Gilbert's disease and atazanavir: From phenotype to UDP-glucuronosyltransferase haplotype


  • Potential conflict of interest: Nothing to report.


Gilbert's disease leads to intermittent non-hemolytic hyperbilirubinemia by a reduction of hepatic bilirubin glucuronidation associated with the presence of the UDP-glucuronosyltransferase (UGT) 1A1*28 polymorphism. It is considered benign because it does not result in hepatocellular damage. However, pharmacogenetic analyses have linked UGT1A1*28 to drug toxicity and cancer predisposition. The protease inhibitor atazanavir (ATV) is an inhibitor of hepatic UGT activity leading to hyperbilirubinemia in individual patients. Whether this is linked specifically to UGT1A1*28 or to more complex variants influencing glucuronidation is unclear. One hundred and six ATV-treated patients were characterized and genotyped for UGT1A1*28, the UGT1A3 (-66C) and UGT1A7 (-57G) promoter variants, and UGT1A7129K/131K. ATV treatment increased median bilirubin levels from 10 to 41 μmol/L (P = .001) with hyperbilirubinemia exceeding 43 μmol/L in 37%. Hyperbilirubinemia over 43 μmol/L was significantly associated not only with UGT1A1*28 but also with UGT1A3-66C, UGT1A7-57G, and UGT1A7129K/131K, although these variants do not naturally occur in linkage dysequilibrium in blood donors. Homozygous combinations of UGT1A1*28 with the other variants increased from 7.4% (normal bilirubin to 42 μmol/L) to 41% to 46.1% (43 to >85 μmol/L), and 100% (>85 μmol/L). All six patients with hyperbilirubinemia greater than 85 μmol/L were homozygous for all four variants identifying a haplotype inherited on a single allele. In conclusion, the genetic variant associated with Gilbert's disease is identified as part of a haplotype of four UGT1A variants spanning three genes at the UGT1A gene locus. This haplotype predisposes to hyperbilirubinemia in ATV treatment and may have an additional role as a pharmacogenomic risk factor for drug therapy. (HEPATOLOGY 2006;44:1324–1332.)

Jaundice represents one of the most striking clinical symptoms. It is caused by a variety of hepatic and non-hepatic conditions, including hepatocellular or cholangiocellular toxicity, infection, ischemia, biliary obstruction, hemolysis, as well as genetic diseases that determine hepatic metabolism and transport. Among the genetic diseases, Gilbert's syndrome1 is the most common inheritable condition leading to transient unconjugated hyperbilirubinemia.2 It is considered benign because it does not lead to chronic hepatic destruction. The underlying cause is linked to a TA insertion into the promoter region of the hepatic UDP-glucuronosyltransferase (UGT) 1A1 termed UGT1A1*28.3, 4 The UGT1A1*28 promoter variant leads to a 70% reduction of UGT1A1 transcription and to reduced bilirubin glucuronidation, for which UGT1A1 represents the only efficient UGT isoform.5 The complete loss of UGT1A1 function is fatal. It is present in Crigler-Najjar type 1 disease, usually requiring liver transplantation as an option for a genetic cure.6 Although benign in nature from a hepatological point of view, Gilbert's disease has been identified as a risk factor for the development of cancer7 based on the ability of UGT1A1 to detoxify mutagenic arylamines8 and benzo(α)pyrenes.9 In addition, UGT1A1*28 has been identified as a risk factor for unwanted drug side effects exemplified by the possibility of severe toxicity of the anti-cancer drug irinotecan.10–13 Gilbert's disease can therefore be regarded as a sentinal phenotype pointing to a pharmacogenetic risk constellation of UGT1A1*28-carrying individuals.13

UGT1A1 is a member of the UGT1A family of transferases encoded on human chromosome 2. The human UGT1A gene locus encodes nine functional isoforms with overlapping as well as specific substrate activities.14 The human UGT1A genes are regulated in a tissue-specific fashion, which is considered to represent the biochemical basis of organ-specific glucuronidation activity.15–17 Glucuronidation is further influenced by induction,18, 19 phosphorylation,20 the presence of genetic variants in the UGT coding regions leading to catalytically altered proteins,21, 22 by variants of the promoter regions leading to reduced UGT transcription,4, 13, 23 as well as inhibitory effects of therapeutic drugs. Commonly administered drugs with inhibitory potential include amitriptyline,24 protease inhibitors,25 ketoconazole,26 and acyl glucuronide adducts of ketoprofen.27, 28 Therefore, the combination of genetic variants in addition to environmental or therapeutic xenobiotic exposure defines glucuronidation activity and thus the potential risk profile of an individual.

