Epigenomic Profiles of African American Transthyretin Val122Ile Carriers Reveals Putatively Dysregulated Amyloid Mechanisms

Abstract

Background -

The Val122Ile mutation in Transthyretin (TTR) gene causes a rare, difficult to diagnose hereditary form of cardiac amyloidosis. This mutation is most common in the United States and mainly present in people of African descent. The carriers have an increased risk of congestive heart failure, peripheral edema, and several other non-cardiac phenotypes such as carpal tunnel syndrome, and arthroplasty which are top reasons for ambulatory/outpatient surgeries in the country.

Methods -

We conducted first-ever epigenome-wide association study using the Illumina’s EPIC array, in Val122Ile carriers of African descent for heart disease (HD) and multiple outpatient surgeries (OS) - an early disease indicator. Differential methylation across genome wide CpG sites was tested between carriers with and without HD and OS. Significant CpG sites were investigated for cis-mQTLs loci, followed by gene ontology and protein-protein interaction (PPI) network. We also investigated the significant CpG sites in a secondary cohort of carriers for replication.

Results -

Five differentially methylated sites (p≤2.1e-08) in genes – FAM129B, SKI, WDR27, GLS, and an intergenic site near RP11-550A5.2, and one differentially methylated region containing KCNA6 and GALNT3 (p=1.1e-12) were associated with HD. For OS, we observe four sites – two sites in UBE2E3 and SEC14L5, and other two in intergenic regions (p≤1.8e-07) and three regions overlapping SH3D21, EVA1B, LTB4R2 and CIDEB (p≤3.9e-07). Functional protein-interaction module analysis identified ABCA1 (p=0.001) for HD. Six cis-mQTLs were associated with one of the significant CpG sites (FAM129B; p=4.1e-24). We replicated two CpG sites (cg18546846 and cg06641417; p<0.05) in an external cohort of biopsy- confirmed cases of TTR amyloidosis. The genes identified are involved in transport and clearance of amyloid deposits (GLS, ABCA1, FAM129B); cardiac fibrosis (SKI); and muscle tissue regulation (SKI, FAM129B).

Conclusions -

These findings highlight the link between a complex amyloid circuit and diverse symptoms of Val122Ile.

Introduction

Hereditary transthyretin amyloidosis, caused by specific disease-causing mutations, is due to a gradual extracellular deposition of amyloid in multiple tissues primarily leading to several clinical signs and symptoms¹. There are 113 known mutations in the transthyretin (TTR) gene² giving rise to hereditary form of TTR amyloidosis. The tetrameric structure of TTR protein dissociates into dimers and monomers resulting in formation of fibrils. The heterogeneous symptoms of hereditary TTR amyloidosis (hATTR) arises from amyloid deposition in different tissues and organs^2,3. Val122Ile mutation (NP_000362.1:p.Val142Ile; rs76992529) is the most prevalent TTR amyloidogenic variant in the United States and is primarily observed in populations of African descent⁴. This point mutation results in the substitution of an isoleucine with a valine at 122 position of TTR mature protein. According to latest report by International Society of Amyloidosis, the recommended nomenclature is named based on the substitution or deletion of the TTR protein, hence here we use Val122ILe ⁵. Extensive amyloid deposition seems to resemble hypertrophic cardiomyopathy such as enlargement or wall thickening leading to heart failure and atrial fibrillation⁶. These symptoms often are attributed to other population-prevalent cardiovascular risk factors resulting in underestimation of the clinical penetrance of the Val122Ile⁷. The estimated age of onset for non-cardiac precursor phenotypes for hereditary transthyretin amyloidosis is between 30 to 40 years of age⁸. In two retrospective studies, carpal tunnel syndrome preceded hereditary transthyretin amyloidosis diagnosis by 9-10 years^9–11. According to The Transthyretin Amyloid Outcome Survey (THAOS), several cardiac, gait, gastrointestinal, neurological and renal disorders are prevalent in Val122Ile carriers⁴. Parallel to these findings, another study reported atrial fibrillation, myopathy related to ventricular thickness, gastrointestinal and kidney dysfunction including nausea, vomiting, and neuromuscular dysfunction to be associated with TTR mutations¹².

