Admixture mapping in African Americans identifies new risk loci for HCV-related cirrhosis

Abstract

Background & Aims:

Cirrhosis is the main predisposing condition for hepatocellular carcinoma. Host genetic risk factors have been reported for cirrhosis; however, whether there is a genetic contribution to racial disparities in cirrhosis requires further investigation.

Methods:

We used an affected-only Mapping by Admixture Linkage Disequilibrium analysis to characterize the genetic risk of cirrhosis in 227 African American patients with cirrhosis genotyped at 19,804 ancestry-informative marker (AIM) single nucleotide polymorphisms (SNPs). We additionally performed analyses stratified by hepatitis C virus (HCV) infection status. To replicate our findings, we conducted a case-control analysis in an external study population (452 cases and 196 controls).

Results:

Mean age of patients was 63.3 years and 98.2% were male. Risk factors for cirrhosis included HCV infection (83.7%) and alcohol abuse (56.4%). In the admixture mapping analysis, we found that European ancestry on chromosome 2q21.1 and African ancestry on chromosome 6p21.2 were associated with increased risk of cirrhosis in African Americans. In the fine-mapping analysis, we identified regions near POTEKP on 2q21.1 (p=0.0001) and DNAH8 on 6p21.2 (p=0.0017) that were associated with cirrhosis. As the admixture peaks in the HCV-positive patients were the same as those in the overall group, findings in the analysis are reflective of the HCV-positive group. In the replication analysis, the results on chromosome 2 were not significant after adjusting for multiple comparisons, and we could not replicate the results on chromosome 6.

Conclusions:

We used admixture mapping to identify novel genomic regions on 2q21.1 and 6p21.2 that may be associated with HCV-related cirrhosis risk in African Americans.

INTRODUCTION

Cirrhosis, an advanced stage liver disease caused by many diseases, is a common chronic condition that carries high morbidity and mortality, and is also the main predisposing condition for hepatocellular carcinoma (HCC), a rapidly increasing, highly fatal cancer.^{1, 2} The burden of HCC in the United States as measured by age-adjusted mortality rates is approximately 2-fold higher among African Americans compared to non-Hispanic whites (NHW).³ However, the risk of developing cirrhosis and cirrhosis-related mortality among people with cirrhosis risk factors (e.g., hepatitis C virus [HCV] infection, non-alcoholic fatty liver disease [NAFLD]) are lower among African Americans compared with NHWs.^{4, 5} The underlying reasons for these racial/ethnic variations in risks are not completely understood.

Recent genome-wide association studies (GWAS) have identified genetic variants for cirrhosis.^6–11 However, these studies were among predominantly Caucasian populations and it is unclear whether there is a genetic contribution to racial disparities in cirrhosis.

Admixture mapping is a statistically efficient method for studying genetic contributions to ethnic/racial disparities in a disease risk.^{12, 13} Admixture mapping methods leverage ancestry-informative markers (AIMs), which are autosomal single nucleotide polymorphisms (SNPs) with differing allele frequencies between ancestral groups. Admixture mapping is a method for discovery of genomic regions that is based on linkage disequilibrium in populations with recent population admixture and enables detection of loci associated with phenotypic variation using far fewer samples and markers than GWAS.^{14, 15} Additionally, unlike GWAS which are association analyses of cases and controls within the same ancestral group, admixture mapping does not require control subjects because it compares the proportion of ancestry at each locus to the average genome-wide ancestry using only affected persons (cases). Comparisons of admixture mapping and GWAS have demonstrated that each approach can provide unique information, and that ancestry can be more informative than GWAS when the causal variant has large allele frequency differences between ancestral populations.^{16, 17} Admixture mapping studies have successfully identified novel genomic regions associated with complex traits in admixed populations, including cancers with strong racial/ethnic disparities in incidence and mortality rates.^18–20

Genetic admixture coupled with racial differences in cirrhosis risk provides excellent rationale for performing admixture mapping in African Americans with cirrhosis. We therefore performed a genome-wide admixture mapping study in germline DNA from a cohort of genetically classified African American Veterans with cirrhosis to help characterize genetic risk of cirrhosis.

