Abstract
Background: Obesity is an established risk factor for renal cell carcinoma (RCC). Although genome-wide association studies (GWAS) of RCC have identified several susceptibility loci, additional variants might be missed due to the highly conservative selection.
Methods: We conducted a multiphase study utilizing three independent genome-wide scans at MD Anderson Cancer Center (MDA RCC GWAS and MDA RCC OncoArray) and National Cancer Institute (NCI RCC GWAS), which consisted of a total of 3,530 cases and 5,714 controls, to investigate genetic variations in obesity-related genes and RCC risk.
Results: In the discovery phase, 32,946 SNPs located at ±10 kb of 2,001 obesity-related genes were extracted from MDA RCC GWAS and analyzed using multivariable logistic regression. Proxies (R2 > 0.8) were searched or imputation was performed if SNPs were not directly genotyped in the validation sets. Twenty-one SNPs with P < 0.05 in both MDA RCC GWAS and NCI RCC GWAS were subsequently evaluated in MDA RCC OncoArray. In the overall meta-analysis, significant (P < 0.05) associations with RCC risk were observed for SNP mapping to IL1RAPL2 [rs10521506-G: ORmeta = 0.87 (0.81–0.93), Pmeta = 2.33 × 10−5], PLIN2 [rs2229536-A: ORmeta = 0.87 (0.81–0.93), Pmeta = 2.33 × 10−5], SMAD3 [rs4601989-A: ORmeta = 0.86 (0.80–0.93), Pmeta = 2.71 × 10−4], MED13L [rs10850596-A: ORmeta = 1.14 (1.07–1.23), Pmeta = 1.50 × 10−4], and TSC1 [rs3761840-G: ORmeta = 0.90 (0.85–0.97), Pmeta = 2.47 × 10−3]. We did not observe any significant cis-expression quantitative trait loci effect for these SNPs in the TCGA KIRC data.
Conclusions: Taken together, we found that genetic variation of obesity-related genes could influence RCC susceptibility.
Impact: The five identified loci may provide new insights into disease etiology that reveal importance of obesity-related genes in RCC development. Cancer Epidemiol Biomarkers Prev; 26(9); 1436–42. ©2017 AACR.
Introduction
Kidney cancer is the third most common urologic cancer, accounting for approximately 4% of newly diagnosed cancer patients in the United States each year (1). Obesity exhibited consistent and significant associations with higher risk of kidney cancer or renal cell carcinoma (RCC) and is considered to be an established risk factor for the disease (2, 3). A large-scale meta-analysis of prospective cohort studies showed that a 5 kg/m2 increase of body mass index (BMI) was associated with 24% and 34% higher risk of RCC in men and women, respectively (4). However, the underlying biological mechanisms linking obesity to RCC tumorigenesis are not fully understood. With an accumulated body of evidence supporting a genetic basis for obesity (5), it is plausible that genetic variants in the obesity-related genes may alter the susceptibility to RCC.
Six RCC susceptibility loci, i.e., 2p21 (EPAS1), 2q22.3 (ZEB2), 8q24.1 (MYC-PVT1), 11q13.3 (a CCND1 transcriptional-enhancer site), 12p11.23 (SSPN-ITPR2), and 12q24.31 (SCARB1), have been identified by four genome-wide association studies (GWAS; refs. 6–9). According to a recent twin study, a great proportion of variation in kidney cancer risk can be explained by interindividual genetic difference [heritability = 38%; 95% confidence interval (CI) = 21%–55%; ref. 10], which indicates that additional genetic susceptibility loci remained to be identified for RCC. Previous studies have suggested RCC risk associations with genetic polymorphisms in obesity-related genes such as FTO (11), ADIPOQ (12), and genes in the mTOR signaling pathways (13). However, independent validations are largely missing in previous studies. Furthermore, no study has systematically evaluated known and putative obesity-related genes. An overall investigation of genetic polymorphisms in obesity-related genes is therefore necessary for elucidating their impact on susceptibility to RCC.
In the present study, we hypothesized that genetic variants in obesity-related genes may harbor potential associations with RCC risk. We utilized three independent RCC datasets to test and validate our hypothesis.