We previously investigated medical history phenotypes in the Yale-Penn cohort associated with Val122Ile carriers of African descent and found heart disease history and having had 10 or more outpatient (ambulatory) surgeries significantly associated with this amyloidogenic variant¹³. One of the top reasons for ambulatory surgery in the United States is arthroplasty¹⁴, which occurs in TTR-carriers years before the expected cardiac dystrophy at advanced ages¹⁵. These epidemiological findings indicate that atypical phenotypes occurring earlier in life could be connected to the risk of heart failure in Val122Ile carriers. Findings from several studies including ours raise the possibility of non-regulatory molecular factors contributing to the genotype-phenotype correlation^12,16–19. Therefore, understanding the underlying biological changes in Val122Ile carriers is key to explaining the symptom heterogeneity and earlier onset of atypical phenotypes.

DNA methylation is a heritable non-coding regulatory mechanism causing phenotypic variation¹⁸. Epigenetic modifications arising from the addition of methyl groups on cytosine-phosphate-guanosine (CpG) sites²⁰ could contribute to molecular mechanisms involved in TTR amyloidosis. So far, no study has investigated this hypothesis. Aberrant methylation profiles have been implicated in increasing accelerating the progression of common and rare diseases²¹ . The accumulation of amyloid-fibrils within or around cellular structures of the tissue result in damage invoking an immune response²². The inter-individual variation in response to site of damage invokes an acute phase response²². DNA methylation profiles have the potential to capture individual-level variability and highlight mechanisms involved in TTR amyloidosis²³. Thus, we conducted the first epigenome-wide association study of TTR Val122Ile carriers to investigate the association of methylation changes with medical history of heart disease and outpatient surgeries.

Methods

The intensity files of methylation data have been uploaded to Gene Expression Omnibus (GSE154683). The authors declare that all supporting data are available within the article [and its online supplementary files]. For further inquiries, please contact the corresponding author. Author contributions

Yale-Penn study was approved by the institutional review boards at each participating site including receiving participants consent. The current study was approved under the protocol 2000023750 by the institutional review board (IRB) at Yale University School of Medicine. All intensity files of methylation data have been made publicly available at the GSE154683 and can be accessed at https://www.ncbi.nlm.nih.gov/geo/. The authors declare that all supporting data are available within the article [and its supplementary files]. Detailed methods are available in the Supplementary section.

Results

Differentially methylated sites

We investigated differentially methylated sites with respect to two binary outcomes: – a) self-reported heart disease and b) a history of 10 or more outpatient surgeries. After performing the recommended quality control procedure, we investigated 737,385 sites in 96 individuals. In addition to performing standard association analysis, we permuted the phenotypes (p_perm: p-value from permutation), which accounted for the case-control imbalance, yielding nine significant CpG sites (Figure 1). Five sites were hypomethylated in individuals with heart disease: cg06641417 (FAM129B; logFC=−1.822; p_perm=1.6e-08), cg26033908 (SKI; logFC=−1.615; p_perm=1.7e-08), cg14890866 (WDR27; logFC=−2.028; p_perm=3.0e-08), cg15522719 (GLS; logFC=−1.731; p_perm=4.7e-08) and cg18546846 (RP11-550A5.2; intergenic; logFC=−0.786; p_perm=2.2e-08). The CpG sites mapped to FAM129B and SKI are located in gene bodies, cg14890866 is between the 5’UTR (Un-Translated Region) and TSS200 (− 200 nt upstream of Transcription start site²⁴) of WDR27, while cg15522719 is in TSS150 at GLS. Four methylation sites were associated with 10 or more outpatient surgeries: cg13998023 (UBE2E3; logFC=−2.632; p_perm=1.8e-07), cg05189127 (intergenic; logFC=1.885; p_perm=1.4e-07), cg03718655 (SEC14L5; logFC=−2.673; p_perm=1.5e-07) and cg25814327 (intergenic; logFC=−2.075; p_perm=3e-08). Three sites were hypomethylated, while cg05189127 (intergenic) was hypermethylated. The two sites that were annotated to genes were in the 5’UTR (UBE2E3) and TSS200 (SEC14L5). Details of the association result and annotation are reported in Supplementary file (Table S1).

Figure 1:

Differentially methylated sites in African American TTR-Val122Ile carriers. (A) Methylation sites that were significantly associated with medical history of heart disease. (B) Methylation sites that were significantly associated for having had 10 or more outpatient surgeries. Each CpG site is represented as a data point, with the x-axis being the genomic location, grouped by chromosome and wherein colors represent alternating chromosomes. The y-axis is the −log10 of the p-value of the CpG site. Significant sites are shown as triangles and labelled with CpG probe name and genic annotation in parentheses, triangles pointing upwards signify hypermethylation, whereas triangles pointing downwards signify hypomethylation.

