Functional Single-Nucleotide Polymorphisms in the Secretogranin III (SCG3) Gene that Form Secretory Granules with Appetite-Related Neuropeptides Are Associated with Obesity

Abstract

Context: Genetic factors are important for the development of obesity. However, the genetic background of obesity still remains unclear.

Objective: Our objective was to search for obesity-related genes using a large number of gene-based single-nucleotide polymorphisms (SNPs).

Design and Setting: We conducted case-control association analyses using 94 obese patients and 658 controls with 62,663 SNPs selected from the SNP database. SNPs that possessed P ≤ 0.02 were further analyzed using 796 obese and 711 control subjects. One SNP (rs3764220) in the secretogranin III (SCG3) gene showed the lowest P value (P = 0.0000019). We sequenced an approximately 300-kb genomic region around rs3764220 and discovered SNPs for haplotype analyses. SCG3 was the only gene within a haplotype block that contained rs3764220. The functions of SCG3 were studied.

Patients: Obese subjects (body mass index ≥ 30 kg/m², n = 890) and control subjects (general population; n = 658, body mass index ≤ 25kg/m²; n = 711) were recruited for this study.

Results: Twelve SNPs in the SCG3 gene including rs3764220 were in almost complete linkage disequilibrium and significantly associated with an obesity phenotype. Two SNPs (rs16964465, rs16964476) affected the transcriptional activity of SCG3, and subjects with the minor allele seemed to be resistant to obesity (odds ratio, 9.23; 95% confidence interval, 2.77–30.80; χ² = 19.2; P = 0.0000067). SCG3 mRNA and immunoreactivity were detected in the paraventricular nucleus, lateral hypothalamic area, and arcuate nucleus, and the protein coexisted with orexin, melanin-concentrating hormone, neuropeptide Y, and proopiomelanocortin. SCG3 formed a granule-like structure together with these neuropeptides.

Conclusions: Genetic variations in the SCG3 gene may influence the risk of obesity through possible regulation of hypothalamic neuropeptide secretion.

OBESITY HAS BECOME one of the major issues in public health, medicine, and the economy (1). In particular, visceral obesity is considered to be important due to its relation to various complications such as diabetes mellitus, dyslipidemia, and hypertension. A combination of these dysfunctions is now defined as the metabolic syndrome (2), which significantly increases the risk of cardiovascular disease. Adipose tissue secretes various adipokines, and an increase in adipose tissue mass affects the level of adipokines, resulting in the development of dyslipidemia, hypertension, and insulin resistance (3, 4).

Both genetic and environmental factors contribute to the development of obesity. In epidemiological studies, heritability of body weight is estimated to be approximately 70% (5, 6). Genetic studies in mice suggested that mutations in several genes, such as leptin, proopiomelanocortin (POMC), and melanocortin-4 receptor, were implicated in a monogenic form of inherited obesity, whereas mutations in such genes were also reported in human subjects with obesity (6, 7). However, the most prevalent MC4R gene mutations have been found in only 3–5% of obese patients with a body mass index (BMI) of more than 40 kg/m². In general, the vast majority of obesity is considered to be caused by a polygenic disorder, and its genetic susceptibility is likely to differ among various ethnic groups (6, 7). A large number of manuscripts concerning obesity-related genes have been reported (7). However, because there are also many papers reporting controversial results at these candidate loci, the genetic background of obesity still remains unclear.

As one of the Japanese Millennium Projects, a large-scale collaborative effort performed a search for gene-based single-nucleotide polymorphisms (SNPs) in a group of Japanese subjects and discovered approximately 190,000 genetic variations (JSNP database) (8), and subsequently our center developed a high-throughput SNP genotyping system that uses a combination of multiplex PCR and the Invader assay (9, 10) to effectively determine these variations’ frequencies in the Japanese population. We performed an association study using a large number of SNPs selected from the JSNP database (62,663 SNPs in 11,932 genes, covering approximately 30% of the human genome) by genotyping Japanese obese and lean subjects and found that one SNP (SNP-1, rs3764220) showed the smallest P value and was significantly associated with obesity. This SNP existed in the 5′-flanking region of the secretogranin III (SCG3) gene. SCG3 belongs to a family of acidic secretory proteins, known as granins, which are widely expressed in endocrine and neuronal cells (11). SCG3 has been cloned from brain- and pituitary-specific mRNA and is expressed in the paraventricular nucleus (PVN) of the hypothalamus (12), which is known to be an important region for appetite regulation. SCG3 is also expressed in pancreatic β-cells and participates in insulin secretion together with chromogranin (CHG) A (13). Interestingly, SCG3 is located on chromosome 15q21, on which association with obesity has been previously indicated (14). Data from the Framingham Heart Study suggested a moderate linkage of the metabolic syndrome to this general region on chromosome 15q (15) on which the presence of a susceptibility gene for type 2 diabetes in the Japanese population has also been indicated (16).

