Genome-Wide Association Study Identifies Loci for Liver Enzyme Concentrations in Mexican Americans: The GUARDIAN Consortium.

OBJECTIVE: Populations of Mexican American ancestry are at an increased risk for nonalcoholic fatty liver disease. The objective of this study was to determine whether loci in known and novel genes were associated with variation in aspartate aminotransferase (AST) (n = 3,644), alanine aminotransferase (ALT) (n = 3,595), and gamma-glutamyl transferase (GGT) (n = 1,577) levels by conducting the first genome-wide association study (GWAS) of liver enzymes, which commonly measure liver function, in individuals of Mexican American ancestry.
METHODS: Levels of AST, ALT, and GGT were determined by enzymatic colorimetric assays. A multi-cohort GWAS of individuals of Mexican American ancestry was performed. Single-nucleotide polymorphisms (SNP) were tested for association with liver outcomes by multivariable linear regression using an additive genetic model. Association analyses were conducted separately in each cohort, followed by a nonparametric meta-analysis.
RESULTS: In the PNPLA3 gene, rs4823173 (P = 3.44 × 10-10 ), rs2896019 (P = 7.29 × 10-9 ), and rs2281135 (P = 8.73 × 10-9 ) were significantly associated with AST levels. Although not genome-wide significant, these same SNPs were the top hits for ALT (P = 7.12 × 10-8 , P = 1.98 × 10-7 , and P = 1.81 × 10-7 , respectively). The strong correlation (r2  = 1.0) for these SNPs indicated a single hit in the PNPLA3 gene. No genome-wide significant associations were found for GGT.
CONCLUSIONS: PNPLA3, a locus previously identified with ALT, AST, and nonalcoholic fatty liver disease in European and Japanese GWAS, is also associated with liver enzymes in populations of Mexican American ancestry.

Introduction

Liver enzymes are commonly used to assess liver function and disease. High concentrations of aspartate aminotransferase (AST), alanine aminotransferase (ALT), and gamma-glutamyl transferase (GGT) are associated with various conditions, including non-alcoholic fatty liver disease (NAFLD), the most common cause of elevated liver enzymes in adults in the United States (1). NAFLD is often benign; however, it can lead to inflammation, fibrosis, and eventually, cirrhosis (2). NALFD is more common in Hispanic Americans than either African Americans or Caucasians (3, 4).

Plasma concentrations of liver enzymes are highly heritable (20-77%) (5), suggesting a genetic role that may help in interpretation of results and explain variation among individuals. Previous genome-wide association studies (GWAS) of liver enzymes in those of European (6, 7), Indian (7), and Japanese ancestry (8) identified candidate genes involved in various pathways, including glucose and lipid metabolism as well as inflammation and immunity. However, few studies have extensively examined the genetic contributions in Hispanics, a group at increased risk of NALFD (9), and no GWAS studies of liver enzymes or NAFLD have been conducted in individuals of Mexican-American ancestry.

The Genetics Underlying Diabetes in Hispanics (GUARDIAN) Consortium was formed to identify the genetic determinants of type 2 diabetes-related phenotypes in Hispanics (10), and is comprised of five family-based [BetaGene (11), Hypertension-Insulin Resistance Family study (HTN-IR) (12), Insulin Resistance Atherosclerosis Family Study (IRAS FS) (13), Mexican-American Coronary Artery Disease study (MACAD) (14), and NIDDM-Atherosclerosis Study (NIDDM-Athero) (15)] and two non–family-based studies [IRAS (16) and Troglitazone in the Prevention of Diabetes study (TRIPOD) (17)]. The majority of these studies also included measures of liver function, with measures of AST, ALT and GGT. The objective of this study was to determine whether loci in known and novel genes were associated with variation in AST, ALT and GGT levels in Mexican-American ancestry populations in the GUARDIAN Consortium by conducting the first GWAS of liver enzymes in individuals of Mexican-American ancestry.

