Genetic Variation in BEACON Influences Quantitative Variation in Metabolic Syndrome–Related Phenotypes

The samples for the current study were obtained during the course of large-scale epidemiological investigations of obesity/metabolic syndrome in the countries of Mauritius and Nauru. These populations are geographically isolated, relatively homogeneous, and demonstrate a high prevalence of obesity and diabetes (20,21). In the Mauritian sample, we focused our study on 409 individuals of Asian-Indian descent who were comprehensively phenotyped for multiple traits related to the metabolic syndrome. For replication purposes, a sample of 400 Micronesian Nauruans was also studied. The subjects were randomly selected from each population but within a limited age range of 35–65 years. Subjects were excluded if they were current smokers or drinkers, pregnant, were undergoing any treatment for diabetes, or currently on antihypertensive agents. All research protocols were approved by the Inner Eastern Health Care Network institutional review board.

For identification of sequence variants, 40 individuals, 20 from each population, were sequenced. The 40 individuals consisted of 17 obese, diabetic, or obese/diabetic individuals from each population (representing extremes on quantitative measures of BMI, waist circumference, and fasting plasma glucose) and 3 “hypernormal” individuals from each population (age range 50–70 years), who had all measured physical and biochemical parameters within the normal physiological range.

Phenotypes related to the metabolic syndrome were obtained from each individual during the course of a medical exam. Four phenotypes directly related to obesity and body composition were assessed in the Mauritian sample. These included waist-to-hip ratio (WHR), BMI, fat mass (in kilograms), and relative fat mass. Height was measured to the nearest centimeter and weight to the nearest 0.1 kg in light clothes and without shoes. BMI was calculated as weight in kilograms divided by height in meters squared. The waist measures were taken at the midpoint between the iliac crest and the lower rib margin, whereas the hip measurement was taken around the maximum circumference of the buttocks posteriorly and the symphisis pubis anteriorly. Waist and hip circumferences were measured in duplicate to the nearest 0.5 cm with a measuring tape while the participant was standing relaxed and wearing a single layer of light clothing. If the first two measures differed by >2 cm, a third measure was taken. The mean of the closest two measures were used to calculate the waist circumference and WHRs. Body composition measures were taken using a single-frequency bioelectrical impedance meter (BIM4; UniQuest, St. Lucia, Australia), with the participants rested and lying horizontally on a bed. Total cholesterol and triglycerides were determined on fasting plasma by manual enzymatic methods. To measure aspects of glucose metabolism, we obtained fasting levels of insulin using standard enzyme-linked immunosorbent assay methods (Dako, Ely, U.K.) and measures of plasma glucose using a YSI Glucose Analyzer (Yellow Springs, OH). All plasma measures were done in duplicate, and quality control procedures were implemented to maintain accuracy across batches. All of these metabolic syndrome–related phenotypes were available in the Mauritian sample. However, due to limitations in the field, only BMI, WHR, total cholesterol, triglycerides, and fasting plasma glucose were available in the sample from Nauru.

A region spanning 2.8 kb was sequenced encompassing the promoter and all exons and introns of BEACON. Comparative genomics was used to identify conserved sequences. A 130-bp conserved element was identified ∼1.7 kb upstream of the start of transcription in the human gene and 235 bp upstream of the start of transcription in the mouse gene. This element was screened for single nucleotide polymorphisms (SNPs) in the human gene. All gene sequences were analyzed for repetitive DNA using RepeatMasker (A.F.A. Smit, P. Green, http://ftp.genome.washington.edu/RM/RepeatMasker.html) to facilitate primer design. Primers were designed using Primer3 (S. Rozen, H.J. Skaletsky, http://frodo.wi.mit.edu/primer3/primer3_code.html) to be between 20 and 30 bp in length, with annealing temperatures within 1°C of each other and within the range of 58–63°C.

