Whole-Exome Sequencing Suggests LAMB3 as a Susceptibility Gene for Morbid Obesity Free

Obesity is a common chronic disease with a strong heredity component; it is a major contributor to the global burden of chronic disease such as type 2 diabetes (1). Genome-wide association studies (GWAS) have revolutionized the genetic analysis of multifactorial traits and led to the successful discovery of numerous genetic risk markers for obesity (2). However, large gaps in our understanding of the heredity impact on obesity remain. First, identified single nucleotide variants (SNVs) associated with obesity tend to have a minor impact on disease risk, thereby complicating the characterization of underlying functional genetic variants and pathophysiology. Second, together, identified genetic loci explain <3% of variation in BMI in the population (2).

In an attempt to identify additional susceptibility gene loci for complex disease, researchers have begun to apply the recently developed high-throughput sequencing technology, which has shown power to detect low-frequency and rare disease-causing variants by deep sequencing of all known exons (3,4). A few gene variants have been shown to associate with severe obesity (5,6). Whole-exome sequencing (WES) might be most efficient when applied to patients with more extreme forms of common diseases such as morbid obesity for a number of reasons. First, rare and low-frequency variants must have a high penetrance to allow statistical detection. Second, sequencing variants in the protein coding sequence might be more likely to have a stronger impact on phenotypes than variants in intergenic regions (e.g., known monogenic or major gene causes of obesity are associated with a severe phenotype and major impact on BMI) (7).

In this study, we used WES to detect low-frequency gene variants enriched in adult subjects with morbid obesity and validated them against never obese elderly adults. We believed that it would be possible to enrich for obesity genes among cases and to filter away such genes in the control subjects. In this study, we report a low-frequency obesity-associated variant in the coding region of the laminin subunit β3 (LAMB3) gene.

Obese and control subjects were recruited since 1993 through local advertisement for the purpose of studying genes regulating body weight or in association with planned visits to our medical or surgical units for morbid obesity or scoliosis, respectively. Subjects were examined in the morning after a night fast, and anthropometric measurements were performed. Serum leptin was measured by ELISA (Mercodia, Uppsala, Sweden).

At present, the total sample set consists of ∼2,400 nonobese and ∼4,300 obese subjects. From the obese subset, we selected, among the 10% with the highest BMI, 200 subjects with extreme obesity for WES. We have performed two rounds of the sequencing, each containing 100 obese samples. The 100 subjects in the first examination were reported previously (6). The results presented in this study included WES data from all 200 obese subjects. As control subjects, we used 100 subjects with idiopathic scoliosis who were sequenced using the same layout setting for WES from a genetic study (8). For the first genotyping validation, we selected the 484 morbidly obese subjects with the highest BMI in the large sample collection described above, including the 200 subjects who had been used in WES and 491 never-obese subjects with age >40 years and BMI always <30 kg/m². In the second validation, we genotyped 1,427 morbidly obese subjects and 783 never-obese control subjects with age >40 years and BMI always <30 kg/m².

All investigated obese subjects in this study had morbid obesity (i.e., BMI >40 kg/m²). The nonobese control subjects for WES had scoliosis. All other control subjects were healthy according to self-report. There was an overrepresentation of women in the studied cohorts because obese women are more likely to search for medical advice for their obesity, and scoliosis is more common among women. All subjects were of European ancestry and living in Sweden. The study was approved by the Regional Ethics Committees of Stockholm, and all subjects gave their informed consent to participation.

LAMB3 mRNA expression in abdominal subcutaneous white adipose tissue (WAT) was measured in 114 Swedish women without diabetes who were recruited from the general adult population in Stockholm, Sweden; this dataset is described by Dahlman et al. (9). The women displayed a large interindividual variation in BMI.

Genomic DNA was prepared from peripheral blood mononuclear cells using the QiAmp DNA Blood Maxi kit (catalog number 51194; Qiagen, Hilden, Germany). DNA purity and quality was confirmed by A260/280 ratio >1.8 on Nanodrop (Thermo Fisher Scientific, Waltham, MA) and agarose gel electrophoresis. DNA concentration was measured by Qubit (Life Technologies, Stockholm, Sweden). Subsequently, we took 0.8 μg of each DNA sample and randomly divided them into 10 pools, each containing 10 samples of either obese cases or control subjects. The concentrations of pooled DNA samples were measured with Qubit (Life Technologies) and the samples run on agarose gel.