In the present study, we demonstrate a significant association between functional SNPs in the SCG3 gene and obesity. We found that SCG3 was expressed together with appetite-regulating peptides such as orexin and melanin-concentrating hormone (MCH) in the lateral hypothalamic area (LHA) and neuropeptide Y (NPY) and POMC in the arcuate nucleus (ARC), suggesting that SCG3 is a good candidate as an obesity-related gene.

Subjects and Methods

Subjects

The sample size of the first set of Japanese obese subjects (BMI ≥ 30 kg/m²) was 94 (case 1; male to female ratio 39:55; age 47 ± 17 yr; BMI 36.3 ± 5.0 kg/m²). The sample size of the first set of control individuals (control 1) was 658 and consisted of the Japanese general population as described in JSNP database [Institute of Medical Science (IMS)-Japan Science and Technology (JST) Agency Japanese SNP database] (8). The sample size of the second set of Japanese obese subjects (BMI ≥ 30 kg/m²) was 796 (case 2; male to female ratio 379:417; age 49 ± 14 yr; BMI 34.3 ± 5.5 kg/m²), whereas that of the second set of Japanese normal-weight controls (BMI ≤ 25 kg/m²) was 711 (control 2; male to female ratio 267:444; age 52 ± 16 yr; BMI 21.6 ± 2.2 kg/m²). Secondary obesity and obesity-related Mendelian disorders were excluded in this study. Patients with obesity caused by medications were also excluded. Control 2 subjects were Japanese normal-weight volunteers collected from subjects who had undergone a medical examination for common disease screening. We further collected 403 Japanese subjects with various BMIs [male to female ratio 144:259 females; age 48 ± 12 yr; BMI 29.7 ± 7.0 kg/m²; visceral fat area (VFA) 126 ± 81 cm²; sc fat area (SFA) 248 ± 117 cm²] who agreed to undergo computed tomography examinations to measure the VFA and SFA. All subjects except control 1 were newly recruited for this study. Written informed consent was obtained from each subject, and the protocol was approved by the ethics committee of each institution and that of RIKEN.

DNA preparation and SNP genotyping

Genomic DNA was prepared from each blood sample according to standard protocols. Approximately 100,000 Invader probes (Third Wave Technologies, Madison, WI) could be made for SNPs of IMS-JST (8), and the SNPs were genotyped in case 1 by Invader assays as described previously (9, 17). Genotype and allele frequencies of these SNPs were compared with control 1. The SNPs selected by association study using case 1 and control 1 were submitted for further examination using independent case 2 and control 2 groups.

SNP discovery in around SNP-1

To identify additional variations in the genomic region around SNP-1, we generated a reference sequence of approximately 300 kb by assembling the relevant regions from the sequences with GenBank accession no. AC066613, AC020892, AC026770, and AC090971. We amplified appropriate fragments of genomic DNA by PCR and sequenced the products to identify SNPs within 300 kb genomic region using previously described methods (10, 18).

Cell culture

SH-SY5Y and BE(2)-C neuroblastoma cells and HIT-T15 cells were purchased from the American Type Culture Collection (Manassas, VA). Cells were cultured in advanced DMEM (Invitrogen, Carlsbad, CA) with 2 mm glutamine, 5% fetal bovine serum, 100 U/ml penicillin, and 100 μg/ml streptomycin.

Luciferase assay

We synthesized double-stranded oligonucleotides containing either a single copy or four concatenated copies of either the major or minor allele for a 19-bp region centered on SNP-1, SNP-2, SNP-5, SNP-9, SNP-11, or SNP-12 (Fig. 1B), with an NheI restriction site at the 5′ end and an XboI restriction site at the 3′ end. We constructed luciferase reporter plasmids by cloning the oligonucleotides into the pGL3-promoter vector (Promega, Madison, WI) upstream of the Simian virus 40 promoter. pGL3-promoter vectors containing oligonucleotides were transfected into SH-SY5Y neuroblastoma cells together with the phRL-TK vector (Promega), an internal control for transfection efficiency, using lipofectamine 2000 reagent (Invitrogen). After 24 h, we collected the cells and measured luciferase activity with the dual-luciferase reporter assay system (Promega).

Fig. 1.

