Introduction

Family with sequence similarity 3 member C (FAM3C), previously known as interleukin-like epithelial to mesenchymal transition inducer (ILEI), is a cytokine-like protein essential for epithelial to mesenchymal transition, tumor formation and metastasis in normal epithelial cells.1 Researches have demonstrated that FAM3C gene was involved in embryonic development, and the progression of pancreatic cancer.2,3 Furthermore, FAM3C was overexpressed and/or altered in intracellular localization in various human tumors, rendering it a promising molecular marker for the diagnosis and prognosis of some cancers, such as colon and breast cancer.1,2

Recently, Chen et al.4 performed functional studies and revealed a novel function of FAM3C in glucose and lipid metabolism in hepatocytes. Chen et al.4 identified that FAM3C suppressed the mTOR-SREBP1-FAS lipogenic pathway in hepatocytes in vitro and in vivo, which contributed to the beneficial effects of FAM3C on steatosis.4,5 They considered FAM3C as a novel therapeutical target for metabolic diseases, such as nonalcoholic fatty liver disease (NAFLD).4,5 However, whether FAM3C genetic variants are associated lipid traits among humans remain largely unknown. Moreover, previous genome-wide association studies and subsequent replication studies have found that genetic variants near FAM3C gene were associated with bone mineral density (BMD) and fracture in children,6,7,8 which is correlated with childhood overweight and obesity.9,10,11,12 Therefore, we hypothesized that FAM3C genetic variants would also be associated with overweight/obesity and lipid traits.

In the present study, we aimed to analyze the associations of FAM3C genetic variants with overweight/obesity and lipid traits among Chinese children; and assess whether the genetic associations differed in boys and girls.

Methods

Participants

The current study subjects were from four independent studies conducted in Beijing, China, with a total sample size of 3305 children (1029 obese children, 1014 overweight children and 1262 normal-weight children). The study on adolescent lipids, insulin resistance and candidate genes (ALIR) and the Comprehensive Prevention project for Overweight and Obese Adolescents (CPOOA) were both case−control studies with similar subject recruitments and data collection methods, involving 937 (386 obese children, 400 overweight children and 151 normal-weight children) and 1093 children (319 obese children, 318 overweight children and 456 normal-weight children), respectively.13,14 The other two studies were school-based interventions on childhood and adolescent obesity. The School-Based Comprehensive Intervention on Childhood Obesity (SCICO) was performed in eight elementary schools of the Haidian District of Beijing, including 153 obese children, 151 overweight children and 1588 normal-weight children.15 The baseline survey data of all the overweight and obese children, as well as 152 normal-weight children randomly selected from the normal-weight group, were analyzed in the present study. The School-based Physical Activity Intervention on Childhood Obesity (SPAICO) was conducted in two primary schools and two middle schools in Changping District of Beijing.16 The cross-sectional baseline survey data of all children with blood samples in the SPAICO were analyzed in the current study, involving 171 obese children, 145 overweight children and 503 normal-weight children. Detailed descriptions of the four studies have been previously published.13,14,15,16 Subjects with any vital organs diseases and obesity-related comorbidities, including heart disease, liver diseases, pulmonary diseases, kidney disease and diabetes, were excluded, as well as those who were underweight, according to the Chinese national screening criteria for malnutrition of school-age children and adolescents.17

Written informed consents were obtained from all children and their parents. The ALIR, CPOOA and SPAICO were approved by the Ethic Committee of Peking University Health Science Center, and the SCICO was approved by the Ethic Committee of Chinese Center for Disease Control and Prevention. All studies abided by the Declaration of Helsinki principles.

Anthropometric measurements

Anthropometric measurements including height and weight in the four studies were performed according to the same standard protocols.13,14,15,16 We calculated BMI (kg/m2) using weight (in kilograms) divided by the square of height (in meters). Sex-and age-specific BMI standard deviation score (BMI-SDS) was also calculated, according to the growth reference data of the World Health Organization for children aged 5–19 years.18 The body fat percentage was measured using bioelectrical impedance analyzer except in ALIR study (CPOOA: Genuis-220, Jawon, Korea; SCICO and SPAICO: DF50, ImpediMed, Australia).