WES was performed at the Science for Life Laboratory (Stockholm, Sweden) as previously descried (6). Briefly, each DNA library was prepared from 3 μg of the pooled genomic DNA. DNA was sheared to 300 base pairs (bp) using a Covaris S2 instrument (Covaris, Brighton, U.K.) and enriched by using the SureSelectXT Human All Exon 50 Mb kit (Agilent Technologies) and an Agilent NGS workstation according to the manufacturer’s instructions (SureSelectXT Automated Target Enrichment for Illumina Paired-End Multiplexed Sequencing, version A; Agilent Technologies). The clustering was performed on a cBot cluster generation system using a HiSeq paired-end read cluster generation kit according to the manufacturer’s instructions (Illumina). Samples were sequenced on HiSEq 2500 (Illumina) as paired-end reads to 100 bp/read. The sequencing runs were performed according to the manufacturer’s instructions. Demultiplexing and conversion were done by using CASAVA v1.8.2 (http://support.illumina.com/sequencing/sequencing_software/casava.html).

WES data processing was described previously (6). All sequencing reads were first aligned to the human reference genome assembly hg19/GRCH37 using Burrows-Wheeler Aligner (BWA version 0.6.1; http://bio-bwa.sourceforge.net/) with a read-trimming parameter quality score of 20. Sequence variants were called by using the mpileup function of samtools-0.1.18 (http://samtools.sourceforge.net/) with a minimum mapping quality of 20 and read depth between 8× and 1,000× for filtering. PCR duplicates were removed using samtools prior to variant calling. Annovar (http://www.openbioinformatics.org/annovar/ [10]) was used to integrate variant information from public databases, including gene reference (http://hgdownload.soe.ucsc.edu/goldenPath/hg19/database/refGene.txt, 2013), Single Nucleotide Polymorphism database (dbSNP; SNP138; http://hgdownload.soe.ucsc.edu/goldenPath/hg19/database/snp137.txt.gz), and the 1000 Genomes Project (http://www.1000genomes.org/).

Variant enrichment for obesity was achieved through several filtering and comparison steps (Fig. 1). The enrichment was performed based on the WES data from 20 obese pools against 10 control pools. First, we filtered out nonexonic and synonymous SNVs and kept missense and nonsense SNVs or SNVs causing potentially splicing changes. Secondly, we counted appearance frequencies of SNVs in obese and control pools separately and compared differences of the frequencies between the two types of pools. We retained SNVs appearing more often in obese pools using the following criteria: 1) present in at least two obese pools and 2) not present in any of the control pools. Our focus was low-frequency coding SNVs with potential functional impact. The last filtering step was based on minor allele frequency (MAF). With the assumption that a causative variant should be low frequency in the normal population, we filtered for SNVs that were either not found in public databases or known SNVs with an MAF in European origin populations (MAF_Eur) ≤5% (1000 Genomes Project, 2011 May release).

A two-stage validation strategy was used (Fig. 1). In the first stage, all obesity-enriched variants from the above filtering steps that could fit into a 384-well panel were genotyped in 484 obese subjects and 491 control subjects using the Illumina GoldenGate assay (Illumina) performed at the SNP&SEQ Technology Platform at Uppsala University (http://www.molmed.medsci.uu.se/SNP+SEQ+Technology+Platform/Genotyping/). The results were analyzed using the software GenomeStudio 2011.1 from Illumina. For the second step, the SNVs showing nominal significant association (P ≤ 0.05) with obesity in the first step were genotyped in additional 1,427 obese and 783 control subjects. Genotyping was performed using matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (SEQUENOM; Agena Bioscience, San Diego, CA) at the Mutation Analysis Core Facility at Karolinska Institutet (11). Multiplexed assays were designed using MassARRAY Assay Design v4.0 Software (Agena Bioscience). Protocol for allele-specific base extension was performed according to Agena Bioscience’s recommendation.