Materials and Methods
Study population and sample collection
Phase I MDA RCC GWAS (MDA-GWAS): The details of the study population for the MDA-GWAS were reported previously (8). Briefly, newly diagnosed RCC cases and healthy control subjects were recruited from an ongoing RCC case–control study that was initiated in 2002 at The University of Texas MD Anderson Cancer Center (MDACC). All cases were histologically confirmed. The recruitment of cases has no restriction on age, sex, ethnicity, or cancer stage. Random digital dialing (RDD) was used to recruit healthy control subjects (14). Healthy controls were individuals who had no history of cancer (except nonmelanoma skin cancer) at the time of recruitment. A control subject had to have lived for ≥1 year in the same county or socioeconomically matched surrounding counties in which a case subject resided. Additional control subjects from an ongoing bladder cancer case–control study were also included. They were recruited from Kelsey Seybold Clinic and part of a previously published GWAS of bladder cancer (15). Only cases and controls who self-reported to have European ancestry were included for genotyping in this GWAS study. Epidemiologic data and a 40 mL blood samples were collected by trained MDACC staff interviewers. For smoking history, a never-smoker was defined as an individual who had never smoked or had smoked <100 cigarettes. Those subjects who smoked >100 cigarettes and had quit smoking >12 months prior to diagnosis for cases or prior to interview for controls were considered former smokers.
Phase II NCI RCC GWAS (NCI-GWAS): We used the U.S. NCI RCC GWAS to validate nominal significant SNPs (P < 0.05) identified in phase I. The NCI-GWAS consisted of 1,311 cases and 3,424 controls who were recruited from four studies (PLCO, CPS-II, ATBC, and USKC). The details of the study design and population characteristics were previously described (9).
Phase III MDA RCC OncoArray (MDA-OncoArray): Additional RCC cases were also obtained from MDACC, including additional cases from our ongoing case–control study and cases from the MD Anderson Cancer Patient Cohort (16). The healthy controls were recruited using the same RDD method that was used in phase I MDA-GWAS and was described above. MDA-OncoArray was used to validate the significant SNPs (P < 0.05) identified by combined MDA-GWAS and NCI-GWAS.
Informed consent was obtained from all participants. The study was approved by study-specific institutional review boards and/or ethics committees.
Genotyping and quality control
Phase I MDA-GWAS: The primary scan of the study population was performed at MDACC using either the Illumina HumanHap 660W Beadchips or the HumanHap 610 Beadchips. The details of quality control could be found in the previous publication (8). After quality control exclusions, a total of 894 cases and 1,516 controls and 533,191 SNPs remained in the analytic dataset.
Phase II NCI-GWAS: The primary scan of the NCI-GWAS was performed at the NCI Core Genotyping Facility with the Illumina HumanHap 500, 610, or 660w BeadChips. The details of quality control could be found in the previous publication (9). The study was composed of 1,311 cases and 3,424 controls, and 585,576 SNPs remained in the analytic dataset after quality control procedures.
Phase III MDA-OncoArray: The primary scan of the MDA-OncoArray study was performed at MDACC using the Illumina OncoArray-500K BeadChips. Duplicated and first-degree relatives were excluded by estimating the Identical by descent (pihat > 0.5, n = 357). Samples with the reported sex that did not match with X chromosome heterozygosity were excluded (n = 18). SNPs in the X chromosome were either coded as 0 or 2 for men. Subjects deviated from European ancestry (probability of European ancestry < 0.85) were identified using STRUCTURE (Version 2.3.2.1) and were removed from the current analysis (n = 57). The analytic dataset contained 1,325 cases and 774 controls after these exclusions. The quality control procedures on SNPs included exclusions of SNPs with minor allele frequency less than 0.01 (n = 84,416), call rate < 0.90 (n = 174), duplicated SNPs (n = 752), and SNPs that failed the Hardy–Weinberg equilibrium test among controls (P < 0.0001, n = 698). Imputation was performed using IMPUTE2 (17). After imputation and quality control (imputation score > 0.3), 19 of the 21 SNPs with P < 0.05 in both MDA-GWAS and NCI-GWAS were available for further validation.
Compilation of obesity-related genes
We previously compiled a gene list that consists of 2,051 obesity-related genes (18) from four sources: (1). Bioinformatics tool Text-mined Hypertension, Obesity, and Diabetes Candidate Gene Database (19). We restricted genes to those reported by three or more studies and further reviewed these genes in details; 216 candidate genes remained in further analysis. (2) Online database of obesity-related genes: Integratomics TIME (20). (3) Obesity-relevant pathways selected from Biocarta, KEGG, and Reactome pathway databases (15 pathways; Supplementary Table S1). (4) Genes covering/close to GWAS-confirmed loci for BMI or obesity. We downloaded the list consisted of 43 studies from A Catalog of Published Genome-Wide Association Studies (http://www.genome.gov/gwastudies/). The keywords used for searching were as follows: BMI, obesity, obesity (early onset extreme), obesity (extreme), BMI (interaction), adiposity, fat body mass, and weight. We included both the upstream and downstream genes closest to an intergenic SNP. Loci with genome-wide significant SNPs (P < 5 × 10−8) were eligible to be included. After removal of pseudo genes (n = 33), mitochondrial genes (n = 10), and genes withdrawn by National Center for Biotechnology Information or not found in University of California Santa Cruz Genome Browser (n = 7), 2,001 genes remained for further analysis. A total of 32,946 SNPs mapped to ± 10 kb of downstream/upstream of obesity-related genes were extracted from phase I MDA GWAS.
