To ascertain whether distinct chromosomal loci existed that were linked to severe obesity, as well as to utilize the increased heritability of this excessive phenotype, we performed a genome-wide scan in severely obese French Caucasians. The 109 selected pedigrees, totaling 447 individuals, required both the proband and a sibling to be severely obese (BMI ≥35 kg/m2), and 84.8% of the nuclear families possessed ≥1 morbidly obese sibling (BMI ≥40). Severe and morbid obesity are still relatively rare in France, with rates of 2.5 and 0.6%, respectively. The initial genome scan consisted of 395 evenly spaced microsatellite markers. Six regions were found to have suggestive linkage on 4q, 6cen-q, 17q, and 19q for a BMI ≥35 phenotypic subset, and 5q and 10q for an inclusive BMI ≥27 group. The highest peak on chromosome 19q (logarithm of odds [LOD] = 3.59) was significant by genome scan simulation testing (P = 0.042). These regions then underwent second-stage mapping with an additional set of 42 markers. BMI ≥35 analysis defined regions on 17q23.3–25.1 and 19q13.33–13.43 with an maximum likelihood score LOD of 3.16 and 3.21, respectively. Subsequent pooled data analysis with an additional previous population of 66 BMI ≥35 sib-pairs led to a significant LOD score of 3.8 at the 19q locus (empirical P = 0.023). For more moderate obesity and overweight susceptibility loci, BMI ≥27 analysis confirmed suggestive linkage to chromosome regions 5q14.3–q21.3 (LOD = 2.68) and 10q24.32–26.2 (LOD = 2.47). Plausible positional candidate genes include NR1H2 and TULP2.
Significant changes in western lifestyle and increased globalization of these trends over the past 50 years have led to what is now termed an “obesogenic” environment (1). The subsequent rapid escalation in the prevalence of obesity now sees this trait affecting 30.5% of the U.S. adult population (2), and in the last decade, obesity nearly doubled in England and Wales to reach a level of 21% (3). The study of children with early-onset severe obesity has led to the first insights into its genetic causes in humans. The monogenic forms that have been discovered to date implicate the leptin and melanocortin pathways (4), and in >1–4% of all cases, melanocortin-4 receptor mutations can be attributed (5). However, these Mendelian disorders can only be held accountable for a small proportion of the condition. Obesity is a heterogeneous group of disorders, being predominantly a polygenic multifactorial trait, with an interplay of genetic and environmental factors (6). The genetic determinants of common forms of obesity are largely unknown, and although 71 associations with minor gene variants have been observed in different populations, few have been convincingly replicated (7).
Previous genome-wide scans have implicated many regions on various chromosomes; however, linkages to obesity-related traits at 2p21, 7q31–q32, 10p11–p12, and 20q13 have been able to be replicated in different ethnic groups (7). The linkage results in the 10p12 region found in four independent populations have aided the successful localization of the positional candidate gene GAD2, in which functional polymorphisms have been found to be associated with obesity (8). This gene encodes for a GAD enzyme that catalyzes the reaction responsible for the production of the orexigenic molecule γ-aminobutyric acid. Positional work on the X chromosome has also recently led to the association of the SLC6A14 gene encoding an amino acid transporter affecting serotonin synthesis and possibly appetite control (9). Both recent results, together with others in different complex traits (10,11), further support the evidence that positional cloning approaches are efficient in identifying new etiological pathways in the pathogenesis of disease.
In an attempt to enrich our study of obesity for its genetic susceptibility components, the strategy chosen was to focus on the extreme phenotypes. Severe pediatric (12) and severe and morbid adult obesity have been investigated in order to favor genes of increased penetrance and to try to reduce the causal heterogeneity. The determination of familial risk ratios for obesity has shown that the risk in relatives increases proportionally with the degree of obesity in the initial subject (13). The more extreme this risk ratio, as it is with the morbid phenotype, the more statistically powerful the gene mapping study is for detecting linkage in affected pairs (14). Also, we can examine the possibility that distinct loci exist that are causative only in these extreme subsets examined. Using all of the sib-pairs in these pedigrees who possess BMI ≥27 kg/m2, an additional analysis may be used to reveal susceptibility loci to an overweight state or common obesity. The possibility of stronger penetrance of these genes in the excessive phenotype may aid in their detection.
