Owing to its growing prevalence, obesity forms a global threat to public health and a burden to healthcare systems. Interventions by weight loss seem unable to bring this pandemic to a stop. Therefore, finding novel ways for treatment and prevention is a must, but depends on profound knowledge of the aetiology of obesity. Notably, the background of obesity is heterogeneous and complex, with involvement of genetic and environmental factors. Various physiological processes may be involved, including the response to environmental food cues, the hormonal regulation of hunger and satiety, the activity of the central reward system, whole-body energy expenditure and the storage capacity for fat in the adipose tissue. Each of these processes is related to the activity of a particular set of genes with some genes being involved in a broad spectrum of processes. Such genes are of special interest, because the pleiotropic effect of variation in those genes might explain a broader part of the aetiology of obesity. At the same time, those genes might be efficient targets for intervention.
Ciliary genes, which code for proteins required for the compartmentalised cilia biogenesis and function, represent such a group and have recently gained interest in the context of obesity. Cilia are specialised organelles extruding from eukaryotic cells (Fig. 1; more detailed information about the cilium structure and composition is available from the literature)( Reference Benzing and Schermer 1 – Reference Valente, Rosti and Gibbs 7 ), and are the product of a process that is referred to as ciliogenesis( Reference Garcia-Gonzalo and Reiter 8 ). There are two types of cilia, motile and immotile/primary. Besides well-known motile cilia in the respiratory tract and the oviducts and immotile cilia of the inner ear and the nasal epithelium, almost every cell in the body carries a single primary cilium at a certain stage of its life cycle. Cilia regulate the proper differentiation and migration of all kinds of cells in the body, are involved in signal transduction and allow cells to obtain environmental information as part of sensory systems( Reference Boekhoff, Tareilus and Strotmann 9 – Reference Schneider, Clement and Teilmann 13 ). Mutations in ciliary genes can cause monogenic disorders referred to as ciliopathies with pleiotropic consequences for the phenotype, which can include obesity. In this study, we focus on the obesity-linked subset of ciliary genes, review the clinical presentation of mutations, location of the ciliary proteins in the cell and their most likely function, and describe mechanisms by which they might contribute to the genetic background of obesity. In addition, we present our own experimental data of those genes in recent obesity and weight-loss studies.
Fig. 1 Schematic representation of the cilium and the intraflagellar transport. 1, Axoneme; 2, basal body; 3, pericentriolar satellite; 4, transition zone; 5, transition fibre; 6, assembly complex; 
, membrane receptors; 
, structural Bardet–Biedl syndrome (BBS) proteins; 
, BBSome; 
, BBS3; 
, intraflagellar transport (IFT)-A; 
, IFT-B; 
, kinesin; 
, dynein.
Clinical presentation of obesity-linked ciliary genes
Because of the multiplicity of cilia functions, ciliopathy patients usually suffer from a range of symptoms and their condition is often classified as a syndrome. Examples of such conditions are the autosomal recessive Bardet–Biedl syndrome (BBS; MIM209900) and Alström syndrome (MIM203800). Among patients with different clinical entities due to a different mutated gene, symptoms may overlap( Reference Baker and Beales 14 ). For instance, BBS patients can present with retinitis pigmentosa, truncal obesity, renal dysfunction, polydactyly, behavioural dysfunction and hypogonadism as major symptoms, but also with diabetes, hepatic complaints and olfactory and/or auditory deficiencies. Alström syndrome is characterised by blindness, hearing loss, childhood obesity with hyperinsulinaemia and type 2 diabetes mellitus, but also by cardiomyopathy and by renal, pulmonary and/or hepatic dysfunction. Although mutations in different genes may cause similar symptoms, it is also possible that different mutations in the same ciliary gene lead to different clinical diagnoses. For instance, patients with mutations in the Meckel syndrome type 1 (MKS1) gene can present with BBS or Meckel–Gruber syndrome, whereas various mutations in CEP290 may lead to BBS, Joubert or Meckel syndrome as well as Leber congenital amaurosis( Reference Leitch, Zaghloul and Davis 15 ).
Some ciliopathy syndromes are genetically heterogeneous such as BBS, for which nineteen genes have already been reported( Reference Forsythe and Beales 16 ). For other ciliopathy syndromes, only a single gene has been identified, as is the case for Alström syndrome. Other single-gene, obesity-linked ciliopathies are a syndrome with mental retardation, truncal obesity, retinal dystrophy and micropenis (MIM610156)( Reference Hampshire, Ayub and Springell 17 , Reference Jacoby, Cox and Gayral 18 ) and a syndrome with morbid obesity and spermatogenic failure (morbid obesity syndrome 1 or morbid obesity and spermatogenic failure syndrome, MIM615703). The latter is a condition with morbid obesity in humans, and in the mouse presents with obesity, hyperphagia and insulin resistance( Reference Shalata, Ramirez and Desnick 19 ).
A patient is diagnosed with BBS when at least four of the following six major manifestations are scored: retinopathy, obesity, polydactyly, genital abnormalities, cognitive impairment and renal anomalies. It means that BBS patients do not necessarily have to be obese. In Table 1, we have listed various clinical conditions based on mutations in ciliary genes, and we have indicated for which of those genes obesity has actually been reported. It shows that for two of the BBS genes, BBS11 and BBS15, obesity has not been reported as a symptom among patients. Interestingly, Bbs11/Trim32-knockout mice also do not become obese, although adult male mice have a 10 % increased body weight originating from the non-muscle mass( Reference Kudryashova, Wu and Havton 77 ). When a mutated ciliary gene leads to Joubert syndrome, obesity is commonly not part of the symptoms. However, recently, a patient with Joubert syndrome and obesity has been reported with a homozygous missense mutation in the ARL13B gene, which affects ciliogenesis( Reference Thomas, Cantagrel and Mariani 21 ).
Table 1 Clinical syndromes due to mutations in ciliary genes
Id., idem; COACH, cerebellar vermis defect, oligophrenia, ataxia, coloboma, hepatic fibrosis; MOSPFG, morbid obesity and spermatogenic failure; MORM, mental retardation, truncal obesity, retinal dystrophy and micropenis. For expansions of gene names, see the text or the abbreviations list.
