Triallelic Inheritance in Bardet-Biedl Syndrome, a Mendelian Recessive Disorder

See allHide authors and affiliations

Science  21 Sep 2001:
Vol. 293, Issue 5538, pp. 2256-2259
DOI: 10.1126/science.1063525


Bardet-Biedl syndrome (BBS) is a genetically heterogeneous disorder characterized by multiple clinical features that include pigmentary retinal dystrophy, polydactyly, obesity, developmental delay, and renal defects. BBS is considered an autosomal recessive disorder, and recent positional cloning efforts have identified two BBS genes (BBS2 andBBS6). We screened our cohort of 163 BBS families for mutations in both BBS2 and BBS6 and report the presence of three mutant alleles in affected individuals in four pedigrees. In addition, we detected unaffected individuals in two pedigrees who carry two BBS2 mutations but not aBBS6 mutation. We therefore propose that BBS may not be a single-gene recessive disease but a complex trait requiring three mutant alleles to manifest the phenotype. This triallelic model of disease transmission may be important in the study of both Mendelian and multifactorial disorders.

Locus heterogeneity in Mendelian disorders is the phenomenon whereby mutations in different genes result in a similar or identical clinical phenotype. In most reported instances, mutations at a single locus suffice to cause disease, although rare cases have been reported where mutations at two loci are necessary for pathogenesis (1) or exacerbate the severity of the phenotype (2, 3). BBS is a typical example of a rare, genetically heterogeneous disorder. BBS patients manifest a complex and variable phenotype that includes pigmentary retinal dystrophy, polydactyly, central obesity, hypogonadism, learning difficulties, and renal dysplasia; additional features such as asthma and diabetes mellitus may also be present (4, 5). The segregation of the disorder in families and population isolates led to the hypothesis that the syndrome is inherited in an autosomal recessive manner. On the basis of this model, six BBS loci have been identified: BBS1 on 11q13 (6),BBS2 on 16q21 (7), BBS3 on 3p12 (8), BBS4 on 15q22.2-q23 (9),BBS5 on 2q31 (10), and BBS6 on 20p12 (11), with evidence for at least one more locus (12). Two BBS genes have been cloned recently:BBS6 (11, 13), which encodes a putative chaperonin (14), and BBS2, which encodes a protein of unknown function (15).

During our analysis of BBS6 in 163 patients, we identified eight pedigrees with mutations. A high frequency of these (seven of eight) harbored only a single mutant BBS6 allele (12). Furthermore, we described one consanguineous pedigree in which both the affected and unaffected sibs carried a heterozygous A242S mutation in BBS6, but the affected sib also exhibited homozygosity by descent (HBD) across the BBS2 locus. This observation suggests either that BBS arises through multiallelic inheritance or that the A242S allele is a rare polymorphism (12).

The recent identification of BBS2 enabled us to test the former hypothesis by screening the same cohort of BBS patients (16) for coding sequence alterations inBBS2, irrespective of haplotype-inferred chromosomal assignment for any given pedigree or previously obtained mutational data for BBS6. Upon deducing the intron-exon structure ofBBS2 (17), we designed suitable primers and amplified and sequenced each of the 17 exons of this gene in our entire patient cohort (18).

We identified numerous sequence alterations. In 19 unrelated patients (Table 1), the alterations fulfilled our minimal criteria for mutations, in that they caused a nonconservative amino acid change and that they were not found in a minimum of 192 control chromosomes from ethnically matched samples (19). Initial segregation analyses confirmed thatBBS2 mutations were likely to be pathogenic. We identified six pedigrees (AR171, PB005, PB020, PB026, PB058, and K059) in which two independent mutations segregated with the disease (Table 1) (20). However, despite complete DNA sequence coverage of theBBS2 open reading frame (ORF) and exon-intron boundaries, we identified only a single mutant allele in eight pedigrees. We considered four possible explanations for this result: (i) the single alleles found may not be pathogenic but rare polymorphisms, (ii) by chance alone, we were detecting rare BBS2 carriers, (iii) the second mutation might reside in the regulatory regions or introns of BBS2 or might be undetectable by sequencing because of deletions of the whole or part of the gene, or (iv) additional mutations lie in another gene, and it is the combination of mutant alleles at two or more loci that causes the BBS phenotype.

Table 1

Mutational, genetic, and phenotypic analysis of pedigrees with mutations in BBS2. The different aspects of the BBS phenotype in each pedigree are indicated as present (+), normal (Nml), or not recorded (NR). The primary features of BBS were examined: retinitis pigmentosa (RP), obesity (Ob), polydactyly (PD), developmental delay (Dev), and renal (Ren) and gonadal malformations (Gon). HBD, homozygous by descent; the “fs” annotation for some mutations indicates a frameshift; IVS, intervening sequence. *, **, ***, and **** indicate recurrent mutations.

