1 Comparative Methylome Analyses Identify Epigenetic Regulatory

Comparative Methylome Analyses Identify EpigeneticRegulatory Loci of Human Brain Evolution

Isabel Mendizabaldagger12 Lei Shidagger34 Thomas E Keller1 Genevieve Konopka5 Todd M Preuss6

Tzung-Fu Hsieh7 Enzhi Hu38 Zhe Zhang3 Bing Su3 and Soojin V Yi1

1School of Biological Sciences Institute of Bioengineering and Bioscience Georgia Institute of Technology Atlanta GA2Department of Genetics Physical Anthropology and Animal Physiology University of the Basque Country Leioa Spain3State Key Laboratory of Genetic Resources and Evolution Kunming Institute of Zoology Chinese Academy of Sciences KunmingChina4The Molecular amp Behavioral Neuroscience Institute University of Michigan Ann Arbor MI5Department of Neuroscience University of Texas Southwestern Medical Center Dallas TX6Division of Neuropharmacology and Neurologic Diseases amp Center for Translational Social Neuroscience Department of Pathologyand Laboratory Medicine Yerkes National Primate Research Center Emory University School of Medicine Emory University AtlantaGA7Department of Plant and Microbial Biology and Plants for Human Health Institute North Carolina State University Raleigh NC8Kunming College of Life Science University of Chinese Academy of Sciences Beijing ChinadaggerThese authors contributed equally to this work

Corresponding authors E-mails soojinyigatechedu submailkizaccn

Associate editor Dr Connie MulliganThe Gene Expression Omnibus accession numbers for the data reported in this paper are GSE77124 and GSE85868 for WGBS andtargeted validation data respectively

Abstract

How do epigenetic modifications change across species and how do these modifications affect evolution These arefundamental questions at the forefront of our evolutionary epigenomic understanding Our previous work investigatedhuman and chimpanzee brain methylomes but it was limited by the lack of outgroup data which is critical for com-parative (epi)genomic studies Here we compared whole genome DNA methylation maps from brains of humanschimpanzees and also rhesus macaques (outgroup) to elucidate DNA methylation changes during human brain evolu-tion Moreover we validated that our approach is highly robust by further examining 38 human-specific DMRs usingtargeted deep genomic and bisulfite sequencing in an independent panel of 37 individuals from five primate species Ourunbiased genome-scan identified human brain differentially methylated regions (DMRs) irrespective of their associationswith annotated genes Remarkably over half of the newly identified DMRs locate in intergenic regions or gene bodiesNevertheless their regulatory potential is on par with those of promoter DMRs An intriguing observation is that DMRsare enriched in active chromatin loops suggesting human-specific evolutionary remodeling at a higher-order chromatinstructure These findings indicate that there is substantial reprogramming of epigenomic landscapes during human brainevolution involving noncoding regions

Key words DNA methylation human brain evolution transcriptional divergence differentially methylated regionsepigenomes

Introduction

Epigenetic changes such as DNA methylation and histone tailmodifications leave distinctive marks on specific cell lineageswhich otherwise harbor the same genetic material Eventhough the functional significance of epigenetics in develop-ment (Gifford et al 2013 Zhu et al 2013) imprinting(Ferguson-Smith 2011 Hanna et al 2016) and disease(Timp and Feinberg 2013 Ziller et al 2013) is extremelywell appreciated whether and how epigenetic modificationsare inherited across generations and what heritable conse-quences may arise remain as an unresolved and fundamental

research area of epigenetics (Daxinger and Whitelaw 2012Heard and Martienssen 2014) Even more uncertain is howepigenetic signals diverge in evolution and what functionalconsequences if any may arise due to evolutionary epigeneticdivergence

In this paper we investigate these pressing questions in thecontext of human brain evolution Our brains are exception-ally large relative to body size compared to those of otheranimals (Jerison 1973) which occurred via a remarkably rapidexpansion in recent evolutionary past (Kaas and Preuss 2013Somel et al 2013) Previous analyses have demonstrated that

Article

The Author 2016 Published by Oxford University Press on behalf of the Society for Molecular Biology and EvolutionThis is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License(httpcreativecommonsorglicensesby-nc40) which permits non-commercial re-use distribution and reproduction in anymedium provided the original work is properly cited For commercial re-use please contact journalspermissionsoupcom Open AccessMol Biol Evol 33(11)2947ndash2959 doi101093molbevmsw176 Advance Access publication August 25 2016 2947Downloaded from httpsacademicoupcommbearticle-abstract331129472272475

by gueston 11 February 2018

evolution has modified the human brain at almost every levelof organization including changes in long-distance cortico-cortical connectivity histology and local organization (Preuss2011) as well as gene protein and metabolite expression(Brawand et al 2011 Konopka et al 2012 Liu et al 2012)Intriguingly increasing numbers of human brain studies arediscovering the critical importance of epigenetic mechanismsin regulatory functions development and manifestations ofneuropsychiatric diseases (Houston et al 2013 Nikolova et al2014) We thus ask do the human brain specific phenotypictransitions implicate changes at the epigenetic level If sohow did the epigenetic modifications affect evolutionaryinnovations