In the current study, obesity was defined as over 95th percentile of age- and sex-specific BMI and overweight was defined as between 85th and 95th percentile of age- and sex-specific BMI for children aged 7–18 years. The cut-offs were determined in a representative Chinese population.19 For children aged 6 years, the Chinese national screening criteria for overweight and obesity among school-age children and adolescents were applied.20

Biochemical assessment

Laboratory assays were performed on blood samples collected from all children after an overnight fast. Biochemical autoanalyzer (ALIR and CPOOA: Hitachi7060, Tokyo, Japan; SCICO and SPAICO: Olympus AU400, Tokyo, Japan) was used to determine total cholesterol, triglyceride, low-density lipoprotein cholesterol (LDL-C), and high-density lipoprotein cholesterol (HDL-C). Dyslipidemia was defined according to the latest report of National Cholesterol Education Program (NCEP) Expert panel on integrated guidelines for cardiovascular health and risk reduction in children and adolescents,21 as having one or more of the following: (1) an elevated TC level (TC ≥ 5.18 mmol/L); (2) an elevated TG level (children aged 7–9 years old: TG ≥ 1.13 mmol/L; children aged 10–19 years old: TG ≥ 1.47 mmol/L); (3) an elevated LDL-C level (LDL-C ≥ 3.37 mmol/L); (4) or a low HDL-C level (HDL-C ≤ 1.04 mmol/L).

SNPs selection and genotyping

Given the sample size, in order to reach over 80% power to detect an assumed effect size of 1.3, the allele frequency should be equal to or greater than 0.10. Therefore, only 21 intronic SNPs with minor allele frequency (MAF) ≥ 0.10 in the Han Chinese population from 1000 genomes (http://phase3browser.1000genomes.org/index.html) were left, representing by three tag SNPs. We only needed to select three SNPs, because the other SNPs were in strong linkage disequilibrium (LD) (r2 > 0.80) with these three tag SNPs. However, for one tag SNP, multiplex SNP assays design failed and no proxy SNP was available. Therefore, we finally included two intronic SNPs: rs7776725 (chromosome and position (based on NCBI build 38 (NCBI, Bethesda, MD)): 7:121393067) and rs7793554 (7:121393137) within the FAM3C gene (r2 = 0.61), representing >95% SNPs with MAF ≥ 0.10. Both rs7776725 and rs7793554 were previously reported to be associated with BMD and fracture.6,7,8 We used Haploview to estimate LD between the SNPs.22

Genomic DNA of all children was extracted from the blood leukocytes using the phenol-chloroform extraction method (ALIR, CPOOA and SCICO) and salt extraction method (SPAICO). Genotyping of rs7776725 and rs7793554 was conducted using the MassARRAY System (Agena Bioscience Inc., San Diego, CA, USA). The detail genotyping process had been described elsewhere.23,24 The call rates for rs7776725 and rs7793554 were 99.8% and 99.0%, respectively. For quality control, we performed genotyping on 5% randomly selected duplicated samples with 100% concordance.

Statistical analyses

Data were expressed as the mean and standard deviation (SD) for continuous variables and n (%) for categorical variables. Total cholesterol, triglyceride, LDL-C and HDL-C were natural log-transformed before analysis. The differences in general characteristics among normal-weight, overweight and obese children were examined with one-way analysis of variance (ANOVA) (continuous variables) or χ2 test (categorical variables). The Hardy−Weinberg equilibrium (HWE) of the genotype data in the normal-weight children were examined by a χ2 test.