Differentially methylated regions

For heart disease, one region on chromosome 12 overlapping KCNA6 and GALNT3 (p=1.1e-12) was differentially methylated. Associations with more than 10 outpatient surgeries were identified on chromosome 1 (SH3D21; EVA1B; p=1.3e-09), chromosome 10 (intergenic region; p=1.7e-08) and chromosome 14 (LTB4R2; CIDEB; p=3.9e-07). Methylation levels among all sites were positively correlated within each region. (Figure 2; Supplementary file; Table S2)

Figure 2:

Differentially methylated regions in TTR-Val122Ile carriers. (A) Regional association with heart disease (B-D) Regional association with 10 or more outpatient surgeries. Each panel displays the association of sites within each region, followed by genomic location, ENSEMBL gene name, DNAse, regulation, SNP tracks (from UCSC browser) and correlation of CpG sites shown as a heatmap.

Overrepresented gene ontology and PPI networks

Differentially methylated sites and regions were annotated to their respective genes using UCSC RefGene: hg19 genome build. The gene ontology (GO) analysis identified 15 significantly enriched pathways in GO’s biological process. GLS, SKI, GALNT8, and KCNA6 are involved in protein oligomerization (FDR_p-value=4.8e-03) and KCNA6 and GALNT8, which are located near one another, are involved in potassium ion transport (FDR_p-value=4.9e-02). FAM12B and SKI (FDR_p-value=4.8e-03 to 3.2e-02) participate in the development of various tissue types – myotubules, and skeletal and striated muscles (Figure 3; Supplementary file, Table S3).

Figure 3:

Enriched gene ontology (GO) biological processes. The dendrogram shows the FDRpvalue of the pathway associations and are grouped by similarity of function. The genes involved in each of the processes are highlighted in orange bars.

We also investigated the methylation sites for differentially methylated functional modules using the R package FEM ²⁵. The CpG sites are weighted based on their location in the genes, which are then mapped to a protein-protein interaction (PPI) network. For each module (i.e. PPI network) identified, the seed gene is the primary gene to which other functionally related genes are connected. For heart disease, we found the ABCA1 module to be significant (p=0.001) and target genes identified within the module were: ABCA1, SNTB2, BLOC1S2 and LIN7B (p<0.05). The EXOSC4 gene module was associated with the phenotype of 10 or more outpatient surgeries, and it was the only gene that was a target (Figure 4; Supplementary file, Table S4).

Figure 4:

Functional Protein-Protein Interaction (PPI) Networks. The differentially methylated modules consist of a network of genes based on their functional connectivity using protein-protein interaction. Each module has primary gene which is connected to other target genes in the network. Each module was significant p<0.05 using the FEM method (see methods). The genes in blue show hypermethylation and yellow represents hypomethylation. A) ABCA1 module was associated with heart disease and the significant target genes in addition to ABCA1 were SNTB2, BLOC1S2 and LIN7B. B) EXOSC4 module was associated with outpatient surgeries and also was the only significant gene in the network.

Local quantitative trait loci for methylated sites (mQTL)

We tested SNP associations with nine methylation sites that were epigenome-wide significant with the two phenotypes investigated. The cis-mQTL loci were defined as SNPs within ±1 Mb of the significant CpG site. The sites were considered significant based on an FDR_p-value < 0.05 and genomic corrected p-value (p_gc<0.05). We found six SNPs, rs192528579, rs182192023, rs114553373, rs187644239, rs114896522, and rs139996037 significantly associated (Figure 5; p=4.1e-24; Table S5) with site cg06641417. The SNPs are in high linkage disequilibrium (LD>0.8), rs192528579 is in the intronic region of neighboring gene –- GARNL3; rs182192023, rs114553373, rs187644239 and rs114896522 map to LRSAM1. Rs139996037 is a non-coding transcript variant of the FAM129B.

Figure 5:

Local-mQTL associated with site – cg06641417 mapped to the FAM129B gene. The top panel displays single nucleotide polymorphisms (SNPs) associated with CpG site – cg06641417 as a data point and color coded based on linkage disequilibrium with the top lead SNP in purple. The x-axis shows gene annotation (hg19) of the region and the y-axis displays the −log10 of p-value. The following panels present various annotations of the reported SNPs i.e. CADD - Combined Annotation Dependent Depletion, and RegulomeDB – score to identify regulatory variants. The bottom panel highlights the chromatin states of various regulatory features being putatively affected from chromatin markers observed in aorta tissue cell line. Visualization made in FUMA.