Recent studies have begun to elucidate epigenetic differ-ences between brains of humans and nonhuman primates(Enard et al 2004 Farcas et al 2009 Shi et al 2014 Shulhaet al 2012b Zeng et al 2012) In particular we have previouslycompared the whole genome DNA methylation maps (re-ferred to as ldquomethylomesrdquo) of human and chimpanzee brains(Zeng et al 2012) However our previous study focused ongene promoters missing putative differentially methylatedregions at un-annotated transcription start sites IndeedDNA methylation at gene bodies and gene deserts harborsstrong regulatory potential (Hon et al 2013 Huh et al 2013Mendizabal and Yi 2016) Zeng et al (2012) also lacked dataon outgroup species limiting the inference on human-specificmethylation changes Importantly to ascertain species-levelepigenetic divergence with high confidence it is critical tovalidate the observed epigenetic profiles in an independentdataset comprising a large number of brain samples whichwere not available in previous studies in this domainAddressing these concerns here we present an unbiased iden-tification and analysis of human brain specific DNA methyl-ation changes Our novel and robust research design allows usto infer lineage-specific epigenetic changes with high confi-dence Moreover using unbiased approaches we illustratethe significance of epigenetic changes occurring in noncodingregions of the human genome Comparative epigenomicanalyses can offer significant insight into understanding evo-lution of human-specific regulatory processes

Results

Whole-Genome Discovery of DifferentiallyMethylated Regions in Human BrainsOur initial discovery data set consisted of whole genome bi-sulfite sequencing (WGBS) data from prefrontal cortices ofthree humans three chimpanzees (Zeng et al 2012) and twoadditional macaque methylomes from same brain region gen-erated in this study (supplementary table S1 SupplementaryMaterial online) We first identified differentially methylatedregions (DMRs) between human and chimpanzee brainWGBS data sets using BSmooth (Hansen et al 2012)BSmooth is specifically designed to utilize local averagingand information from biological replicates to identify DMRsfrom WGBS data (Hansen et al 2012) We applied additionalcriteria including a minimum DNA methylation difference of03 between humans and chimpanzees We used such a

stringent criterion compared to other differential DNA meth-ylation studies so that we can identify conservative candidateDMRs In addition using this criterion relieves concerns re-garding differential cell composition (see below) We addi-tionally imposed each DMR to harbor at least 10 mappedCpGs to avoid effects of outlier CpGs

The initial candidate DMRs following these proceduresincluded 278 regions (supplementary table S2Supplementary Material online for DMR coordinates) Weused WGBS data from two rhesus macaques (Macacamulatta) to estimate the polarity of humanndashchimpanzee dif-ferences (ie human-specific if DMR is present in human butlacking in chimpanzees and macaques) Using this parsimonyapproach we identified 85 human-specific and 102chimpanzee-specific DMRs (refer ldquoMaterial and Methodsrdquo)Interestingly there was a significant excess of hypo-methylated DMRs in the human lineage (Pfrac14 0005 v2 testsupplementary table S3 Supplementary Material online)Nevertheless given the current lack of knowledge on epige-netic evolutionary rates (refer ldquoDiscussionrdquo) these resultsshould be taken with caution

Validation of DMRs via Expanded Sampling and DeepSequencingPatterns of DNA methylation at some genomic positions arevariable among individuals within the same species In addi-tion the average read depth for our discovery dataset basedon whole-genome sequencing was low for some samples(ranged from 2 to 9 supplementary table S1Supplementary Material online) To examine the consistencyof the DMRs in the presence of these potential variabilitieswe analyzed DNA methylation patterns of selected candidateDMRs from expanded samples using deep sequencingSpecifically we performed targeted PCR and sequencing ofbisulfite converted genomic DNA (average sequencingdepthfrac14 365X supplementary table S4 SupplementaryMaterial online) for a subset of DMRs (38 out of the 278)in the prefrontal cortex of a wider set of individuals andspecies 23 humans 2 chimpanzees 1 Hoolock gibbon(H leuconedys) 7 rhesus macaques and 5 crab-eatingmacaques (M fascicularis)

The validation set is biased toward human-specific DMRsbecause of sample availability (ie chimpanzee samples arenot available in a large number) All validation DMRs showedsignificant humanndashchimpanzee differences (Plt 005 one-sided Wilcoxon test) indicating that our discovery strategysuccessfully identified differentially methylated regions be-tween the two species Moreover all but two validationDMRs showed consistent human brain specific DNA meth-ylation in the extended comparison of humanndashchimpanzeendashmacaques The majority of DMRs were human-specific evenwhen we included two additional primate species (Hoolockgibbon and crab-eating macaques) (supplementary table S5Supplementary Material online) In sum human-specificmethylation patterns of the majority of DMRs were consis-tent in a large number of individuals and in a larger panel ofprimates The consistency between the discovery and valida-tion sets is remarkable considering the age and sex differences

Mendizabal et al doi101093molbevmsw176 MBE

2948Downloaded from httpsacademicoupcommbearticle-abstract331129472272475by gueston 11 February 2018

found in the two sets (the discovery set consisted mostly ofadult males the validation set is composed of individuals ofdifferent ages and both genders supplementary table S1Supplementary Material online) which further support theidea that these DMRs represent true species-specificdifferences

Due to the chemical nature of the bisulfite conversiontechnique DMRs could also potentially reflect underlyinggenetic polymorphisms instead of methylation levels (ie C-to-T polymorphisms not present in the reference genomescould be interpreted as unmethylated sites) To avoid con-founding effects of SNPs we identified variable positionswithin species by sequencing the validation DMR dataset atextremely high coverage (mean coveragegt 1400X supplementary table S4 Supplementary Material online) and ana-lyzing the data following the Genome Analysis Toolkit(GATK) pipeline (McKenna et al 2010) DMRs held thesame methylation patterns after excluding polymorphic po-sitions indicating that they are not artifacts of genetic poly-morphisms Figure 1 illustrates one of those validated DMRs

