Article Text

Original Article
Homozygous mutation in CEP19, a gene mutated in morbid obesity, in Bardet-Biedl syndrome with predominant postaxial polydactyly
  1. Esra Yıldız Bölükbaşı1,
  2. Sara Mumtaz2,
  3. Muhammad Afzal2,
  4. Ute Woehlbier3,
  5. Sajid Malik2,
  6. Aslıhan Tolun1
  1. 1 Department of Molecular Biology and Genetics, Boğaziçi University, Istanbul, Turkey
  2. 2 Human Genetics Program, Department of Animal Sciences, Faculty of Biological Sciences, Quaid-i-Azam University, Islamabad, Pakistan
  3. 3 Center for Integrative Biology, Faculty of Science, Universidad Mayor, Santiago, Chile
  1. Correspondence to Dr Sajid Malik, Human Genetics Program, Department of Animal Sciences, Faculty of Biological Sciences, Quaid-i-Azam University, Islamabad, Pakistan; malik{at} and Professor Aslıhan Tolun, Department of Molecular Biology and Genetics, Boğaziçi University, Istanbul, Turkey-34342; tolun{at}


Background Bardet-Biedl syndrome (BBS) is a ciliopathy with extensive phenotypic variability and genetic heterogeneity. We aimed to discover the gene mutated in a consanguineous kindred with multiple cases of a BBS phenotype.

Methods SNP genotype data were used for linkage analysis and exome sequencing to identify mutations. Modelling and in silico analysis were performed to predict mutation severity.

Results Patients had postaxial polydactyly plus variable other clinical features including rod-cone dystrophy, obesity, intellectual disability, renal malformation, developmental delay, dental anomalies, speech disorder and enlarged fatty liver. The 4.57 Mb disease locus harboured homozygous, truncating CEP19 c.194_195insA (p.Tyr65*) mutation. We also found glioma-associated oncogene homolog 1(GLI1) c.820G>C (p.Gly274Arg) in the homozygous state in most patients. In silico modelling strongly suggests that it is damaging. Also, different combinations of four possible modifier alleles in BBS-related genes were detected. Two are known modifier alleles for BBS, splicing variant CCDC28B c.330C>T and missense MKKS/BBS6 p.Ile339Val, and the others are C8ORF37/BBS21 p.Ala178Val and TMEM67/BBS14 modifier p.Asp799Asp. Some patients carry all those five known/possible modifier alleles. Such variants are highly significantly more abundant in our patients than in a control group.

Conclusion CEP19 encodes a centrosomal and ciliary protein, as all BBS genes do. Another truncating mutation p.Arg82* has been reported as responsible for morbid obesity in a family; however, in the family we present, not all homozygotes are obese, although some are severely obese. The variant in GLI1, encoding a transcription factor that localises to the primary cilium and nucleus and is a mediator of the sonic hedgehog pathway, possibly exacerbates disease severity when in the homozygous state.

  • bardet-biedl syndrome
  • cep19
  • gli1
  • ccdc28b
  • modifier

Statistics from


Bardet-Biedl syndrome (BBS; MIM 209900) is a rare ciliopathy resulting from dysfunction of primary cilia and exhibits extensive clinical variability. The primary features are retinal dystrophy, obesity, postaxial polydactyly, urogenital abnormalities, renal anomalies and intellectual disability (ID).1 2 Additionally, a variety of secondary features are associated with the syndrome including enlarged fatty liver, type 2 diabetes, speech disorder, dental anomalies, developmental delay, brachydactyly and syndactyly. A generally recognised diagnostic criterion is the presence of at least either four primary features or three primary and two secondary features.1 Besides the substantial inter- and intra-familial variability in the types of clinical features,1 3 variable expressivity in the clinical features and reduced penetrance of the causal mutation have been reported.4 5

