GR243584Jef 1057..1066 Growth disrupting mutations in epigenetic regulatory molecules are associated with abnormalities of epigenetic aging Aaron R. Jeffries,1 Reza Maroofian,2 Claire G. Salter,1,3,4 Barry A. Chioza,1 Harold E. Cross,5 Michael A. Patton,1,2 Emma Dempster,1 I. Karen Temple,3,4 Deborah J.G. Mackay,3 Faisal I. Rezwan,3 Lise Aksglaede,6 Diana Baralle,3,4 Tabib Dabir,7 Matthew F. Hunter,8,13 Arveen Kamath,9 Ajith Kumar,10 Ruth Newbury-Ecob,11 Angelo Selicorni,12 Amanda Springer,8,13 Lionel Van Maldergem,14 Vinod Varghese,9 Naomi Yachelevich,15 Katrina Tatton- Brown,2,16,17 Jonathan Mill,1 Andrew H. Crosby,1 and Emma L. Baple1,18 1Institute of Biomedical and Clinical Science, University of Exeter Medical School, RILD Wellcome Wolfson Centre, Royal Devon and Exeter NHS Foundation Trust, Exeter, EX2 5DW, United Kingdom; 2Genetics Research Centre, Molecular and Clinical Sciences Institute, St. George’s University of London, London SW17 0RE, United Kingdom; 3Human Genetics and Genomic Medicine, Faculty of Medicine, University of Southampton, Southampton, SO16 6YD, United Kingdom; 4Wessex Clinical Genetics Service, Princess Anne Hospital, Southampton, SO16 5YA, United Kingdom; 5Department of Ophthalmology and Vision Science, University of Arizona School of Medicine, Tucson, Arizona 85711, USA; 6Department of Clinical Genetics, Copenhagen University Hospital, Blegdamsvej 3B, 2200 Copenhagen N, Denmark; 7Northern Ireland Regional Genetics Centre, Clinical Genetics Service, Belfast City Hospital, Belfast, BT9 7AB, United Kingdom; 8Monash Genetics, Monash Health, Clayton, Victoria, VIC 3168, Australia; 9Institute of Medical Genetics, University Hospital of Wales, Cardiff, CF14 4XN, United Kingdom; 10North East Thames Regional Genetics Service and Department of Clinical Genetics, Great Ormond Street Hospital, London, WC1N 3JH, United Kingdom; 11University Hospitals Bristol, Department of Clinical Genetics, St Michael’s Hospital, Bristol, BS2 8EG, United Kingdom; 12UOC Pediatria ASST Lariana, Como, Italy; 13Department of Paediatrics, Monash University, Clayton, Victoria, VIC 3168, Australia; 14Centre de génétique humaine and Clinical Investigation Center 1431 (INSERM), Université de Franche-Comté, 25000, Besançon, France; 15Clinical Genetics Services, New York University Hospitals Center, New York University, New York, New York 10016, USA; 16Division of Genetics and Epidemiology, Institute of Cancer Research, London SM2 5NG, United Kingdom; 17South West Thames Regional Genetics Service, St. George’s University Hospitals NHS Foundation Trust, London SW17 0QT, United Kingdom; 18Peninsula Clinical Genetics Service, Royal Devon and Exeter Hospital, Exeter, EX1 2ED, United Kingdom Germline mutations in fundamental epigenetic regulatory molecules including DNA methyltransferase 3 alpha (DNMT3A) are commonly associated with growth disorders, whereas somatic mutations are often associated with malignancy. We pro- filed genome-wide DNA methylation patterns in DNMT3A c.2312G > A; p.(Arg771Gln) carriers in a large Amish sibship with Tatton-Brown–Rahman syndrome (TBRS), their mosaic father, and 15 TBRS patients with distinct pathogenic de novo DNMT3A variants. This defined widespread DNA hypomethylation at specific genomic sites enriched at locations annotated as genes involved in morphogenesis, development, differentiation, and malignancy predisposition pathways. TBRS patients also displayed highly accelerated DNA methylation aging. These findings were most marked in a carrier of the AML-asso- ciated driver mutation p.Arg882Cys. Our studies additionally defined phenotype-related accelerated and decelerated epi- genetic aging in two histone methyltransferase disorders: NSD1 Sotos syndrome overgrowth disorder and KMT2D Kabuki syndrome growth impairment. Together, our findings provide fundamental new insights into aberrant epigenetic mecha- nisms, the role of epigenetic machinery maintenance, and determinants of biological aging in these growth disorders. [Supplemental material is available for this article.] DNA methylation is an essential epigenetic process involving the addition of a methyl group to cytosine. It is known to play a role in many important genomic regulatory processes, including X-Chromosome inactivation, genomic imprinting, and the repres- sion of tumor suppressor genes in cancer, mediating transcription- al regulation as well as genomic stability (Jones 2012). Three catalytically active DNA methyltransferases (DNMTs) are involved Corresponding authors: e.baple@exeter.ac.uk, a.h.crosby@exeter.ac.uk, j.mill@exeter.ac.uk Article published online before print. Article, supplemental material, and publi- cation date are at http://www.genome.org/cgi/doi/10.1101/gr.243584.118. Freely available online through the Genome Research Open Access option. © 2019 Jeffries et al. This article, published in Genome Research, is available un- der a Creative Commons License (Attribution 4.0 International), as described at http://creativecommons.org/licenses/by/4.0/. Research 29:1057–1066 Published by Cold Spring Harbor Laboratory Press; ISSN 1088-9051/19; www.genome.org Genome Research 1057 www.genome.org mailto:e.baple@exeter.ac.uk mailto:a.h.crosby@exeter.ac.uk mailto:j.mill@exeter.ac.uk http://www.genome.org/cgi/doi/10.1101/gr.243584.118 http://www.genome.org/cgi/doi/10.1101/gr.243584.118 http://genome.cshlp.org/site/misc/terms.xhtml http://creativecommons.org/licenses/by/4.0/ http://creativecommons.org/licenses/by/4.0/ http://genome.cshlp.org/site/misc/terms.xhtml in the methylation of cytosine: DNMT1, which is mainly respon- sible for the maintenance of DNA methylation over replication, and DNMT3A and DNMT3B, which generally perform de novo methylation of either unmethylated or hemimethylated DNA. An absence of these enzymes in mice results in embryonic (DNMT1 and 3B) or postnatal (DNMT3A) lethality (Okano et al. 1999), confirming their essential roles in development. In line with knockout mouse models, pathogenic variants affecting the chromatin binding domains of DNMT1 have been shown to cause two separate progressive autosomal dominant adult-onset neuro- logic disorders (Klein et al. 2011). Biallelic pathogenic variants in DNMT3B have been associated with immunodeficiency, centro- mere instability, and facial anomalies (ICF) syndrome (Jiang et al. 2005). To date, DNMT3A has been linked to a number of physiological functions, including cellular differentiation, malig- nant disease, cardiac disease, learning, and memory formation. Somatically acquired pathogenic variants in DNMT3A are associat- ed with >20% of acute myeloid leukemia (AML) cases, whereas het- erozygous germline pathogenic loss-of-function variants have been found to underlie Tatton-Brown–Rahman syndrome (TBRS; also known as DNMT3A-overgrowth syndrome, OMIM 615879) (Challen et al. 2011; Tatton-Brown et al. 2014). TBRS is character- ized by increased growth, intellectual disability (ID), and dysmor- phic facial features. More recently, heterozygous gain-of-function DNMT3A missense variants affecting the DNMT3A PWWP domain have been shown to cause microcephalic dwarfism and hypermethylation of Polycomb-regulated regions (Heyn et al. 2019). There is an emerging group of epigenetic regulatory mole- cule-associated human growth disorders in which the underlying molecular defect is a disruption to the DNA methylation and his- tone machinery. There are now over 40 disorders identified within this group, which can be further subgrouped into diseases result- ing from disruption of the “writers,” “readers,” and “erasers” of epigenetic modifications (Bjornsson 2015). Example disorders in each group include Kabuki, Sotos, and Weaver syndromes (“writ- ers”); Smith-Magenis, Rett, and Bohring–Opitz syndromes (“read- ers”); and Wilson–Turner and Cleas–Jensen syndromes (“erasers”). The final subgroup occurs because of disruption of chromatin remodelers, with example resulting disorders including CHARGE and Floating–Harbor syndromes. Neurological and cognitive impairment are common features of these conditions, suggesting that precise epigenetic regulation may be critical for neuronal homeostasis. However, a true understanding of the pathogenic mechanism underlying these conditions remains poorly understood. In the current study, we investigated the methylomic conse- quences of a DNMT3A pathogenic variant (NC_000002.12: g.25240312C > T; NM_022552.4:c.2312G > A; p.