Additionally, rs2076349 was validated by using Sanger sequencing at Eurofins (http://www.eurofins.com/en.aspx) and TaqMan genotyping (C__22274317_10; Applied Biosystems, Foster City, CA) to confirm the genotypes of all homozygote for the rare allele T of the variant and a few heterozygote subjects.

All genetic association results in validation cohorts 1 and 2 have been submitted to GWAS Central.

Patients were investigated in the morning after an overnight fast when abdominal subcutaneous WAT was obtained by fine-needle aspiration. Adipocytes were isolated from abdominal subcutaneous WAT and adipocytes isolated as described (12). By comparing the size of the adipocytes with the total amount of body fat, the morphology of WAT could be quantitatively determined (delta values), as described in detail (13). Positive values are indicative of hypertrophy (few but large fat cells) and negative values by hyperplasia (many small fat cells).

WAT LAMB3 expression in the population of healthy subjects was assessed by Gene 1.0 and 1.1 ST Affymetrix arrays (Affymetrix) as described (9).

Isolation, growth, and differentiation of human mesenchymal stem cells (hMSCs) were previously described (14). hMSCs were reverse transfected 24 h before induction of adipogenesis using ON-TARGETplus SMARTpool small interfering RNAs (siRNAs) targeting LAMB3 or nontargeting siRNA pool (Dharmacon, Lafayette, CO). For the knockdown of gene expression, 6.25 pmol of siRNA (giving a final concentration of 50 nmol/L) was incubated together with 0.35 μL of Dharmafect-3 (Thermo Fisher Scientific) and medium (final volume of 25 μL) in a 96-well plate for 30 min at room temperature. Thereafter, 10,000 of cells in a volume of 100 μL were added. When transfection was performed in a 24-well plate, 3.2 μL of Dharmafect-3 was mixed with 50 nmol/L of siRNA in a final volume of 100 μL following addition of 100,000 cells in a volume of 400 µL. Cells were incubated for 24 h in a proliferation medium without fibroblast growth factor 2, and then adipogenesis was induced.

The hMSCs were collected at days 2, 7, 9, and 12 after the induction of differentiation for isolation of RNA. Total RNA was extracted using NucleoSpin RNA II kit (Macherey-Nagel, Düren, Germany) according to the manufacturer’s instructions. Concentration and purity of RNA were measured using a Nanodrop ND-1000 Spectrophotometer (Thermo Fisher Scientific). Reverse transcription was performed using the iScript cDNA synthesis kit (Qiagen) and random hexamer primers (Invitrogen, Carlsbad, CA). Quantitative RT-PCR of coding genes was performed using commercial TaqMan probes (Applied Biosystems). Gene expression was normalized to the internal reference gene 18s. Relative expression was calculated using the 2(−ΔΔ threshold cycle) method (15).

The hMSCs were transfected and differentiated in 96-well plates for 12 days as described above, washed with PBS, and fixed with 4% paraformaldehyde solution containing 0.123 mol/L NaH₂PO₄ × H₂O, 0.1 mol/L NaOH, and 0.03 mol/L glucose for 10 min at room temperature. Fixed cells were washed with PBS. Cell nuclei were stained with Hoechst 33342 (2 μg/mL; Molecular Probes, Thermo Fisher Scientific), and neutral lipids were stained with Bodipy 493/503 (0.2 μg/mL; Molecular Probes) diluted in PBS for 20 min at room temperature.

After washing, accumulation of intracellular lipid droplets and cell number (amount of stained nuclei) were quantified with the Acumen eX3 imager (TTP Labtech, Hertfordshire, U.K.). Total Bodipy (lipid droplet) fluorescence was normalized to the amount of nuclei in each well (Hoechst staining).

Genetic association analyses and odds ratio (OR) calculations were performed for A1 (minor alleles) based on genotypes using PLINK (http://pngu.mgh.harvard.edu/∼purcell/plink/). An χ² test was used in statistical analysis. Hardy-Weinberg equilibrium among case and control subjects was checked separately for each round of validation prior to the association analyses. A nominal P value of 0.01 in control subjects was used as a cutoff value for exclusion of SNVs from further association analyses. Obesity status was characterized by BMI and reported as mean ± SD for obese and control subjects separately (Table 1). Quantitative phenotypes were analyzed using standard linear regression implemented via PLINK. Sex and age were used as covariates to evaluate their influence on the association of a variant with BMI. Correction for multiple tests was performed using PLINK adjust function, with genomic control–corrected p value calculated based on genotypes of all variants in the final analysis. Standard least square algorithm was used in regression analysis of mRNA expression in the population of healthy subjects. Student t test was used in fat cell functional assays.

