A Quantitative Trait Locus for Body Fat on Chromosome 1q43 in French Canadians: Linkage and Association Studies

Authors

  • Brahim Aissani,

    1. Human Genomics Laboratory, Pennington Biomedical Research Center, Louisiana State University System, Baton Rouge, Louisiana
    2. Present address: Department of Epidemiology, School of Public Health, University of Alabama, Birmingham, Alabama
    Search for more papers by this author
  • Louis Perusse,

    1. Division of Kinesiology, Department of Social and Preventive Medicine, Faculty of Medicine, Laval University, Sainte-Foy, Quebec, Canada
    Search for more papers by this author
  • Gilles Lapointe,

    1. Division of Kinesiology, Department of Social and Preventive Medicine, Faculty of Medicine, Laval University, Sainte-Foy, Quebec, Canada
    Search for more papers by this author
  • Yvon C. Chagnon,

    1. Molecular Psychiatric Genetic Unit, Robert-Giffard Laval University Research Center, Beauport, Quebec, Canada
    Search for more papers by this author
  • Luigi Bouchard,

    1. Division of Kinesiology, Department of Social and Preventive Medicine, Faculty of Medicine, Laval University, Sainte-Foy, Quebec, Canada
    Search for more papers by this author
  • Brandon Walts,

    1. Human Genomics Laboratory, Pennington Biomedical Research Center, Louisiana State University System, Baton Rouge, Louisiana
    Search for more papers by this author
  • Claude Bouchard

    Corresponding author
    1. Human Genomics Laboratory, Pennington Biomedical Research Center, Louisiana State University System, Baton Rouge, Louisiana
      Human Genomics Laboratory, Pennington Biomedical Research Center, Louisiana State University System, 6400 Perkins Road, Baton Rouge, LA 70808. E-mail: BouchaC@pbrc.edu
    Search for more papers by this author

Human Genomics Laboratory, Pennington Biomedical Research Center, Louisiana State University System, 6400 Perkins Road, Baton Rouge, LA 70808. E-mail: BouchaC@pbrc.edu

Abstract

Objective: To explore a quantitative trait locus (QTL) on human chromosome 1q affecting BMI, adiposity, and fat-free mass phenotypes in the Quebec Family Study cohort.

Research Methods and Procedures: Non-parametric sibpair and variance component linkage analyses and family-based association studies were performed with a dense set of chromosome 1q43 microsatellites and single-nucleotide polymorphism markers in 885 adult individuals.

Results: Linkage was observed between marker D1S184 and BMI (p = 0.0004) and with body fat mass or percentage body fat (p ≤ 0.0003), but no linkage was detected with fat-free mass. Furthermore, significant linkages (p < 0.0001) were achieved with subsamples of sibpairs at both ends of phenotype distributions. Association studies with quantitative transmission disequilibrium tests refined the linkage to a region overlapping the regulator of G-protein signaling 7 (RGS7) gene and extending to immediate upstream gene loci.

Discussion: The present study indicates that the QTL on chromosome 1q43 specifically affects total adiposity and provides a genetic mapping framework for the dissection of this adiposity locus.

Introduction

Obesity is a common multifactorial trait resulting from an imbalance between energy intake and energy expenditure. Its prevalence increased over the last decades, reaching epidemic proportions in the U.S. adult population (1). Importantly, obesity is a major risk factor for chronic diseases such as type 2 diabetes, hypertension, and coronary heart disease.

The heterogeneity of the obesity phenotype suggests a complex etiology involving interactions among behavioral, environmental, and biological factors (2). Although the existence of a genetic component has been suspected for decades, the genetic basis of obesity is still not well understood. Segregation analyses carried out in different populations provided evidence for major genes for BMI (3)(4)(5)(6) and fat mass (FM)1(7)(8)(9); however, other lines of evidence supported multiple genetic determinants (10)(11). Genetic studies over the last decade have been driven primarily by the belief that there were major genes, perhaps influenced by the single-gene deficiencies found to cause obesity in rodents and in some people.

Genome-wide approaches to human obesity have emerged in the last decade. A good number of genome-wide scans for obesity quantitative trait loci (QTLs) have been reported over the last 3 years, with 204 human QTLs reported as of October 2004 (12). However, follow-up fine mapping studies of these QTLs are scarce.

