Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

Cytochrome P-450 (CYP)¹1A2 is constitutively expressed in liver in mice and humans (Kimura et al., 1986; Sesardic et al., 1990). Hepatic expression of this enzyme also is inducible through Aryl hydrocarbon receptor (AHR)-dependent (Poland and Glover, 1973) and -independent (Ryu et al., 1996; Corcos et al., 1998) pathways. CYP1A2 is involved in the metabolism of a number of clinically significant drugs (Spatzenegger and Jaeger, 1995; Olesen and Linnet, 1997), as well as the bioactivation of heterocyclic and arylamine procarcinogens (Shimada et al., 1989).

Caffeine (1,3,7-trimethylxanthine) has been extensively studied as an in vivo metabolic probe for CYP1A2 activity in mice (Buters et al., 1996) and humans (Kalow and Tang, 1993; Fuhr and Rost, 1994; Rostami-Hodjegan et al., 1996). The 3-demethylation of caffeine to form paraxanthine (1,7-dimethylxanthine) accounts for >80% of the clearance of caffeine in humans and mice (Lelo et al., 1986; Buters et al., 1996). This activity has been attributed almost exclusively to CYP1A2 activity in humans, based on in vivo inhibition of caffeine metabolism by the CYP1A2-specific inhibitor furafylline (Tarrus et al., 1987) and by in vitro experiments with individual human CYP isoforms expressed from cDNAs (Gu et al., 1992). A Cyp1a2 (/) knockout line was used to demonstrate that 87% of the clearance of caffeine was attributable to CYP1A2 activity in mice (Buters et al., 1996).

CYP1A2 activity is most commonly assessed either by estimating clearance of caffeine from urinary metabolite ratios (Kalow and Tang, 1991; Butler et al., 1992) or by determining the ratio of paraxanthine to caffeine in plasma at a single time point (Fuhr and Rost, 1994).

A number of population studies using caffeine metabolites have provided evidence for a genetic polymorphism of CYP1A2 expression in humans based on the determination of bi- or trimodal phenotypic distributions (Butler et al., 1992; Fuhr and Rost, 1994; Nakajima et al., 1994) Other population studies with caffeine metabolites have reported unimodal population distributions for this trait (Kalow and Tang, 1991; Vistisen et al., 1992). It has been proposed that confounding variables such as altered renal function could contribute to the apparent bimodal distribution of CYP1A2 activity seen in some studies that used urinary caffeine metabolite ratios (Tang et al., 1994; Rostami-Hodjegan et al., 1996).

The study of genetic variation of CYP1A2 expression in human populations is somewhat confounded by the environmental responsiveness of the expression of the gene. Because numerous environmental exposures, including foods, drugs, smoking, and industrial pollutants may influence gene expression, it is extremely difficult to discriminate these factors from genetic variation in expression. The use of caffeine metabolite indices as surrogate markers of CYP1A2 expression may further complicate the analysis due to variations in factors other than CYP1A2 that may be contributing to the phenotype.

The laboratory mouse offers the opportunity to study genetic variation in gene expression in the absence of differential exposure to environmental inducers. Phenotypic differences among inbred mouse strains can be exploited to identify genetic differences in the expression of genes underlying the trait under consideration. We had previously demonstrated interstrain variation in the phenotypic trait of serum paraxanthine/caffeine index among inbred mice. We further demonstrated that this phenotypic parameter predicted significant differences in Cyp1a2 gene expression between the acetaminophen nonsusceptible (APN) inbred strain developed in our laboratory (Casley et al., 1997a) and a common inbred laboratory strain, C3H/HeJ (Casley et al., 1997b).

In the present study, we have tested the hypothesis that the paraxanthine/caffeine index is a polygenic trait in the mouse. We have undertaken a genome-wide scan to identify and map quantitative trait loci (QTLs) affecting this phenotype with an interval mapping approach. This approach tests the likelihood of an association between the quantitative value of a phenotypic trait and specific regions of the genome defined by codominant genetic markers of fixed position (Lander and Botstein, 1989).