LD mapping, polymorphisms, and P values identified around the SCG3 gene. A, LD mapping around the SCG3 gene. LD coefficients (Δ) between every pair of SNPs around SNP-1 (rs3764220, −1492A→G) were calculated. Minor allele frequencies of all SNPs used in this analysis are greater than 10%. Genomic structure is shown at the bottom. SNP rs2124879, SNP-1, SNP-29, and SNP-40 are indicated. B, Genetic variations and P values in the SCG3 gene. #, SNP-1; *, insertion/deletion polymorphisms; †, SNPs analyzed in the first screening; no symbol, SNPs identified in the extensive search of the gene’s genomic sequence. P values are represented as −logarithm of P values of genotype mode. Each SNP is labeled with its rs number, except for novel SNPs, which are indicated by JSNP ID (ssj0011008-0011013).

Gel-shift assay

We prepared nuclear extract from SH-SY5Y cells using NE-PER extraction reagents (Pierce, Rockford, IL) and then incubated the extracts with 33-bp double-stranded oligonucleotides containing SNP-1, SNP-2, SNP-5, SNP-9, SNP-11, or SNP-12 (Fig. 1B) labeled with digoxigenin-11-ddUTP using the digoxigenin gel-shift kit (Roche Diagnostics, Indianapolis, IN). For competition studies, we incubated nuclear extract with unlabeled oligonucleotides (100-fold excess before adding digoxigenin-labeled oligonucleotide). Protein-DNA complexes were separated on a 5% nondenaturing polyacrylamide gel in 0.5 × Tris-borate-EDTA buffer. The gel was transferred to nylon membrane, and the signal was detected with a chemiluminescent detection system (Roche Diagnostics) according to the manufacturer’s instructions.

Double-labeling immunohistochemistry for SCG3, orexin, MCH, NPY, and POMC

Male mice (B57BL/6, 8 wk old) were purchased from CLEA Japan (Tokyo, Japan). After being anesthetized with sodium pentobarbital (100 mg/kg), mice were perfused with 10% neutral buffered formalin. The hypothalamic region was dissected from the brain, further fixed with tissue fixative (Genostaff, Tokyo, Japan), embedded in paraffin, and sectioned. Tissue sections (4 μm) were dewaxed and incubated at 4 C overnight with polyclonal goat anti-SCG3 (1:200; Santa Cruz Biotechnology, Santa Cruz, CA) together with either rabbit polyclonal antibody to orexin B (1:500; Chemicon, Temecula, CA), MCH (1:500; Phoenix Pharmaceuticals, Belmont, CA), NPY (1:200; Chemicon), or POMC (1:5000; Phoenix Pharmaceuticals). After washing, the sections were incubated at room temperature for 2 h with Alexa Fluor 568 donkey antigoat IgG (1:2000; Molecular Probes, Eugene, OR) and Alexa Fluor 488 donkey antirabbit IgG (1:2000; Molecular Probes) secondary antibodies. Double-immunofluorescence detection was carried out using a BX51 microscope (Olympus, Tokyo, Japan).

Expression of SCG3, orexin, MCH, NPY, and POMC in BE(2)-C neuroblastoma

The coding sequence of human SCG3 was amplified by RT-PCR from hypothalamus cDNA using primers with an added N-terminal PstI restriction site located before the start codon and a C-terminal SalI restriction site located after the stop codon. The PCR product was cloned between the PstI and SalI sites of the pBI vector (CLONTECH, Palo Alto, CA). The coding region of human preproorexin, pro-MCH, POMC, and pro-NPY were also amplified but with primers that included an N-terminal MluI site located before the start codon and a C-terminal EcoRV site located after the stop codon. The PCR products were cloned between the MluI and EcoRV sites of the pBI-SCG3 plasmid. The pBI-SCG3-preproorexin, pBI-SCG3-pro-MCH, pBI-SCG3-POMC, and pBI-SCG3-pro-NPY were transfected using lipofectamine 2000 reagent (Invitrogen) into a previously established cell line of BE(2)-C cells containing the pTet-Off vector (CLONTECH). For immunocytochemical detection, cells were fixed with 4% paraformaldehyde for 15 min then treated with 0.5% Triton X-100. Cells were incubated with polyclonal goat anti-SCG3 (1:200; Santa Cruz Biotechnology) together with rabbit polyclonal antibody to orexin B (1:500; Chemicon), MCH (1:500; Phoenix Pharmaceuticals), NPY (1:500; Progen Biotechnik, Heidelberg, Germany), or POMC (1:5000; Phoenix Pharmaceuticals) in PBS containing 1% BSA overnight at 4 C. We washed and incubated the cells at room temperature for 2 h with Alexa Fluor 488 donkey antigoat IgG (1:2000; Molecular Probes) and Alexa Fluor 568 donkey antirabbit IgG (1:2000; Molecular Probes) secondary antibodies. The cells were examined using an Olympus FV300 confocal laser-scanning microscope.

