). The influence of tyrosine brain levels on LORR was tested by intraperitoneal injection of tyrosine (200 mg/kg) or saline (control) immediately before or 30 min before the propofol injections. Brain levels of propofol at awakening (BLA) for the RIs were determined by gas chromatography of pooled samples of brain cortex, as previously described by Simpson and Blednov. [18] 

DNA was extracted and purified from spleen, kidney, or liver using a Super-Quik Gene DNA Isolation Kit (AGTC Research, Denver, CO). Mit simple-sequence length polymorphism (SSLP) primers were obtained from Research Genetics (Huntsville, AL) and the polymerase chain reaction was carried out as previously described. [22]These primers are genetic markers that detect alleles at specific chromosomal locations differing in their polymerase chain reaction product length due to different numbers of cytosine and adenine (CA) dinucleotide repeats. [31–33]Coat color was scored by eye on living mice; pigmented RI mice were homozygous C/C and albino RI mice were homozygous c/c [34](the c locus was renamed Tyr because it encodes for tyrosinase [35,36]).

Broad-sense heritability measures dominant, additive, and between-gene interactions, whereas narrow-sense heritability measures only additive interactions. [21]Broad-sense heritability was estimated as 100% x (V^F2- V^F1)/V^F2, where V^F2is the variance of the F2 population and V^F1is the weighted average of the variances of ILS, ISS, and ILS by ISS F¹s. [21]Narrow-sense heritability was estimated as 100% x 0.5 sigma^2B/(sigma^2w+ 0.5 sigma^2B), in which sigma^2Bis the between-strain variance (the variance of the strain means minus the weighted average of the within-strain variances, which is then divided by the number of strains) and sigma^2wis the weighted average of the within-strain variances. [37]This formula assumes that interactions between genes are negligible. Within-strain variance (sigma^2w) is equivalent to environmental variance because the RI strains had undergone > 20 generations of inbreeding and thus genetic variability within strains would be expected to average < 1.4%. [21]In contrast to F²s, in which one half the mice are homozygous and the other half are heterozygous at any given locus, all of the RIs are homozygous at all loci; therefore, the between-strain variance (sigma sup 2^B) of the RIs is multiplied by one half to obtain a standardized genetic variance.

Although the heritability estimates were made from RI and F sub 2 mice derived from different parental strains (noninbred LS and SS vs. inbred ILS and ISS, respectively), the parental strains are closely related. This inference is based on the amount of inbreeding (> 60%) at the time of RI derivation. [20]This inbreeding is confirmed by the predominance of ISS and ILS alleles in the RI strain distribution pattern of SSLP markers throughout the genome. [15]It also can be seen from Table 1(section 1) that LS and ILS, SS and ISS have very similar LORRs in response to propofol.

The QTL maps were anchored using a map position for c and D7Mit31 of 44 centimorgans (cM), which is based on the Mouse Genome Database (MGD, World Wide Web URL:http://www.informatics.jax.org/mgd.html). The positions of other SSLP markers were placed relative to c and D7Mit31 based on the Whitehead Institute/MIT Center for Genome Research map (WI/MIT/CGR, World Wide Web URL:http://www.genome.wi.mit.edu), which is the more recently updated and more accurate map regarding the relative position of the SSLP markers. [38] 

The RI map was constructed using an interval-mapping procedure previously described by Markel et al. [26]The F²linkage map was constructed using Mapmaker 3.0 and QTL mapping was done using Mapmaker/QTL Version 1.1. For the F²s, two mice were excluded because LORR was not induced, and four mice were excluded because of LORRs greater than 2 SDs from the mean (15.9, 16.8, 17, and 44.7 min). Including these mice had no effect on the LOD (logarithm of the odds of linkage) profile, but the peaks were reduced by about 0.5 LOD.

Heritability is an estimate of the relative importance of heredity versus the environment. [21]We estimated heritability for the propofol sensitivity difference between LS and SS using two approaches. The first estimate was based on the variance of LORR for the ILS by ISS F²mice (Table 1, section 3). The phenotypic variance of F²mice is due to heredity and environment. However, the variance of ILS, ISS, or ILS by ISS F¹mice (Table 1, sections 1 and 2) is due only to environment, because the mice within each group are genetically identical. Therefore, the F²variance (heredity and environment) minus the average of the variance of the ILS, ISS, and F sub 1 mice (environment) provides an estimate of genetic variance. The ratio of the genetic variance to the total variance (F²variance) is heritability. In this manner, the propofol sensitivity difference between LS and SS was found to be about 55% heritable. In a similar manner, the LSXSS RI strains yielded a heritability estimate of 32%(see Materials and Methods).