Statistical analysis
Additive genetic model was tested for each SNP in multivariable logistic regression with the adjustment of age, sex, and the top two eigenvectors in cases and controls to derive the OR and 95% CI. SNPs with P < 0.05 were selected for validation using NCI-GWAS. Sex, studies, and the top two eigenvectors were included as covariates in the logistic regression model for NCI-GWAS. SNPs achieving significant and consistent associations in phases I and II (both P < 0.05) were selected for further validation in MDA-OncoArray. The model adjustment included age, sex, and top two eigenvectors for MDA-OncoArray.
Proxies were identified for the nominal significant SNPs that were not genotyped in NCI-GWAS (R2 > 0.8). SNPtest (v2.4.1, https://mathgen.stats.ox.ac.uk/genetics_software/snptest/snptest.html) was used to perform the missing data likelihood score test for imputed SNPs (21). Meta-analysis was conducted for the 19 SNPs in the validation phases. If the P value of meta-analysis was <0.05 in the combined validation phases, the SNP was further tested in the combined three phases. A fixed-effect model was used if test of heterogeneity was not significant (P > 0.05). Otherwise, a random-effect model was selected. Expression quantitative trait loci (eQTL) analysis was assessed for the identified SNPs (including SNPs located within the same LD block, R2 > 0.8) in the TCGA-KIRC tumor tissue using the methods described by Li and colleagues (22). We performed eQTL analysis for our index SNPs and surrounding SNPs in LD. Three of five SNPs were successfully evaluated in the TCGA dataset. Enhancer and DNase enrichment analyses were evaluated using Haploreg (http://www.broadinstitute.org/mammals/haploreg/haploreg_v2.php; ref. 23).
Results
In the present multiphase study, we included three independent datasets which consisted of 3,530 cases and 5,714 controls. Nominal significant associations were found for 1,709 SNPs in the discovery phase (P < 0.05). We evaluated 1,661 SNPs with direct genotyping or proxies in phase II NCI-GWAS. Twenty-one SNPs showed consistent associations in both sets. The P values of meta-analysis ranged from 10−3 to 10−5 (Supplementary Table S2). No evidence of strong heterogeneity was found. We further evaluated these 21 SNPs using the MDA-OncoArray. Five SNPs were directly genotyped and 14 were successfully imputed (info score between 0.50 and 0.99; Supplementary Table S3), whereas rs1780361 and rs2237896 were excluded due to low imputation quality (imputation score < 0.3). Five SNPs were validated in the validation phases (Pmeta < 0.05 in the combined validation phases; Table 1). No consistent associations were found for other SNPs.
Meta-analysis of identified SNPs for all three phases
SNP . | Effect allele . | Chr . | Mapped gene . | OR (95% CI) . | Pmeta . | Pheterogeneity . |
---|---|---|---|---|---|---|
rs10521506 | G | X | IL1RAPL2 | |||
Discovery | ||||||
MDA-GWAS | 0.82 (0.71–0.95) | |||||
Validation | ||||||
NCI-GWAS, rs5962542, r2 = 0.97 | A | 0.84 (0.74–0.95) | ||||