(Arg771Gln)) in a large Amish family comprising four individuals affected with TBRS arising as a result of a mosaic pathogenic DNMT3A variant in their father (Xin et al. 2017). The occurrence of multiple affected and unaffected individuals in the same sibship, together with the combined genetic and environmental homogeneity of the Amish, permitted an in-depth investigation of the genome-wide patterns of DNA methylation associated with pathogenic variation in DNMT3A. We subsequently extended our analyses to other (non-Amish) TBRS patients harboring distinct pathogenic de novo DNMT3A variants, as well other methyltransferase-associat- ed overgrowth and growth deficiency syndromes, defining altered epigenetic profiles as common key themes of these growth disorders. Results Reduced DNA methylation at key sites involved in morphogenesis, development, and differentiation in TBRS patients DNMT3A encodes a DNMT with both de novo and maintenance activity (Okano et al. 1999; Chen et al. 2003). We first looked for global changes in DNA methylation in whole blood obtained from DNMT3A c.2312G > A; p.(Arg771Gln) carriers, using the methylation-sensitive restriction enzyme–based luminometric methylation assay (LUMA) (Karimi et al. 2006) to quantify DNA methylation across GC-rich regions of the genome, finding no ev- idence for altered global DNA methylation (LUMA: mean DNMT3A c.2312G > A carriers = 0.274, wild type = 0.256, t-test P- value = 0.728). We next quantified DNA methylation at 414,172 autosomal sites across the genome using the Illumina 450K array. Globally, a subtle decrease in mean DNA methylation was noted in available age/sex-matched DNMT3A heterozygous c.2312G > A; p.(Arg771Gln) individuals compared with their matched unaffect- ed sibling samples, although this was not statistically significant (Wilcoxon rank-sum test P-values for two matched pairs = 0.24 and 0.14) (Supplemental Fig. S1). In contrast, an analysis of site- specific DNA methylation differences in DNMT3A c.2312G > A; p.(Arg771Gln) carriers (including the mosaic father) versus wild- type individuals in the Amish pedigree identified 2606 differen- tially methylated positions (DMPs; Benjamini–Hochberg false discovery rate [FDR] < 0.05) (Fig. 1A,B; Supplemental Table S1), of which 1776 DMPs were characterized by a >10% change in DNA methylation. Supplemental Figure S2 also highlights DNA methyl- ation levels at these DMPs across all carriers and control individu- als profiled in this study. Technical validation of Illumina 450K array data was performed using bisulfite pyrosequencing for three top-ranking DMPs, confirming significant differences in DNMT3A c.2312G > A; p.(Arg771Gln) carriers at each of the tested loci (Supplemental Fig. S3). The DMPs identified were highly enriched for sites character- ized by reduced DNA methylation in DNMT3A c.2312G > A; p.(Arg771Gln) heterozygotes (n = 2576 DMPs, 98.85%, sign-test P-value <2.2 × 10−16). Although there were no statistically signifi- cant differences between DNA methylation-based blood cell com- position estimates derived from our data (Supplemental Table S2), we examined the extent to which the identified DMPs were poten- tially influenced by cell-type differences between DNMT3A c.2312G > A; p.(Arg771Gln) carriers and wild-type family mem- bers. There was a highly significant correlation (r = 0.876, P-value <2.2 × 10−16) (Supplemental Fig. S4) in effect sizes at the 2606 DMPs between models, including and excluding cell types as co- variates, indicating that the observed patterns of differential DNA methylation are not strongly influenced by cell-type varia- tion. We used DMRcate (Peters et al. 2015) to identify spatially cor- related regions of differential DNA methylation significantly associated with the DNMT3A c.2312G > A; p.(Arg771Gln) variant, identifying 388 autosomal differentially methylated regions (DMRs) (for an example DMR, see Supplemental Fig. S5), all characterized by hypomethylation in DNMT3A c.2312G > A; p.(Arg771Gln) carriers apart from one 739-bp DMR that showed increased DNA methylation (Supplemental Table S3). The mean size of the identified DMRs was 625 bp (range = 6–5522 bp), span- ning an average of six probes (Supplemental Fig. S6). We next investigated whether DNMT3A p.(Arg771Gln)-asso- ciated DMPs are enriched in specific genic locations (see Methods). We found a modest enrichment of DMPs in regions ≥1500 bp Jeffries et al. 1058 Genome Research www.genome.org http://genome.cshlp.org/lookup/suppl/doi:10.1101/gr.243584.118/-/DC1 http://genome.cshlp.org/lookup/suppl/doi:10.1101/gr.243584.118/-/DC1 http://genome.cshlp.org/lookup/suppl/doi:10.1101/gr.243584.118/-/DC1 http://genome.cshlp.org/lookup/suppl/doi:10.1101/gr.243584.118/-/DC1 http://genome.cshlp.org/lookup/suppl/doi:10.1101/gr.243584.118/-/DC1 http://genome.cshlp.org/lookup/suppl/doi:10.1101/gr.243584.118/-/DC1 http://genome.cshlp.org/lookup/suppl/doi:10.1101/gr.243584.118/-/DC1 http://genome.cshlp.org/lookup/suppl/doi:10.1101/gr.243584.118/-/DC1 http://genome.cshlp.org/lookup/suppl/doi:10.1101/gr.243584.118/-/DC1 upstream of the transcriptional start site (chi-squared Yates-correct- ed P-value = 0.047) and more prominent enrichment in intergenic regions (chi-squared Yates corrected P-value = 1.54 × 10−14) (Sup- plementalFig.S7).DMPswerealsosignificantlyenrichedinCpGis- land shore regions (chi-squared Yates-corrected P-value = 8 × 10−30) (Supplemental Fig. S7). We also examined DMP occurrence in ex- perimentally determined cancer and reprogramming-specific DMR locations (Doi et al. 2009), finding a 2.4-fold and B A C D E Figure 1. TBRS DNMT3A variants are associated with widespread DNA hypomethylation. (A) Simplified pedigree indicating the genotyping of individuals in the Amish family investigated: (+/−) heterozygous carriers of the DNMT3A c.2312G > A p.(Arg771Gln) variant; (+/− Mosaic) the DNMT3A c.2312G > A p.(Arg771Gln) mosaic father; (−/−) wild-type individuals. Black shading indicates individuals with a phenotype consistent with TBRS, gray shading, the father with macrocephaly and mild intellectual impairment; and white shading, unaffected individuals. Each of these samples was profiled on the Illumina 450K DNA methylation array. (B) Volcano plot showing site-specific DNA methylation differences (x-axis) and −log10 P-values (y-axis) from an analysis comparing Amish DNMT3A c.2312G > A; p.(Arg771Gln) pathogenic variant carriers and wild-type family members using the Illumina 450K array. Red values indicate the 2606 differentially methylated positions (DMPs) detected at a Benjamini–Hochberg FDR < 0.05. (C) Top 20 Gene Ontology enrich- ment analysis categories associated with the 2606 DMPs identified in DNMT3A c.2312G > A; p.(Arg771Gln) pathogenic variant carriers versus wild-type family members. (D) Comparison of DNMT3A c.2312G > A; p.(Arg771Gln) identified DMPs (log2 fold change) relative to other DNMT3A TBRS-associated variants assessed in this study (all variants grouped and measured relative to controls). Pearson correlation coefficient = 0.6620, P-value <2.2 × 10−16. (E) Boxplot illustrating the DNA methylation changes observed in association with the DNMT3A TBRS variants studied at the DMPs identified in the Amish DNMT3A c.2312G > A p.