Association between liver-specific gene polymorphisms and their expression levels with nonalcoholic fatty liver disease

Authors


  • Potential conflict of interest: Nothing to report.

  • Supported by National Health and Medical Research Council (NHMRC) Project Grants 634445, 403981, 572613, NHMRC Program Grant 353514, Canadian Institute of Health Research MOP-82893, Gastroenterology Association of Australia Career Development Award (to L.A.A.).

Abstract

Genetic factors account for a significant proportion of the phenotypic variance of nonalcoholic fatty liver disease (NAFLD); however, very few predisposing genes have been identified. We aimed to (1) identify novel genetic associations with NAFLD by performing a genome-wide association study (GWAS), and (2) examine the biological expression of the strongest genetic associations in a separate cohort. We performed GWAS of a population-based cohort (Raine Study) of 928 adolescents assessed for NAFLD by ultrasound at age 17. Expression of genes with single nucleotide polymorphisms (SNPs) that were associated with NAFLD at a significance level of P < 10−5 was examined in adults with NAFLD and controls by quantifying hepatic messenger RNA (mRNA) expression and serum levels of protein. After adjustment for sex and degree of adiposity, SNPs in two genes expressed in liver were associated with NAFLD adolescents: group-specific component (GC) (odds ratio [OR], 2.54; P = 1.20 × 10−6) and lymphocyte cytosolic protein-1 (LCP1) (OR, 3.29; P = 2.96 × 10−6). SNPs in two genes expressed in neurons were also associated with NAFLD: lipid phosphate phosphatase-related protein type 4 (LPPR4) (OR, 2.30; P = 4.82 × 10−6) and solute carrier family 38 member 8 (SLC38A8) (OR, 3.14; P = 1.86 × 10−6). Hepatic GC mRNA was significantly reduced (by 83%) and LCP1 mRNA was increased (by 300%) in liver biopsy samples from patients with NAFLD compared to controls (P < 0.05). Mean serum levels of GC protein were significantly lower in patients with NAFLD than controls (250 ± 90 versus 298 ± 90, respectively; P = 0.004); GC protein levels decreased with increasing severity of hepatic steatosis (P < 0.01). Conclusion: The association between GC and LCP1 SNPs and NAFLD as well as altered biological expression implicate these genes in the pathogenesis of NAFLD. (HEPATOLOGY 2013;)

The development of nonalcoholic fatty liver disease (NAFLD) is closely related to increasing adiposity and insulin resistance, largely attributed to excess caloric intake and reduced energy expenditure. It is increasingly recognized that there are strong genetic influences in the development of NAFLD. The underlying risk factors for NAFLD of diabetes mellitus and obesity have numerous predisposing genetic polymorphisms.1, 2 Furthermore, the heritability of hepatic steatosis has been estimated at 39% after controlling for factors such as age, gender, race, and body mass index (BMI).3 Identifying genetic associations with NAFLD may offer insight into novel mechanisms of disease pathogenesis, provide new diagnostic tools, and identify new therapeutic targets.

Genome-wide association studies (GWAS) offer a powerful technique to discover novel genetic associations between single nucleotide polymorphisms (SNPs) and disease phenotype. Romeo et al.4 performed a GWAS in the Dallas Heart Study cohort and found a novel triglyceride metabolism gene (PNPLA3) associated with hepatic triglyceride content. Subsequent studies have confirmed the importance of this discovery in determining histological severity in NAFLD as well as in other diseases, including chronic viral hepatitis C and alcoholic liver disease.5–7 The second GWAS of hepatic steatosis performed to date found two additional novel genetic associations (NCAN, PPP1R3B) which were also associated with altered serum lipid profiles.8

To date, no GWAS has examined the genetic associations with NAFLD in adolescents. In contrast to adults with hepatic steatosis, adolescents with NAFLD are at the earliest stage of disease development and potential confounding factors such as alcohol, type 2 diabetes mellitus, and comorbid illness are much less likely to be present than in adulthood. Furthermore, pediatric NAFLD is associated with a different histological phenotype and may be rapidly progressive,9, 10 suggesting childhood and adolescent NAFLD may have unique pathogenic factors or respond differently to metabolic risk factors compared to adults. These observations coupled with a population-based cohort with NAFLD provide a unique opportunity to study genetic variants associated with NAFLD in the absence of many of the established confounders. Identifying genetic risk factors associated with early onset NAFLD may also identify those at greatest risk of future disease morbidity.