Besides clinical presentation, animal studies may also reveal new interesting genes in this cilia–obesity context. Recently, it was shown that knocking out the ankyrin repeat domain 26 (Ankrd26) gene in the mouse leads to defects of primary cilia in regions of the central nervous system, accompanied by hyperphagia, obesity and gigantism( Reference Acs, Bauer and Mayer 78 ). The human ANKRD26 gene is situated in a locus for obesity on chromosome 10p with a maternal parent-of-origin effect( Reference Dong, Li and Geller 79 ). However, mutations in this gene have not yet been described in connection to obesity in humans. Another gene of which mutations lead to symptoms including obesity in the mouse is Tub ( Reference Noben-Trauth, Naggert and North 80 ). The protein of this gene is probably involved in the selective import of G protein-coupled receptors into cilia( Reference Sun, Haley and Bulgakov 81 ).
Functions of obesity-linked ciliary proteins and their closely interacting partners
The BBSome
The elucidation of the underlying genes in BBS led to the discovery of a special protein complex, the BBSome( Reference Nachury, Loktev and Zhang 82 ). The function of the BBSome includes, but may not be limited to, the intraciliary (intraflagellar) trafficking of a diverse set of molecules (Fig. 1). This includes membrane-embedded receptors for signal transduction as well as structural components of the cilium. As such, the BBSome is important for the establishment, growth, turnover and functioning of cilia( Reference Jenkins, McEwen and Martens 83 , Reference Williams, McIntyre and Norris 84 ). BBS1, BBS2, BBS4, BBS5, BBS7, BBS8 and BBS9 are part of the BBSome together with the 10 kDa BBSome interacting protein 1 (BBIP1), also referred to as BBIP10 or BBS18( Reference Scheidecker, Etard and Pierce 44 , Reference Jin and Nachury 85 ).
Chaperonin activity
Not all BBS genes code for components of the BBSome; three of the BBS genes code for chaperonin-like proteins: BBS3, BBS10 and BBS12. They share sequence homology with genes coding for members of the chaperonin containing TCP1 (CCT)/TCP1 ring complex (TRiC) family of chaperonins, which take part in ATP-dependent protein folding. The three BBSome proteins associate with six CCT-chaperonins, CCT1, CCT2, CCT3, CCT4, CCT5 and CCT8, to form a complex that is involved in the assembly of the BBSome( Reference Seo, Baye and Schulz 86 ). The assembly starts with the interaction of BBS2 with BBS7 and BBS9, forming a core particle to which the other structural BBS proteins attach( Reference Zhang, Yu and Seo 87 ). BBS18 appears to be important for enabling the incorporation of BBS4 into the BBSome( Reference Loktev, Zhang and Beck 88 ). It was observed that BBS2 is subject to ubiquitination. Therefore, turnover of BBS2 via the ubiquitin-proteasome pathway might be important for the regulation of BBSome quantity. BBS11/tripartite motif containing 32 (TRIM32), which has been shown to be an E3 ubiquitin ligase for actin and dysbindin, was suggested to be the processor of BBS2( Reference Zhang, Yu and Seo 87 ).
Intraflagellar transport
To enable intraflagellar transport (IFT), two other protein complexes are needed, IFT-A and IFT-B. The IFT-A particle is composed of IFT43 (c14orf179), IFT121 (WDR35), IFT122 (WD repeat domain 10; WDR10), IFT139, IFT140 (uncharacterized KIAA gene 0590; KIAA0590) and IFT144 (WDR19)( Reference Behal, Miller and Qin 89 ), whereas the B-particle is composed of IFT20, IFT25, IFT27, IFT46, IFT52, IFT57, IFT72, IFT74, IFT80, IFT81, IFT88 and IFT172 (selective LIM-binding factor; SLB)( Reference Cole, Diener and Himelblau 60 ). Mediated by IFT144, the BBSome can interact with the IFT-A and IFT-B particles to form the basal transport complex (BTC), which moves along the central axoneme (Fig. 1). Membrane-embedded receptor proteins at the base of the cilium can dock to the BTC, and are then transported to the tip of the cilium. Although much of what we know about the function of these proteins has been collected from studies on Chlamydomonas and Caenorhabditis elegans, a similar model seems to be operative in mammalian cilia( Reference Williams, McIntyre and Norris 84 ).
In a more advanced model, the IFT-B particle in connection with kinesins such as kinesin II and its component kinesin family member 3A (KIF3A)( Reference Rosenbaum and Witman 90 ) drives the anterograde transport of proteins to the tip of the cilium, from which the IFT-A particle in connection with dyneins drives the retrograde transport( Reference Davis and Katsanis 91 ). Observations in the Ift27-knockout mouse indicate that IFT27, a component of the IFT-B particle, together with BBS17/leucine zipper transcription factor-like 1 (LZTFL1), facilitates the retrograde transport of BBS proteins and sonic hedgehog (SHH) receptors from the cilium( Reference Eguether, San Agustin and Keady 92 ). In vitro, Liew et al.( Reference Liew, Ye and Nager 93 ) found that IFT27 can bind to unloaded BBS3/ADP-ribosylation factor-like 6 (ARL6). BBS3/ARL6, after being loaded with GTP, is assumed to link the BBSome to the membrane( Reference Nachury, Seeley and Jin 94 , Reference Wiens, Tong and Esmail 95 ). What may happen is that, when the BTC reaches the top of the cilium, BBS3/ARL6-bound GTP is hydrolysed, perhaps by the action of the Rab-like GTPase IFT27, and as a consequence BBSomes are released from the membrane. Next, the release of IFT27 allows the activation of BBS3/ARL6 by GTP binding and attachment of BBSomes to the membrane, in order to start the export of cargo proteins such as SHH receptors from the cilium( Reference Liew, Ye and Nager 93 ). Studies in the mouse and in zebrafish have suggested that another protein, clusterin associated protein 1 (CLUAP1), is also associated with IFT, possibly as part of the IFT-B particle, and may be involved in regulating the transport at the base and tip of the cilia( Reference Botilde, Yoshiba and Shinohara 96 , Reference Lee, Wallingford and Gross 97 ). The knockout mouse of this gene has a severe phenotype and dies at embryonic mid-gestation( Reference Botilde, Yoshiba and Shinohara 96 ).