Epigenetic age

The epigenetic age (DNAm) was measured using the biological clock developed by Horvath and colleagues, which uses 353 CpG sites²⁶ and also with a second clock based on 71 CpG sites from Hannum and colleagues²⁷. The ‘Horvath’ clock is considered to be a pan-tissue epigenetic clock, while the ‘Hannum’ clock is considered to be accurate for whole-blood tissue²⁸. Both clocks estimated that carriers with heart disease are of older epigenetic age than carriers without heart disease (p_Horvath=0.007 and p_Hannum=0.0009) (Supplementary file; Table S6). However, the (delta) ΔDNAm age (difference in chronological and biological age) was not significant between the two groups (p_ΔHorvath=0.31 and p_ΔHannum=0.57; Supplementary file; Table S7).

Replication of methylation sites in the Italian cohort

We tested the nine CpG sites identified in Val122Ile carriers in an independent cohort of biopsy-confirmed TTR amyloidosis cases and healthy controls. We replicated cg18546846 (intergenic; near to RP11-550A5.2; p= 0.021) and cg06641417 (FAM129B; p=0.016) at nominal significance (p<0.05).

Discussion

The clinical consequences of the TTR Val122Ile mutation remain underappreciated and the syndrome that accompanies this risk mutation, under-diagnosed. Individuals exhibiting early TTR-amyloidosis symptoms are more likely to be diagnosed with another condition prior to receiving the diagnosis of TTR-amyloidosis²⁹. There is nonetheless a greater burden over time towards developing ventricular hypertrophy, reduced left ventricular ejection fraction, and atrial dilation, at a later age^3,6. We previously showed that African-American carriers of the Val122Ile mutation had a higher prevalence of heart disease and having multiple outpatient surgeries than individuals without the mutation¹³. In the present study, we identified methylation changes associated with these same phenotypes in Val122Ile carriers. We also replicated two CpG sites (RP11-550A5.2; cg18546846 and FAM129B; cg06641417) at nominal significance in an external cohort including biopsy confirmed cases of TTR amyloidosis³⁰. Thus, we hypothesize that the epigenetic changes associated with the pathogenesis heart disease differs from the methylation profile of carriers who are not affected by the disease. Lastly, we used GeneMANIA³¹ to interpret the interaction among the significant genes (Supplementary file). We observed that major genes identified in the present study physically interact and share pathways with TTR (Figure 6).

Figure 6:

Gene network of significant genic hits and their relationship with TTR. The genes circled in red are the query genes that were identified from the reported analysis. The network shows intermediate genes that connect the query genes based on different interactions as shown in the legend. The lines are colored based on the type of network domain shown in the figure.

ABCA1 (ATP binding cassette transporter A1) identified via the functional network analysis encodes a transporter of cholesterol from apolipoproteins³². ABCA1 regulates Apolipoprotein E (ApoE) levels, with lower expression of ABCA1 reducing ApoE levels. However, ApoE with ApoA1 (Apolipoprotein A) reduces amyloid deposition twice as fast as inhibition of the expression of ApoE. Additionally, amyloid-beta levels were the lowest for the dual-knockout of APP (which encodes amyloid precursor protein) and ABCA1³³.

GLS (glutaminase) is a key contributor to the metabolizing of glutamine to glutamate³⁴. Amyloid-beta-treated neurons show elevated glutaminase expression, which increases glutamate levels and disrupts calcium neural regulation³⁵. Additionally, neurofibrillary tangles have been shown to coexist with higher glutaminase activity³⁶. The hypomethylated site in the transcription start site of the GLS gene may indicate its potential involvement in the central nervous system, which supports the recent finding of cerebral amyloid angiopathy in individuals with mutated TTR cardiac amyloidosis³⁷. FAM129B (aliases; MEG-3 and NIBAN2) is downregulated in tissues with amyloid deposition and animal studies have shown that enhancing the expression of FAM129B reduces oxidative damage by reducing amyloid-beta production via PI3K/Akt signaling³⁸. Cardiac hypertrophy increases the risk of heart failure. FAM129B is overexpressed in heart failure samples, and rodent experiments indicate a potential role of the gene in the apoptosis of cardiac myocytes after myocardial infarction³⁹. Rescuing the expression levels of FAM129B reverses hypertrophic responses, thus the hypomethylation of the CpG site in FAM129B observed in our finding supports the overexpression of the gene in heart failure⁴⁰. The African-American population has a high prevalence of diabetes⁴¹. FAM129B is also overexpressed in cardiomyocytes under high glucose concentration reflecting its role in diabetic cardiomyopathy⁴². Although SKI is an inhibitor of TGF-beta-induced fibrosis and is under expressed in cardiac fibrosis, other epigenetic modulators such as miRNAs-34a and 93-c affect both SKI and TGF-beta, targeting the inhibitory factors of SKI, which could rescue cardiac fibrosis⁴³. The gene enrichment analysis identified a role for FAM129B and SKI in the development of myotube cells and skeletal muscle fiber and organ, and striated muscle cell development.