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

Standards and Reagents. Caffeine and 8-chlorotheophyllline were purchased from Sigma Chemical Co. (St. Louis, MO). Polymerase chain reaction primers were purchased from Research Genetics (Huntsville, AL). Taq DNA polymerase, deoxynucleotides, and reaction buffers were purchased from Boehringer Mannheim Canada Inc. (Laval, Quebec, Canada). All other chemicals were reagent grade or higher quality.

Mouse Lines. APN inbred mice were derived by selection from an outbred Swiss-Webster colony on the basis of resistance to acetaminophen-induced hepatotoxicity as described previously (Casley et al., 1997a). APN mice used in the present study were from the 21st and 23rd inbred generations. C3H/HeJ mice were obtained from The Jackson Laboratory (Bar Harbor, ME), and were maintained in our animal care facility for 3 weeks before paraxanthine/caffeine index phenotyping and initiating crosses. All animals were housed and cared for in the animal care facility of the Health Protection Branch of Health Canada (Ottawa, Ontario, Canada). Housing and experimental protocols were in accordance with Canadian Council on Animal Care guidelines.

Paraxanthine/Caffeine Index Phenotyping. Animals at 7 weeks of age, with free access to food and water, were dosed by oral gavage with an aqueous solution of caffeine (1 mg/ml; 10 mg caffeine/kg b.wt.). Serum caffeine and paraxanthine levels were determined 2 h after dosing by HPLC and paraxanthine/caffeine peak area ratios were calculated as described previously (Casley et al., 1997a). Paraxanthine/caffeine indices were determined for all parental, F₁, and F₂ animals.

Genetic Analysis. Reciprocal crosses of APN and C3H/HeJ mice generated an F₁ of 153 animals. F₁ animals were crossed to produce 441 F₂ animals. The number of genes affecting the paraxanthine/caffeine index segregating in the F₂ intercross was estimated separately for males and females according to the formula: N = [(1/k)  ·  (P₁  P₂)² ]/V_g, where N = no. of segregating loci; 1/k = coefficient of difference between the square of parental means (k = 8 for an F₂ intercross); P₁, P₂ = phenotypic means of parental strains; and V_g = genetic variance (calculated as: V_g = V_F2  [1/4 (V_P1 + V_p2) + 1/2 (V_F1)] (Wright, 1968). Phenotypic data from crosses were compared by application of the Mann-Whitney rank sum test to detect significant differences between groups. The rank sum test was used to allow comparisons between groups within which the data were not normally distributed or had unequal variances. The Kolmogorov-Smirnov analysis was used to test for normality of distribution within groups. All analyses were done with the SigmaStat software system (Jandel Sci., Corte Madera, CA). Phenotypic data from the F₂ males and females were log-transformed to obtain a normal distribution for subsequent analysis. To account for gender differences in analyzing F₂ data, the male and female data sets were merged as described by Taylor and Phillips (1996).

Genotyping. Genomic DNA was extracted from ~1-cm tail clippings taken post mortem from the 438 F₂ animals with a commercial spin column procedure (QIAamp; QIAGEN Inc., Mississauga, Ontario, Canada) according to the manufacturers instructions. DNA was eluted in 10 mM Tris, pH 8.8, and stored at 20°C. Short tandem repeat (STR) genetic markers were amplified in a 96-well microplate format in a 10-µl reaction containing 40 ng of genomic DNA; 2.0 mM MgCl₂; 200 µM each of deoxy-ATP, deoxy-cytidine 5'-triphosphate, deoxy-ribothymidine 5'-triphosphate, and deoxy-GTP; 10 mM Tris-HCl, pH8.3; 50 mM KCl; 0.25 U Taq DNA polymerase; and 240 µM each of forward and reverse primers. Polymerase chain reaction conditions were 2 min at 94°C followed by cycles of 20 s at 94°C, 45 s at 55°C, and 1 min at 72°C, followed by a final extension of 5 min at 72°C, in a PTC-200 thermocycler (MJ Research, Watertown, MA). Samples to be analyzed by ethidium bromide staining in agarose gels were amplified for 35 cycles. All others were amplified for 28 cycles. All markers used in the genome-wide scan were part of the Dietrich et al. (1996) murine STR map. No allele size data were available for any markers for the novel APN parental strain. Consequently, it was necessary to screen markers for polymorphism between the APN and C3H/HeJ parental strains. We screened 536 primer pairs to obtain the 170 marker loci used in the present study. Markers were genotyped either by agarose gel electrophoresis and ethidium bromide staining, autoradiographic detection of ³²P-labeled products after 6% polyacrylamide urea gel electrophoresis, or by semiautomated detection and genotyping of fluorescently labeled products with an ABI 310 genetic analyzer and Genotyper software (PE Biosystems, Mississauga, Ontario, Canada). The majority of the markers used in the genomic screen (143/170) were genotyped with the ABI 310 system.