(Arg771Gln) carriers. The predicted protein consequence of each DNMT3A variant studied is indicated: Pink indicates in-frame deletion; yellow, single-nucleotide variant; cyan, duplications predicted to result in a frameshift; green, Amish c.2312G > A; p.(Arg771Gln) variant. Epigenetic aging abnormalities in growth disorders Genome Research 1059 www.genome.org http://genome.cshlp.org/lookup/suppl/doi:10.1101/gr.243584.118/-/DC1 http://genome.cshlp.org/lookup/suppl/doi:10.1101/gr.243584.118/-/DC1 http://genome.cshlp.org/lookup/suppl/doi:10.1101/gr.243584.118/-/DC1 4.7-fold overrepresentation (chi-squared Yates-corrected P-value = 4.24 × 10−6 and chi-squared Yates-corrected P-value <1.3 × 10−41, respectively), as well as predicted enhancer elements that showed a 1.4-fold overrepresentation (chi-squared Yates-corrected P-value = 7.57 × 10−16). We then undertook Gene Ontology analysis, ac- countingfor the background distribution of probes on the Illumina 450K array, to functionally annotate the DNA methylation differences observed in the DNMT3A c.2312G > A; p.(Arg771Gln) carriers. The 2606 DMPs identified in this study showed a significant overrepresentation in functional pathways related to morphogenesis, development, and differentiation (top hit GO:0007275, multicellular organism development, contains 474 genes associated with DMPs; FDR Q-value = 3.7 × 10−7) (Fig. 1C; Supplemental Table S4). We also performed a functional overlap analysis to identify cell- or tissue-specific chromatin signals associ- atedwiththeseDMPsusingeFORGE(Breezeetal.2016).Significant overlap (FDR Q-value < 0.01) was found with DNase I sensitivity hotspots, most apparent with pluripotent cells in ENCODE (The ENCODE Project Consortium 2012; Davis et al. 2018) and fetal tis- sues within the NIH Roadmap Epigenomics Consortium data set (Supplemental Data S1; Roadmap Epigenomics Consortium et al. 2015). Chromatin states fromthe NIH Roadmap Epigenomics Con- sortium data set show an enrichment of DMPs in regions defined as active transcriptional start sites in brain tissue and embryonic stem cells. Of particular interest, given the established importance of DNMT3A during embryonic development, eFORGE analysis of blood cell types highlighted an enrichment of DMPs in regions characterized by repressed Polycomb and enhancer activity. To provide additional evidence to support the notion that DNMT3A c.2312G > A; p.(Arg771Gln) carriers show disruption to developmental pathways, we used the Genomic Regions Enrichment of Annotation Tool (GREAT) (McLean et al. 2010) to explore functional pathways enriched in genes annotated to DNMT3A c.2312G > A; p.(Arg771Gln)–associated DMRs. This re- vealed a significant effect on genes implicated in developmental pathways (first ranked GO biological process = skeletal system de- velopment, fold enrichment = 2.5, binomial FDR Q-value = 9.22 × 10−7), with a specific enrichment for Homeobox protein domain encoding genes (InterPro; fold enrichment = 239.59, binomial FDR Q-value = 4.15 × 10−23), fundamental for normal developmen- tal processes. An enrichment for malignancy terms was also noted (from the Molecular Signatures Database; first ranked term = Genes with promoters occupied by PML-RARA fusion protein in acute promyelocytic leukemia [APL] cells NB4 and two APL primary blasts, based on ChIP-seq data, fold enrichment = 3.14, binomial FDR Q-value = 6.18 × 10−7) (see Supplemental Data S2; Liberzon et al. 2011). To establish whether these DMPs are a consistent feature of TBRS, we profiled a further 15 non-Amish patients carrying dis- tinct previously published DNMT3A pathogenic variants (Table 1; Fig. 2) using the Illumina EPIC DNA methylation array. Examination of the DMPs identified in DNMT3A c.2312G > A; p.(Arg771Gln) carriers revealed that the majority of DMPs were common to all of the TBRS patients regardless of the underlying causative DNMT3A variant (Fig. 1D), with a Pearson correlation co- efficient of 0.6620 (P-value <2.2 × 10−16) for effect sizes across all DMPs. Each variant showed some heterogeneity in effect size (Fig. 1E), with DNMT3A c.2644C > T p.(Arg882Cys) associated with the greatest overall changes in DNA methylation. This data leads us to conclude that TBRS patients show loss of methylation at sites annotated to key genes involved in development and growth pathways, mirroring the well-characterized overgrowth and neurocognitive features that characterize this disorder. DNMT3A mutations are associated with highly accelerated epigenetic aging, particularly the cardinal AML driver mutation p.Arg882Cys DNA methylation at a specific set of CpG sites, representing a so-called “epigenetic clock,” has been shown to be strongly Table 1. TBRS DNMT3A variants are associated with epigenetic age acceleration ID Nucleotide change Protein change Chronological age (yr) Epigenetic age (yr) Epigenetic age acceleration (fold change) Epigenetic age acceleration (percentage increase) Single nucleotide variants 1 c.929T > A p.(Ile310Asn) 9.27 24.2 2.61 161% 2 c.1594G > A p.(Gly532Ser) 5.92 22.7 3.83 283% 3 c.1645T > C p.(Cys549Arg) 9.36 25 2.67 167% 4 c.1943T > C p.(Leu648Pro) 19.34 32.1 1.66 66% 5 c.2099C > T p.(Pro700Leu) 13.45 32.5 2.42 142% 6 c.2245C > T p.(Arg749Cys) 13.87 19.6 1.41 41% 7 c.2246G > A p.(Arg749His) 8.9 33.6 3.78 278% 8 c.2312G > A p.(Arg771Gln) 8–23 15.8–36.1 1.41 41%a 9 c.2512A > G p.(Asn838Asp) 14.44 38.8 2.69 169% 10 c.2644C > T p.(Arg882Cys) 2.28 21.9 9.61 861% 11 c.2705T > C p.(Phe902Ser) 9.84 25.4 2.58 158% 12 c.2711C > T p.(Pro904Leu) 7.78 35.7 4.59 359% In-frame deletions 13 c.889_891delTGG p.(Trp297del) 5.82 23.2 3.99 299% 14 c.2255_2257delTCT p.(Phe752del) 3.05 10.6 3.48 248% Duplications resulting in a frameshift 15 c.2297dupA p.(Arg767fs) 3.12 14.4 4.62 362% 16 c.934_937dupTCTT p.(Ser312fs) 21 38.9 1.85 85% DNMT3A genotype (p.(Arg771Gln), shown in blue; p.(Arg882Cys), shown in red), chronological age, predicted epigenetic age, and percentage of age acceleration calculated for TBRS syndrome cases included in this study. aEpigenetic age acceleration taken from the linear regression model applied to four individuals carrying the mutation. Jeffries et al. 1060 Genome Research www.genome.org http://genome.cshlp.org/lookup/suppl/doi:10.1101/gr.243584.118/-/DC1 http://genome.cshlp.org/lookup/suppl/doi:10.1101/gr.243584.118/-/DC1 http://genome.cshlp.org/lookup/suppl/doi:10.1101/gr.243584.118/-/DC1 correlated with chronological age (Horvath 2013). Deviations from chronological age have been associated with several measures of ac- celerated biological aging and age-related phenotypes (Johnson et al. 2012; Levine et al. 2015; Marioni et al. 2015; Chen et al. 2016). We investigated the DNA methylation age of DNMT3A c.2312G > A p.(Arg771Gln) carriers using the DNA age calculator (Horvath 2013; http://dnamage.genetics.ucla.edu/), finding that DNMT3A c.2312G > A; p.(Arg771Gln) carriers show evidence for highly accelerated aging—an increase of ∼40% beyond their chro- nological age—compared with wild-type family members (ANCOVA P-value = 0.004) (Fig. 3A). Only one of 353 probes used in the epigenetic clock (Horvath 2013) overlapped with the DMPs significantly associated with the DNMT3A c.2312G > A; p.(Arg771Gln) pathogenic variant, leading us to conclude that this finding represented a true acceleration of epigenetic age. Furthermore, compared with an extensive number (322) of wild- type control samples profiled in a previous study from our group (Hannon et al. 2016), DNMT3A c.2312G > A; p.(Arg771Gln), carri- ers were consistent outliers for epigenetic age, suggesting their profiles fall outside the normal distribution of variance observed in the general population (Supplemental Fig. S8). Consistent with this, the mosaic Amish father was found to have an interme- diate level of epigenetic age acceleration, with a 23% increase over his chronological age. This age acceleration was a cumulative pro- cess as indicated by the increased slope of DNMT3A c.2312G > A; p.