The genome scan approach can reveal genetic loci influencing a common trait independently of our understanding of its physiology and may generate useful leads. In the present study, we undertook a follow-up study of a QTL influencing total body adiposity that has been identified on chromosome 1q43 through a genome scan performed on the Quebec Family Study (QFS) cohort. The identified QTL has not been detected in previous genome scans for obesity-related phenotypes in humans. Genetic loci influencing quantitative variations in lipid (13) and sex hormone-binding globulin (SHBG) plasma levels (14) have also been mapped to chromosome 1q43. The clustering of multiple genetic loci affecting obesity and metabolism within this region of chromosome 1q is intriguing and deserves further analyses. The purpose of the present study was to narrow down the location of the gene associated with total body adiposity in the QFS.

Research Methods and Procedures

The QFS is based on French-Canadian families from the greater Quebec City area (15). The Laval University Medical Ethics Committee approved the study protocol, and written informed consents were obtained. The initial genome scan involved a sample of 780 adult individuals (age ≥ 17.5 years), which was increased in the present study to 885 individuals clustered in 292 families ranging in size from 2 to 8 members. The families were enrolled either through an obese proband (BMI > 30) or at random. BMI, FM, percentage body fat (PCTFAT) and fat-free mass (FFM) were available. Body composition phenotypes were assessed by underwater weighing and were available on 85% of the participants. A description of the unadjusted phenotypes for individuals stratified by generation and sex can be found in Table 1.

Table 1.  Descriptive statistics for the unadjusted phenotypes
 PCTFAT (%)FM (kg)FFM (kg)BMI (kg/m2)
  1. PCTFAT, percentage body fat; FM, fat mass; FFM, fat-free mass; SD, standard deviation.

Family groupMeanSDnMeanSDnMeanSDnMeanSDn
Grandfathers28.80.0129.90.0174.10.0127.44.715
Grandmothers42.64.3435.56.9447.33.1428.28.137
Fathers26.66.914623.010.714660.08.114628.55.9179
Mothers35.78.916927.014.216944.97.016928.78.0226
Sons20.19.717217.914.517262.58.717226.46.8185
Daughters29.710.622422.515.122447.06.722427.18.9243
Total28.210.871122.714.371152.910.871127.77.7885

Genotyping

Genomic DNA was prepared from lymphoblastoid cell lines derived from white blood cells. DNA preparation, polymerase chain reaction (PCR) conditions, and genotyping were as described elsewhere (16).

Briefly, microsatellite genotyping was carried out by means of automatic DNA sequencing and the computer software SAGA from LI-COR (Lincoln, NE). Single-nucleotide polymorphism (SNP) markers were genotyped using the Taqman fluorogenic 5′-nuclease assay, and the marker alleles were assigned using the FLUOstar galaxy software (BMG LABTECH GmbH, Offenburg, Germany). The fluorogenic probes were purchased from Biosearch Technologies (Novato, CA). Taqman primers and probes were designed on the basis of high annealing temperatures (>60 °C) using Primer 3 software (Whitehead Institute/Massachusetts Institute of Technology Center for Genome Research, Cambridge, MA; http:www-genome.wi.mit.edu). Taqman assays were run using the following PCR conditions: 1× QIAGEN PCR buffer (QIAGEN, Valencia, CA), 2 mM MgCl2, 25 μM deoxyribonucleotide triphosphate (dNTP), 0.01% Tween, 4% glycerol, 200 nM each of the forward and reverse primers, 100 μM each of the probes, and 0.3 units of Taq polymerase (QIAGEN). Optimizations of the Taqman assays were achieved by varying the annealing temperature and/or MgCl2 concentration and the probe concentration. The results of microsatellite and SNP genotyping were stored in a local dBase IV database, GENMARK, which checks for Mendelian incompatibilities. The search for erroneous genotypes was also carried out by a more sensitive test implemented in the Multiple Engine for Rapid Likelihood Inference (Merlin; http:www.sph.umichcsgAbecasis) (17), which is based on a multipoint likelihood statistic that detects inconsistencies between neighboring markers.

Marker Map

The initial genome-wide scan for obesity phenotypes was performed with 335 microsatellite markers (average heterozygosity > 0.7). The order and the intermarker distance in centimorgans (cM) were taken from version 9.0 of the Marshfield Institute map. The final marker map included 107 additional polymorphisms or restriction fragment-length polymorphisms from 63 gene loci found in the Genome Database and the Online Mendelian Inheritance in Man (OMIM) gene map. The sex-averaged genetic distance between markers for the whole set of 442 autosomal markers was 7.2 cM. Physical distances between markers were based on Builds 25–33 of the National Center for Biotechnology Information and versions hg8 to 14 of the University of California at Santa Cruz Golden Path.