The aim of this study was to identify genetic variants associated with adolescent NAFLD. We performed a GWAS in the population-based Western Australian Pregnancy (Raine) Study where detailed liver assessment occurred at age 17. Second, we examined the biological significance of the most significant loci found in the genome-wide screen. To achieve this we examined the phenotypic expression of these loci by investigating hepatic messenger RNA (mRNA) levels in liver biopsies and protein levels in the serum of adult subjects with proven NAFLD compared to non-NAFLD subjects.

Materials and Methods

GWAS Study Population

The Western Australian Pregnancy Cohort (Raine) Study is an ongoing longitudinal pregnancy cohort study of 2,868 live-born children born between 1989 and 1992 in Perth, Western Australia. Recruitment and follow-up of the cohort has been described in detail previously.11, 12 Since birth, the cohort has undergone serial cross-sectional assessments with the current study performed when the cohort had reached late adolescence. The Raine cohort is predominantly of European descent, with both parents indicating Caucasian ethnicity for 82% of participants and is representative of the Western Australia adolescent general population.13

At age 17 years, 1,170 participants underwent assessment including (1) a detailed health questionnaire; (2) anthropometric assessment; (3) abdominal ultrasonography according to a validated protocol14; and (4) fasting serum biochemistry. The diagnosis of NAFLD required ultrasound-confirmed hepatic steatosis and daily alcohol consumption <10 g for females and <20 g for males. Ultrasound was performed by trained sonographers using a Siemens Antares ultrasound machine with a CH 6-2 curved array probe (Sequoia, Siemens Medical Solutions, Mountain View, CA) according to a standardized protocol.14 A single radiologist then interpreted standard images and graded severity of NAFLD based upon scoring of echotexture, deep attenuation, and vessel blurring with a range from zero to six. A score of 0-1 indicated no NAFLD, two mild NAFLD and a score of 3-6 moderate-severe NAFLD. The intraobserver reliability (κ statistic) for fatty liver was excellent at 0.78 (95% confidence interval [CI] 0.73-0.88).15

Of the 1,170 adolescents assessed, 951 met the inclusion criteria for genetic association studies, being unrelated, of European descent, and having DNA available for genotyping. Twenty-three participants were excluded due to excessive alcohol consumption.

The study was conducted with appropriate institutional (King Edward Memorial Hospital and Princess Margaret Hospital for Children) ethics approval. Written informed consent was obtained from the primary caregiver and assent obtained from adolescents.

Genotyping

DNA was extracted from 5-mL samples of ethylenediaminetetraacetic acid (EDTA) anticoagulated blood using a Puregene DNA isolation kit based on a simple salting out technique.16 Genotyping was performed on the Illumina (San Diego, CA) BeadArray Reader at the Centre for Applied Genomics (Toronto, Ontario, Canada) using 250 ng of DNA and the Illumina Human660-W Quad Array. Further genotype imputation was performed using MACH v. 1.0.1617, 18 against a reference of the CEU samples of HapMap phase 2, build 36, release 22. A total of 97,718 SNPs were removed at genotyping QC because they had a call rate of <95%. A further 919 SNPs were removed with Hardy-Weinberg P-values <5.7 × 10−7. SNPs with a minor allele frequency less than 5% (35,7502 SNPs) were excluded due to lack of power to detect reasonable effect sizes in association with less common SNPs. Finally, SNPs with imputation quality less than 0.80 (10,7880 SNPs) were removed from analysis. A total of 2,078,505 SNPs were included in the final analysis. Further details of the quality control procedures are included in Supporting Table 1.

Table 1. Demographic and Biochemical Phenotype of NAFLD and Non-NAFLD Adolescents in the Raine Cohort
CharacteristicNon-NAFLD(n=802)NAFLD(n=126)P
  1. Parametric variables are presented as mean (standard deviation), nonparametric variables are presented as median (interquartile range) and binomial variables are presented as number (percentage). *Tanner stage not available on 8 subjects.