The prefrontal cortices of primates consist of different celltypes most notably glia and neurons which may exhibit dis-tinct epigenetic patterns at some loci Furthermore the ratiosof neurons versus glial cells are divergent between primate

brains (Herculano-Houzel 2014 Sherwood et al 2006)However our stringent cutoff value of methylation difference03 makes the DMRs robust against cellular heterogeneitybetween human and chimpanzee brains Specifically giventhe mean glianeuron ratios of 16 and 12 in humans andchimpanzees (Sherwood et al 2006) even in the most ex-treme case of a region being completely divergent betweenthe two cell types (eg 0 and 100 methylation in one celltype versus the other) the expected methylation differencebetween human and chimpanzee is only 8 In addition weanalyzed neuron- and non-neuron (mostly glia) -specific DNAmethylation markers from cell-sorted methylation datasets(Guintivano et al 2013 Kozlenkov et al 2014) and foundno evidence of significant effect of cell type compositionbias in our DMR (supplementary fig S1 and table S6Supplementary Material online)

Genomic Annotation of DMRsDifferentially methylated regions between human and chim-panzee brains identified in this study were on average 584 bplong (ranging between 67 and 2015 bp) They were signifi-cantly CpG enriched compared to genomic background (av-erage CpG OE ratio of 065gt13-fold enrichment comparedto control regions Plt 0001 based on 1000 bootstraps) We

FIG 1 DMR identification (A) Phylogenetic tree of the species analyzed in this study Sample sizes are shown for whole-genome bisulfite andtargeted bisulfite sequencing (N and n respectively) (B) Example of a DMR located 19 kb upstream of the promoter of Alpha-2C adrenergicreceptor (ADRA2C) Lines indicate smoothed methylation values from whole-genome bisulfite and each dot represents raw methylation values ofeach CpG site in the targeted validation dataset at high coverage Genomic coordinates correspond to the hg19

Epigenetic Loci of Human Brain Evolution doi101093molbevmsw176 MBE


first sought to examine the genomic locations of DMRs ac-cording to the annotation of transcripts in the UCSC GenomeBrowser Approximately half of DMRs (46) were found atknown promoters (fig 2) Interestingly slightly over half (54)of DMRs were located outside annotated gene promotersincluding many (30) DMRs found in distal intergenic re-gions (gt3kb away from known transcription start sites) and21 of DMRs within gene bodies

To gain further insights into the functional role of DMRswe examined enrichment of gene ontology (GO) andgenome-wide association studies (GWAS) at DMRs andneighboring regions Genes containing DMRs or closest toDMRs (supplementary table S2 Supplementary Material on-line) were significantly enriched for several GO developmen-tal processes including nervous system developmentforebrain development and embryonic morphogenesis (supplementary table S7 Supplementary Material online) Interms of GWAS 146 DMR regions (DMR 6 50 kb) over-lapped with GWAS signals Among these 24 DMR regionsharbored variants associated with brain-related traits (supplementary table S7 Supplementary Material online) includingthose enriched in neurological disorders such as Asperger andParkinsonrsquos disease (Plt 005 binomial test and FDR correc-tion using HapMap SNPs as a null distribution see ldquoMaterialand Methodsrdquo section and supplementary table S8Supplementary Material online) Significantly enriched traitsexclusively at human-specific DMRs (supplementary table S9Supplementary Material online) included immune responseskeleton and tooth development all with marked impacts onrecent human evolution (Nielsen et al 2007 Lachance andTishkoff 2013) These patterns suggest that DMRs are locatedin genomic regions involved in brain development and othertraits that are particularly relevant for our recent evolutionaryhistory

Coordinated Species-Specific Epigenetic Marks atDMRsTo understand the extent of coordinated evolutionary signa-tures of epigenetic modifications we analyzed human-specifichistone H3-trimethyl-lysine 4 (H3K4me3) modification datafrom prefrontal cortex neurons (Shulha et al 2012b) DMRsare significantly enriched in regions harboring human-specificenrichment or depletion of H3K4me3 (51 DMRs overlap withhuman specific signatures of H3K4me3 17-fold enrichmentcompared to control regions Plt 0001 based upon 1000bootstraps) In concordance with the known antagonisticdistribution of H3K4me3 and DNA methylation human-specific hypo- and hyper-DNA methylation associated withsignificant H3K4me3 enrichment and depletion respectively(Plt 106 v2 test fig 3) These observations illustrate coor-dinated epigenetic changes (of H3K4me3 and DNA methyl-ation) during the evolution of human brains

DMRs Cluster in Interacting Chromatin LoopsRecent analyses of three-dimensional configuration of eukary-otic genomes solidified the concept that the basic unit ofgenome organization is large (Mb scale) topologically as-sociated domains (TADs) that form hierarchical structures(Dixon et al 2012 Ea et al 2015) At the sub-TAD levelinteractions between distal sequences (such as between en-hancers and promoters) are accomplished by chromatinloops (Dixon et al 2012 Ea et al 2015) Chromatin loopsfacilitate transcriptional regulation by enabling enhancerndashpromoter interactions (Dixon et al 2012 Ea et al 2015Whalen et al 2016)

We hypothesized that some nearby DMRs may comprisechromatin loops that are potentially co-regulated IndeedDMRs were significantly clustered at megabase (Mb)-scalewhen compared to control regions with the similar CpG con-tent length and chromosomal distribution (supplementaryfig S3 Supplementary Material online) We further analyzedchromatin interaction maps compiled in the 4D GenomeDatabase (Teng et al 2015) We found 19 significant interac-tions among 32 unique DMRs as detected by Hi-C and ChiA-PET experiments (297-fold excess Plt 0001 1000 boot-straps supplementary table S10 Supplementary Material on-line) The majority of these interactions were found withinthe same DMR or between two adjacent DMRs but oneinteracting loop included 15 consecutive DMRs on chromo-some 2 Of note a smaller scale interaction within this largeloop was also captured by 3C assay of prefrontal cortex sam-ples (Shulha et al 2012b) Furthermore the enrichment waslargely driven by human hypo-methylated DMRs (122-foldenrichment Pfrac14 005 vs 045-fold of hyper-methylated re-gions Pfrac14 01) Considering that DMRs are on average anorder of magnitude shorter than chromatin loops (average584 bp vs 5145 bp for DMRs and chromatin loops respec-tively) we further investigated the degree of interactions ofDMRs including flanking regions (63 kb) We found thatnearly half of DMRs (Nfrac14 132) were involved in a total of227 significant chromatin interactions among themselves(odds ratio of 115 Pfrac14 0024 supplementary table S2Supplementary Material online) and 247 out of 278 DMRs