Statistical analysis

For each case-control study, the frequencies of the genotypes or the alleles were compared between cases and controls in four different modes. In the first mode (allele frequency mode), allele frequencies were compared between cases and controls using a 2 × 2 contingency table, whereas in the second mode (genotype mode), frequencies of the three genotypes were compared between cases and controls using a 2 × 3 contingency table. In the third mode (minor allele homozygotes mode), the frequencies of the homozygous genotype for the minor allele were compared using a 2 × 2 contingency table, whereas in the fourth mode (major allele homozygotes mode), the frequencies of the homozygotes for the major allele were compared using a 2 × 2 contingency table. Odds ratio and its 95% confidence interval (CI) were calculated by Woolf’s method. Hardy-Weinberg equilibrium was assessed using the χ² test (19). We used the correlation coefficient Δ, calculated as reported previously (20), as the measure to evaluate the strength of linkage disequilibrium (LD). Haplotype phasing was estimated using the Expectation-Maximization algorithm (21). Haplotype blocks were estimated using Haploview 3.2 (22). Multiple linear regression analysis was performed using StatView 5.0 (SAS Institute Inc., Cary, NC) to test an independent effect of SNP-2 genotypes on SFA or VFA, considering the effects of other variables (age, BMI, and gender) that were assumed to be independent of the effect of the SNP. The significance of the association between an independent variable and the dependent variable was tested by t test. The relative luciferase activities and clinical data are expressed as mean ± sd. Differences in luciferase activities were analyzed with the unpaired t test.

Results

Case-control association study

A total of 62,663 IMS-JST SNPs covering 11,932 gene loci were successfully genotyped in 94 obese subjects (case-1). The genotype and allele frequencies were compared with 658 random Japanese subjects. According to the National Nutrition Survey, the proportion of the subjects with BMI of 30 kg/m² or greater was estimated to be 0.023 in males and 0.034 in females aged 20 yr and older (23), and the mean BMIs are approximately 23 kg/m² for ages 15–84 yr in Japan (24). Therefore, control 1 that was randomly selected from the Japanese subjects was not an inappropriate control for the initial analysis. A total of 2261 SNPs that possessed P values less than or equal to 0.02 by a test of independence using either genotype mode or allele frequency mode were further analyzed using another set of obese (case 2) and control subjects (control 2). Among the 2261 SNPs, we successfully completed genotyping of 2115 SNPs and identified a strong association with the obesity phenotype for SNP-1 (rs3764220, −1492A→G), which lies in the 5′ flanking region of the SCG3 gene (Table 1). There were no gender- or age-related differences with respect to SNP-1 alleles. Because the P value of SNP-1 was the smallest (P = 0.0000019, genotype mode) among the 2115 SNPs, we considered this gene as a good candidate for further investigation.

TABLE 1.

Association of SNP-1 (rs3764220, 5′ flanking −1492) in the SCG3 gene with obesity in the first (case 1 vs. control 1) and the second (case 2 vs. control 2) set of experiments

Population	No. of subjects (%)			No. of chromosomes (%)		HWE testa
	AA	AG	GG	A	G	χb	P value
Case 1 (n = 94)	81 (86.2)	13 (13.8)	0 (0.0)	175 (93.1)	13 (6.9)	0.5	0.47
Control 1 (n = 634)	486 (76.7)	134 (21.1)	14 (2.2)	1106 (87.2)	162 (12.8)	1.7	0.19
Case 2 (n = 796)	639 (80.3)	154 (19.3)	3 (0.4)	1432 (89.9)	160 (10.1)	3.9	0.05
Control 2 (n = 711)	522 (73.4)	164 (23.1)	25 (3.5)	1208 (85.0)	214 (15.0)	6.8	0.009

	Genotype modeb		Allele frequency modeb			Major allele homozygotes modeb			Minor allele homozygotes modeb
	χb	Pc	χb	P	OR (95% CI)	χb	P	OR (95% CI)	χb	Pc	OR (95% CI)
Case 1 vs. control 1	5.2	0.09	5.3	0.02	1.97 (1.10–3.54)	4.3	0.04	1.90 (1.03–3.51)	2.1	0.24	ND
Case 2 vs. control 2	24.7	0.000002	17.3	0.00003	1.59 (1.27–1.97)	9.9	0.002	1.47 (1.16–1.88)	20.3	0.000004	9.63 (2.90–32.05)

The position of SNP in the 5′ flanking region is counted from the transcription initiation site. OR, Odds ratio; ND, not determined.