(Arg771Gln) carriers versus wild-type. Epigenetic age could therefore be predicted by the linear regression model as follows: epigenetic age = 4.81 + 1.405 × chronological age. The cumulative increase of epigenetic age relative to chronological age is also nota- ble compared with a recent meta-analysis of longitudinal cohort data that shows the trajectory of epigenetic age in different popula- tions progresses at a slightly slower rate compared with increasing chronological age (Marioni et al. 2019). We next looked for evidence of elevated epigenetic aging in TBRS patients carrying one of the 15 additional de novo DNMT3A pathogenic variants. All TBRS patients showed accelerat- ed epigenetic aging, although the position and type of each variant result in differing degrees of accelerated epigenetic aging (Table 1). The greatest rate of epigenetic age acceleration (>800%) was ob- served in association with the germline p.(Arg882Cys) substitu- tion, somatic mutation of DNMT3A Arg882 being the most commonly associated with AML. Altered epigenetic aging in methyltransferase-associated human growth disorders To determine whether altered epigenetic aging is a characteristic of other growth disorders associated with disruption of epigenetic regulatory molecules, we extended our study using publicly avail- able Illumina 450K DNA methylation data. We first analyzed the data from individuals with Sotos syndrome, a congenital over- growth syndrome that results from mutation of the epigenetic modifier NSD1 (Supplemental Table S5), a lysine histone methyl- transferase (Kurotaki et al. 2002; Qiao et al. 2011). Consistent with DNMT3A pathogenic variant carriers, these individuals are characterized by an epigenetic age acceleration of ∼40% (linear re- gression model R2 = 0.869, P-value = 6.4 × 10−9) (Fig. 3B,D). We then examined data from Kabuki syndrome patients carrying path- ogenic variants in the KMT2D gene (Supplemental Table S6), which also encodes a lysine histone methyltransferase (Ng et al. 2010; Butcher et al. 2017). Kabuki syndrome is a multisystem dis- order. Patients typically present with postnatal growth deficiency (rather than overgrowth), a characteristic facial gestalt, ID, and other variable phenotypic features. Although there is more hetero- geneity in epigenetic age when compared with the NSD1 patho- genic variant carriers, there was a significant reduction in epigenetic age of ∼40% seen across these individuals (linear regres- sion model R2 = 0.418, P-value = 0.023) (Fig. 3C,D). Discussion To date, 78 individuals have been described with the overgrowth condition TBRS. Within this group, a wide variety of germline DNMT3A pathogenic variants have been reported, including 33 missense, eight stop-gain, seven frameshift and two splice site var- iants, two in-frame and five whole-gene deletions (including a set of identical twins) (Tatton-Brown et al. 2014; Okamoto et al. 2016; Tlemsani et al. 2016; Hollink et al. 2017; Kosaki et al. 2017; Lemire et al. 2017; Shen et al. 2017; Spencer et al. 2017; Tatton-Brown et al. 2017, 2018; Xin et al. 2017). Clinically, the predominant 900aa PRC2/EED-EZH2 complex S-adenosyl-L-methionine 199 292 350 403 482 494 586 614 634 912 641-645 686-688 891-893 AA posi�ons PWWP ADD MTaseDomains Interac�ons DNMT1 and DNMT3B 100aa 200aa 300aa 400aa 500aa 600aa 700aa 800aaScale: Missense variant In-frame dele�on Frameshi� variant Figure 2. Schematic representation of DNMT3A. The positions of the disease-associated variants included in this study are indicated relative to the pro- tein domain architecture. PWWP, proline-tryptophan-tryptophan-proline domain; ADD, ATRX-Dnmt3-Dnmt3L domain; MTase, Methyltransferase domain; AA, amino acid. Epigenetic aging abnormalities in growth disorders Genome Research 1061 www.genome.org http://dnamage.genetics.ucla.edu/ http://dnamage.genetics.ucla.edu/ http://dnamage.genetics.ucla.edu/ http://dnamage.genetics.ucla.edu/ http://dnamage.genetics.ucla.edu/ http://dnamage.genetics.ucla.edu/ http://genome.cshlp.org/lookup/suppl/doi:10.1101/gr.243584.118/-/DC1 http://genome.cshlp.org/lookup/suppl/doi:10.1101/gr.243584.118/-/DC1 http://genome.cshlp.org/lookup/suppl/doi:10.1101/gr.243584.118/-/DC1 features of TBRS are overgrowth, a characteristic facial gestalt, and neurocognitive impairment. These features show phenotypic overlap with conditions associated with germline pathogenic var- iants in other epigenetic regulatory genes, including Sotos and Weaver syndromes caused by variants in NSD1 and EZH2 histone methyltransferases, respectively (Tatton-Brown et al. 2017). These genes encode essential epigenetic regulatory proteins, with a dual somatic/germline role in the pathogenesis of hematological malig- nancies and overgrowth syndromes with variable degrees of intel- lectual impairment (Tatton-Brown et al. 2014). The majority of DNMT3A pathogenic variants in TBRS have been found to be de novo, with five individuals inheriting the pathogenic variant from two mosaic parents (Tlemsani et al. 2016; Xin et al. 2017) and two individuals inheriting the patho- genic variant from their affected father (Lemire et al. 2017). Extensive studies of the role of DNMT3A in hematopoietic stem cell (HSC) differentiation are also reported, including the regular occurrence of somatic DNMT3A variants in patients with acute myeloid leukemia (AML). The most common somatic pathogenic variant reported in patients with AML affects the amino acid resi- due Arg882. To date, pathogenic variants predicted to affect this residue have been described in the germline of 12 TBRS patients, five with p.(Arg882His) and seven with p.(Arg882Cys) (Tlemsani et al. 2016; Hollink et al. 2017; Kosaki et al. 2017; Shen et al. 2017; Spencer et al. 2017; Tatton-Brown et al. 2018). Despite these studies, the underlying biological mech- anism and outcomes of DNMT3A gene mutation in TBRS, as well as the potential risks of hematological malignancy, re- main largely unclear. Here we investigated variation in DNA methylation associated with a germline heterozygous DNMT3A mis- sense pathogenic variant c.2312G > A; p.(Arg771Gln), affecting the catalytic MTase domain, in a large Amish family comprising four children with TBRS, un- affected siblings, and their mosaic father who displayed an intermediate clinical phenotype (Xin et al. 2017). Affected in- dividuals were characterized by wide- spread hypomethylation, with DMPs enriched in the vicinity of genes/reg- ulatory regions associated with growth and development, tissue morphogene- sis, and differentiation. The magnitude of hypomethylation typically exceeded 10%, a level often considered to show biological significance (Leenen et al. 2016). The accelerated epigenetic age ob- served did not appear to be driven by overlap of the DNMT3A c.2312G > A; p.(Arg771Gln) variant–associated DMPs with the probes that comprise the epige- netic clock as shown by overlap with only one out of the 353 probes used in the epigenetic age estimation (Horvath 2013). Although the relevance of blood cells to understanding the etiology of TBRS is not yet known, we hypothesize that our findings will be generalizable across cell types given the ubiquitous developmental expression of DNMT3A and given that many age-associated DMPs are shared across different cell types (Zhu et al. 2018). Nevertheless, it would still be prudent to undertake epigenetic age assessment of other tissues from TBRS patients to determine whether epigenetic agetrulyisaccelerated across all cell typesorafindingthatislimited to blood. Although dysregulation of growth control has been linked to numerous developmental disorders and malignancy, the specific molecular basis of this relationship is not fully understood. The as- sessment of DMPs associated with the DNMT3A c.2312G >A; p.