Promoter (2minus3kb)


Promoter (lt=1kb)

5 UTR

Exon

Intron

3 UTR

Downstream (lt3kb)

Distal Intergenic

30

13

27

6

14

4

32

FIG 2 Genic annotation of DMRs Annotation of DMRs with respectto known genes Promoter region is divided into three according todistance to the TSS (lt1 kb 1ndash2 kb and 2ndash3 kb) Downstream region isconsidered up to 3 kb downstream of gene end and distal intergenicis defined asgt3 kb away from any gene



(889) showed at least one significant interaction with othernon-DMR regions (107 fold enrichment Pfrac14 0007) ThusDMRs appear to preferentially locate in regions participatingin chromatin loops and therefore could potentially affecttranscriptional regulation

DMRs Mark Active PromoterEnhancers withEnriched TF Binding SignaturesWhile evolutionary data on epigenomic marks are sparseinformation on multiple epigenetic modifications from di-verse cell types and developmental stages is available for hu-mans In order to understand the regulatory potential ofDMRs we explored the chromatin states of DMRs in 98 dif-ferent tissues inferred from ChIP-Seq experiments on sixchromatin marks (H3K4me3 H3K4me1 H3K36me3H3K27me3 H3K9me3 and H3K27ac) (Bernstein et al2010) We found that human-specific hypo-methylatedDMRs were conspicuously marked by active regulatory chro-matin states mainly promoter and enhancer states (fig 4) Incontrast human hyper-methylated regions harbored a largenumber of DMRs classified as quiescent chromatin statesacross tissue types (supplementary fig S4 SupplementaryMaterial online) Interestingly the promoter and enhancerstates of hypo-methylated DMRs were highest at brain

samples although not limited to this tissue type In contrastmany DMRs currently annotated in gene deserts exhibit ac-tive promoter chromatin marks nearly exclusively in braintissues (supplementary fig S5 Supplementary Materialonline)

We further analyzed binding of specific transcription fac-tors (TFs) using ChIP-Seq data of 161 transcription factors in91 human cell lines from diverse tissues in the ENCODE proj-ect This analysis revealed that DMRs were highly enriched inexperimentally validated functional transcription factor bind-ing sites (TFBS) A majority of DMRs (845) overlapped withTFBS (194 fold enrichment versus control regions Plt 0001)Human-specific hypo-methylated DMRs were significantlymore enriched with TFBS than hyper-methylated ones(Pfrac14 00015 v2 test) Moreover we found that four specifictranscription factors were significantly more frequentlybound in DMRs than expected whereas bindings of 18 tran-scription factors were significantly depleted in DMRs(Plt 005 fig 5A) Among the enriched transcription factorswe found NRF1 which associates with neurodegenerativediseases (Lee et al 2011) and BRF1 implicated in neurode-velopmental processes (Borck et al 2015) The experimentallysupported transcription factor binding activity of DMRs andtheir promoter and enhancer-like epigenetic features

FIG 3 Coordinated epigenetic changes at DMRs (A) Example of human-hypomethylated DMR in human prefrontal cortex (hg19-based coor-dinates) (B) The region in panel A co-localizes with a human-specific H3K4me3 peak in cell sorted neurons (C) Human-specific hyper-meth-ylation of a DMR (D) The region in panel C overlaps with a human-specific depletion of H3K4me3 histone mark whereas chimpanzee andmacaque show significant enrichment Additional examples shown in supplementary figure S1



advocate an active role of DMRs as regulatory hotspots inhuman brains

DMRs Associate with Gene Expression DifferencesWe tested whether species-specific epigenetic changes andtranscription factor binding potential of DMRs directly trans-lated into gene expression differences We analyzed threedifferent expression datasets generated via different methods(RNA-seq SuperSAGE and microarray) from prefrontal cor-tices of human chimpanzee and macaques (Brawand et al2011 Konopka et al 2012 Liu et al 2012) Specifically weexamined whether gene expression differences if any wereconsistent with the silencing effect of DNA methylation Ofnote the consistency across datasets was low since only 66out of 6725 genes showed significant and consistent human-specific gene expression in all three datasets [in part due tolow sample size eg only 3 and 4 macaques were studied inBrawand et al (2011) and Konopka et al (2012) respectively]Nonetheless we found that 13 out of 61 human-specificDMRs with both methylation and expression informationassociated with significant human-specific expression pat-terns in at least one dataset (Plt 005 in human-chimpanzee and human-macaque comparisons Wilcoxontests) including genes with key roles on neurological processessuch as CSPG5 SPG7 and COBL (supplementary table S11

Supplementary Material online) We next tested whether theDMR-genes show an excess of human-specific expressioncompared to genes selected at random (1000 randomiza-tions) in each of the datasets independently as well as inthe combined dataset We observed a significant associationbetween human-specific DMR-genes and human-specific ex-pression patterns in one of the three datasets (the one withthe largest sample sizes per species) as well as when all data-sets were considered together (with 37 humans 24 chimpan-zee and 38 macaques supplementary table S12Supplementary Material online) Therefore despite the limi-tations of the available datasets these results suggest thathuman-specific DMRs associate with gene expression changesof adjacent genes