MDA-OncoArraya | 0.90 (0.82–0.99) | |||||
Meta | 0.88 (0.81–0.95) | 5.86 × 10−4 | 0.419 | |||
All stages, meta | 0.87 (0.81–0.93) | 2.33 × 10−5 | 0.533 | |||
rs2229536 | A | 9 | PLIN2 | |||
Discovery | ||||||
MDA-GWAS | 0.74 (0.56–0.97) | |||||
Validation | ||||||
NCI-GWAS | 0.75 (0.60–0.93) | |||||
MDA-OncoArraya | 0.82 (0.60–1.22) | |||||
Meta | 0.77 (0.64–0.92) | 4.57 × 10−3 | 0.649 | |||
All stages, meta | 0.76 (0.65–0.88) | 3.87 × 10−4 | 0.872 | |||
rs4601989 | A | 15 | SMAD3 | |||
Discovery | ||||||
MDA-GWAS | 0.85 (0.74–0.99) | |||||
Validation | ||||||
NCI-GWAS | 0.85 (0.75–0.95) | |||||
MDA-OncoArray | 0.91 (0.77–1.07) | |||||
Meta | 0.87 (0.79–0.95) | 3.11 × 10−3 | 0.507 | |||
All stages, meta | 0.86 (0.80–0.93) | 2.71 × 10−4 | 0.789 | |||
rs10850596 | A | 12 | MED13L | |||
Discovery | ||||||
MDA-GWAS | 1.17 (1.04–1.32) | |||||
Validation | ||||||
NCI-GWAS | 1.16 (1.05–1.28) | |||||
MDA-OncoArraya | 1.05 (0.89–1.25) | |||||
Meta | 1.13 (1.04–1.23) | 5.24 × 10−3 | 0.336 | |||
All stages, meta | 1.14 (1.07–1.23) | 1.50 × 10−4 | 0.551 | |||
rs3761840 | G | 9 | TSC1 | |||
Discovery | ||||||
MDA-GWAS | 0.87 (0.77–0.98) | |||||
Validation | ||||||
NCI-GWAS | 0.9 (0.82–0.99) | |||||
MDA-OncoArray | 0.95 (0.84–1.09) | |||||
Meta | 0.92 (0.85–0.99) | 0.032 | 0.489 | |||
All stages, meta | 0.9 (0.85–0.97) | 2.47 × 10−3 | 0.607 |
SNP . | Effect allele . | Chr . | Mapped gene . | OR (95% CI) . | Pmeta . | Pheterogeneity . |
---|---|---|---|---|---|---|
rs10521506 | G | X | IL1RAPL2 | |||
Discovery | ||||||
MDA-GWAS | 0.82 (0.71–0.95) | |||||
Validation | ||||||
NCI-GWAS, rs5962542, r2 = 0.97 | A | 0.84 (0.74–0.95) | ||||
MDA-OncoArraya | 0.90 (0.82–0.99) | |||||
Meta | 0.88 (0.81–0.95) | 5.86 × 10−4 | 0.419 | |||
All stages, meta | 0.87 (0.81–0.93) | 2.33 × 10−5 | 0.533 | |||
rs2229536 | A | 9 | PLIN2 | |||
Discovery | ||||||
MDA-GWAS | 0.74 (0.56–0.97) | |||||
Validation | ||||||
NCI-GWAS | 0.75 (0.60–0.93) | |||||
MDA-OncoArraya | 0.82 (0.60–1.22) | |||||
Meta | 0.77 (0.64–0.92) | 4.57 × 10−3 | 0.649 | |||
All stages, meta | 0.76 (0.65–0.88) | 3.87 × 10−4 | 0.872 | |||
rs4601989 | A | 15 | SMAD3 | |||
Discovery | ||||||
MDA-GWAS | 0.85 (0.74–0.99) | |||||
Validation | ||||||
NCI-GWAS | 0.85 (0.75–0.95) | |||||
MDA-OncoArray | 0.91 (0.77–1.07) | |||||
Meta | 0.87 (0.79–0.95) | 3.11 × 10−3 | 0.507 | |||
All stages, meta | 0.86 (0.80–0.93) | 2.71 × 10−4 | 0.789 | |||
rs10850596 | A | 12 | MED13L | |||
Discovery | ||||||
MDA-GWAS | 1.17 (1.04–1.32) | |||||
Validation | ||||||
NCI-GWAS | 1.16 (1.05–1.28) | |||||
MDA-OncoArraya | 1.05 (0.89–1.25) | |||||
Meta | 1.13 (1.04–1.23) | 5.24 × 10−3 | 0.336 | |||
All stages, meta | 1.14 (1.07–1.23) | 1.50 × 10−4 | 0.551 | |||
rs3761840 | G | 9 | TSC1 | |||
Discovery | ||||||
MDA-GWAS | 0.87 (0.77–0.98) | |||||
Validation | ||||||
NCI-GWAS | 0.9 (0.82–0.99) | |||||
MDA-OncoArray | 0.95 (0.84–1.09) | |||||
Meta | 0.92 (0.85–0.99) | 0.032 | 0.489 | |||
All stages, meta | 0.9 (0.85–0.97) | 2.47 × 10−3 | 0.607 |
aSNPs were imputed by IMPUTE2. Info score: rs2229536 = 0.918; rs10521506 = 0.983; rs10850596 = 0.54.