Female (%)368 (45.9%)76 (60.3%)0.004
Age (years)17.0 (0.25)17.0 (0.23)0.26
Tanner stage*2/3/4/5, n (%)5(1%)/13(2%)/215 (27%)/565 (71%)1(1%)/1(1%)/31 (25%)/89(73%)0.8
Body mass index (kg/m2)21.7 (19.9, 23.7)27.5 (22.9, 32.4)<0.001
z-BMI0.13 (0.91)1.22 (0.97)<0.001
Suprailiac SFT (mm)12.3 (8.0, 18.3)54.4 (17.3, 33.7)<0.001
Waist circumference (cm)76.2 (71.6, 81.4)89.3 (78.0, 103.0)<0.001
ALT (IU/l)18 (14, 24)21 (15, 37)<0.001
GGT (IU/l)13 (10, 16)15 (11, 22)<0.001
AST (IU/l)23 (20, 27)23 (19, 28)0.49
Cholesterol (mg/dl)158 (28)162 (36)0.32
Triglyceride (mg/dl)80 (62, 104)94 (67, 131)0.004
HDL-cholesterol (mg/dl)51 (12)46 (11)<0.001
Fasting insulin (μU/l)7.2 (4.8, 10.3)10.5 (7.1, 16.0)<0.001
Fasting glucose (mg/dl)85 (81, 90)85 (81, 90)0.10
HOMA-IR1.50 (0.98, 2.18)2.27 (1.39, 3.46)<0.001
C-reactive protein (mg/l)0.05 (0.02, 0.13)0.14 (0.05, 0.35)<0.001
Leptin (ng/ml)7.9 (2.0, 21.8)29.0 (11.8, 56.8)<0.001
Adiponectin (mg/ml)9.0 (6.3, 12.3)7.4 (5.1, 10.6)<0.001

Biological Validation

To investigate the biological validity of the most significant loci from the GWAS, we examined the hepatic gene expression of the two loci whose respective genes are predominately expressed in liver: (1) group-specific component (GC) that encodes vitamin D binding protein (VDBP), and (2) lymphocyte cytosolic protein-1 (LCP1) gene. Hepatic gene expression was examined in human adult subjects with biopsy-proven NAFLD and compared to biopsies from control subjects without NAFLD. To benchmark the biological validation in human samples, hepatic gene expression for PNPLA3 was also examined, as genetic variants from PNPLA3 have consistently been associated with hepatic steatosis in adults.4, 8 Lastly, we examined the protein expression of VDBP in human subjects with biopsy-proven NAFLD compared to healthy controls.

Hepatic Gene Expression Methods.

Liver biopsy specimens from adults with NAFLD undergoing liver biopsy as part of routine clinical practice (n = 9) or subjects undergoing bariatric surgery (n = 4) were stored in RNAlater (Ambion, Austin, TX) at 4°C for 12 hours before being frozen at −80°C. Patients undergoing bariatric surgery had liver biopsies using a 16-G core biopsy needle (Bard Biopsy Systems, Tempe, AZ). Subjects with liver biopsies demonstrating <5% hepatic steatosis were labeled controls (n = 8).

NAFLD and control liver biopsies were homogenized and total RNA was extracted with TRI Reagent (Ambion) according to the manufacturer's protocols. First-strand cDNA synthesis was performed from 70 ng total RNA using the Quantitect Reverse Transcription kit (Qiagen, Valencia, CA). Primers were designed using NCBI Primer Design and Primer3 (detailed in Supporting Table 2). Primer efficiency was between 97%-99%. Real-time polymerase chain reaction (PCR) was performed in a Rotorgene 3000 cycler (Qiagen, Australia) in a 10-μL reaction. Each 10-μL PCR reaction contained 2 μL of complementary DNA (cDNA) template, 1 μL of 10× ImmoBuffer, 0.05 μL IMMOLASE DNA Polymerase (Bioline, Alexandria, NSW, Australia), 2 mM MgCl2, 0.2 μL of 10 mM dNTP mix, 0.3 μM of each primer, 0.1 μL of 20× EvaGreen dye (Biotium, Hayward, CA), and deionized sterile water. Gene expression data were normalized to three stably expressed endogenous reference genes: hypoxanthine phosphoribosyltransferase-1 (HPRT1), succinate dehydrogenase complex subunit-A flavoprotein (SDHA), and peptidylprolyl isomerase-A (PPIA).19 Single product amplification was confirmed by melt curve analysis.

