SLC25A24 as a Novel Susceptibility Gene for Low Fat Mass in Humans and Mice

Context:

Genetic factors contribute to the development of obesity.

Objective:

The aim of this study was to identify novel genes that regulate body fat mass.

Design:

We performed a search for single nucleotide polymorphisms (SNPs) associated with body fat percentage using SNP arrays and undertook a replication study and animal study.

Setting and Patients:

Baseline examinations were conducted in 251 (first-stage analysis), 499 (second-stage analysis), and 732 (additional analyses) Japanese postmenopausal women. The mean age (mean ± SD) of the subjects was 66.5 ± 9.4 years. We also analyzed the fat-related phenotypes of a candidate gene in knockout mice.

Results:

In the analysis of total body fat, we focused on an SNP of SLC25A24 that showed the lowest significant P value obtained from multiple comparison tests among Japanese postmenopausal women. A significant association was also found between SLC25A24 SNPs and body mass index in the 1482 Japanese postmenopausal women examined in the study. The SLC25A24 SNPs affected the mRNA expression of SLC25A24 in human preadipocytes. Compared with wild-type mice, Slc25a24-KO mice had significantly lower body weights and white adipose tissue weights. Adipocyte differentiation was inhibited in Slc25a24-KO adipose tissues and Slc25a24-knockdown adipocytes.

Conclusions:

Genetic analyses in human and mouse models revealed the importance of SLC25A24/Slc25a24 in the regulation of body fat mass and adipogenesis.

Obesity is a key feature of metabolic disorders such as type 2 diabetes and cardiovascular disease (1, 2). Twin and adoption studies have shown that genetic factors play a role in the development of obesity (3–5). Genome-wide association studies (GWAS) are hypothesis-free approaches that involve scanning the entire human genome for novel genes/genome regions that have modest effects on complex human diseases. With the development of high-throughput genotyping techniques and the implementation of GWAS, common variations have been associated with obesity and body mass index (BMI). GWAS have identified at least 75 obesity susceptibility loci (6–11). Given the increasing prevalence of obesity and the aging of the population, understanding the factors associated with obesity in old age is critical. Whether these relationships persist in old age or in the Japanese population is unclear. Okada et al (11) have reported GWAS in East Asian subjects, including Japanese, which identified BMI-associated loci. BMI is a good indicator of adiposity risk, but it does not distinguish between fat and lean body mass. Body fat percentage, which is measured using dual-energy x-ray absorptiometry, is a more accurate measure of body composition. The purpose of this study was to conduct an unbiased analysis of genetic associations with body fat percentage in Japanese postmenopausal women to identify new loci associated with obesity in this population. We performed a large-scale analysis of these associations combined with a replication study between single nucleotide polymorphisms (SNPs) and body fat percentage in Japanese postmenopausal women. We noticed a common intronic SNP in the SLC25A24 gene; the association between the SNP and body fat percentage had the combined lowest P values in the first- and second-stage analyses. We also observed that Slc25a24-knockout mice displayed obesity-resistant phenotypes. Thus, genetic analyses in both human and mouse models revealed that SLC25A24/Slc25a24 is a novel candidate gene for obesity susceptibility.

Subjects and Methods

Study population

The present study was a prospective observational study conducted between 1993 and 2006. Baseline examinations were performed in 251 (first-stage analysis), 499 (second-stage analysis), and 732 (additional analyses) unrelated ambulatory postmenopausal volunteers living in the central area of Japan. The exclusion criteria were endocrine disorders such as hyperthyroidism, diabetes mellitus, renal disease, prior extensive gastrointestinal surgery, and use of medications known to affect fat metabolism. The mean age (mean ± SD) of the subjects was 66.5 ± 9.4 years. All subjects gave informed consent prior to the study which was approved by the ethics committees of the University of Tokyo and the participating clinical institutes.

Associations between fat percentage, BMI, and SLC25A24 polymorphisms

Dual-energy x-ray absorptiometry scans (DPX-L machine; GE Medical Systems Lunar Corporation, Madison, WI) of the total body were performed to determine fat percentage. We calculated the z scores of the total body fat percentage using a linear model optimized with the Akaike information criterion (12) to adjust for the effects of confounding factors such as age on body fat percentage variation.

