Epigenetic signatures of Werner syndrome occur early in life and are distinct from normal epigenetic aging processes

Abstract Werner Syndrome (WS) is an adult‐onset segmental progeroid syndrome. Bisulfite pyrosequencing of repetitive DNA families revealed comparable blood DNA methylation levels between classical (18 WRN‐mutant) or atypical WS (3 LMNA‐mutant and 3 POLD1‐mutant) patients and age‐ and sex‐matched controls. WS was not associated with either age‐related accelerated global losses of ALU, LINE1, and α‐satellite DNA methylations or gains of rDNA methylation. Single CpG methylation was analyzed with Infinium MethylationEPIC arrays. In a correspondence analysis, atypical WS samples clustered together with the controls and were clearly separated from classical WS, consistent with distinct epigenetic pathologies. In classical WS, we identified 659 differentially methylated regions (DMRs) comprising 3,656 CpG sites and 613 RefSeq genes. The top DMR was located in the HOXA4 promoter. Additional DMR genes included LMNA, POLD1, and 132 genes which have been reported to be differentially expressed in WRN‐mutant/depleted cells. DMRs were enriched in genes with molecular functions linked to transcription factor activity and sequence‐specific DNA binding to promoters transcribed by RNA polymerase II. We propose that transcriptional misregulation of downstream genes by the absence of WRN protein contributes to the variable premature aging phenotypes of WS. There were no CpG sites showing significant differences in DNA methylation changes with age between WS patients and controls. Genes with both WS‐ and age‐related methylation changes exhibited a constant offset of methylation between WRN‐mutant patients and controls across the entire analyzed age range. WS‐specific epigenetic signatures occur early in life and do not simply reflect an acceleration of normal epigenetic aging processes.


| INTRODUC TI ON
Aging is a universal biological process, leading to an overall decline of organ functions, tissue homeostasis, and the ability to successfully respond to internal and external stresses, which takes place at highly different rates within members of a species and between species. Segmental progeroid syndromes are very rare monogenic human disorders showing clinical features of premature aging involving more than one tissue or organ (Martin, 1978). Werner syndrome (WS; OMIM 277700) is an autosomal recessive adult-onset segmental progeria that is characterized by ocular cataracts, scleroderma-like skin changes, subcutaneous calcification and ulceration, premature graying and loss of hair, short stature from the second decade of life, and an elevated risk for age-associated diseases such as atherosclerosis, diabetes mellitus, and osteoporosis. Cancer (especially sarcomas) and myocardial infarction are the leading causes for early death at an average age of 54 years.
More than 80 different homozygous or compound heterozygous mutations in the Werner syndrome (WRN) gene have been associated with WS (Yokote et al., 2017). The WRN protein is a member of the RecQ family of helicases possessing both 3'->5' DNA helicase and 3'->5' exonuclease activities (Yu et al., 1996). Cells from WS patients show a prolonged S phase of the cell cycle, hypersensitivity to agents causing DNA crosslinks and double-strand breaks, elevated frequencies of micronuclei, and a reduction in recombinational double-strand break repair (Dhillon et al., 2007;Poot, Jin, Hill, Gollahon, & Rabinovitch, 2004). WS fibroblasts show a limited proliferative lifespan due to clonal attenuations and successions, a variegated chromosomal translocation mosaicism, and multiple spontaneous deletions (Fukuchi, Martin, & Monnat, 1989;Salk, Au, Hoehn, & Martin, 1981). It has therefore been hypothesized that WRN plays a role in the resolution of potentially damaging, complex DNA structures accidentally formed during DNA replication, recombination, repair, and transcription as well as in preventing chromothripsis (Poot, 2018). Atypical WS is characterized by a WS-like phenotype without WRN mutations. Some patients with atypical WS carry heterozygous mutations in the lamin A (LMNA) gene (Chen et al., 2003). These nuclear intermediate filaments are major structural components of the mammalian nuclear lamina, contributing to nuclear shape, mechanical stability, nuclear assembly, and positioning. They are involved in chromatin organization, transcription regulation, and DNA replication (Mattout, Dechat, Adam, Goldman, & Gruenbaum, 2006). Ten to 15% of patients initially diagnosed with WS displayed mutations neither in WRN nor in LMNA. A subset of them presented with mandibular hypoplasia, deafness, and progeroid features (MDPL syndrome) and heterozygous mutations in the polymerase delta 1 (POLD1) gene (Weedon et al., 2013). POLD1 has both DNA polymerase and 3'->5' exonuclease activities and is involved in DNA synthesis of the lagging strand, mismatch repair, and resolution of DNA replication-blocking structures. It functionally and physically interacts with WRN during DNA replication and repair (Kamath-Loeb, Shen, Schmitt, & Loeb, 2012). Skin fibroblasts of a MDPL patient exhibited increased fractions of senescent markers and persistent DNA damage after genotoxic treatment (Fiorillo et al., 2018).
Although in most segmental progerias the underlying mutations are known, the molecular mechanisms causing a plethora of aginglike phenotypes remain to be elucidated. Impairment of genome stability explains some symptoms; however, other mechanisms, in particular epigenetic dysregulation, may also play important roles, as WS fibroblasts and WRN-depleted cells show extensive alterations of gene expression (Cheung et al., 2014;Kyng, May, Kolvraa, & Bohr, 2003;Zhang et al., 2015). The most thoroughly studied epigenetic modification is DNA methylation at the carbon 5' atom of cytosine, mainly in the context of CpG dinucleotides. Methylation of CpG islands, which are present in the promoter and/or first exon of most mammalian genes, leads to an inactive chromatin structure and gene silencing during development, differentiation, and disease. In contrast, gene body methylation is usually associated with active genes (Jones, 2012). Methylated CpGs are enriched in repetitive DNA elements to prevent retrotransposition and to maintain genome integrity (Yoder, Walsh, & Bestor, 1997). There is to date only limited information on DNA methylation patterns associated with premature aging diseases. One methylation array study (Heyn, Moran, & Esteller, 2013) compared lymphoblasts of four patients with WS (two with WRN, one with LMNA, and one without a known mutation) and three related nonmutant patients with Hutchinson-Gilford progeria with normal lymphoblasts, naive B cells, and peripheral blood mononuclear cells. The samples with WRN and LMNA mutations clustered together and were distinct from nonmutant patients and controls.
A conceptually related study (Guastafierro et al., 2017) found profound blood methylation differences between three classical WS patients and controls; however, these results were not statistically significant. We now report a more comprehensive methylome analysis of 24 independent patients with segmental progeria (18 with WRN, 3 with LMNA, and three with POLD1 mutations) together with carefully matched controls.