BBS is notable also for genetic heterogeneity. Only approximately 80% of all patients with BBS studied to date have mutations in known BBS genes, all of which encode proteins involved in the structure, assembly and/or function of the primary cilium.1 6 Twenty-three BBS-related genes are known, and biallelic mutations in 21 (BBS1-21) of them cause BBS (for reviews, see refs 2 3). Heterozygous alleles in one of those, ARL6/BBS3 and in the remaining two BBS genes, that is, CCDC28B and TMEM67, have been reported as modifier alleles that in the heterozygous state could increase the severity of BBS caused by a homozygous mutation in another BBS gene.7–9

Besides the genetic heterogeneity and modifier alleles, genetics of BBS is also complicated by the observation in some individuals that a homozygous mutation can cause BBS only when occurring together with a heterozygous mutation in another BBS gene. The hypothesis that BBS is a digenic, triallelic disease has gained support,3 5 10–12 but there has been opposition also.13–16

We present a Pakistani family with highly variable BBS phenotype. We mapped the disease locus and identified a novel homozygous truncating mutation in 19-kd Centrosomal Protein (CEP19) showing that the CEP19 mutation is responsible for the BBS in the family. Recently, in a large Arab family, another homozygous truncating CEP19 mutation was reported to cause morbid obesity (body mass index (BMI) >40) and azoo/oligospermia, and Cep19 knockout mice were similarly morbidly obese and azoospermic.17 In contrast, the majority of the BBS members of the family we present are not even obese, although some are severely obese.

The large size of the consanguineous family allowed us to search for possible contributing mutations in known BBS-related genes. A very rare homozygous missense mutation is detected in glioma-associated oncogene homologue (GLI1), encoding a transcriptional factor activated by the sonic hedgehog (SHH) signal transduction cascade, which has vital functions during brain development (reviewed in ref 18 19) and is functionally related to the primary cilium.20–22 The protein localises to the distal tip of the primary cilium in addition to the nucleus,22 23 as reviewed in ref24. Impairment of its function is expected to affect the downstream targets of the pathway.18 Furthermore, GLI1 is suggested to interact with GLI3 (STRING tool),25 deficit of which can cause postaxial polydactyly types 1A and B or preaxial polydactyly type IV. Also, a link between postaxial polydactyly observed in patients with BBS and SHH signalling has been reported.26


The study was conducted in accordance with the Declaration of Helsinki and national guidelines.

Family and clinical evaluation

The highly consanguineous kindred is from Southern Punjab, Pakistan. The five-generation pedigree constructed showed that all parents originated from the same village and almost all marriages were consanguineous, suggestive of autosomal recessive inheritance (figure 1). Initial physical and medical examination of the three male and five female patients and their seven unaffected relatives was carried out with the help of physicians at Nishtar Hospital in Multan. Later, five patients (501, 503, 505, 508 and 509) were subjected to chorioretinal evaluation and ultrasonography of vital organs including kidneys, spleen and liver, whereas three (505, 508 and 509) of those also underwent ECG evaluations at the District Headquarters Hospital in DG Khan. Brain MRI was available for one patient (503).

Figure 1

Pedigree of the kindred. A horizontal line above the symbol indicates that the individual was physically examined. *Available for all genetic studies; +Available for mutation analysis; #Exome sequenced individuals.

Genetic studies

DNA extracted from peripheral blood samples of seven affected and seven unaffected relatives were available for the genetic study (see figure 1). SNP genotype data were generated using Illumina Human OmniExpress-24 Bead Chip. Multipoint logarithm of odds (LOD) score calculations were performed assuming a disease frequency of 0.0001 and using marker spacing of 0.01 Mb and sets of 30. Multipoint linkage analysis in autosomal recessive model with full penetrance was performed separately for the two branches of the family because the pedigree exceeded the capacity of software Allegro. Unaffected sibs were not included in order to not overlook regions possibly harbouring candidate variants with reduced penetrance. A new linkage analysis (fine mapping) was performed for the regions yielding LOD scores >3 using all markers and in sets of 10, and identity by descent (IBD) for the haplotypes in the regions with new scores >3 was investigated by genotype comparison on MS Excel and/or haplotype construction. Final linkage analysis at the identified gene locus included the genotype data of all participants. Linkage analysis in a dominant model (80% penetrance) was performed similarly but for branch A only, as the whole pedigree exceeded the capacity of Allegro. For the two regions yielding the highest LOD score of 1.6, linkage analysis was performed also for branch B.