Fine Mapping Genotyping

For the follow-up studies, we selected a 10-megabase (Mb) subregion on chromosome 1q42-43 that was identified in the initial genome scan (Y.C. Chagnon, L. Perusse, C. Leblanc, C. Bouchard, unpublished data). We extended the coverage by typing 23 additional microsatellite markers, resulting in an average marker spacing of ∼0.58 Mb in the critical region. We genotyped a total of 41 SNPs clustered into two peaks of linkage (Figure 1). All SNPs were from the dbSNP consortium except three polymorphisms: kynurenine 3-monooxygenase (KMO) 1 and 2, derived by multiple alignments of mRNA transcripts for the KMO gene (GenBank) using the Clustal program (European Bioinformatics Institute), and AGTM235T, a restriction fragment-length polymorphism within the angiotensinogen gene. Among the 41 SNPs, six departed from Hardy-Weinberg equilibrium, and one had a minor allele frequency of <0.1%; these polymorphisms were excluded from the analysis. The remaining 34 SNPs all had a minor allele frequency > 0.12 (Table 2).

Figure 1.

Genome-wide scan for anthropometric measures of obesity on chromosome 1q. Single-point (dotted curves) and multipoint (solid curves) linkage of BMI, PCTFAT, FM, and FFM on chromosome 1q are shown. These results were obtained in a genome scan with an average marker density of 8 cM in QFS using the traditional Haseman-Elston (H-E) approach [sibpair analysis (SIBPAL)]. A non-uniform scale of the distances was used for a better visualization of the marker distribution. The dotted horizontal lines indicate the significance level at the nominal p value of 0.05.

Table 2.  Description of the genotyped nucleotide polymorphisms on human chromosome 1q
Locus identificationLocus nameDistance (kb)HW (p value)*Allele frequency
  • HW, Hardy-Weinberg; nd, not determined; SNP, single nucleotide polymorphism.

  • *

    Departure from HW equilibrium. SNPs with genotype frequencies different from expected HW equilibrium proportions were excluded from all analyses and were not assigned a locus identification.

  • Frequency of the rare allele in Quebec Family Study vs. dbSNP (in parentheses).

SNP1rs9323352065440.96710.20 (0.26)
SNP2rs18921232214510.84550.38
SNP3rs15727502217870.69380.29
SNP4rs15552902267680.36890.15
SNP5rs9104972268240.49640.43 (0.44)
 rs14719152268750.00270.11
SNP6rs7292982269230.23730.28
SNP7rs8654702269800.70690.25
SNP8rs7011522271010.35590.21
SNP9rs5186862278500.35450.39 (0.39)
SNP10AGTM235T2273200.6090.39
SNP11rs20663802371980.1770.28
SNP12rs14184792372280.22310.24
SNP13rs14184772372770.15820.23
SNP14rs7859762373070.2460.21 (0.19)
 rs1877832373490.02080.48
SNP15rs13822442373680.55670.38 (0.28)
SNP16rs4222562373840.69240.41
SNP17rs4784522374880.50010.48
 rs18787292375390.01090.44
 rs19158732376590.00090.12
SNP18rs19158802376620.16320.12
SNP19rs21771162376720.40960.25
SNP20rs19968052377300.67860.15
SNP21rs23416922377860.91240.19
SNP22rs13414672378140.36340.25
SNP23rs13414462378380.83580.46 (0.49)
SNP24rs13414512378730.15720.35
SNP25rs19542022379080.12120.23
SNP26rs9473062379560.35880.43 (0.32)
SNP27rs19324402379990.41420.11
SNP28KMO22380080.25380.31
SNP29KMO12380510.88090.18
 rs6407182380520.0010.12 (0.10)
 rs20505102380640.79830.02
SNP30rs6463552380650.27650.32
 rs6767502380930.05170.33
SNP31rs6638142381060.49640.37
SNP32rs8517772383010.70930.16
SNP33rs695236nd0.25530.17
SNP34rs1915007nd0.29690.25

Statistical Analyses

The phenotypes were adjusted by multiple regressions for the effects of age and its higher order terms (age2 + age3) on the mean and variance within generation/sex groups (18); a minimal heteroscedasticity was observed. The residual phenotypes were further standardized to a mean of zero and 1-unit variance within the four groups. Extreme outliers, defined by phenotypic values of 4 standard deviations away from the population mean, were removed in the development of the regressions but were kept in the final study group.