Signatures of Human-Specific Mutations at TFBSMotifs at DMRsHow do DMRs arise between species One clue may begained from recent studies demonstrating the effect ofgenetic variation on epigenetic divergence (Hernando-Herraez et al 2015) Genetic changes can cause epigeneticchanges is by generating or depleting functional transcrip-tion factor binding sites since binding of transcription fac-tors are known to reduce DNA methylation (eg Schubeler2015) Consequently we tested whether human-specific

FIG 4 Chromatin states at human hypo-methylated DMRs Heatmap of the fraction of DMRs exhibiting distinctive chromatin states (rows) indifferent tissues (columns) Each cell of the heatmap indicates the proportion of DMRs classified as that specific chromatin state in a given tissuetype



hypo-methylated DMRs associate with novel TFBS by iden-tifying single nucleotide substitutions at functional TF mo-tifs within or around DMRs (63kb) We focused onexperimentally validated TFBS sites (ChIP-Seq peaks withinthe ENCODE project) and fixed substitutions in thehuman lineage (ielt1 frequency in the 1000 Genomesdataset) We further restricted to substitutions that gener-ate a major TFBS (ie the position weight matrix of thehuman-specific nucleotide is at least 08)

We identified a total of 41 human substitutions within 51TFBS motifs from this analysis (supplementary table S13Supplementary Material online) One such mutation affectsthe E-box canonical binding sequence (enhancer-boxCANNTG) in the promoter of the PSMB2 gene where ahuman-specific allele represents a putative de novo bindingsite for BHLHE40 (fig 5B and C) The transcription factorBHLHE40 is a well-known circadian gene with strong cyclicexpression in human brains that is deregulated in major de-pressive disorders Consistent with the presence of the E-boxmotif and human-specific hypo-methylated DMR in the pro-moter of PSMB2 human brains have increased PSMB2 ex-pression levels compared to chimpanzees and macaques (fig5D) Even though the motif was only predicted for BHLHE40E-boxes are known to be utilized by a variety of transcriptionfactors for regulation of gene expression in diverse tissuesincluding neurons (Massari and Murre 2000) Indeed a largenumber of additional TFs (a total of 71) show binding to thatspecific region in ChIP-Seq experiments (supplementary tableS14 Supplementary Material online) Notably one of these

TFs that bind to this region is FOXP2 a gene important forspeech and language that regulates a suite of transcriptionalnetworks in the human brain (Konopka et al 2009) Anotherexample of human-specific TFBS mutation with associatedexpression divergence is the stearoyl-CoA desaturase (SCD)gene (supplementary fig S6 Supplementary Material online)Interestingly mutations that generate functional transcrip-tion binding sites in DMRs are observed on average at amuch lower rate compared to binding sites outside DMRs(nearly 6-fold less frequent Plt 002 1000 bootstraps) sug-gesting that such mutations do not occur frequently at re-gions of strong functional constraint

DiscussionThe human brain represents a challenging yet extremely in-triguing system to study the intersection of evolution epige-netics and diseases Human brains are characterized bydramatic evolutionary innovations such as increased brainsize and cortical reorganization Genetic and transcriptomicstudies have revealed extensive changes in gene expressionand gene-regulatory networks that could provide the basis forevolutionary specializations of human brain structure andfunction (Brawand et al 2011 Konopka et al 2012Bauernfeind et al 2015) Epigenetic modifications can causesuch transcriptional changes even in the absence of substan-tial sequence changes In fact epigenetics is known to bedeeply involved in regulation of brain functions in humans(Akbarian and Huang 2009 Cheung et al 2010 Zhu et al

FIG 5 Transcription factor binding sites at DMRs (A) Transcription factors significantly enriched or depleted at DMRs (BndashC) Human-specificnucleotide change in position 3 of the binding site by transcription factor BHLHE4 This position is 11 kbp upstream of a human hypo-methylatedDMR in the promoter of PSMB2 gene compared to chimpanzee and macaque (D) Human-specific expression increase of PSMB2 gene observed inthree different data sets



2013 Kozlenkov et al 2014) as well as in several neuropsy-chiatric diseases (Shulha et al 2012a Jaffe et al 2016)

To understand the role of epigenetic regulation on theevolution of human brains we compared the methylomesfrom cortices of human and several nonhuman primates inparticular with chimpanzees We combined genome-scansnonparametric testing and stringent cutoff criteria and iden-tified nearly 300 differentially methylated regions betweenhuman and chimpanzee cortices To further confirm the ob-served patterns we performed high quality targeted experi-ments on a validation panel which included one of the largestnumber of brain specimens used for comparative epigeneticstudies Differential DNA methylation between human andchimpanzee brains was confirmed in all the validation DMRstested and most (gt70) were human-specific in the largerpanel of all five species These results provide a strong supportfor evolutionary divergence of DNA methylation Our crite-rion of methylation differencegt 03 is extremely stringentcompared to the known examples of putatively functionalmethylation differences (Dias and Ressler 2014 Tobi et al2014 Jaffe et al 2016) For example we validate regions pre-viously identified by targeted analyses (Farcas et al 2009 Shiet al 2014) although the observed fractional methylationdifferences fall below our criteria Future studies with largernumbers of samples (specially for chimpanzees) and ad-vanced statistical tools will likely identify many additionalgenomic regions of epigenetic divergence