We also examined the nonheritable portion of the difference between LS and SS to determine if there might be a maternal effect. Table 1(sections 2 and 4) shows that the F1 and N²(F¹s backcrossed with ILS or ISS) offspring of ILS mothers had a mean LORR about 0.7 min longer than the mean LORR of the offspring of ISS mothers. The probability that this difference was due to chance alone was less than 0.1 (P < 0.1, one-tailed test), thereby suggesting that a difference between ILS and ISS mothers could be having a small effect on LORR. The LORR for RIs and F²s did not covary with age, weight, sex, or time of day of injection.

The LS and SS selected lines are completely differentiated at the albino locus (Tyr), with all SS mice being albino and all LS mice being pigmented. When we examined the LSXSS RIs we found that the 11 albino RI strains had a mean LORR of 4.8 min (SEM, 0.3) in response to propofol, whereas the 13 pigmented strains had a significantly (P <0.0001, two-tailed test) longer mean LORR of 9.3 min (SEM, 0.4), with no difference between males and females (Figure 1(A)). (RI 32 was excluded from this analysis because it was still segregating for the albino mutation at that time.) Furthermore, the LORRs of the albino strains did not overlap with the LORRs of the pigmented strains (Figure 1(B)), suggesting a major QTL. The albino strains also had a significantly higher BLA (P < 0.0001;Figure 1(A)), suggesting the the difference in LORR was due to differential CNS sensitivity and not pharmacokinetic differences. Plotting BLA versus LORR for the RI strains (not shown) confirmed a strong negative correlation (Pearson r correlation coefficient, -0.85) that was highly significant (P <0.001).

The region around Tyr was genetically mapped using the regression approach described by Haley and Knott, [39]which we previously used in mapping QTLs for alcohol sensitivity. [26]We assessed the genotypes of seven SSLP markers spanning 50 cM of chromosome 7, obtaining peak LOD scores of 19 and 21 for males and females, respectively. The combined LOD score for males and females was about 28 (Figure 2), far exceeding the value of 3.3 recommended for significant linkage in a whole-genome scan. [40]This assignment was highly specific and positioned the QTL to a 2.5-cM region encompassing Tyr with 2-LOD support, indicating that the probability that the QTL being within this region is greater than 0.99. The LOD peak turned out to be at Tyr itself. This QTL, which we have called Lorp1 (Loss of righting due to propofol), explained about 80% of the genetic variance between LS-like and SS-like RI strains.

We also were interested to see if F²mice derived by intercrossing the ILS and ISS strains would have the same QTL indicated by the LSXSS RIs. A panel of 164 F²s was genotyped with 16 SSLP markers from murine chromosome 7 that included all of the informative markers within 10 cM of Tyr. [38]A peak LOD score of 4.4 was obtained between marker D7Mit31 within Tyr and D7Mit123 less than 2 cM distal to Tyr (Figure 3), thus confirming Lorp1. F²mice homozygous for the SS alleles of D7Mit123 had an LORR of 5.7 min (SEM, = 0.3), heterozygotes had an LORR of 6.5 min (SEM, 0.3), and LS homozygotes an LORR of 8 min (SEM, 0.4). The 2.3-min difference in LORR between mice homozygous for the S allele and those homozygous for the L allele matches the 2.2-min difference previously observed between LS and SS. [18]This difference is consistent with all of the genetic difference between LS and SS for propofol sensitivity having been captured in the ILS and ISS strains. This difference also suggests that most or all of the genetic difference between ILS and ISS is due to Lorp1. Lorp1 explained about 23% of the F²genetic variance.

Mapping of the RI BLA data also supported a QTL linked to Tyr, with a statistically significant peak LOD score of 3.3 occurring at the same location as Lorp1 (not shown). RIs homozygous for the L allele (long LORR) of Lorp1 had a propofol BLA of 8.4 (g/g brain), whereas those homozygous for the S allele (short LORR) had a BLA of 17.2 (g/g brain), a highly significant difference (P = 0.001, one-tailed test assuming equal variances) previously seen in the LS and SS lines. [18]That mice with the S allele had a significantly higher BLA indicated that their CNSs were less sensitive to propofol and not that they were better able to metabolize propofol or sequester propofol outside the CNS (as in fat tissue). F²s could not be used to confirm BLA due to the need to pool samples from several mice to meet the minimum sensitivity requirements of the propofol assay (another advantage of RIs).