Linkage Analysis

Model-free quantitative linkage between 49 microsatellite markers on chromosome 1q and obesity phenotypes was tested using the regression-based approach sibpair analysis (SIBPAL) [Statistical Analysis for Genetic Epidemiology (SAGE), releases 4.1 to 4.3; computer program package available from the Department of Epidemiology and Biostatistics, Rammelkamp Center for Education and Research, Metro Health Campus, Case Western Reserve University, Cleveland, OH], maximum likelihood variance component analysis (Merlin), and the LODPAL method (a program to perform linkage analysis) for selected samples, also implemented in SAGE. For follow-up linkage analyses, we have used a more powerful version of SIBPAL (19). Single and multipoint estimates of alleles shared identical-by-descent were generated with the GENIBD program (SAGE) using all possible sibpairs in sibships. Because the Student's t statistic of the Haseman-Elston (H-E) regression is valid only asymptotically, empirical p (pemp) values for critical results (asymptotic p < 0.05) were obtained by permutation tests. Sufficient permutations (up to 500,000 replications) were performed to reach 90% confidence that the estimated value is within 20% of the true value. Linkage was also tested in a variance component framework (20) as implemented in Merlin. Linkage was further carried out with the LODPAL method (21) using the discordant sibpair (DSP) and concordant sibpair (CSP) designs (22). For all three methods, linkage was deemed to be suggestive or significant if it reached the standard thresholds (23) except for the results obtained in the initial genome scan, which used less stringent thresholds.

Linkage Disequilibrium (LD) Mapping

Fine mapping was performed by testing allelic associations across moderately dense marker maps in the QTL interval. We employed a family-based test of association for quantitative traits [quantitative transmission disequilibrium test (QTDT; http:www.sph.umich.educsgabecasisQTDT)] (24). This approach allows for multiple siblings and for partitioning the phenotypic variance into environmental, polygenic, and additive variance components.

For more accuracy and confidence in our results, p values for association were confirmed by comparison with random samples of genotypes generated by conditioning on the observed phenotypes using the gene dropping simulation method in Merlin (17). p values for association were not adjusted for multiple testing because of the correlated phenotypes and the tight linkage between marker alleles (25) (see also “Discussion”).

Multiallelic tests (as opposed to single allele/haplotype) were performed, and alleles/haplotypes with frequencies ≤0.05 were pooled to reduce multiple testing. Haplotypes were generated from unrelated individuals (founders) using Merlin (17) and recoded before their use in QTDT.

Pair-wise disequilibria across the map were assessed using the expectation maximization algorithm (26) implemented in the Graphical Overview of LD (GOLD; http:www.sph.umichcsgAbecasis) software package (27). The level of LD was assessed by Lewontin's parameter D′ for marker loci with a minor allelic frequency >5%.

Results

Linkage Analyses

An initial genome scan for obesity QTLs in the QFS (Y.C. Chagnon, L. Perusse, C. Leblanc, C. Bouchard, unpublished data) identified a region on chromosome 1q43 with suggestive evidence for linkage with BMI, PCTFAT, and FM. Figure 1 shows these results on 1q. Of 49 microsatellite markers, 12 were localized in a linked region spanning ∼10 Mb and delimited by D1S251 and D1S547, the proximal and distal markers, respectively. Within this region, linkage to BMI fell short of suggestive significance at two marker positions centered on the angiotensinogen gene, near D1S251 (p = 0.020), and a more distal marker D1S180 (p = 0.012). Similar linkage patterns were obtained with the adiposity phenotypes, PCTFAT and FM, but with slightly increased linkage significance. The peaks of linkage centered on D1S3462 (p = 0.017 and 0.006, for PCTFAT and FM, respectively) and on D1S180 (p = 0.0039 and 0.005, respectively). No linkage to the FFM phenotype was found with any of the markers.

In the present study, the QTL on 1q was reevaluated on the basis of a denser map that included 23 additional markers, which brought the average marker density over chromosome 1q to 2 Mb and 0.5 Mb over the critical regions. Linkage was performed using larger sample sizes (Table 1) and the new H-E approach (28). Single-point and multipoint interval linkage analyses were performed with a maximum of 392 sibpairs for BMI and 336 sibpairs for PCTFAT, FM, and FFM phenotypes. Improved linkages with all phenotypes but FFM were obtained with the new H-E method (Table 3). Suggestive linkages to PCTFAT (p = 0.0002; pemp = 0.0013), FM (p = 0.0003; pemp = 0.0086), and BMI (p = 0.0004; pemp = 0.003) were identified at D1S184.

Table 3.  Fine mapping of the obesity QTL on chromosome 1q
  p
  1. QTL, quantitative trait loci; PCTFAT, percentage body fat; FM, fat mass; h2, heritability; FFM, fat-free mass; H-E, Haseman-Eltson; VC, variance-components.