Therefore, appropriate pedigrees were selected by the requirement that both the proband and one sibling were at least severely obese (BMI ≥35), with the vast majority (84.8%) of the nuclear families in fact possessing at least one sibling that was morbidly obese (BMI ≥40). Two prior studies have benefited from selection of extreme phenotypes and identified linkages at 20q13 in U.S. Caucasians and African Americans (15) and 4p15–p14 in U.S. female subjects (16). However, these studies were performed in a country that is under the most severe environmental pressure to gain weight, where morbid obesity is now at a rate of 4.7% of the population (2). Within European countries France is among those with the lowest prevalence of obesity (17), and its differing environmental influence can be demonstrated by the fact that portion sizes are found to be smaller in restaurants, supermarkets, and cookbooks in comparison to the U.S. (18). Therefore, although these external influences are changing, as they are worldwide, currently only 0.6% of the population is morbidly obese (19), and these are in general those with an early age of onset in childhood or adolescence. Here, we present results from a genome-wide scan for severe obesity in a French Caucasian affected sib-pair population. Our results show suggestive evidence for linkage with this phenotype to regions on chromosomes 17q and 19q as well as for susceptibility to moderate obesity or overweight status at 5q and 10q. Replication of prior study results and plausible candidate genes with regard to the pathophysiology of obesity are subsequently discussed.
RESEARCH DESIGN AND METHODS
A total of 447 subjects were gathered from 109 French Caucasian multiplex families by way of a multimedia campaign run by the Centre National de la Recherche Scientifique in France and the Department of Nutrition of the Paris Hôtel-Dieu Hospital. The approval of the ethics committee of the Paris Hôtel-Dieu Hospital was obtained. Pedigrees were eligible for the study if two siblings possessed BMI ≥35 kg/m2. In total there were 109 pedigrees comprising 116 nuclear families. Only two of the families overlapped with the previous common obesity genome scan performed in the French Caucasian population (20). After they had given informed written consent, family participants were submitted to detailed personal and medical questionnaires and anthropometric measurements. The major phenotypic attributes of the study participants are given in Table 1.
Phenotyping.
Weight was measured in a nonpostprandial state and with an empty bladder and was determined to the nearest 0.1 kg on a standard physician’s beam scale, with the subject dressed only in light underwear and without shoes. Height was measured to the nearest 0.5 cm on a standard height board, again without shoes. The BMI was calculated as weight (in kilograms) divided by height (in meters) squared. Genomic DNA was extracted from peripheral blood cells by Pure-Gene D50K DNA isolation kits (Gentra Systems).
Initial genome-wide scan.
