Combined effect of regulatory polymorphisms on transcription of UGT1A1 as a cause of Gilbert syndrome

Background Gilbert syndrome is caused by defects in bilirubin UDP-glucuronosyltransferase (UGT1A1). The most common variation believed to be involved is A(TA)7TAA. Although several polymorphisms have been found to link with A(TA)7TAA, the combined effect of regulatory polymorphisms in the development of Gilbert syndrome remains unclear. Methods In an analysis of 15 patients and 60 normal subjects, we detected 14 polymorphisms and nine haplotypes in the regulatory region. We classified the 4-kbp regulatory region of the patients into: the TATA box including A(TA)7TAA; a phenobarbital responsive enhancer module including c.-3275T>G; and a region including other ten linked polymorphisms. The effect on transcription of these polymorphisms was studied. Results All haplotypes with A(TA)7TAA had c.-3275T>G and additional polymorphisms. In an in-vitro expression study of the 4-kbp regulatory region, A(TA)7TAA alone did not significantly reduce transcription. In contrast, c.-3275T>G reduced transcription to 69% of that of wild type, and the linked polymorphisms reduced transcription to 88% of wild type. Transcription of the typical regulatory region of the patients was 56% of wild type. Co-expression of constitutive androstane receptor (CAR) increased the transcription of wild type by a factor of 4.3. Each polymorphism by itself did not reduce transcription to the level of the patients, however, even in the presence of CAR. Conclusions These results imply that co-operation of A(TA)7TAA, c.-3275T>G and the linked polymorphisms is necessary in causing Gilbert syndrome.


Background
The elimination pathway of bilirubin in humans is catalyzed exclusively by bilirubin UDP-glucuronosyltransferase (UGT1A1) [1]. Gilbert syndrome is a mild hereditary unconjugated hyperbilirubinemia without liver dysfunction or hemolytic anemia. It is caused by defects of UGT1A1 and is one of the most prevalent congenital metabolic disorders; Gilbert syndrome is found clinically in 3-10% of the population [2][3][4][5].
The UGT1A1 gene (UGT1A1) has the TATA box and a phenobarbital responsive enhancer module (gtPBREM) in its regulatory region [6,7]. The wild type of TATA box has six repeats of TA, and is conventionally written as A(TA)6TAA [6]. The gtPBREM is an important enhancer module of UGT1A1, comprising a region of 290-bp. It is located approximately 3.5 kbp upstream from the coding region and works in the presence of nuclear receptors, such as constitutive androstane receptor (CAR) [7].
UGT1A1 activity in patients with Gilbert syndrome was about 30% of normal, based on studies involving human liver samples [16][17][18]. In subsequent studies using human liver samples of homozygous A(TA)7TAA, UGT1A1 enzyme activity was 48% of wild type, and the expression level of UGT1A1 protein was 42-50% of wild type [18][19][20]. A(TA)7TAA reportedly has only one-third of the transcriptional activity of wild type, based on a gene expression study with the regulatory region from c.-546 to c.-4 [11]. No statistically significant difference in activity between A(TA)6TAA and A(TA)7TAA was detected in our own earlier study, however, using six different lengths of regulatory region varying from 111 to 3188-bp in length [21]. Also, other groups reported that transcriptional activity was not so drastically reduced by A(TA)7TAA (to 70-86% of wild type, not by one-third) [14,22].
The frequency of c.-3275T>G in gtPBREM is 0.26 in Japanese, 0.47 in Caucasians and 0.85 in African-Americans [12,23]. We found that all patients with Gilbert syndrome having homozygous A(TA)7TAA were also homozygous for c.-3275T>G, and we suggested that a combined effect of these polymorphisms on transcription is essential to the syndrome [24]. Furthermore, the c.-3152G>A and c.-364C>T polymorphisms, as well as c.-3275T>G, were recently found to link with A(TA)7TAA [12,23], and numerous single nucleotide polymorphisms (SNPs) of UGT1A1 were also reported on GeneSNPs [25].
To determine the role of the regulatory polymorphisms in Gilbert syndrome, we first analyzed haplotypes of the patients and normal subjects. Based on the results, the transcriptional activity of a continuous 4-kbp regulatory region was studied in the presence and absence of CAR.

