Epigenetic profiling of Italian patients identified methylation sites associated with hereditary transthyretin amyloidosis

Hereditary transthyretin (TTR) amyloidosis (hATTR) is a rare life-threatening disorder caused by amyloidogenic coding mutations located in TTR gene. To understand the high phenotypic variability observed among carriers of TTR disease-causing mutations, we conducted an epigenome-wide association study (EWAS) assessing more than 700,000 methylation sites and testing epigenetic difference of TTR coding mutation carriers vs. non-carriers. We observed a significant methylation change at cg09097335 site located in Beta-secretase 2 (BACE2) gene (standardized regression coefficient = −0.60, p = 6.26 × 10–8). This gene is involved in a protein interaction network enriched for biological processes and molecular pathways related to amyloid-beta metabolism (Gene Ontology: 0050435, q = 0.007), amyloid fiber formation (Reactome HSA-977225, q = 0.008), and Alzheimer’s disease (KEGG hsa05010, q = 2.2 × 10–4). Additionally, TTR and BACE2 share APP (amyloid-beta precursor protein) as a validated protein interactor. Within TTR gene region, we observed that Val30Met disrupts a methylation site, cg13139646, causing a drastic hypomethylation in carriers of this amyloidogenic mutation (standardized regression coefficient = −2.18, p = 3.34 × 10–11). Cg13139646 showed co-methylation with cg19203115 (Pearson’s r2 = 0.32), which showed significant epigenetic differences between symptomatic and asymptomatic carriers of amyloidogenic mutations (standardized regression coefficient = −0.56, p = 8.6 × 10–4). In conclusion, we provide novel insights related to the molecular mechanisms involved in the complex heterogeneity of hATTR, highlighting the role of epigenetic regulation in this rare disorder.


Background
Hereditary transthyretin amyloidosis (hATTR; OMIM#105210) is a life-threatening disorder caused by transthyretin (TTR) protein misfolding. This causes amyloid fibril deposition in several tissues (e.g., peripheral nerves, heart, and gastrointestinal tract) [1,2]. hATTR is characterized by extreme clinical heterogeneity including age of onset, penetrance, and clinical display [3][4][5]. To date, more than 130 amyloidogenic mutations have been identified in the coding regions of the TTR gene, which are the cause of hATTR [6]. The prevalence of hATTR is estimated to be approximately 1/100,000 [7]. hATTR endemic areas have been identified in Portugal and Sweden [4,5]. Although both of these regions are affected by the same amyloidogenic mutation, Val30Met (rs28933979), the penetrance and age of onset are different: early age of onset and high penetrance in Portugal [4,5,8,9] vs. late age of onset and low penetrance in Sweden and in non-endemic countries [3,10,11]. hATTR phenotypic heterogeneity is likely due to the contribution of genetic and non-genetic factors involved in the complex genotype-phenotype correlation observed [12][13][14][15][16][17][18]. Recent data strongly support the role of non-coding regulatory variation on TTR gene expression, as one of the mechanisms affecting the phenotypic manifestations observed in carriers of TTR amyloidogenic mutations [19][20][21][22]. Among genomic regulatory features, epigenetic modifications are key mechanisms in modulating a wide range of molecular functions and potential targets to develop novel treatments [23][24][25]. Of several epigenetic modifications, DNA methylation is the most studied with respect to human traits and diseases [23]. In the context of monogenic disorders, methylation studies investigate the role of epigenetic changes involved in the phenotypic expression observed among carriers of disease-causing mutations [26][27][28]. While epigenetic modifications have the potential to be involved in hATTR pathogenic mechanisms, to our knowledge no study has explored methylation changes of patients affected by this lifethreatening disease. In the present study, we conducted an epigenome-wide association study (EWAS) to identify DNA methylation sites associated with hATTR, investigating 48 carriers of TTR amyloidogenic mutations and 32 controls, comparing hATTR affected patients, asymptomatic carriers, and non-carriers.
Within TTR gene region, we observed that Val30Met mutation disrupts a methylation site, cg13139646, causing a drastic hypomethylation in Val30Met carriers when compared with carriers of other TTR mutations (standardized regression coefficient = −2.18, p = 3.34 × 10 -11 ; Additional file 3: beta-value and M value plots). Since Val30Met mutation disrupts the probe extension for the cg13139646 methylation site (the Met allele removes the G nucleotide from the CpG site), its impact on the methylation changes in TTR gene should be confirmed using alternative typing methods. However, to conduct an initial exploratory analysis of the potential functional implications of cg13139646 disruption, we performed a co-methylation analysis with respect to cg13139646 in the non-carrier sample. We identified 34 methylation sites that are correlated with cg13139646 site (Pearson's r 2 > 0.20; Additional file 4). Considering these co-methylated CpG sites, we investigated epigenetic differences among hATTR patients, asymptomatic carriers of TTR mutations, and controls (Additional file 5). Applying a Bonferroni correction accounting for the number of CpG sites tested, we observed a significant methylation difference between hATTR patients and asymptomatic carriers of TTR amyloidogenic mutations at cg19203115 (standardized regression coefficient = −0.555, p = 8.6 × 10 -4 ; Additional file 6: beta-value and M value plots). Nominally significant methylation differences were observed (i) between carriers vs. non-carriers at cg11481443 (standardized regression coefficient = −0.306, p = 3.4 × 10 -3 ; Additional file 6: beta-value and M value plots) and cg02936398 (standardized regression coefficient = 0.177, p = 4.9 × 10 -2 ; Additional file 6: beta-value and M value plots); (ii) hATTR patients and asymptomatic carriers at cg14311811 (standardized regression coefficient = −0.273, p = 3.8 × 10 -2 ; Additional file 6: beta-value and M value plots). Considering symptoms reported by hATTR patients, we identified CpG sites co-methylated with cg13139646 (i.e., the site disrupted by Val30Met mutation) nominally associated with cardiac involvement (cg27392998, standardized regression coefficient = −0.235, p = 7.5 × 10 -3 ; cg18038361, standardized regression coefficient = 0.227, p = 5 × 10 -2 ; Additional file 6: beta-value and M value plots), carpal tunnel syndrome (cg16492377, standardized regression coefficient = −0.229, p = 1.8 × 10 -2 ; Additional file 6: betavalue and M value plots), and peripheral nervous system involvement (cg14719951, standardized regression coefficient = 0.249, p = 3.5 × 10 -2 ; Additional file 6: beta-value and M value plots). Since some of these CpG sites were mapped to loci located near TTR gene (Additional file 5), we analyzed TTR transcriptomic profile in hATTR target organs, observing a different pattern when compared to the expression of the surrounding genes ( Fig. 2, Bottom Panel).