The basal area of the cilium
On the basal area of the cilium, three substructures can be distinguished: the basal body with transition fibres, the transition zone and the pericentriolar satellite. It is the area where many BBS proteins are active, and where the BBSome is assembled and the cargo proteins are uploaded or unloaded; two obesity-linked ciliary proteins, Alström syndrome protein 1 (ALMS1) and centrosomal protein 19 kDa (CEP19), are located at the basal body of the cilium and the subdistal centriolar appendage (transition fibre), respectively( Reference Shalata, Ramirez and Desnick 19 , Reference Gupta, Coyaud and Goncalves 98 – Reference Jagger, Collin and Kelly 100 ). Absence of Alms1 in the mouse leads to changes in shape and orientation of cilia in hair cells( Reference Jagger, Collin and Kelly 100 ). Disruption of Alms1 in drosophila induces over-activation of the Notch signalling pathway, similar to after disruption of BBS1, BBS3 or BBS4, with the accumulation of Notch receptors in endosomes( Reference Leitch, Lodh and Prieto-Echague 101 ). This observation suggests that ALMS1 and BBS proteins are involved in endosomal cycling and breakdown of signal transduction receptors occurring at the base of the cilium.
The transition zone is a region that forms the border between the cilium and the cell and prevents the free exchange of proteins between the cilium membrane and the plasma membrane( Reference Otto, Hurd and Airik 42 ). Several proteins have been localised to this area, such as BBS13, BBS14, BBS15 and BBS16. In addition, particular complexes have been identified in this region such as the TCTN complex, which contains tectonic family member 1-3 (TCTN1-3), B9 protein domain 1 (B9D1), BBS13, BBS14, coiled-coil and C2 domain containing 2A (CC2D2A), transmembrane protein 67 (TMEM67) and TMEM216, and the HPNP complex composed of HPNP1, 4, 8 and retinitis pigmentosa GTPase regulator interacting protein 1-like (RPGRIP1L)( Reference Garcia-Gonzalo, Corbit and Sirerol-Piquer 2 , Reference Sang, Miller and Corbit 4 , Reference Huang, Szymanska and Jensen 76 , Reference Williams, Li and Kida 102 ). It is suggested that they may function as gatekeepers for proteins that, being linked to the BBSome, are to be transported into and out of the cilium.
Investigations in embryos of knockout mice suggest that Rpgrip1l, which is located at the transition zone, in interaction with proteasome 26S subunit, non-ATPase 2 (Psmd2) regulates proteasome activity at the basal body of the cilium( Reference Gerhardt, Lier and Burmuhl 103 ). Although in humans no mutation in the RPGRIP1L gene has been reported in connection to obesity, the heterozygotes of the Rpgrip1l-knockout mouse are hyperphagic, have more fat mass and have a reduced suppression of food intake in response to leptin. In the hypothalamus of the heterozygous mice, the number of adenylate cyclase 3 (ADCY3)-positive cilia is decreased, with impaired localisation of the leptin receptor near the cilia, and reduced leptin signalling. A similar phenotype was observed in human fibroblasts with hypomorphic mutations in RPGRIP1L( Reference Stratigopoulos, Martin Carli and O’Day 104 ).
The BBSome may transiently interact with PCM-1, a major protein of the pericentriolar satellite. Through BBS1, the BBSome interacts with RAB interacting protein 8 (RABIN8), the RAS oncogene family member 8 (RAB8) nucleotide exchange factor. It stimulates GTP binding to RAB8, which in turn directs vesicles to the cilium for ciliary membrane elongation( Reference Nachury, Loktev and Zhang 82 , Reference Heymsfield, Avena and Baier 105 ).
Inositol transduction and cyclic AMP signalling
The obesity-linked enzyme inositol polyphosphate-5-phosphatase E (INPP5E) has a function in inositol metabolism, hydrolysing phosphatidylinositol (4,5)-diphosphate (PI(4,5)P2) and phosphatidylinositol (3,4,5)-triphosphate (PI(3,4,5)P3), and as such is an important mediator of inositol signal transduction( Reference Bielas, Silhavy and Brancati 106 ). In addition, INPP5E activity regulates the phosphoinositide composition of the cilium membrane, which may have an influence on ciliary protein trafficking( Reference Chavez, Ena and Van Sande 107 – Reference Park, Lee and Kavoussi 109 ). In the mouse, Inpp5e has been spotted in the axoneme( Reference Jacoby, Cox and Gayral 18 ). Targeting of INPP5E into the cilium requires farnesylation and interaction with phosphodiesterase 6D( Reference Thomas, Wright and Le Corre 66 , Reference Humbert, Weihbrecht and Searby 110 ). Inactivation of INPP5E was shown to lead to cilium instability, which can be restored by blocking phosphoinositide 3-kinase (PI3K)( Reference Jacoby, Cox and Gayral 18 ). Inactivation results in the accumulation of PI(4,5)P2 at the tip of the cilium and in depletion of phosphatidylinsositol 4-phosphate (PI4P). It attracts PI(4,5)P2-associated proteins such as Tubby-related protein (Tulp) and G protein-coupled receptor 161 (Gpr161), which induce changes in cyclic AMP (cAMP) production and Shh signalling( Reference Chavez, Ena and Van Sande 107 , Reference Garcia-Gonzalo, Phua and Roberson 108 ).
Interaction of INPP5E with phosphodiesterase 6D indicates a link between inositol transduction and turnover of the second messenger cAMP. In fact, cilia contain various adenylate cyclases, of which type III (ADCY3) is relatively abundant. Links between the ADCY3 gene and obesity have been reported both in humans and in mice. In a Swedish cohort, genetic association was found between variants of the ADCY3 gene and men with obesity and type 2 diabetes( Reference Nordman, Abulaiti and Hilding 111 ). In the Han population, genetic association with obesity was also observed( Reference Wang, Wu and Zhu 112 ). Adcy3-knockout mice present with obesity, hyperphagia, low locomotor activity and leptin insensitivity( Reference Wang, Li and Chan 113 ). Such mice also demonstrate anosmia towards IP3-and cAMP-generating odourants( Reference Wong, Trinh and Hacker 114 ), which is not surprising, as olfaction depends on sensory cilia with a functional ADCY3 gene.