PCR was performed with 5 ng genomic DNA in a 20-μl reaction using 0.5 units TaqDNA polymerase (Qiagen, Hilden, Germany) on a PCR Express thermal cycler (Thermo Hybaid, Ashford, Middlesex, U.K.). PCR products were purified using the spin procedure for the QIAquick 96 PCR Purification Kit (Qiagen). Cycle sequencing was performed on both sense and antisense DNA strands in 15-μl reactions with 0.75 μl ABI Prism Big Dye Terminators version 3.0 Ready Reaction Mix (PE Biosystems, Boston, MA) and 5.25 μl halfTERM Dye Terminator Sequencing Reagent (Genetix, Hampshire, U.K.), using the Applied Biosystems dye terminator cycle sequencing protocol, on a PCR Express thermal cycler. Sequencing products were purified using genCLEAN 96-well Dye Terminator Removal Plates (Genetix).

Purified sequencing products were dehydrated and resuspended in 10 μl Hi-Di Formamide (PE Biosystems) and run through ABI Prism 3100 POP-6 polymer (PE Biosystems) in a 50-cm ABI Prism 3100 capillary array on an ABI Prism 3100 Genetic Analyzer (PE Applied Biosystems). Sequence analysis was performed using ABI Prism SeqScape software version 1.1, which allows the analysis of raw data and provides quality values to indicate the confidence of the automated base calls (PE Applied Biosystems).

Genotyping was carried out using the MassARRAY system (Sequenom, San Diego, CA), as has been previously described (22). Briefly, PCR primers were designed using SpectroDESIGNER to amplify from ∼100 bp surrounding the variant sites, and reactions were performed using 2.5 ng genomic DNA, 2.5 mmol/l MgCl², standard concentrations of other PCR reagents, and 0.1 unit HotStar TaqDNA polymerase (Qiagen) in a total reaction volume of 5 μl. For multiplex reactions, primers were diluted and mixed to simplify PCR cocktail assembly.

For the two SNPs with sufficient variation to allow analysis, we obtained allele frequencies and the linkage disequilibrium coefficient (D′) using a maximum likelihood method that allows for missing genotypic data. Standard statistical genetic methods were used to verify that the assumption of Hardy-Weinberg equilibrium was appropriate.

To assess the influence of the polymorphism in the BEACON gene on phenotypes related to the metabolic syndrome, we utilized a measured genotype approach (23,24). Measured genotype analysis is a form of association analysis that uses the genotypic information in a fixed-effect model for the mean effects. Thus, it can be used to test the hypothesis that phenotypic means differ among genotypes. Because these traits were nonnormally distributed, we used a robust method for estimating parameters using a multivariate t distribution (25). For the Indo-Mauritian sample, 11 phenotypes were examined, including BMI, fat mass, relative fat mass, WHR, total cholesterol, LDL cholesterol, HDL cholesterol, triglycerides, fasting insulin, and fasting glucose. These phenotypes were simultaneously adjusted for a number of covariates (sex, age, age², sex*age, and sex*age²) before genetic analysis using standard regression methods.

All parameter estimation was done by maximum likelihood under the assumption of a multivariate t distribution using the computer package, SOLAR (26). A formal test of association was obtained by calculating a robust likelihood ratio test statistic comparing a model in which the genotypic means were held equal against a model in which the genotypic means could vary. Because of the known phenotypic correlations among the multiple metabolic syndrome–related phenotypes that we examined, we also performed a multivariate test of association by testing for differences among the multivariate distributions of phenotypes across BEACON genotypes.

The human homologue, BEACON, is located on human chromosome 19p13.3 and consists of five exons, the first of which is noncoding, and spans 2,194 nucleotides from the transcription initiation codon to the polyadenylation tail. It is 100% identical to the P. obesus and M. musculus orthologs at the amino acid level, suggestive of an important function common to these mammalian species. We resequenced ∼2.8 kb of DNA (Fig. 1) of the BEACON gene in samples from 40 individuals (20 Mauritians and 20 Nauruans). All exons and introns were sequenced together with a part of the putative promoter. As shown in Table 1, we identified a total of four SNPs. Three of these variants (T-447C, A183G, and A818G) were novel, as defined as being absent from publicly available databases at the time of publication. The C1801G SNP was present in the dbSNP database (rs2287839; www.ncbi.nlm.nih.gov/SNP/index.html).