Table 1 showed five significant SNPs in the overall meta-analysis (phases I–III). The most significant SNP, rs10521506 at ChrX:104770109 (GRCh37.p13), maps to an intron of IL1 receptor accessory protein-like 2 (IL1RAPL2). Minor allele G of rs10521506 was associated with a 13% reduced risk of RCC [ORmeta = 0.87 (0.81–0.93), Pmeta = 2.33 × 10−5, Phet = 0.533; Supplementary Fig. S1A]. Moreover, minor alleles A of rs2229536 at Chr9:19116543, A of rs4601989 at Chr15:67451954, and G of rs3761840 at Chr9:135804735 were also associated with a decreased RCC risk in the overall meta-analysis [for rs2229536: ORmeta = 0.76 (0.65–0.88), Pmeta = 3.87 × 10−4, Phet = 0.872, Supplementary Fig. S1B; for rs4601989: ORmeta = 0.86 (0.80–0.93), Pmeta = 2.71 × 10−4, Phet = 0.789, Supplementary Fig. S1C; for rs3761840: ORmeta = 0.90 (0.85–0.97), Pmeta = 2.47 × 10−3, Phet = 0.607, Supplementary Fig. S1E]. Finally, minor allele A of rs10850596 at Chr12: 116579736 was associated with a 14% increased RCC risk [ORmeta = 1.14 (1.07–1.23), Pmeta = 1.50 × 10−4, Phet = 0.551, Supplementary Fig. S1D]. No significant heterogeneity was observed in the overall meta-analysis. We further performed meta-analysis stratified by smoking status and observed similar associations among these five SNPs (Table 2).
Stratified analysis of identified SNPs for all three phases by smoking
. | Never-smoker . | Ever smoker . | ||||
---|---|---|---|---|---|---|
SNP . | OR (95% CI) . | Pmeta . | Pheterogeneity . | OR (95% CI) . | Pmeta . | Pheterogeneity . |
rs10521506 | ||||||
Discovery | ||||||
MDA-GWAS | 0.92 (0.80–1.06) | 0.89 (0.79–1.00) | ||||
Validation | ||||||
NCI-GWAS | 0.78 (0.60–1.01) | 1.10 (0.85–1.42) | ||||
MDA-OncoArray | 0.86 (0.74–1.00) | 1.00 (0.86–1.17) | ||||
Meta | 0.84 (0.74–0.95) | 7.39 × 10−3 | 0.53 | 1.03 (0.90–1.17) | 0.72 | 0.54 |
All stages, meta | 0.88 (0.80–0.96) | 5.80 × 10−3 | 0.52 | 0.95 (0.87–1.03) | 0.23 | 0.25 |
rs2229536 | ||||||
Discovery | ||||||
MDA-GWAS | 0.55 (0.34–0.88) | 0.87 (0.61–1.24) | ||||
Validation | ||||||
NCI-GWAS | 0.58 (0.38–0.90) | 0.77 (0.59–1.00) | ||||
MDA-OncoArray | 1.06 (0.66–1.71) | 0.79 (0.47–1.32) | ||||
Meta | 0.76 (0.55–1.05) | 9.83 × 10−2 | 0.07 | 0.77 (0.61–0.97) | 2.92 × 10−2 | 0.94 |
All stages, meta | 0.69 (0.53–0.90) | 5.94 × 10−3 | 0.10 | 0.80 (0.66–0.97) | 2.48 × 10−2 | 0.85 |
rs4601989 | ||||||
Discovery | ||||||
MDA-GWAS | 0.87 (0.70–1.09) | 0.85 (0.70–1.03) | ||||
Validation | ||||||
NCI-GWAS | 0.97 (0.79–1.02) | 0.79 (0.68–0.91) | ||||
MDA-OncoArray | 0.77 (0.60–0.99) | 1.07 (0.83–1.38) | ||||
Meta | 0.88 (0.75–1.04) | 0.12 | 0.17 | 0.90 (0.67–1.22)a | 0.49 | 0.04 |
All stages, meta | 0.88 (0.77–1.00) | 4.84 × 10−2 | 0.39 | 0.85 (0.76–0.94) | 1.96 × 10−3 | 0.11 |
rs10850596 | ||||||
Discovery | ||||||
MDA-GWAS | 1.06 (0.88–1.28) | 1.26 (1.07–1.47) | ||||
Validation | ||||||
NCI-GWAS | 0.98 (0.82–1.17) | 1.23 (1.09–1.39) | ||||