Cilium stability
It has been observed that BBS18 is involved in microtubule stabilisation, for which its interaction with histone deacetylase 6 (HDAC6) is of importance( Reference Loktev, Zhang and Beck 88 , Reference Diaz-Font and Beales 115 ). HDAC6 deacetylates α-tubulin, and thereby de-stabilises the cilium( Reference Loktev, Zhang and Beck 88 , Reference Diaz-Font and Beales 115 ). It is activated by aurora kinase A (AURKA), which co-localises at the base of the cilium with neural precursor cell expressed, developmentally down-regulated 9 (NEDD9)/human enhancer of filamentation 1 (HEF1), a factor involved in the cilia-related cancer disorder von Hippel–Lindau syndrome( Reference Xu, Li and Wang 116 ). In fact, the phosphorylation of HDAC6 depends on the interaction of AURKA with NEDD9( Reference Pugacheva, Jablonski and Hartman 117 ). Interestingly, INPP5E can also be activated by AURKA, which in turn down-regulates transcription of the AURKA gene( Reference Plotnikova, Seo and Cottle 118 ). Although activation of HDAC6 destabilises the cilium( Reference Loktev, Zhang and Beck 88 , Reference Diaz-Font and Beales 115 ), activation of INPP5E has a stabilising effect( Reference Jacoby, Cox and Gayral 18 ). In this regard, AURKA may play a key role in cilium turnover, and as such in the risk for obesity( Reference Loktev, Zhang and Beck 88 , Reference Chavez, Ena and Van Sande 107 , Reference Diaz-Font and Beales 115 , Reference Plotnikova, Seo and Cottle 118 ).
Pinpointing obesity to the basal area of the cilium
Mutations disrupting the function of the structural BBS proteins are supposed to hamper IFT, giving rise to obesity. As the IFT particles are associated with the BBSome, one would expect that mutations in the IFT proteins would also often lead to obesity. However, as can be seen in Table 1, this is only the case for BBS19/IFT27 and IFT172. The lack of obesity might be explained by a more severe phenotype of mutations, leading to prenatal or early postnatal death. Skeletal and renal abnormalities are common, in line with the function of primary cilia in those tissues( Reference Yuan, Serra and Yang 119 ). It is also possible that with modern efficient sequencing methods, mutations in milder phenotypes of obesity are expected to be detected more often in the coming time. Alternatively, obesity is not so much a consequence of ciliary transport itself. As Table 2 shows, many of the obesity-linked ciliary proteins are located at the basal area of the cilium. In this region, the assembly of the BBSome takes place, its attachment to the membrane and its loading with the proper cargo protein. It is therefore tempting to speculate that obesity results from an altered initiation of the ciliary transport at the basal area of the cilium.
Table 2 Location and function of ciliary proteins
SHH, sonic hedgehog; IFT, intraflagellar transport; AURKA, aurora kinase A; cAMP, cyclic AMP. For expansions of gene names, see the text or the abbreviations list.
* Mutation in the gene associated with obesity in human and/or mouse.
Genetic studies regarding obesity-linked ciliary genes
BBS is associated with truncal obesity. Compared with BMI-matched controls, BBS patients have a similar energy metabolism with higher visceral fat mass and higher leptin levels( Reference Feuillan, Ng and Han 125 ). Apparently, these higher values are related to the underlying genetic defect( Reference Grace, Beales and Summerbell 126 ). BBS1 is for 80 % caused by the M390R missense mutation and patients show diet-responsive obesity( Reference Cox, Kerr and Kedrov 127 ). For this mutation, a knock-in mouse has been generated( Reference Davis, Swiderski and Rahmouni 128 ). The homozygous mice are obese, hyperphagic, have increased leptin levels and reduced locomotor activity. It proves that obesity is indeed a consequence of this gene mutation. Ultrastructural examination showed elongated cilia and swollen distal ends, but an intact axonemal structure( Reference Davis, Swiderski and Rahmouni 128 ). Notably, phenotypic differences occur between the BBS subtypes. BBS10 patients have a significantly higher visceral fat mass than BBS1 patients( Reference Feuillan, Ng and Han 125 ). Differences have also been observed with regard to the severity and age-of-onset of obesity. Comparing BBS2 patients and BBS4 patients, Carmi et al.( Reference Carmi, Elbedour and Stone 129 ) found that for BBS2 obesity was relatively mild, whereas BBS4 was associated with early-onset morbid obesity. Further, genetic variants of BBS genes were found to be associated with different types of common obesity. In a study among French-Caucasian individuals( Reference Benzinou, Walley and Lobbens 130 ), the association was found between BBS2 (rs4784675) and common adult obesity, between BBS4 (rs7178130) and BBS6 (rs6108572) and early-onset childhood obesity, and between BBS6 (rs221667) and adult severe obesity. Other studies did not show genetic associations. Variation in the BBS6 gene among Danish obese subjects did not show significant association with common types of obesity( Reference Andersen, Echwald and Larsen 131 ). Studying SNP for fourteen BBS genes in large cohorts of women did not show association with body weight or body fat, suggesting that common variation in BBS genes does not have a significant influence on body weight and fat( Reference Birk, Ermakov and Livshits 132 ). This is in line with the outcome of a segregation analysis from 1995 by Reed et al.( Reference Reed, Ding and Xu 133 ) in 207 sibling pairs, showing that BBS genes do not co-segregate with extreme obesity. Possibly, the penetrance of BBS gene mutations with regard to obesity is limited and/or the contribution to obesity risk depends more on rare alleles.
Previously, we sequenced thirty subjects with extreme obesity( Reference Mariman, Bouwman and Aller 134 , Reference Mariman, Szklarczyk and Bouwman 135 ). We checked for relatively rare variants (frequency<0·01) with a predicted damaging impact on protein function in sixty-six genes that code for proteins, which are needed for proper cilia function (online Supplementary Table S1), including the genes mentioned in Tables 1 and 2. This resulted in the identification of twenty-four variants in seventeen subjects (Table 3). The BBS5 N184S variant in subject 628 represents a serine substitution of an asparagine residue that is highly conserved within the two DM16 domains of the protein( Reference Hjortshoj, Gronskov and Philp 29 , Reference Li, Gerdes and Haycraft 136 ); this has been reported before in two families with BBS5( Reference Li, Gerdes and Haycraft 136 ). However, the two heterozygous patients of one of those families were also homozygous for the common M390R mutation in BBS1. Therefore, it was suggested that BBS5 interacts with BBS1 and that the N184S variant acts a modifier of the BBS phenotype. Besides these rare variants, we observed a non-synonymous SNP in BBS10 (rs35676114, P539L) with a minor allele frequency of 0·07. For this SNP, we found association with extreme obesity in the examined cohort (χ 2, P=0·0004).
Table 3 Exome sequencing results of rare variants with a predicted impact on protein function in extremely obese subjects
SIFT, scale-invariant feature transform prediction method; PolyPhen2: LRT, likelihood ratio test; B, benign; P, possibly damaging; D, probably damaging; mutation taster: N, non-disease causing; D, disease causing; LTR: U, unknown; N, neutral; D, deleterious; ND, not determined; no, not present in these databases. For expansions of gene names, see the text or the abbreviations list.