Linkage and Interval Mapping. Genotype data were used to construct a genetic linkage map spanning the genome with MapMaker 3.0b (Lander et al., 1987) software. Calculated marker orders and genetic distances were compared with published data compiled in the mouse genome database (Blake et al., 1998). Interval mapping to generate log-likelihood (LOD) plots for statistically significant associations between genotype and phenotype for each chromosome based on a maximum likelihood algorithm was undertaken with the MapMaker QTL 1.1 program (Paterson et al., 1988). LOD scores are the logarithm of the ratio of the odds of a QTL affecting the phenotype occurring at a given position over the odds of no QTL at that position. LOD plots were produced by calculating LOD scores at 2 centiMorgan (cM) intervals along a linkage group. Calculations were based on genotypic data from 92 animals comprising the 46 most phenotypically extreme F₂ males and females and gender-merged phenotypic data from the entire F₂ generation. As well, gender-specific LOD plots with genotypic data from the 46 phenotypically extreme males or females of the F₂, with their respective log-transformed F₂ phenotypic data were tested. Where statistically suggestive associations were found, the remainder of the F₂ was genotyped for a subset of markers spanning the location of the putative QTL to increase the power of the analysis for that region.

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

Analysis of Paraxanthine/Caffeine Index Data from Crosses. Crosses of the APN and C3H/HeJ parental strains produced a F₁ population with gender-specific paraxanthine/caffeine index means that were significantly different from either of the parental strain gender-specific means (p < .001). The mean for the F₁ did not vary significantly from that of the F₂ among males (p = .083) or females (p = .249). These data are consistent with additive inheritance involving multiple loci. Significant differences between male and female progeny were found in all classes except APN males and females that were not significantly different (p = .509). These data are summarized in Fig. 1.

View larger version (28K): [in this window] [in a new window]   Fig. 1.   Distribution of serum 1,7-dimethylxanthine/caffeine indices among APN, C3H/HeJ, F1, and F2 intercross mice 2 h after oral dosing with 10 mg/kg caffeine. Values to the right of each histogram are means ± S.D and total number of mice in each class.

Genotypic Data. From the 536 STR markers screened, 170 were selected based on their informativeness in the experimental cross, both in terms of polymorphism between the parental strains and their map positions within the genome. A genetic linkage map was constructed that spanned the genome with an average interval between markers of ~13 cM. A higher marker density was achieved in regions of putative QTLs. Linkage groups and most likely marker orders determined with the MapMaker 3.0b program agreed with those published in the latest chromosome committee reports (Blake et al., 1998). Map distances between markers were generally similar to reported values. Two markers, D7 Mit215 and D8 Mit114, were found to be unlinked with respect to the other markers on their assigned chromosomes after duplicate genotyping and were excluded from the interval mapping analysis.