The genetic background in Gilbert's disease may not be as simple as suggested by the association of unconjugated hyperbilirubinemia with UGT1A1*28. Recent data have identified single nucleotide polymorphisms (SNPs) at the human UGT1A gene locus in high numbers and high frequencies of up to 50%, which suggests that UGT variant combinations are not rare, and that combinations are likely to influence an individual's response to drug therapy.13, 21 In patients with Gilbert's disease simultaneously present SNPs of UGT1A6 and UGT1A7 have been reported.13, 29, 30 From these data we hypothesize that a UGT1A haplotype combining several genetic UGT variants may account for the susceptibility to side effects of drug therapy, which has not been conclusively demonstrated.

An interesting model to study this hypothesis is the observation of hyperbilirubinemia in patients treated with anti-retroviral protease inhibitors, such as atazanavir or indinavir.31 In many instances drug reactions occur when a single enzyme/pathway is the primary route of elimination of a drug. Recently, this has been prominently recognized with cerivastatin and gemfibrozil. Glucuronidated gemfibrozil is a potent inhibitor of cytochrome P450 (CYP) 2C8, which in turn inhibits CYP2C8 catalyzed cerivastatin metabolism leading to toxicity.32 In the case of atazanavir, the compound does not undergo significant glucuronidation but is capable of inhibiting UGT proteins, which include UGT1A1, UGT1A3, and UGT1A4.25 This offers the possibility to study the outcome of two potential mechanisms: genetic alteration of catalytic UGT activity and pharmacological UGT inhibition. Even in the presence of the UGT1A1*28 variant, atazanavir-treated patients develop variable degrees of hyperbilirubinemia. In this study we analyzed variants of the UGT1A3 and UGT1A7 genes in addition to UGT1A1*28 in atazanavir-treated patients. Our data demonstrate that ATV treatment side effects are associated with the presence of a haplotype of four functional UGT variants spanning three UGT1A genes.


SNP, single nucleotide polymorphism; UDP, Uridine diphosphate; UGT, UDP-glucuronosyltransferase; ATV, atazanavir; ULN, upper limit of normal; HIV, human immunodeficiency virus.

Patients and Methods

Study Population

HIV-Positive Patients.

A total of 106 patients were recruited from the outpatient clinic of the Department of Clinical Immunology, Hannover Medical School, Germany, from March 2003 to April 2005. Blood samples were collected from antiretroviral-treated and experienced patients receiving 300 mg protease-inhibitor atazanavir (ATV) and 100 mg Ritonavir once daily for at least 1 month. Total bilirubin levels were measured before and 30 days after the beginning of treatment. Hyperbilirubinemia was graded in accordance to the Division of AIDS table for grading the severity of adverse events (total bilirubin levels): grade 0 (normal), <19 μmol/L; grade 1 (mild), 19 to 26 μmol/L [1.1–1.5 × upper limit of normal (ULN)]; grade 2 (moderate), 26 to 43 μmol/L (1.6–2.5 × ULN); grade 3 (severe), 43 to 85 μmol/L (2.6–5.0 × ULN); grade 4 (serious), ≥ 85 μmol/L (≥ 5.0 × ULN). Jaundice was defined as a total bilirubin level greater than 43 μmol/L.

Healthy Blood Donors.

Blood samples were obtained from a total of 427 anonymous healthy blood donors from the Department of Transfusion Medicine/Blood Bank of Hannover Medical School, Germany.

Informed consent was obtained from all patients, and the study was approved by the Ethics Committee of Hannover Medical School.


Genomic DNA.

Genomic DNA was isolated from full blood samples by the NucleoSpin Blood XL Kit according to the recommendations of the manufacturer (Machery & Nagel, Dueren, Germany).

Allelic Discrimination Genotyping for UGT1A-57T/G, UGT1A3 -66 T/C, UGT1A7 N129K/R131K.