We used the Affymetrix 50-K Hind SNP GeneChip (57,244 SNPs) to examine the genetic association of SNPs with obesity-related quantitative traits (total body fat percentage) according to the manufacturer's protocol. We used the data from the group of 251 subjects for the 50-K SNP array analysis (first-stage analysis) and selected 15 662 autosomal SNPs with that met the following criteria: genotypic call rate at least 95%, minor allele frequency (MAF) at least 10%, and Hardy-Weinberg equilibrium at least 0.0001. We then analyzed the association between total body fat percentage and the SNPs under the assumption of the dominant and recessive models for a minor allele in each SNP using the quantitative trait loci estimation model with the threshold model of Weller et al (13). For each SNP in the quantitative trait loci analysis, the genotypic value (total body fat percentage) was divided into two parts: the set of genotypic values corresponding to individuals with two minor alleles and others. To evaluate the effects of the genotypes on fat percentage, we categorized the z scores of total body fat percentage into two groups: scores lower and scores higher than the threshold value of the total body fat percentage z score. The following formula was used for the determination of the total body fat percentage z score: z score = total body fat percentage (%) − 37.73 + 0.077 × age (years). We performed a large-scale association test in 251 Japanese postmenopausal women to select SNPs based on the retrospective power analysis (14, 15). The significance levels for the statistical tests under the assumptions of dominant and recessive models were determined on the basis of the higher power score. We ultimately analyzed 15 662 SNPs selected using the criteria described above.

Thereafter, we recruited an additional 499 subjects for second-stage analysis. After the first and second analyses, we selected one SNP that was both significantly associated with the z scores of total body fat percentage and had the lowest P value. To examine the linkage disequilibrium (LD) region of SLC25A24 SNPs, the selected rs491785 SNP in intron 2 of the SLC25A24 and 12 additional SNPs in or near the SLC25A24 were genotyped using a fluorescence probe-based PCR method (TaqMan SNP Genotyping Assays; Applied Biosystems, Foster City, CA) in 750 subjects that were recruited for the first- and second-stage analyses. The 12 additional SNPs did not overlap with the SNPs of the Affymetrix 50-K Hind SNP GeneChip. For third-stage analysis, we measured the BMI of 1482 Japanese postmenopausal women recruited in the present study. The 1482 females included the 251 females from the first-stage analysis, the 499 females from the second-stage analysis, and additional 732 females. We analyzed the association of BMI with the rs491785 SNP and two additional SNPs (rs547364 and rs519129) that were located in intron 1 of the SLC25A24. All analyses were performed using the JMP software (SAS Institute, Cary, NC).

LD analysis

We genotyped rs491785 and 12 additional SNPs in close proximity around the SLC25A24 locus in 750 Japanese postmenopausal women. We used the Haploview software (MIT Broad Institute, Cambridge, MA) to analyze and visualize the state of LD among several SNPs (r²) and between adjacent SLC25A24 SNPs (16).

Human cell culture and qRT-PCR

Human primary preadipocytes obtained from mesenteric adipose tissue using collagenase as described previously (17) were purchased from the Japan Human Sciences Foundation (Osaka, Japan). Briefly, fat tissue was minced and digested in Hank's balanced salt solution containing 1 mg/mL collagenase and 7.5% fetal bovine serum (FBS) in a 37°C shaking water bath until fragments were no longer visible and digests had a milky appearance. The digests were filtered, centrifuged, and treated with an erythrocyte lysis buffer to enhance subsequent differentiation. The cells were then maintained in phosphoglucomutase-2 medium (Lonza, Basel, Switzerland). This study was approved by the ethics committee of the University of Tokyo and Japan Human Sciences Foundation. DNA samples were collected from preadipocytes, were obtained from nine subjects. The SNPs in the SLC25A24 gene were genotyped using a fluorescence probe-based PCR method. The expression levels of SLC25A24 mRNA in human primary preadipocytes were examined by qRT-PCR using the Prism 7000 System (Applied Biosystems) and SYBR Green I fluorescence as previously described (18, 19). cDNA was synthesized from 1 μg of total RNA using a First Strand cDNA Synthesis Kit (GE Healthcare, Life Sciences, Pittsburgh, PA). The relative SLC25A24 mRNA level normalized to that of a reference gene, GAPDH, was determined using the comparative cycles at threshold fluorescence (Ct) method. All experiments were independently repeated three times. The sequences of the PCR primers were listed in Supplemental Table 1.

Slc25a24 targeting strategy and generation of knockout mice

The SV40polyA-Neo cassette replaced exon 2 of the mouse Slc25a24 gene, which contains the ATG initiation codon in exon 1. The linearized targeting vector was electroporated into C57BL/6J embryonic stem cells (BRUCE-4 ES cells). G418-resistant ES cell–targeted clones were isolated and confirmed by PCR and Southern blotting on the two homologous arms with gene-specific probes. In addition, a Neo probe was used to screen for random insertion events. Chimeras were generated by injecting the positive ES cells into the blastocysts of BALB/c mice and were identified by coat color. Chimeras were mated to C57BL/6J females to generate F1 heterozygous mice for the knockout (KO) line. Heterozygous F1 animals from each clone were intercrossed to produce homozygous mutants. Experiments were conducted on C57BL/6 (wild type [WT]) and Slc25a24-knockout female mice maintained in a temperature-controlled room (28°C) with a 12-hour light/dark cycle. mRNA and protein expression levels of Slc25a24 in inguinal white adipose tissue (WAT) removed from WT and Slc25a24-knockout (KO) mice were analyzed. Relative amounts of Slc25a24 mRNA were determined by qRT-PCR. Rabbit antisera against slc25a24 were raised against synthetic peptides. The protein levels of Slc25a24 were analyzed on immunoblots using an anti-SLC25A24 antibody.