View this table:

The first hypothesis is unlikely because changes caused by several of the “singleton” mutant alleles are predicted to be severe (Table 1). For instance, conceptual translation of the single mutation in pedigree AR124 causes a frameshift affecting a valine at position 158 and then results in premature termination of translation at position 200 (V158fsX200; Fig. 1 and Table 1). Likewise, the mutation found in pedigree AR050 was an arginine to STOP (R275X), also causing premature termination (Table 1). Furthermore, this R275X mutation was also found in the homozygous state and segregated with the disorder in pedigree PB005 (20). None of these alterations were found in 192 ethnically matched control chromosomes.

Figure 1

Pedigrees with one BBS2 mutant allele that have been excluded genetically from the BBS2 region. In AR124, the two affected siblings carry different maternal chromosomes. In AR238, the pedigree is excluded from BBS2 because the single affected sib AR238-04 has an identical haplotype to that of unaffected sib AR238-06. The genotypes across the BBS1critical interval (31) of both pedigrees are also shown, where the haplotypes are consistent with linkage.

The possibility of incidental detection of BBS carriers is even less likely. The frequency of BBS is 1:125,000 to 1:150,000 in the European and North American populations (4). Furthermore, the contribution ofBBS2 to all BBS has been estimated at about 10 to 20% (12, 15, 21), which at Hardy-Weinberg equilibrium leads to an expected carrier frequency of 1:400 to 1:700 for BBS2. Therefore, the association of five mutations with BBS2 (in North American pedigrees AR050, AR124, AR237, AR238, and AR596; Table 1) in a sample of 163 patient DNAs is significant (χ2 = 52.9; P < 0.0001).

We explored the possibility that the second mutation was not located in the ORF or was a large deletion by genotyping microsatellites around BBS2 and examining whether the segregation pattern of the parental chromosomes in any given pedigree with a single mutation was consistent with linkage toBBS2 (22). Six (of eight) pedigrees provided sufficient information for this investigation. Only one pedigree, AR029, was consistent with linkage to BBS2. The remaining five pedigrees could be excluded genetically from BBS2either because the affected individuals shared only one parental chromosome at the locus (e.g., AR124; Fig. 1) or because the affected and unaffected sibs had inherited the same parental chromosomes and both were carriers for the single mutation found. Despite the dense microsatellite spacing, there remained a remote chance that a recombination between any marker and the locus might have led to a deceptive exclusion; we eliminated this possibility by genotyping all individuals with a highly informative single nucleotide polymorphism in exon 3 of BBS2 (Fig. 1). When we used polymorphic microsatellites from all BBS loci to monitor the segregation of the parental chromosomes, we observed that three out of five pedigrees could be excluded from all BBS loci except forBBS1, the major BBS locus, suggesting that these three pedigrees may harbor BBS1 mutations. One pedigree was consistent with mapping only to BBS3, and one pedigree, AR237, could be assigned to BBS6 (Fig. 2A).

Figure 2

Pedigrees carrying three BBSmutations. (A) Pedigree AR237 has two BBS6mutations and one BBS2 mutation; (B andC) the converse is the case for pedigrees NFB14 and AR579.

These data suggest that mutations at more than one locus may be present in BBS patients. We therefore integrated and expanded all sequence data from the BBS2 and BBS6 mutation screens that yielded a complete scan of the ORFs of the two genes, and we constructed detailed haplotypes across both genomic regions. Of the 19 pedigrees with one or two BBS2 mutations, we identified four pedigrees with a third potential mutation (AR237, AR579, NFB14, and PB043). AR237 was reported in our initial discovery ofBBS6, because it harbored a homozygous Tyr37 → Cys (Y37C) mutation inBBS6 (11). We then found a heterozygous Asn70 → Ser (N70S) alteration in BBS2 in this family (Fig. 2A and Table 1). In AR259, we found two heterozygousBBS2 nonsense mutations [Tyr24 → X (Y24X) and Gln59 → X (Q59X)] and oneBBS6 nonsense mutation [Gln147 → X (Q147X)]. Similar findings were seen in pedigrees AR579 and NFB14 (Table 1and Fig. 2, B and C).