MDA-OncoArray | 1.26 (0.97–1.63) | 0.88 (0.66–1.19) | ||||
Meta | 1.06 (0.92–1.23) | 0.41 | 0.12 | 1.07 (0.78–1.48)a | 0.67 | 0.04 |
All stages, meta | 1.06 (0.95–1.19) | 0.31 | 0.30 | 1.20 (1.10–1.32) | 6.75 × 10−5 | 0.09 |
rs3761840 | ||||||
Discovery | ||||||
MDA-GWAS | 0.81 (0.67–0.98) | 0.91 (0.77–1.07) | ||||
Validation | ||||||
NCI-GWAS | 0.96 (0.80–1.15) | 0.89 (0.79–1.00) | ||||
MDA-OncoArray | 0.82 (0.68–1.00) | 1.06 (0.86–1.31) | ||||
Meta | 0.89 (0.78–1.02) | 8.38 × 10−2 | 0.25 | 0.93 (0.84–1.03) | 0.15 | 0.15 |
All stages, meta | 0.86 (0.77–0.96) | 7.79 × 10−3 | 0.38 | 0.92 (0.84–1.01) | 6.88 × 10−2 | 0.35 |
. | Never-smoker . | Ever smoker . | ||||
---|---|---|---|---|---|---|
SNP . | OR (95% CI) . | Pmeta . | Pheterogeneity . | OR (95% CI) . | Pmeta . | Pheterogeneity . |
rs10521506 | ||||||
Discovery | ||||||
MDA-GWAS | 0.92 (0.80–1.06) | 0.89 (0.79–1.00) | ||||
Validation | ||||||
NCI-GWAS | 0.78 (0.60–1.01) | 1.10 (0.85–1.42) | ||||
MDA-OncoArray | 0.86 (0.74–1.00) | 1.00 (0.86–1.17) | ||||
Meta | 0.84 (0.74–0.95) | 7.39 × 10−3 | 0.53 | 1.03 (0.90–1.17) | 0.72 | 0.54 |
All stages, meta | 0.88 (0.80–0.96) | 5.80 × 10−3 | 0.52 | 0.95 (0.87–1.03) | 0.23 | 0.25 |
rs2229536 | ||||||
Discovery | ||||||
MDA-GWAS | 0.55 (0.34–0.88) | 0.87 (0.61–1.24) | ||||
Validation | ||||||
NCI-GWAS | 0.58 (0.38–0.90) | 0.77 (0.59–1.00) | ||||
MDA-OncoArray | 1.06 (0.66–1.71) | 0.79 (0.47–1.32) | ||||
Meta | 0.76 (0.55–1.05) | 9.83 × 10−2 | 0.07 | 0.77 (0.61–0.97) | 2.92 × 10−2 | 0.94 |
All stages, meta | 0.69 (0.53–0.90) | 5.94 × 10−3 | 0.10 | 0.80 (0.66–0.97) | 2.48 × 10−2 | 0.85 |
rs4601989 | ||||||
Discovery | ||||||
MDA-GWAS | 0.87 (0.70–1.09) | 0.85 (0.70–1.03) | ||||
Validation | ||||||
NCI-GWAS | 0.97 (0.79–1.02) | 0.79 (0.68–0.91) | ||||
MDA-OncoArray | 0.77 (0.60–0.99) | 1.07 (0.83–1.38) | ||||
Meta | 0.88 (0.75–1.04) | 0.12 | 0.17 | 0.90 (0.67–1.22)a | 0.49 | 0.04 |
All stages, meta | 0.88 (0.77–1.00) | 4.84 × 10−2 | 0.39 | 0.85 (0.76–0.94) | 1.96 × 10−3 | 0.11 |
rs10850596 | ||||||
Discovery | ||||||
MDA-GWAS | 1.06 (0.88–1.28) | 1.26 (1.07–1.47) | ||||
Validation | ||||||
NCI-GWAS | 0.98 (0.82–1.17) | 1.23 (1.09–1.39) | ||||
MDA-OncoArray | 1.26 (0.97–1.63) | 0.88 (0.66–1.19) | ||||
Meta | 1.06 (0.92–1.23) | 0.41 | 0.12 | 1.07 (0.78–1.48)a | 0.67 | 0.04 |
All stages, meta | 1.06 (0.95–1.19) | 0.31 | 0.30 | 1.20 (1.10–1.32) | 6.75 × 10−5 | 0.09 |
rs3761840 | ||||||
Discovery | ||||||
MDA-GWAS | 0.81 (0.67–0.98) | 0.91 (0.77–1.07) | ||||
Validation | ||||||
NCI-GWAS | 0.96 (0.80–1.15) | 0.89 (0.79–1.00) | ||||
MDA-OncoArray | 0.82 (0.68–1.00) | 1.06 (0.86–1.31) | ||||
Meta | 0.89 (0.78–1.02) | 8.38 × 10−2 | 0.25 | 0.93 (0.84–1.03) | 0.15 | 0.15 |
All stages, meta | 0.86 (0.77–0.96) | 7.79 × 10−3 | 0.38 | 0.92 (0.84–1.01) | 6.88 × 10−2 | 0.35 |
aOR and 95% CI estimated from random effect meta-analysis since P for heterozygosity < 0.05.