* Stopcodon.
† SIFT: value<0·05 is regarded as damaging.
‡ Mutation in the gene associated with obesity in human and/or mouse.
§ ExAc database non-Finish European frequencies.
║ Genome of the Netherlands.
All subjects were heterozygous for the rare altered allele in line with the absence of syndromic symptoms. On the other hand, heterozygous variants may contribute to the phenotype as reported by McEwen et al.( Reference McEwen, Koenekoop and Khanna 37 ), who found reduction of smell perception in homozygotes and heterozygotes of the Cys998X mutation in CEP290. In this respect, the observed heterozygous variants might exert a small phenotypic effect. An accumulation of small effects from various rare alleles would fit with a multifactorial genetic background. Notably, subject 869 carried marked variations in four genes: BBS1, IFT46, IFT88 and ANKRD26. Indeed, our observation shows the presence of a considerable number of rare alleles with a predicted damaging impact on the proteins related to cilia function in extremely obese subjects. However, the actual involvement of this genetic variation in the risk for (extreme) obesity remains to be shown.
Another case (yet unreported data) concerns a male individual with early-onset severe obesity and anosmia. Exome sequencing revealed compound heterozygosity with two mutations in the ADCY3 gene. One mutation is a frameshift (Gly423fs) retaining only one-third of the correct polypeptide sequence, whereas the other mutation is a deletion of a phenylalanine (Phe1118del). Using Provean prediction software (provean.jcvi.org), this mutation was classified as ‘deleterious’. This finding in a male patient confirms the link between obesity and cilia function via the ADCY3 gene as previously reported for Adcy3-knockout mice( Reference Wang, Li and Chan 113 , Reference Wong, Trinh and Hacker 114 ).
Although we did not find marked variation in the gene for RPGRIP1L in the thirty extremely obese subjects, from a genetic point of view this ciliary gene is particularly interesting, because its 5' end is only 100 bp from that of the gene for FTO, with overlapping promoters. FTO is one of the most studied genes in relation to the genetic risk for obesity. However, it has become clear that the FTO gene is part of a chromosomal segment, in which several genes are located that influence weight regulation( Reference Rask-Andersen, Almen and Schioth 137 , Reference Tung, Yeo and O’Rahilly 138 ).
Mechanistic role of cilia in obesity
Several mechanisms of how mutations in ciliary genes can contribute to increased body weight have been proposed( Reference Chennen, Scerbo and Dollfus 139 – Reference Sen Gupta, Prodromou and Chapple 143 ). In this study, we focus on three possible mechanisms: adipogenesis, central signalling of food intake and odour perception. Although these mechanisms are separately discussed, it should be kept in mind that mutations in cilia genes are pleiotropic and can increase the risk for obesity via more than one mechanism.
A primary cilium for adipogenesis
Adipogenesis occurs as a result of two opposing forces based on pro-adipogenic factors such as insulin-like growth factor 1 receptor (IGF1-R), CAATT/enhancer binding protein α-β (CEBP/A-B) and PPARγ and on anti-adipogenic signalling pathways such as SHH, wingless-type MMTV integration site regulatory gene/pathway (Wnt) and Notch. Ciliary proteins may influence either one or both of these forces, and have therefore been referred to as gatekeepers of adipocyte differentiation( Reference Marion, Mockel and De Melo 144 ). Marion et al.( Reference Marion, Mockel and De Melo 144 ) showed that the reduced expression of the BBS12 gene in mesenchymal stem cells down-regulated the anti-adipogenic pathways but promoted the pro-adipogenic factors. On the other hand, a decrease in Alms1, Ift88 or Kif3a expression inhibits cilium formation, as well as also adipocyte differentiation in mouse 3T3-L1 cells( Reference Huang-Doran and Semple 145 , Reference Zhu, Shi and Wang 146 ).
When human pre-adipocytes in vitro were induced to differentiate to mature white adipocytes, it was observed that a primary cilium appeared on the pre-adipocytes when cell cultures became confluent. Immunostaining showed in the cilium the presence of receptors involved in the SHH and Wnt signalling( Reference Marion, Stoetzel and Schlicht 147 ). Similar observations were made in cultures of mouse 3T3-L1 pre-adipocytes, where a primary cilium together with α-tubulin acetylation was induced in growth-arrested confluent cells( Reference Zhu, Shi and Wang 146 ). In those cells, a sensitised form of the IGF1-R, an important pro-adipogenic factor, was also detected in the cilium. Cilium formation could be inhibited by the suppression of Ift88 or the kinesin Kif3a( Reference Zhu, Shi and Wang 146 ).
Recently, the process of cilium formation was studied in more detail using in vitro differentiation of human mesenchymal stem cells into adipocytes. Within the first 2 d of differentiation, the primary cilium was observed to elongate together with increased trafficking of IGF1-Rβ into the cilium. This elongation process could be inhibited by insulin or by reduced IFT88 expression( Reference Dalbay, Thorpe and Connelly 148 ). Similar information was obtained with in vitro differentiation of human adipose stem cells( Reference Forcioli-Conti, Lacas-Gervais and Dani 149 ). During the first few days after confluence, the primary cilium appeared and elongated, but thereafter it decreased in size to the stage where cells began to accumulate lipids. At that stage, the cilium completely disappeared. On day 3, approximately at the maximal length of the cilium, SHH signalling was reduced by 50 % as compared with undifferentiated cells. However, there is no definite proof that cilium length and SHH signalling are linked. Final disassembly of the cilium may involve the deacetylation of α-tubulin with microtubule destabilisation, by enzymes such as sirtuin 2 (SIRT2) and HDAC6( Reference North, Marshall and Borra 150 ). Recently, the transient occurrence of the primary cilium during differentiation of human adipose stem cells was observed with disappearance of the cilium at the beginning of lipid accumulation( Reference Forcioli-Conti, Lacas-Gervais and Dani 149 ).
Altogether, a picture emerges in which the primary cilium behaves like a sensory system that initially elongates and extends through the extracellular matrix (ECM) to monitor signals from the cellular environment( Reference Singla and Reiter 151 ). This allows the cells to make a go/no go decision for differentiation. As the cells mature into lipid-loaded adipocytes, the ECM develops into a strong supportive layer and the cilium disappears.