Body weight study

All female mice were fed a chow diet (21.6% fat, 23% protein, and 55.4% carbohydrate) until 8 weeks of age. To generate diet-induced obesity, some mice were subsequently fed a high-fat diet (60% fat, 20% protein, and 20% carbohydrate) until the end of the experiment (high-fat diet group). Other mice were fed a chow diet (normal diet group). For body weight study, body weights were measured weekly until 20 weeks of age. Differences between the mean values of mouse samples were analyzed using the unpaired Student t test.

qRT-PCR in mouse tissues and adipocytes

Testis, heart, liver, kidney, muscle, brown adipose tissue, and inguinal WAT from the WT mice are also collected. Total RNA was extracted by the RNeasy Lipid Tissue Kit (QIAGEN, Valencia, CA) according to the protocol supplied by the manufacturer. As suggested by this protocol, samples were digested with DNase. The quality and quantity of total RNA were determined by spectrophotometry (model ND-1000, NanoDrop Technologies, Wilmington, DE). First-strand cDNA was synthesized from 1 μg total RNA using a First Strand cDNA Synthesis Kit (GE Healthcare Life Sciences). The relative levels of mouse Slc25a24 mRNA normalized to a reference gene (Hprt1) were determined using the comparative Ct method. Samples without reverse transcriptase were used as negative controls.

We also analyzed the effect of a high-fat diet on the mRNA expression of SLC25A24. Mice were fed regular laboratory chow (normal diet; n = 7) or a high-fat diet (high-fat diet; n = 7) for 4 weeks. Relative amounts of Slc25a24 mRNA were determined using qRT-PCR and normalized to those of mouse Hprt1.

Murine 3T3-L1 adipocytc cells were obtained from American Type Culture Collection (Manassas, VA). Cells were maintained in DMEM supplemented with 10% fetal bovine serum (FBS), 50 U/mL penicillin, and 50 μg/mL streptomycin. For adipocyte differentiation, the cells were grown to confluence (day 2), and differentiation was induced 2 days postconfluence (day 0) by substituting the medium with a differentiation growth medium (standard medium containing a 10% FBS and MDI mixture, which is a mixture of 0.5mM isobutylmethylxanthine (Sigma, St. Louis, MO), 1μM dexamethasone (Sigma), and 10 μg/mL bovine insulin (Sigma). After 2 days, this medium was substituted with a postdifferentiation medium (standard medium containing 10% FBS and 10 μg/mL insulin).

Synthetic small interfering RNAs (siRNAs) targeting mouse Slc25a24 were purchased from Life Technologies Incorporation (Rockville, MD). 3T3-L1 cells were incubated with 400nM siRNA against either Slc25a24 (siSlc25a24) or the negative control using HiPerFect transfection reagent (QIAGEN) for 48 hours. Transfected cells were subjected to adipogenic induction as described above. The mRNA was collected at day 5, and the mRNA expression levels of murine ap2, Pparγ, C/ebpα, C/ebpβ, and C/ebpδ in 3T3-L1 cells were evaluated by qRT-PCR on the Prism 7000 System and SYBR Green I fluorescence as previously described (18, 19). Tissues and mRNAs from the Slc25a24-KO and WT mice are also collected.

cDNA was synthesized from 1 μg total RNA using a First Strand cDNA Synthesis Kit (GE Healthcare Life Sciences). The relative levels of mouse Slc25a24, ap2, Pparγ, C/ebpα, C/ebpβ, and C/ebpδ mRNA normalized to that of a reference gene (Gapdh) were determined using the comparative Ct method.

The sequences of the PCR primers are listed in Supplemental Table 1.

Measurement of triglycerides content

Liver from the WT and Slc25a24-KO mice are collected. The 3T3-L1 preadipocytes were differentiated with differentiation medium (Takara, Tokyo, Japan) for 6 days and collected. Tissues and cells were washed with PBS and extracted with Lipid Extraction Solution (BioVision, Inc., Milpitas, CA). Extracted cellular triglycerides content was measured using the Adipogenesis Assay Kit according to the manufacturer's protocols. Proteins were quantified by The bicinchoninic acid assay methods (Pierce, Rockford, IL) and total cellular triglycerides content was normalized to the cellular protein content.