  2. Multipoint linkage analyses were performed with 49 microsatellite markers (average density < 2 Mb) and a maximum of 392 sibpairs using the revisited H-E (SIBPAL) and VC (Merlin) methods. p values are shown for only those markers located across the peak of linkage. Permutation tests (maximum of 5 × 105 runs) were performed for the critical results (nominal, p < 0.01) to report evidence for linkage with SIBPAL. For VC-based linkage, simulation studies with 100 replicates assessed the significance of the observed linkage to p < 0.01. The h2 of each obesity trait in the population studied and the number of families analyzed (n) are shown below each trait name. ns, not significant (p > 0.05).

  BMI (h2 = 0.61; n = 273)PCTFAT (h2 = 0.49; n = 244)FM (h2 = 0.52; n = 244)FFM (h2 = 0.62; n = 244)
MarkerIntermarker distance (kb)H-EVCH-EVCH-EVCH-EVC
D1S1149 0.0180.020.010ns0.013nsnsns
D1S2785729ns0.0030.0060.0030.0160.004nsns
D1S3044120.00160.00050.00030.00070.00130.001nsns
D1S180770.00100.00030.00090.00080.00200.0011nsns
D1S1841990.00040.000090.00020.00020.00030.0004nsns
D1S204420.00160.000080.00110.00090.00100.0003nsns
D1S5471500.04390.00020.00190.00090.0100.0004nsns
D1S32165ns0.0003ns0.03ns0.0008nsns

Linkage analysis was also undertaken with the variance components method, assuming no dominance. As shown in Table 3, the results were consistent with those obtained with the new H-E approach, namely that the linkage peak centered at D1S184 for all phenotypes but FFM. The observed linkages were not confined to a single marker but were spread over four markers flanking D1S184. Moreover, a significant linkage was found between the proximal marker D1S3462 (see Figure 1 for the location of this marker) and BMI (p = 2 × 10−5) and with the same marker and FM (p = 0.0007) (data not shown).

The substantial proportion (36%) of sibships with at least three siblings in QFS prompted an analysis of linkage based on sibpairs selected from opposite ends of the phenotypic distributions or from either end. Multipoint linkage using all sibpair types or a combination of DSPs and CSPs for high trait values (CSPH) was performed with LODPAL assuming a recessive model. The results shown in Figure 2 were obtained with a pool of DSPs and CSPH; all of the DSPs are composed of one sibling from the top 25% and the other from the bottom 25% of the distributions, and both siblings in all CSPH are from the top 25% (moderate sampling, 60 ≥ n ≥ 48 sibpairs). Linkage to BMI was observed at D1S184 (p = 6 × 10−6) and 42 kb further at D1S204 (p = 4 × 10−6). Linkage to PCTFAT was detected at five contiguous marker positions (5 × 10−4p ≥ 2 × 10−5). The peak of linkage centered on D1S180, a marker located ∼200 kb upstream of D1S184. A linkage pattern very similar to that of PCTFAT was seen with FM. Consistent with the results of the H-E and variance components approaches, no linkage to FFM was observed. Extreme sampling from the top and bottom 10% of the distributions yielded very small sample sizes (12 ≥ n ≥ 8); nonetheless, strong linkage signals (1 × 10−4p ≥ 5 × 10−6) were again found with PCTFAT (data not shown).

Figure 2.

Fine mapping of the QTL on chromosome 1q in selected samples (LODPAL). Multipoint linkage with a pool of DSPs, one sibling from the upper 25% and the other from the lower 25% of the phenotypic distributions, and CSPs (CSPH)—both siblings in a pair selected from the upper 25%. Linkage was carried out with the LODPAL program assuming a recessive mode of inheritance. The dotted horizontal lines indicate the genome-wide standard threshold for declaring a significant linkage (p = 2.2 × 10−5). No linkage to FFM was obtained with any of the markers tested in a pool of 51 sibpairs (data not shown). As calculated with the EDAC algorithm (31) for the smallest sample (n = 48), assuming a gene frequency, a residual correlation, and a trait heritability of 0.20, 0.40, and 0.52, respectively, the power to detect linkage (α = 0.05) was 0.41 for CSPH alone (n = 30), 0.85 for DSP alone (n = 18), and 0.91 for the combination CSPH + DSP. n, number of sibpairs.