Sequencing of the UGT1A1 transcriptional regulatory region
The upstream region of UGT1A1 were analyzed in a random population of 50 Japanese subjects and in four Japanese with Gilbert syndrome having homozygous A(TA)7TAA, as well as in ten healthy normal Caucasians and 11 unrelated Caucasians with Gilbert syndrome; all gave informed consent. The study was approved by the ethics committee of Shiga University of Medical Science. We amplified the regulatory region as four separate DNA fragments, using the polymerase chain reaction (PCR  Tables -A1-A and -D1-D show all primers and conditions used in this study.

Construction of expression vectors
To study the effect of the TATA box, gtPBREM, and the other polymorphisms in the regulatory region, we constructed luciferase expression vectors including the regulatory regions with differing combinations of these polymorphisms. We also constructed a CAR expression vector, to investigate the effect of CAR on transcription of the regulatory region with gtPBREM.

Cloning of the regulatory region in two separate DNA fragments
Two types of genomic DNA were chosen to construct expression vectors. One is a Gilbert-syndrome type from a Japanese patient with Gilbert syndrome having homozygous A(TA)7TAA. The other is wild type from a normal subject having identical sequence, reported in GenBank Accession No. NG_002601. The distal region (c.-4076 to c.-1945) and proximal region (c.-2483 to c.247) were amplified with PCR, using TaKaRa LA Taq ® Hot Start Version (TaKaRa BIO INC) (Table -B1-B). The PCR products were cloned into a pCR ® -XL-TOPO ® vector with a TOPO XL PCR cloning kit (Invitrogen Corporation), and the entire regions involved were checked by sequencing so as to eliminate any mistakes arising in the PCR amplification (Table -E1-E).

Construction of expression vectors
Restriction sites for ligation into an expression vector were introduced into the proximal and distal regulatory regions by PCR of the cloned regulatory region, using PCR primers with the restriction sites (Table -C1 Table 2). The effect of the region on the transcriptional activity was determined as a unit. We classified the regulatory region of UGT1A1 with the detected polymorphisms into the following three regions: the TATA box including

Construction of CAR expression vector
CAR-cDNA was isolated from a human liver cDNA library (TaKaRa BIO INC) by PCR, as described by Auerbach et al [26]. The PCR product was cloned into pCR2.1 vector (Invitrogen Corporation). CAR-cDNA was cut out by EcoR I and EcoR V, and ligated into the pCR3.1 vector (Invitrogen Corporation).

Transfection and luciferase assay
HepG2 cells from the Riken Cell Bank (Tsukuba, Japan) were cultured in the minimal essential medium, which includes Earle's salts, non-essential amino acids but not -L-Glutamine (MEM, Invitrogen Corporation), supplemented with 10% fetal bovine serum (Invitrogen Corporation), 50 U of penicillin/ml and 50 μg of streptomycin/ ml. The cells were transferred to a 24-well culture plate (7 × 10 4 cells/well) at 25 hours before transfection.

Data analysis
Haplotype analysis was carried out using Haploview 3.32 software [27]. Changes in transcriptional activity due to the polymorphisms were analyzed by multiple regression analysis (see Additional file 1, Table S1), using SPSS16.0.1J software (SPSS Japan Inc., Tokyo, Japan). For dummy coding, we defined the wild-type and mutanttype polymorphisms as 0 and 1, respectively.

Sequencing and haplotype analysis in the regulatory region of UGT1A1
The upstream region of UGT1A1 was analyzed in patients with Gilbert syndrome and in normal Japanese and Caucasian subjects. The regulatory region of wild type was consistent with NG_002601, which is stated to be the sequence of UGT1A1 by the US National Center for Biotechnology Information (   Table S2).