Methods

Study Population

All participants provided written informed consent, and the institutional review boards at the respective institutions approved the study. Five cohorts of self-reported individuals of Mexican-American ancestry, ascertained on varying conditions, had measures of liver enzymes and were used in these analyses. Cohort descriptions can be found in the Supplementary Materials.

Outcome Variables and Covariates

Concentrations in IU/L of ALT, AST, and GGT were determined by enzymatic colorimetric assays. Age and sex were self-reported. Body mass index (BMI) was calculated as weight (kg)/height (m2).

Genotyping

All samples were genotyped on the Illumina HumanOmniExpress BeadChip, and alleles were called using GenomeStudio software (Illumina, San Diego, CA) (18, 19). Samples with call rates >0.98, single nucleotide polymorphisms (SNPs) with call rates >0.99, and minor allele frequency (MAF) >0.001 passed laboratory quality control (20), with 22,000 additional SNPs manually reviewed for clustering accuracy. Samples were removed from analysis if the overall call rate was <0.98, if the samples were genetic outliers for sex and admixture proportions, if the samples were monomorphic , or if there was inconsistent fingerprinting from existing SNP data (21).The primary inferential SNPs did not exhibit differential missingness by trait, had a SNP call rate >98%, and did not depart from Hardy- Weinberg equilibrium expectations. Pedigrees were examined for consistency of stated family structure. Each SNP was examined for Mendelian inconsistencies using PedCheck (Program for Detecting Marker Typing Incompatibilities in Pedigree Data,http://watson.hgen.pitt.edu/register/docs/pedcheck.html). and inconsistencies were converted to missing. A maximum of 659,222 SNPs were meta-analyzed.

Population Stratification

Population substructure was estimated using ADMIXTURE version 1.21 (http://www.genetics.ucla.edu/software/admixture) at each study site based on SNPs that passed quality control [n = 117,347 linkage disequilibrium (LD)–pruned SNPs]. Data from the HapMap Project, including Mexican ancestry, were used as reference populations (21). Admixture proportions were included as covariates in the tests of association with liver enzymes.

Statistical Analysis

All cohorts had measures of AST and ALT, while only HTN-IR and IRAS FS had data available for GGT. To best approximate normality and homogeneity of variance in all cohorts, levels of AST, ALT and GGT were log-transformed. There are 3757 individuals for genotyping. Variance component analyses as implemented in the GWAF (Genome-Wide Association analyses with Family) (22) or SOLAR (Sequential Oligogenic Linkage Analysis Routines) (23) programs were used to test for association in family cohorts and linear regression analyses as implemented in QSNPGWA (http://github.com/guyrt/WFUBMC) were used to test for association in the non-family-based cohort. SNPs were tested for association with liver enzymes using an additive genetic model adjusted for age, sex, BMI, admixture proportions, family structure, and recruitment center (for multi-center studies). Conditional analyses were performed for significant loci with multiple uncorrelated variants by including the most significant variant as an additional covariate. For significant uncorrelated loci, SNP-age and SNP-sex interactions were examined.

Association analyses were carried out separately in each cohort followed by a non-parametric meta-analysis as implemented in METAL (http://www.sph.umich.edu/csg/abecasis/metal). For each liver enzyme outcome, a weighted, fixed-effects meta-analysis was calculated, weighting by cohort sample size due to different study design and ascertainment criteria, and correcting for genomic control inflation factors. A MAF >0.01 filter resulted in 659,222 SNPs for AST, 659,211 SNPs for ALT, and 655,023 SNPs for GGT. Genome-wide significance was determined at P<1×10−8, allowing for a Bonferroni adjustment for 106 SNPs tested and three liver enzyme outcomes.