METHODS

Study Population

We used baseline data and blood samples from patients with cirrhosis recruited into the Veterans Affairs Cirrhosis Surveillance Cohort, a prospective cohort study conducted at the Michael E. DeBakey Veterans Affairs Medical Center in Houston, Texas between August 2014 and December 2016. Cirrhosis diagnosis was based on histology, radiology, liver elastography, or serum biomarkers.²¹

Genotyping and Quality Control

Germline DNA was extracted from blood samples using a commercially available Macherey-Nagel Nucleospin Blood Kit (Macherey-Nagel Inc., PA, USA) and performed accordingly to the manufacturer’s instructions. Genotyping was performed using the Axiom Precision Medicine Research Array by Affymetrix (Affymetrix, CA, USA), which queried 849,552 common tagging SNPs, and captured common genetic variation in populations of mixed continental ancestry. The CEL files were processed using the Axiom Analysis Suite v4.0.3.3. We excluded 22,152 SNPs that were not in Hardy-Weinberg equilibrium, as indicated by a P-value of <1×10⁻⁶, or with high missing rate (i.e., per sample missingness >5%).²²

Specification of the AIMs SNP Panel

For the 827,400 SNPs that passed initial quality control, we performed linkage disequilibrium (LD) pruning with r-squared threshold 0.2 to restrict analysis to markers that were weakly correlated with each other. After pruning, 178,650 SNPs were retained. We further reduced the number of SNPs by requiring a Fixation index (F_ST)>0.1 (to identify SNPs that differentiate well between populations as ancestry informative markers),²³ calculated using overlapped SNPs with the Northern European (the Utah Residents with Northern and Western European Ancestry: CEU) (n=99) and African (Yoruba in Ibadan from Nigeria: YRI) (n=108) samples from the 1000 Genomes Project.²⁴ This resulted in 19,804 AIM SNPs that were used in the admixture mapping analyses presented here.

Statistical Analysis

Admixture mapping was conducted using the program ADMIXMAP to compare observed versus expected European and African ancestry allele copies across the genome conditional on a priori ancestral allele frequencies.²⁵ ADMIXMAP fits a Bayesian probability model with a computationally intensive Markov Chain Monte Carlo parameter estimation algorithm. In our analysis, ADMIXMAP was run for 1100 iterations of the Markov chain, including 100 for burn-in.²⁵ We supplied to ADMIXMAP the prior allele frequencies for each locus for the CEU and YRI sub-populations using data from the third phase of the International HapMap project.²⁶ Principal component analysis (PCA) was performed on our cohort to evaluate the ancestry inference and we compared it to self-reported race/ethnicity to ensure that our study included only genetically African ancestry patients.²⁷

The admixture mapping analysis was conducted in two steps. First, admixture mapping analysis was conducted using ADMIXMAP software to identify regions that suggested an association between genetic ancestry and cirrhosis. Consistent with prior admixture mapping studies among African Americans^{19, 20, 28, 29}, regions of the genome with an absolute value Z-score of 3.00 or greater (corresponding to a p-value ~0.0025) were considered statistically significant and indicated an ancestry-dependent association. In ADMIXMAP, the direction of the Z-score indicates the ancestry that confers the risk allele in African Americans. Specifically, a negative value suggests that excess European ancestry at that locus increases the risk of cirrhosis, while a positive value suggests that African ancestry at the locus increases cirrhosis risk. The admixture mapping results from ADMIXMAP are presented as a series of admixture graphs, first in the overall sample and then by HCV status. The line indicating absolute Z-score >3.0 is included to identify regions that were suggestive of admixture association.

In the second step of the analysis, the genomic regions surrounding the most extreme Z-scores were fine mapped using ADMIXMAP to identify specific variants that are associated with increased risk of cirrhosis. The fine-mapping results are presented for the variants that were tested in the regions suggesting admixture evidence; these variants included genes within 10kb of the peak SNP, as determined by gene annotation using the microarray annotation files.³⁰

Sample size calculation in admixture mapping depends upon the difference in allele frequencies between populations, the risk effect of the alleles, and the desired size of the interval to be mapped.³¹ With 227 African American cirrhosis patients and allele frequency between African and European descent populations of 0.8 and 0.2 and a relative risk per allele of 2.0 to increase cirrhosis risk, a sample size of 227 gives 80% power to narrow the region(s) containing disease-causing loci to less than 10 million base pairs.³² ADMIXMAP software version 3.8.3103³² was used for Admixture mapping and statistical analyses were conducted using R version 3.2.2.³³

Replication analysis

To confirm our findings in admixture mapping, we also genotyped additional 452 African Americans with cirrhosis from the Texas Hepatocellular Carcinoma Consortium³⁴ and 107 controls at MEDVAMC using the same Axiom Precision Medicine Research Array by Affymetrix. In this case-control study, we conducted logistic regression analyses adjusted for age, sex, and mean global African ancestry to calculate odds ratio for associations with cirrhosis risk using the significant genetic loci identified in the admixture mapping analysis.