Table 2. Leading SNPs Associated With Adolescent NAFLD Identified by GWAS
SNPChromosomePosition (Mb)LocationGeneMajor/ Minor AllelesMAFOdds RatioStandard ErrorP
  1. MAF, minor allele frequency; LPPR4, lipid phosphate phosphatase-related protein type 4; GC, genome-specific component (encodes vitamin D binding protein); LCP1, lymphocyte cytosolic protein-1; SLC38A8, solute carrier family 38 member 8.

rs12743824199.6IntergenicLPPR4A/C0.4412.301.204.82x10-6
rs222054472.8IntergenicGCG/C0.3012.541.211.20x10-6
rs73248451345.6IntronLCP1G/A0.0963.291.292.96x10-6
rs118641461682.6IntronSLC38A8G/A0.1003.141.271.86x10-6

Human VDBP Serum Levels.

Subjects with biopsy-proven NAFLD (n = 74) and healthy controls (n = 19) had fasting serum samples stored at −80°C. Subjects with NAFLD underwent liver biopsy as part of routine clinical management or at the time of bariatric surgery. Normal controls were healthy volunteers recruited from the community with normal serum liver biochemistry and hepatic ultrasound examinations. Serum was analyzed for VDBP levels using an enzyme-linked immunosorbent assay (ELISA; R&D Systems, Minneapolis, MN). The severity of hepatic steatosis on liver biopsy samples was graded 1-3, inflammation was graded 0-2, and fibrosis was staged 0-4.20 Normal controls were graded as zero.

All adult subjects donating either liver tissue or blood for characterization of gene expression had provided informed written consent for use of their biological samples for the purposes of research with approval by the Human Research Ethics Committee of Sir Charles Gairdner Hospital.

Statistical Analysis

Univariate comparisons of continuous demographic and biochemical variables with NAFLD status in the Raine Cohort were compared using Student's t test or Welch's one-way test if normally distributed and a Kruskal-Wallis test or the Wilcoxon rank sum test if the distribution was skewed. The association of binary variables with NAFLD status were assessed using the chi-square test or Fisher's exact test as appropriate for group sizes.

Multivariate models for risk of NAFLD were selected by: (1) identifying adiposity, blood pressure, liver-function, lipid-metabolism, and insulin-metabolism factors that were associated with NAFLD in univariate analyses (significance level 0.05); (2) using the backwards elimination technique (stepwise regression based on the likelihood ratio test) on the multivariate model which included factors identified in the previous step. Suprailiac skinfold thickness (SSF) had the strongest association with NAFLD risk compared to all other measures of adiposity. SSF was highly predictive of the risk of NAFLD in both males and females, such that no associations were detected between NAFLD and any other factors after adjustment for SSF. Adiposity factors were BMI, waist circumference, hip circumference, ratio of waist to hip circumference, weight and skinfold thicknesses for suprailiac, subscapular, tricep, and abdomen. Blood pressure factors were systolic and diastolic blood pressure and pulse rate. Liver function factors were bilirubin, γ-glutamyltransferase (GGT), alanine aminotransferase (ALT), aspartate aminotransferase (AST), and alkaline phosphatase (ALP). Lipid-metabolism factors were high-density lipoprotein (HDL), LDL, and triglycerides. Insulin-metabolic factors were fasting glucose, insulin, homeostasis model assessment of insulin resistance (HOMA-IR), and adiponectin.

Multivariate associations between each SNP and the dichotomous NAFLD diagnosis were assessed using logistic regression (under an additive genetic model), adjusting for adiposity (SSF), gender, and the first two principal components for ethnicity. The genome-wide analysis was performed using the MACH v. 1.0.16 software.17, 18 Results were quality checked for formatting errors, duplicate SNPs, and for the range (validity) of odds ratios, standard errors, and P-values. SNPs that were available for less than half of the analysis population were excluded. SNPs that reached a P-value threshold of less than 1 × 10−5 in the GWAS screen (n = 12 SNPs, representing four distinct genomic regions on four separate chromosomes) were considered for biological validation. Associations between the 12 most significant SNPs and the ultrasound-assessed severity scores for NAFLD were also assessed using Fisher's exact test.