Using newly generated brain methylomes of rhesus ma-caques we inferred evolutionary origins of DMRs following asimple parsimony rule Rhesus macaques are often used asoutgroups to assign evolutionary polarity to either the humanor chimpanzee lineages in DNA sequence and gene expres-sion comparisons (Uddin et al 2004 Kim et al 2006 Konopkaet al 2012 Rogers and Gibbs 2014) However accuratelyquantifying how rapidly DNA methylation evolves and deter-mining whether the use of rhesus macaques is adequate re-quire future research Large-scale phylogenetic analyses ofDNA methylation indicate some degree of conservation(Mendizabal et al 2014 Keller et al 2016) Indeed when wetest the difference between human and macaque DMRs allbut two putatively human-specific DMRs fit the expectationbased upon parsimony With few exceptions data from crab-eating macaques (M fascicularis) are largely consistent withthose from rhesus macaques These observations begin toprovide information on the rates of epigenetic changes duringevolution

Even though many DMRs are found proximal to knownTSS the majority of DMRs locate outside promoters Thisfinding highlights the importance of employing unbiasedwhole-epigenome approaches in contrast to those that in-terrogate predetermined regions such as gene promoters orclassical CpG islands (Enard et al 2004 Farcas et al 2009 Zenget al 2012 Shi et al 2014) Genomic and epigenomic analysesof DMRs in this study suggest that non-promoter DMRs havesimilar transcriptional potential as promoter DMRs as judgedby their epigenetic states TF binding signatures and associ-ation with gene expression (supplementary fig S7Supplementary Material online) In particular many DMRs

found in intergenic regions also exhibit chromatin features ofactive promoters which interestingly show high brain specif-icity Taken together human brain-specific DNA methylationchanges may contribute to regulation of transcriptionIndeed in light of the available expression datasets fromthe cortices of the three species human-specific DMRs aresignificantly enriched for human-specific gene expression di-vergence at adjacent genes in the direction that is expectedaccording to the methylation patterns of the DMRs Finally inthe human-specific DMRs there was a significant excess ofhypo-methylated DMRs This could potentially be in line withincreased gene expression levels in human brains (Cacereset al 2003 Preuss et al 2004) although some studies castdoubt on the generality of this pattern (Uddin et al 2004Babbitt et al 2010)

Genes overlapping with DMRs and DMR-adjacent geneswere enriched for several functional categories related tobrain development Moreover many GWAS loci of neuropsy-chiatric diseases were over-represented in DMR regions sug-gesting that some of these GWAS SNPs may affect epigeneticmodifications On the other hand DMRs also associated withGWAS variants and GO categories not related with brainfunctions but with other traits known to have played signif-icant roles in human evolution In fact many DMRs showedchromatin features of promoters and enhancers in nonbraintissues as well This can be explained by the fact that DNAmethylation patterns are often stable across tissues and de-velopmental stages (Hon et al 2013 Ziller et al 2013 Zenget al 2014) For example hypo-methylated embryonic en-hancers can maintain their distinctive methylation profilesin adult tissues even without direct enhancer activities inthose tissues (Hon et al 2013) Moreover some of the func-tional ontology categories enriched in the current study over-lap with those identified in previous comparative epigenomicstudies despite different methodologies andor different tis-sues used (Pai et al 2011 Hernando-Herraez et al 2013Hernando-Herraez et al 2015) We propose that some ofthe DMRs are broadly used as regulatory regions in diversetissues or active regulatory elements in embryonic or fetaltissues that retained their epigenetic memories in adultbrains

Our study highlights the synergistic nature of epigeneticchanges during evolution DMRs extensively overlap withhuman-specific signatures of H3K4me3 modifications identi-fied in a previous study (Shulha et al 2012b) Moreover manyof the DMRs cluster locally and exhibit evidence of physicalinteractions by participating in chromatin loops indicatingthat human brain-specific epigenetic changes affect large-scale spatial organization of three-dimensional interactionsamong chromosomal regions (Shulha et al 2012b) Notablya strong signature of human-specific hypo-methylation over-laps with a human-specific H3K4me3 peak in the secondintron of DDP10 gene (supplementary fig S1Supplementary Material online) which also harbors the TSSof a human expressed sequence tag (BF979467) DPP10 en-codes a dipeptidyl peptidase-related protein involved in neu-ronal excitability that plays roles in brain developmentGenetic variation in DPP10 is associated with autism bipolar



disorder and schizophrenia (Djurovic et al 2010) DMRsfound in the vicinity of the DPP10 locus participate in inter-acting domains of chromatin loops in a fetal lung derived cellline (Jin et al 2013) as well as in human prefrontal cortexsamples (Shulha et al 2012b) These observations indicatethat hotspots of human-specific epigenetic modificationmay exist in this region

We identified several fixed nucleotide substitutions in thehuman lineage that associate with human-specific epigeneticand gene expression divergence The patterns of nucleotidesubstitution DNA methylation changes transcription factorbinding and gene expression profiles of these mutations areall in accord with the well-established functional roles of dif-ferent components of gene expression regulationComparative epigenetic studies such as the current investi-gation can provide critical information to prioritize candidateregulatory changes for necessary experimental validation ul-timately to elucidate the functional roles of human-specificsequence changes

Materials and Methods

Whole Genome Bisulfite Sequencing MethodBriefly genomic DNA from macaque prefrontal cortices wasextracted following the Qiagen DNeasy Blood amp Tissue Kit Intotal 300ndash500 ng of genomic DNA were then fragmented bysonication end repaired and ligated to custom-synthesizedmethylated Illumina PE adapters (Eurofins MWG OperonHuntsville AL) according to manufacturerrsquos (Illumina SanDiego CA) instructions for gDNA library constructionAdaptor-ligated libraries were subjected to two successivetreatments of sodium bisulfite conversion using the EpiTectBisulfite kit (Qiagen Valencia CA) as outlined in the manu-facturerrsquos instructions Five to ten nanograms of bisulfite-converted libraries was PCR amplified with the following con-dition 25 U of ExTaq DNA polymerase (Takara) 5ml of 10XExtaq reaction buffer 25mM dNTPs 1ml of 10 mM PE PCRprimer 10 and 1ml of 10 mM PE PCR Primer 20 in a 50ml PCRreaction The thermocyling was as follows 95 C 3 min then10 cycles of 95 C 30 s 65 C 30 s and 72 C 60 s The enrichedlibraries were purified twice with SPRI method using 08x vvAMPure beads (Beckman Coulter Inc Brea CA) andsequenced using an Illumina HiSeq2000 at the Vincent JCoates Genomic Sequencing Laboratory at the University ofCalifornia Berkeley that was supported by NIH S10Instrumentation Grants S10RR029668 and S10RR027303WGBS data from human and chimpanzee prefrontal corticeswere downloaded from GEO (GSE37202)