Further functional characterization of rs3761840 (proxy: rs1076160, r2 = 0.92) showed a suggestive eQTL effect on tuberous Sclerosis 1 (TSC1; Table 3). TSC1 expressions were lower in renal tissues with GG+GA genotypes, when compared with tissues with AA genotypes, although the association was borderline significant (β = −0.108, P = 0.078). The regions where rs3761840 resides had no overlap with CpG islands (data described in text only). No statistical significance in the enhancer and DNase enrichment analyses was observed for rs3761840 (Supplementary Table S4).
eQTL analysis of identified SNPs in the TCGA KIRC tumor tissues
SNP . | Proxy SNP . | r2 . | Gene . | β (95% CI) . | P . |
---|---|---|---|---|---|
rs2229536 | — | ||||
AA+AG vs. GG | — | — | PLIN2 | 0.266 (−0.134 to 0.665) | 0.191 |
rs4601989 | rs2033785 | ||||
AA+AG vs. GG | CC+CG vs. GG | 0.96 | SMAD3 | 0.018 (−0.08 to 0.115) | 0.71 |
rs3761840 | rs1076160 | ||||
GG+GA vs. AA | GG+GA vs. AA | 0.92 | TSC1 | –0.108 (−0.229 to 0.012) | 0.078 |
SNP . | Proxy SNP . | r2 . | Gene . | β (95% CI) . | P . |
---|---|---|---|---|---|
rs2229536 | — | ||||
AA+AG vs. GG | — | — | PLIN2 | 0.266 (−0.134 to 0.665) | 0.191 |
rs4601989 | rs2033785 | ||||
AA+AG vs. GG | CC+CG vs. GG | 0.96 | SMAD3 | 0.018 (−0.08 to 0.115) | 0.71 |
rs3761840 | rs1076160 | ||||
GG+GA vs. AA | GG+GA vs. AA | 0.92 | TSC1 | –0.108 (−0.229 to 0.012) | 0.078 |
NOTE: The r2 indicates LD in the 1000 Genome phase III CEU population. A dominant model was tested due to small number of subjects with homozygous minor allele.
Discussion
In the current study, we explored the potential associations between common genetic variants in obesity-related genes and RCC risk. Our analyses showed that five potential new loci mapped to Chr9 p22.1, Chr9 q34.13, Chr12 q24.21, Chr15 q22.32, and ChrX q22.3 may confer susceptibility to RCC. To the best of our knowledge, this is the first study that comprehensively evaluated the chromosomal regions covering known and putative obesity-related genes to search for potential susceptibility loci of RCC.
Relatively high heritability was observed for kidney cancer, which indicates an important role of genetic variations in the disease etiology (10). The estimated heritability of kidney cancer was higher than that of cancer overall, breast, bladder, lung, and colon cancer. Few susceptibility loci have been identified by GWAS for RCC compared with other major cancers. Almost all of these RCC susceptibility loci were confirmed in this study except rs4765623 at 12q24.31 (Supplementary Table S5). Considering the relatively high heritability estimated for RCC and the stringent significance level set for published GWAS, additional susceptibility loci might be ignored due to their moderate significance in the association tests.
Being obese is partially attributed to genetic variations. For example, knocking out the fat mass and obesity-associated (FTO) gene resulted in a significant decrease of body weight and white adipose mass in mouse (24), and mutations in melanocortin 4 receptor (MC4R) were associated with obesity in humans (25, 26). Obesity is associated with the etiology of multiple cancers (27) and is considered an established risk factor for RCC (3, 28). It is likely that genetic variations in the shared pathways are essential for etiology of both obesity and RCC. One study reported that A allele of rs9939609 (FTO) was weakly associated with increased risk of kidney cancer (11). However, we did not observe such association in our datasets, although the SNP was not directly genotyped (proxy: rs3751812, R2 = 0.99, data described in text only). Another study specifically assessed three genetic variants in the adiponectin gene (ADIPOQ) in a case–control study, and nominal significant association was found for rs182052 (12). Neither rs182052 nor good proxies were genotyped in the discovery phase. Therefore, the association was not evaluated in the current study. Instead, four other SNPs located in ADIPOQ were tested and did not have significant associations with RCC risk (P > 0.30, data described in text only). Previously, we showed that genetic variations in the mTOR signaling pathway were in association with RCC risk (13). The top significant SNPs (rs4132509 and rs3766673) identified in MDA RCC GWAS was not further evaluated in phase III study as only suggestive associations were found in the NCI-GWAS (P = 0.18 for rs4132509 and P = 0.09 for rs3766673). However, the associations were consistent with OR in the same direction for MDA-GWAS and NCI-GWAS, which deserve further investigations in future studies.