Results

Large-scale association of SNPs with total body fat percentage

We used the Affymetrix 50-K Hind SNP GeneChip (57 244 SNPs) to examine the genetic association of SNPs with total body fat percentage in 251 subjects. The basic characteristics of the 251 subjects are shown in Supplemental Table 2. We analyzed 57 244 SNPs, from which we selected 15 662 with genotypic call rates (SNP CR) of at least 95%, MAF of at least 10%, and Hardy-Weinberg equilibrium of at least 0.0001. Table 1 shows the top 20 SNPs of the dominant model and the top 20 SNPs of the recessive model (Table 1). Among these 40 SNPs, we focused on the 21 SNPs located near and in the RefSeq protein-coding genes to search for the novel genes that regulate fat percentage. Twenty-one SNPs were located in or adjacent to (within 60 kb) the genes. We successfully genotyped the selected 21 SNPs during the second-stage analysis in an additional 499 postmenopausal women not included in the first-stage analysis. The basic characteristics of the 499 subjects are shown in Supplemental Table 3. Among these SNPs, we identified four SNPs (rs491785, rs2237682, rs8005131, and rs1475524) that were associated with a low body fat percentage phenotype and had significant combined P values (2.04 × 10⁻⁸ to 2.73 × 10⁻⁷) for the first- and second- stage analyses (Table 2 and Supplemental Table 4). The significance level was set at 3.19 × 10⁻⁶, which was calculated as 0.05/15,662 using Bonferroni's correction. Among these four SNPs, rs491785 showed the lowest P value associated with total body fat percentage in postmenopausal women (P = 2.04 × 10⁻⁸; Table 2). Rs491785 was located in intron 2 of SLC25A24. We therefore considered this gene a good candidate for further investigation.

Table 1.

First-Stage Analysis of Single Nucleotide Polymorphisms (SNPs) for Association With Total Body Fat Percentage

SNP	Gene Symbol	Chromosome	MAF	P Value	Power
Dominant model
rs2139681		12q21.1	0.17	3.91 × 10−5	0.994
rs761002	TGM2	20q21.1	0.40	1.54 × 10−4	0.984
rs530629		Xq27.3	0.26	1.93 × 10−4	0.969
rs10512844	FGF10	5p12	0.19	2.32 × 10−5	0.968
rs1989458		21q21.1	0.43	2.14 × 10−4	0.967
rs8005131	NPAS3	14q13.1	0.21	4.78 × 10−4	0.962
rs6955887	LOC441282	7q33	0.46	4.53 × 10−5	0.957
rs10509490	LRCC22	10q23.1	0.23	1.04 × 10−4	0.946
rs9293311		5q14.2	0.46	4.09 × 10−5	0.944
rs4940770	ONECUT2	20q21.3	0.33	1.19 × 10−4	0.943
rs10511208		3q13.1	035	1.25 × 10−4	0.937
rs491785	SLC25A24	1p13.3	0.33	1.87 × 10−4	0.936
rs9309361	AFTPH	2p14	0.42	1.66 × 10−4	0.936
rs10489251		1q24.3	0.18	8.16 × 10−5	0.931
rs10520541	ODZ3	4q35.1	0.22	1.14 × 10−4	0.928
rs1475524	DAPK1	9q21.3	0.28	8.72 × 10−4	0.924
rs437381		14q31.1	0.49	7.87 × 10−4	0.921
rs10506368		12q14.1	0.37	6.41 × 10−4	0.921
rs10509628		10q23.3	0.13	2.78 × 10−4	0.919
rs7897937	KIAA1600	10q25.3	0.21	9.75 × 10−4	0.918
Recessive model
rs514536		11q14.1	0.37	3.41 × 10−5	0.997
rs1887068	NPAS3	14q13.1	0.20	1.61 × 10−6	0.996
rs6877544		5q13.1	0.33	1.83 × 10−5	0.991
rs258862	FAM44B	5q35.2	0.30	5.32 × 10−5	0.985
rs1436932		9q21.3	0.46	1.29 × 10−5	0.983
rs453687		21q21.3	0.43	1.71 × 10−5	0.983
rs1467527		14q22.1	0.42	6.02 × 10−6	0.974
rs7224569		17p13.3	0.43	1.89 × 10−4	0.971
rs2237682	DLD	7q31.1	0.41	8.96 × 10−5	0.970
rs7029362		9q21.3	0.50	3.48 × 10−5	0.969
rs10511684	KIAA1797	9p21.3	0.44	4.02 × 10−6	0.969
rs4674696		2q36.1	0.10	4.49 × 10−6	0.963
rs2171107	LRP1B	2q22.1	0.47	3.81 × 10−4	0.959
rs1521849		15q14	0.30	1.57 × 10−4	0.958
rs10511685	KIAA1797	9p21.3	0.47	7.74 × 10−6	0.957
rs517735	LOC391048	1p31.1	0.39	2.72 × 10−4	0.955
rs9301789	GPC5	13q31.3	0.37	3.53 × 10−4	0.955
rs867410	ESRRG	1q41	0.21	1.41 × 10−4	0.954
rs9309977		3p12.1	0.19	2.42 × 10−5	0.953
rs2074215	ACNN1	17q11.2	0.43	9.34 × 10−5	0.953

Genome-wide SNPs (57 244 SNPs) were screened in 251 Japanese postmenopausal women. The top 20 SNPs of the dominant model and the top 20 SNPs of the recessive model in the first-stage analysis are shown.