Human hepatic gene expression of GC, LCP1, and PNPLA3 were analyzed as fold-change in mRNA expression levels of the target gene in the NAFLD group relative to expression in the control group. Hepatic gene expression data are presented graphically as mean and error normalized to housekeeping genes using the comparative Ct method. Human serum VDBP levels were compared between NAFLD and non-NAFLD subjects using Student's t test. The association between human serum VDBP levels and liver steatosis severity was assessed using analysis of variance with a contrast for a linear trend to account for the ordinal nature of the severity score. Two-tailed P-values were considered significant if <0.05.

Abbreviations

BMI, body mass index; C-RP, c-reactive protein; GC, group-specific component; GWAS, genome wide association study; HDL, high-density lipoprotein; HPRT1, hypoxanthine phosphoribosyltransferase 1; LCP-1, lymphocyte cytosolic protein-1; LPPR4, lipid phosphate phosphatase-related protein type 4 gene; NAFLD, nonalcoholic fatty liver disease; PNPLA3, patatin-like phospholipase domain containing 3; PPIA, peptidylprolyl isomerase A; SDHA, succinate dehydrogenase complex, subunit A flavoprotein; SLC38A8, solute carrier family 38 member 8; SNP, single nucleotide polymorphism; SST, suprailiac skin fold thickness; VDBP, vitamin D binding protein.

Results

Raine Cohort: NAFLD Phenotype Prevalence and Characteristics

The overall prevalence of NAFLD in the Raine Cohort was 13.7%, (126/928). NAFLD was significantly more common in females than in males (17.1% versus 11.3%, P = 0.004) and was associated with higher levels of adiposity, insulin resistance, serum triglyceride, and hepatic amino-transaminases and lower HDL-cholesterol levels (Table 1). In multivariate analysis suprailiac skin fold thickness (SST) was the major predictor of NAFLD independent of other metabolic or clinical factors. For a 1 mm increase in SST, there was a 15% increase in the risk of NAFFLD (odds ratio [OR] 1.15, 95% CI 1.12-1.18 mm, P < 2.2 × 10−16).

Genome-Wide Associations With Adolescent NAFLD

The most significant associations from the genome-wide analysis of NAFLD (Fig. 1, Table 2) were: rs12743824 near the lipid phosphate phosphatase-related protein type-4 gene (LPPR4) on chromosome 1 (P = 4.825 × 10−6); rs222054 in the group-specific component gene (GC) on chromosome 4 (P = 1.205 × 10−6); rs7324845 in the lymphocyte cytosolic protein-1 (LCP1) gene on chromosome 13 (P = 2.962 × 10−6); and rs11864146 in the solute carrier family-38 member 8 (SLC38A8) gene on chromosome 16 (P = 1.858 × 10−6). The association between each SNP and NAFLD was not significantly effected after adjustment for adiponectin or insulin resistance (HOMA-IR). The LPPR4 protein is involved in intercellular signaling in neuronal tissue. The SLC38A8 protein is a putative sodium-coupled neutral amino acid transporter. Neither protein is expressed in the liver nor adipose tissue, thus they were not able to be pursued for biological validation. Both GC (which encodes the VDBP) and LCP1 are expressed in hepatic tissue and are biologically plausible candidates for a role in NAFLD pathogenesis. These loci were pursued for biological validation of the observed NAFLD genomic association.

Figure 1.

(A) Manhattan plot of genome-wide associations. Each point represents the P-value for the association between an SNP and NAFLD. The “peaks” represent the leading associations between polymorphic loci and NAFLD: for chromosome 1 representing the LPPR4 gene locus (tagged by SNP rs12743824); chromosome 4, GC (rs222054); chromosome 13, LCP1 (rs7324845); and chromosome 16, SLC38A8 (rs11864146). (B,C) Regional association plot of GC (B) and LCP1 (C) regions. Each point represents a SNP. The left y-axis represents the P-value for the association between the SNP and NAFLD. The x-axis represents the position of the SNP on its respective chromosome, with annotations of genes within the region. The right y-axis represents the rate of chromosomal recombination at that position. The color of the points represents the strength of linkage disequilibrium between the SNP and the strongest signal in the region, which is depicted as a purple diamond (rs222054, P = 1.205 × 10−6 for GC and rs7324845, P = 2.962 × 10−6 for LCP1). (B,C) Generated at LocusZoom (https://statgen.sph.umich.edu/locuszoom).

rs222054 in the GC Gene.