RESULTS

After PCA, we included 227 genetically-defined African American patients with cirrhosis in the analysis (Table 1; Supplementary Figure 1); 99.6% of whom were self-reported African American (1 patient self-identified as Hispanic). The mean global African ancestry was 81.1%, which is consistent with other studies of African Americans^18–20 (Supplementary Figure 2). The mean age was 63.3 years (standard deviation, 5.5 years). The vast majority of patients were male (98.2%), and 37.9% were obese at study enrolment. Risk factors, in order of frequency, were HCV infection (83.7%), alcohol abuse (56.4%), NAFLD/NASH (4%), hepatitis B virus (HBV) infection (0.9%), and primary biliary cholangitis (0.4%).

Table 1.

Select characteristics of 227 African American cirrhosis patients included in the admixture mapping analysis.

Variables	N (%)
Age, years
Mean (SD)	63.3 (5.5)
Sex
Male	223 (98.2)
Female	4 (1.8)
Body mass index, kg/m2
<30	141 (62.1)
≥30	86 (37.9)
Risk factor for cirrhosis*
HBV	2 (0.9)
HCV	190 (83.7)
Co-HBV and HCV infection	4 (1.8)
Alcohol abuse	128 (56.4)
NAFLD/NASH	9 (4.0)
Primary biliary cholangitis	1 (0.4)
Unclassified	2 (0.9)

^*Multiple etiologies are possible.

In the overall admixture mapping analysis, we identified two peaks that suggested association between ancestry and risk of cirrhosis (Figure 1A). The admixture regions at 2q21.1 (127905840–137554051 bp) and 6p21.2 (35465786–39174038 bp) were significantly associated with increased risk of cirrhosis. Among African Americans, European ancestry at the 2q21.1 locus was associated with increased risk of cirrhosis, and African ancestry at the 6p21.2 locus was associated with increased risk of cirrhosis.

Figure 1.

Results of African American Admixture Mapping. On the X-axis are genomic positions by chromosome. On the Y-axis are Z-scores for the association between locus-specific ancestry and cirrhosis risk. A Z-score >3 implies that the loci have statistically significant increase in African ancestry and a Z-score <−3 implies that the loci have statistically significant increase in European ancestry. 1A: Results for all sample (n=227), 1B: Results for HCV positive patients (n=194), 1C: Results for HCV negative patients (n=33).

To further localize the variants of interest, we conducted fine mapping analyses in the two admixture regions (Supplementary Figure 3) and the fine-mapping results are presented in Supplementary Table 1 . In the admixture region on chromosome 2, the most significant locus (132358911–132585990 bp) is a region near pseudogene POTEKP (POTE Ankyrin Domain Family K) (Z = −3.792, p = 0.000149). In the admixture region on chromosome 6, one of the most significant risk genes is DNAH8 (Z = 3.141, p = 0.00168).

Stratified admixture mapping analysis

In the stratified admixture mapping analyses, subjects were analyzed by HCV infection status. Among cirrhosis patients with HCV infection (Figure 1B), the regions that were suggestive of admixture association were the same as those for the overall analysis; that is, European ancestry at locus 2q21.1 and African ancestry at locus 6q21.2 conferred increased risk of cirrhosis in African Americans with HCV infection. In the fine-mapping analysis of the HCV infection positive group, the most significant results were noted for POTEKP on chromosome 2 (p = 0.000027) and DNAH8 on chromosome 6 (p = 0.00058), as was found in the overall fine-mapping analysis (Supplementary Table 2). Interestingly, the cirrhosis risk loci were not the same for African Americans without HCV infection. African ancestry at genomic regions at 7p14.1 and 18p11.31 conferred increased risk of cirrhosis in African Americans without HCV infection (Figure 1C). In the fine-mapping analysis of the HCV infection negative group, the most significant results were found in ELMO1 on chromosome 7 (p = 0.0012) and DLGAP1 on chromosome 18 (p = 0.0008); however, these results are obtained from a very small cohort and should be considered with caution (Supplementary Table 3).