The Linkage Marker Set MD 10 (Applied Biosystems, Foster City, CA) formed the core marker set for the initial microsatellite genome-wide screen. These 400 microsatellite markers, labeled with fluorescent dyes (FAM, HEX, and NED), are distributed at an average marker spacing of 10 cM and have an average heterozygosity of 75%. The internal size standard was fluorescently labeled with a fourth dye (ET-ROX 400; Amersham Biosciences). We used a robust protocol allowing the coamplification of up to six of these markers in a single reaction using dual 384-well GeneAmp PCR 9700 cyclers (Applied Biosystems) and an automated procedure for PCR and purification set-up. All PCR mixes were prepared with a 96-tip head Automation Partnership BasePlate liquid handling robot. PCR was carried out with 4 μl of DNA (diluted to 5 ng/μl) + 6 μl of PCR mix. The PCR fragments obtained were pooled and purified before separation on automatic sequencers. All of the steps for pooling and G50 purification were performed using a 96-tip head Automation Partnership BasePlate robot. Then, 2 μl of the purified product was transferred to a 96-well plate and mixed with 3 μl of MegaBace loading cocktail (for one reaction: 2.75 μl H2O + 0.25 μl ET-ROX 400). The purified dye-labeled fragments were separated according to size on Amersham Biosciences MegaBace 1000 96-capillary sequencers. After injection (45 s at 3 kV), the samples were run for 65 min at 10 kV, using data collection software (Instrument Control Manager, version 2.1). After collection, data from runs were transferred to an independent team for blind analysis. The raw MegaBace data were automatically genotyped by the Genetic Profiler software (version 1.1, 1999–2000; Molecular Dynamics), which involved trace processing, fragment sizing, allele calling, and assigning of genotype quality scores. DNA fragment sizes were estimated by identifying the peak intensities for each sample using an internal DNA standard of known fragment sizes (ET-ROX 400). Fragments of similar estimated size for the same markers in the 96 samples of a single run were clustered to provide an allele assignment. A quality score was assigned to each genotype based on a measure of the deviation of the observation from the median value for the allele, and it was weighted by the standard deviation of the distribution over all observations. Five markers consistently erroneous for Mendelian inheritance were removed from study.
Second-stage mapping.
The six regions that showed linkage MLS of >2.2 for one of the BMI ≥35 or ≥27 phenotypes were then investigated. Further microsatellite markers were genotyped within these regions, leading to a reduction in the mean intermarker distance from 9.6 to 4.3 cM in order to further support and refine these linkage results. Microsatellite markers were selected from Genome Data Base (www.gdb.org) and were required to have a minimum heterozygosity of 0.7. A total of 42 markers were chosen, with a breakdown of 7, 6, 11, 9, 4, and 5 for chromosomes 4, 5, 6, 10, 17, and 19, respectively. Primers were designed for these markers using Primer 3 (www-genome.wi.mit.edu/genome software/other/primer3.html). Fluorescent dyes (FAM, HEX, or TAMRA) were attached to the 5′ end of forward primers. Primers were optimized with control DNA reactions for the temperature and MgCl2 concentration. Next, we added 0.5 units AmpliTaq Gold, 1.5 μl 10 × PCR Buffer II, 0.3 μl deoxy-nucleoside triphosphate 10 mmol/l, and 0.9–2.4 μl of 25 mmol/l MgCl2 to 20 ng of DNA and then made up to a volume of 15 μl with H2O. Then, 30 ng of both forward and reverse primers were added to this reaction volume, and the DNA was subsequently amplified by PCR using MJ Research PTC-225 Peltier Thermal Cycler tetrad machines. Cycling conditions were 95° for 15 min followed by 35 cycles of 95° for 30 s, 50–65° for 30 s, 72° for 30 s, and then a final step of 72° for 10 min. The resultant product was pooled using a Beckman Biomek FX robot. The first set of 26 fine-mapping markers were run on an Amersham Biosciences MegaBace 1000 96-capillary sequencer and analyzed using Genetic Profiler (version 1.1, 1999–2000; Molecular Dynamics). A second subsequent set of 16 markers were genotyped using the ABI 3700 machine and analyzed using Genotyper 3.7 (1993–2000; Applera). Genotyping analysis calls were determined independently by two researchers blind to the other’s results, and then inconsistencies between the two were analyzed further.
Genotyping quality control.
Before statistical analysis, rigorous genotype quality assurance was performed to ensure accurate binning of alleles. Mendelian inheritance inconsistencies were checked with the program Pedcheck (21). Six individuals showing frequent inconsistencies were removed from further analysis. The program Merlin was used to track suspected double recombinations (22). If the genotype error could not be solved, it was removed from analysis. Markers were also checked for Hardy-Weinberg equilibrium and removed if they diverged significantly (P < 0.05) (23). Finally, the recombination rates were compared between the Marshfield maps and the map built from our data by using the program Vitesse (24). A marker showing more than a twofold discrepancy was removed.