The addition of 31–49 CSPs for low trait values (CSPL)—both members in all of the CSPL are selected from the bottom 25% of the distributions—to the pool of DSP and CSPH did not affect the linkage results for PCTFAT under either sampling scheme but led to marginal linkages to FM (n = 80; p = 0.0012) and to BMI (n = 97; p = 0.0015) (Figure 3). Finally, tests of linkage were undertaken with LODPAL assuming dominance variance or intermediary models between recessive and dominant models, but no linkage evidence was found. Overall, the above studies support the presence of a QTL on chromosome 1q43 affecting specifically body adiposity (but not lean mass) and defined a target region for fine mapping by association studies.

Figure 3.

Fine mapping of the obesity QTL on chromosome 1q in selected samples (LODPAL). Multipoint linkage with a pool of DSPs, CSPs from the top 25% (CSPH), and CSPs from the bottom 25% of the distributions (CSPL), both siblings selected from the bottom 25%. Other conditions are as in Figure 2.

Association Studies

For LD mapping, of the 34 SNPs that have been validated (see “Research Methods and Procedures”), 24 were distributed over an ∼1-Mb distance around the distal peak of linkage. The remaining SNPs, which initially mapped to the proximal peak of linkage, were subsequently found outside this region after the reordering of the clones and marker loci in updated genome assemblies. For this reason, only the SNPs pertaining to the distal peak are reported here. LD was evaluated with single markers, SNPs, and microsatellites or two-marker haplotypes assuming no dominance variance. No evidence for genetic admixture or deviations from Hardy-Weinberg equilibrium as potential sources of disequilibrium was found. Table 4 summarizes the significance levels of the transmission disequilibrium tests for the marker alleles/haplotypes that were found in association (p < 0.05) with the obesity phenotypes. BMI was moderately associated with SNP22 (pemp = 0.03), whereas the associations of PCTFAT and FM with this marker were even more significant (pemp ≤ 0.008). We have generated haplotypes for SNP22 and D1S184 and conducted multiallelic tests of association. A lack of association of BMI and a moderate association of FFM with hapG23 (haplotype of allele G of SNP22 and allele 23 of D1S184) were observed. In contrast, stronger associations were seen with FM (pemp = 0.002) and with PCTFAT (pemp = 0.006). These results defined hapG23 as the haplotype carrying the putative Chr1q43 variants influencing adiposity in the QFS cohort.

Table 4.  Variance components QTDT
  PCTFATFMFFMBMI
  1. QTDT, quantitative transmission disequilibrium test; PCTFAT, percentage body fat; FM, fat mass; FFM, fat-free mass.

  2. QTDTs with chromosome 1q43 markers from the obesity-linked region. Only markers from the quantitative trait loci interval defined by D1S2785 and D1S321 (see Table 3) that remained significantly (p < 0.05) associated with the obesity phenotypes after the permutation tests are shown. n, number of informative probands; p, empirical p value obtained by gene dropping simulation (1000 replications); hapG23 refers to a two-marker haplotype (allele G of SNP22 and allele 23 of D1S184). ns, not significant (p > 0.05).

MarkerIntermarker distance (Kb)npnpnpnp
SNP22 2140.0062140.0082140.0132370.035
D1S184473810.0143810.030381ns419ns
hapG23 3190.0063190.0023190.023356ns

To substantiate the QTDT results and gain more insights into the overall picture of LD in the critical region, an LD map was constructed for this region. Figure 4, A and B, show the distribution of the pair-wise disequilibrium as measured by the disequilibrium statistic D′. As depicted by the square and rectangular boxes (Figure 4B), the extent of useful LD (D′ > 0.40) barely exceeded 100 kb. This suggests that the location of the causal variant(s) might locate between SNP22 and D1S184 or in the immediate flanking regions because markers located 29 kb apart were not in LD (D′ < 0.30) with any of these two markers.

Figure 4.

LD map of the candidate QTLs. (A) Physical map and pattern of LD across the linked interval on chromosome 1q43 (GOLD plot). Four islands of LD defined by the disequilibrium measure D′ are shown in the QTL interval delimited by the long brackets. Within each island, areas of strong LD are distinguished by black and white. The QTL interval encompasses a large gene, RGS7, and a smaller one, FH, located near a cluster of three known genes. The thick arrows indicate the extent of the genes and the direction of their transcription. The thick vertical bars show the positions of SNPs (22–24) from the region associated with the adiposity phenotypes. (B) Schematic representation of the pair-wise LD across the QTL map. The squared and rectangular boxes show the extent of the haploblocks. The thick vertical bars indicate microsatellite (top lane) and SNP markers (bottom lane). The exon-intron structure of the RGS7 gene is shown. (C) Fine-scale map of the 5′-end of the RGS7 gene. The location of the nucleotide site (SNP22) associated with the body fat phenotypes is shown. (Filled box) Translated exon. (White box) Untranslated exon. (Gray box) CpG-island.