We searched for genomic deletions or duplications in BBS-related genes or in any other part of the genome, shared by at least some of the patients using cnvPartition (V.3.2.0) CNV Analysis Plug-in for Genome Studio. Hg19 build GRC37 map was used throughout the study.

Exome sequencing was performed for one patient in each branch of the family (patients 503 and 509) using the Agilent SureSelect Target Enrichment System and the Illumina HiSeq2000 platform. The reads were mapped to the reference genome using Burrows-Wheeler alignment (0.5.9). Variant calling was performed with Sequence Alignment/Map tools (0.1.14) and Annotate Variation to annotate variants. In addition, the aligned reads in the linked region were scrutinised by Integrative Genomics Viewer for any variants that possibly escaped detection. Bedtools was employed to ascertain that each of the target exons was sufficiently covered in at least one of the sequenced exomes. For a recessive model, variants with frequencies <0.005, as reported in the Exome Aggregation Consortium (ExAC) South Asian samples that contain many Pakistani exome files, and the number of the reads with the variant allele over the number of total reads >0.6, and for a dominant model <0.01 and >0.25, respectively, were filtered. Splicing variants and exonic variants deduced to affect protein function were evaluated. In the search for modifier alleles, we considered both homozygous and heterozygous variants possibly damaging to protein with population frequencies <0.05 in BBS-related genes, as recommended by Richards et al.27 Variants were validated by Sanger sequencing in many participants, and segregation in the family was investigated by Single Strand Conformation Polymorphism (SSCP) analysis. All variants mentioned herein have been submitted to ClinVar database.

ExAC and ESP6500 databases were used to interrogate variant novelty. We used computational prediction algorithms Sorting Intolerant from Tolerant (SIFT), Polymorphism Phenotyping V.2 (PolyPhen-2) and Mutation Taster to assess the overall impact of the variants.28–30

To test significance of difference, we used unpaired one-tailed Student’s t-test to compare the total number of possibly damaging variants with frequencies <0.05 in the BBS-related genes (we detected in the two exome files) in our patients to the total number of similar alleles in the same genes in 10 unrelated Pakistani control exomes.

Structural analysis of GLI1 Gly274Arg mutant

Sequence alignment was performed using online tool PRALINE ( Structural analysis was performed using the SWISSMODEL server ( and Chimera1.11rc imaging software. Protein Data Bank structure 2gli.1.C of the five-finger GLI1/DNA complex was used as a template to model the structures of wild-type and mutant GLI1 proteins. To display the structure of DNA-bound wild-type GLI1, tfmodeller was used.31 32


Clinical findings

Patients had variable clinical features including polydactyly, rod-cone dystrophy, obesity, ID, renal malformation, dental anomalies and developmental delay. Clinical findings in the eight affected and seven unaffected members of the family are compiled in table 1. All patients had polydactyly, mostly postaxial type A in the feet and postaxial type B in the hands, which were bilateral and most commonly symmetrical (figure 2). The second prominent feature was rod-cone dystrophy, found in four of the five patients evaluated, one of whom also had optic nerve atrophy. Three of those five patients also had exotropia of right eye, and three had enlarged fatty liver. Obesity was observed in three of the eight patients, and three others were overweight, whereas only one of the seven unaffected relative was obese and two were overweight. Four patients had ID, two with severe ID, developmental delay and unclear speech/dysarthria and the other two with mild ID and obesity. The three patients who had developmental delay also had hypoplastic teeth with irregular dentation; one of them additionally had renal parenchymal disease and another had aggressive and hypertensive behaviour, hyperphagy, insomnia, low hairline, 2/3 toes syndactyly and an over-riding toe. Six of the seven patients investigated were unable to perform strenuous activities and had shortness of breath. Three of those patients underwent ECG, with unremarkable results. Lipid profiling performed in two of those six patients revealed high blood cholesterol, and one of those patients also had high blood glucose, indicative of diabetes (see online supplementary table S1). Blood cell count performed for three patients revealed normal or near-normal levels for most of the variables except that peculiarities in blood cell morphology indicated anaemia (see online supplementary table S2). In the female patient assayed (508), serum estradiol level was found normal (see online supplementary table S1), indicative of normal ovarian function. Another female patient (501) had retroverted uterus.