Genetic variation or mutation in each of the ciliary genes may shift the balance between anti- and pro-adipogenesis differently, which may explain the variation in obesity phenotype between ciliopathy subtypes as mentioned before. In addition, the effect on the level of hyperplasia and hypertrophy may differ per gene. By studying Bbs12-knockout mice( Reference Marion, Mockel and De Melo 144 ), it was observed that those mice had a higher number of small-sized to normal-sized adipocytes between hypertrophic cells than the wild-type mice. As small adipocytes are supposed to have a more healthy metabolic activity( Reference Skurk, Alberti-Huber and Herder 152 ), this was seen as a possible explanation for the low risk of BBS12 patients to develop type 2 diabetes, whereas in Alström syndrome early-onset type 2 diabetes is common( Reference Marshall, Maffei and Collin 20 , Reference Marshall, Muller and Collin 153 ). Knockdown of Alms1 in murine 3T3-L1 pre-adipocytes reduced pre-adipocyte differentiation by 2-fold( Reference Huang-Doran and Semple 145 ), suggesting that obesity in Alström patients is accompanied mainly by hypertrophy.
Assuming that cilia monitor signals from the environment, it would be interesting to know how the genes mentioned here respond to changes in energy availability, but not much data have been reported. For eleven patients with obesity and type 2 diabetes, who underwent bariatric surgery, microarray analysis of blood cell RNA was performed before and 6 months after the surgery( Reference Berisha, Serre and Schauer 154 ). IFT121/WDR35 was among the seven genes, of which the change in gene expression was strongly correlated with change in body weight, fasting plasma glucose and glycosylated Hb content.
To obtain further insight, we analysed gene expression data of fifty-three obese subjects who had lost approximately 10 % of their body weight either rapidly in 5 weeks or more gradually in 12 weeks on an energy-restricted diet, as reported elsewhere( Reference Vink, Roumans and Arkenbosch 155 ). Adipose tissue RNA levels were measured by Affimetrix microarrays at the start of energy restriction (t 0), at the end of energy restriction (t 1, 5 or 12 weeks) and after subsequently having been on a weight-stable diet for 4 weeks (t 2). Fig. 2 shows the relative expression levels over time of fifteen genes coding for components of the BTC. In all, fourteen of the fifteen BTC genes were up-regulated during energy restriction. For half of the genes, the measured up-regulation was significant (P≤0·01). The average up-regulation was 8 %, with the genes for BBS2, BBS7 and BBS9, from which the BBSome assembly starts( Reference Zhang, Yu and Seo 87 ), showing the highest up-regulation of 11 % (P<0·001), 17 % (P<0·001) and 11 % (P=0·002), respectively. This suggests that energy restriction stimulates IFT. Remarkably, the gene for BBS19/IFT27 was down-regulated by 7 %. This would leave BBS3/ARL6 more available for binding GTP, which would promote retrograde transport with possible (functional) decline of the cilium( Reference Liew, Ye and Nager 93 ). Such BTC profile might shift the balance between pro- and anti-adipogenic processes towards adipogenesis. An increased adipogenic capacity of the adipose tissue after energy restriction has been observed( Reference Rossmeislova, Malisova and Kracmerova 156 ).
Fig. 2 Changes in expressions of genes coding for proteins of the basal transport complex (BTC) during energy restriction (t
0–t
1) and balanced energy intake (t
1–t
2). Expression changes are indicated as average (n 53) fold changes at t
1 and t
2 compared with t
0. 
, BBSome components; 
, components of the intraflagellar transport (IFT) particles; 
, changes in gene expression of BBS19/IFT27.
The cilium in neuronal food intake regulation
The hypothalamic neurons of genetic mouse models for obesity (ob/ob, db/db) have shorter cilia compared with lean mice (C57BL/6)( Reference Han, Kang and Byun 157 ). The same observation was made for diet-induced obese mice with leptin resistance after a 14-week, high-energy diet. When 36-h fasted C57BL/6 mice were compared with mice that were re-fed afterwards for 6 h, an increased frequency of short cilia was seen after the fast and a higher frequency of longer cilia after the re-feeding( Reference Han, Kang and Byun 157 ). Cilia length in these genetic models could be corrected by leptin supplementation, suggesting that the cilium length is increased by leptin. Using N1 hypothalamic neuronal cells in vitro, it was confirmed that leptin does not promote the number of ciliated cells, but increases cilium length( Reference Kang, Han and Ko 158 ). This process seemed to depend on destabilisation of F-actin.
Rahmouni et al.( Reference Rahmouni, Fath and Seo 159 ) developed knockout mice for Bbs2, Bbs4 and Bbs6, which all showed increased food intake, body mass and fat mass. Administration of leptin could not bring down food intake and body weight, showing that these Bbs–/– mice are leptin resistant. Gene expression measurements pointed to a defect in the pro-opiomelanocortin (POMC) neurons of the hypothalamus. Indeed, disruption of the cilia on the POMC neurons of the hypothalamus led to the obesity phenotype in mice( Reference Davenport, Watts and Roper 160 ). In an additional study, it was shown that the leptin receptor interacts with BBS1, and that down-regulation of BBS1 or BBS2 in ARPE-12 cells induces mislocalisation of the leptin receptor( Reference Seo, Guo and Bugge 161 ). Altogether, this suggests that the leptin receptor is transported by the BBSome into the cilia of hypothalamic POMC neurons and that a defect in this system results in hyperphagia, obesity and leptin resistance. Studying a conditional Ift88 deletion mutant with absence of cilia in the hypothalamus, Berbari et al. showed that leptin resistance only occurred in adult mice after they had become obese( Reference Berbari, Pasek and Malarkey 162 ). This observation was confirmed in the Bbs4-knockout study. It was concluded that leptin resistance is a secondary effect of obesity and not the cause.
Hypothalamic neurons also function in signalling pathways other than leptin signalling related to energy intake and energy homoeostasis. Probably, the cilia are key players in those processes as well. Loktev and Jackson showed that the cilia of hypothalamic neurons contain receptors for the orexigenic neuropeptide Y( Reference Loktev and Jackson 163 ). Moreover, the melanin-concentrating hormone receptor 1 (Mchr1), which is involved in food intake regulation, is normally present in the primary cilium. According to Berbari et al.( Reference Berbari, Lewis and Bishop 164 ), Mchr1 is not properly taken up in the cilia of Bbs2- and Bbs4-knockout mice, causing over-activation of the orexigenic melanin-concentrating hormone (MCH)-signalling pathway. In humans, association studies between the MCHR1 gene and parameters of obesity are conflicting, which may be because of epigenetic effects( Reference Stepanow, Reichwald and Huse 165 ).
