Evidence for a Gene Influencing Serum Bilirubin on Chromosome 2q Telomere: A Genomewide Scan in the Framingham Study

Abstract

There is an inverse relationship between serum bilirubin concentrations and risk of coronary artery disease. The strength of the association is similar to that of smoking, systolic blood pressure, and HDL cholesterol. We carried out a genomewide scan in a Framingham Heart Study. Our study sample consisted of 330 families with 1,394 sibling pairs, 681 cousin pairs, and 89 avuncular pairs. Using variance-component methods, the heritability was estimated to be 49%±6%, and the genome scan demonstrated significant evidence of linkage of serum bilirubin to chromosome 2q, with a LOD score of 3.8 at location 243 cM. The peak multipoint LOD score is located 1 cM away from the uridine diphosphate glycosyltransferase 1 (UGT1A1) gene. UGT1A1 catalyzes the conjugation of bilirubin with glucuronic acid and thus enhances bilirubin elimination; therefore, it is an important candidate gene for serum bilirubin. Gilbert syndrome, a hyperbilirubinemic syndrome, has a population frequency of 2%–19% and is mainly due to a TA insertion at the promoter region of UGT1A1. Only one other region in the genome produced a multipoint LOD score >1 (LOD = 1.3). Our findings suggest that UGT1A1 may be a major gene controlling serum bilirubin levels in the population.

Recent studies, including a Framingham Offspring Study, have shown that there is an inverse relationship between serum bilirubin concentrations and risk of coronary artery disease (CAD) (Schwertner et al. 1994; Breimer et al. 1995; Hopkins et al. 1996; Levinson 1997; Djousse et al. 2001; Hunt et al. 2001). The strength of the association with CAD was similar to that of smoking, systolic blood pressure, and HDL cholesterol (Schwertner 1998; Schwertner and Fischer 2000). Several studies indicate bilirubin is an effective antioxidant that efficiently scavenges peroxyl radicals; suppresses the oxidation of lipids and lipoproteins, especially low density lipoprotein; and thus acts against plaque formation and subsequent atherosclerosis (Wu et al. 1991, 1996; Miller et al. 1993; Lamont et al. 1997). Thus, there is increasing interest in factors that regulate levels of serum bilirubin.

Serum bilirubin is derived primarily from the degradation of hemoglobin, catalyzed by heme oxygenase, and is transported to the liver by binding to albumin. Within hepatocytes, the solubility of bilirubin is increased by the addition of one or two molecules of glucuronic acid, catalyzed by UGT1A1 (MIM 191740). Bilirubin monoglucuronide and diglucuronide metabolites are then actively transported into the bile.

In humans, the heritability of total serum bilirubin has been estimated to be 58% (Williams et al. 1991). Two segregation analyses in large population-based cohorts have suggested that there is a major gene with the rarer genotype associated with elevated bilirubin levels (Hunt et al. 1996, 2001). Very recently, the authors of a genome scan on serum bilirubin in a white population ascertained by the presence of early CAD reported a locus on chromosome 2q telomere, 2q34–37, that influences variation in serum bilirubin levels (Kronenberg et al. 2002). Here, we report the results of a serum bilirubin genome scan in unselected families, the Framingham Heart Study.

The Framingham Heart study is a population-based cohort study. In brief, the Framingham Heart Study began in 1948 with the recruitment of 5,209 residents from Framingham, MA. The subject individuals have undergone biennial examinations since the study began. In 1971, the Framingham Offspring Study (Kannel et al. 1979) was initiated. In total, there were 5,124 subjects recruited. The offspring subjects have been examined every 4 years (except the first two examinations, which have 8 years intervening). The current analysis consists of 1,702 genotyped subject individuals from the 330 largest extended families within the study.

Total serum bilirubin was measured during the first and second examinations of the offspring, using the ultramicromethod (Walters and Gerarde 1970). Our analysis was restricted to bilirubin measurements in the first examination, since far fewer individuals had serum bilirubin measured in the second examination. The correlation coefficient between the two examinations is 0.6, P<.01.

Genomic DNA was isolated from nucleated blood cells. DNA samples were sent to the Marshfield Mammalian Genotyping Service. At an average of 10 cM density, 399 microsatellite markers (Screening Set 8) (Yuan et al. 1997) covered the genome with an average marker heterozygosity of 0.77.