Replication analysis

Cases and controls had mean age of 62.0 and 60.2 years, respectively, and were mostly male (83.8% and 93.5%, respectively). Over 40% of cases (40.4%) and controls (46.7%) were obese at study enrolment. In the case-control replication analysis, we were not able to replicate the admixture findings on chromosome 2 after adjusting for multiple comparisons (Supplementary Table 4). One of the most interesting risk variants near POTEK is SNP rs788186 (OR = 1.78, p = 0.01). The risk allele is more frequent in cirrhosis cases (0.206) than controls (0.126) and in similar proportion in the VA cirrhosis cases (0.196); the alternative allele is also more common in CEU (0.2626) than YRI (0.162) which is consistent with the admixture mapping result that this is a region in which European ancestry confers risk in African Americans. Additionally, nearby SNP rs2707645 (OR = 1.4, p = 0.04) is a significant risk variant that is associated with European ancestry in this region. Cirrhosis cases (0.522) have a greater frequency of the minor allele compared to controls (0.439) and there are similar proportions in the VA cirrhosis cases (0.518). The alternative allele is more common in CEU (0.753) than YRI (0.38), which is consistent with the admixture result. This SNP is not exonic and is located near the gene INPP5B. On chromosome 6, eight SNPs in DNAH8 were moderately associated with increased cirrhosis risk in African Americans (all OR = 1.1–1.8) in the replication analysis, but none were statistically significant (all p > 0.1) (Supplementary Table 4).

DISCUSSION

Admixture analysis revealed novel genomic regions on chromosomes 2 and 6 associated with cirrhosis risk in African Americans. Sub-group analyses found that these regions might be specific to HCV-related cirrhosis risk, such that both 2q21.1 and 6p21.2 may harbor cirrhosis susceptibility alleles that increase the risk of HCV-related cirrhosis for African Americans.

For the admixture region on 2q21.1, the most significant variants were located near POTE Ankyrin Domain Family Member K (POTEKP). POTEKP, also known as kappa-actin, was previously reported to be elevated in HCC tissue.³⁵ In addition, mutations in CCDC115 (coiled-coil domain containing 115) on 2q21.1 have been previously reported for disorders of Golgi homeostasis, which may lead to hepatosplenomegaly and liver failure in children.³⁶ Comparative genomic hybridization of multinodular liver cancer also identified 2q21.1 as locus associated with metastasis.³⁷ Although we could not replicate the results in the independent dataset, the findings suggest that this region is of interest for additional liver cancer studies in African Americans.

On chromosome 6, we found association with increased cirrhosis risk for the SNPs in DNAH8, and mutations in DNAH8 have been previously reported for worse disease-free survival in patients with HCC.³⁸ As the admixture results indicated this is a region in which African ancestry confers risk for cirrhosis, it was not surprising that loss-of-function variants in DNAH8 occur more frequently in African and African American populations than in European populations (https://varsome.com/gene/dnah8). However, we could not fully replicate the DNAH8 result from the fine-mapping analysis.

Interestingly, the 6p21.2 cytoband also includes CDKN1A (p21), which has been previously reported for association with alcoholic liver disease in humans³⁹ and hepatocarcinogenesis in mice.⁴⁰ In addition, loss-of-function variants in CDKN1A do not occur often, but they are still more frequent in African and African American populations than European populations,⁴¹ and this is consistent with our identification of African ancestry in this region conferring increased cirrhosis risk. While the fine-mapping results were not definitive in identifying novel genes of risk, the admixture mapping findings narrowed down genomic regions of interest on chromosomes 2 and 6 that may harbor causal variants for cirrhosis in African Americans.

Interestingly, the admixture findings differed by HCV status in the stratified analysis. While the HCV-negative group was much smaller than the HCV-positive, the genomic regions identified in this sample were consistent with known cirrhosis and cancer-predisposing genes. Specifically, African ancestry at genomic regions at 7p14.1 and 18p11.31 conferred increased risk of cirrhosis in African Americans without HCV infections. Locus 7p14.1 is ELMO1 which has been previously reported for primary biliary cirrhosis.⁴² In addition, locus 18p11.31 has been previously reported as possibly containing tumor suppressor genes,⁴³ which is of interest as cirrhosis is a risk factor for HCC.⁴⁴ It would be of interest to conduct a follow-up analysis of these patients to determine whether these findings are seen in a larger cohort.