Table 2.

Results of the Second-Stage Analysis of SNPs for Association With Total Body Fat Percentage

SNP	Gene Symbol	Chromosome	P Valuea	P Valueb
rs491785	SLC25A24	1p13.3	1.09 × 10−4	2.04 × 10−8
rs2237682	DLD	7q31.1	3.69 × 10−3	3.31 × 10−7
rs8005131	NPAS3	14q13.1	4.67 × 10−3	2.23 × 10−6
rs1475524	DAPK1	9q21.3	3.13 × 10−3	2.73 × 10−6

^a

P value in the population of second-stage analysis.

^b

Combined P values between the first- and second- stage analyses.

Association of SLC25A24 SNPs with BMI and mRNA expression

In the second-stage analysis of the 21 top hits in and around genes, rs491785 SNP showed the lowest P value (Table 2: P = 2.04 × 10⁻⁸). We focused on this SNP for further analysis. Rs491785 is located in intron 2 of the SLC25A24. To determine whether rs491785 was located in an LD block in the SLC25A24, we genotyped 12 additional SNPs close to rs491785 that were either within or near the SLC25A24 in 750 subjects that were recruited for the first- and second-stage analyses (Supplemental Figure 1).

We focused on the rs491785 SNP, which was identified after the first- and second-stage analyses, for further study. To investigate whether the rs491785 SNP was associated with BMI, we used the data from the 1482 Japanese postmenopausal women whose BMI was measured. The 1482 females include the subjects studied in the first- and second-stage analyses, and additional 732 females. Although P values have not been corrected for multiple testing, we observed an association of rs491785 with BMI in the 1482 Japanese postmenopausal women (P = .022; Figure 1A). As shown in Supplemental Figure 1A, rs519129 and rs547364 are located in the same LD region as rs491785. They are located in intron 1 of the SLC25A24. GWAS have shown that genetic variations in intron 1 play important roles in transcriptional regulation of gene expression (20–23). Therefore, we also genotyped these two additional SNPs (rs519129 and rs547364) in the 1482 Japanese postmenopausal women and analyzed their association with BMI. We observed associations of rs547364 and rs519129 with BMI (P = .024 and 0.044, respectively; Figure 1, B and C). The basic characteristics of the 1482 subjects are shown in Supplemental Table 5. To investigate the association of SLC25A24 SNPs with mRNA expression, we used qRT-PCR, which allowed the study of the expression levels of SLC25A24 mRNA in primary human preadipocytes from nine subjects (Figure 1D). The SLC25A24 mRNA levels in human preadipocytes with the GG genotype of the rs491785 SNP were higher than those in preadipocytes with the AA or AG genotypes (Figure 1D). Three SNPs (rs491785, rs547364, and rs519129) were tightly associated with each other among the individuals due to high LD in the region. The SLC25A24 mRNA levels in human preadipocytes with the TT genotype of the rs519129 SNP and the GG genotype of the rs547364 SNP were higher than those in preadipocytes with other genotypes.

Figure 1.

SNPs in SLC25A24 affect the BMI and mRNA expression of SLC25A24.

A–C, SNPs in SLC25A24 affect BMI. A, BMIs are shown for the AA, AG, and GG genotypes of the rs491785 SNP in intron 2 of SLC25A24. BMIs are expressed as means ± s.e. The number of subjects is shown in parentheses. The significant association between genotypes (the AA vs GG groups) and BMI was determined using the unpaired Student t test (P = .022). B, BMIs are shown for the CC, CT, and TT genotypes of the rs519129 SNP in intron 1 of SLC25A24. The significant association between genotypes (the CC vs TT groups) and BMI was determined using the unpaired Student t test (P = .024). C, BMIs are shown for the CC, CG, and GG genotypes of the rs547364 SNP in intron 1 of SLC25A24. The significant association between genotypes (the CC vs GG groups) and BMI was determined using the unpaired Student t test (P = .044). D, SNPs in SLC25A24 affect SLC25A24 mRNA expression in human primary preadipocytes. We used qRT-PCR to analyze the SLC25A24 mRNA levels in primary human preadipocytes of nine subjects with the AA, AG, or GG genotypes. The SLC25A24 mRNA levels in human preadipocytes with the GG genotype of the rs491785 SNP were higher than those in human preadipocytes with the AA or AG genotypes.