Approximately 10 ng genomic DNA was used as a template in Taqman 5′-nuclease assays. Primers and Probes specific for each SNP were designed with Primer Express software (Applied Biosystems, Foster City, CA) and labeled with either 6-FAM or VIC as reporter dyes and MGB-NFQ (Applied Biosystems) as a quencher (all sequences summarized in Table 1) as described previously.13 The Taqman assays were performed using 600 nmol/L primer concentrations and 200 nmol/L probe concentrations (Applied Biosystems) and qPCR Mastermix Plus (Eurogentec, Seraing, Belgium). The run consisted of a hot start at 95°C for 10 minutes and 35 cycles of 94°C for 15 seconds and 61°C for 1 minutes. All assays were performed in 25-μL reactions in 96-well trays using an ABI 7000 instrument (Applied Biosystems). UGT1A3 and UGT1A7 genotyping was performed on all 104 human immunodeficiency virus (HIV) samples, and on 427 (UGT1A7) as well as 322 (UGT1A3) healthy blood donors.

Table 1. Primer/Taqman Probes Sequences Used for Genotyping
 UGT1A7–57 T>G
 UGT1A7 N129K R131K
Reverse primer5′-AGG ATC GAG AAA CAC TGC ATC A-3′
 UGT1A3 - 66 T/C
Forward primer5′- CGT TGA TTT GCT AAG TGG CTC AG -3′
Reverse primer5′- GCC ATC TCA GCA GAA GAC ACG-3′
Probe variant6-VIC- TTA AGA CGA AGA AAG C -MGB

Real-Time Polymerase Chain Reaction and Melting Curve Analysis for Genotyping of the UGT1A1*28 Promoter Polymorphism.

Real-time polymerase chain reaction (PCR) technology was set up for the detection of TATA-box variants in the promoter region of the UGT1A1 gene.33 The reaction mixture contained 5 μL genomic DNA, 1 μL forward primer (5′-GTCACGTGACACAGTCAAAC-3′, 10 pmol/μL), 1 μL reverse primer (5′-CAGCATGGGACACCACTG-3′, 10 pmol/μL), 2 μL FRET probes (4 pmol/μL) (5′-GCCATATATATATATATATAA-FITC-3′, 5′-Cy5.5-AGGGC- GAACCTCTGGCAGGA-Ph-3′; MWG), 2 μL Fast Start DNA Master Hybridization Mix, 1.6 μL MgCl2 (25 mmol/L) and 1 μL DMSO (Roche Diagnostics, Mannheim, Germany). The mixture was adjusted to 20 μL using DEPC-ddH2O. For denaturation, a temperature of 95°C was maintained for 30 seconds. Amplification was performed at 95/56/72°C for 20/13/10 seconds (50 cycles, slope 20°C/s) without color compensation. After amplification a final melting curve was recorded by heating to 95°C once and then cooling to 31°C at 20°C/s, followed by a 60-second hold before heating slowly at 0.2°C/s up to 50°C. Fluorescence was measured continuously during the slow temperature rise to monitor the dissociation of the fluorescein isothiocyanate–labeled probe from the amplicon. The fluorescence signal (F) was plotted in real time against the temperature (T) to produce melting curves for each sample (F versus T). Melting curves were converted to melting peaks by plotting the negative derivative of F with respect to T against T (-dF/dT vs T) using version 3.5 of the LightCycler software. Individuals with an homzygous 6/6 UGT1A1 genotype had a single mealting peak with Tm = 38.6 °C ± 0.3 °C, those with Gilbert′s syndrome (7/7 UGT1A1) had a single mealting peak with Tm = 45.2 °C ± 0.3 °C. UGT1A1 genotyping was performed on all 106 HIV samples and 104 healthy blood donors. In addition, UGT1A1*28 genotyping was performed on a historic cohort of 52 healthy blood donors in whom sex was known so as to analyze gender dependency.

Statistical Analyses.

The analyses of allelic variants were conducted using the two-tailed Fisher's exact test or chi-squared test. Statistical analyses were conducted using SPSS (SPSS; 2000: SigmaPlot for Windows. Ver. 13, SPSS, Chicago, IL) or Epicalc (version 5.0/2000 for windows,