All obese subjects used in WES and validations had morbid obesity (BMI >40 kg/m²). Genotyped control subjects were chosen from never-obese subjects with age >40 years and BMI <30 kg/m². The characteristics of the studied subjects for WES and validations are shown in Table 1 and Jiao et al. (6).

WES generated ∼100 million nonduplicated sequencing reads for each of 20 obese pools and 10 control pools (Supplementary Table 1). The vast majority of paired reads were uniquely mapped onto the same chromosome with a pairing rate up to 90%. These reads were aligned to the human reference genome (hg19/GRCh 37) with mapping rates >95% (except one of the control pools) (Supplementary Table 1), and >90% of the target regions were completely covered with at least 8× depth (Supplementary Table 2).

Potentially interesting variants were identified through several filtering steps (Fig. 1). Obtained variants were summarized in Supplementary Table 3. The majority of the variants (between 111,589 and 201,653 per pool) were SNVs, of which >20,000 were exonic or at potential splicing sites. Our focus was SNVs with functional impact (i.e., missense, stop codon, or potential splicing site variants). We therefore filtered for exonic SNVs with the exclusion of synonymous variants. At this stage, we got >10,000 (10,511–13,297) such functional SNVs from each pool (Supplementary Table 3). Next, we filtered for SNVs shared by at least two obesity pools and not present in any control pools. Finally, after removing common SNVs with MAF_Eur >0.05 (1000 Genomes Project, http://www.1000genomes.org/), 439 SNVs were retained for the validation studies.

Two rounds of validations by genotyping were performed. In the first round, 382 SNVs, out of 439, that successfully passed quality criterion in primer design (final score >0.4) for the Illumina GoldenGate assay (Illumina) were genotyped in 484 morbidly obese subjects and 491 never-obese elderly control subjects. Two additional SNVs (with a score close to 0.4) were added to fill a 384-well genotyping panel. Genotyping of 19 SNVs failed; this was likely due to technical problems with the assay because each SNV had 0% in call rate. For remaining genotyped SNVs, the average sample call rate was 98.6%. Among successfully genotyped SNVs, eight were nonpolymorphic. Hardy-Weinberg equilibrium was checked prior to association analysis. After removal of eight more SNVs with skewed genotypes (P < 0.01), 349 SNVs were used in association analysis.

Twenty-five SNVs showed nominal association (P ≤ 0.05) with obesity in the first validation (Table 2). They were subsequently subjected to a second round of genotyping in an additional large case-control cohort consisting of 1,427 morbidly obese subjects and 783 never-obese elderly control subjects. In this round, three SNVs failed in genotyping, and one failed in primer design. Out of the remaining 21 SNVs (Supplementary Table 4), rs2076349 (P = 0.011 and 0.003 for the first and second validations, respectively) and rs142678624 (P = 0.026 and 0.047, respectively) consistently displayed a significant association with obesity (Supplementary Table 4) in the two rounds of validations. In the final joint data analysis with all genotyped subjects, rs2076349 showed strong association with obesity (P = 9.67 × 10⁻⁵; OR [95% CI], 1.84 [1.35–2.53]), whereas the association of rs142678624 with obesity was diminished. Moreover, the association of rs2076349 was demonstrated in both females (P = 0.002; OR [95% CI], 1.69 [1.21–0.2.36]) and males (P = 0.003; OR [95% CI], 4.19 [1.48–11.85]) (Table 3). Besides rs2076349, six other SNVs also showed nominal significant (P ≤ 0.05) associations with obesity in the final joint analysis of all genotyped subjects (Table 3).