Two (T-447C and A183G) of the four variants were rare in both populations and did not provide enough copies of the rarer allele to allow further statistical investigation. The remaining two SNPs (A818G and C1801G) were nearly in complete linkage disequilibrium (D′ = 0.973 ± 0.015) and exhibited virtually identical allele frequencies. In the Indo-Mauritian sample, the frequency of the rarer G allele was 0.180 ± 0.012 for the A818G polymorphism and 0.177 ± 0.015 for the C1801G SNP. Thus, the effective correlation between these two SNPs was extremely high and prohibits statistical separation of their effects. The frequencies of the rarer G allele were effectively identical in the Nauruan population (0.180 ± 0.013 for the A818G SNP and 0.180 ± 0.013 for the C1801G SNP). Similarly, these two SNPs were in complete linkage disequilibrium in the Nauruan sample.

Table 2 lists the descriptive statistics for the phenotypes examined in the Indo-Mauritian sample. A total of 409 individuals were examined. Of these, 293 (71.6%) were women and 116 (28.4%) were men. The average age was ∼47 years. As expected, women exhibited greater adiposity than that observed in men, as illustrated by BMI (3 units higher in women) and fat mass (7.5 kg higher in women). Similarly, female body composition tended to favor the prominence of the hips over the waist. Women also exhibited twice the prevalence of diabetes (10.9%) as that seen in men (5.2%). However, men exhibited a somewhat worse lipid profile than women.

Table 3 provides the results of the robust measured genotype analyses in the Indo-Mauritian sample. It shows the maximum likelihood estimates of the genotype-specific means and the P value obtained from a robust likelihood ratio test. Because of the relative paucity of the GG genotype, it was combined with the AG genotype for analysis. Seven of the 11 phenotypes considered showed significant evidence for the effect of the BEACON polymorphisms. Significantly associated obesity-related phenotypes included the amount of total body fat (P = 0.026), relative body fat (P = 0.001), and WHR (P = 0.034). In all obesity-related phenotypes, the trend was for the rarer G variant to be associated with a better risk profile (i.e., genotypes with at least one G allele had less relative body fat and less visceral fat as indicated by WHR). We observed that while the correlation between BMI and fat mass was substantial (r = 0.86), the association results were qualitatively different, with BMI failing to show a significant association. This can be explained by the fact that while the correlation between BMI and fat mass is significant, a large proportion of variance within either trait (1 − 0.86² = 0.26) is independent of the other trait. Thus, there must be aspects of the underlying biology (and genetics) of fat mass that are independent of crude BMI. Additionally, these two traits have substantially different distributions in the observed sample. BMI was distributed nonnormally, with significant positive skewness and leptokurtic tails, whereas fat mass was normally distributed. In summary, measures of body composition appear to be influenced by these polymorphisms within BEACON or by other nondetected variants that are in high linkage disequilibrium with the observed variants.

Measures of cholesterol metabolism were also significantly associated with BEACON polymorphisms, including total cholesterol (P = 0.029), LDL cholesterol (P = 0.022), and plasma triglycerides (P = 0.011). As with the obesity measures, genotypes having at least one of the rarer G variants exhibited a lipid profile that would be less associated with risk of cardiovascular disease. For example, triglycerides were markedly lower in the AG + GG genotypes versus those observed in the common AA (CC) genotype. The observed associations could potentially be due to a primary effect of increased fat mass; therefore, to test this we reanalyzed the data, controlling for fat mass. We found that significant associations were retained for all three of the previously associated traits: total cholesterol, LDL cholesterol, and triglycerides (data not shown). For the two phenotypes most directly related to glucose metabolism, fasting plasma insulin showed a significant association with BEACON polymorphisms (P = 0.018). Fasting glucose was not significantly associated, although the trend in the genotypic means was consistent with that seen for insulin.