Study Population

The 106 patients infected with the human immunodeficiency virus (HIV) were predominantly male (83%) and represented a homogeneous Northern German Caucasoid cohort with only five patients of African and three patients of Asian descent (Table 2). The analysis of viral coinfection performed on all subjects identified five patients with hepatitis B (HBV) virus infection who were hepatitis B surface antigen positive and 11 patients with hepatitis C virus infection who were anti–hepatitis C virus antibody positive. Biochemical workup of all patients included bilirubin, alanine aminotransferase, aspartate aminotransferase, as well as gamma glutamyl transferase (γGT) determinations. Although mean aspartate aminotransferase, alanine aminotransferase, and γGT levels did not differ before and 30 days after the administration of ATV therapy, a significant increase in mean bilirubin levels and bilirubin range was observed. These data clarify that ATV therapy did not lead to hepatic toxicity indicated by aminotransferase elevations but rather resulted in a significantly higher rate of hyperbilirubinemia. Among the ATV-treated patients, 16 (15%) had normal bilirubin levels (<19 μmol/L), 12 (11%) had grade 1 (mild 19-26 μmol/L), 39 (37%) had grade 2 (moderate 26–43 μmol/L), 33 (31%) had grade 3 (severe 43–85 μmol/L), and 6 (6%) had grade 4 (serious ≥85 μmol/L). Jaundice (hyperbilirubinemia >43 μmol/L) was present in 39 (37%) of atazanavir-treated patients. All data are summarized in Tables 2 and 3. Ethical protocol at Hannover Medical School required the anonymous genetic analysis of the blood donor group of 427 subjects and therefore comparative data are not given in Table 2. The blood donor group consisted of 427 consecutive blood donors of the blood bank of Hannover Medical School who were recruited from the surrounding county of Hannover and also represent a Northern German Caucasoid population. To document possible gender differences a historic cohort of 52 healthy blood donors was additionally genotyped for UGT1A1*28.

Table 2. Characteristics of Atazanavir-Treated Patients
  • Abbreviations: HIV, human immunodeficiency virus; HbsAg, hepatitis B virus surface antigen; HCV, hepatitis C virus; ATV, atazanavir; AST, aspartate aminotransferase; ALT, alanine aminotransferase.

  • *

    P = .001, other comparisons were not significant.

  • Before and after ATV denotes analyses before and 30 days after initiation of atazanavir treatment.

HIV-positive patients106
Median age (range) [years]45 (29–76)
Male88 (83%)
Female18 (17%)
Caucasian white98 (92.5%)
African5 (4.7%)
Asian3 (2.8%)
HBsAg positive5 (4.7%)
Anti-HCV positive11 (10.4%)
 Before ATVAfter ATV
Median total bilirubin [μmol/L] (range)10 (3–43)*41 (4–118)*
Median AST [IU/mL] (range)46 (10–488)45 (15–241)
Median ALT [IU/mL] (range)45 (9–176)48 (9–317)
Gamma glutamyltransferase [IU/mL] (range)99 (12–686)83 (11–526)
Table 3. Genotyping Results of All 106 Atazanavir-Treated Patients and Healthy Blood Donors
ATV no.Bilirubin [grade]UGT1A7– 57UGT1A7 129K/131KUGT1A3 - 66UGT1A1*28
  • NOTE. The overall genotyping analysis shows that the frequency of UGT polymorphisms increased with the grade of hyperbilirubinemic phenotype under atazanavir (ATV) therapy. −/−, wildtype; −/+, heterozygous; +/+, homozygous; *, in the grade 0 and 1 group one patient was found with the rare genotype (TA)5TAA/(TA)7TAA.

  • n = 427;

  • n=322;

  • +

    n= 104.

16 (15%)07 (44%)7 (44%)2 (13%)3 (19%)6 (38%)7 (44%)5 (31%)6 (38%)5 (31%)9* (56%)6 (38%)1 (6%)
12 (11%)15 (42%)3 (25%)4 (33%)3 (25%)4 (33%)5 (42%)4 (33%)3 (25%)5 (42%)5* (42%)5 (42%)2 (16%)
39 (37%)213 (33%)20 (51%)6 (15%)5 (13%)15 (38%)19 (49%)10 (26%)17 (44%)12 (31%)17 (44%)20 (51%)2 (5%)
33 (31%)38 (24%)15 (45%)10 (30%)016 (48%)17 (52%)6 (18%)12 (36%)15 (45%)9 (27%)12 (36%)12 (36%)
6 (6%)4006 (100%)006 (100%)006 (100%)006 (100%)
Totaln = 106 33 (31%)45 (42%)28 (27%)11 (10%)41 (39%)54 (51%)25 (24%)38 (36%)43 (40%)40* (38%)43 (40%)23 (22%)
Blood donors0160 (37%)203 (48%)64 (15%)64 (15%)194 (45%)169 (40%)98 (30%)161 (50%)63 (20%)44+ (42%)49+ (47%)11+ (11%)

Frequency of UGT Variants in HIV Patients and in Blood Donors.