Additionally, association analyses with quantitative traits were performed to observe possible influence of rs2076349 on obesity-related phenotypes. rs2076349 demonstrated strong associations with BMI (P = 0.0002; β = 2.8). The SNV also showed nominal associations with height (P = 0.0054; β = −1.66), body weight (P = 0.01; β = 6.00), and fasting serum leptin (P = 0.032; β = 6.93) (Table 4). Homozygous subjects for the rs2076349*T allele had the highest average BMI (mean 44.28 kg/m²) when compared with subjects who were heterozygous or homozygous for the rs2076349*C allele (Supplementary Table 5). The T allele was overrepresented in obese subjects (MAF 4%), especially in the homozygous form (10 obese vs. 1 control). To further evaluate any confounding effect of age or sex, we added these variables as covariates in regression models and tested their individual effects separately. The association of rs2076349 with BMI maintained strong with P = 0.0001 after adjustment by sex. The association between rs2076349 and BMI was with P = 0.0079 after adjustment for age (Supplementary Table 6). Analysis of association of rs2076349 with waist after adjustment for BMI was significant with P = 0.003 (Supplementary Table 7).

rs2076349 located at 209,800,230 bp on chromosome 1 (hg19) is a missense variant (V527M; valine to methionine) in the exon 12 (NM_001017402 and NP_001017402) of the LAMB3 gene. The variant is located in the laminin-type epidermal growth factor–like 5 domain of LAMB3. We used PolyPhen and PredictSNP to predict the possible impact of rs2076349 of protein function (16,17). According to Poplyphen, the SNV was predicted to be possibly damaging, with a score of 0.6. According to PredictSNP, which applies several independent algorithms to predict the pathogenic potential, rs2076349 is with 61–83% likelihood neutral in its effect on protein function.

rs2076349*T homozygous genotypes were confirmed by Sanger sequencing and TaqMan genotyping. The association of rs2076349 with obesity maintained significant with a Bonferroni adjusted P value of 0.002 (Supplementary Table 8).

LAMB3 mRNA expression was investigated in human subcutaneous WAT. The gene expression showed strong positive correlation with BMI (r = 0.38; P < 0.0001) (Fig. 2A). The expression was also positively correlated with adipose morphology (r = 0.25; P = 0.008) (Fig. 2B), meaning that hypertrophy (few large cells) was associated with high expression and the other way around for hyperplasia (many small cells). These correlations remained unchanged in multiple regression including age and array batch as independent factors. For 21 women, we had information about both LAMB3 genotype and adipose tissue LAMB3 expression levels. LAMB3 expression levels did not differ between genotypes (values not shown).

To study if LAMB3 affected adipogenesis in vitro, its expression was monitored during differentiation of hMSCs. The expression was increased almost twofold (day 13 of differentiation compared with the beginning, day 0) (Fig. 3A). To evaluate the possible role of LAMB3 in the differentiation of adipocyte cells, the gene was knocked down in vitro using siRNA in the hMSCs 24 h before induction of differentiation (day −1). Expression levels of LAMB3 3 days after the transfection of siRNA (day 2 of differentiation) were decreased by ∼90%. The expression of LAMB3 gradually recovered during the differentiation, but was still inhibited by 60% at day 12 of differentiation (Fig. 3B). Knockdown of LAMB3 also significantly decreased expression of adipocyte-specific gene ADIPOQ at days 7 (P < 0.01) and 9 (P < 0.05) of differentiation (Fig. 3B). Differentiation grade was also negatively affected, as evidenced by 15% decreased accumulation of neutral lipids (Fig. 3C).

This study has identified a previously unreported low-frequency coding SNV, rs2076349 (V527M) in the LAMB3 gene, associated with morbid obesity in a Swedish population using WES followed by genotyping. LAMB3 mRNA in WAT correlated with BMI and adipose morphology, and reduction of LAMB3 expression by siRNA inhibited fat cell differentiation, pointing to a specific role of LAMB3 in adipose tissue.

The minor allele T of rs2076349 was low frequency (MAF_Eur 2%) in the European population but enriched in obese subjects, especially among those homozygous, when compared with control subjects (10 vs. 1). Our bioinformatics analysis predicted that the LAMB3 V527M could possibly be damaging, indicating that it could be a causal marker for obesity. It is located in a domain of the gene for which function is not defined in detail. To our knowledge, rs2076349 has not been reported as a cause of other diseases.