Discussion
hATTR is a rare multi-organ disorder caused by TTR misfolding and consequently amyloid deposition in several tissues [30]. This life-threatening condition is characterized by high clinical heterogeneity with respect to age of onset, penetrance, and phenotypic manifestation [1][2][3][4][5][6][7][8][9][10]30]. Although TTR amyloidogenic mutations are the cause of TTR misfolding, non-coding variation and modifier genes are hypothesized to be involved in the wide variability of phenotypic manifestations observed in carriers of TTR disease-causing mutations [12,15,[17][18][19][20][21][22]. Epigenetic modifications (e.g., DNA methylation changes) could also play an important role in the molecular network regulating the hATTR amyloidogenic process [25]. To explore this hypothesis, we conducted an EWAS investigating more than 700,000 methylation sites in 48 carriers of TTR amyloidogenic mutations and 32 non-carriers. A CpG site (cg09097335) located in BACE2 gene was significantly hypomethylated in carriers when compared to non-carriers. This gene encodes Beta-secretase 2, a protein mainly known for its role in cleaving APP protein in amyloid-beta, which is a key factor involved in AD pathogenesis [31][32][33]. Differently from BACE1, which is the primary β-secretase protein cleaving APP to amyloid-beta, BACE2 is poorly expressed in the brain and its cleaving ability increases following an inflammatory response [34]. APP processing occurs via three proteolytic cleavages caused by αβ-and γ-secretase [35]. In non-amyloidogenic processes, α-and γ-secretases lead to the production of a smaller P3 fragment and APP intracellular domain, while, in the amyloidogenic pathway, β-secretase and γ-secretase produce amyloid-beta [35][36][37][38][39]. Our results also showed a high-confidence interaction between APP and TTR. Numerous studies explored the interactions between these two amyloidogenic proteins, displaying a relevant biological role of TTR in amyloid-beta aggregation and clearance in AD patients [40][41][42][43][44]. Specifically, TTR instability reduces the clearance of amyloid-beta, increasing amyloid toxicity in the brain [40][41][42]. Metal ions and interaction with other proteins could also affect TTR stability [40]. Although TTR genetic reduction did not alter APP processing, immunohistochemical and biochemical studies showed that genetic reduction of TTR elevates Aβ deposition in the brains of transgenic mice harboring APPswe/PS1ΔE9TTR + / − transgenes [45]. Under physiological conditions, the APP intracellular domain appears to be involved in epigenetically up-regulation of TTR to increase its amyloid-beta clearance activity [46]. In AD patients with TTR Val30Met, a significant association between amyloid-beta levels and AD was identified [40,44]. A putative amyloidogenic role of amyloid-beta in hATTR was also identified in a post-mortem analysis of a Val30Met carrier where both TTR and amyloid-beta were deposited in the cerebral leptomeningeal and cortical blood vessel walls with a part of the vessel wall occupied by a combination of TTR and amyloid-beta aberrant proteins [43]. These previous findings strongly indicate an interplay between the pathogenic mechanisms involved in hATTR and AD. Our epigenome-wide study identified BACE2 as a potential key factor in this interaction. As previously discussed, BACE2 protein plays a minor role in APP cleaving in the brain [33,34], while its activity increases in peripheral tissues under inflammatory response [34]. Our transcriptomic analysis showed that BACE2 is expressed in tissues affected by TTR amyloid deposits (i.e., heart, nerves, colon, small intestine, and adipose tissues). Accordingly, the methylation change observed in the TTR-mutation carriers is possibly due to the role of BACE2 in response to the inflammation induced by TTR amyloidogenic process in peripheral tissues [47].
Within TTR gene region, we observed that Val30Met disrupts a CpG site, causing a drastic hypomethylation in the carriers of this mutation. SNPs at CpG sites disrupting the methylation reactions can be associated with changes in regulatory function. [48,49]. Although the impact of Val30Met on cg13139646 has to be confirmed by alternative methods because of the cg13139646 prove extension disruption, we conducted an initial exploratory analysis, testing the co-methylation of this site with CpG sites in the surrounding regions (NC_000018.9: 28,171,000-30,171,500). Indeed, co-methylation patterns reflect specific molecular mechanisms responsible for the regulation of multiple genes located in the same region [50]. In our analysis, some of the CpG sites identified map to TTR gene region, while others map in nearby loci. Considering hATTR target organs, these surrounding loci with co-methylated CpG sites have higher gene expression than TTR . We speculate that these genes may be involved in the formation of TTR amyloid deposits. This hypothesis is supported by the fact that comethylated CpG sites are associated with hATTR traits. The