After the univariate robust association analyses revealed strong evidence for the potential influence of genetic variation in the BEACON gene across several of the physiological domains related to the metabolic syndrome, we performed a more global multivariate analysis to test whether there were substantial differences in the multivariate phenotypic distributions among the BEACON genotypes. Using all of the 10 available quantitative phenotypic traits, we performed classical multivariate ANOVA. The resulting likelihood ratio test provided a P value of 0.0004, further strengthening our evidence that BEACON genetic variation influences metabolic syndrome–related phenotypes. More specifically, the rarer G variant was found associated with a metabolic profile that has the potential to be protective for diabetes and heart disease.

In order to replicate the association between metabolic syndrome–related phenotypes and genetic variation within BEACON, we genotyped the two polymorphisms in the Nauruan sample. Table 4 shows the descriptive clinical profile of the Nauruan sample. Unfortunately, a greatly reduced set of phenotypes was available in this sample, with data available on only two (total cholesterol and plasma triglycerides) of the seven phenotypes that showed significant associations in the Mauritian sample. Table 4 clearly shows that the Nauruan population exhibits substantially more obesity and diabetes than that seen in the Indo-Mauritian sample. The average BMI for both men and women is ∼37 kg/m², one of the highest in the world. Similarly, the prevalence of diabetes in this sample was >30%. Thus, the Nauruan sample represents a population with a much greater burden of diabetes and metabolic syndrome than the Indo-Mauritians, in whom we first observed our associations.

Table 5 shows the results of the robust measured genotype analysis in the Nauruan sample. As in the Indo-Mauritian sample, total cholesterol appeared to be significantly lower in those genotypes, with at least one copy of the rarer G variant (P = 0.013). This replicates the original association observed in the Mauritius cohort for total cholesterol. However, although the results for triglycerides exhibit the same trend as that seen in the Mauritian sample, this phenotype was not significantly associated (P = 0.221). Results for the other two phenotypes that were available in this cohort were negative and consistent with the results from the Mauritian sample.

Three novel SNPs were identified in the human BEACON gene and genotyped, together with a fourth SNP available from public databases, in two large unrelated cohorts. Two of the SNPs, A818G and C1801G, were significantly associated with obesity-related measures, total body fat, relative fat mass, and WHR, with a trend noted for the association of the rarer G allele with a better disease risk profile. Fasting plasma insulin was also associated with BEACON sequence variants, as were several measures relating to the lipid profile, including total cholesterol, LDL cholesterol, and plasma triglycerides. One of two significant associations observed in the Mauritian cohort (total cholesterol) was replicated in a second cohort (Nauru), although the set of available phenotypes differed between the two populations, somewhat weakening the strength of this replication study.

The association of human BEACON gene variants with several obesity-related traits is consistent with our previously published animal model data (12) showing that BEACON gene expression levels were altered in the hypothalamus of obese animals and intracerebroventricular administration of BEACON protein increased food intake and body weight gain compared with controls. This apparent conservation of function across these two mammalian species is also supported by complete conservation (100% identity) of the primary amino acid sequence of Beacon between humans and P. obesus.

The influence of BEACON variation on lipid metabolism is particularly noteworthy because of its chromosomal location at 19p13.3. Numerous linkage studies of lipid and lipid-related phenotypes have provided evidence for a quantitative trait locus (or loci) (QTL) in this genomic region (15–19). The gene for the LDL receptor (LDLR) resides ∼1 MB centromeric to BEACON and has been largely assumed to be the likely gene underlying the widely observed QTL. However, our association evidence suggests that BEACON may also play a role in determining a variety of metabolic syndrome–related phenotypes, including various lipid parameters. Because of the substantial physical (1 MB) and genetic distance (∼2–3 cM), it is extremely unlikely that our observed genetic association is due to linkage disequilibrium with variants in the LDLR gene. Thus, BEACON and LDLR together may jointly influence quantitative variation in lipids. The presence of multiple genes underlying a QTL is expected, particularly in genomic regions where QTL localization is replicated widely, because it is probable that multiple variants are contributing and thus provide a greater chance for any single study to obtain a sufficient genetic signal for detection. Our findings of a significant role for the BEACON gene in metabolic syndrome–related phenotypes suggest that it may be the source of at least a portion of the QTL variance widely attributed to this genomic region on chromosome 19p13.3.