Known variants (n=31) from previous liver enzyme and NAFLD GWAS that were available in the GUARDIAN Consortium genotype data were examined for association with AST, ALT and GGT. Associations were reported for enzyme specific variants (e.g. AST associations for reported AST variants), and associations with all liver enzymes were reported for previously reported NAFLD/NASH variants. If reported variants in the same gene had a LD > 0.80 in the HapMap Project Mexican population, then only the variant with the highest p-value was reported. Significance for the known variants was determined at P<0.05. We did not correct for multiple testing due to a priori hypotheses based on well-known associations.

Results

Characteristics of the study participants are shown in Table 1. Mean age ranged from 35-41 years, the majority was female, and mean BMI ranged between 28-31 kg/m2. Mean levels of AST, ALT and GGT were within normal ranges (10-40 IU/L, 7-56 IU/L, and 0-45 IU/L, respectively.

Table 1.

Characteristics of the GUARDIAN Consortium meta-analysis cohorts.

	BetaGene	HTN-IR	IRAS FS	MACAD	TRIPOD
Sample Size	1195	689	1034	714	125
Age	34.69 ± 7.92	37.28 ± 14.13	40.60 ± 13.7	34.49 ± 8.86	34.72 ± 6.26
Female (%)	71.72	58.90	58.90	56.74	100.00
BMI (kg/m2)	29.58 ± 6.09	28.73 ± 5.45	28.30 ± 5.80	28.92 ± 5.11	30.50 ± 5.43
AST (IU/L)	23.21 ± 11.60	24.57 ± 13.09	19.10 ± 8.90	24.86 ± 10.92	18.98 ± 8.10
ALT (IU/L)	36.83 ± 24.56	35.19 ± 23.88	11.30 ± 9.90	28.86 ± 20.65	26.21 ± 16.40
GGT (IU/L)	NA	31 ± 25.51	39.40 ± 43.70	NA	NA

Data are mean ± SD unless otherwise indicated. NA, not applicable. BMI, body mass index. AST, aspartate aminotransferase. ALT, alanine aminotransferase. GGT, gamma-glutamyl transferase. Data on GGT was not available in BetaGene, TRIPOD nor MACAD.

The P-P plot of P values (Supplementary Figure S1) from the meta-analysis did not suggest any systematic biases, such as those related to population stratification. We identified three SNPs, within the PNPLA3 (patatin-like phospholipase domain containing 3) locus on chromosome 22, attaining genome-wide significance (P < 1 × 10−8) with AST levels (Table 2, Figure 1A, Supplementary Table S1) with a positive effect direction. These same three SNPs, rs4823173, rs2896019, and rs2281135, were also the top signals for ALT levels (Table 2, Figure 2B, Supplementary Table S2) in the same positive effect direction, approaching genome-wide significance (P<7.12 × 10−8). There were no genome-wide significant associations found with GGT levels (Table 2, Figure 2C, Supplementary Table S3), with the top variants representing ZNF804A (Zinc protein finger 804A) and USP24P1 (ubiquitin specific peptidase 24 pseudogene 1). There were no significant rs482173xage (p=0.60 and p=0.47) or rs482173xsex interactions (p=0.54 and p=0.53) for AST and ALT, respectively.

Table 2.

Top hits in the GUARDIAN Consortium liver enzyme meta-analysis by trait (P < 1 × 10−7).

SNP	Chromosome	Position[1]	Gene	Alleles[2]	MAF	Trait	P-value
rs4823173	22	44328730	PNPLA3	A/G	0.42	AST	3.44 × 10−10
rs2896019	22	44333694	PNPLA3	G/T	0.43	AST	7.29 × 10−9
rs2281135	22	44332570	PNPLA3	A/G	0.43	AST	8.73 × 10−9
rs178571	6	136346130	PDE78	T/C	0.03	AST	3.21 × 10−8
rs1883350	22	44328043	PNPLA3	T/C	0.49	AST	5.22 × 10−7
rs4823173	22	44328730	PNPLA3	A/G	0.42	ALT	7.12 × 10−8
rs2281135	22	44332570	PNPLA3	A/G	0.43	ALT	1.81 × 10−7
rs2896019	22	44333694	PNPLA3	G/T	0.43	ALT	1.98 × 10−7
rs9601485	13	81475442	intergenic	T/C	0.12	ALT	3.17 × 10−7
rs10497655	2	185462041	ZNF804A	C/T	0.30	GGT	3.10 × 10−7
rs9511099	13	19591944	USP24P1	G/T	0.21	GGT	3.72 × 10−7
rs10458877	11	15941099	intergenic	T/C	0.13	GGT	8.59 × 10−7