Expression of mouse Slc25a24 in white adipose tissue

The genomic structure of human SLC25A24 and mouse Slc25a24 are shown in Supplemental Figure 1, A and B. These human and mouse genes encode the same 10 exons. The mouse Slc25a24 protein is very similar to the human SLC25A24 protein in terms of total amino acid sequence length, EF hand, calcium binding motif (93% homology) and mitochondrial carrier motif (93% homology) (Supplemental Figure 2, A and B). The tissue distribution of mRNAs for mouse Slc25a24 determined qRT-PCR is summarized in Figure 2A. We found that Slc25a24 mRNA was most strongly expressed in the testis, which agrees with previous reports in human tissues (24, 25). Slc25a24 mRNA was strongly expressed in the white adipose tissue as well. Slc25a24 expression was increased in the white adipose tissues of mice that consumed the high-fat diet (Figure 2B) and this was accompanied by an increase in the weight of the white adipose tissue. These data suggest that the amount of SLC25A24 may affect body fat percentage.

Figure 2.

Expression of Slc25a24 and targeted disruption of the mouse Slc25a24 locus.

A, Tissue distribution of mRNAs of the mouse Slc25a24. Slc25a24 mRNA levels in the mouse testis, heart, liver, kidney, muscle, brown adipose tissue, and inguinal WAT were normalized to those of Hprt1, analyzed by qRT-PCR, and expressed as means ± SD values. The highest levels of Slc25a24 mRNA expression were found in the testis. High levels of Slc25a24 mRNA were also found in WAT. B, Effect of a high-fat diet on the mRNA expression of SLC25A24. Mice were fed regular laboratory chow (normal diet; n = 7) or a high-fat diet (high fat diet; n = 7) for 4 weeks. Relative amounts of Slc25a24 mRNA were determined using qRT-PCR and normalized to those of Hprt1, which was used as internal control. Data are presented as means ± SD. *, P < .01 (unpaired Student t test). C and D, mRNA and protein expression levels of Slc25a24 in inguinal WAT removed from wild-type (WT) and Slc25a24-knockout (KO) mice. Relative amounts of Slc25a24 mRNA were determined by qRT-PCR (C). The protein levels of Slc25a24 were analyzed on immunoblots using an anti-SLC25A24 antibody (D). **, P < .01 (unpaired Student t test).

Resistance to high-fat diet-induced obesity in Slc25a24-knockout mice

To investigate the direct effect of Slc25a24 on fat mass and body weight, we generated Slc25a24-knockout mice. The strategy for targeted deletion of the endogenous mouse gene Slc25a24 is outlined in Supplemental Figure 2. To disrupt Slc25a24, we constructed a targeting vector to replace exon 2 of Slc25a24 with a neomycin-resistant gene (neo) cassette (Supplemental Figure 2). The mRNA and protein expression of Slc25a24 were depleted in the Slc25a24-knockout white adipose tissue (Figure 2, C and D). Slc25a24 homozygous mutants grew normally and without obvious external phenotypic defects. Both sexes were fertile. The Slc25a24-knockout and WT mice were fed a high-fat diet from 8–20 weeks of age. The initial weights of the Slc25a24-knockout and WT mice were not significantly different. The Slc25a24-knockout mice weighed less than the WT mice when consuming the high-fat diet and this difference became statistically significant at 13 weeks of age (Figure 3A). We also analyzed the organs of the mice at 20 weeks of age. The weights of the WATs and livers of the Slc25a24-knockout mice were significantly lower than those of the WT mice (Figure 3, B–D). Slc25a24-knockout mice exhibited reduced triglyceride deposition in the liver (Supplemental Figure 4). The weights of the kidneys of the Slc25a24-knockout and WT mice were similar (Figure 3E).

Figure 3.

Total-body and organ weights of Slc25a24-knockout (KO) mice.

A, Body weight growth curves in Slc25a24-KO mice (n = 5) and WT (n = 7) mice. Slc25a24-KO and WT mice were fed a high-fat diet from 8 weeks to 20 weeks of age. The initial weights of the Slc25a24-KO and WT mice were not significantly different. Slc25a24-KO mice showed an obesity-resistant phenotype. The body weight of Slc25a24-KO mice was lower than that of WT mice on the high-fat diet; the difference in weights became statistically significant at 13 weeks of age. *, P < .05 (unpaired Student t test). B–E, Organ weight of WT and Slc25a24-KO mice. Compared with those of WT mice, the weights of inguinal WAT, kidney WAT, and liver of Slc25a24-KO mice were significantly lower (unpaired Student t test). B–D, The kidney weights of Slc25a24-KO and WT mice were similar (E). N.S., not significant.