^a

Hardy-Weinberg equilibrium test.

^b

Association test was performed in four different modes as described in Subjects and Methods, and the results in the three modes are shown.

^c

Fisher’s exact test.

LD blocks of the SCG3 locus

We identified 112 genetic variations (107 SNPs and five insertions/deletions) by sequencing in the approximately 300-kb genomic region around SNP-1, of which 38 SNPs and two insertions/deletions resided in the SCG3 gene. Among the 107 SNPs, Invader probes could be synthesized for 81 SNPs, and 79 SNPs were successfully genotyped. Seven SNPs had minor allele frequency less than 5% and were excluded from LD analysis, whereas 10 SNPs had minor allele frequency less than 10% and were excluded from case-control association study. LD analysis revealed that SNP-1 in the SCG3 gene was located in a 40-kb LD block (block 2, Fig. 1A), which did not contain any gene apart from SCG3. Because no association with obesity was observed for SNPs located outside this LD block (Fig. 1B), we judged SCG3 to be a candidate susceptibility gene for obesity. P values of 39 SNPs located in block 2 and the adjacent blocks 1 and 3 are indicated in Fig. 1B. Among 40 genetic polymorphisms within the SCG3 gene that we found and genotyped, 11 SNPs [SNP-2 (rs16964465), 5′ flanking −1203; SNP-5 (rs3809498), 5′ flanking −65; SNP-9 (rs16964476), intron 1 + 190; SNP-11 (ssj0011012), intron 1 + 478; SNP-12 (rs3214014), intron 1 + 605; SNP-16 (rs2305709), exon 4 + 351(I117I); SNP-17 (rs3816544), intron 4 + 127; SNP-20 (rs2305715), intron 5 + 677; SNP-26 (rs2305719), intron 6 + 2677; SNP-27 (ssj0011013), intron 8 + 25; SNP-29 (rs3765067), intron 9 + 52] were in almost complete linkage disequilibrium (Δ = 0.99–1.0) with SNP-1 and also revealed significant associations with obesity (Fig. 1B). For example, the frequency of the subjects with the C/C genotype at SNP-2 was significantly lower in the obesity group than the control group (odds ratio 9.23; 95% CI 2.77–30.80, χ² 19.2, P = 0.0000067) (Supplemental Table 1, published as supplemental data on The Endocrine Society’s Journals Online Web site at http://jcem.endojournals.org). The remaining SNPs showed no significant association with obesity. SNPs in the 5′-flanking region are counted from the transcription initiation site. For SNPs in introns, nucleotide positions are counted from the first intronic nucleotide at the exon-intron junction and for SNPs in exon regions, from the first exonic nucleotide (transcription initiation site) according to sequence accession no. AC020892.7 and NM_013243.2.

Regulatory effect of SNPs on SCG3 expression

Three SNPs (SNP-1, SNP-2, SNP-5) were located in the 5′ flanking region and three SNPs (SNP-9, SNP-11, SNP-12) were located in intron 1 of SCG3 gene, regions that could putatively affect transcriptional activity. To examine whether these six SNPs would affect the transcriptional activity, we performed a luciferase assay using the neuroblastoma cell-line SH-SY5Y, which has previously been shown to express SCG3 (25). Between the major and minor alleles at each locus, only the clones containing SNP-2 or SNP-9 showed significant differences in transcriptional activity (Fig. 2A), and these differences were enhanced using the plasmids containing four concatenated copies of these DNA fragments, suggesting that SNP-2 and SNP-9 were able to affect the transcriptional activity of the SCG3 gene. SCG3 was also reported to be expressed in pancreatic β-cells (13); thus, we performed the same experiments using the hamster pancreatic β-cell line, HIT-T15 (Fig. 2A). We observed similar results, although the differences in the transcriptional activity between SNPs using HIT-T15 cells were smaller than those seen in the SH-SY5Y cells, probably due to the species difference.

Fig. 2.