Maps and allele frequencies.
Maps were built according to Marshfield data (http://research.marshfieldclinic.org/genetics) and were used in the multipoint linkage analysis. Population allele frequencies were estimated with an estimation-maximization algorithm implemented in the program FBAT (Family Based Association Test) (25).
Linkage analysis.
Affected sib-pairs (i.e., both possessing BMI greater than or equal to the considered threshold) were analyzed using the maximum likelihood score (MLS) method to assess for linkage with the qualitative trait of BMI ≥35 or ≥27. This is a function of the identity-by-descent (IBD) probabilities p(ibd = 0), p(ibd = 1), and p(ibd = 2), i.e., z0, z1, and z2 (26). The outcome is compared with the null hypothesis likelihood values of 0.25, 0.5, and 0.25, respectively, and significant deviation from this, which is assessed by χ2 with a mixture of 1 and 2 degrees of freedom, can be taken as evidence of linkage. The MLS program is included in the Genehunter 2.1 package (27). The X chromosome was analyzed with the program ASPEX (Affected Sib Pairs Exclusion map, version 1.88). Because of multiple testing, we estimated genome-wide empirical P values by using the program Simulate (28) to create genotypes under the hypothesis of no linkage for the individuals initially genotyped. We retained the same phenotype information, marker allele frequencies, map distances, and missing genotypic data. A total of 1,000 simulations were conducted for BMI ≥27 and ≥35 on each autosome. The P value computed (empirical P value) is for each studied phenotype. For each set of simulations (from chromosome 1 to 22), the highest MLS was saved for both phenotypes. Thus, for each chromosome, we have 1,000 MLS results for BMI ≥27 and 1,000 MLS results for BMI ≥35. From these scores, we counted the number of MLS results above the given threshold of our study result, and we divided that number by 1,000 to get the empirical P value.
RESULTS
Initial genome-wide scan.
The phenotypic breakdown of the subjects’ BMI is revealed in Table 1 with the number of sib-pairs in each category. The average age in our study was 49.6 years, with minimal differences between the sexes, with men on average very slightly younger (49.2 years) than women (49.8 years). Average BMI was 46.95 kg/m2 for the initial proband and 38.13 kg/m2 for the sibling in the BMI ≥35 category and thus very close to a true morbid obesity sib-pair analysis. The first stage of the study involved the genome scan of all autosomes plus the X chromosome to search for linkage with BMI delineated into two phenotypic groups: BMI ≥27 and ≥35. The first subset was an all-inclusive group of overweight and obese subjects from gathered pedigrees, and the second subset focused on our extreme sib-pairs. The graphical results of MLS logarithm of odds (LOD) values for all autosomes are shown in Figs. 1A and B. Six regions revealed suggestive linkage for sib-pair analysis, with LOD scores >2.2 as defined by Lander and Kruglyak (29). The locations of the regions of linkage are displayed on Table 2 with their cytogenic location, peak location (in cM), adjacent marker, 1 LOD unit, MLS LOD score, empirical P value, and the phenotypic category for which they were found. Four of the regions were found with our extreme subset, and these were located at 4q22.2-q25, 6Cen-q21, 17q23–q25, and 19q13.3–q13.4, with LOD scores of 2.55, 2.49, 3.00, and 3.59, respectively, with the latter giving the highest score just below that of a genome-wide significant linkage of 3.6. The intervals on 5q14.2–q22.1 and 10q25-qter were found for BMI ≥27, with LOD scores of 2.28 and 2.90. The highest peak from our study, that on chromosome 19, was also found to be significant when the empirical P value (0.042) for this result for the BMI ≥35 analysis was estimated by simulation. No evidence of linkage was found on chromosome X.
Second-stage mapping results.