Our multiphase study discovered five potential novel susceptibility loci (represented by rs10521506, rs2229536, rs4601989, rs10850596, and rs3761840) for RCC which achieved statistical significance in the meta-analysis. Suggestive associations with depression, amyotrophic lateral sclerosis, and rheumatoid arthritis were previously found for rs10521506 (https://grasp.nhlbi.nih.gov/Search.aspx). rs10521506 is located in the intron of IL1RAPL2, which belongs to the IL1 receptor (IL1R) family (29). A large body of evidence links obesity to insulin resistance through obesity-associated inflammation in which IL1 and IL1R families are heavily involved (30). The IL1R family is composed of ten molecules. The biological impact of IL1R molecules on the immune and inflammatory responses varies (e.g., ILR1 initiates the signaling, whereas IL1R2 inhibits it when binding to the IL1 agonist ligands; ref. 31). In contrast, less is known for the role played by IL1RAPL2 or IL1R10, but speculations suggested that it may have downregulating functions on inflammation (31).
Another locus identified (rs2229536) by our study links PLIN2, which encodes perilipin 2, with RCC susceptibility. Perilipin 2, previously named adipophilin or adipose differentiation–related protein, is one of lipid droplet proteins (32). Abundant intracellular lipid droplets are considered an essential feature of clear cell RCC (33). It closely correlated with the activation of hypoxia-inducible factor 2 alpha in clear-cell RCC patient samples (33). Furthermore, urinary perilipin 2 has been proposed as an early-detection screening biomarker for RCC, which has shown promising predictive performance in differentiating cancer patients from healthy controls, benign kidney tumors, noncancerous kidney diseases, and other cancers including bladder and prostate cancer (34–36). In a recent study, urinary perilipin 2 was evaluated prospectively in a population undergoing routine abdominal CT, which showed its potentials in the screening setting (34). Our results further support the biological plausibility of rs2229536 in association with RCC risk. Previous studies have shown that rs2229536 was also associated with various obesity-related phenotypes such as BMI, total cholesterol, and cholesterol changes with statin use, but none reached genome-wide significance (https://grasp.nhlbi.nih.gov/Search.aspx).
SMAD3 also plays an important role in regulating glucose energy homeostasis. It has been shown that Smad3-deficient mice are protected from diet-induced insulin resistance, obesity, and diabetes (37, 38). These observations are in line with our findings that minor allele A of rs4601989 is associated with lower expressions of SMAD3 and reduced risk of RCC. Another interesting gene that rs3761840 mapped to is TSC1. It is well known that TSC1 is a critical tumor suppressor in the mammalian target of rapamycin complex (mTOR) pathway (39). Previous studies including our work have shown that the mTOR pathway is associated with both energy balance and RCC risk (13, 40). Finally, less is known about MED13L and its role in the genetics–obesity–RCC link. One previous study showed that SNPs located downstream of MED13L were suggestively associated with end-stage kidney disease and chronic kidney disease in Mexican Americans (41).
In summary, our analyses identified five potential susceptibility loci, Chr9 p22.1 (rs2229536), Chr9 q34.13 (rs3761840), Chr12 q24.21 (rs10850596), Chr15 q22.32 (rs4601989), and ChrX q22.3 (rs10521506), for RCC through a multiphase design. Although the associations are moderate, the identified loci may provide new insights into the disease etiology that reveal importance of obesity-related genes in RCC development. More studies are demanded to validate our results and elucidate the underlying biological mechanisms.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Authors' Contributions
Conception and design: C.G. Wood, N. Rothman, X. Wu
Development of methodology: M.P. Purdue, N. Rothman
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.P. Purdue, C.G. Wood, N.M. Tannir, D. Albanes, S.M. Gapstur, N. Rothman, S.J. Chanock
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): X. Shu, M.P. Purdue, Y. Ye, H. Tu, Z. Wang, X. Wu
Writing, review, and/or revision of the manuscript: X. Shu, M.P. Purdue, Y. Ye, H. Tu, C.G. Wood, N.M. Tannir, Z. Wang, D. Albanes, S.M. Gapstur, V.L. Stevens, N. Rothman, S.J. Chanock, X. Wu
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): X. Wu
Study supervision: C.G. Wood, X. Wu
Other (acquisition of grant funding): X. Wu
Grant Support
This work was supported in part by the NIH (grant R01 CA170298 to X. Wu), and the Center for Translational and Public Health Genomics, Duncan Family Institute for Cancer Prevention, The University of Texas MD Anderson Cancer Center. The NCI RCC GWAS was supported by the Intramural Research Program of the NIH, NCI.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.