Transcriptional activities affected by SNPs. A, Comparison of allelic variants of SCG3 analyzed by relative luciferase activity in SH-SY5Y cells and HIT-T15 cells. The values are mean ± sd. pGL3-promoter, the empty vector. The gray boxes indicate the oligonucleotide unit around the SNPs, and white and black small boxes represent major and minor allele of each SNP, respectively. SV40, Simian virus 40. B, Binding of unknown nuclear factor(s) to the SCG3 gene. Gel-shift assay was performed with digoxigenin-labeled 33-bp oligonucleotides corresponding to two SCG3 polymorphic sites (SNP-2 and SNP-9) in SH-SY5Y cells. An arrow indicates the band that shows binding of nuclear proteins to the oligonucleotides containing minor alleles of SNP-2 (C allele, left panel) and SNP-9 (G allele, right panel).

To further investigate whether the regions containing each of these six SNPs can act as target binding sites of nuclear protein(s), we performed a gel-shift assay using SH-SY5Y cell extract and oligonucleotides corresponding to genomic sequences that included major or minor alleles of each of the six SNPs (SNP-1, SNP-2, SNP-5, SNP-9, SNP-11, and SNP-12). The band corresponding to the minor allele (C allele) of SNP-2 was more intense than that corresponding to the major allele (A allele) (Fig. 2B), indicating that some nuclear factor(s) has higher binding affinity to the minor allele. Although we observed shifted bands for the oligonucleotides corresponding to SNP-5 and SNP-12, no significant difference in the intensity of the bands between the major and minor alleles was observed (data not shown). No shifted band was observed in the case of SNP-1 and SNP-11. In the case of SNP-9, the band corresponding to the minor allele was more intense than that corresponding to the major allele, as observed in SNP-2 (Fig. 2B). The combination of the results of the luciferase assay and the gel-shift assay suggested that the genetic variations corresponding to SNP-2 and -9 were the most likely candidates to affect the transcriptional activity of SCG3 and perhaps susceptibility to the development of obesity.

Expression of SCG3 in the hypothalamus

SCG3 was reported to be expressed in the hypothalamus, but its physiological roles have not yet been clarified (12). To further elucidate this role, we performed in situ hybridization and immuohistochemical analysis for SCG3 in the murine hypothalamus and observed that SCG3 was expressed in the LHA, PVN, ventromedial hypothalamus, and ARC (data not shown). SCG3 immunoreactivity was also observed in various other regions of the mouse brain as reported previously (12); however, the most intense immunoreactivities were observed in the ARC and LHA as well as the PVN and ventromedial hypothalamus (Fig. 3). The ARC neurons that express and secrete NPY and POMC are regulated by leptin and transfer their neuronal signal to orexin-expressing neurons in the LHA (26). To investigate the relationship between SCG3 and these neuronal peptides, we performed double-labeling immunohistochemical analysis and found that SCG3 was coexpressed with POMC and NPY in ARC cells (Fig. 3). We also examined the relationship between SCG3 and two major neuropeptides in the LHA that inhibit food intake, orexin and MCH, and detected that many orexin-expressing neurons and MCH-expressing neurons coexpressed SCG3 (Fig. 3).

Fig. 3.

Colocalization of SCG3 with POMC, NPY, MCH, or orexin in mouse hypothalamus. Immunostained tissue sections were double labeled for SCG3 with POMC, NPY, MCH, or orexin B. Scale bar, 50 μm.

Granins, such as CHGA, CHGB, and secretogranin II, form granule-like structure when they are expressed in cultured cells (11, 27). To examine whether SCG3 would also form granule-like structures and interact with each of these neuropeptides, we transfected pBI-SCG3-preproorexin, pBI-SCG3-pro-MCH, pBI-SCG3-POMC, and pBI-SCG3-pro-NPY into established BE(2)-C cell lines that were stably transfected with the pTet-Off vector system. The results indicated that SCG3 formed granule-like structures, like other granins, and colocalized with orexin, MCH, NPY, and POMC (Fig. 4). Immunoelectron microscopic analysis revealed that the granules were detected in BE(2)-C cells transfected with SCG3 but not in those transfected with vector alone (data not shown). The granules stained with anti-SCG3 antibody (data not shown), suggesting that SCG3 forms secretory granules in neuroblastoma cells. These in vivo and in vitro data suggest that SCG3 may play some role in the secretion of neuropeptides that are related to appetite.

Fig. 4.

Granular accumulation of SGG3 protein overexpressed in BE(2)-C cells. BE(2)-C cells expressed SCG3 (green stain) and POMC (red stain), NPY (red stain), orexin (red stain), or MCH (red stain). Scale bars, 10 μm.