There are several strengths to this study. The admixture mapping approach to gene discovery is a powerful method that takes advantage of the LD patterns among the recently admixed African American population so that fewer individuals and fewer SNPs are required than for a GWAS. In addition, this study focused on African Americans, a population that is known to have different exposure patterns for known risk factors and possibly different genetic risks for cirrhosis than other racial/ethnic subgroups. Although racial differences in LD and cirrhosis risk factors warrant the inclusion of African Americans in genetic studies, few studies have been done in this population. It is noteworthy to mention that both 2q21.1 and 6p21.2 regions were not identified as associated with cirrhosis risk in a recent GWAS by Emdin et al.¹⁰ In that study, the largest GWAS to date and conducted among a primarily Caucasian population, 5 established (rs2642438, rs6834314, rs28929474, rs58542926, rs739409) and 7 novel SNPs (rs12904, rs888655, rs9398804, rs7029757, rs1799992, rs429358, and rs1883711) were associated with cirrhosis risk. None of these SNPS were located in 2q21.1 and 6p21.2.

Although we identified two novel genomic regions associated with cirrhosis risk in African Americans, our study has several limitations as well. First, the affected-only (i.e., cirrhosis) study design ideally requires strong prior information on the frequencies of each allele given locus ancestry. Second, admixture analysis depends on differences in genetic factors between the parental populations of the admixed samples. Our study, using African Americans can map genetic factors that vary between African and European populations. Our results may not capture variability for other populations like Hispanics who have different parental populations.⁴⁵ Third, although we performed sample size and power calculations, our sample size may not be sufficient to detect all significant genetic alleles, particularly those of moderate effect size, and conduct subgroup analysis. Also, admixture mapping takes advantage of recombination that has occurred ancestrally for the study population and is efficient for initial identification of regions of interest but requires larger sample size to refine the specific genes contributing to disease risk. Therefore, larger studies or targeted mechanistic studies will help to specifically identify the genetic factors that are contributing to disease risk. Fourth, the subgroup analysis by HCV infection status suggested that there may be a different genetic susceptibility profile for cirrhosis risk by cirrhosis etiology. However, the number for non-HCV cirrhosis African American patients in our analysis was small and this therefore limited the interpretation of the findings for this subgroup. Last, our admixture mapping study was conducted among majority male Veterans. As such, the generalizability of our findings to African American females needs further investigation.

It is exceedingly important to better understand the underlying biology of cirrhosis to develop targeted treatments and to identify populations at highest risk. It is likely that many genes contribute to an individual’s cirrhosis (and HCC) risk, and our finding that these variants may differ by HCV status is an important piece to the equation. These results add to the findings from the GWAS in Caucasian populations and suggest novel candidate gene areas. These regions need to be confirmed in additional studies with lager sample size and finer mapping.

Supplementary Material

1

Supplementary Figure 1. PCA plot of the 227 African American cirrhosis patients and reference populations from the 1000 Genomes project, including African Ancestry in Southwest US, ASW (n=61); Utah residents with Northern and Western European Ancestry, CEU (n=99); Mexican Ancestry in Los Angeles, MXL (n=64); and Yoruba in Ibadan, Nigeria, YRI (n=108).

Supplementary Figure 2. Distribution of mean global African ancestry among the 227 African American cirrhosis patients in the study.

Supplementary Figure 3. Admixmap z-scores for chromosomes 2 and 6. The right panels focus on regions that suggested association with ancestry (|z-score|>3, red-colored regions in the left panels).

What You Need to Know.

Background

Cirrhosis is the final common pathway of all chronic liver disease, and is the main precursor lesion for hepatocellular cancer. Multiple germline variants have been reported for cirrhosis risk; however, it remains unclear whether host genetic factors contribute to racial disparities in cirrhosis.

Findings

In a genome-wide admixture mapping study among African American cirrhosis patients, we identified novel genomic regions in 2q21.1 and 6p21.2 that may be associated with HCV-related cirrhosis risk.

Implications for patient care

These data provide preliminary insight into the genetic contribution to racial/ethnic disparities in cirrhosis risk, and, when combined with findings from genetic studies among Caucasian populations, could lead to methods for identification of high-risk patients.