Studying the obese/hyperphagic Ankrd26-knockout mouse, Acs et al.( Reference Acs, Bauer and Mayer 78 ) recently demonstrated absence of Adcy3-containing primary cilia in the paraventricular nucleus, a part of the hypothalamus involved in food intake regulation. Similarly, disrupting the function of the Alms1 gene leads to a large decrease of ciliated neurons in the hypothalamus as determined by the absence of Adcy3, Mchr1 and somatostatin receptor 3 (Sstr3)( Reference Heydet, Chen and Larter 166 ).
In summary, the findings indicate that cilia on hypothalamic neurons, particularly on the POMC neurons, contain the receptors for hormones that regulate food intake such as leptin, neuropeptide Y and MCH. Therefore, those cilia and their proper functioning are important for weight regulation. Either a reduction in the number of cilia or a change in length, which may be influenced by leptin, may induce hyperphagia and obesity.
Sensory cilia for odour perception
A reduction or ablation of odour perception referred to as hyposmia and anosmia, respectively, is a symptom of various ciliopathies. Hyposmia and anosmia have been shown to be cardinal and constant features in BBS( Reference Braun, Noblet and Durand 167 ). A disturbance of the olfactory system has been shown in patients with BBS1, BBS3, BBS4, BBS5, BBS9, BBS10, BBS12 and BBS17, syndromes that are all associated with obesity (Table 1)( Reference Kulaga, Leitch and Eichers 24 , Reference Iannaccone, Mykytyn and Persico 28 , Reference McEwen, Koenekoop and Khanna 37 , Reference Braun, Noblet and Durand 167 ).
Additional information on the link between ciliary genes, disturbed olfaction and weight regulation comes from animal studies. For several genes including Bbs1, Bbs2, Bbs4, Bbs6 and Bbs8 ( Reference Kulaga, Leitch and Eichers 24 , Reference Rahmouni, Fath and Seo 159 , Reference Tadenev, Kulaga and May-Simera 168 ), genetic manipulation is accompanied by olfactory dysfunction, hyperphagia and/or increased weight. Knockout mice of Bbs1 and Bbs4 are runts of the litter at birth, but 10 % of them become obese at week 10( Reference Kulaga, Leitch and Eichers 24 ). Their odourant signalling is disturbed and the cilia of the olfactory epithelium show structural abnormalities with severe affection of the axoneme. In addition, the microtubular organisation of the dendrites is damaged( Reference Kulaga, Leitch and Eichers 24 ). Bbs2-knockout mice present with a deficit of olfaction and altered social behaviour( Reference Nishimura, Fath and Mullins 169 ). Olfactory dysfunction presenting as partial or complete anosmia is also observed in the knockout of Bbs6/Mkks (McKusick–Kaufman syndrome)( Reference Ross, May-Simera and Eichers 170 ). In the Bbs8 knockout, a loss of olfactory cilia from the olfactory sensory neurons is observed as well as an altered pattern of axon targets( Reference Tadenev, Kulaga and May-Simera 168 ). Weight gain is slow at young age but eventually the adult mice develop obesity, which is especially pronounced in the females.
The IFT proteins have been far less studied than the BBS proteins for their possible involvement in olfaction and obesity. Knock-down of Ift46 in the zebrafish led, among others, to disturbed ciliogenesis in the olfactory pits( Reference Lee, Hwang and Oh 171 ). Similar observations have been made by Halbritter et al.( Reference Halbritter, Bizet and Schmidts 124 ) concerning a knock-down of Ift172 with defective and shortened cilia in the zebrafish olfactory placode. A nonsense mutation in the gene for CLUAP1, supposed to be a part of the IFT-B particle, leads to absence of cilia from the olfactory pit of the zebrafish( Reference Lee, Hwang and Oh 171 ). However, in none of those studies, a link with weight regulation was made. Mice homozygous for a hypomorphic mutation in the Ift88 gene display polycystic kidney disease, underweight in litters and olfactory dysfunction, but do not develop obesity. However, this is suggested to be due to early death and health issues from multiple organ malfunction( Reference Lehman, Michaud and Schoeb 172 ). Ciliopathy in animals is not always accompanied by obesity, which is comparable with the situation in man. Moreover, direct evidence between cilia-related olfactory dysfunction and overweight/obesity is not yet available.
In more general sense, olfactory dysfunction may cause weight change, but does not always lead to overweight/obesity. Acquired hyposmia and anosmia may lead to all possible outcomes: either weight loss, weight gain or no change in weight( Reference Lee, Tucker and Tan 173 ). Aschenbrenner et al.( Reference Aschenbrenner, Hummel and Teszmer 174 ) found weight gain in 21 % of patients with acquired reduced sense of smell, whereas 11 % lost weight. Duration and severity of the affection and age may be of influence. Weight gain under olfactory deficiency is explained by a compensatory intake of nutrients, such as a higher amounts of sugar. Acquired reduction of smell in elderly women was observed to be associated with increased fat intake, suggesting risk for obesity( Reference Duffy, Backstrand and Ferris 175 ). In a small sample of persons with congenital anosmia (n 41), fifteen were overweight, which was significantly less than the expected twenty-six based on the frequency of overweight in the general population( Reference Aschenbrenner, Hummel and Teszmer 174 ).
Although the relationship between the cilia-mediated olfactory system and overweight/obesity can be complex, a link on the molecular basis can be demonstrated. The olfactory sensory neurons form a dendritic knob from which sensory cilia protrude through the mucus layer of the nasal cavity. Those sensory cilia carry the olfactory receptors (OR), which are transported into the cilia by the BTC( Reference Williams, McIntyre and Norris 84 , Reference Dwyer, Adler and Crump 176 ). As such, the cilium forms a link between the environment and the brain. Odourous compounds bind to the receptors, leading to depolarisation of the olfactory sensory cell( Reference Stephan, Tobochnik and Dibattista 177 ). This signal is transferred to the mitral cells, to which the sensory neuron is attached in the glomerulus of the olfactory bulb. From there, the signal is sent on to the hypothalamus and other regions in the brain. Experiments in the mouse have indicated that fluctuations in the level of secondary messengers, which in part are controlled by cyclases and phosphodiesterases, influence the way in which signals from odour-binding receptors are transferred( Reference Antunes, Sebastiao and Simoes de Souza 178 ). Cilia-based olfactory defects may also relate to changes in the grey matter of the brain as shown by Braun et al.( Reference Braun, Noblet and Durand 167 ).