strongest evidence was observed with respect to cg19203115 mapped in B4GALT6 gene. Considering a Bonferroni correction accounting for the number of comethylated CpG sites tested, cg19203115 showed a significant difference in methylation levels between hATTR patients and asymptomatic carriers of TTR mutations. B4GALT6 gene encodes beta-1,4-galactosyltransferase 6, a type II membrane-bound glycoprotein that has exclusive specificity for the donor substrate UDP-galactose. B4GALT6 enzyme activity changes in response to inflammatory processes [51]. B4GALT6 stimulates astrocyte activation through the catalyzation of lactosylceramide synthesis, which in turn controls the production of proinflammatory cytokines and chemokines [51]. Hence, observing methylation changes in B4GALT6 may be associated with the inflammatory response to TTR amyloid deposits. Nominally significant differences were observed for CpG sites mapped in other surrounding loci: DSC2 (cg02936398, Carriers vs. Controls); DSG2 (cg14311811, hATTR patients vs. asymptomatic carriers); DSC3 (cg16492377, carpal tunnel syndrome in hATTR patients). DSC2 and DSG2 encode components of the desmosome. This protein complex is specialized for cell-to-cell adhesion in myocardial tissue and mutations in DSC2 and DSG2 genes are associated with arrhythmogenic right ventricular cardiomyopathy [52]. In an in vivo study of myocardial inflammation, DSC2 overexpression was observed to lead to tissue necrosis, fibrosis, and calcification of ventricles [53]. This process alters homeostasis among desmosomal proteins, inducing a cascade of different cell-cell interactions leading to cardiac remodeling [53]. In hATTR, cardiac amyloid fibril depositions also led to tissue dysfunctions [7]. Heart failure, restrictive cardiomyopathy, and rhythm disturbances (i.e., conduction system diseases, atrial fibrillation, and ventricular tachycardia) are the main clinical signs that occur after the accumulation of misfolded TTR protein [54][55][56]. Furthermore, transcriptomic interaction is observed between TTR and DSG2 to induce hypertrophic cardiomyopathy in animal models [57]. In this context, methylation changes in DSC2 and DSG2 genes could reflect pathogenic processes in hATTR target organs. We also identified two CpG sites co-methylated with cg13139646 (i.e., the methylation site disrupted by Val30Met mutation) that are nominally associated with hATTR symptoms. Cg18038361 is located in TTR gene promoter region and is associated with cardiac involvement in hATTR patients. DNA methylation changes in promoter regions are well known to play an important role in gene expression regulation [58,59]. Cg18038361 association may be linked to regulatory changes in TTR gene expression. Lastly, two CpG sites-cg16492377 and cg14719951, map to DSC3 transcription start site and gene body, respectively. In our analysis, methylation changes in these sites were associated with carpal tunnel syndrome and peripheral nervous system involvement in hATTR patients, respectively. DSC3 gene encodes the desmocollin-3 a calcium-dependent glycoprotein. Low DSC3 expression in human epidermis leads to a loss of tissue integrity [60]. We speculate that cg16492377 methylation association with carpal tunnel syndrome may be related to changes in DSC3 transcriptomic regulation.
Although we provide novel findings regarding the role of methylation changes in hATTR, our study presents several limitations. Since hATTR is a rare disease, we investigated a relatively small sample size. Our calculation showed that the sample size investigated in our main analysis should provide > 80% statistical power to detect medium effect sizes (Δ β = 0.2; Additional file 7). However, large samples will be needed to investigate how epigenetic changes affect hATTR symptoms and differences across TTR amyloidogenic mutations. Our cohort showed age and sex differences between carriers of TTR amyloidogenic mutations and controls. Our analysis was adjusted for these confounding variables together with blood cell types, genetic principal components, and epigenetically determined smoking status. More balanced case-control groups are needed to investigate how epigenetic differences are associated differently between sexes and across age groups. We used transcriptomic data from GTEx project to explore the potential mechanisms related to the epigenetic associations identified. Further studies generating transcriptomic and epigenomic information across multiple informative tissues will provide a more comprehensive understanding of the molecular processes involved in hATTR. Although we provide some preliminary evidence showing that Val30Met mutation disrupts cg13139646 which appears co-methylated with CpG sites potentially associated with hATTR, our findings should be considered exploratory. Indeed, since Val-30Met disrupts the probe extension for the cg13139646 methylation site, the impact of this mutation on the methylation changes in TTR gene should be confirmed using alternative methods.