(Arg771Gln) variant identified an enrichment of pluripotent and fetal DNase I sensitivity hotspots, as well as brain and embry- onic stem cell–associated chromatin sites according to The ENCODE Project Consortium and Epigenomic Roadmap Consortium data sets (Breeze et al. 2016). Similarly, functional an- notation based on Gene Ontology terms showed an overrepresen- tation of pathways related to morphogenesis, development, and differentiation annotations. DNMT3A loss of function has previ- ously been reported to result in up-regulated multipotency genes A B C D w Figure 3. Altered epigenetic aging is observed in methyltransferase-associated human growth disor- ders. (A) Scatter plot comparing “DNA methylation age” derived from the Illumina 450K data (y-axis) and actual chronological age (x-axis) in DNMT3A c.2312G > A p.(Arg771Gln) pathogenic variant carriers (red) versus wild-type family members (blue). Green indicates the mosaic individual. The linear regression model is also shown. (B) Scatter plot comparing DNA methylation age versus chronological age in patients with Sotos syndrome. In-frame legend illustrates the different NSD1 pathogenic variants studied. (C) Scatter plot comparing DNA methylation age versus chronological age in patients with Kabuki syn- drome. In-frame legend illustrates the different KMT2D pathogenic variants studied. (D) Boxplot compar- ing the epigenetic age acceleration rates found in association with TBRS DNMT3A variants, KMT2D Kabuki syndrome variants, and NSD1 Sotos syndrome variants. Each age acceleration observation is plot- ted as a circle. The dotted red line denotes no age acceleration. Jeffries et al. 1062 Genome Research www.genome.org and impaired differentiation of neural stem cells and HSCs (Wu et al. 2010; Challen et al. 2011; Jeong et al. 2018) compared with a gain-of-function DNMT3A variant that may increase cellular dif- ferentiation (Heyn et al. 2019). It is thus conceivable that TBRS-as- sociated DNMT3A variants may promote increased proliferation of stem/progenitor cell pool, resulting in increased cell numbers dur- ing organ morphogenesis and clinical overgrowth. Our finding of altered epigenetic outcomes in TBRS prompted us to consider similar investigations in other growth disorders as- sociated with epigenetic dysfunction: Sotos syndrome, a neurode- velopmental disorder with features overlapping TBRS and with association with overgrowth in childhood owing to histone meth- yltransferase NSD1 gene alterations, and Kabuki syndrome, a dis- tinct neurodevelopmental disorder associated with poor growth and histone methyltransferase KMT2D gene alterations. This work defined clear aberrations in epigenetic aging appropriate to the specific nature of each condition. In both overgrowth condi- tions, TBRS and Sotos syndrome, we identified accelerated epige- netic aging as measured by the DNA methylation age calculator (Horvath 2013). Conversely, patients with Kabuki syndrome, clin- ically characterized by poor growth, displayed decelerated epige- netic age. Epigenetic age has been strongly correlated with chronological age in unaffected individuals in previous studies of a variety of tissue types (Hannum et al. 2013; Horvath 2013). The observation of accelerated epigenetic aging in both TBRS and Sotos syndrome potentially results from reduced methyltrans- ferase activity in addition to increased cell turnover associated with the overgrowth seen with these disorders, with the converse being the case for Kabuki syndrome. Accelerated epigenetic aging has been associated with age-related clinical characteristics and mortality in epidemiological studies. For example, accelerated epi- genetic age in lymphocytes correlates with reduced physical and cognitive function in the elderly and with increased overall mor- tality independent of other variables such as BMI, sex, and smok- ing status (Marioni et al. 2015; Chen et al. 2016). The molecular basis of TBRS has only been determined relatively recently, and as such, most of the affected individuals reported are children and young adults. There is therefore still only very limited data available relating to the progression and prognosis of this disorder, meaning that it is not yet possible to determine whether there might be any clinical evidence of multimorbidity indicative of pre- mature aging or a reduction in average life span in TBRS. Further long-term natural history studies of TBRS patients will be extreme- ly helpful for determining the clinical implications of the epige- netic age acceleration observed as a feature of this disorder. Accelerated epigenetic age has previously been reported in as- sociation with specific diseases such as Huntington’s disease (+3.4 yr) (Horvath et al. 2016), Down syndrome (+6.6 yr) (Horvath et al. 2015), and Werner’s syndrome (+6.4 yr) (Maierhofer et al. 2017). The accelerated epigenetic aging described in association with these disorders is an average increase in epigenetic age, which is relatively consistent throughout lifespan. A distinguishing feature of carriers of the Amish DNMT3A c.2312G > A; p.(Arg771Gln) var- iant is the year-on-year or cumulative increase of accelerated epige- netic aging over life time course, in other words, a true acceleration of epigenetic aging. Although it was not possible to undertake these studies for other DNMT3A variants, this may be indicative of a similar effect on cumulative epigenetic age acceleration over the life course in TBRS. It is also noted that the gene encoding DNMT3L is located on Chromosome 21; given the previous report of an average DNA methylation age acceleration of 6.6 yr in blood and brain tissue in individuals with Down syndrome (Horvath et al. 2015) and the role of DNMT3L in stimulating DNMT3A de novo methylation, further investigations are needed to explore the potential relevance of this observation. There are currently only four reported cases of an AML tumor carrying the DNMT3A p.Arg771Gln substitution. Biochemical measurements of DNMT3A show that mutations at both the Arg771 and Arg882 residues result in reduced methyltransferase activity, with a greater degree of reduction resulting from Arg882 variants compared with Arg771 variants (2.4-fold difference) (Holz-Schietinger et al. 2012). Given this reduced methyltransfer- ase activity, we may expect to observe more pronounced changes in DNA methylation in patients with germline variants affecting Arg882 compared with variants affecting other amino acid resi- dues such as Arg771. Our data reflected this notion, with alteration Arg882 displaying markedly greater methylation changes com- pared with the other DNMT3A mutations investigated in this study. Currently, available literature suggests that the risk of hema- tological malignancy in TBRS individuals may vary depending on the specific pathogenic variant underling their condition (Hollink et al. 2017). The significantly advanced epigenetic age that we ob- served in association with p.Arg882Cys may explain why hemato- logical malignancy has to date only been reported in two TBRS patients, one harboring this germline variant and the second the p.Tyr735Ser variant, the latter not being assessed in this study (Hollink et al. 2017; Tatton-Brown et al. 2018). In summary, our findings identify widespread DNA hypome- thylation in genes involved in morphogenesis, development, dif- ferentiation, and malignancy in TBRS patients. TBRS patients also displayed highly accelerated DNA methylation aging. Our studies additionally defined phenotype-related altered epigenetic aging in two histone methyltransferase disorders: NSD1 Sotos syn- drome overgrowth disorder and KMT2D Kabuki syndrome growth impairment. Taken together, these findings provide important new insights into the role of DNMT3A during development and of relevance to hematological malignancy, and define perturba- tion to epigenetic machinery and biological aging as common themes in overgrowth and growth deficiency syndromes. Methods Genetic and clinical studies The phenotypic features of the four affected siblings (three females and one male, aged 10–25 yr) (Fig. 1A, individuals III:3, III:5, III:7, and III:13) include macrocephaly, tall stature, hypotonia, mild to moderate ID, behavioral problems, and a distinctive facial appear- ance. Whole-genome SNP genotyping and exome sequencing of DNA samples taken with informed consent under regionally ap- proved protocols excluded pathogenic variants in known genes, or candidate new genes, associated with neurodevelopmental dis- orders. Subsequent studies defined a heterozygous c.2312G > A variant in DNMT3A, resulting in a p.