Supplementary file 1

Figure 2

Pictures of autopods of patients.

Table 1

Characteristics of the family members

The most severely affected of all is index patient 503, a 25-year-old man with severe ID, early features of retinitis pigmentosa, renal paranchymal disease, unclear speech, dental anomalies and developmental delay. He crawled and walked at the age of 12 years. He has weak lower limb muscles and difficulty standing up and can walk only with support. He also had delayed puberty, but a hormonal assay could not be performed. The features he does not share with any relative are sparse hair, enuresis, enlarged head and frontal bossing. He moves his head in a circular motion while sitting. He eats only bread and no cooked food, and cries when hungry. Usually he is friendly towards strangers. Brain MRI results are normal (see online supplementary figure S1).

Supplementary file 2

Genetic findings

Locus and gene identification

Initial multipoint linkage analysis in a recessive model using SNP genotype data of three affected and two unaffected relatives excluded all known BBS loci. We then launched a search for the disease locus. Linkage analysis not including the unaffected sibs and subsequent fine mapping (using all markers) at the loci with high LOD scores yielded only one region >1 Mb, with the highest cumulative LOD score of 5.40 (see online supplementary figure S2, table 2), mapping the disease gene to a 4.57 Mb region between rs2630239 and rs1147240 at 3q29. When all sibs were included, a maximal LOD score of 5.65 was obtained (see online supplementary figure S3). We did not find any additional candidate locus by homozygosity mapping using Homozygosity Mapper. In the exome files of two affected cousins (503 and 509), we investigated with priority the homozygous rare and novel variants in the identified gene region that could potentially affect the protein (see online supplementary table S3). Novel truncating mutation CEP19 c.194_195insA (p.Tyr65*) (NM_032898) appeared as the best candidate, because it is deduced as damaging (the insertion of A creates a termination codon) and the gene encodes a ciliary protein as all other BBS genes do. All tested patients but no unaffected relative was homozygous for it (figure 3). Examples of Sanger sequencing results for all variants considered relevant to BBS phenotype are presented in online supplementary figure S4.

Supplementary file 3

Supplementary file 4

Figure 3

Genotypes of family members for the six variants that we considered relevant to the disease. *Available for genetic studies; +Available for mutation analysis; #Exome sequenced individuals.

Table 2

LOD scores for branch A plus branch B of the kindred in the maximal shared homozygosity regions at loci yielding relatively high LOD scores

Searching for modifier variants

We first searched for possible modifier alleles possibly increasing disease severity in the two exome files in regions yielding LOD scores >2 in branch A plus >0 in branch B (table 2). The most striking is the very rare missense GLI1 c.820G>C (p.Gly274Arg) (NM_005269) at 12q13.3. Four of the seven patients tested were homozygous for it, and two others plus three unaffected relatives were heterozygous (figure 3). Computational algorithms SIFT, PolyPhen-2 and Mutation Taster predicted the variant to be deleterious, possibly damaging and disease causing, respectively. It is reported in only ExAC Latino and non-Finnish European samples, with frequencies of 0.00025 (three alleles in 11572) and 0.000015 (one allele in 66280), respectively. The substituted glycine residue at position 274 as well as the neighbouring 103 residues of the 1106 amino acid proteins are completely conserved across mammals. Conservation of the amino acids altered by the mutations we detected are presented in online supplementary figure S5.

