Introduction

Crigler–Najjar syndrome type I (CN-I, MIM no. 218800) is a rare and severe autosomal disorder (less than 1/106 live births). CN-I is caused by deficiency of the liver enzyme responsible for bilirubin elimination, the uridine diphosphate glucuronosyltransferase 1A1 (UGT1A1; EC 2.4.1.17). Biologically, the disease manifests itself with severe and persistent unconjugated hyperbilirubinemia. Clinically, affected newborns are at high risk for bilirubin-induced brain damage (kernicterus). The UGT1A locus complex on chromosome 2 (located on 2q37.1) encodes nine functional UDP-glucuronosyltransferases type 1A.1, 2 The locus is organized with 13 different exons 1 (responsible for the specificity of the enzyme activity) and 4 common exons. Recently, a new regulation system of the UGT1A1 protein activity was identified with an alternative exon 5.3 As for the UGT1A1 gene, about 70 different mutations have been reported since 1992, in the five exons or in splicing sites.4, 5, 6 In our laboratory, the recurrent mutation c.1070A>G responsible for the modification in the peptide sequence p.Gln357Arg has been only observed in patients from Tunisia, suggesting a Tunisian founder effect.7 Recently, this mutation has also been reported in two Kuwaiti Bedouin families, suggesting a wider Arabian founder effect.8

The recurrence of a particular mutation may result from three different conditions. First, the mutation occurs in a region of high risk for mutation (‘hot spot’), such as CpG dinucleotides.9 Second, an external event (such as an infectious disease) has exerted a pressure on the population and selected individuals carrying the morbid allele (eg, the mutation responsible for sickle cell anemia was preserved in the African population by conferring on heterozygote patients, a resistance to malaria).10 Finally, the mutation appeared in one individual and diffused through generations in isolated communities because of consanguineous marriages. In this situation, the chromosomal fragment bearing the mutation is transmitted from a common ancestor to many persons in successive generations. When a founder effect is suspected, several tandem repeats or microsatellites must be analyzed in a restricted genomic region to confirm the single origin (common ancestor) of a recurrent mutation in a population. During meiosis, chromosomal recombination can randomly occur, so that the more recent the mutation is, the more conserved the genetic region around the mutation will be. In these cases, microsatellites around the mutation will disclose the same or at least a similar haplotype in all affected patients.

In this study, genetic markers on both centromeric and telomeric sides of the c.1070A>G mutation in the UGT1A1 gene were analyzed in 21 Tunisian and 2 Kuwaiti CN-I patients. We have determined that this mutation probably appeared in a Bedouin nomad group and then diffused in Tunisia around eight centuries ago after population migrations from Middle East to Maghreb.

Patients and methods

To determine the origin of the c.1070A>G mutation in the UGT1A1 gene, 28 healthy unrelated Tunisians (group 1), 21 CN-I Tunisian patients (group 2) and 2 CN-I Kuwaiti patients (group 3) were included. Seven microsatellite markers (centromere-D2S2344, D2S331, D2S1279, D2S2348, D2S2234, D2S2205 and D2S336-telomere) were selected from the Ensembl Genome Project Database. Haplotypes were constructed by amplification of microsatellites with fluorescent primers 5′-labeled in a multiplex PCR except for D2S336 and D2S2205, each amplified in a single PCR. The samples were run on an ABI PRISM 3130 DNA analyzer (Applied Biosystems). The results were analyzed by using Genescan 3.1 analysis software to determine the PCR fragment sizes of PCR products, and the allele sizes were carried out automatically with Genotyper 2.5 analysis software. Single PCRs were performed using the same program for D2S336 and D2S2205. Haplotypes were classified as 1 for the shortest to 20 for the longest.