SNP, single nucleotide polymorphism. MAF, minor allele frequency. AST, aspartate aminotransferase. ALT, alanine aminotransferase. GGT, gamma-glutamyl transferase.

For BetaGene, there are n=1191 for AST and n=1193 for ALT, respectively; for IRAS FS, there are n=925 n=878, and n=888 samples with AST, ALT, and GGT respectively; sample size for HTN is n=688 for ALT; and sample size for ALT is n=711 in MACAD.

Build hg19.

Minor/Major allele.

Figure 1.

Genome-wide Manhattan plots for the GUARDIAN Consortium meta-analysis. SNPs above the red line were genome-wide significant at P < 1 × 10−8. A. Results for AST of 659,222 SNPs indicate a genome-wide significant loci for PNPLA3 (P=3.44 × 10 −10). B. Results for ALT of 659,211 SNPs. C. Results for GGT of 655,023 SNPs.

There was perfect LD between rs4823173, rs2896019, and rs2281135 (r2=1.0 for all correlations) in the HapMap Project Mexican population; combined with the regional plot using the 1000 genomes AMR population, this indicates a single hit on the PNPLA3 gene (Figure 2). For the significant hit in the PNPLA3 gene, the OmniExpress imputed data from the PNPLA3 gene region on Chromosome 22 was pulled as it included rs738409, the most commonly reported significant SNP in the literature. To assess the evidence for multiple independent association signals within PNPLA3 for AST levels, association with rs4823173 (the most significant SNP in the meta-analysis) were tested after adjusting for rs738409. After adjusting for rs738409 along with the standard covariates, rs4823173 was no longer statistically significant (p=0.32). The loss of significance of rs4823173 after adjustment for rs738409, the top literature SNP, was consistent with the LD among the SNPs in this population (r2=0.84) and does not provide evidence for multiple association signals.

Figure 2.

Regional plot of the PNPLA3 locus in the GUARDIAN Consortium meta-analysis for AST. Genotyped SNPs are plotted with their meta-analysis p-values (−log10) according to their chromosomal position (hg19). The purple diamond represents rs4823173, the top variant in the meta-analysis for AST; red circles represent rs2896019 and rs2281135 (r2≥0.80).

Thirty one variants reported in other liver enzyme or NAFLD/NASH GWAS in European (6, 7, 24-28), Indian(7, 28), Japanese(8, 29), and Korean (30) populations (Table 3) were examined. Beyond the PNPLA3 gene, several SNPs associated with NAFLD showed significant associations with liver enzymes in the Mexican-American ancestry populations. rs3761472 in SAMM50 was significantly associated with AST, ALT and GGT (P=9×10−4, 3×10−4, and 0.01, respectively), and variants in PARVB were significantly associated with AST (P=3.0×10−4) and GGT (P=0.02) levels. For previous enzyme specific variants, prior ALT-associated variants rs11597086 in CHUK (P=0.03) and rs11597390 in CPN1 (P=0.02) were significantly associated with ALT. In addition, several prior GGT-associated variants were associated with GGT levels in individuals of Mexican-American ancestry: 7310409 in HNF1A (P=0.02), rs4820599 in GGT1 (P=0.04), rs4547811 in ZNF827 (P=0.03), rs944002 in C14orf73 (P=1.0×10−4), and rs339969 and rs340005 in RORA (P=0.02 for both). Aside from the variants in PNPLA3 and SAMM50, no prior AST variants were significantly associated with AST in individuals of Mexican-American ancestry.