View larger version (14K): [in this window] [in a new window]   Fig. 2.   LOD plots for linkage to the caffeine 3-demethylation phenotype with an unconstrained genetic model. The 1-LOD confidence intervals for the positions of the QTLs, determined for the free genetics model, are indicated by the vertical lines to the right of the chromosomes. The lengths of these intervals, in cM, are indicated. Markers are indicated to the left of the chromosomes, with reported map positions in parentheses. Map intervals between markers, determined in the APN × C3H/Hej F2, are indicated to the right of the chromosomes. A, chromosome 9; B, chromosome 1; C, chromosome 4.

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

The data presented herein demonstrate that the phenotype of paraxanthine/caffeine ratio in serum determined 2 h after oral dosing with caffeine is a polygenic trait in the mouse. Estimation of the number of genetic determinants segregating in an F₂ intercross of APN and C3H/HeJ mice by the method of Wright (1968) indicated that at least three loci contributed to the observed variation in this trait. This method depends on certain assumptions that typically cannot be tested before undertaking the interval mapping analysis. These include the assumptions that each genetic factor is contributing equally to the trait, that each locus is contributing additively (uniform dominance effects across loci) without epistatic interactions, and that each allele is contributing to the trait in the quantitative direction of the parental strains deviation from the F₁ mean. That is, all alleles from the parental strain having the higher phenotypic value contribute additively to increase the phenotypic value. Any violation of these assumptions leads to an underestimation of the number of loci contributing to the trait (Paterson et al., 1988; West et al., 1994). In the present study, the QTLs identified on chromosomes 1 and 9 do not make equal contributions to the phenotypic variation observed, whereas a dominant model for the contribution of the APN allele at the putative QTL on chromosome 4 could not be excluded.

The present study has identified a statistically significant association between the paraxanthine/caffeine index trait and loci on chromosomes 1 and 9, and suggestive linkage to a locus on chromosome 4, according to proposed criteria for the significance of linkage statistics in genomic scans (Lander and Kruglyak, 1995). Collectively, these loci account for 63.2% of the variance observed for this trait in the F₂ (Table 1). The major QTL identified on chromosome 9 accounts for 41.3% of the observed phenotypic variance. The 1-LOD confidence interval for the map location of this QTL encompasses the marker D9 Mit21. This marker is located less than 1 kilobase from the translation start codon of the Cyp1a2 gene. Cyp1a2 is an obvious candidate gene for this QTL, based on its predominant role in caffeine 3-demethylation, and the previously observed differences in CYP1A2 enzyme protein and mRNA expression between the two parental strains (Casley et al., 1997b). No expression polymorphisms of the murine Cyp1a2 gene have been reported. A previously identified allelic variant was not correlated to altered gene expression (Kimura and Nebert, 1986). In humans, a C₂₈₆₆G mutation in exon 2 that caused a Phe ₂₁-Leu₂₁ amino acid change was identified in <1% of Chinese subjects (Huang et al., 1999). The frequency of this mutation in other populations, as well as its functional significance, has yet to be determined. Single nucleotide polymorphisms in the 5'-flanking region (Nakajima et al., 1999) and in intron 1 (Sachse et al., 1999) have recently been reported that may affect inducibility, but not basal expression, of the gene based on caffeine metabolite assays of CYP1A2 activity among smokers and nonsmokers.