The BEACON gene variants were located in noncoding regions in intron 4 and the 3′ flanking region. It is not clear how these genetic variations influence Beacon activity or if they are merely in linkage disequilibrium with more distant variants that influence Beacon activity. However, if these variants are indeed functional, how might they exert their effects on Beacon function within the cell? There would be potential to influence Beacon activity through modulation of gene expression or splicing. Intronic enhancer sequences have previously been reported, and variations within these transcriptional control units may influence their activity; however, these are more likely to occur within the first 1,000 nt of intron 1 for expression and within 20 bp of the intron/exon boundary for splicing (27). The SNP in intron 4 (A818G) does not lie in either of these two regions, thus there is little a priori evidence that this variant may be functional. The SNP present in the 3′ flanking region (C1801G) falls outside the putative transcribed region, and the current gene annotation suggests it would not be present in the mRNA sequence where it could potentially influence mRNA transport and stability, a well-documented phenomenon. However a recent study (28) revealed that up to 50% of currently annotated genes use more distant polyadenylation signals with varying efficiencies that were not necessarily dependent on distance from the open reading frame. It is therefore possible that the use of an alternate polyadenylation signal further downstream would include the 3′ flanking SNP in the mRNA strand, presenting a potential mechanism for the variant to influence mRNA transport and stability. These putative changes in the expression of BEACON may influence the cellular pools of Beacon protein and therefore have the potential to affect its total cellular activity.

The human BEACON gene is a member of the family of ubiquitin-like modifiers (UBLs), which are involved in reversible modulation of protein function. The UBLs modulate their target proteins through a number of mechanisms, including alterations to subcellular localization, conformation, susceptibility to degradation, and ability to interact with other molecules (29). BEACON is highly conserved across species, with 100% amino acid identity with M. musculus, 80% with C. elegans, and 65% identity with S. cerevisiae. However, among the UBLs it is the most divergent of the group, with only 22% amino acid identity with ubiquitin. Additionally, the Beacon protein lacks the typical di-glycine motif common to other UBL family members; although, it was still found to be able to conjugate to target proteins in budding yeast following the usual processing of removal of a single carboxy-terminal amino acid (30). These target proteins were Spa2 homologue (SPH1) and Hub1p substrate (HBT1) and are thought to be involved in cytoskeletal regulatory protein binding activity in budding yeast. It is not yet clear whether this interaction relates to Beacon function in mammalian cells.

Beacon, acting as a ubiquitin-like protein, could potentially be involved in modulating the activity of target proteins. An interaction between Beacon and CLK4 was recently demonstrated (31) and subsequently confirmed following the determination of Beacon’s three-dimensional structure (32). While the physiological significance of this interaction remains to be elucidated, two related members of the CDC-like kinase (CLK) family, CLK1 and CLK2, have previously been shown (33,34) to phosphorylate and activate protein tyrosine phosphatase 1B, a negative regulator of insulin and leptin signaling pathways. It is therefore plausible that the Beacon-CLK4 interaction may be involved in regulation of these or related pathways known to affect traits associated with the metabolic syndrome.

In summary, the evidence presented here suggests that BEACON may play a role in influencing a variety of metabolic syndrome–related phenotypes, in particular those governing the control of lipid levels and obesity development. These data are supported by our previously published studies showing that the BEACON gene is involved in energy homeostasis in the polygenic animal model of the metabolic syndrome P. obesus. Although the mechanism behind the influence of BEACON on traits of the metabolic syndrome has yet to be fully elucidated, these findings suggest that BEACON may be one of many genetic factors that combine to predispose an individual to the development of the polygenic complex disease, the metabolic syndrome.

Funding for this study was provided by Chemgenex Pharmaceuticals, Australia.

The authors would like to acknowledge the valuable contributions of our colleagues from the Mauritian Ministry of Health and Quality of Life, Dr. Pierrot Chitson, Dr. Hassam Gareeboo, and Sudhir Kowlessur, and from the Nauru General Hospital, Dr. Kiki Thoma.