Role of Slc25a24 in adipocyte differentiation

Slc25a24 mRNA was expressed in high amounts in the WAT, suggesting that the amount of Slc25a24 may affect the fat metabolism. We therefore analyzed gene expression in the WAT of Slc25a24-knockout mice (Figure 4, A–E). Interestingly, expression of adipocyte differentiation markers, including ap2, Pparγ, C/ebpα, C/ebpβ, and C/ebpδ, in the Slc25a24-knockout white adipose tissue was reduced compared with that in the WT white adipose tissue. Under high-fat-diet conditions, the expression of mature adipocyte marker ap2 and master regulators of adipocyte differentiation Pparγ and C/ebpα was reduced in the Slc25a24-knockout WAT. Next, we investigated the effects of the knockdown of Slc25a24 by siRNA on the expression of adipogenesis-related genes in vitro. mRNA and protein expression of Slc25a24 were dramatically reduced by the knockdown of Slc25a24 by siRNA in 3T3L1 cells (Supplemental Figure 5, A and B). The expression of mature adipocyte markers ap2, Pparγ, C/ebpα, C/ebpβ, and C/ebpδ in Slc25a24-knockdown 3T3L1 cells was lower than that in control siRNA-treated 3T3-L1 cells after differentiation (Figure 4, F–J). After the induction of differentiation, we also analyzed the intracellular lipid accumulation. Compared with the control siRNA–treated 3T3L1 cells, the triglyceride content was significantly reduced in Slc25a24-knockdown 3T3L1 cells. These results are in line with the hypothesis that low fat mass is a direct consequence of Slc25a24 depletion in the WAT, which may explain the reduced adiposity in the mutant mice.

Figure 4.

Effect of Slc25a24 knockout and Slc25a24 knockdown on differentiation markers of the WAT and 3T3–L1 cells.

A–E, Murine ap2, Pparγ, C/ebpα, C/ebpβ, and C/ebpδ mRNA expression in inguinal WAT removed from WT and Slc25a24-knockout (KO) mice was analyzed by quantitative qRT-PCR. F–J, Murine ap2, Pparγ, C/ebpα, C/ebpβ, and C/ebpδ mRNA expression in 3T3–L1 cells treated with control siRNA (siCont) or Slc25a24-specific siRNA (siSlc25a24) was analyzed by qRT-PCR. Relative mRNA levels were normalized to those of the reference gene, Hprt1, determined using the comparative Ct method. Data are presented as means ± SD. **, P < .01 (unpaired Student t test).

Discussion

In the present study, we conducted large-scale first- and second- stage analyses that helped identify rs491785 located in SLC25A24 as a novel candidate SNP associated with body fat percentage in Japanese postmenopausal women. LD analysis also revealed the presence of nine SNPs in a 30-kb region of high LD from the 5′-flanking region to intron 6 in SLC25A24. These SNPs were also associated with BMI in 1482 Japanese postmenopausal women. The data revealed that the subjects with major alleles had lower body fat percentage and BMI phenotypes than those with minor alleles. Human primary preadipocytes with major alleles expressed lower levels of SLC25A24 mRNA than those with minor alleles. These data suggest that the expression level of human SLC25A24 may affect body fat percentage and BMI.

This study provided important and new information defining not only the association between SLC25A24 SNPs and body fat percentage in humans but also the important role of Slc25a24 in mouse fat metabolism. Slc25a24-knockout mice had an obesity-resistant phenotype. The current results also demonstrate that systemic Slc25a24 deficiency results in low adipose tissue content, thus confirming the importance of SLC25A24/Slc25a24 as a regulator of fat homeostasis. Concordant with reduced body weight, decreased amounts of WATs were observed in Slc25a24-knockout mice. The mean body weight of heterozygous knockout mice was lower than that of WT mice and higher than that of homogenous knockout mice (WT vs heterozygous vs KO (mean ± SE): 34.1 ± 1.5 g vs 26.9 ± 1.7 g vs 25.7 ± 2.3 g). These data suggest that, similar to the human gene, the expression level of mouse Slc25a24 may affect body weight. Despite these findings, the function of SLC25A24/Slc25a24 in fat tissue generation remains to be elucidated. qRT-PCR analysis showed that mouse Slc25a24 was dominantly expressed in the WAT. However, we noted that C/ebpα and Pparγ expression in the WAT derived from the Slc25a24-knockout mice was lower than that in the WT white adipose tissue. C/ebpα and Pparγ are essential for adipogenesis (26–28). These data suggest that SLC25A24/Slc25a24 may be critical for adipocyte differentiation. Several findings support the role of SLC25A24/Slc25a24 in terminal adipocyte differentiation. First, the mature adipocyte marker ap2 was significantly decreased during in vitro adipocyte differentiation in Slc25a24-knockdown 3T3-L1 cells. Second, Slc25a24 expression was increased in the white adipose tissue after the feeding of a high-fat diet.