Analysis of various quantitative phenotypes with SNP-2 and SNP-9

Because SCG3 is expressed in pancreatic β-cells and involved in insulin secretion (13), SCG3 may play a role in metabolic disorders as well as in obesity. Therefore, to investigate whether the genotypes of SNP-2 and SNP-9 are related to the phenotypes of the metabolic disorders, we compared BMI, blood insulin, glucose, cholesterol, triglycerides, and high-density lipoprotein-cholesterol, and blood pressure among the different genotypes in cases and controls. We detected no relationship between these quantitative phenotypes and the genotypes at SNP-2 and SNP-9 in either the case or control groups.

The most important phenotype of the metabolic syndrome is visceral fat accumulation. Thus, we performed multiple linear regression analysis to further define the role of this gene in the amount of visceral and/or sc fat. The SNP-2 genotype was transformed to a multidichotomous variable, i.e. homozygosity with the A alleles vs. the other genotypes, heterozygosity vs. the other genotypes, or homozygosity with the C alleles vs. the other genotypes. Stepwise multiple regression analysis (both forward selection and backward elimination) revealed that gender, age, and BMI were significantly associated with VFA. However, no genotypes were significantly associated with VFA. In contrast, gender, BMI, and genotype (homozygosity with the A allele or heterozygosity with A and C alleles) were significantly associated with SFA. Neither age nor homozygosity with the C allele was significantly associated with SFA. Table 2 shows the data of multiple regression analysis using gender, age, BMI, and genotype as independent variables. Among the independent variables, homozygosity with the A allele, female gender, and increase in BMI were significantly associated with increases in SFA. In concordance, each of the three parameters, heterozygosity with A and C alleles, male gender, and decrease in BMI, was significantly associated with a decrease in SFA. Because the number of homozygotes with the C allele was very small (n = 9), we were unable to validate its association with either SFA or VFA. SNP-2 and SNP-9 were in complete linkage disequilibrium (Δ = 1.0); thus, the same results were observed. These data suggested that the genotypes of SNP-2 and SNP-9 have an effect on the amount of SFA independent of the effects of the other independent variables.

TABLE 2.

Multiple linear regression analysis for VFA or SFA using SNP-2 (5′ flanking −1203) and other features as independent variables

Independent variables	AA vs. the other genotype			AC vs. the other genotype			CC vs. the other genotype
	Regression coefficient	se	P	Regression coefficient	se	P	Regression coefficient	se	P
VFA (dependent variable)
Gender (men/women, 1/0)	61.271	6.583	<0.0001	61.398	6.572	<0.0001	61.352	6.570	<0.0001
Age (yr)	1.064	0.264	<0.0001	1.063	0.264	<0.0001	1.06	0.264	<0.0001
BMI (kg/m2)	5.723	0.454	<0.0001	5.728	0.454	<0.0001	5.705	0.456	<0.0001
Genotype (1/0)	3.193	7.332	0.66	−2.293	7.577	0.76	−8.755	21.229	0.68
Rb	41%			41%			41%
SFA (dependent variable)
Gender (men/women, 1/0)	−69.550	7.167	<0.0001	−68.998	7.157	<0.0001	−67.623	7.244	<0.0001
Age (yr)	−0.329	0.289	0.26	−0.326	0.289	0.26	−0.357	0.293	0.22
BMI (kg/m2)	13.044	0.495	<0.0001	13.095	0.496	<0.0001	12.998	0.503	<0.0001
Genotype (1/0)	25.499	7.952	0.0015	−25.761	8.221	0.0019	−11.438	23.275	0.62
Rb	67%			67%			66%

Discussion

Epidemiological studies have provided evidence indicating the involvement of genetic factors in the development of obesity (5, 6). Through case-control association studies using gene-based SNPs, our center has successfully discovered candidate genes that confer susceptibility to various common diseases (myocardial infarction, diabetic nephropathy, type 2 diabetes mellitus) (9, 28–30). Using this approach, we identified novel functional SNPs associated with obesity, which are located in the SCG3 gene. Our approach should prove effective and useful in searching for genes related to common diseases; however, the set of SNPs that we used covered only 11,932 gene loci. Recently a haplotype map of the human genome has been constructed (31). Despite the relatively high SNP density in genomic region, our SNP set only covered approximately 30% of the human genome by counting HapMap phase II SNPs that are: in LD (r² > 0.5) at least with one SNP in our set, with minor allele frequencies greater than 0.05, and at distances less than 500 kb from at least one SNP in our set. Because of this low genomic coverage for studies up to now, further investigations will be necessary as high-throughput genotyping products achieve higher SNP densities.