For the six chromosomal regions that showed suggestive linkage results from the initial scan, additional second-stage mapping was then used to further support these findings and to more accurately define their peak regions. Results of linkage analysis after the addition of the second group of markers are shown in Table 3. This increased level of marker density led to a reduction in the peak linkage score in four of the regions on chromosomes 4, 6, 10, and 19, and in the initial two of these four regions, they subsequently dropped below the suggestive value for linkage of 2.2 to 2.03 on chromosome 4 and 1.72 on chromosome 6. For the severe phenotypic group of BMI ≥35, the regions at 17q and 19q gave linkage scores of 3.16 and 3.21, respectively. The 1- LOD confidence intervals were reduced by 4 cM on chromosome 17 and increased by 2 cM on chromosome 19. To also search for any unique loci or possible replication of existing susceptibility loci for propensity to overweight status and moderate obesity, the inclusive phenotypic group of BMI ≥27 was also used in the second stage linkage analysis. This revealed a locus on chromosome 5q with a LOD score of 2.68 and another on chromosome 10q with a LOD score of 2.9. There was a slight increase in the former and decrease in the latter; however, both showed reductions in their 1-LOD confidence intervals of 6 and 2 cM, respectively. The results of these suggestive linkages after second-stage mapping on chromosomes 5, 10, 17, and 19 are displayed in Figs. 2A–D. The empirical P value in chromosome 19 drops to 0.09, reflecting the small decrease in LOD score.
In the previous French Caucasian genome scan reported by Hager et al. (20), subjects were not recruited for excessive phenotypes, but nevertheless a small proportion of the sib-pairs of this population (n = 66) had BMI ≥35. Retrospective analysis of these data for a BMI ≥35 linkage, although underpowered, resulted in a LOD score of 1.15 at the same chromosome 19q locus. The data from both populations were then combined (with the two common families accounted for only once) and analyzed using the method from Cordell et al. (30). This allowed for estimations of the LOD scores based on pair-specific IBDs because the maps were slightly dissimilar as a result of the markers used (data not shown). This led to the discovery of a significant LOD score of 3.8 located near the common marker of D19S571. This marker is located at 85.94 cM, which is 8.63 cM from D19S418. An empirical P value for this combined result was calculated by simulation and revealed a significant value of 0.023.
DISCUSSION
We have previously reported in the French Caucasian population three regions of linkage with obesity on chromosomes 2, 5, and 10p, which have been confirmed in other ethnic groups (31,32). However, this first French sample set only included a small number (n = 66) of sibships with both individuals possessing the severe form of obesity (BMI ≥35 kg/m2), which in our view justified a specific and adequately powered genome-wide analysis focused at the more extreme end of the phenotype. The two regions we report here on 17q23.3–q25.1 and 19q13.33–q13.43, respectively, are unique loci with regard to the two previous morbid obesity genome scans performed in the U.S. population (15,16). The chromosome 19q result was further supported by being a significant linkage, as calculated by simulation, as well as by reanalysis and combination with the BMI ≥35 sib-pairs from the previous study. In France, <0.4% of the general adult population was morbidly obese in 1995–1998 (19), the time during which the pedigrees used for the present study were collected. Also, in the Hôtel-Dieu obesity clinic, it has been observed that the recurrence rate of morbid obesity within sibships was as high as 20% (K.C., unpublished data), suggesting a familial aggregation. Thus, working in pedigrees enriched for familial cases of extreme forms of obesity in a (still) “lean” population may help to identify new obesity genes or those specific to the extreme phenotype. The power of the study strongly depends on the underlying true genetic effect, i.e., the λs. By using a not-too-optimistic estimate 2.5 for λs, the power to detect linkages with a scores of 3.6 and 3 can be estimated as 60 and 79%, respectively.