Supplementary file 5

We searched for possibly damaging variants with frequencies <0.05 in the 23 known BBS-related genes that could increase disease severity and detected four such variants (see online supplementary table S4). Two were known BBS modifier alleles: splicing CCDC28B c.330C>T (formerly c.430C>T; p.Phe110Phe; rs41263993) (NM_024296) and missense MKKS/BBS6 c.1015A>G (p.Ile339Val; rs137853909) (NM_018848). Two other variants are also in BBS-related genes. C8ORF37 c.533C>T (p.Ala178Val; rs375314973) (NM_177965) is predicted as deleterious, possibly damaging and disease causing, respectively. Residue Ala178 is completely conserved across species. Adjacent Arg177 has been reported as altered in BBS and isolated recessive retinal dystrophy.33 The other is synonymous TMEM67 (BBS 14, modifier) c.2397T>C (p.Asp799Asp) (NM_153704), which was predicted as disease causing (donor marginally increased) by Mutation Taster; SIFT and PolyPhen-2 are not applicable for synonymous changes. Father 403 homozygous for this variant did not have an obvious phenotype and died at the age of 53 years of infection of a wound. These four variants are present in different combinations in patients and unaffected relatives (figure 3). Patients are either heterozygous or non-carriers, except that patient 502 is homozygous for the CCDC28B allele.

To assess whether the mutational load in BBS-related genes in our patients are heavier than in a control group, we screened 10 unrelated Pakistani exomes for exonic and splicing variants with frequencies <0.05, as we did to search for modifier alleles, in the known BBS-related genes plus GLI1. The total number of those known/possible modifier alleles ranged from zero to three, and the average was 1.10 (11/10) (see online supplementary table 5). The number of such variants ranged from one to six in the seven patients and from one to four in their total of three unaffected siblings. The mean was 4.43 per patient (31/7) and 2.67 (8/3) per unaffected sibling (figure 3). Applying unpaired one-tailed Student’s t-test to the patient and control groups, t value is calculated as 4.58 and P value as 0.00018. Thus, the mean of variants in patients was extremely significantly higher (P<0.001) as compared with the mean of the control group.

Supplementary file 6

We also investigated whether any candidate variant with dominant inheritance was shared by at least some of the patients. Multipoint linkage analysis in branch A yielded a maximal LOD score of 1.6 at 8p23.1 and 12p11.2, but branch B did not yield positive LOD scores in any of those regions (data not presented). Nonetheless, in those regions, neither of the exome files contained any candidate modifier alleles.

Structural analysis of GLI1 p.Gly274Arg

We analysed the available three-dimensional structures for GLI1 to define the possible structural and functional consequences of the identified mutation. The amino acid change at position 274 from neutral non-polar glycine to polar hydrophilic arginine results in the gain of a positive charge in the second of the five zinc fingers (figure 4A). GLI1 is a sequence-specific DNA-binding protein acting as a transcription factor that recognises specific sites in the genome.34 The structure of the five zinc fingers of human GLI1 bound to a 21 bp DNA fragment has been reported35; zinc fingers 2–5 fit into the major groove of the DNA and wrap around the helix (figure 4B), whereas zinc finger 1 does not bind to DNA but rather interacts extensively with zinc finger 2. To assess the overall impact of the Gly274Arg substitution on the structure of the five zinc fingers, we overlaid the backbones of the reported wild-type model and the predicted mutant model (figure 4C). The arrangement of the zinc fingers is significantly shifted in the mutant, indicating possible functional consequences. Mutant Gly274Arg displays a more closed structure, where specifically the substitute arginine seems to block access of the DNA binding domain to DNA molecule, which is mainly dependent on residues Lys350, Arg354, Lys380 and Arg381, reported to be essential for DNA binding (figure 4D).36 This observation suggests that mutant GLI1 might be severely impaired in its DNA-binding ability and, therefore, in its function as a transcriptional activator.