Effect of regulatory polymorphisms on transcription
Based on haplotype analysis of Gilbert patients in Japanese and Caucasians, we selected the most typical regulatory regions of wild and Gilbert-syndrome type for invitro expression studies: in Table 2, type I and type V respectively. The regulatory region for Gilbert syndrome was cloned from a Japanese patient because of the uncer- The gray bars in Figure 2 show transcriptional activities without co-expression of CAR. A(TA)7TAA did not significantly reduce transcriptional activity (94% of wild type, p = 0.14). In contrast, c.-3275T>G and the ten linked polymorphisms significantly reduced the activity, to 69% and 88% of wild type respectively (p < 0.001 and p = 0.001). Our study shows that c.-3275T>G exerts the greatest reduction in transcription. The typical haplotype of Gilbert patients (G-m-(TA)7_Gilbert in Figure 2) reduced transcriptional activity to 56% of wild type.
The white bars in Figure 2 show transcriptional activities with co-expression of CAR. The activity of wild-type haplotype was increased with CAR by a factor of 4.27 (S.D. = 0.18, p < 0.001). The ten linked polymorphisms alone did not reduce transcriptional activity. In contrast, c.-3275T>G and A(TA)7TAA significantly reduced the activity, to 85% and 83% of wild type respectively (p < 0.001 and p < 0.001). The typical haplotype of Gilbert syndrome patients (G-m-(TA)7_Gilbert in Figure 2) reduced transcriptional activity to 68% of wild type. These results suggest that AT(TA)7TAA by itself does not reduce transcriptional activity enough to cause Gilbert syndrome, and that co-operation of c.-3275T>G and the ten linked polymorphisms with AT(TA)7TAA is necessary to cause the syndrome.