(Arg771Gln) substitution, as the cause of the condition. Full clinical details are previously de- scribed (Xin et al. 2017). Further testing revealed mosaicism for the DNMT3A c.2312G > A variant in the father, and Xin et al. (2017) showed pathogenic variant load varied in different tissue types. DNA methylation profiling Genomic DNA from blood was sodium bisulfite converted using the EZ-96 DNA methylation kit (Zymo Research) and DNA meth- ylation quantified across the genome using the Illumina Epigenetic aging abnormalities in growth disorders Genome Research 1063 www.genome.org Infinium HumanMethylation450 array (Illumina 450K array) (Illumina). The additional 15 DNMT3A pathogenic variants were profiled using the Illumina Infinium EPIC array (Illumina). The Bioconductor package wateRmelon (Pidsley et al. 2013) in R 3.4.1 (R Core Team 2017) was used to import IDAT files, and after check- ing for suitable sodium bisulfite conversion (bisulfite control probe median >90%), the DNA methylation data were imported and quantile normalized using the dasen function in wateRmelon and methylation beta values produced (ratio of intensities for methylated versus unmethylated alleles). Probes showing a detec- tion P-value >0.05 in at least 1% of samples or a beadcount <3 in 5% of samples were removed across all samples. Any samples showing low quality, indicated by a detection P-value >0.05 in ≥1% of probes within a sample, were removed from analysis. Probes containing common SNPs within 10 bp of the CpG site were removed (minor allele frequency >5%). Nonspecific probes and probes on the sex chromosomes were also removed (Chen et al. 2013; Price et al. 2013). Identification of DMPs DMPs were identified using a limma-based linear model based on pathogenic variant genotype and sex as a covariate (Smyth 2004) and a Benjamini–Hochberg FDR of 5% applied (Benjamini and Hochberg 1995). When the epigenetic age was used as a covariate, a similar level of DMPs were detected (2557 DMPs) with an 80% overlap to the limma model without age as a covariate. Changes in methylation were calculated based on comparison between DNMT3A c.2312G > A; p.(Arg771Gln) carriers versus wild-type in- dividuals in the Amish pedigree. The additional 15 DNMT3A path- ogenic variants were assessed relative to seven wild-type control samples run on the same EPIC array run. Blood cell counts were unknown and so were estimated using the DNA methylation age calculator (Horvath 2013; Koestler et al. 2013) and assessed in the linear model. To identify DMPs, the package DMRcate was used with the same limma-based design (Peters et al. 2015). Gene Ontology and functional enrichment analyses Gene Ontology enrichment analysis was performed using genes annotated to FDR corrected DMPs using the gometh function of the missMethyl package (Phipson et al. 2016), which takes into account potential bias of probe distributions on the beadchip ar- ray. KEGG pathway analysis was performed using the gsameth command of missMethyl and KEGG annotation files from the Bioconductor KEGGREST package (http://bioconductor.org/ packages/release/bioc/html/KEGGREST.html). Regional enrich- ment analysis based on Illumina annotations was performed using a chi-squared test with Yates correction in R. DMRs were function- ally annotated using the webtool GREAT (http://great.stanford .edu/public/html/). The top P-value–ranked 1000 DMPs were also annotated using the eFORGE tool (https://eforge .altiusinstitute.org/) to perform functional overlap analysis for identifying any cell- or tissue-specific epigenetic signals. Quantification of global DNA methylation Global DNA methylation measurements were made using the luminometric methylation assay (LUMA) (Karimi et al. 2006) based on cleavage by a methylation-sensitive restriction enzyme followed by polymerase extension assay via pyrosequencing on the PyroMark Q24 (Qiagen). Peak heights were obtained using the Pyro Q24 CpG 2.0.6 software and a t-test applied in R 3.4.1. Global methylation estimates from the Illumina 450K array were assessed through R 3.4.1 using summary statistics and a Wilcoxon rank-sum test on two pairs of samples matched for age and sex. DNA methylation age estimation Epigenetic age calculations were made using the DNA methylation age calculator (https://dnamage.genetics.ucla.edu/) for Illumina 450K data, and Illumina EPIC arrays were assessed using the agep function of the wateRmelon Bioconductor package, the latter based on the original calculator developed by Steve Horvath (2013). Accelerated age was calculated for the Amish TBRS DNMT3A c.2312G > A; p.(Arg771Gln) carriers and wild-type family members and Sotos syndrome and Kabuki syndrome patients and compared with data from 322 control individuals taken from a previous study (Hannon et al. 2016), using linear models of recorded chronolog- ical age and calculated epigenetic age. Estimates of age acceleration for the additional 15 TBRS cases were calculated by dividing the calculated epigenetic age with their chronological age. Additional NSD1 Sotos syndrome patient Illumina DNA methylation files were obtained from GEO accession GSE74432, with corresponding chronological ages derived from the associated paper (Choufani et al. 2015). KMT2D Kabuki syndrome DNA methylation data and chronological age were obtained from GEO accession GSE97362 (Butcher et al. 2017). Validation of DMPs using bisulfite-pyrosequencing Bisulfite pyrosequencing was used to validate specific differentially methylated CpG sites originally identified using the Illumina 450K array. Primers, designed using PyroMark Assay Design soft- ware (Qiagen), and PCR conditions are provided in Supplemental Table S7. Bisulfite conversion was performed on ∼500 ng of DNA using a bisulfite-gold kit (Zymo Research). PCR was performed with HOT FIREPol DNA polymerase (Solis Biodyne) for 15 min at 95°C followed by 37 cycles of 15 sec at 95°C, 15 sec at annealing temperature (shown in Supplemental Table S7), and 30 sec at 72°C. A final extension of 10 min at 72°C was then applied. DNA methylation was then assessed using the resulting bisulfite PCR amplicons, together with a pyrosequencing primer on the PyroMark Q24 system (Qiagen) following the manufacturer’s standard instructions and the Pyro Q24 CpG 2.0.6 software. Data access The raw and processed primary data sets generated in this study have been submitted to the NCBI Gene Expression Omnibus (GEO; https://www.ncbi.nlm.nih.gov/geo/) under accession num- ber GSE128801. R scripts are provided as Supplemental Code S1 and at the following repository: https://github.com/arjeffries/ TBRS2019. Acknowledgments We thank the Amish families for participating in this study and the Amish community for their continued support of the Windows of Hope project. The work was supported by the Newlife Foundation for Disabled Children (Ref: SG/16-17/02, to C.G.S., A.R.J., A.H.C., and E.L.B.), Medical Research Council (MRC) grant G1001931 (to E.L.B.), MRC grant G1002279 (to A.H.C.), and MRC grants MR/ M008924/1 and MR/K013807/1 (to J.M.). Author contributions: E.L.B., A.H.C., and J.M. conceived the study. R.M. and B.A.C. performed the genetic analysis. A.R.J. and R.M. performed LUMA global