| Global DNA methylation of repetitive elements in WS
Aging is associated with hypo-and hypermethylation events at specific regions of the genome (Unnikrishnan et al., 2018). Recently, we showed that the methylation of various repeat families decreased, whereas that of rDNA increased during in vitro aging of fibroblast clones (Flunkert et al., 2018). Here, we used the same bisulfite pyrosequencing assays to quantify mean methylation of ALU, LINE1, and α-satellite repeats in WRN-, LMNA-, and POLD1-mutant patients and controls (Table 1). Both interspersed repeats and centromeric α-satellite DNA showed almost identical methylation levels in WS patients and controls. Methylations of the rDNA promoter and upstream control element were increased by 1-2 percentage points in WRN-and POLD1-mutant patients and decreased by approximately 5% in LMNA-mutant patients (Table 1); however, due to the large interindividual variations these results were not significant (Table S1).

| Differentially methylated sites and regions in WS
Infinium MethylationEPIC BeadChips were used to compare genome-wide DNA methylation patterns at a single CpG level between WS patients and controls. After initial filtering, 816,980 probes were included in the analysis. Although estimations of the proportions of blood cell types did not reveal significant differences between patient and control samples ( Figure S1), exploratory analyses clearly indicated cell composition as a major factor explaining array methylation variation ( Figure S2). Therefore, these scores were included into the linear model. A correspondence analysis of the 10,000 most variable methylation sites over all 48 samples clearly separated  (Table 3), including one biological process for intercellular signal transduction and three molecular functions related to transcription factor activity and sequence-specific DNA binding. Analysis of a much smaller published EPIC array data set (GSE10 0825) using our bioinformatics pipeline did not yield significantly differentially methylated CpGs. However, the β differences between WS patients and controls correlated well between the published study (Guastafierro et al., 2017) and our study (r = 0.182, p < 0.001), consistent with a common signal in both data sets.
From previous studies (Table S3), we obtained six partially overlapping lists of genes that were differentially expressed between WS patient-derived or WRN-depleted cells (mainly fibroblasts) and controls. 132 of our 613 DMR genes showed differential expression in at least one, 36 in at least two, and seven in at least three studies. Expression of SGK1 was consistently upregulated, whereas DEPDC1, E2F8, HIST1H1A, POLD1, SMC4, and PKMYT1 were transcriptionally downregulated in classical WS.
In LMNA-mutant patients, we identified 72 genome-wide significant DMRs encompassing 399 CpG sites and 67 RefSeq genes ( Table   S2b). The top promoter DMR was associated with the CARNS1 gene