Discussion
By reducing the transcription of UGT1A1, A(TA)7TAA is acknowledged to be one of the most important causes of Gilbert syndrome. Further polymorphisms located upstream of the UGT1A1-coding region have recently been detected but, apart from A(TA)7TAA and -3275T>G at gtPBREM, the role of these polymorphisms in the development of Gilbert syndrome is currently unclear.
In the present analysis of the regulatory region of UGT1A1, nine haplotypes were detected with differing combinations of 14 polymorphic variations ( Table 2). We classified the regulatory region into three categories so as to simplify the analysis, taking account of previous studies and the locations found in the regulatory sequence. Those three categories are: gtPBREM includ-    Haplotypes with A(TA)6TAA Haplotypes with A(TA)7TAA Boldfaces indicate variations. a JRS, Japanese random subjects; b JG7, Japanese patients with Gilbert syndrome having homozygous A(TA)7TAA; c NC, normal Caucasians; d CG, Caucasian patients with Gilbert syndrome; e Numbers in parentheses indicate the number of total alleles in each group *Ten linked polymorphisms between gtPBREM and TATA box ing c.-3275T>G; the region including ten linked polymorphisms; and the TATA box including A(TA)7TAA. We further examined their effect on transcription, both in the presence and absence of CAR. This was done for three reasons. First, enhancer activity of gtPBREM is elevated by CAR in in-vitro experiments using culture cells [10]. Second, CAR has been proposed as a key regulator of bilirubin clearance in the liver in wild type and humanized CAR transgenic mice [29]. Third, c.-3275T>G is reported to be a further factor for the development of hyperbilirubinemia [30]. Transcription of the wild-type regulatory region was enhanced 4.3 times by co-expression of CAR (Figure 2). This was a similar factor to that in previous studies of UGT1A1 [10]. There were differences in transcription between A(TA)7TAA, c.-3275T>G and the ten linked polymorphisms, both in the presence and the absence of CAR. A(TA)7TAA by itself reduced transcription to 94% of wild type (p = 0.14, not significant) without CAR, and to 83% (p < 0.001) with CAR. By itself, then, A(TA)7TAA does not reduce transcriptional activity to 42-50%, which is the value estimated from the enzyme activity and the amount of enzyme protein in human liver samples of homozygous A(TA)7TAA [18][19][20] The ten linked polymorphisms located between A(TA)7TAA and c.-3275T>G together reduced transcription to 88% (p = 0.001) of wild type without CAR. This reduction might be due mainly to c.-1352A>C, since this variation reduced the activity to 85% of wild type in our previous study with a 3-kbp regulatory region [21]. The ten linked polymorphisms did not induce any significant change in transcription with CAR. It is not clear why there was no significant reduction of transcription in the presence of CAR, but the 4.3-fold enhancement of the activity with CAR might mask the inhibitory effect of c.-1352A>C.
Unlike A(TA)7TAA and the ten linked polymorphisms, c.-3275T>G noticeably reduced transcriptional activity both with and without CAR, to 69% (p < 0.001) and 85% (p < 0.001) of wild type respectively. This result is consistent with the report that c.-3275T>G is a factor for the development of hyperbilirubinemia [30].
Sugatani et al. recently reported that the TATA box reduced transcription to a greater extent than c.-3275T>G (c.-3279T>G on their sequence), in a study using the regulatory region from c.-3570 to c.-1 [28]. This regulatory region has two significant differences in their work and ours. First, they did not consider the ten polymorphisms between gtPBREM and the TATA box. Second, their study used the sequence of AF297093 whereas we used NG_002601. At least part of the discrepancy between their results and ours might stem from these differences.
Without CAR, the typical haplotype of Gilbert syndrome patients was effective in reducing transcription, to 56% of wild type, and with CAR to 68% of wild type (Figure 2). The results without CAR are almost compatible with previous in vivo data: UGT1A1 enzyme activity was 48%, and the expression level of UGT1A1 protein was 42-50%, of wild type in studies using human liver samples of carriers with homozygous A(TA)7TAA [18][19][20]. There is only a small amount of CAR in the nuclei of human primary hepatocytes [31], but it has been reported that defects in bilirubin clearance are not observed in CARnull mice until the appropriate stress is applied, and that nuclear CAR increases in the presence of activators such  as a large concentration of bilirubin [29]. Consequently the CAR state of the patients may be similar to those in the experiment without CAR, in which case CAR is not important to serum bilirubin levels of the patient. Even if the amount of nuclear CAR in the patients is intermediate between the values in the experiments with and without CAR, A(TA)7TAA by itself is not enough to reduce transcription to the level found in the liver of patients with homozygous A(TA)7TAA [18][19][20]. Bilirubin levels are found to be related to polymorphisms of UGT1A1; c.-3275T>G, A(TA)7TAA and c.211G>A in epidemiological genetic studies [30,32].
Epidemiological evaluation of the effect of each polymorphism on the development of Gilbert syndrome is difficult, because almost all haplotypes with A(TA)7TAA also have c.-3275T>G and other polymorphisms such as c.-1352A>C.
Our in-vitro studies imply the existence of combined effects of the polymorphisms on transcriptional activity in the development of Gilbert syndrome; these studies use vectors with combinations of the polymorphisms designed specifically to analyze the role of the individual polymorphisms. Our results may explain why almost all patients with Gilbert syndrome are double homozygous for A(TA)7TAA and c.-3275T>G [24,33,34]. Since c.-3275T>G alone is not enough to reduce the transcription level to that of the patients, less frequent A(TA)7TAA may play a determinant role in the etiology of Gilbert syndrome. Exceptional patients, such as one who is homozygous for A(TA)7TAA and heterozygous for c.-3275T>G, heterozygous for A(TA)7TAA and homozygous for c.-3275T>G, and homozygous only for c.-3275T>G, may have further variations of UGT1A1 that have not yet been investigated, or have variations of the other enzyme(s) related to bilirubin metabolism [28,33,34]. Further studies of the relation between serum bilirubin concentration and genotype that include enough patients with minor haplotypes, based on this analysis, should identify accurately the involvement of each polymorphic mutation in the development of Gilbert syndrome.

Conclusions
All haplotypes of Gilbert syndrome patients with A(TA)7TAA have both c.-3275T>G and 8-11 polymorphisms between gtPBREM and A(TA)7TAA. CAR increases transcription and changes the amount of inhibition exerted by each polymorphism on transcription. Regardless of the presence of CAR, A(TA)7TAA did not by itself reduce transcriptional activity enough to cause Gilbert syndrome. Our study shows that c.-3275T>G and the ten polymorphisms detected in the patients are also important for the decrease of transcription of UGT1A1 in addition to A(TA)7TAA, and that Gilbert syndrome is likely to be caused by the combined effects of these polymorphisms.