methylation assay. A.R.J. performed the microarray-based DNA methylation analysis and all associated statistical analyses. A.R.J. and E.D. performed the pyrosequencing Jeffries et al. 1064 Genome Research www.genome.org http://bioconductor.org/packages/release/bioc/html/KEGGREST.html http://bioconductor.org/packages/release/bioc/html/KEGGREST.html http://bioconductor.org/packages/release/bioc/html/KEGGREST.html http://bioconductor.org/packages/release/bioc/html/KEGGREST.html http://bioconductor.org/packages/release/bioc/html/KEGGREST.html http://bioconductor.org/packages/release/bioc/html/KEGGREST.html http://great.stanford.edu/public/html/ http://great.stanford.edu/public/html/ http://great.stanford.edu/public/html/ http://great.stanford.edu/public/html/ http://great.stanford.edu/public/html/ https://eforge.altiusinstitute.org/ https://eforge.altiusinstitute.org/ https://eforge.altiusinstitute.org/ https://eforge.altiusinstitute.org/ https://eforge.altiusinstitute.org/ https://dnamage.genetics.ucla.edu/ https://dnamage.genetics.ucla.edu/ https://dnamage.genetics.ucla.edu/ https://dnamage.genetics.ucla.edu/ https://dnamage.genetics.ucla.edu/ https://dnamage.genetics.ucla.edu/ http://genome.cshlp.org/lookup/suppl/doi:10.1101/gr.243584.118/-/DC1 http://genome.cshlp.org/lookup/suppl/doi:10.1101/gr.243584.118/-/DC1 http://genome.cshlp.org/lookup/suppl/doi:10.1101/gr.243584.118/-/DC1 https://www.ncbi.nlm.nih.gov/geo/ https://www.ncbi.nlm.nih.gov/geo/ https://www.ncbi.nlm.nih.gov/geo/ https://www.ncbi.nlm.nih.gov/geo/ https://www.ncbi.nlm.nih.gov/geo/ https://www.ncbi.nlm.nih.gov/geo/ https://www.ncbi.nlm.nih.gov/geo/ http://genome.cshlp.org/lookup/suppl/doi:10.1101/gr.243584.118/-/DC1 https://github.com/arjeffries/TBRS2019 https://github.com/arjeffries/TBRS2019 https://github.com/arjeffries/TBRS2019 https://github.com/arjeffries/TBRS2019 https://github.com/arjeffries/TBRS2019 and associated analysis. A.R.J., C.G.S., R.M., A.H.C., J.M., and E.L.B. wrote the manuscript. H.E.C., M.A.P., I.K.T., D.J.G.M., F.I.R., K.T.-B., L.A., D.B., T.D., M.F.H., R.N.-E., A.K., A.K., A.S., A.S., L.VM., V.V., and N.Y. contributed samples and clinical data. References Benjamini Y, Hochberg Y. 1995. Controlling the false discovery rate: a prac- tical and powerful approach to multiple testing. J R Stat Soc Ser B 57: 289–300. doi:10.2307/2346101 Bjornsson HT. 2015. The Mendelian disorders of the epigenetic machinery. Genome Res 25: 1473–1481. doi:10.1101/gr.190629.115 Breeze CE, Paul DS, van Dongen J, Butcher LM, Ambrose JC, Barrett JE, Lowe R, Rakyan VK, Iotchkova V, Frontini M, et al. 2016. eFORGE: a tool for identifying cell type-specific signal in epigenomic data. Cell Rep 17: 2137–2150. doi:10.1016/j.celrep.2016.10.059 Butcher DT, Cytrynbaum C, Turinsky AL, Siu MT, Inbar-Feigenberg M, Mendoza-Londono R, Chitayat D, Walker S, Machado J, Caluseriu O, et al. 2017. CHARGE and Kabuki syndromes: Gene-specific DNA meth- ylation signatures identify epigenetic mechanisms linking these clini- cally overlapping conditions. Am J Hum Genet 100: 773–788. doi:10 .1016/j.ajhg.2017.04.004 Challen GA, Sun D, Jeong M, Luo M, Jelinek J, Berg JS, Bock C, Vasanthakumar A, Gu H, Xi Y, et al. 2011. Dnmt3a is essential for hema- topoietic stem cell differentiation. Nat Genet 44: 23–31. doi:10.1038/ng .1009 Chen T, Ueda Y, Dodge JE, Wang Z, Li E. 2003. Establishment and mainte- nance of genomic methylation patterns in mouse embryonic stem cells by Dnmt3a and Dnmt3b. Mol Cell Biol 23: 5594–5605. doi:10.1128/ MCB.23.16.5594-5605.2003 Chen YA, Lemire M, Choufani S, Butcher DT, Grafodatskaya D, Zanke BW, Gallinger S, Hudson TJ, Weksberg R. 2013. Discovery of cross- reactive probes and polymorphic CpGs in the Illumina Infinium HumanMethylation450 microarray. Epigenetics 8: 203–209. doi:10 .4161/epi.23470 Chen BH, Marioni RE, Colicino E, Peters MJ, Ward-Caviness CK, Tsai PC, Roetker NS, Just AC, Demerath EW, Guan W, et al. 2016. DNA methyl- ation-based measures of biological age: meta-analysis predicting time to death. Aging (Albany NY) 8: 1844–1865. doi:10.18632/aging.101020 Choufani S, Cytrynbaum C, Chung BH, Turinsky AL, Grafodatskaya D, Chen YA, Cohen AS, Dupuis L, Butcher DT, Siu MT, et al. 2015. NSD1 mutations generate a genome-wide DNA methylation signature. Nat Commun 6: 10207. doi:10.1038/ncomms10207 Davis CA, Hitz BC, Sloan CA, Chan ET, Davidson JM, Gabdank I, Hilton JA, Jain K, Baymuradov UK, Narayanan AK, et al. 2018. The encyclopedia of DNA elements (ENCODE): data portal update. Nucleic Acids Res 46: D794–D801. doi:10.1093/nar/gkx1081 Doi A, Park IH, Wen B, Murakami P, Aryee MJ, Irizarry R, Herb B, Ladd- Acosta C, Rho J, Loewer S, et al. 2009. Differential methylation of tissue- and cancer-specific CpG island shores distinguishes human induced pluripotent stem cells, embryonic stem cells and fibroblasts. Nat Genet 41: 1350–1353. doi:10.1038/ng.471 The ENCODE Project Consortium. 2012. An integrated encyclopedia of DNA elements in the human genome. Nature 489: 57–74. doi:10 .1038/nature11247 Hannon E, Dempster E, Viana J, Burrage J, Smith AR, Macdonald R, St Clair D, Mustard C, Breen G, Therman S, et al. 2016. An integrated genetic- epigenetic analysis of schizophrenia: evidence for co-localization of ge- netic associations and differential DNA methylation. Genome Biol 17: 176. doi:10.1186/s13059-016-1041-x Hannum G, Guinney J, Zhao L, Zhang L, Hughes G, Sadda S, Klotzle B, Bibikova M, Fan JB, Gao Y, et al. 2013. Genome-wide methylation pro- files reveal quantitative views of human aging rates. Mol Cell 49: 359– 367. doi:10.1016/j.molcel.2012.10.016 Heyn P, Logan CV, Fluteau A, Challis RC, Auchynnikava T, Martin CA, Marsh JA, Taglini F, Kilanowski F, Parry DA, et al. 2019. Gain-of-func- tion DNMT3A mutations cause microcephalic dwarfism and hyperme- thylation of Polycomb-regulated regions. Nat Genet 51: 96–105. doi:10 .1038/s41588-018-0274-x Hollink I, van den Ouweland AMW, Beverloo HB, Arentsen-Peters S, Zwaan CM, Wagner A. 2017. Acute myeloid leukaemia in a case with Tatton- Brown–Rahman syndrome: the peculiar DNMT3A R882 mutation. J Med Genet 54: 805–808. doi:10.1136/jmedgenet-2017-104574 Holz-Schietinger C, Matje DM, Reich NO. 2012. Mutations in DNA methyl- transferase (DNMT3A) observed in acute myeloid leukemia patients dis- rupt processive methylation. J Biol Chem 287: 30941–30951. doi:10 .1074/jbc.M112.366625 Horvath S. 2013. DNA methylation age of human tissues and cell types. Genome Biol 14: R115. doi:10.1186/gb-2013-14-10-r115 Horvath S, Garagnani P, Bacalini MG, Pirazzini C, Salvioli S, Gentilini D, Di Blasio AM, Giuliani C, Tung S, Vinters HV, et al. 2015. Accelerated epi- genetic aging in Down syndrome. Aging Cell 14: 491–495. doi:10.1111/ acel.12325 Horvath S, Langfelder P, Kwak S, Aaronson J, Rosinski J, Vogt TF, Eszes M, Faull RL, Curtis MA, Waldvogel HJ, et al. 2016. Huntington’s disease ac- celerates epigenetic aging of human brain and disrupts DNA methyla- tion levels. Aging (Albany NY) 8: 1485–1512. doi:10.18632/aging .101005 Jeong M, Park HJ, Celik H, Ostrander EL, Reyes JM, Guzman A, Rodriguez B, Lei Y, Lee Y, Ding L, et al. 2018. Loss of Dnmt3a immortalizes hemato- poietic stem cells in vivo. Cell Rep 23: 1–10. doi:10.1016/j.celrep.2018.03 .025 Jiang YL, Rigolet M, Bourc’his D, Nigon F, Bokesoy I, Fryns JP, Hultén M, Jonveaux P, Maraschio P, Megarbane A, et al. 2005. DNMT3B mutations and DNA methylation defect define two types of ICF syndrome. Hum Mutat 25: 56–63. doi:10.1002/humu.20113 Johnson AA, Akman K, Calimport SR, Wuttke D, Stolzing A, de Magalhães JP. 2012. The role of DNA methylation in aging, rejuvenation, and age-related disease. Rejuvenation Res 15: 483–494. doi:10.1089/rej .2012.1324 Jones PA. 2012. Functions of DNA methylation: islands, start sites, gene bodies and beyond. Nat Rev Genet 13: 484–492. doi:10.1038/nrg3230 Karimi M, Johansson S, Ekstrom TJ. 2006. Using LUMA: a luminometric- based assay for global DNA-methylation. Epigenetics 1: 45–48. doi:10 .4161/epi.1.1.2587 Klein CJ, Botuyan MV, Wu Y, Ward CJ, Nicholson GA, Hammans S, Hojo K, Yamanishi H, Karpf AR, Wallace DC, et al. 2011. Mutations in DNMT1 cause hereditary sensory neuropathy with dementia and hearing loss. Nat Genet 43: 595–600. doi:10.1038/ng.830 Koestler DC, Christensen B, Karagas MR, Marsit CJ, Langevin SM, Kelsey KT, Wiencke JK, Houseman EA. 2013. Blood-based profiles of DNA methyl- ation predict the underlying distribution of cell types: a validation anal- ysis. Epigenetics 8: 816–826. doi:10.4161/epi.25430 Kosaki R, Terashima H, Kubota M, Kosaki K. 2017. Acute myeloid leukemia- associated DNMT3A p.Arg882His mutation in a patient with Tatton- Brown–Rahman overgrowth syndrome as a constitutional mutation. Am J Med Genet A 173: 250–253. doi:10.1002/ajmg.a.37995 Kurotaki N, Imaizumi K, Harada N, Masuno M, Kondoh T, Nagai T, Ohashi H, Naritomi K, Tsukahara M, Makita Y, et al. 2002. Haploinsufficiency of NSD1 causes Sotos syndrome. Nat Genet 30: 365–366. doi:10.1038/ ng863 Leenen FA, Muller CP, Turner JD. 2016. DNA methylation: conducting the orchestra from exposure to phenotype? Clin Epigenetics 8: 92. doi:10 .1186/s13148-016-0256-8 Lemire G, Gauthier J, Soucy JF, Delrue MA. 2017. A case of familial transmis- sion of the newly described DNMT3A-overgrowth syndrome. Am J Med Genet A 173: 1887–1890. doi:10.1002/ajmg.a.38119 Levine ME, Lu AT, Bennett DA, Horvath S. 2015. Epigenetic age of the pre- frontal cortex is associated with neuritic plaques, amyloid load, and Alzheimer’s disease related cognitive functioning. Aging (Albany NY) 7: 1198–1211. doi:10.18632/aging.100864 Liberzon A, Subramanian A, Pinchback R, Thorvaldsdottir H, Tamayo P, Mesirov JP. 2011. Molecular signatures database (MSigDB) 3.0. Bioinformatics 27: 1739–1740. doi:10.1093/bioinformatics/btr260 Maierhofer A, Flunkert J, Oshima J, Martin GM, Haaf T, Horvath S. 2017. Accelerated epigenetic aging in Werner syndrome. Aging (Albany NY) 9: 1143–1152. doi:10.18632/aging.101217 Marioni RE, Shah S, McRae AF, Ritchie SJ, Muniz-Terrera G, Harris SE, Gibson J, Redmond P, Cox SR, Pattie A, et al. 2015. The epigenetic clock is correlated with physical and cognitive fitness in the Lothian Birth Cohort 1936. Int J Epidemiol 44: 1388–1396. doi:10.1093/ije/dyu277 Marioni RE, Suderman M, Chen BH, Horvath S, Bandinelli S, Morris T, Beck S, Ferrucci L, Pedersen NL, Relton CL, et al. 2019. Tracking the epigenet- ic clock across the human life course: a meta-analysis of longitudinal co- hort data. J Gerontol A Biol Sci Med Sci 74: 57–61. doi:10.1093/gerona/ gly060 McLean CY, Bristor D, Hiller M, Clarke SL, Schaar BT, Lowe CB, Wenger AM, Bejerano G. 2010. GREAT improves functional interpretation of cis-reg- ulatory regions. Nat Biotechnol 28: 495–501. doi:10.1038/nbt.1630 Ng SB, Bigham AW, Buckingham KJ, Hannibal MC, McMillin MJ, Gildersleeve HI, Beck AE, Tabor HK, Cooper GM, Mefford HC, et al. 2010. Exome sequencing identifies MLL2 mutations as a cause of Kabuki syndrome. Nat Genet 42: 790–793. doi:10.1038/ng.646 Okamoto N, Toribe Y, Shimojima K, Yamamoto T. 2016. Tatton-Brown– Rahman syndrome due to 2p23 microdeletion. Am J Med Genet A 170: 1339–1342. doi:10.1002/ajmg.a.37588 Okano M, Bell DW, Haber DA, Li E. 1999. DNA methyltransferases Dnmt3a and Dnmt3b are essential for de novo methylation and mammalian de- velopment. Cell 99: 247–257. doi:10.1016/S0092-8674(00)81656-6 Epigenetic aging abnormalities in growth disorders Genome Research 1065 www.genome.org Peters TJ, Buckley MJ, Statham AL, Pidsley R, Samaras K RVL, Clark SJ, Molloy PL. 2015. De novo identification of differentially methylated re- gions in the human genome. Epigenetics Chromatin 8: 6. doi:10.1186/ 1756-8935-8-6 Phipson B, Maksimovic J, Oshlack A. 2016. missMethyl: an R package for an- alyzing data from Illumina’s HumanMethylation450 platform. Bioinformatics 32: 286–288. doi:10.1093/bioinformatics/btv560 Pidsley R, Wong CCY, Volta M, Lunnon K, Mill J, Schalkwyk LC. 2013. A data-driven approach to preprocessing Illumina 450K methylation array data. BMC Genomics 14: 293. doi:10.1186/1471-2164-14-293 Price ME, Cotton AM, Lam LL, Farré P, Emberly E, Brown CJ, Robinson WP, Kobor MS. 2013. Additional annotation enhances potential for biologically-relevant analysis of the Illumina Infinium HumanMethylation450 BeadChip array. Epigenetics Chromatin 6: 4. doi:10.1186/1756-8935-6-4 Qiao Q, Li Y, Chen Z, Wang M, Reinberg D, Xu RM. 2011. The structure of NSD1 reveals an autoregulatory mechanism underlying histone H3K36 methylation. J Biol Chem 286: 8361–8368. doi:10.1074/jbc.M110 .204115 R Core Team. 2017. R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R- project.org/. Roadmap Epigenomics Consortium, Kundaje A, Meuleman W, Ernst J, Bilenky M, Yen A, Heravi-Moussavi A, Kheradpour P, Zhang Z, Wang J, et al. 2015. Integrative analysis of 111 reference human epigenomes. Nature 518: 317–330. doi:10.1038/nature14248 Shen W, Heeley JM, Carlston CM, Acuna-Hidalgo R, Nillesen WM, Dent KM, Douglas GV, Levine KL, Bayrak-Toydemir P, Marcelis CL, et al. 2017. The spectrum of DNMT3A variants in Tatton-Brown–Rahman syndrome overlaps with that in hematologic malignancies. Am J Med Genet A 173: 3022–3028. doi:10.1002/ajmg.a.38485 Smyth GK. 2004. Linear models and empirical Bayes methods for assessing differential expression in microarray experiments. Stat Appl Genet Mol Biol 3: Article3. doi:10.2202/1544-6115.1027 Spencer DH, Russler-Germain DA, Ketkar S, Helton NM, Lamprecht TL, Fulton RS, Fronick CC, O’Laughlin M, Heath SE, Shinawi M, et al. 2017. CpG island hypermethylation mediated by DNMT3A is a conse- quence of AML progression. Cell 168: 801–816.e13. doi:10.1016/j.cell .2017.01.021 Tatton-Brown K, Seal S, Ruark E, Harmer J, Ramsay E, Del Vecchio Duarte S, Zachariou A, Hanks S, O’Brien E, Aksglaede L, et al. 2014. Mutations in the DNA methyltransferase gene DNMT3A cause an overgrowth syn- drome with intellectual disability. Nat Genet 46: 385–388. doi:10 .1038/ng.2917 Tatton-Brown K, Loveday C, Yost S, Clarke M, Ramsay E, Zachariou A, Elliott A, Wylie H, Ardissone A, Rittinger O, et al. 2017. Mutations in epigenetic regulation genes are a major cause of overgrowth with intellectual dis- ability. Am J Hum Genet 100: 725–736. doi:10.1016/j.ajhg.2017.03.010 Tatton-Brown K, Zachariou A, Loveday C, Renwick A, Mahamdallie S, Aksglaede L, Baralle D, Barge-Schaapveld D, Blyth M, Bouma M, et al. 2018. The Tatton-Brown–Rahman syndrome: a clinical study of 55 indi- viduals with de novo constitutive DNMT3A variants. Wellcome Open Res 3: 46. doi:10.12688/wellcomeopenres.14430.1 Tlemsani C, Luscan A, Leulliot N, Bieth E, Afenjar A, Baujat G, Doco-Fenzy M, Goldenberg A, Lacombe D, Lambert L, et al. 2016. SETD2 and DNMT3A screen in the Sotos-like syndrome French cohort. J Med Genet 53: 743–751. doi:10.1136/jmedgenet-2015-103638 Wu H, Coskun V, Tao J, Xie W, Ge W, Yoshikawa K, Li E, Zhang Y, Sun YE. 2010. Dnmt3a-dependent nonpromoter DNA methylation facilitates transcription of neurogenic genes. Science 329: 444–448. doi:10.1126/ science.1190485 Xin B, Cruz Marino T, Szekely J, Leblanc J, Cechner K, Sency V, Wensel C, Barabas M, Therriault V, Wang H. 2017. Novel DNMT3A germline mu- tations are associated with inherited Tatton-Brown–Rahman syndrome. Clin Genet 91: 623–628. doi:10.1111/cge.12878 Zhu T, Zheng SC, Paul DS, Horvath S, Teschendorff AE. 2018. Cell and tissue type independent age-associated DNA methylation changes are not rare but common. Aging (Albany NY) 10: 3541–3557. doi:10.18632/aging .101666 Received September 2, 2018; accepted in revised form May 24, 2019. Jeffries et al. 1066 Genome Research www.genome.org https://www.R-project.org/ https://www.R-project.org/ https://www.R-project.org/ https://www.R-project.org/ https://www.R-project.org/ https://www.R-project.org/