A genotyping analysis was performed (Table 3) that included 4 UGT variants: the UGT1A1*28 promoter polymorphism associated with Gilbert's disease, a previously described promoter variant of the UGT1A7 gene (−57 T>G) associated with changes in irinotecan metabolism,13 coding variants at amino acid position 129/131 of the first exon of UGT1A7,22 as well as a recently detected single-nucleotide polymorphism in the promoter of the UGT1A3 gene (−66 UGT1A3)34 (Fig. 1A).

Figure 1.

(A) Schematic representation of the human UGT1A gene locus and the location of the single nucleotide polymorphisms (SNP) analyzed in this study (not drawn to scale). The first exons are aligned in series at the 5′ end of the UGT1A gene locus on chromosome 2. They are numbered sequentially from 3′ to 5′. Each first exon individually initiates transcription to form a transcript sharing exon 2–5 sequences at its 3′ end. These are identical in every UGT1A mRNA and encode the identical carboxyl terminus of each resulting UGT1A protein. Acceptor and donor splice sites at the 3′ end of first exon cassettes and at the 5′ end of exon 2 lead to splicing and elimination of sequence that lies in between to form individual UGT1A transcripts.14 (B) Graphic representation of the allelic frequencies of UGT1A1, UGT1A3, and UGT1A7 polymorphisms in atazanavir-treated patients (white bars) and normal blood donors (black bars). In atazanavir-treated patients, the allelic frequencies of UGT1A3 −66T>C, UGT1A7 −57T>G, and UGT1A7129K/131K but not of UGT1A1*28 were significantly higher. ATV, atazanavir, *P < .05; **P < .001; NS, not statistically significant. (C) Graphic representation of the percentage of homozygous carriers of UGT polymorphisms in the two groups. Homozygous variants were significantly associated with the atazanavir group. ATV, atazanavir; *P < .05; ***P < .01 ; P < .0001.


Several studies have demonstrated that the allelic frequency of UGT1A1*28 in Europeans is approximately 0.4, with homozygosity amounting to approximately 10%.2 HIV-infected patients had a higher allelic frequency of 0.43, which was not significantly different from blood donors (0.34). Homozygotes were more prevalent in this group (22%) (Fig. 1B–C). To address the issue of sex differences, 52 (39 men, 13 women) healthy blood donors were genotyped. Males showed wild-type, heterozygosity, and homozygosity in 41%, 46%, and 13%; females in 46%, 46%, and 8%, respectively, which did not exhibit significant sex differences. Male ATV-treated patients showed wild-type, heterozygosity, and homozygosity in 39%, 39%, and 22%, and women in 33%, 50%, and 17%, respectively, also not demonstrating sex differences. The high rate of homozygous carriers of UGT1A1*28 in the ATV-treated group is therefore not likely to result from a sex bias.

UGT1A7N129K/R131K and UGT1A7 –57 T>G.

Genetic analyses of UGT1A7 gene variants have demonstrated that −57 T>G and an SNP at position 208 (W208R) are always simultaneously present (linkage dysequilibrium),22 which is not observed with N129K/R131K and −57 T>G.13 Although UGT1A7N129K/R131K has a high allelic frequency of 0.62, −57 T>G is present in 0.39.13 In blood donors UGT1A7 −57 T>G and UGT1A7 N129K/R131K were detected at rates in agreement with previously published data. In contrast, ATV-treated HIV patients had significantly higher allelic frequencies of both variants (0.48 and 0.7) as well as higher rates of homozygotes.

UGT1A3 −66T>C.

The UGT1A3 −66T>C variant was significantly more frequent in the patient group (0.58) compared with blood donors (0.45). Individuals homozygous for UGT1A3 −66C were more prevalent in the treatment group.

These data indicate that polymorphisms of UGT1A1 but also of UGT1A3 and UGT1A7 are more frequent in ATV-treated HIV patients (Fig. 1B–C).