One of the clinical findings associated with cardiac amyloidosis is increased left ventricular wall thickness, which can lead to heart failure⁴⁴. Electrical perturbations resulting from lower potassium repolarizing current leads to a prolonged action potential in heart failure⁴⁵. The role of KCNA6 and GALNT8 is associated with potassium ion transport and the transmembrane transport complexes. One of the cardiovascular symptoms of the TTR amyloidosis is pronounced diastolic hypertension⁷, and diastolic dysfunction is one of the symptoms associated with transthyretin amyloidosis⁴⁶ . WDR27 was reported to be differentially methylated in individuals with significant differences in diastolic blood pressure²⁶.

Aging is a common denominator to the symptomology of Val122Ile and DNA methylation⁴⁷. Age-related methylation changes measured via “epigenetic clocks” help to identify molecular aging and its disconnect with chronological age. The Horvath clock based on 353 CpG sites and the Hannum clock based on 71 CpG sites have been extensively replicated in various tissues⁴⁸. While these clocks were developed using blood tissues^26,27, Horvath’s clock is validated across multiple tissues, while Hannum’s clock is more consistent in samples originating from blood tissues. Higher epigenetically derived age has been associated with several cardiovascular disease traits. Hypermethylation of genes that are protective against heart disease, lead to cardiovascular aging and increased risk for coronary disease⁴⁹. The dysregulation of the ABCA1 gene, the product of which is involved in the transport of cholesterol from the periphery to liver tissue⁵⁰ has been associated with different cardiovascular pathologies. The hypermethylation of the ABCA1 promoter region silences its expression and is associated with coronary artery disease⁵¹. In contrast, the increased expression of ABCA1 regulated by ApoA1 leads to reverse cholesterol efflux in hepatic tissue ⁵². Elevated high density lipoprotein (HDL) in the liver is a target site for serum amyloid A, an acute phase response protein that is expressed during amyloidosis⁵³. The observed hypomethylation of ABCA1 and putative increase in gene expression underscores its likely involvement in shifting the methylation milieu and could perhaps explain the cardiac symptomology in a comparatively younger group of Val122Ile carriers. These findings reflect the observational symptomology of carpal tunnel, a common denominator to arthroplasty and hATTR⁵⁴.

These findings provide unique insights into epigenomic contrasts related to symptomology in Val122Ile carriers. However, our study has limitations. First, we investigated the Val122Ile polymorphism only for heart disease, though it is possible that we could identify additional differences with individuals who are non-Val122Ile carriers or who present with wild-type transthyretin amyloidosis. Additionally, due to the low frequency of the disease-causing mutation investigated, our study suffers from an imbalance in the ratio of cases to controls. Although the permutation analysis accounting for this imbalance confirmed our results and we replicated two associations in an independent cohort (with mostly different risk variants). Our findings would benefit from replication in a larger, more balanced study to further dissect the underlying disease mechanisms. While the DNAm age was significantly different in the two groups, the ΔDNAm age was not different. It is possible that our study is underpowered to detect delta-DNAm age. Additionally, there may be some biases in applying DNAm age measures developed largely on individuals of European descent to individuals of other ancestries, and such ancestry-stratified DNAm differences have been reported by other studies as well ^55–57.

Conclusions

Our study is the first to explore the epigenetic changes in TTR Val122Ile carriers. Certain Val122Ile carriers in our study presented with heart disease earlier than usually reported by individuals affected by cardiac amyloidosis. This could be due to modifier effects accelerating the pathogenicity of Val122Ile mutation. Due to the underestimated clinical penetrance of the mutation in the African American population, we leveraged an external secondary dataset with confirmed clinical phenotype as a random population sample. The purpose of this study was to understand possible non-coding mechanisms that may explain the heterogeneous phenotypes observed in Val122Ile carriers with history of heart disease. The epigenetic changes identified affect the regulation of genes involved in the transport of amyloid and regulating striated and smooth muscle, which form key components of amyloidosis and cardiac tissue susceptibility. These findings provide higher resolution on mechanisms underlying the TTR-Val122Ile mutation.

Nonstandard Abbreviations and Acronyms

TTR

Transthyretin

OS

Outpatient surgery

HD

Heart Disease

CpG

Cytosine-phosphate Guanine

hATTR

hereditary TTR amyloidosis

SNP

Single Nucleotide Polymorphisms

TSS

Transcription Start Site