The probability that we had detected three mutant alleles by chance is minuscule, given the rarity of this disorder and the observation that BBS6 contributes only ∼4% to BBS (11, 12). That the third allele is not a mutation is possible but is not likely. First, nonsense mutations are unlikely polymorphisms. Second, the Ala242 → Ser (A242S) allele identified in NFB14 has been observed on two other occasions, once in the Old Order Amish, associated with McKusick-Kaufman syndrome (14), and subsequently in a BBS patient, but not in 384 control chromosomes (Table 2). The N70S (BBS2) and Cys499 → Ser (C499S) (BBS6) alleles are more challenging to interpret in the absence of a functional assay for either gene, but neither has been found in 384 control chromosomes. More importantly, however, the cumulative data indicate that, among 19 pedigrees with BBS2mutations, sequence or genetic evidence for involvement of another locus occurs in at least nine families (47.3%). The same is also true for BBS6: Of the eight outbred pedigrees where mutations have been identified (12), three harbor BBS2mutations (37.5%). The combination of the genetic data described earlier and the mutational data for the only two genes known to be mutated in BBS suggest that multiple alleles may act in concert to cause pathogenesis. We therefore propose a model of triallelic inheritance, in which three mutant alleles segregate with the disorder.

Table 2

Summary of the number of the third BBS mutations found in our cohort of 163 patients, and 384 control chromosomes. The exon affected by splice junction mutations is indicated as “x.”

View this table:

In the absence of either complete gene identification for allBBS loci or an animal model, it is not clear whether two mutations at a single locus may sometimes suffice to cause BBS. The recessive segregation pattern in several BBS2 pedigrees initially suggests that this may be the case and that “biallelic” inheritance may occur for some instances. However, the BBS6homozygous Y37C mutation in pedigree AR237 segregates perfectly with the disease, yet this pedigree harbors an additional BBS2mutation (Fig. 2A). In fact, in three additional pedigrees, three disease alleles are identified in affected individuals: twoBBS2 alleles with one BBS6 allele (Figs. 2, B and C, and 3). Consistent with our triallelic hypothesis, individuals with two mutant alleles at BBS2 who are either wild type forBBS6 (AR259-05; Fig. 3) or map genetically to another locus (BBS4 in PB043; Table 1) are unaffected. The absence of these alterations in 384 control chromosomes and the nature of theBBS2 mutations in AR259-05, in particular (Y24X and Q59X), suggest that these are not polymorphisms (Table 2).

Figure 3

Pedigree AR259 carrying three nonsenseBBS mutations. Individual AR259-05 carries two nonsenseBBS2 mutations but no BBS6 mutations and is phenotypically normal.

Inheritance of multiple alleles has been documented previously. Paired single mutations in ROM1 and RDS, for example, are the cause of some cases of retinitis pigmentosa (1). Subsequent functional studies confirmed these genetic observations by modeling the mutations in vitro and showing defective assembly of the ROM/RDS complex (23). Dominant Hirschsprung disease also provides a similar, potentially digenic example, where mutations in both RET and GDNF were found in patients (24, 25). However, given that knockout mice for either gene display the Hirschsprung phenotype (26–29), it is likely that the combination of mutations modulates the severity of the phenotype. Likewise, in recessive junctional epidermolysis bullosa (JEB), mutations in bothCOL17A1 and LAMB3 lead to a phenotype with features of both severe JEB (Herlitz form, associated withLAMB3 mutations) and generalized atrophic benign epidermolysis bullosa (caused by COL17A mutations) (30). In contrast, we have no evidence that the differentBBS mutations modulate the phenotype (Table 1), although in principle, the phenotype/genotype relationship may be the combinatorial result of both the type of mutation at each locus and the number of loci involved.

The frequency of multiallelic inheritance in recessive, genetically heterogeneous disorders is unclear. A model of a dominant loss-of-function mutation requisite to unmask recessive mutations may explain, however, the genetic heterogeneity seen in many instances in inbred populations, where both one dominant and two recessive mutations are required for pathogenesis. This model is similar to the inheritance patterns seen in some complex traits, such as type I diabetes (IDDM1), where a dominant HLA genotype is necessary but not sufficient to cause disease; a second, recessively inherited alteration must also be present (31). The presence of a common dominant “master locus” but variable “partners” may also explain the inconsistency in replicating genetic linkage data from discrete populations in various complex traits such as asthma, diabetes, and schizophrenia, because each population isolate will be enriched for a limited number of the recessive mutations that, by themselves, will not satisfy all genetic criteria for pathogenicity. A modification of the theoretical model of disease transmission in these disorders may provide important insights.

To our knowledge, there is only one other clearly documented mechanistic example with a three-allele requirement to manifest a Mendelian trait, exemplified by the segmental duplication associated with Charcot-Marie-Tooth neuropathy type 1A (CMT1A). That model, however, represents a dominant trait requiring three copies (alleles) of a normal gene to manifest disease (32). Our observations on BBS described herein are distinct from the CMT1A mechanism and different from a digenic model in that a combination of three distinct alleles from two (or more) loci appears necessary and sufficient for trait manifestation. As such, triallelic inheritance may represent a transmission model that bridges classic Mendelian disorders with complex traits.

  • * To whom correspondence should be addressed. E-mail jlupski{at}


Stay Connected to Science

Navigate This Article