Conclusions
Our study provided novel insights regarding hATTR pathogenesis, supporting the involvement of methylation changes in the amyloidogenic process induced by TTR disease-causing mutations. Further studies will be needed to characterize specific mechanisms underlying the epigenetic associations with particular attention to the potential role of amyloid-beta metabolic process and inflammatory response. The understanding of how methylation changes modulate the penetrance and the severity of TTR mutations could lead to the identification of novel targets to develop treatments and screening tools for the carriers. Additionally, similarly to what was observed with respect to genetic variation [21], it will be important to estimate the epigenetic similarity between hATTR and wild-type transthyretin amyloidosis.

Methods
Thirty-eight symptomatic patients and 10 asymptomatic TTR mutations carriers were recruited from three Italian centers for the treatment of systemic amyloidosis: "San Giovanni Calibita" Fatebenefratelli Hospital, Isola Tiberina-Rome, Fondazione Policlinico Universitario "A. Gemelli"-Rome, and Careggi University Hospital-Florence [16][17][18][19][20]. Thirty-two controls were recruited by the Department of Biology-University of Rome "Tor Vergata" (Table 2). hATTR diagnosis was based on the presence of clinical signs and symptoms and the presence of an amyloidogenic mutation on TTR gene. One hATTR patient is a carrier of a mutation (rs36204272) in an intronic region with a putative clinical impact [61]. Carpal tunnel syndrome and cardiac involvement with confirmed TTR amyloid deposits are present in rs36204272 carrier. Information regarding the organ involvements was collected for each patient: peripheral and nerve involvement (nerve conduction study); cardiac involvement (electrocardiographic and echocardiography anomalies); gastrointestinal involvement (gastric paresis, stypsis, or diarrhea); autonomic neurological involvement (orthostatic hypotension and urinary incontinence); ocular involvement (vitreous opacities): and carpal tunnel syndrome (median nerve decompression) [11,[62][63][64]. In our previous analysis in this cohort [20], we observed that certain mutations were associated with specific clinical manifestations: Ile68Leu and Val122Ile with cardiac involvement; Val30Met with ocular symptoms; Phe64Leu with renal involvement. Conversely, autonomic neurological, peripheral neurological, and gastrointestinal symptoms were observed in most of the amyloidogenic mutations [20]. The present study was approved under the protocol 39/18 by the Comitato Etico Indipendente, Fondazione Policlinico Tor Vergata-Rome, Italy.