Within the linkage interval on chromosome 19q13.33–q13.43, the muscle glycogen synthase gene (GYS1) resides, and a polymorphism of this gene has been shown to be associated with the metabolic syndrome and specifically type 2 diabetes in Finnish, French, Japanese, and Pima Indian populations (33). It had been considered a good postulation for inherited skeletal muscle insulin resistance because of the observation of reduced activity in normoglycemic first-degree relatives. However, the polymorphism appears not to be functional, so it has been hypothesized to be in linkage disequilibrium with a causative variant within the region. Situated at 19q13.33 is a plausible physiological candidate, the NR1H2 gene, also known as LXR-β (liver X receptor-β [LXR-β]). Activation of this receptor has been shown to be important in the regulation of lipid and lipoprotein metabolism, and it has also been found to be a modulator of glucose homeostasis by the downregulation of genes involved in liver gluconeogenesis (34). The expression of the insulin-sensitive glucose transporter GLUT4 (or SLC2A4), which when impaired has been associated with insulin resistance and diabetes (35), is found to be controlled through the action of LXRs on the conserved LXR response element within its promoter (36). In an in vitro model of diet-induced insulin resistance, activation of LXR by a synthetic agonist has been found to improve glucose tolerance (34). Also at this location is TULP2, a member of the tubby gene family. Mutation in the tub gene in mice causes progressive retinal and cochlear degeneration, maturity-onset obesity, and impaired glucose tolerance (37). This phenotypic picture is similar to that seen in human obesity syndromes such as Bardet-Biedl (BBS) (38) and Alstrom (39). Tubby proteins have also been found to be highly expressed in the appetite control centers of the hypothalamus, including the arcuate nucleus (40).
With regard to chromosome 17q23.3–q25.1, Norman et al. (41) reported a pointwise linkage with percentage fat in Pima Indians at marker D17S785 located in 17q25.1. Also within this region, near the marker D17S1301, a linkage peak with the LOD score of 1.29 has been found in type 2 diabetes in Finland and southern Sweden (42). In an ethnic group with similar ancestors to our own, Caucasians from Quebec, a linkage was found with a quantitative trait locus influencing abdominal subcutaneous fat that overlaps this region (43). The Gly40Ser mutation in the glucagon receptor gene, which is positioned at 17q25, has been associated with central adiposity in men (44), type 2 diabetes (45), and another member of the metabolic syndrome, essential hypertension (46). The growth hormone gene is also located at 17q23.3. The secretion of this hormone is found to be lowered in those with obesity. There is a reduction in its half-life, the number of secretory episodes, and the total daily production, all of which are progressively reversed with weight loss (47). The reasons for this hyposomatotrophism are yet to be fully understood, and although the therapeutic use of recombinant hormone (due to its lipolytic effect) has only led to equivocal results, it still remains a plausible metabolic candidate (48). Positive results include a trial in children with Prader-Willi syndrome, where growth hormone administration was found to be beneficial, leading to a reduction in body fat mass and percentage body fat (49). An association with a polymorphism in ACE with obesity in Italian men has also been reported, and this gene is found within the region at 17q23.3 (50).
Chromosome 10q24.32–10q26.2 shows suggestive linkage for the overweight and broad-ranging obesity phenotype BMI ≥27. This region has been linked with waist-to-hip ratio in the U.S. population (51), and in the previous genome scan for generalized obesity in the French population, a linkage to serum leptin levels with a LOD score of 2.5 was found here (20). Percentage fat was found to be linked with the microsatellite marker D10S587 and BMI with D10S670 in a U.S. sample cohort, both of which are within this region (15). Linkage to the chromosome 5q14.3–q21.3 region was also found to be suggestive for the BMI ≥27 category. The above-mentioned previous French scan revealed suggestive linkage with serum leptin levels at 5cen-q, which overlaps with this region (20). The prohormone convertase 1 gene (PCSK1) is found within this segment of overlap between the two studies at 5q15. It is responsible for the cleavage of proinsulin to its active form as well as many other prohormones and neuropeptides, e.g., hypothalamic proopiomelanocortin. A monogenic form of obesity in humans has been described in an individual who was a compound heterozygote for a complete loss of function mutation of this gene and who presented with extreme childhood obesity and very low insulin levels (52). Also, linkage has been reported to type 2 diabetes in an extended French pedigree near marker D5S428, close to this gene (53).