| Age-related methylation changes
Both a site-wise and a region-wide analysis based on site-wise p values demonstrated that age was one of the strongest contributing factors in the data set. Of 812,996 interrogated CpG sites, 13,616 (1.7%) showed genome-wide significant age-related methylation changes, among them 241 sites with WRN-specific methylation signatures ( Figure 2b). Although this is a fourfold enrichment (Fisher's exact test, p < 0.001), the vast majority (>90%) of differentially methylated CpGs in WS did not show age effects. There were no sites showing a significant difference in DNA methylation changes with age between WS patients and controls.
Altogether, 1,340 genes were endowed with age-related DMRs (Table S2d) HOXA4, the top DMR in WRN-mutant patients, the correlation between promoter methylation and age was borderline significant F I G U R E 2 Venn diagrams showing the overlaps of genome-wide significant CpG sites (a, b) and DMR-containing genes (c, d) between WRN-, LMNA-, and POLD1mutant patients (a, c) as well as between WS-and age-related changes (b, d)  stress that is involved in the pathogenesis of many human diseases including cancer, and metabolic and cardiovascular disorders . Impaired ribosome biogenesis (Rattan, 1996) and rDNA hypermethylation (Flunkert et al., 2018) have also been linked to aging. During in vitro aging of fibroblasts, rDNA hypermethylation was reported to be more pronounced in two WS patients, compared with four controls (Machwe, Orren, & Bohr, 2000 TA B L E 2 (Continued)

| Is classical WS a primary transcription disease?
With a prevalence of 1/50,000 (in Japan and Sardinia) to 1/200,000 (in most populations worldwide), WS is a very rare disease. Due to our increased sample sizes compared to earlier methylation array analyses (Guastafierro et al., 2017;Heyn et al., 2013), we identified genome-wide significant DMRs in 613 RefSeq genes, the majority (78%) of which were hypermethylated in WS. The observed methylation differences were of the order of several percentage points

| Differentially methylated genes in classical WS
Our top DMR (Table 2, Figure 3a) covering 25 CpGs in the promoter-associated region of the homeobox A4 (HOXA4) gene was hypermethylated (on average by 8 percentage points) in blood of WS patients. It belongs to a family of homeodomain-containing transcription factors that play an important role during developmental processes and hematopoietic differentiation. HOXA4 promoter hypermethylation and reduced expression have been linked to acute leukemias and a shorter survival (Strathdee et al., 2007).

Hypomethylation of the HOXA4 promoter was associated with
Silver-Russell syndrome and patients with severe growth retardation of unknown etiology, suggesting a role for HOXA4 in growth regulation (Muurinen et al., 2017). These genes have been linked to key cellular processes (transcription, replication, repair, cell cycle progression, chromosome condensation, stress response, apoptosis, and senescence), genome stability, and cancer; however, their possible relationship with premature aging is presently unclear.

| WS and atypical WS are epigenetically distinct disorders
Despite small sample sizes, 67 and 37 genes with genome-wide significant DMRs, respectively, were associated with LMNA and  (Wilson, Cheung, Martindale, Scherer, & Koop, 2006). PILRB was hypermethylated in WRN-and LMNA-mutant blood, whereas in Alzheimer brain samples, it was found to be hypomethylated (Humphries et al., 2015). The active BCR-related (ABR) protein interacts with members of the RhoGTP-binding protein family, regulating cellular signaling (Chuang et al., 1995). ABR was hypermethylated in blood of WRN-and POLD1-mutant patients and downregulated in WS fibroblasts (Kyng et al., 2003).
In a correspondence analysis, LMNA-and POLD1-mutant patients