DNA methylation analysis
DNA was extracted using the phenol/chloroform protocol [65] and purified through Amicon Ultra-0.5 mL Centrifugal Filters (EMD Millipore) to achieve a DNA concentration of 100 ng/µL. DNA concentration was  checked via NanoDrop technology (ND-1000, Thermo Fisher Scientific) and Qubit Quantitation technology (High Accuracy & Sensitivity, Thermo Fisher). DNA methylation analysis was executed in two phases: the EZ DNA Methylation kit (Zymo Research) was used to perform sodium bisulfite conversion; the Illumina Infinium Methylation EPIC Chip (with over 850,000 methylation sites; Illumina Inc.) was used to quantify DNA methylation according to the standard Illumina protocol. The methylation array analysis was performed at the Connecting bio-research and Industry Center, Trieste-Italy.

Preprocessing, quality control, and normalization
The raw signal intensity files were processed and cleaned using R 3.6 and ChAMP package [66]. The ratio of methylated and unmethylated intensities from idat files was converted into beta-values for further processing. The probes failing thresholds on detection p value (< 0.01) bead count, sites near SNPs-this method performed extensive characterization of probes on the EPIC and HM450 microarrays, including mappability to the latest genome build, genomic copy number of the 3΄ nested subsequence, and influence of polymorphisms including a previously unrecognized color channel switch for Type I probes, probes that align to multiple positions, sex chromosomes and outliers were removed. None of the samples failed quality control. The remaining 718,509 probes for 80 individuals were normalized with BMIQ. Batch effects were assessed using singular vector decomposition and corrected with ComBat method [67]. The genomic lambda of the case-control association was 1.03, calculated using QQPerm package (https ://cran.rproje ct.org/web/packa ges/QQper m/index .html).

Blood cell type composition, genetic variability estimation, and smoking prediction
A reference-based method was employed to adjust for the heterogeneity due to the cell type composition of the whole blood samples investigated [68]. This method uses specific DNA methylation signatures derived from purified whole blood cell type as biomarkers of cell identity, to correct the beta-values. Cell proportions for five cell types (B cells, granulocytes, monocytes, natural killer cells, and T cells) were detected, and a linear regression was applied [66,68]. To account for the genetic variability among the samples investigated, principal components (PCs) were calculate using the method proposed by Barfield, Almli [69]. This approach allowed us to compute PCs based on CpGs selected for their proximity to SNPs. The data obtained can be used to adjust for population stratification in DNA methylation studies when genomewide SNP data are unavailable [69]. Cigarette smoke has a very large effect on DNA methylation profile, triggering alteration at multiple CpGs [70]. Consequently, smoking status needs to be considered as a potential confounder in epigenetic association studies. EpiSmokEr package was used to classify the smoking status of each participant on the basis of their epigenetic profile [71]. Briefly, EpiSmokEr is a prediction tool that provides smoking probabilities for each individual (never smoker, former smoker, and current smoker) using a set of 121 informative CpG sites [70].

Data analysis
We conducted an epigenome-wide analysis testing 718,509 methylation sites. First, we investigated the methylation changes (measured as M values, i.e., the log2 ratio of the intensities of methylated probe versus unmethylated probe; beta-values are plotted to provide more biologically interpretable data) between 48 carriers of a TTR amyloidogenic mutation) and 32 controls (i.e., non-carriers). Considering CpG sites that survive epigenome-wide multiple testing correction, we verified whether the associations observed were due to diseaseassociated genetic differences or post-disease processes, comparing (i) patients affected by hATTR vs. controls, (ii) asymptomatic carriers of TTR mutations vs. controls, and iii) patients affected by hATTR vs. asymptomatic carriers of TTR mutations. Additionally, we also tested whether the methylation changes observed in the casecontrol analysis were different between (i) Val30Met carriers vs. controls, (ii) carriers of other TTR mutations vs. controls and, iii) Val30Met patients vs. carriers of other TTR mutations.
To investigate the functionality of the CpG disrupted by Val30Met mutation, we analyzed its comethylation with CpG sites in the surrounding region (NC_000018.9: 28,171,000-30,171,500). This region was selected based on the TTR regulatory mechanisms observed in previous studies [20][21][22]. We used cor() R function to calculate Pearson's correlation coefficient testing 367 sites and considering the methylation levels (M values) using as reference the non-carriers. CpG sites with high co-methylation (Person's r 2 > 0.2) were investigated with respect to hATTR-related traits (i.e., carrier status, disease status, and symptoms). In all association analyses, we implemented a linear regression analysis including cell composition proportions, top three genetic PCs, epigenetically determined smoking status, age, and sex as covariates. The results of the regression analyses were reported as standardized regression coefficients and p values. FDR method [72] was applied to adjust the results for epigenome-wide testing and the q value < 0.05 was considered as the significance threshold. Co-expression analysis was conducted using GTEx v8 [29] via the Multi-Gene Query