Mapping and DMR IdentificationRaw files were processed by FastQC for quality controlQuality and adapter trimming was performed usingTrimGalore (Babraham Institute) Reads were mapped tothe respective reference genomes (hg19 pantro4 andrheMac3 for human chimpanzee and macaque respectively)using Bismark (Krueger and Andrews 2011) and BsMap (Xiand Li 2009) We chose to remove all reads with identical startand ends considering them as duplicates This strategy is

likely to be conservative given the uncertainty on the bestpractice to deal with such reads (Balzer et al 2013) After de-duplication we calculated fractional methylation levels atindividual CpGs (Lister et al 2009)

We identified differentially methylated regions (DMRs) be-tween human and chimpanzee brain methylomes usingBSmooth (Hansen et al 2012) We used quantile cutoff of1 of the t-statistics using only CpGs supported by at leasttwo reads in two or more individuals per species We addi-tionally filtered the detected DMRs to select those with atleast 10 CpGs per DMR and a minimum difference of frac-tional methylation level of 03 between the two species DMRsidentified from the two mappers were combined

DMR PolarizationWe used the macaque methylome data to polarize the DMRsA DMR was classified as human-specific when the methyla-tion difference between the human and macaque was greaterthan the difference between chimpanzee and macaque Usingthis simple parsimony approach 100 DMRs could be assignedas human-specific and 127 DMRs as chimpanzee-specificThe remaining 51 DMRs did not have matching CpGsmapped thus remained unclassified We then imposed addi-tional criteria to make our inference more conservativeSpecifically the difference between human and macaqueshould be greater than 015 which led to 85 human-specificand 102 chimpanzee-specific DMRs A stricter threshold of030 was also applied which resulted in 62 human-specific 72chimp-specific and 144 unknown DMRs (supplementarytable S2 Supplementary Material online)

Targeted Validation AnalysesWe performed genomic and bisulfite sequencing for 38 DMRson the prefrontal cortices of 23 humans 2 chimpanzees 1gibbon 7 rhesus macaques and 5 crab eating macaques usinga MiSeq machine (supplementary table S4 SupplementaryMaterial online) Frozen tissues were obtained from the pre-frontal cortices of the individuals all of which had no knownneuronal diseases or history of drug abuse For the humansubjects informed written consent was obtained from therelatives of the human subjects prior to sample collection oranalysis The internal review board of Kunming Institute ofZoology Chinese Academy of Sciences approved all proto-cols of this study

Genomic DNA was extracted according to the DNeasyBloodampTissue Kit (QIAGEN) protocol We used the EpiTectBisulfite Kits (Qiagen Valencia CA) to conduct bisulfiteconversions of genomic DNA following the manufacturerrsquosinstructions Sodium bisulfite converts unmethylated cy-tosine to uracil which is then PCR amplified as thymidineand methylated cytosine remain cytosine PCR primerswere designed using Methyl Primer Express 10 (ABI) Todetect potential SNPs we also performed the PCR to theuntreated genomic DNA PCR primers were designed usingPrimer Premier 5 The primer information is available un-der request

The targeted bisulfite PCR products from single reactionswere first quantified before proceeding with sequencing For



quality control purposes we also performed qPCR Followingthe Nextera XT DNA Sample Preparation protocol we used1 ng of total DNA (02ngml) per sample to prepare MiSeqlibraries Genomic libraries of individuals used for DMR vali-dation were also generated and subjected to MiSeq sequenc-ing All of the MiSeq sequencing was performed in theKunming Institute of Zoology sequencing core

Mapping and Analyses of Validation DatasetFor validation we performed bisulfite sequencing on 38DMRs using a MiSeq machine One region failed to be am-plified in all individuals After quality control reads weremapped using Bismark (Krueger and Andrews 2011) to thefollowing reference genomes hg19 pantro4 rheMac3macFas5 and nomLeu3 for human chimpanzee rhesusMacaque crab eating macaque and gibbon respectively

We also sequenced the corresponding genomic regions ofthe specific individuals analyzed to identify any potentialSNPs that may affect our inference Genome sequence readswere aligned using bwa (Li and Durbin 2009) Duplicate readswere removed using Picard (httppicardsourceforgenet)We then identified SNPs using GATK v34 (McKenna et al2010) and using standard hard filtering parameters (QDlt 20k FSgt 600 k MQlt 400 k MQRankSumlt125 kReadPosRankSumlt80)

Functional Annotation and Epigenetic Features ofDMRsTo test the significance of different genomic features ofDMRs we selected 1000 control regions of same lengthand chromosome location as the DMRs and containing atleast 10 CpGs (the same criteria we used for DMR discovery)P-values were computed by counting the number of simula-tions showing values as extreme as the observed oneEnrichments were computed as the ratio between the ob-served value and the mean of the control regions