UGT Variants Are Associated With Hyperbilirubinemia of Atazanavir-Treated Patients

Stratification of ATV-treated patients according to the grade of hyperbilirubinemia demonstrated that the frequency of UGT variants among patients with grade 3 and 4 hyperbilirubinemia was higher than in patients showing only grade 0 to 2 (Table 3). Most strikingly, all six patients with grade 4 hyperbilirubinemia were homozygous not only for the UGT1A1*28 variant of Gilbert's disease but also for the other three tested genetic markers. Patients with normal to moderate hyperbilirubinemia were found not to differ significantly from the allelic frequencies observed in healthy blood donors. In contrast, patients displaying the hyperbilirubinemic phenotype were distinguishable by a high frequency of UGT polymorphisms. Stratification identified the presence of four different UGT variants as risk factors for ATV-associated hyperbilirubinemia.

Hyperbilirubinemia Is Associated With a UGT1A1-UGT1A3-UGT1A7 Haplotype

An association of hyperbilirubinemia with UGT1A1*28 is an expected finding that has previously been reported. However, the simultaneous association of a drug reaction with four single-nucleotide polymorphisms at the human UGT1A gene locus that are not in linkage dysequilibrium represents a significant observation that suggests UGT variant combinations as a pharmacogenomic risk factor. Table 4 demonstrates the combination of UGT1A1*28 with UGT1A7 −57G, UGT1A3 −66C as well as with UGT1A7 129K/131K, respectively. The combination of homozygosity for these four variants indicates that in the affected individuals all variants lie on a single allele. Patients with bilirubin levels grades 0 to 2 and healthy blood donors exhibit a comparable number of homozygous marker combinations of approximately 10%. This is dramatically different when patients with grade 3 and 4 hyperbilirubinemia are studied, showing a fourfold higher number of homozygotes for two markers, and is finally increased to 10-fold in the six patients with serious hyperbilirubinemia. Although UGT1A1*28 may be a risk factor for hyperbilirubinemia in ATV treatment, a UGT variant haplotype spanning the three studied UGT1A genes seems more probable. This is further substantiated by the number of individuals carrying UGT1A1*28, UGT1A7 −57G, UGT1A3 −66C, and UGT1A7 129K/131K on a single allele (i.e., homozygotes) (Fig. 2). In healthy blood donors, the homozygous haplotype of four variants is present in 9.6%, in all ATV-treated HIV patients in 19.8%, in grade 3 and 4 hyperbilirubinemia patients in 41.2%, and finally in patients with grade 4 hyperbilirubinemia in 100%. ATV levels were measured in 15 patients and showed an average level of 1,339 ng/mL (368–3,710) but did not correlate with hyperbilirubinemia (data not shown). In addition, when baseline bilirubin levels are studied before ATV therapy no obvious association with the UGT1A haplotype is seen, indicating that jaundice was not present before drug therapy (Table 5). Combined, these data identify a UGT1A gene haplotype as pharmacogenomic risk factor in hyperbilirubinemia associated with ATV treatment.

Table 4. Concurrent Identification of Homozygous UGT Polymorphisms in Atazanavir-Treated Patients Stratified According to Grade of Hyperbilirubinemia* and in Healthy Blood Donors
 ATV-Treated PatientsHealthy Blood Donors
  • NOTE. With increasing grade of hyperbilirubinemia (defined in Patients and Methods), the number of patients who are homozygous for two UGT polymorphisms increases. The rate in healthy blood donors is similar to patients with hyperbilirubinemia grade 0–2. In grade 4 hyperbilirubinemia, all six patients were homozygous for each combination. ATV, atazanavir; (+/+), homozygous.

  • *

    See definition in Patients and Methods.

HyperbilirubinemiaNormal-moderate (grade 0–2)Severe-serious (grade 3–4)Serious (grade 4)(grade 0)
UGT1A1*28 (+/+)UGT1A757G (+/+)10/104 (9.6%)
 5/67 (7.4%)16/39 (41%)6/6 (100%) 
UGT1A1*28 (+/+)UGT1A366C (+/+)11/104 (10.7%)
 5/67 (7.4%)18/39 (46.1%)6/6 (100%) 
UGT1A1*28 (+/+)UGT1A7129K/131K(+/+)10/104 (9.6%)
 5/67 (7.4%)18/39 (46.1%)6/6 (100%) 
Figure 2.

Prevalence of a homozygous UGT haplotype consisting of UGT1A1*28, UGT1A3 −66C, UGT1A7 −57G, UGT1A7 129K/131K in patients with different degrees of hyperbilirubinemia under atazanavir (ATV) treatment and in blood donors. The clinical phenotype of severe and serious hyperbilirubinemia in ATV treatment is significantly associated with the four single nucleotide polymorphisms analyzed in this study. The finding of high rates of homozygosity indicates that this haplotype is inheritable on a single allele and represent a pharmacogenetic risk factor for drug toxicity. *P < .05; **P < .0001.