Possession of at least one copy of the variant C allele of SNP rs222054 was associated with a 2.54-fold increase (SE = 1.21) in risk of NAFLD compared to the wild GG genotype (Table 2). The prevalence of NAFLD increased from 9.7% in wildtype homozygotes, to 16.8% in heterozygotes, and to 21.4% (P = 0.002) in those with the homozygous CC (variant) genotype (Fig. 2A, Supporting Table 3). After stratification by sonographic severity of hepatic steatosis, moderate to severe hepatic steatosis increased significantly with each C allele (P = 0.003, Supporting Table 3, Fig. 2B). This locus was not significantly associated with measures of adiposity, serum lipid levels, insulin resistance, or adipocytokines (Supporting Table 3).

Figure 2.

Prevalence of NAFLD increases significantly in the presence of SNPs in the GC and LCP1 genes in the Raine Cohort (n = 928). Genotypes of SNP rs222054 in the GC gene are associated with an increase in the prevalence of NAFLD (9.7% versus 16.8% versus 21.4%, *P = 0.002) and severe NAFLD (6.1% versus 10.8% versus 16.7%, **P = 0.003) (A,B). Similarly, genotypes of SNP rs7324845 in the LCP1 gene are associated with an increase in the prevalence of NAFLD (11.6% versus 21.4% versus 28.6%, *P = 0.002) and severe NAFLD (7.1% versus 15.3% versus 28.6%, **P = 0.004) (C,D). Associations analyzed by Fisher's exact test. Numbers within the bars represent the subject number per genotype.

Table 3. Demographic, Metabolic, and Histological Characteristics of NAFLD and Normal Healthy Control Adult Subjects Tested for Serum VDBP Levels
CharacteristicNAFLD(n=74)Normal Controls(n=19)P
  1. Parametric variables are presented as mean (standard deviation), nonparametric variables are presented as median (interquartile range) and binomial variables are presented as number (percentage). Histology scored according to Kleiner et al. (20) Steatosis grade in normal controls coded as zero based upon normal ultrasound and normal liver enzymes.

Female (%)48 (65%)13 (68%)0.771
Age (years)54.5 (44, 60)51.0 (32, 58)0.136
Body mass index (kg/m2)38.5 (9.2)26.0 (3.3)<0.001
Waist circumference (cm)118.8 (16.5)88.4 (10.1)<0.001
ALT (IU/l)49.5 (28, 76)21 (18, 28)<0.001
GGT (IU/l)61 (34, 132)16 (12, 22)<0.001
AST (IU/l)34.5 (26, 53)26 (19, 31)0.002
Triglyceride (mg/dl)162 (92)89 (30)0.001
HDL-cholesterol (mg/dl)51 (57)59 (12)0.575
Fasting insulin (μU/l)15.0 (10.0, 28.5)5.5 (4.2, 9.8)<0.001
Fasting glucose (mg/dl)102 (86, 120)91 (90, 97)0.024
HOMA-IR3.6 (2.1, 8.4)1.3 (0.9, 2.0)<0.001
Steatosis grade (0,1,2,3)0/21/31/2219/0/0/0<0.001
Inflammation grade (0,1,2)26/40/8NA
Fibrosis stage (0,1,2,3,4)35/15/3/12/8NA

rs7324845 in the LCP1 Gene.

Possession of at least one copy of the variant A allele of SNP rs7324845 was associated with a 3.29-fold (SE = 1.29) increase in risk of NAFLD compared to the wildtype GG genotype (P = 2.962 × 10−6, Table 2). The prevalence of NAFLD increased from 11.2% in wildtype homozygotes to 21.4% in heterozygotes and 28.6% in variant allele homozygotes (P = 0.002, Fig. 2C, Supporting Table 4). There was also a significant association between the number of minor allele copies for rs7324845 and moderate to severe hepatic steatosis as determined by ultrasound (P = 0.004, Fig. 2D). The rs7324845 SNP in LCP1 was also associated with a reduction in fasting insulin and HOMA-IR (P = 0.042 and P = 0.040, respectively) but not with any of the other markers of metabolic disease.

Association Between PNPLA3 and Other SNPs With NAFLD.