| WS and normal aging
The "epigenetic clock" is a DNA methylation-based biomarker of aging that is defined as a weighted average across several hundred CpG sites (Horvath, 2013). The resulting epigenetic age estimate can predict lifespan and has been found to be increased in Alzheimer disease and other age-related conditions Maierhofer et al., 2017). We previously reported an epigenetic age acceleration in blood of adult-onset WS patients (Maierhofer et al., 2017) and in fibroblasts of childhood-onset Hutchinson-Gilford progeria syndrome patients . In contrast to the epigenetic clock, which is based on a highly selected subset (<0.05%) of array CpGs, we interrogated 816,980 CpGs representing the entire epigenome. In our data set, only 241 of 13,616 CpGs and 117 of 1,340 genes with age-related methylation changes exhibited significant methylation differences between WS and controls. The vast majority of CpGs (3,629 of 3,870; 94%) and genes (496 of 613; 81%) with WS-specific epigenetic signatures were not affected by aging.
There was not a single CpG site showing a significant difference in methylation change with age between WS patients and controls.
Since aging is a slow and highly multifactorial process, one  percentage points per year). The transcription factors ZIC1 and ZIC4 are involved in brain development (Aruga & Millen, 2018) and  . Methylation changes preceding disease manifestations argue in favor of a causal relationship. We propose that epigenetic misregulation of downstream genes transcribed by RNAPII contributes to disease onset and premature aging symptoms in WS and may not be merely a secondary phenomenon. In addition to genome instability, which is a hallmark of WS, cancer, and aging, WRN may play an important role in epigenetic maintenance systems (Maierhofer et al., 2017;Zhang et al., 2015), affecting methylation patterns of hundreds of downstream genes.
Collectively, our data suggest that the WS epigenome(s) is shaped early in life by processes which are largely distinct from normal epigenetic aging. Confirmations of that conclusion, however, will require comparable investigations of gene regulation during earlier stages of life.

| Bisulfite pyrosequencing
ALU, LINE1, and α-satellite DNA repeats were first amplified in a multiplex PCR, followed by second-round nested PCRs for each repeat.
Two amplicons of the rDNA promoter were amplified separately, region 1 covering the distal rDNA promoter and region 2 the core promoter element and the upstream control element. Primer sequences and PCR conditions have been published previously (Flunkert et al., 2018). Bisulfite pyrosequencing was done on a PyroMark Q96MD pyrosequencing system (Qiagen) using the PyroMark Gold Q96 CDT reagent kit (Qiagen) and the Pyro Q-CpG software (Qiagen). Overall methylation differences were modeled by a linear model adjusting for patient age (Table S1). At the individual CpG level, methylation differences between patients and controls were modeled using a linear model in the limma framework (Ritchie et al., 2015), adjusting for age, gender, and cell composition. The scores of the first two axes of a correspondence analysis on the estimated cell composition were included in the linear model to account for differences in cell composition. In a second step, we additionally introduced and tested an interaction term between methylation and age. Here, we investigated whether we could detect a different linear association (i.e., different slope) of age and methylation percentage between the classical WS and control groups. In contrast to both covariates (WS/CTRL or age) alone, the effect of the interaction term between aging and WS showed only a very weak signal on a genome-wide scale. After multiple testing corrections, there were no significant sites left.

| DNA methylation arrays
To derive DMRs from probewise p values, we used the approach implemented in the comb-p package (Pedersen, Schwartz, Yang, & Kechris, 2012). First, a Stouffer-Liptak-Kechris (SLK)-corrected p value for each probe was calculated based on the autocorrelation on neighboring p values. In a second step, regions enriched with SLK-corrected p values were identified by a peak-finding algorithm. Finally, the significance of each identified region was then determined by applying a Stouffer-Liptak correction to the original p values of all probes in the region. To correct for multiple testing, a Sidak correction, based on the number of possible regions of the same size, was applied to all identified regions. A region was extended if another p value within a genomic distance of 1,000 nucleotides was found (dist = 1,000). Sites with p < 0.05 (seed = 0.05) were considered as a starting point for a potential region.
Functional relevance of the genes covered by DMRs was analyzed using the Database for Annotation, Visualization and Integrated Discovery (DAVID), version 6.8 (https ://david.ncifc rf.gov/). p Values for enrichment were calculated using Fisher's exact test and corrected for multiple testing with the Benjamini-Hochberg procedure.

ACK N OWLED G M ENTS
This work was supported by the grants, NIH/NCI R01CA210916 (to G.M.M. and J.O.) and JSPS KAKENHI 17H04037 (to J.O.).

CO N FLI C T O F I NTE R E S T
None declared.