Table 5. Baseline Bilirubin Levels (Before ATV Therapy) in Patients With the UGT Haplotype and UGT1A1*28
 UGT1A1-UGT1A3-UGT1A7 HaplotypeUGT1A1-UGT1A3-UGT1A7 Haplotype (Grade 4 Hyperbilirubinemia)UGT1A1*28
  1. −/−, wildtype; +/+, homozygous variants; all comparisons were not significant.

No. of patients1021-64023
Median bilirubin level [μmol/L] (range)8.8 (4–15)10.0 (4–22)-10.7 (4–15)10.3 (3–43)10.1 (4–22)


In clinical hepatology, the concept of Gilbert's disease is that of a benign cholestatic entity because it carries little to no risk of developing chronic liver damage. In contrast to this view, earlier descriptions had already noted that functional tests such as indocyanine green clearance and hepatic metabolism were significantly altered in patients with Gilbert's disease.35, 36 The availability of more detailed knowledge surrounding the molecular and genetic aspects of bilirubin metabolism and the UGTs has led to a considerable body of evidence linking the principle defect of Gilbert's disease, a glucuronidation activity reducing TA insertion into the promoter of the human hepatic UGT1A1 gene (UGT1A1*28), to a disposition to cancer,7, 37 and severe unwanted drug reactions such as leukopenia and diarrhea in irinotecan therapy.12, 38 Even this view may be too simple, however, because it emphasizes the function of a single gene and its corresponding protein. Genetic analyses of the human UGT1A gene locus have shown that functional SNPs have been detected in each of the first exons encoding the nine functional UGT1A proteins.21 Their frequencies have been observed to range between 2% and over 50%, predicting that combinations of individual genetic variants of glucuronidation enzymes are likely to exist in higher numbers than previously believed. Against this background Gilbert's disease is an interesting disease that exhibits a clinically apparent phenotype of benign intermittent jaundice but also may be an indicator of more complex alterations of metabolism.13 In this study, we analyzed patients treated with the protease inhibitor atazanavir (ATV), which is not a significant substrate for glucuronidation but acts as an inhibitor of UGT proteins, including UGT1A1, UGT1A3, and UGT1A4.25 This situation can be expected to affect individuals who already suffer from reduced UGT activity such as homozygous carriers of UGT1A1*28 with only approximately 30% of UGT1A1 activity.31 However, this also affects genetically determined low-activity alleles of other UGT proteins inhibited by ATV, the effects of which will not necessarily be evident as a hyperbilirubinemic phenotype linked to the specific bilirubin activity of UGT1A1. This view is further strengthened by the clinical observation that the degree of hyperbilirubinemia in Gilbert's patients and in ATV-treated patients is variable even in those patients who exhibit the UGT1A1*28 genotype, suggesting that the pharmacogenetic risk profile may involve additional factors.

For this reason, UGT1A1*28 was determined in 106 ATV-treated HIV-positive patients together with an analysis of UGT1A3 −66C, UGT1A7 −57G, as well as UGT1A7 129K/131K. This study focused on a haplotype analysis using patient genomic DNA and did not analyze tissue glucuronidation activities. Genetic variants of UGT1A1 and UGT1A3 were chosen because of the inhibitory effects of atazanavir on these proteins. UGT1A7 was chosen because it represents a well-characterized UGT with reduced function coding sequence (exon 1) polymorphisms,22 and a functional TATA-box polymorphism leading to approximately 30% of wild-type transcriptional activity.13 The comparison of UGT SNPs in ATV-treated patients and healthy blood donors from the same Northern German Caucasoid background demonstrated that homozygous individuals for each of the four markers were significantly more frequent in the treatment group (Fig. 1C). This fact may be reflective of the higher rate of male patients (83%) in the HIV group, although a significant sex difference was not detected in blood donors or HIV patients. Moreover, in an analysis after stratification according to the grade of hyperbilirubinemia, those ATV-treated patients with normal to moderate hyperbilirubinemia did not differ significantly from blood donors (Table 3). In addition, the allelic frequencies of UGT1A1*28 in blood donors and ATV-treated patients was not significantly different (Fig. 1C). However, this does not rule out a disease-related bias in the patient group.