We assessed the associations of NAFLD in our adolescent cohort with the previously reported risk variants at the PNPLA3 (rs738409), NCAN (rs2228603), and PPP1R3B (rs4240624) loci.4, 8 The PNPLA3 polymorphism rs738409 (P = 0.010) and an “rs738409 × BMI” interaction term (P = 0.01), were significantly associated with NAFLD among obese individuals, in keeping with recent reports confirming an interaction between adiposity and PNPLA3 in children and adults.21, 22 There were no significant associations with NAFLD for any of the other candidate loci in the cohort.

GC and LCP1 Expression in Human NAFLD

To examine the biological significance of the GC and LCP1 genes in human NAFLD, we quantified hepatic mRNA levels of these genes, as well as PNPLA3, in liver biopsies from subjects with NAFLD (n = 13) and controls (n = 8) (Fig. 3). Hepatic GC mRNA was down-regulated by 83% in NAFLD subjects compared to controls (P = 0.026). In contrast, LCP1 and PNPLA3 gene expression were significantly increased (300%, P = 0.002 and 161%, P = 0.02, respectively).

Figure 3.

Hepatic gene expression of GC, LCP1, and PNPLA3 is significantly different in adult NAFLD subjects versus controls. Hepatic expression of GC mRNA is significantly reduced in human NAFLD versus controls (A), whereas LCP1 and PNPLA3 mRNA is significantly increased (B,C). Gene expression quantified by rt-PCR of liver biopsy specimens. Gene expression is normalized to housekeeping genes and presented as mean (95% CI bars). *P < 0.05. **P < 0.01.

To further examine the significance of GC in NAFLD, we compared the serum levels of VDBP in humans with biopsy-proven NAFLD (n = 74) to healthy controls without NAFLD (n = 19). NAFLD subjects and healthy controls were of similar age and gender; however, those with NAFLD had higher measures of adiposity, insulin resistance, serum triglyceride, and liver enzyme levels (Table 4). Subjects with NAFLD also had significantly lower serum VDBP levels compared to controls (250 ng/mL [SD 90] versus 298 ng/mL [SD 90], P = 0.04). Furthermore, serum VDBP levels were reduced significantly as the severity of steatosis increased (P = 0.012, Fig. 4). Serum VDBP levels did not correlate significantly with age, BMI, HOMA-IR, fasting triglycerides, or HDL-cholesterol, degree of fibrosis, or inflammation on liver biopsy (P > 0.1 for all comparisons, data not shown).

Figure 4.

Serum levels of VDBP are reduced in adults with NAFLD (n = 74) versus healthy controls (n = 19). Steatosis graded in NAFLD subjects according to Kleiner et al.,20 with healthy controls being graded as zero. Data presented as mean ± SEM. **P < 0.01 using a linear trend test.

Discussion

We undertook for the first time a GWAS in a well-characterized Caucasian adolescent cohort in order to identify potential genetic determinants of ultrasound-defined NAFLD. We found associations between NAFLD and SNPs in four novel genes. Two of the SNPs (rs222054, rs7324845) involved genes that are expressed in the liver (GC gene and LCP1 gene, respectively). To determine whether the most significant genetic associations from the GWAS were biologically plausible, we examined their expression in human adults with NAFLD and found the hepatic expression of both genes was significantly altered when compared with healthy controls. Furthermore, the systemic protein expression of the GC gene, reflected by serum levels of VDBP, was significantly lower in adult subjects with NAFLD compared with controls. Serum VDBP levels reduced as the severity of hepatic steatosis increased.

GC gene expression is predominately in the hepatocytes where it codes for VDBP.23 VDBP is the main carrier protein for vitamin D, which has been implicated in the development of obesity and diabetes. Specifically, low vitamin D levels may increase adipocyte intracellular calcium, stimulating lipogenesis, whereas vitamin D supplementation improves insulin resistance and down-regulates inflammatory cytokines such as tumor necrosis factor-α and interleukin-6 in cell models.24, 25 Epidemiological studies have demonstrated vitamin D status is inversely associated with obesity,26 insulin resistance,27 and risk of incident type 2 diabetes mellitus and hyperglycemia.28, 29 In addition, vitamin D levels are low in NAFLD patients compared to matched controls and were associated with histological severity of disease.30 Vitamin D levels are influenced by GC gene polymorphisms,31 which have been associated with glucose intolerance in Pima Indians and hyperinsulinemia and type 2 diabetes in Japanese.32, 33 Thus, it is conceivable that the novel rs222054 SNP described in this cohort may be a marker of altered pathways involved in adipogenesis or insulin resistance, thereby predisposing to NAFLD.