In summary, we performed a genome-wide linkage analysis for affected sib-pairs with severe obesity (BMI ≥35) that was further enriched for obesity susceptibility genes by the large proportion of morbidly obese individuals (BMI ≥40), and we found one significant and one suggestive linkage at chromosomes 19q and 17q, respectively. This study further confirms the heterogeneous nature of the polygenic background of obesity. It suggests that specific genes that predispose to morbid obesity, or those involved with common susceptibility that have only been revealed by analyzing extreme subjects, may be identified through positional cloning.
APPENDIX
Electronic database information
Marshfield Medical Research Foundation (research.marshfieldclinic.org/genetics), Généthon (www.genethon.fr), The Genome Database (www.gdb.org), National Centre for Biotechnology Information (www.ncbi.nlm.nih.gov), Primer 3 (www-genome.wi.mit.edu/genome_software/other/primer3.html), Unified Database for Human Genome Mapping (genecards.weizmann.ac.il/udb), Ensembl (www.ensembl.org), Human Genome Browser Gateway (genome.ucsc.edu/cgi-bin/hgGateway), and Obesity Gene Map (obesitygene.pbrc.edu).
A: Autosomal results MLS LOD for initial scan for BMI ≥27 phenotype. B: Autosomal results MLS LOD for initial scan for BMI ≥35 phenotype.
A: Autosomal results MLS LOD for initial scan for BMI ≥27 phenotype. B: Autosomal results MLS LOD for initial scan for BMI ≥35 phenotype.
A: Chromosome 5 MLS LOD after second-stage mapping for BMI ≥35 (solid line) and BMI ≥27 (dashed line). B: Chromosome 10 MLS LOD after second-stage mapping for BMI ≥35 (solid line) and BMI ≥27 (dashed line). C: Chromosome 17 MLS LOD after second-stage mapping for BMI ≥35 (solid line) and BMI ≥27 (dashed line). D: Chromosome 19 MLS LOD after second-stage mapping for BMI ≥35 (solid line) and BMI ≥27 (dashed line).
A: Chromosome 5 MLS LOD after second-stage mapping for BMI ≥35 (solid line) and BMI ≥27 (dashed line). B: Chromosome 10 MLS LOD after second-stage mapping for BMI ≥35 (solid line) and BMI ≥27 (dashed line). C: Chromosome 17 MLS LOD after second-stage mapping for BMI ≥35 (solid line) and BMI ≥27 (dashed line). D: Chromosome 19 MLS LOD after second-stage mapping for BMI ≥35 (solid line) and BMI ≥27 (dashed line).
BMI classification of study population subjects
| BMI | Number of individuals | Male/female | Number of sib-pairs |
|---|---|---|---|
| ≥40 | 142 | 23/119 | 44 |
| ≥35 | 240 | 50/190 | 128 |
| ≥30 | 332 | 93/239 | 254 |
| ≥27 | 385 | 118/267 | 366 |
| <27 | 62 | 26/36 | 12 |
| Total* | 447 | 144/303 | 378 |
| BMI | Number of individuals | Male/female | Number of sib-pairs |
|---|---|---|---|
| ≥40 | 142 | 23/119 | 44 |
| ≥35 | 240 | 50/190 | 128 |
| ≥30 | 332 | 93/239 | 254 |
| ≥27 | 385 | 118/267 | 366 |
| <27 | 62 | 26/36 | 12 |
| Total* | 447 | 144/303 | 378 |
Phenotypic breakdown of French Caucasian subjects by BMI. Sex differential male/female and number of sib-pairs within each category.
Total number of individuals or sib-pairs.