Considering the tight linkage between Lorp1 and Tyr, we asked whether the resistance of SS mice to propofol might result from the albino phenotype. This was tested by crossing ISS (c/c) with C57BL/6 mice from the Jackson Laboratory that were completely inbred but heterozygous for a spontaneous albino mutation (c^2j/C). The F¹progeny were thus coisogenic, identical at all loci except Tyr; all F sub 1 progeny derive a c allele from the ISS parent and either a c^2jor C allele from the other parent. Thus one half of the F¹mice are albino (c/c^2j) and the half are normally pigmented (c/C). If propofol resistance were due to albinism, then albinos would be expected to have a mean LORR about 0.8 min less than their pigmented littermates based on the ILS by ISS F²results. However, the albino LORR was not significantly less, and a 0.8-min difference was ruled out at the 95% confidence level after accounting for sex (Figure 4(A)).

Because Tyr encodes for the enzyme tyrosinase, we also tried to determine whether tyrosine injections might alter LORR. The connection between Tyr and tyrosine levels was appealing because tyrosine is converted to L-dopa in the rate-limiting step of catecholamine synthesis, and tyrosine shares structural similarities with propofol. It also was intriguing that tyrosine availability was previously shown to alter the ethanol-induced LORRs of LS and SS, [41]and a QTL (Lore7) for ethanol-induced LORR has been found to be linked to Tyr. [28]It was found, however, that coinjecting the same dose of tyrosine that influenced ethanol-induced LORR had no influence on propofol-induced LORR for either ILS or ISS mice (Figure 4(B)).

Differential sensitivity to propofol between LS and SS has a substantial genetic component, as indicated by testing both LSXSS RIs and ILS by ISS F²s. The F²estimate of heritability was 55%, which is considerably greater than the more limited estimate of 32% from the RIs, which detects only one type of genetic effect (additive). The difference between the two estimates suggests that in the F²s there are additional dominance or between-gene interactions, or both. Part of the difference could be due to the RIs having been derived from noninbred LS and SS mice, whereas the F²s were derived from inbred ILS and ISS mice. However, the difference between LS and ILS and between SS and ISS is small (see Materials and Methods) and thus unlikely to be a major factor.

A QTL, Lorp1, is tightly linked to Tyr in the middle of murine chromosome 7. This QTL accounts for about 80% of the genetic variance in the RIs, indicating that this is the major locus specifying differential propofol sensitivity between LS and SS. The LORR difference of 2.3 min observed between ILS by ISS F²mice homozygous for the S allele of Lorp1 compared with those homozygous for the L allele also is sufficient to account for all of the LORR difference observed between LS and SS. Although the involvement of other genes is suggested by the heritability estimates, their effects were not detected in the LSXSS RIs, and no whole-genome map of F²s has yet been generated.

The LORR of F²s heterozygous at Lorp1 (6.5 min) was more similar to that of S allele homozygotes (5.7 min) than L homozygotes (8 min). Therefore the S allele appears to be partially dominant over the L allele, although a strictly additive model cannot be ruled out. It is curious, however, that heterozygotes and S allele homozygotes were more similar in LORR to their ILS progenitor (7.4 min) rather than ISS (3.8 min). This could be due to a difference in gene interactions, but it also could be due to environmental differences (e.g., room temperature, noise level, injection proficiency) because the F²s were not tested at the same time as ISS and ILS. In this regard, we have observed that the mean LORR of ILS or ISS can vary from day to day by 1–2 min (in our experiments, however, ILS and ISS were almost always tested together on the same day).

Two methods of mapping were used to position Lorp1: RI interval mapping and F²mapping. Unlike the usual situation, the RI mapping proved to be extremely powerful, resulting in an LOD score of 28. Therefore, the identification of Lorp1 was not merely provisional, but rather highly significant. Further, the RI mapping localized Lorp1 to a 99% confidence interval of just 2.5 cM region. By comparison, large-scale F²mapping of QTLs for ethanol-induced LORR resulted in 99% confidence intervals that were an order of magnitude larger. [28] 

Given the RI mapping result, it was surprising that the F²result did not approach the same significance or sharpness of LOD peak. This probably occurred for several reasons. First, the number of F sub 2 s used (164) is relatively small, much smaller than the number used (approximately 1,000) to confirm ethanol sensitivity QTLs. [28]Second, Lorp1 is a major QTL nearly equivalent to a single-gene effect. Therefore, the ability to test an RI strain repeatedly to reduce environmental variance can drive even a single recombination event between the QTL and a tightly linked marker to very high statistical significance, thus excluding the region beyond that marker. Such independent, repeated testing cannot be done with an F²population. Third, the QTL effect size (the amount of the genetic effect resulting from variation at a particular locus) between SS-like and LS-like RIs was 4.5 min compared with only 2.3 min between F²homozygotes. This suggests additional gene interactions in the F²s that attenuated the influence of Lorp1; however, it also could reflect environmental differences given that RIs and F²s were not tested at the same time.