The primary outcomes were overweight/obesity and dyslipidemia, and the secondly outcomes were anthropometric and lipid quantitative traits. We tested additive, dominant and recessive genetic models for the primary outcomes, and found that the additive model could best explain the effect of rs7776725 and rs7793554 based on the Akaike information criterion and the Bayesian information criterion. Therefore, the associations of FAM3C genetic variants with overweight/obesity and dyslipidemia were analyzed using logistic regressions, and their associations with anthropometric and lipid quantitative traits were analyzed using general linear regressions, assuming additive model adjusting for study groups, sex, age, age squared and BMI (except for overweight/obesity, BMI, and BMI-SDS). In addition, we performed sex-specific analyses and tested the gene-by-sex interaction effects. Since the participants were recruited from four independent studies, we also performed sensitivity analyses using meta-analysis with the inverse variance method.15,25

A two-sided P value of 0.05 was considered nominally significant. P < 0.003 (i.e. 0.05/[2SNPs × (2 categorical variables + 7 continuous variables)] = 0.05/18 = 0.003) was considered statistically significant after multiple testing using Bonferroni correction. We used Quanto 1.2.4 (University of Southern California, Los Angeles, CA, USA; http://biostats.usc.edu/cgi-bin/DownloadQuanto.pl) to calculate power and estimate the detectable effect sizes. All statistical analyses were performed with SPSS 24.0 (IBM Corp., Armonk, NY, USA). Meta-analyses were conducted with Revman 5.3.

Results

Table 1 presents the characteristics of the 3305 children in the normal-weight and overweight/obese groups. There were significant differences in sex and age between the two groups (P < 0.001). Compared with normal-weight children, overweight/obese children had higher BMI, body fat percentage, total cholesterol, triglyceride, LDL-C, and lower HDL-C (all P < 0.001).

Table 1 General characteristics of the 3305 Han Chinese children.
Full size table

The two SNPs (rs7776725 and rs7793554, located in the intron of the FAM3C gene) were in moderate LD (r2 = 0.61) in our study, and the minor allele frequencies were 0.112 and 0.168, respectively (Supplementary Table 1), similar to those reported in Han Chinese from 1000 genomes (http://phase3browser.1000genomes.org/index.html, 0.102 and 0.170 respectively). Among normal-weight children, the genotype frequencies of the two SNPs were both in HWE (P > 0.05, Supplementary Table 1).

Logistic regression analyses were performed to examine the associations of rs7776725 and rs7793554 with overweight/obesity and dyslipidemia. Both rs7776725 and rs7793554 were nominally associated with the risk of dyslipidemia (OR (95% CI) = 1.402 (1.116, 1.761), P = 0.004 and OR (95% CI) = 1.291 (1.069, 1.560), P = 0.008, respectively, Table 2), after adjustment for study groups, sex, age, age squared and BMI. However, these associations were no longer statistically significant after multiple correction.

Table 2 Association of FAM3C genetic variants with overweight/obesity and dyslipidemia in Chinese children.
Full size table

We further analyzed the associations of the two FAM3C genetic variants with various anthropometric and lipid quantitative traits. As shown in Table 3, rs7776725 was nominally associated with triglyceride (β (SE) = 0.036 (0.015), P = 0.014) and HDL-C (β (SE) = −0.019 (0.008), P = 0.015), while rs7793554 was significantly associated with triglyceride (β (SE) = 0.040 (0.013), P = 0.001) and HDL-C (β (SE) = −0.019 (0.006), P = 0.003) after multiple correction. No significant associations were observed between the two FAM3C genetic variants and other anthropometric and lipid traits (all P > 0.05).

Table 3 Associations of FAM3C genetic variants with anthropometric and lipid traits in Chinese children.
Full size table

Sex-specific analysis

Subsequently, we performed sex-specific analysis. As presented in Fig. 1, rs7776725 was significantly associated with higher risk of dyslipidemia (OR (95% CI) = 2.119 (1.456, 3.083), P = 8.78 × 10–5) only in girls, and the interaction between rs7776725 and sex was nominally significant on the risk of dyslipidemia (Pfor interaction = 0.004). Similarly, rs7793554 was nominally associated with dyslipidemia only in girls (OR (95% CI) = 1.495 (1.108, 2.019), P = 0.009). However, the interaction between rs7793554 and sex was not significant (Pfor interaction = 0.183).