To complete this panel, two other markers were added: the UGT1A1 TATA box polymorphism and a new polymorphic genetic marker ∼77 000 bp upstream from UGT1A1. The polymorphism A(TA)6TAA or A(TA)7TAA in the promoter of the UGT1A1 gene was determined by PCR followed by electrophoresis on polyacrylamide gel as described previously.11 The second genetic marker – called MARK01 – was a CA-repeat sequence whose variability had previously been assayed on 10 French healthy unrelated volunteers. Its physical location was calculated to be 77 440 pb upstream from the mutation c.1070A>G. Primers to amplify this specific region were designed using Primer3 (Primer3 web site: http://frodo.wi.mit.edu/cgi-bin/primer3/primer3_www.cgi). The primer sequences were as follows: forward primer 5′-GGAGCTACTCTTTAGGGATCG-3′; reverse primer 5′-TGTGAGCTTTGACTGTACTAAG-3′. After PCR amplification, fragments were separated by electrophoresis onto a 10% polyacrylamide gel and then were visualized after ethidium bromide staining. As for the other microsatellites studied, haplotypes were classified according to their electrophoretic profile as 1 for the shortest to 4 for the longest.

To determine linkage disequilibrium (LD) score, the appropriate physical-to-genetic distance conversion relation was determined by linear regression analysis (SPSSv8.0 program, SPSS) of genetic versus physical map position.12 Megabase (UniSTS, NCBI) and centimorgan (Marshfield Comprehensive Human Genetic Map) information for 14 markers positioned in a 10-Mb genomic region around UGT1A1 gene was used (Table 1). Applying this relation, genetic distances were calculated in relation to the mutation c.1070A>G for the nine studied genetic markers (Table 2). Kosambi's function was applied to convert genetic distance (cM) into recombination fraction (θ): θ=0.5(e(cM/25)−1)/(e(cM/25)+1) (Table 2).13 Statistical comparison of allele frequencies between disease and normal chromosomes was based on Mantel–Haenszel common odds ratio estimate with alleles classified into two groups: one for the associated allele and all others combined into a single group. The LD was calculated applying the formula δ=(p11p22−p12p21)/(p+1p22), where p11 is the frequency of the associated allele on disease chromosomes, p22 the frequency of the normal alleles on normal chromosomes, p12 the frequency of the associated allele on normal chromosomes, p21 the frequency of the normal alleles on disease chromosomes and p+1 the frequency of disease chromosomes.13

Table 1 Physical and genetic distance of 14 markers on chromosome 2 to determine the appropriate physical-to-genetic ratio in the UGT1A1 region
Full size table
Table 2 Linkage disequilibrium between D2S2344 and D2S336 markers and age estimation of the founder UGT1A1 c.1070A>G mutation
Full size table

The number of generations (g) since the appearance of the mutation was calculated with the formula g=ln(δ)/ln(θ−1), where δ is the LD and θ the recombination fraction.14

Results

Allele frequencies for the nine genetic markers group are presented in Tables 3a, b (for group 1 and group 2, respectively).

Table 3a Allele frequencies for the nine genetic markers in the Tunisian healthy population
Full size table
Table 3b Allele frequencies for the nine genetic markers in the Tunisian CN-I population
Full size table

The polymorphism observed in the CN-I Tunisian population was strongly restricted in comparison with the healthy Tunisian population (Table 2). For example, for D2S2348, only 4 different alleles out of 20 observed in the healthy Tunisian population were identified in the CN-I Tunisian population. In the healthy Tunisian population, frequency of heterozygotes for the A(TA)7TAA allele was 0.393 and frequency of homozygotes was 0.178. For all Tunisian CN-I patients, the c.1070A>G mutation was associated with the homozygous A(TA)7TAA mutant allele in the promoter. For MARK01, the allele 4 frequency was 0.952 in the Tunisian CN-I population and only 0.357 in the healthy Tunisian population. The ancestor haplotype (D2S2344: 5, D2S331: 9, D2S1279: 9, D2S2348: 5, D2S2234: 5, D2S2205: 1, MARK01: 4, TATA box: A(TA)7TAA and D2S336: 4) was observed in three Tunisian CN-I patients and was not observed in the healthy Tunisian population.

LD calculated with the Delvin and Risch formula decreased from 1.000 to 0.4524 with increasing distance of markers from the mutation, except for D2S2234 and D2S2348. The age of the mutation was estimated to be between 21 and 60 generations (mean 32 generations).