The LOD score for BLA peaked at the same location as Lorp1, suggesting that Lorp1 affects neurosensitivity. However, the LOD for BLA was much lower at 3.3 than the LOD for Lorp1 at 28. This is most likely due to the smaller number of BLA measurements, a higher coefficient of variation (SD/mean) for BLA, and not including RI strain 1 having a recombination close to Tyr. The difference also could suggest that the LORR difference is not entirely determined by BLA, although variation in BLA did explain about 72% of the variation in LORR.

Provisional QTLs for LORR induced by ethanol, propanol, and chloral hydrate also have been linked to Tyr in LS and SS mice. [25,26]The ethanol QTL, confirmed by Markel and Corley, [27]is called Lore7. [28]It is possible that Lore7 and Lorp1 are the same; however, the map position of Lore7 has a 90% confidence interval of 35 cM and thus is not nearly as sharply defined as the position of Lorp1.

Given the very tight linkage between Lorp1 and Tyr, we considered whether albinism could be responsible for the decreased sensitivity of the SS and ISS strains. However, comparison of coisogenic albino and nonalbino F¹mice from the ISS x C57BL/6 c^2j/C cross argues that albinism per se is not a causal factor. Albinism also seems to be an unlikely factor given that the c allele is completely recessive with respect to albinism, whereas the S allele of Lorp1 is partially dominant or additive. Although the coisogenic result does not rule out an effect unrelated to albinism and unique to the c allele, such an effect seems unlikely given that the c and c^2jalleles abolish tyrosinase activity [42–45]and thus would appear to be functionally equivalent in all aspects.

The coisogenic experiments also do not exclude the possibility that Tyr affects LORR due to the C allele of LS. If the C allele is responsible, a simple hypothesis is that this allele leads to an increase in tyrosinase activity that reduces the availability of tyrosine in the brain for other purposes such as synthesizing catecholamines. If it does, then increasing tyrosine levels might alleviate this competition and cause LS mice to have a shorter LORR similar to SS mice. This is exactly the result obtained when LS mice were coinjected with tyrosine and measured for ethanol-induced LORR (SS mice were unaffected). [41]In our study, however, coinjecting the same dose of tyrosine as that used in the ethanol study had no effect on propofol-induced LORR in the LS or SS mice. Although these results still do not rule out unusual effects of Tyr, they argue against Tyr's involvement due to modulating tyrosine levels. These results also imply that, even though propofol and tyrosine are structurally similar, propofol does not act through competing with tyrosine.

Many other genes besides Tyr fall within the 99% confidence interval of 2.5 cM for Lorp1. Assuming an average gene density, 2.5 cM would be expected to harbor about 150 genes. So far, however, only about ten genes have been either physically or genetically mapped to this region with better than 1-cM resolution (MGD map). These genes are Cckbr, encoding the cholecystokinin B receptor; Cht, a coat color gene; eed, essential for embryonic ectoderm development; fit1, essential for normal growth and fertility; jdf, essential for normal juvenile development and fertility; mod2, encoding mitochondrial malic enzyme; mod2r, encoding a regulator of mod2, and Rps17-rs6 and Rps4-rs2, encoding ribosomal-related proteins. Of these, the cholecystokinin B receptor gene is interesting because it encodes a neurotransmitter receptor expressed throughout the CNS that is known to influence arousal and excitability. [46]However, the position of the human CCKBR gene [47]seems to suggest this gene is closer to Omp and Hbb than it is to Tyr and thus just outside the 99% confidence interval. The TYR locus of humans is on chromosome 11, within cytogenetic band q21 [48]; therefore, other genes being identified in this region as part of the Human Genome Project will suggest additional candidates for Lorp1.

Although studies describe lines of mice differentially sensitive to nitrous oxide, [49]diazepam, [50]and opiates, [51]no loci specifying these differences have been mapped. Therefore, Lorp1 is the first locus specifying differential sensitivity to a clinically relevant general anesthetic agent to have been mapped in a mammalian species. In nonmammalian species, single-gene mutants in the nematode Caenorhabditis elegans and in the fruit fly Drosophila melanogaster have been found to be differentially sensitive to the anesthetic effects of halothane, [52–56]and QTL studies in C. elegans [57]have revealed additional loci underlying halothane sensitivity. There is optimism, therefore, that identifying these gene products will pinpoint molecular pathways directly relevant to the action of general anesthetics.

The authors thank M. Beeson, L. Gordon, and A. Keller for technical support; and J. DeFries and B. Bennett for advice; and Drs. C. P. Gibbs and R. A. Deitrich for their support.