Fig. 1: Interaction between FAM3C genetic variants and sex on overweight/obesity and dyslipidemia under additive model.
figure 1

Logistic regressions were performed to explore the associations of rs7776725 (a) and rs7793554 (b) within the FAM3C gene with overweight/obesity and dyslipidemia in boys and girls, separately. Gene-by-sex interactions were also tested, assuming additive model and adjusting for study groups, age, age square and BMI (except for overweight/obesity). CI confidence interval, OR odds ratios.

Full size image

Besides, we also found nominally significant interactions between rs7776725 and sex on total cholesterol and LDL-C (Pfor interaction = 0.012 and 0.008, respectively, Fig. 2). Girls with T-allele of rs7776725 were associated with higher total cholesterol (β = 0.030, P = 0.002) and LDL-C (β = 0.050, P = 0.001). Meanwhile, the interactions between rs7793554 and sex on total cholesterol and LDL-C were close to nominally significant level (Pfor interaction = 0.053 and 0.084, respectively, Supplementary Fig. 1). No significant gene-by-sex interactions were observed between the two FAM3C genetic variants and sex on other blood lipids, and anthropometric traits (all P > 0.05).

Fig. 2: Interaction between rs7776725 within the FAM3C gene and sex on blood lipids under additive model.
figure 2

Linear regressions were performed to explore the associations of rs7776725 with total cholesterol (a), triglyceride (b), LDL-C (c), and HDL-C (d) in boys and girls, separately. Gene-by-sex interactions were also tested, assuming additive model and adjusting for study groups, age, age square and BMI. HDL-C high-density lipoprotein cholesterol, LDL-C low-density lipoprotein cholesterol, TC total cholesterol, TG, triglyceride.

Full size image

Sensitivity analyses

We also performed sensitivity analyses to pool the results of the four studies using meta-analysis. The results of sensitivity analyses were consistent with the primary results (Supplementary Tables 2, 3).

Discussion

In this association study among 3305 Chinese children, we examined the associations of FAM3C genetic variants with anthropometric and lipid traits. We found significant associations of rs7776725 and rs7793554 with dyslipidemia and lipid traits. In addition, we observed significant interactions between rs7776725 and sex on dyslipidemia, and sex-stratified analyses showed that it was significantly associated with dyslipidemia only in girls. The variant also showed significant interactions with sex on total cholesterol and LDL-C, with T-allele of rs7776725 associated with higher total cholesterol and LDL-C only in girls.

FAM3C is a member of the family with sequence similarity 3 family and expressed in almost all tissues.26 It encodes a secreted protein with a GG domain, which is a cytokine involved in epithelial to mesenchymal transition, tumor formation and metastasis in normal epithelial cells.1 In 2010, Katahira et al.27 reported its function in retinal laminar formation processes. Besides, previous studies also indicated that FAM3C gene could play an important role in bone metabolism.7,8,28,29,30,31 Both rs7776725 and rs7793554 are located in the first intron of the FAM3C gene, and the variant rs7793554 was suggested to be an intronic enhancer. The A to G change of rs7793554 will delete a binding site of the transcription factor CRE-BP (cyclic AMP response element (CRE)- binding protein (BP)).31