Table 3.

GUARDIAN Consortium meta-analysis P values for regions reported in other liver enzyme or NAFLD/NASH GWAS. P value is listed for the GUARDIAN Consortium meta-analysis.

Gene	SNP	PreviouslyAssociated GWASTraits	ASTGUARDIANConsortium P-value	ALTGUARDIANConsortium P-value	GGTGUARDIANConsortium P-value
PNPLA3	rs2896019rs381062	AST [9,30], ALT[9,10,30],NAFLD/NASH[9,29,31]	P=7.29×10−9P=0.78	P=1.98×10−7P=0.62	P=0.008P=0.89
SAMM50	rs738491rs3761472	ALT[30]NAFLD/NASH[29]	P=0.11P=9×10−4	P=0.08P=3×10−4	P=0.37P=0.01
CHUK	rs11597086	ALT[30]		P=0.03
CPN1	rs11597390	ALT[9,30]		P=0.02
ST6GALNAC3	rs4949718	ALT[32]		P=0.81
CELF2	rs596406	AST[28],ALT[32]	P=0.92	P=0.59
GOT1	rs17109512	AST[28]	P=0.28
ST6GALNAC3	rs4949718	AST[28]	P=0.64
ZP4	rs2499604	AST[26]	P=0.64
PZP	rs6487679	AST[26]	P=0.45
HNF1A	rs1169313rs1169288rs7310409	GGT[9,27,30]			P=0.10P=0.29P=0.02
GGT1	rs4820599rs5751901	GGT [9,27,30]			P=0.04P=0.09
C2ofr16	rs1260326	GGT[9]			P=0.23
ZNF827	rs4547811	GGT[9]			P=0.03
ITGA1	rs4074793	GGT[9]			P=0.93
MLXIPL	rs17145750	GGT[9]			P=0.81
C14orf73	rs944002	GGT[9,27]			P=1.0×10−4
RORA	rs339969rs340005	GGT[9,27]			P=0.02P=0.02
CD276	rs8038465	GGT[9]			P=0.59
FUT2	rs516246	GGT[9]			P=0.74
FDFT1	rs2645424	NAFLD/NASH[26]	P=0.09	P=0.77	P=0.70
PARVB	rs5764455rs6006611	NAFLD/NASH[29]	P=0.72P=3.0×10−4	P=0.26P=0.08	P=0.02P=0.75
SLC38A8	rs11864146	NAFLD/NASH[26]	P=0.77	P=0.80	P=0.72
GCKR	rs780094	NAFLD/NASH[29]	P=0.33	P=0.39	P=0.34

Discussion

We conducted the first GWAS of liver enzymes in individuals of Mexican-American ancestry, a population at increased risk of NAFLD. This study provided evidence that PNPLA3 is associated with liver function in Mexican-Americans. We report genotyped data and therefore may limit our power to detect associations not included in our genotyping array by not imputing. However, while previous candidate gene studies in Hispanics have reported associations between variants in PNPLA3 and increased AST and ALT levels (31-35), we report a genome-wide significant association (P < 1 × 10−8) with AST and near significant association with ALT. There was high LD between rs4823173 with rs738409 in this population (r2=0.84), the most commonly reported PNPLA3 SNP in the literature, indicating that there is likely one association signal between PNPLA3 and levels of AST and ALT. These results demonstrate that PNPLA3, a locus previously identified by GWAS for ALT and AST in European(6, 7), Indian(7), and Japanese ancestry (8), is associated with liver function in Hispanics as well.

The PNPLA3 gene is located on chromosome 22 and has been implicated in genetic studies of liver function in both GWAS (6-8, 24, 26-30) and candidate gene (33, 36-40) studies across several ancestries. The PNPLA3 protein is expressed in both adipocytes and hepatocytes (41). PNPLA3 mediates triacylglycerol hydrolysis in adipocytes (42), and may be involved in energy storage and usage.