Variation in serum bilirubin due to known factors was tested and removed, using SOLAR, to enhance the ability of linkage analysis to detect genetically determined variation (Almasy and Blangero 1998). The covariates selected were all statistically significantly associated in a regression model and were incorporated in both the heritability estimation and the linkage analysis. These were sex, height, weight, total cholesterol, hematocrit, albumin, SGOT, smoking (no. of cigarettes/d), and alcohol intake (oz./wk). At first, BMI (instead of height and weight) was tested but was not significant in the multiple-regression model. When BMI was replaced by height and weight, both variables were highly significant. In essence, SOLAR uses the residuals for heritability estimation and two-point and multipoint linkage analysis. Further analyses were conducted using normalized deviates. SOLAR makes use of all information in pedigrees of any size and structure. Residual heritability is the proportion of total phenotypic variation due to additive genetic effects, after removing the variation attributable to covariates. Marker allele frequencies were estimated from the study participants and then used to estimate the proportion of alleles shared identical by descent (IBD) among all relative pairs. SOLAR uses a likelihood-ratio test to evaluate linkage by comparing a purely polygenic model (without consideration of genetic marker information) with a model that incorporates marker information at a specific locus (two-point analysis) or several nearby markers in a chromosome (multipoint analysis). We also repeated the variance-components linkage analysis using the normalized deviates, calculated from ranked standardized residuals in regression models conducted in SAS.

Since this method is based on the assumption of a multivariate normal distribution, violations of the assumption may result in inflated type I error rates (Allison et al. 1999; Blangero et al. 2000, 2001). To ensure a more accurate result, we used an empirical P value method implemented in SOLAR for LOD score adjustment (Blangero et al. 2000). A fully informative marker, which was not linked to the trait studied, was simulated. The IBD information for this marker was calculated, and then linkage analysis was performed. The empirical P values for the trait were obtained on the basis of the LOD score distribution generated from 10,000 replicates. This method provides robust LOD scores for data with nonnormal distributions. As a second approach to account for the nonnormality in these data, the analysis of normalized deviates was also pursued.

There were 1,702 individuals in the 330 largest extended families who were genotyped. Among them, 394 individuals belonged to the original cohort and did not have serum bilirubin measured. The total number of individuals with serum bilirubin measured was 1,612. The mean values of the phenotypes of the 1,612 (50% male) individuals are displayed in table 1. The total number of individuals with serum bilirubin measured and genotyped was 1,264, including 1,394 sibling pairs, 681 cousin pairs, and 89 avuncular pairs.

Table 1.

Characteristics of the 1,612 Framingham Offspring with Bilirubin Measured

	Value in Subjects
	Male		Female
Variable	Mean ± SD	Range	Mean ± SD	Range
Total bilirubin (mg/dl)	8.8 ± 3.8	2–35	7.1 ± 2.9	2–27
Age (years)	32.3 ± 10.4	11–64	33.3 ± 10.5	10–62
Height (inches)	68.9 ± 3.3	53–78	63.6 ± 2.7	52–71
Weight (pounds)	176.1 ± 29.4	54–264	137.1 ± 27.8	67–281
Total cholesterol (mg/dl)	193.5 ± 38.6	96–138	188.4 ± 37.9	100–346
Hematocrit (%)	45.5 ± 2.9	34–62	40.6 ± 2.8	28–52
Albumin (mg/dl)	48.1 ± 3.1	32–57	45.8 ± 3.3	26–59
SGOTa (mg/dl)	47.9 ± 20.0	18–221	40.4 ± 17.7	8–285
Smoking (no./d)	15.0 ± 16.4	0–90	11.0 ± 14.1	0–88
Alcohol (oz./wk)	4.9 ± 5.9	0–43	2.2 ± 3.0	0–34

^aLiver serum glutamic oxaloacetic transaminase.

The distribution of serum bilirubin was not normal, P<.01, and the estimates of skewness and kurtosis were 2.41±0.03 and 9.50±0.07, respectively, without covariates adjustment.

The heritability estimate from the serum bilirubin, after adjusting for covariates, was 49%±6%. The proportion of variance due to all final covariates is ∼11%.

For two-point linkage analysis, LOD scores of 1 or higher were observed for markers on chromosomes 4, 5, and 7 and for three adjacent markers on chromosome 2 (table 2).

Table 2.

Markers with Two-Point LOD Score of 1 or Higher

Marker	Chromosome	Location(cM)	LOD Score
D2S1363	2	227	1.7
D2S427	2	237	1.7
112yd2	2	261	1.5
D4S1647	4	105	1.2
D5S820	5	160	1.1
GATA118G10	7	79	1.4

The highest multipoint LOD scores for each of the 22 autosomes are presented in figure 1. For multipoint analysis, a maximum LOD score of 3.8 (P=.00001) was observed on the chromosome 2q telomere, with the peak location in 243 cM (fig. 2) with comprehensive human genetic maps (Broman et al. 1998). Using normalized deviates also produced very similar results, with a maximum LOD score of 3.9 located in 244 cM. We also carried out a regular LOD score analysis without adjusting for nonnormality; a maximum LOD score of 5.7 was obtained at the same location, 2q 243 cM. Except for this, there was only one region on chromosome 4 (96 cM) in the genome that had a maximum multipoint LOD score ⩾1 (LOD=1.3).

Figure 1.