Intracellular granins are costored and cosecreted with peptide hormones (11). Our results suggest that SCG3 forms secretory granules together with orexin, MCH, NPY, and POMC in the hypothalamus. We demonstrated that SNP-2 and SNP-9 might have an effect on the transcriptional activity of the SCG3 gene. Transcriptional activity of the major allele, the frequencies of which were higher in obese subjects than normal controls, was shown to be lower, which indicates that decreased SCG3 expression levels may increase the risk of obesity. These results seem to be complicated. Many granins are known to work as inhibitors of endocrine secretion (11); for example, extracellular CHGA undergoes proteolytic processing into several bioactive peptides such as pancreastatin and catestatin (11). Pancreastatin inhibits insulin secretion from pancreatic β-cells (32), and catestatin inhibits the release of catecholamines from sympathoadrenal chromaffin cells (33). CHGA also inhibits POMC-derived peptide secretion (34). CHGB-derived peptides inhibit the secretion of PTH and insulin (35, 36). A secretogranin II-derived peptide, secretoneurin, inhibits serotonin and melatonin release from pinealocytes (37). SCG3 also undergoes proteolytic processing and is secreted from cells (38). It needs to be investigated whether the peptides derived from proteolytic processing of SCG3 are bioactive and whether they may also inhibit the secretion of orexin, MCH and NPY, like other granins. Hence, we consider that increased expression of SCG3 in the subjects with the minor allele of SNP-2 and SNP-9 may result in a decrease in the secretion of orexin, MCH, and NPY and thereby inhibit food intake and accumulation of sc fat.

Food intake control is complicated (26) because in addition to many neuropeptides in the central nervous system, peptides secreted from other tissues, such as adipose tissue and gastrointestinal organs, participate in the control of food intake. The neural circuits in the hypothalamic region are also complicated, and the whole network is not well understood. There have been no reports indicating the involvement of SCG3 in appetite regulation, but in light of our data, it is interesting to speculate that SCG3 may be a potential factor in the regulation of food intake. Nevertheless, because fat accumulation is also affected by other variables like physical activity as well as food intake, it is also necessary to investigate whether SCG3 interacts with other variables.

Functional SNP-2 and SNP-9, which we have shown to be associated with obesity, are located on the chromosome 15q21 locus in which a positive linkage to SFA was indicated using the Québec Family Study (39). In concordance with this previous result, our study showed an association of SNP-2 and SNP-9 with SFA.

In summary, we identified the genetic variations in SCG3 that may influence the risk of obesity (particularly sc fat obesity) by a large-scale case-control association study. We found that SNP-2 and SNP-9 posses moderate effect sizes (Supplemental Table 1) and affect the expression levels of SCG3 and that SCG3 forms secretory granules with hypothalamic neuropeptides. Our present data suggest that SCG3 is a good target for the development of new medicine to aid in the prevention and treatment of obesity.

Acknowledgments

The authors express their appreciation to Dr. Chisa Nakagawa, Dr. Hideki Asakawa, Dr. Hiroaki Masuzaki, Dr. Kazuwa Nakao, Ms. Kaoru Nakene, Ms. Taeko Okubayashi, Ms. Yuko Ohta, Ms. Emiko Takada, Mr. Fumitaka Sakurai, and Mr. Masahiro Uchibatake and all the members of the SNP Research Center for their contribution to our study. The authors also thank Dr. Todd Johnson for critical reading of the manuscript.

This work was supported by a grant from the Japanese Millennium Project, Takeda Science Foundation (to K.Ho.), and Chiyoda Mutual Life Foundation (to K.Ho.).

Disclosure statement: T.Y., A.I., S.S., A.S., A.Tak., T.N., T.T., Y.Nakat., K.K., R.K., N.I., I.M., J.W., T.F., S.M., K.To., K.Ha., T.Sh., K.Tan., K.Y., T.H., S.O., H.Y., T.Sa., Y.M., N.K., Y.Nakam. have nothing to declare. S.K. consults for DHC Corporation Laboratories. K.Ho. and A.Tan. are inventors on (Japan and PCT) (PCT/JP2005/023674).

Abbreviations:

 
• ARC,

  Arcuate nucleus;

 
• BMI,

  body mass index;

 
• CHG,

  chromogranin;

 
• CI,

  confidence interval;

 
• IMS,

  Institute of Medical Science;

 
• JST,

  Japan Science and Technology;

 
• LD,

  linkage disequilibrium;

 
• LHA,

  lateral hypothalamic area;

 
• MCH,

  melanin-concentrating hormone;

 
• NPY,

  neu-ropeptide Y;

 
• POMC,

  proopiomelanocortin;

 
• PVN,

  paraventricular nucleus;

 
• SFA,

  sc fat area;

 
• SNP,

  single-nucleotide polymorphism;

 
• VFA,

  visceral fat area.