Computer Software

Computer software used in this study included: SAGE, Merlin, QTDT (http:www.sph.umichcsgAbecasis), GOLD (http:www.sph.umichcsgAbecasis), Primer 3 software (Whitehead Institute/MIT Center for Genome Research; http:www-genome.wi.mit.educgi-binprimer), and the European Conference on Design Automation (EDAC) power calculator (http:www.biostat.wustl.edugccgi-bincalcpwr.cgiJform2).

Electronic Database Information

Electronic database information is as follows: Marshfield Clinic Research Foundation, Center for Medical Genetics (http:research.marshfieldclinic.orggenetics); The Genetic Epidemiology Research Group, Human Genetics Division, University of Southampton School of Medicine (Location Database map; http:cedar.genetics.soton.ac.ukpublic_html); The Genome Database: An International Collaboration in Support of the Human Genome Project (hosted by The Hospital for Sick Children, Toronto, Ontario, Canada; http:gdbwww.gdb.org); OMIM (On-line Mendelian Inheritance in Man; http:www.ncbi.nlm.nih.govOmimgetmap.cgi); The University of California at Santa Cruz Golden Path (UCSC versions hg8 to 14; http:genome.ucsc.edu); The National Center for Biotechnology Information (Builds 25 to 33; http:www.ncbi.nlm.nih.gov); the European Bioinformatics Institute (http:www.ebi.ac.uk); and the EDAC power calculator (http:www.biostat.wustl.edugccgi-bincalcpwr.cgi).

Discussion

We reported the results of linkage and association studies of a QTL for obesity-related traits on chromosome 1q43. We postulated that improved linkage signals could be achieved by partitioning the heterogeneous BMI phenotype into its adiposity and fat-free components. It turned out that higher linkage and association signals were observed for FM and PCTFAT phenotypes compared with BMI with H-E, QTDT, and LODPAL analytical strategies. Linkage to these adiposity phenotypes occurred at the same genetic positions as with BMI. However, no linkage signal was uncovered with FFM, suggesting that the underlying QTL affects total adiposity.

When variance components-based linkage analysis was performed, a significant linkage to BMI was found at D1S184 (p = 9 × 10−5), and comparable linkage significances were found for the body fat phenotypes. The convergence of H-E and variance components procedures argues against a method-dependent bias.

The DSP and CSP designs have been the focus of several studies since the seminal article by Risch and Zhang (22), who provided theoretical arguments for the high efficiency of such designs. Our combination of DSP and CSPH resulted in significant increases of the linkage signals. The results remarkably matched those obtained with the total sample except for the putative signal at D1S3462, which was not detected. The non-normality of the dichotomized distributions should not be a major concern because LODPAL is essentially a regression-based approach. Furthermore, the power of detecting linkage (at α = 0.05) was >91.4% for either sampling scheme or for any of the phenotypes.

The possibility that the set of microsatellites used in this study might produce inflated multipoint linkage due to LD can be ruled out because most markers are not in LD with each other (Figure 4B), and there is a good match between single-point and multipoint linkages.

Multiallelic tests of associations with single markers and two-marker haplotypes in 214 to 381 informative probands suggested that the candidate locus affecting body fat localizes near markers SNP22 and D1S184. With correlated phenotypes and linkage between markers, Bonferroni or false discovery rate correction for multiple testing would be overly conservative because both methods assume independent tests. For note, because only two of 24 null hypotheses were rejected at the 5% α-level in our association studies, these two methods would be nearly equivalent. Correcting for multiple testing with either method nonetheless translates to p values close to the α level (0.002 to 0.006 × 24 = 0.048 to 0.15).

The observation that the extent of LD was in the 50- to 100-kb range and the fact that SNP22 was not found in LD with flanking markers SNP21 and SNP23 (D′ < 0.3) suggests that the causal variant(s) might be located in the vicinity of SNP22 and D1S184 and in the range of the common haplotype G23. Extended genotyping and haplotyping across the critical region overlapping SNP22 and D1S184 in cases carrying hapG23 and in controls are needed for a more precise location of the causal variant(s). There is no simple explanation, however, for the apparent inconsistency between the positive association found between SNP22 and D1S184 but not with SNP23, which is located between these two markers. One could speculate that a putative hypermutability effect of the CpG dinucleotide (C>T polymorphism at SNP23) could be involved.