In the two Kuwaiti patients, the c.1070A>G mutation was also associated with the homozygous A(TA)7TAA mutant allele in the promoter. Their haplotype was identical to the ancestor one.

Discussion

CN-I is a rare genetic disease whose frequency is estimated to be 1/106 births and it affects boys and girls in the same proportion. Founder effects have already been suspected for CN-I in isolated communities such as in France, Portugal or Sardinia.15, 16

In our laboratory, clinical and/or biochemical CN-I was genetically confirmed for 56 patients. Twenty-five of these patients originated from different part of Tunisia (Tunis, Sfax and Sousse) and 21 were homozygous for the c.1070A>G mutation in exon 3 associated with the A(TA)7TAA/A(TA)7TAA polymorphism in the TATA box (three of the other patients were homozygous for the deletion c.396_401delCAACAA associated with the A(TA)7TAA/A(TA)7TAA polymorphism and the fourth was homozygous for a large deletion including the promoter and the exon 1). The implication of this mutation in the CN-I phenotype has never been determined by in vitro expression studies, but convincing arguments support this relation, particularly the frequency of this mutation in the CN-I Tunisian population. However, the c.1070A>G mutation has always been found associated with the A(TA)7TAA/A(TA)7TAA polymorphism, so that it is, to date, impossible to determine whether this mutation totally abolishes UGT1A1 activity by itself or whether its association with the A(TA)7TAA/A(TA)7TAA genotype in the TATA box is necessary. In the Tunisian population, the Gilbert allele A(TA)7TAA frequency was found to be 0.393 and 0.178 for heterozygotes and homozygotes, respectively. These data were comparable to those observed in France (0.385 and 0.17, respectively) or Greece (0.328 and 0.186, respectively).17, 18

CN-I or II cases have been described in several regions of the Middle East and the Maghreb such as Morocco, Algeria, Tunisia, Saudi Arabia, Lebanon or Kuwait.7, 8, 15, 19, 20 The high prevalence of this disease, particularly in Saudi Arabia or Tunisia can be explained by a high rate of consanguineous marriages in restricted communities.19 In the Tunisian population, the large prevalence of the c.1070A>G mutation in exon 3 had suggested an ancestral common origin.7 The absence of particular susceptibility of the exon 3 to this mutation (only observed in Tunisia and in two Kuwaiti Bedouins families) reinforces the founder effect hypothesis.7, 8 On the other hand, the question of a protective effect of Crigler–Najjar causing alleles at the heterozygote state against oxidative damages such as heart coronary diseases is not clear. Vitek et al21 pointed that patients with mild increase of serum bilirubin could be protected against heart disease. But on the other hand, Bosma et al22 and Gajdos et al23 in two different studies did not find any correlation between mutations in UGT1A1 gene promoter and protection against heart disease. Moreover an eventual protective effect would be observed after reproductive age. These observations could not explain the high prevalence of Crigler–Najjar syndrome in the Tunisian population.

In this study, genetic analysis allowed us to confirm for the first time the founder effect hypothesis in the Tunisian population. The c.1070A>G mutation appeared some 32 generations ago (nearly eight centuries ago), and its diffusion was strongly limited, probably thanks to consanguineous marriages. The systematic presence of the A(TA)7TAA/A(TA)7TAA genotype in the promoter in c.1070A>G mutation carriers suggests that this mutation appeared after the promoter polymorphism, reinforcing the founder effect hypothesis. Genetically, the mutation c.1070A>G and the TATA box are very close (7702 bp or 0.021 cM). In a population of 21 patients, the number of generations necessary to observe one meiotic recombination between these two points is estimated to be 191, spanning about 4000 years.

Unfortunately, the Kuwaiti population studied was too restricted to determine LD, but genetic marker analysis allowed us to identify the ancestor haplotype in the two patients.

Referring to the Bedouin population history, the mutation probably appeared in the Bedouin community and was introduced in the Tunisian population nearly eight centuries ago after human migrations from east to west. Indeed, Bedouin nomads spread out from the Arabian Peninsula into all countries between the Arabic Gulf and the Atlantic. Bedouins arrived in Tunisia very early during the first Arab-Muslim invasions and married the native population, leading to today's population.