Maximum multipoint LOD scores for serum bilirubin for the 22 autosomes. X-axis values are the number of 22 autosomes. Y-axis values are maximum multipoint LOD scores. Linkage analyses were conducted using multivariable residuals, adjusted for sex, height, weight, total cholesterol, hematocrit, albumin, SGOT, smoking, and alcohol intake.

Figure 2.

Multipoint LOD scores for serum bilirubin on chromosome 2. X-axis values are cM. Y-axis values are multipoint LOD scores. Linkage analyses were conducted using multivariable residuals, adjusted for sex, height, weight, total cholesterol, hematocrit, albumin, SGOT, smoking, and alcohol intake.

Peak evidence for linkage on chromosome 2 occurred in the region bounded by markers D2S427 and GATA178G09. The LOD-one supporting interval (the region corresponding to maximum LOD score minus 1) for the location of serum bilirubin QTL spans a 22-cM interval flanked by markers D2S1363 and 112yd4, which are 34 cM apart. The results provided substantial evidence for a serum bilirubin QTL on 2q telomere.

Our results are consistent with a recently reported genome scan, conducted in a population enriched for early-onset cardiovascular disease, in which a locus with a multipoint LOD score of 3.01 was identified in this region (Kronenberg et al. 2002). Thus, our finding complement and extend the provocative findings of Kronenberg et al., conducted in a subset of families selected for disease. Kronenberg et al. also provided evidence for loci on chromosome 9, 10, and 18 that may possibly influence serum bilirubin levels. None of the loci on chromosome 9, 10, and 18 overlap with our findings. In addition to coming from a different study population, the different results may be due to differences in linkage analysis methods. Our study was conducted using empirical P value and normalized deviate methods to adjust LOD score for nonnormality of data, whereas Kronenberg et al. used both parametric and nonparametric linkage methods without adjustment for nonnormality. Without adjustment for nonnormality, our results showed LOD=5.7 at 2q, 243 cM.

Within the chromosome region of 2q telomere lies an important candidate gene, UGT1A1. This gene is ∼5 Mb (∼5 cM) telomeric from marker D2S427. The peak maximum LOD score in this area is located 6 cM telomeric from marker D2S427. Therefore, UGT1A1 and the peak maximum LOD score almost overlap. To date, no other genes known to regulate serum bilirubin metabolism have been found in this region.

Uridine diphosphoglucuronate glucuronosyltransferases (UGTs) are a family of enzymes that conjugate various substrates with glucuronic acid and enhance their elimination through the bile. UGT1A1 is the physiologically relevant isoform and significantly contributes to bilirubin glucuronidation. Mutations in UGT1A1 lead to complete or partial inactivation of the enzyme, causing three nonhemolytic unconjugated hyperbilirubinemia syndromes, type I and type II Crigler-Najjar syndromes and Gilbert syndrome. Crigler-Najjar syndromes are rare and lethal forms of autosomal inherited diseases, with a complete lack of bilirubin UGT activity (type I) or a severe deficiency of the enzyme (type II). Type I and type II Crigler-Najjar syndromes are mainly associated with missense mutations in the coding region of the UGT1A1 gene and have a frequency of 1 in 1 million births. Gilbert syndrome, a more prevalent (2%–19% in population studies) and essentially benign condition, is principally caused by a TA insertion at the TATAA promoter region upstream of the UGT1A1 exon. The gene frequency of this mutation among the white population is extraordinarily high, 0.34–0.40 (Bosma et al. 1995; Monaghan et al. 1996; Beutler et al. 1998; Morsche et al. 2001) and is very close to the estimated major gene allele frequency from the two segregation analyses (Hunt et al. 1996, 2001).

Combining the results of previous association and segregation studies, the results of this genome scan, and the physiological function of the gene, we speculate that UGT1A1 may be a major gene controlling serum bilirubin levels.

Two segregation studies indicated that a major gene may protect ∼12% of the population against CAD, and most individuals with CAD lacked the protective effect from this gene associated with bilirubin levels (Hunt et al. 1996, 2001). If ⩾12% of the population have a strong protective effect against CAD owing to the UGT1A1 gene, this would have significant population implications in future CAD prevention. Given the substantial magnitude of variability in bilirubin explained by the chromosome 2 region of linkage, our data support further study of the role of the UGT1A1 gene variants in bilirubin determination and in risk for CAD and related cardiovascular diseases.

In conclusion, our results provide evidence that a major gene with a large effect influencing bilirubin levels exists on chromosome 2q telomere. We have identified a chromosomal region implicated in several Mendelian hyperbilirubinemic disorders, and within this region resides an important candidate gene, UGT1A1. Further research is warranted to examine the linkage between the UGT 1A1 gene and serum bilirubin levels, as well as association studies to examine for associations of polymorphisms in the UGT1A1 gene with serum bilirubin and CAD.