The adiposity QTL revealed in the present study has not been detected in previous genome scans. There could be several possible reasons for this. First, it could be due to the heterogeneity of obesity phenotypes (i.e., BMI vs. direct measures of adiposity) among the family collections and populations involved in these studies. Second, different genetic loci may operate in childhood, adolescence, and adult age obesity. Our study included only adult subjects (>17.5 years) and might not compare with other studies. Third, the QTL is located within a genomic region of high recombination rate (∼4 cM/2 Mb), implying that denser marker maps than those employed in past genome scans (∼10 cM) might be needed for its detection.

An important observation that deserves further discussion relates to the clustering of the adiposity QTL as defined here with those for low-density lipoprotein-cholesterol (logarithm of the odds ratio = 2.5) and SHBG (logarithm of the odds ratio > 3.6) plasma levels in Chr1q43. Sex steroid hormones and SHBG have both been associated with variation in adiposity (29), and candidate loci for lipid-related traits, sex steroid hormones, and SHBG are present on chromosome 1q43 and surrounding regions. Intriguingly, common regions of linkage to body composition and steroid hormones were reported earlier (14) and refs. cited therein). Fine mapping with denser sets of tag SNPs across the linkage peaks for body composition, steroid hormone, and SHBG variations is needed to gain more insights into the genomic organization of Chr1q43 QTLs.

Our linkage and association study converges on a region overlapping the 5′-end of the regulator of G-protein signaling 7 (RGS7) gene and extending to the immediate upstream gene sequences. The RGS7 gene covers 581 kb on 1q43 and contains 19 exons encoding a member of the family of regulators of G-protein signaling proteins that function as inhibitors of G-protein-coupled receptor signaling (for review, see ref. (30). To our knowledge, there has been no report suggesting a role for the RGS7 gene in metabolism and obesity. In addition to the RGS7 gene, the linked region harbors a distal cluster of four known genes: FH (fumarate hydratase), KMO (kynurenine 3-monooxygenase), OPN3 (opsin 3, encephalopsin), and CHML (choroideremia-like Rab escort protein 2) genes, flanked proximally by a hypothetical gene of unknown function, LOC388755, expressing a 1.6-kb mRNA. Because the causal variant is yet to be uncovered, these genes should be considered as candidates for the adiposity QTL.

In summary, we have presented the results of linkage and association studies of a new QTL on chromosome 1q43 affecting body fat in QFS. Our results reveal that a region overlapping the 5′-end of the RGS7 gene and immediate upstream gene loci harbor a gene or sequence influencing variation in human adiposity. Further replications of this linkage in other family studies and populations and generation of finer LD maps across the critical region defined here should be an important goal of future studies.

Acknowledgments

This work was supported by a contract from the Institutes for Pharmaceutical Discoveries (Brampton, CT; to C.B. and L.P.). We thank Tuomo Rankinen for help with the SAGE and QTDT software packages and Marc Boudreaux for assistance with the Taqman genotyping and troubleshooting. L.B. was supported by the Fonds de Recherche en Santé du Québec. We also thank the other colleagues who have been associated with the development and the management of the QFS, particularly A. Tremblay, J.-P. Déspres, G. Theriault, A. Nadeau, and J. Weisnagel, and also G. Fournier, L. Allard, M. Chagnon, and C. Leblanc, without whom the project could not have been kept alive for 25 years. Thanks are also expressed to the other collaborators of the project supported by Institute for Pharmaceutical Discovery, particularly Eric E. Snyder, Tuomo Rankinen, Marc Boudreaux, Anik Boudreau, Jessica Watkins, and Christina Riley. C.B. is partially supported by the George A. Bray Chair in Nutrition.

Footnotes

  • 1

    Nonstandard abbreviations: FM, fat mass; QTL, quantitative trait locus; QFS, Quebec Family Study; SHBG, sex hormone-binding globulin; PCTFAT, percentage body fat; FFM, fat-free mass; PCR, polymerase chain reaction; SNP, single-nucleotide polymorphism; Merlin, Multiple Engine for Rapid Likelihood Inference; cM, centimorgan; Mb, megabase(s); KMO, kynurenine 3-monooxygenase; SIBPAL, sibpair analysis; SAGE, Statistical Analysis for Genetic Epidemiology; H-E, Haseman-Elston; pemp, empirical p; DSP, discordant sibpair; CSP, concordant sibpair; LD, linkage disequilibrium; QTDT, quantitative transmission disequilibrium test; GOLD, Graphical Overview of LD; CSPH, CSPs for high trait values; CSPL, CSPs for low trait values; EDAC, European Conference on Design Automation; RGS7, regulator of G-protein signaling 7.

  • The costs of publication of this article were defrayed, in part, by the payment of page charges. This article must, therefore, be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Ancillary