We confirmed several previously reported NAFLD SNPs and their associations with AST, ALT, and GGT. Variants in SAMM50, which is in the same genetic region as PNPLA3 and is involved in the biogenesis of mitochondria, and PARVB, involved in lipid accumulation and fibrosis, were previously associated with NAFLD/NASH and were associated with AST, ALT and/or GGT in our cohort of individuals of Mexican-American ancestry. Recently, in a candidate gene study in a Mexican population with NAFLD, variants in SAMM50 have been associated with increased with AST levels (35) Interestingly, using the browsers set up by the Genotype-Tissue Expression (GTEx) project (43), we found that our associated intronic variant rs4823173, which is in strong LD with PNPLA3 rs738409, is located near an exon, and SAMM50 expression, not PNPLA3, is affected by rs4821373in subcutaneous adipose tissue. These pathways could indicate that fatty liver is still ongoing and possibly progressing in this largely overweight/obese population with increasing accumulation of adipose tissue. In addition, we confirmed several prior enzyme-specific associations in our populations. For ALT, associations were found for variants in or near genes involved in inflammation and possibly modulation of insulin resistance through NF-kB (CPN1-CHUK gene cluster). For GGT, associations in or near genes involved in glutathione metabolism (GGT1), inflammation (HNF1A, RORA) and lipid metabolism (HNF1A), and several other genes with cardiombetabolic associations but with currently unknown functionality (ZNF827, C14orf73). Interestingly, we did not find any other significant associations with AST levels, indicating that PNPLA3, SAMM50 and PARVB, all previously associated with NAFLD, may be the largest contributors to AST levels of the variants studied in individuals of Mexican-American ancestry.

Similar to other studies (36, 37), we report a higher minor allele frequency in individuals of Mexican-American ancestry compared to individuals of European, African American, or Asian ancestry. The similar effect size reported across ancestries (36) but the higher affect allele frequency may explain, in part, the higher AST and ALT levels in Hispanics (4, 44), as well as Hispanics greater risk of NAFLD (3, 4, 44, 45). One limitation to our study is that we only have liver density measures in one cohort (IRAS FS) so we could not extend our GWAS to include NAFLD as an outcome. However, the IRAS FS has previously reported highly significant associations between rs738409 in PNPLA3 and NAFLD (36, 40). To the extent that our associations with higher AST or ALT levels could be indicative of pre-clinical liver inflammation in Hispanics, our findings suggest that the presence of the PNPLA3 risk alleles could indicate increased risk of NAFLD within this population.

Several studies have reported sex and age differences in the association of PNPLA3 variants and liver enzymes, where younger females had an association with higher ALT and males had an association with higher AST levels (32, 38). We did not see any significant sex or age interactions with rs4823173. This could be due to differences in ancestry, or the younger age of our population compared to the other studies, as the age effect was seen mostly in women over 50 years of age. In addition, our population was majority female, so differences by sex may have been harder to discern.

We did not see any genome-wide significant associations with GGT. This is most likely due to the smaller sample size and resulting decreased power. When comparing our results to previously published loci, we found an association between rs944002 in C14orf73 (7, 25) for GGT (P=5.83 × 10−5). C14orf73, also known as EXOC3L4 (exocyst complex component 3 like 4), is expressed in the liver, and could indicate another genetic link to variation in liver enzymes.

In conclusion, we found that SNPs in the PNPLA3 gene, rs4823173, rs2896019, and rs2281135 were significantly associated with AST levels. Although they did not reach the threshold for significance in this study, these same three SNPs in the PNPLA3 gene were the top hits with ALT. The strong LD for these SNPs, as well as with rs738409, indicates a single hit in the PNPLA3 gene. These results demonstrate that PNPLA3, a locus previously identified with ALT, AST, and NAFLD/NASH in other ancestries, is associated with liver function in Hispanics as well.