Figure 4

Prediction of the effect of Gly274Arg mutation on the structure of GLI1. (A) Amino acid sequence alignment of residues 235–388 (154 residues) in wild-type GLI1 Gly274 and mutant GLI1 Arg274. Red and blue highlighted residues are negatively and positively charged amino acids, respectively. The site of Gly274Arg is indicated by a red box and an arrow. Substitution of Gly274 by Arg leads to a gain of a positive charge. (B) Structure of the GLI1/DNA complex. Location of Gly274 is indicated in green. (C) Structural overlay of residues 235–388 of wild-type GLI1 Gly274 (green on white backbone) and mutant GLI1 Arg274 (red on yellow backbone). A shift in the backbone structure was predicted for the mutant (D) Surfaces of wild-type and mutant GLI1, represented in white and grey, respectively. Locations of wild-type Gly274 (left image, in green) and mutant Arg274 (right image, in purple) are indicated. Lysine residues Lys350 and Lys380 and arginine residues Arg354 and Arg381 known to be essential for binding of GLI1 to DNA are highlighted in red, orange, dark orange and yellow, respectively.


The feature consistent with BBS in all patients was polydactyly, and other primary BBS features were variable. Three of the five patients who underwent extended clinical investigations (503, 508 and 509) presented with four or five primary BBS features plus two to five secondary features investigated for, and another (505) presented with three primary and four secondary features, all easily meeting BBS diagnosis.

Our results show that the identified CEP19 mutation underlies the BBS afflicting the family. The truncating, recessive mutation is novel and deduced to lead to the deletion of 103 of the total 167 native amino acids. The amino acid sequence of CEP19 is highly conserved across species, with total conservation among humans, chimpanzee and Rhesus macaque, whereas conservation between human and mouse is 97% and human and rat is 86%. No orthologue of CEP19 is found in Drosophila, Caenorhabditis elegans or yeast, using the program Geneious V. Another homozygous truncating mutation (p.Arg82*) in CEP19 caused morbid obesity (maximal BMI range was 36.7–61) developing at age 3 years in the family identified.17 In contrast, the mutation we identified did not cause morbid obesity in any of our homozygous patients. The gene is possibly not widely associated with morbid obesity, as no other morbidly obese individual with CEP19 mutation has been reported to date. The homozygous men in the previous family were additionally azoo/oligospermic; unfortunately, we were unable to perform sperm count in the presented family. Of note, coincidingly, some of the homozygotes (3/11) in the morbidly obese family had ID as in the family we present, and similarly, serum cholesterol was high in some members of both families.

Mutations in the same BBS gene can cause diseases with overlapping phenotypes. For example, biallelic mutations in BBS14/CEP290 can cause either isolated retinopathy or BBS.38 The situation for CEP19 is not similar. The previous homozygous truncating mutation (p.Arg82*) caused morbid obesity,17 but the mutation we identified did not cause morbid obesity in any of the homozygotes. Moreover, morbid obesity is not considered a feature of BBS.

Two very recent studies provide an explanation of how a CEP19 mutation can cause BBS. CEP19 was found in a stoichiometric complex with RAB2LB, a highly conserved guanosine triphosphatase (GTPase), which is recruited to the centriole by CEP19 at the base of cilia. RABL2B binds GTP, and then binds to intraflagellar transport B (IFT-B) complex, triggering its entry into the primary cilium.39 Thus, CEP19-RABL2B-IFT pathway is proposed as a new molecular mechanism for directed ciliary traffic, needed for primary cilium formation. These new findings suggest that the loss of function by truncation at Tyr65* in our patients is detrimental for proper primary cilium function. Interestingly, Rabl2 knockout mouse displays features characteristic of ciliopathies that include infertility, obesity, polydactyly and retinal degeneration.39 As these phenotypes are reminiscent of BBS, we propose that loss of function of CEP19 can cause BBS due to failure to recruit RABL2B, mimicking a RABL2B knockout. RAB2LB binding site within CEP19 is at the C-terminus of CEP19, which is lacking in the Y65* truncation mutant.40 Curiously, Shalata et al 17 reported truncation of CEP19 at Arg82 as the cause of obesity not associated with other ciliopathic features. It remains to be investigated why the mutation we identified (Tyr65*) results in a different and more complex phenotype in humans. Could it be due to the additional loss of amino acids and thereby loss of more functions, or instead, has a gain of function occurred? Both hypotheses could be concordant with the CEP19 knockout mouse model. Another hypothesis would be that truncation of CEP19 causes BBS only in epistasis with a mutation in another gene (a triallelic model), and the patients of the reported family do not have such a mutation.