As the cilium is the intermediate in the signal from the environment to the brain, variation in the functioning of OR can be regarded as a mimetic for variation in the functioning of sensory cilia. In this regard, it is interesting to notice that genetic studies have shown a link between OR and food intake/obesity. Genetic association was demonstrated between OR7 genes and eating behaviour and adiposity( Reference Choquette, Bouchard and Drapeau 179 ). Further, a copy number variant at chromosome 11q11 covering three OR genes, OR4P4, OR4S2 and OR4C6, was found to be associated with obesity( Reference Jarick, Vogel and Scherag 180 ). In addition, we have reported associations between genetic variation in the OR14C36 gene and extreme obesity( Reference Mariman, Szklarczyk and Bouwman 135 ).
Besides the OR, the neuronal connections between olfactory sensory neurons and the mitral cells in the glomerulus are important for food intake regulation. Removing the olfactory bulb from Kv1·3–/– mice abolishes their resistance to diet-induced obesity( Reference Tucker, Overton and Fadool 181 ). It has been proposed that the development of the neuronal connections in the glomerulus depends on two opposing processes, axon guidance and repulsion, and involves both OR and clustered protocadherin genes( Reference Hasegawa, Hirabayashi and Kondo 182 ). We recently reported a relationship between genetic variation in the clustered protocadherin genes on chromosome 5q and extreme obesity( Reference Mariman, Bouwman and Aller 134 ). Moreover, a genetic interaction between OR genes on chromosome 1q and the protocadherin-β genes has been observed in this extreme obesity cohort( Reference Mariman, Szklarczyk and Bouwman 135 ).
Notably, food odour can stimulate appetite, food-seeking behaviour and food ingestion, but strong or prolonged exposure to a food odour can have a satiating effect( Reference Lushchak, Carlsson and Nassel 183 ). In this regard, stimulation of sensory cilia may induce increased appetite and food intake. On the other hand, diminished ciliary signalling of the OR neurons in ciliopathy syndromes might reduce satiating cues by food odours and promote food intake and obesity as well. This duality in the response to food odours complicates studies on the relationship between olfaction and obesity risk. Moreover, food perception and response is subject to the metabolic status( Reference Palouzier-Paulignan, Lacroix and Aime 184 ). Stafford & Whittle( Reference Stafford and Whittle 185 ) showed that the preference and sensitivity to the odour of chocolate were different between obese and non-obese subjects. The property to vividly image flavours and aromas was found to be associated with BMI( Reference Patel, Aschenbrenner and Shamah 186 ). Despite this complex interaction between food odour and food intake, it is clear that cilia, particularly the sensory cilia, are essential for this process, and in this respect olfactory sensory cilia function may influence the risk for obesity.
Perspective on general obesity
As cilia are relevant for the proper development and performance of many cell types, organs and tissues, a mutation or, more in general, variation in a ciliary gene usually gives rise to pleiotropic effects. As a consequence, variation of a ciliary gene may contribute to the obesity risk from a broad spectrum of processes, including pre-adipocyte differentiation, hypothalamic regulation of food intake and olfactory perception and response. Despite the limited impact of variation in ciliary genes on obesity in general, the fact that those genes are players in various aetiological processes makes them an interesting target for intervention. The obesity phenotype in ciliopathies is a consequence of mutations that impair protein function. Therefore, overexpression of ciliary genes may provide a gene-therapeutic way to prevent obesity. Experiments with overexpression have been carried out to rescue the retinal phenotype of the Bbs1 knock-in mouse( Reference Davis, Swiderski and Rahmouni 128 ). Overexpression of Bbs1 by AAV-Bbs1 injection into the retina improved symptoms of retinal degeneration. However, injection into the retina of WT mice led to outer retinal degeneration, demonstrating the potential risk of overexpression toxicity( Reference Seo, Mullins and Dumitrescu 187 ). Perhaps a safer approach to explore the possibility of overexpression would be to aim for odour perception as part of the obesity risk via the olfactory sensory cells. The epithelium of the nose can be readily treated by gene therapy using non-invasive intranasal gene delivery( Reference Guemez-Gamboa, Coufal and Gleeson 140 ). As a demonstration of this application, adenovirus-mediated delivery of Ift88 to olfactory sensory nerves has led to restoration of the olfactory function in the Oak Ridge polycystic kidney disease (ORPK) mouse( Reference McIntyre, Davis and Joiner 51 ).
In a more optimal approach for obesity treatment, manipulation of ciliary gene expression should be performed by addressing various tissues to reduce simultaneously food craving, energy intake and fat storage in the adipose tissue. For this, drug therapy would be an attractive method. An example of potential drugs in this respect, although perhaps not very specific for mutations in ciliary genes, is the group of phosphodiesterase inhibitors, which can influence the level of second messengers. Not only can they have an effect on olfaction( Reference Antunes, Sebastiao and Simoes de Souza 178 ) but also on pre-adipocyte differentiation( Reference Alinejad, Shafiee-Nick and Sadeghian 188 ) and on leptin signalling in the hypothalamus( Reference Sahu, Anamthathmakula and Sahu 189 ). Moreover, phosphodiesterase inhibitors exist as natural food components( Reference Rahimi, Ghiasi and Azimi 190 ), although they have to be used with caution( Reference Pendleton, Brown and Thomas 191 ). More knowledge on cilia and ciliary genes in relation to the risk of obesity should provide more specific ways for prevention and treatment in the future.
Acknowledgements
The authors thank the Department of Human Genetics, Maastricht University Medical Centre (MUMC+), for providing the necessary infrastructure.
This study was funded by Netherlands Organisation for Scientific Research (NWO, TOP grant no. 200500001). NWO had no role in the design, analysis or writing of this article.
E. C. M. M. contributed to the design, contents and writing of the review, by analysing genetic data and by drawing the figures; R. G. V., N. J. T. R. and M. A. v. B. contributed by providing and analysing gene expression data; F. B., E. E. J. G. A., M. A. v. B. and C. T. R. M. S. contributed by providing genetic data; P. W. contributed to the design, to the scientific contents and writing. All the authors helped to improve the manuscript by critical evaluation.
There are no conflicts of interest.
Supplementary material
To view supplementary material for this article, please visit http://dx.doi.org/10.1017/S0007114516002282