Another study has proposed that ATP-Mg²⁺/Pi carriers located in the inner mitochondrial membrane are involved in the uptake and accumulation of adenine nucleotides (29). These carriers catalyze the electroneutral exchange of ATP-Mg²⁺ for Pi (30). Importantly, adenine nucleotide transport activity can be activated by cytosolic Ca²⁺ (31, 32). Although structural analysis of purified ATP-Mg²⁺/Pi carriers from the mitochondrial membrane has not been achieved (33), structural and functional evidence suggests that SLC25A24 encodes properties of ATP-Mg2+/Pi carriers (34, 35). SLC25A24 is a member of a family of inner membrane mitochondrial carriers (MCs) that shuttle metabolites, nucleotides, and cofactors between the cytosol and mitochondria (24, 25). SLC25A24 belongs to a subgroup of short calcium-binding MCs (SCaMCs) with four paralogs in mammals: SCaMC-1/SLC25A24, SCaMC-2/SLC25A25, SCaMC-3/SLC25A23, and SCaMC-3-like/SLC25A41 (24, 25, 36). The transporter consists of a C-terminal domain comprising six transmembrane helices homologous to the MC proteins and an N-terminal domain with Ca²⁺-binding EF hands, which confer Ca²⁺ sensitivity to the carrier. Recent studies have shown that SLC25A24 is the dominant isoform of the ATP-Mg/Pi carriers in cancer cells and is highly overexpressed in a series of in vivo tumors and cell lines (37). These structural properties, together with the finding that SLC25A24/Slc25a24 is expressed at higher levels in the white adipose tissue and tumors, suggest that SLC25A24 has an important function in controlling ATP and Ca²⁺ homeostasis in these tissues.

Previous reports have documented the phenotypes of Slc25a25- and Slc25a23-knockout mice. These genes are paralogs of Slc25a24 (38, 39). Slc25a25-knockout mice are resistant to diet-induced obesity and show reduced physical endurance on a treadmill, indicating reduced metabolic efficiency (38). In this previous report, food intake was significantly higher in WT mice than Slc25a25-knockout mice at 28°C. Here, we also showed that Slc25a24-knockout mice were resistant to diet-induced obesity. However, we did not record metabolic efficiency from body weight and food intake data. Similar to Slc25a25-knockout mice, Slc25a24-knockout mice may have shown reduced metabolic efficiency and food intake. Furthermore, we did not measure glucose tolerance, insulin resistance, or hyperlipidemia in the present study. Future studies are required to uncover the mechanisms underlying this metabolic disorder. In this study, we analyzed inguinal and kidney WAT. Analyses of other WAT samples including mesenteric, peri-renal, and sc fat will be useful to understand the role of Slc25a24 in metabolic disorders. Slc25a23-knockout mice (Slc25a23 is also called SCaMC-3) have reduced plasma urea levels that may be due to a limited function of the urea cycle in the liver (39). Slc25a25 and Slc25a23 are dominantly expressed in the muscle and liver, respectively. Taken together, the present results suggest that the Slc25a24, Slc25a25, and Slc25a23 signaling pathways have distinct roles and may cooperate in the regulation of fat, muscle, and liver metabolism.

A limitation of this study was the small number of markers genotyped in the first discovery stage. It is likely that some associations could have been overlooked due to the rather low density of the array. Thus, additional genes that were not included in this study may be associated with obesity. Another limitation of this study was the small number of samples used to estimate mRNA levels from human preadipocytes. A larger number of human adipocyte samples are required to clarify the correlation between human SLC25A24 SNPs and SLC25A24 mRNA levels.

In conclusion, we demonstrated an association between SNPs in SLC25A24 and body fat percentage in Japanese women. Therefore, SLC25A24 genotyping may be beneficial in the prevention and management of obesity. The present findings of the correlation of SLC25A24 polymorphism and fat percentage provide a promising new direction for the clinical management of obesity that may herald the development of new diagnostic markers as well as therapeutic options based on this molecular target.

Acknowledgments

We thank E. Sakamoto for technical assistance, and S. Kamitsuji (Stagen) for statistical analysis. We appreciate all the volunteers and participating institutions for precious clinical data and samples.

This study was supported by Cell Innovation Program and Support Project of Strategic Research Center in Private Universities from the Ministry of Education, Culture, Sports, Science and Technology, grants from the Japan Society for the Promotion of Science, grants-in-aid from the Ministry of Health, Labour and Welfare, and the Advanced Research for Medical Products Mining Programme of the National Institute of Biomedical Innovation.

Disclosure Summary: The authors have nothing to disclose.

Abbreviations

 
• BMI

  body mass index

 
• Ct

  threshold fluorescence

 
• FBS

  fetal bovine serum

 
• GWAS

  genome-wide association studies

 
• LD

  linkage disequilibrium

 
• MAF

  minor allele frequency

 
• MC

  mitochondrial carrier

 
• siRNA

  small interfering RNA

 
• SNP

  single nucleotide polymorphism

 
• WAT

  white adipose tissue

 
• WT

  wild type.