We found that FAM3C genetic variants (rs7776725 and rs7793554) were significantly associated with dyslipidemia and lipid profiles in children. Chen et al.4 firstly revealed that FAM3C played an important role in regulating lipid metabolism in hepatocytes in mice. Chen et al.4 performed functional studies and found that FAM3C overexpression reduced phosphorylated mechanistic target of rapamycin (mTOR) complex 1 (pmTORC1), SREBP-1c, and fatty acid synthase (FAS) protein levels in hepatic cells. Our findings provided supportive evidence for the function of FAM3C gene in lipid metabolism in humans. Moreover, we analyzed their associations with four quantitative lipid traits, including total cholesterol, triglyceride, LDL-C and HDL-C. We found that FAM3C rs7776725 T-allele and rs7793554 A-allele were associated with higher triglyceride and lower HDL-C in a total sample size of 3305 children. Although the associations of FAM3C genetic variants with triglyceride and HDL-C attenuated or abolished after stratified by sex, the effect sizes pointed into the same direction in boys and girls as in the total sample. This failed replication after stratification by sex might be a problem of reduced statistic power due to smaller sample size in each subgroup (power 34−65%). However, neither of the SNPs had been previously reported to be associated with lipids traits by large-scale genome-wide association studies.32,33,34,35 The reason might be that this is a children-specific effect, because the previous studies were performed in adult populations. The genetics of these lipids traits in children remained largely unknown. Therefore, our study could help to provide more information about the genes associated with lipid metabolism, especially in Chinese children. More validation studies with larger sample sizes are needed.

In addition, we found the genetic associations showed significant sex differences. We noted that rs7776725 was significantly associated with dyslipidemia, total cholesterol and LDL-C only in girls (power > 90%). Similar genetic associations were also observed in rs7793554, although the gene-by-sex interactions were only marginally significant. However, the underlying molecular mechanism remains unclear, though it was suggested that sexual dimorphism might be at least partly explained by the differences in levels of sex-hormones.36 Other researchers provided a novel insight into sex-specific differences of cell regulatory processes.37 More researches are required before a clear understanding of biochemical mechanisms underlying sexual dimorphism in the future.

Although some evidences have shown that genetic variants near FAM3C gene were associated with BMD and fracture in children,6,7,8 which were correlated with childhood overweight and obesity.9,10,11,12 In the present study, we did not find rs7776725 and rs7793554 were associated with the risk of overweight/obesity. And no previous studies have reported the associations between FAM3C genetic variants and childhood overweight and obesity. More studies are awaited to confirm our findings.

Strengths and limitations

To the best of our knowledge, this is the first study to identify significant associations of FAM3C genetic variants with dyslipidemia and lipid traits among Chinese children. Our study provided supportive evidence for the function of FAM3C gene in regulating lipid metabolism.

There were also several limitations in our study. Firstly, the heterogeneity among the four independent studies should be considered. However, we have adjusted for the study groups in the regression models. And our previous studies also showed that there was little heterogeneity among the four independent studies.15,25 Secondly, there might be population stratification. However, we thought the population stratification effect was likely to be small because of the following two reasons. On the one hand, the heterogeneity among the four independent studies was generally small (I2 < 30%, P for heterogeneity < 0.05). And it is reasonable to assume that all participants were selected from a random mating population. Therefore, in theory, we thought participants in our study should have high homogeneity in genetic background. On the other hand, almost all children in our study were Han Chinese (about 94%); thus the systematic difference in allele frequencies between subpopulations would be small. Thirdly, as no previous studies have reported the association between the FAM3C genetic variants and overweight/obesity or dyslipidemia, we could only estimate the statistic power. Based on the sample size and effect estimates in the current study, the power to detect associations of FAM3C genetic variants with dyslipidemia and gene-by-sex interaction were >80%. But the power for overweight/obesity was <60%, which need further validation studies with larger sample size.

Conclusions

In conclusion, we identified for the first time that FAM3C genetic variants (rs7776725 and rs7793554) were associated with dyslipidemia and lipid traits among Chinese children. In addition, we found significant interactions between rs7776725 and sex on dyslipidemia, total cholesterol and LDL-C traits. Our findings provided evidence supporting the role of FAM3C gene in regulating lipid metabolism in humans. More studies are awaited to confirm our findings in the future.