LCP-1 is an actin bundling protein expressed predominately in hematopoietic cells and is involved in leukocyte activation and tumor cell proliferation.34 Interestingly, a polymorphism in the LCP1 gene has recently been associated with chemotherapy-induced hypertriglyceridemia, suggesting it may have a role in lipid metabolism.35 The SLC38A8 protein product is a putative sodium-coupled neutral amino acid transporter whose expression is limited to the brain, whereas LPPR4 catalyzes the dephosphorylation of biologically active lipids and is expressed predominately in the hypothalamus.36 While the functional significance of neuronally expressed genes such as SLC38A8 and LPPR4 with NAFLD is not apparent, there is convincing evidence that the nervous system and particularly the hypothalamus play an important role in lipid homeostasis in the liver.37 Neuronally expressed genes have been associated with NAFLD and plasma lipid levels in a previous GWAS study highlighting the need for functional studies to expand our understanding of these genes.8, 38 Alternatively, it may be that we have identified loci in linkage disequilibrium with other genes that are associated with hepatic lipid metabolism. We also cannot exclude that these associations may be due to type 1 error.

A limitation of our study is the modest size of the cohort. Although the Raine Cohort is currently the largest adolescent population-based cohort with ultrasound assessment for NAFLD, our sample size is modest with concomitant modest statistical power to detect small effect sizes. Furthermore, although we utilized a validated and standardized imaging protocol for the diagnosis of NAFLD, ultrasound is less sensitive for the detection of minor hepatic steatosis and it is possible some cases of NAFLD were not diagnosed. Our sample was also homogenous and overwhelmingly European in ancestry. Therefore, our genotypic associations require replication in other adolescent and adult cohorts and also in other ethnic groups to confirm their potential significance. The higher prevalence of NAFLD in female adolescents in our cohort is seemingly at odds with some other population-based studies.39–41 Notably in our cohort, the sex-related difference in prevalence was eliminated after adjusting for suprailiac skin fold thickness. The male predominance of fatty liver in other adolescent cohorts, however, may be due to previous studies relying on nonsex-specific ALT cutoffs to diagnose NAFLD, which would underestimate the prevalence in females, or failing to exclude subjects with significant alcohol intake, who were more likely to be males.

We confirmed the association between PNPLA3 gene polymorphism rs738409 and NAFLD for the first time in a population-based sample of adolescents including a full range of normal weight and obese individuals. Similar to recent reports, we found a significant interaction between rs738409 and BMI, suggesting the impact of this polymorphism is increased with obesity.21, 22 Functional studies are required to clarify the mechanisms leading to this observation.

We were unable to replicate some previous associations from an earlier GWAS study in an adult population.8 This may be due lack of power in our cohort, the previous associations were false-positive results, or that genetic associations differ between adolescent and adult phenotypes of NAFLD. Although adolescents and adults share common pathogenic factors of obesity and insulin resistance, it is important to note that the typical histological patterns of type 1 (adult) and type 2 (pediatric) NASH differ. Furthermore, chronic portal inflammation, which is the hallmark of type 2 NASH, is associated with adiposity and insulin resistance in adults, but not children,9 suggesting alternative pathogenic factors may be important in pediatric NAFLD. Therefore, it is possible that genetic associations with NAFLD may also differ between adults and children.

In summary, using a two-tiered genetic epidemiological and functional biology approach, we identified two novel pathogenic genes involved in NAFLD. First, we explored genetic polymorphisms associated with NAFLD in adolescents using a genome-wide approach. Second, the expression of the biologically plausible loci (GC and LCP1) was subsequently examined in adult human NAFLD compared to control subjects. In addition to demonstrating altered hepatic gene expression of GC and LCP1 in NAFLD compared to controls, this is the first study demonstrating increased hepatic PNPLA3 gene expression in human NAFLD compared to controls. Further studies are required to explore how GC and LCP1 predispose to NAFLD; in particular, whether manipulation of vitamin D metabolism may influence the presence or severity of disease.

Ancillary