Some BBS families have their own prominent features. For example, LZTFL1 (BBS17) mutation causes mesoaxial polydactyly as the main feature, and SDCCAG8 (BBS16) mutation causes absent polydactyly but penetrant renal disease.41 42 The family we present exhibits polydactyly as the prominent feature; all patients have it. Polydactyly is considered the third primary BBS feature, manifesting in 63%–81% of all BBS cases and is usually postaxial and rarely mesoaxial, similar to the phenotype in the presented family.1 The first primary BBS feature is moderate-to-severe vision loss and affects >90% of all patients leading to blindness generally before the age of 20 years. Progressive rod-cone dystrophy is detected in four of our five patients investigated. None of those patients is totally blind at present, but one patient (505) has lost most of her sight in the right eye. The second primary BBS feature is obesity, manifesting in 72%–92% of the cases, but less than half of our patients are obese, despite that CEP19 has been reported as responsible for morbid obesity in a family.17 The prevalence of ID, another primary feature, is 50%–61% and is found in one-third of our patients. We did not have the opportunity to investigate for genital anomalies. As for secondary BBS features, three of the five patients investigated have hepatic disease of enlarged fatty liver. Two of our eight patients have speech disorder, and three have dental anomalies.

Several researchers have reported BBS cases that carried a third trans mutation that modulate disease penetrance and expressivity through an additive and/or epistatic effect.7–9 43 The large size of the presented family provided an opportunity to investigate for such mutations. GLI1 p.Gly274Arg is possibly a modifier allele, a rare example of a homozygous one. Both the altered residue Gly274 and the amino acid sequence of the entire protein are highly conserved among species. Amino acid conservation between human and chimpanzee is 99%, Rhesus macaque is 97%, mouse is 85% and rat is 86%. Structural modelling indicated that this amino acid substitution in the second of the five zinc fingers may lead to the loss of the DNA-binding function of the protein (figure 4). Two of our four homozygous patients (503 and 508) were severely affected, whereas the other two (501 and 507) were not so severe. The patient who does not carry the mutation at all (508) was also rather severely affected. Thus, a genotype–phenotype correlation was not observed, as was also the case with the known CCDC28B modifier in our patients (discussed below). GLI1 is implicated in a variety of cancers (MIM 165220), but there is no cancer history in the family.

Two other variants we found are known BBS modifier alleles (figure 3). Patients with BBS carrying splicing CCDC28B c.330C>T as a third BBS allele were reported to be more severely affected.8 As compared with control individuals a significantly higher fraction of the total 226 patients with BBS screened were found to carry this variant, and a strong overtransmission of the variant was observed. In contrast, the variant had not led to a more severe phenotype in 52 other patients.44 We did not find a genotype–phenotype correlation; homozygote 502 had a rather mild phenotype, as he had polydactyly only in the feet, did not have ID or developmental delay and was not obese. The variant resulted in only approximately 30% of wild-type mRNA molecules in an unaffected homozygote.8 ExAC reports its frequency in South Asians as 0.01075 (102 alleles in 9484). MKKS c.1015A>G (p.Ile339Val) was found in two unrelated patients with BBS in the heterozygous state but not in an affected sibling.45 46 Residue Ile339 is not highly conserved in mammals. Computational algorithms SIFT, PolyPhen-2 and Mutation Taster predicted the variant to be tolerated, benign and polymorphism, respectively, and ClinVar lists it as likely benign (see online supplementary table S4). ExAC reports its frequency as 0.006275 (41 alleles in 6534, three homozygotes) in Europeans and as 0.00604 (98 alleles in 16 224, two homozygotes) in South Asians. It might be a modifier allele in our patients as well, as the severe cases are among the heterozygotes for the allele and those with milder phenotypes are non-carriers. Missense C8ORF37 (BBS21) variant and synonymous TMEM67 variant are both predicted to be damaging by algorithms and have ExAC South Asian frequencies of 0.000548 (nine alleles in 16 422) and 0.02291 (378 alleles in 16 498), respectively. We were unable to make any prediction of a potential structural and/or functional impact of these alleles since the protein structures were not available.