The chromosome 1 QTL, that mapped to a 11.1-cM interval on the distal portion of chromosome 1, between markers D1 Mit505 and D1 Mit356, accounted for 16.5% of the observed phenotypic variation. A comparison of LOD maxima for this locus when the effects of the chromosome 9 locus were fixed, did not indicate the presence of epistatic interaction between the two loci. This conclusion was supported by the observation of significant differences in the phenotypic means between double and single homozygotes for markers tightly linked to the predicted positions of both QTLs. This analysis favored a model in which each locus contributed to the phenotype in an additive fashion (Peirce et al., 1998), consistent with allelic differences in enzyme expression, rather than variants of regulatory loci. These results indicate that, although a Cyp1a2-linked locus explained the greatest proportion of the phenotypic variance among identified QTLs, a substantial amount of the variation in the paraxanthine/caffeine index used herein is attributable to a factor other than CYP1A2 activity. No obvious candidate gene that might be involved in the paraxanthine/caffeine phenotype has been mapped to this region of chromosome 1. This suggests that this QTL, which we have tentatively identified as Cafq1, may represent a novel drug metabolism locus. There is also no obvious candidate gene to explain the QTL putatively located on chromosome 4, but the statistical association achieved at this locus is not sufficiently robust to warrant the declaration of a paraxanthine/caffeine index locus at this position without confirmatory data, preferably from a quantitative trait mapping experiment involving different parental strains. Interestingly, the contribution of Cafq1 to the phenotypic variation of the F₂ appears to be sex determined. This locus demonstrated a significant association with the phenotype among males, but achieved only a suggestive association among females (LOD = 6.3 versus 3.6, respectively). The contribution of this locus to the phenotypic variance among males was also much greater than in females (Table 1). There was no such evidence of gender specificity for either the chromosome 9 or chromosome 4 locus. The existence of a gender-specific factor influencing the paraxanthine/caffeine index is consistent with our previously reported observations of a significantly higher value in males compared with females for this trait among a number of inbred mouse strains (Casley et al., 1997b). In humans, gender and oral contraceptive use have been identified as affecting the determination of CYP1A2 activity with urinary metabolite ratios (Kalow and Tang, 1991).

A number of enzyme activities other than CYP1A2 have been identified as being involved in caffeine metabolism in humans. CYP2A6, CYP2E1, and CYP3A4 have all been shown to be involved in the metabolism of this drug (Gu et al., 1992; Tassaneeyakul et al., 1994). In addition, N-acetyltransferase 2 is involved in the production of the urinary metabolites 5-acetylamino-6-formylamino-3-methyluracil and 1,3,7-trimethylurate (Grant et al., 1984). Xanthine oxidase also is involved in the conversion of 1-methylxanthine to 1-methylurate (Grant et al., 1986). Of the candidate genes known to be involved in caffeine metabolism in humans, only CYP1A2 showed linkage to the paraxanthine/caffeine trait among the mouse orthologs (Table 2). These results do not necessarily imply that these activities do not influence the paraxanthine/caffeine index in the mouse. The interval mapping approach can only identify those genes that exhibit polymorphism between the two parental strains. A candidate gene may strongly influence the trait in question, but will not be detected if there is no genetic polymorphism affecting expression between parental strains at that locus. Alternatively, a gene that is differentially expressed might not be detected in this analysis if its contribution to the phenotypic variation is relatively small. For example, CYP2E1 accounts for <5% of the clearance of caffeine in the absence of enzyme induction (Gu et al., 1992). The Ahr locus, which has been shown to affect the basal expression of CYP1A2 as well as its induction by xenobiotics (Fernandez-Salguero et al., 1995), did not show a significant association with the paraxanthine/caffeine trait in this cross (Table 2). Although the Ahr genotype has not been established for the APN strain, we had previously demonstrated that both parental strains were sensitive to CYP1A2 induction by 3-methylcholanthrene (Casley et al., 1997b).

Of the various approaches to assaying CYP1A2 activity in humans using caffeine, the plasma paraxanthine/caffeine ratio has been identified as having among the best correlation to the "gold standard" of caffeine clearance from plasma, and being among the least sensitive to confounding factors such as variable renal clearance rates between individuals or populations (Kalow and Tang, 1993; Fuhr and Rost, 1994; Rostami-Hodjegan et al., 1996). The present study suggests that, in the mouse, at least one genetic factor other than CYP1A2 influences this assay. The possibility of a human ortholog of the gene underlying the Cafq1 QTL should be considered in evaluating paraxanthine/caffeine data for evidence of genetic polymorphism in CYP1A2 expression.

The work presented herein represents, to our knowledge, the first application of the interval mapping approach to characterize polygenic quantitative genetic variation in drug metabolism. The identification of QTLs affecting variations in caffeine metabolism has implications for pharmacogenetic studies involving CYP1A2 expression, as well as the search for novel determinants of xenobiotic metabolism in toxicology.