LAMB3 forms one subunit of the Laminin 5 (also called Laminin-332) heterotrimer, which is a component of the extracellular matrix and thought to mediate the organization of cells into tissues (18). Laminin 5 is typically found in the dermoepidermal junction of the skin (19). Mutations in each of the three subunits of Laminin 5 are present in patients with junctional epidermolysis bullosa, an autosomal-recessive disease (20,21). Moreover, Laminin 5 induces osteogenic gene expression and promotes osteogenic differentiation of MSCs (18,22). Laminin 5 is also important for β-cell function (23) and has a role in various types of cancer (24–26). Whether Laminin 5 or LAMB3 has functions in organs of importance for the control of obesity such as the brain or WAT has not been described before.

Adipocyte number and size are the major determinants for fat mass, and obese subjects have a higher absolute production of new fat cells and a higher number of fat cells in total (27). About 10% of adipocytes in adult humans are renewed every year, suggesting that adipogenesis might contribute quite substantially to body fat mass (13,27). In agreement with this, GWAS support a role for adipogenesis in obesity susceptibility (2). Our observation that WAT LAMB3 mRNA levels correlate with BMI and that knockdown of LAMB3 in vitro inhibits of adipogenesis give additional support to the notion that production of fat cells is important for various aspects of development of obesity

The hypothesis underlying our study is that more severe forms of obesity might be due to low-frequency gene variants with a larger impact on the phenotype. A minor portion of morbidly obese individuals has been shown to be due to genetic variants with high penetrance (7,28). The association of rs2076349 in LAMB3 with obesity provides another supporting evidence for the notion that low-frequency variants contribute to morbid obesity. One explanation why rs2076349 has not been reported before as a cause of obesity is that this SNV has not been assayed in commonly used genome-wide genotyping platforms. Another explanation is that most GWAS have focused on BMI or common obesity, rather than morbid obesity. However, a noncoding SNP, rs12130212, located distal to the 3′ end of LAMB3, has been reported to be associated with extreme obesity in a Danish cohort (29).

One important limitation of the current study is relatively low sequencing depth and thereby low power to detect rare (MAF <1%) genetic variants causing morbid obesity in the general population. In contrast, analysis of association between such rare SNVs and obesity requires a very large sample size to achieve significance, which is beyond reach using available cohorts of morbidly obese cases. Another limitation is that although the association between SNV rs2076349 and obesity was prominent, it did not reach the level of genome-wide statistical significance. However, it is likely that the present level of association, in combination with functional data for the gene, represents a biological relevant link between LAMB3 and obesity.

In conclusion, a low-frequency SNV rs2076349 (V527M) in the LAMB3 gene is strongly associated with morbid obesity. This gene may be involved in the development of excess body fat at least in part by controlling adipogenesis and the production of new fat cells.

Acknowledgments. The authors thank the mutation analysis facility at the Karolinska Institutet and the SNP&SEQ Technology Platform at Uppsala University for performing genotyping and technical assistance. The authors also thank the Science for Life Laboratory, the National Genomics Infrastructure, and Uppmax for providing assistance in massive parallel sequencing and computational infrastructure; Anna Grauers (Clinical Science, Intervention and Technology, Karolinska Institutet) for scoliosis sample collection and characterization; Elisabeth Dungner (Lipid Laboratory, Department of Medicine, Huddinge, Karolinska Institutet) for DNA preparation; and Christel Björk (Lipid Laboratory, Department of Medicine, Huddinge, Karolinska Institutet) for the reverse-transfection protocol of hMSCs.

Funding. The project was supported by grants from Stockholm County (ALF), the Swedish Research Council, Novo Nordisk Foundation, the Diabetes Strategic Research Program at Karolinska Institutet, Center for Innovative Medicine, and the Erling-Persson Family Foundation.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. H.J. provided study design, data analysis, and wrote the first version of the manuscript. A.K. provided research data. E.N. and A.T. performed sample collection. P.G. provided the scoliosis sample collection. J.K., P.A., and I.D. provided study design. All authors contributed to revision of the first version of the manuscript and approved the final version. I.D. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.