All of our patients carry at least one modifier/possible modifier allele, and the most severe of them (503) carries all five of those alleles, one in the homozygous state, but patient 501 with mild phenotype has the same genotype. It might well be that there are yet other modifier alleles in the family. Hence, our findings do not refute oligogenic inheritance (as no patient is without a modifier allele), but whether it supports modifier hypothesis is not clear, as a straightforward genotype–phenotype correlation was not observed. We attempted to assess whether the modifiers/possible modifiers we found might be circumstantial and evaluated 10 unrelated Pakistani exomes as controls. We searched for exonic and splicing variants similarly in the same genes and found that patients had a highly significantly more of such variants (P<0.001). Of note, this difference is possibly an underestimation, as in five patients we considered only those variants we detected in the exome files of the other patients, which were the only patient exome files available.

Whether a trans allele in a patient just exacerbates the BBS phenotype or is essential for the biallelic mutations in a BBS gene to manifest BBS can be answered only after all BBS genes are identified and the mutation search is performed by exome or genome sequencing and not by candidate approach. Also, deletion and duplication analysis in the known and candidate BBS genes needs to be performed, as exon-destructive copy-number variations also contribute to the mutational load in BBS.43 Analysis of the SNP data of our patients did not reveal any deletions or duplications in such genes.

It has been proposed that in some patients an increase in the ciliary mutational load renders the biallelic mutations penetrant. For example, a father homozygous for splicing variant CCDC28B c.330C>T was not affected, and another unaffected father was homozygous for BBS1 Met390Arg (as his BBS sibs) but not carrying the CCDC28B variant.8 We cannot speculate whether the GLI1 variant we identified is damaging in the homozygous state in humans, as we did not observe homozygosity in any unaffected individual. Of note, Gli1 knockout mice are viable and seem normal.47 We did not find any reported evidence for interaction between CEP19 and GLI1. Future studies could unravel whether the proteins of the three possible modifier genes we identified herein interact with CEP19 and/or one another and the variants contribute to the disease phenotype. The finding of a high number of modifier/possible modifier alleles in our patients supports cumulative mutational load hypothesis. A model of total mutational load of ciliary signalling was proposed to explain complex inheritance in BBS,8 and phenotypic variability was proposed to be due to a different mutational load in cilia-associated genes among siblings.48

In summary, we identified CEP19 p.Tyr65* as responsible for BBS in the family and GLI1 Gly274Arg as a possible modifier of severity, and the three variants in other BBS-related genes could be exacerbating the disease severity. We hope that our findings would facilitate the furthering of our understanding of the molecular processes underlying BBS. Including CEP19 in BBS screening panels could benefit families, considering that only approximately 80% of all patients with BBS studied to date have mutations in known BBS genes,1 6 and usually polydactyly is the only obvious feature at birth, and hence, the disease is diagnosed in late childhood.


We would like to thank the family members for their cooperation. We would also like to thank Rashid Qureshi and Muhammad Sadiq for making the extended clinical evaluation possible.


View Abstract


  • Contributors AT and SajM were responsible for the concept and design of the study. EYB, SarM and UW generated and interpreted the data. SajM and MA contributed clinical data. AT, SajM, EYB and UW drafted the manuscript. All authors revised the manuscript.

  • Funding This research was supported by Boğaziçi University Research Fund (project 10860; to AT), the Scientific and Technological Research Council of Turkey (114Z829; to AT) and URF-QAU, Pakistan (to SajM). UW is supported by FONDECYT no. 1150743.

  • Competing interests None declared.

  • Patient consent Guardian consent obtained.

  • Ethics approval Informed consent was obtained from/for participants in accordance with the regulations of the Ethical Review Committee of Quaid-i-Azam University and Boğaziçi University Institutional Review Board for Research with Human Participants, both of which approved the study protocol.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.