The stratification of patients by grade of hyperbilirubinemia led to the remarkable observation that not only all of the individual markers were more frequent in patients with severe to serious hyperbilirubinemia but that these individuals were also characterized by a high degree of concurrence of UGT1A1, UGT1A3, and UGT1A7 polymorphisms, which are not in linkage dysequilibrium. In blood donors and ATV-treated patients with hyperbilirubinemia grade 0 to 2, the simultaneous presence of homozygous UGT1A1*28 with UGT1A3 −66 C, UGT1A7 −57 G, and UGT1A7 129K/131K was approximately 10% in comparison with 41% to 46.1% in grades 3 and 4, and 100% in grade 4 hyperbilirubinemia. The identification of homozygous carriers with severe hyperbilirubinemia further indicated that all four variants can exist on the same allele of these patients and are therefore inherited as a haplotype (Fig. 2). This haplotype of UGT1A1-UGT1A3-UGT1A7 variants was present as a homozygous trait in 9.6% of blood donors, 41.2% of grade 3 and 4 hyperbilirubinemic patients, and in all (100%) of the six patients in this study with grade 4 hyperbilirubinemia. These data provide direct evidence of a UGT variant haplotype as a risk factor for ATV-induced and possibly other drug-induced side effects. The identified haplotype does not mechanistically explain why hyperbilirubinemia is associated with additional UGT1A variants that are not directly involved in bilirubin metabolism. It does indicate that in those patients with additional variants the risk is increased. This variant haplotype may additionally include other genes beyond the UGT1A gene locus as part of this risk profile. Such variants may include hepatobiliary transport proteins, as suggested by the early descriptions of Gilbert's disease.35, 36

The haplotype of UGT1A variants identified in this study because of the inhibitory effects of ATV may be of additional pharmacogenomic relevance. UGT1A1 is a hepatic and intestinal UGT capable of glucuronidating bilirubin and estradiol as endogenous substrates together with xenobiotics including buprenorphine, the irinotecan metabolite SN-38, N-hydroxy PhIP, hydroxy-benzo(α)pyrenes, simvastatin, gemfibrozil, and valproate.39 Inhibition of UGT1A1 occurs by ATV25 but also by amitriptyline, canrenoic acid, sulfinpyrazone,24 and ketoconazole.26 The combination of genetic variant and drug treatment that includes individual or combinations of the previously mentioned lipid-lowering, anti-fungal, anti-viral, or diuretic drugs therefore has the potential for drug interactions in UGT1A1*28 carriers. Inhibition and genetic variants of the hepatic and intestinal UGT1A3 isoform affect the metabolism of substrates which include common drugs such as amitriptyline, imipramine, ketoprofen, lamotrigine, simvastatin, gemfibrozil, and valproate.39–42 Some of these are UGT inhibitory such as amitriptyline and acyl glucuronides of ketoprofen that form adducts with the UGT protein.27, 28 UGT1A7 expressed primarily in stomach, esophagus,43 and pancreas44 is characterized by the glucuronidation of mutagenic polyaromatic hydrocarbons, and the irinotecan metabolite SN-38.13, 43 It is additionally associated with a predisposition to cancer.22, 45 In view of these functional data, the inheritance of a haplotype combining multiple functional SNPs of the three UGTs analyzed in this study may alter the metabolism of a number of drugs. This may be the case when therapeutic drugs are employed that either have a high specificity for an individual UGT, are inhibitory as parent compound or as glucuronide, or are combined with other drugs that have one or both of these properties. This is a possible scenario in clinical practice, that is, in ATV-treated HIV patients but not restricted to this particular treatment situation. Because this study investigated the association of a clinical phenotype by genetic haplotype analysis and not tissue metabolism, future studies are needed to additionally study the mechanistic impact of the identified UGT haplotype associated with jaundice in ATV-treated individuals.

In summary, this study analyzing ATV-treated patients provides evidence for a haplotype of four genetic variants at the human UGT1A gene locus linked to the easily distinguishable common clinical phenotype of Gilbert's disease. This haplotype is identified as a pharmacogenomic risk factor. The improvement of drug safety and the establishment of individualized pharmacotherapy depend on the identification of this and similar risk constellations, which are likely to involve additional enzyme systems such as cytochrome P450s and other conjugases. Suggestions to incorporate pharmacogenomic data and tests into the drug approval process are a first step that may eventually lead to individualized therapy and risk assessment.