Suggestive linkage for multipoint analysis of initial genome scan
| Cytogenic location | Peak (cM) | Adjacent marker | 1-LOD unit peak (cM) | LOD MLS | Empirical P value* | Phenotype (BMI) |
|---|---|---|---|---|---|---|
| 4q22.2–q25 | 126.87 | D4S406 | 114.87–134.87 | 2.55 | 0.266 | ≥35 |
| 5q14.2–q22.1 | 117.75 | D5S433 | 101.75–129.75 | 2.28 | 0.397 | ≥27 |
| 6Cen–q21 | 109.42 | D6S462 | 85.42–119.42 | 2.49 | 0.297 | ≥35 |
| 10q25–qter | 156.16 | D10S587 | 132.16–168.16 | 2.90 | 0.153 | ≥27 |
| 17q23–q25 | 96.63 | D17S949 | 86.63–104.63 | 3.00 | 0.119 | ≥35 |
| 19q13.3–q13.4 | 95.41 | D19S418 | 87.41–101.41 | 3.59 | 0.042 | ≥35 |
| Cytogenic location | Peak (cM) | Adjacent marker | 1-LOD unit peak (cM) | LOD MLS | Empirical P value* | Phenotype (BMI) |
|---|---|---|---|---|---|---|
| 4q22.2–q25 | 126.87 | D4S406 | 114.87–134.87 | 2.55 | 0.266 | ≥35 |
| 5q14.2–q22.1 | 117.75 | D5S433 | 101.75–129.75 | 2.28 | 0.397 | ≥27 |
| 6Cen–q21 | 109.42 | D6S462 | 85.42–119.42 | 2.49 | 0.297 | ≥35 |
| 10q25–qter | 156.16 | D10S587 | 132.16–168.16 | 2.90 | 0.153 | ≥27 |
| 17q23–q25 | 96.63 | D17S949 | 86.63–104.63 | 3.00 | 0.119 | ≥35 |
| 19q13.3–q13.4 | 95.41 | D19S418 | 87.41–101.41 | 3.59 | 0.042 | ≥35 |
Computed on initial genome-wide scan.
Suggestive linkage for multipoint analysis of fine-mapped genome scan
| Cytogenic location | Peak (cM) | Adjacent marker | 1-LOD unit | LOD MLS | Empirical P value* | Phenotype (BMI) |
|---|---|---|---|---|---|---|
| 5q14.3–1.3 | 109.75 | D5S1463 | 103.75–125.75 | 2.68 | 0.212 | ≥27 |
| 10q24.32–6.2 | 138.16 | D10S1267 | 130.16–164.16 | 2.47 | 0.291 | ≥27 |
| 17q23.3–q25.1 | 90.63 | D17S944 | 86.63–100.63 | 3.16 | 0.100 | ≥35 |
| 19q13.33–q13.43 | 93.41 | D19S418 | 83.41–99.41 | 3.21 | 0.091 | ≥35 |
| Cytogenic location | Peak (cM) | Adjacent marker | 1-LOD unit | LOD MLS | Empirical P value* | Phenotype (BMI) |
|---|---|---|---|---|---|---|
| 5q14.3–1.3 | 109.75 | D5S1463 | 103.75–125.75 | 2.68 | 0.212 | ≥27 |
| 10q24.32–6.2 | 138.16 | D10S1267 | 130.16–164.16 | 2.47 | 0.291 | ≥27 |
| 17q23.3–q25.1 | 90.63 | D17S944 | 86.63–100.63 | 3.16 | 0.100 | ≥35 |
| 19q13.33–q13.43 | 93.41 | D19S418 | 83.41–99.41 | 3.21 | 0.091 | ≥35 |
Computed on genome-wide scan with fine mapped regions.
C.G.B., M.B., A.S., and C.L. contributed equally to the study.
Article Information
This work was funded by the Medical Research Council (G0000477). Council DNA bank and clinical data collection was granted by the Direction de la Recherche Clinique/Assistance Publique-Hopitaux de Paris, the Contrat de recherche Clinique (1993), and the Programme Hospitalier de Recherche Clinique (AOM 96088).
We also thank Dr. Andrew Walley for helpful comments on the manuscript.