We used ChIPSeeker (Yu et al 2015) to annotate DMRs togenes using hg19 KnowGene table from UCSC GO analyseswere performed using GoStat Rpackage (Falcon andGentleman 2007) and using 13455 humanndashchimpanzeeorthologous genes (The Chimpanzee Sequencing andAnalysis Consortium 2005) GWAS enrichment analyseswere performed with traseR package (Chen and Qin 2015)Specifically we considered 44078 SNP-trait associations fromthe Association Results Browser (httpwwwncbinlmnihgovprojectsgapplusprevsgap_plushtm last accessed 24August 2016) that combines associations from both dbGaPand NHGRI GWAS Catalog We also considered linked SNPs(LDgt08 and located within 100 kb of GWAS SNPs inEuropean populations from Thousand Genomes hg19)Including linked variants we have 90700 SNP-trait associationsand 78247 unique trait-associated SNP and 573 unique traitsFor enrichment testing we considered background SNPs fromHapMap CEU population (HapMap phases Ithorn IIthorn III4029840 variants excluding those on Y-chromosome)Specifically a contingency table was created using counts ofthe number of trait-associated SNPs found within DMR andnon-DMR regions and compared those numbers with the

counts for HapMap SNPs that fall in DMR vs non-DMR re-gions Significance was tested using binomial test and FDRcorrection was applied for multiple testing We tested eachof the original 573 traits independently as well as grouped into26 categories (supplementary table S8 SupplementaryMaterial online) The rationale for joining traits into relatedcategories was to gain statistical power as well as to accountfor biases due to heterogeneity on reported trait names acrossGWAS studies (ie the independent entries in the databasealcohol drinking alcoholism or drinking behavior were consid-ered within Addictive Behavior category supplementary tableS8 Supplementary Material online)

We identified the overlap between our DMR dataset andthe human-specific H3K4me3 peaks (Shulha et al 2012b) Forplotting purposes we re-mapped raw reads to hg19 pantro4and rheMac3 genomes We analyzed the chromatin interac-tions (those with Confidence Score 1 005) from the4DGenome Database (Teng et al 2015) We analyzed dataderived from ENCODE ChIP-Seq experiments (UCSCwgEncodeRegTfbsClusteredV3 table for peaks andfactorbookMotifCanonical for motifs)

We explored the 18 chromatin state-model maps from theRoadmap Epigenomics dataset (Roadmap EpigenomicsConsortium et al 2015) We joined several categories relatedto similar states such as active Tss (TssA TssFlnk TssFlnkUand TssFlnkD) enhancers (EnhG1 EnhG2 EnhA1 EnhA2and EnhWk) transcription (Tx and TxWk) and repressedPolyComb (ReprPC and ReprPCWk) yielding a total of ninefinal states We followed the tissue classification in the originaldataset In addition we combined related tissues to the fol-lowing categories ldquofetalrdquo (IMR90 fetal lung fibroblasts CellLine Fetal Adrenal Gland Fetal Brain Male Fetal BrainFemale Fetal Heart Fetal Intestine Large Fetal IntestineSmall Fetal Kidney Fetal Lung Fetal Muscle Trunk FetalMuscle Leg Fetal Stomach and Fetal Thymus) and ldquostem-cellrdquo (H1 Derived Mesenchymal Stem Cells Adipose DerivedMesenchymal Stem Cell Cultured Cells Bone MarrowDerived Cultured Mesenchymal Stem Cells Primary hemato-poietic stem cells Primary hematopoietic stem cells shortterm culture Primary hematopoietic stem cells G-CSF-mobi-lized Female and Primary hematopoietic stem cells G-CSF-mobilized Male)

We further analyzed binding of specific transcription fac-tors (TFs) using ChIP-Seq data of 161 transcription factors in91 human cell lines from diverse tissues in the ENCODE proj-ect (Gerstein et al 2012) For human-specific mutations po-tentially generating novel TFBS within DMRs we excludedDMRs that were chimp-specific Then we searched for posi-tions within and around DMRs (3 kb) where human refer-ence genomes showed a different allele compared to that inmacaque and chimpanzee genomes (which need to be iden-tical) We only considered positions within TFBS motifs wherethe human nucleotide represented at least 80 frequencywithin the motif Finally we excluded any positions thatwere polymorphic within humans (ie non-reference al-leles withgt1 frequency in worldwide populations in the1000Genome Dataset (The 1000 Genomes ProjectConsortium 2015)



Analyses of Gene ExpressionWe focused on human-specific DMRs and tested if humansshowed significant gene expression values (Wilcoxon testPlt 005) in both humanndashchimpanzee and humanndashmacaquecomparisons (assuming repressive effect of DNA methylationon expression ie human hypo-methylated DMR increasedexpression in humans) To test for significance at genome-wide level we randomly selected the same number of genesas in the DMR gene sets from the total genes in a givendataset (while maintaining the observed numbers of hypo-and hyper-methylation) and examined whether they showincrease or decrease of gene expression We repeated thisprocess 1000 times and counted the number of genes ex-hibiting human-specific expression values with Plt005 Wetested each dataset independently as well as using a com-bined data set (using standardized z-scores and the subset ofgenes overlapping among the three datasets)

Supplementary MaterialSupplementary figures S1 to S7 and tables S1 to S14 are avail-able at Molecular Biology and Evolution online (httpwwwmbeoxfordjournalsorg)

AcknowledgmentThis work was supported by the National Science Foundation(SBE-131719 to SVY) the National Institutes of Health(1R01MH103517-01A1 to SVY GK and TMP) the StrategicPriority Research Program of the Chinese Academy ofSciences (XDB13010000 to BS) the National NaturalScience Foundation of China (NSFC 31130051 to BS andNSFC to LS) the Youth Innovation Promotion Associationof the Chinese Academy of Sciences (to LS) the ResearchPersonnel Improvement Program by the Department ofEducation Language Policy and Culture by the BasqueGovernment (POS_2013_1_130 to IM) and the YerkesNational Primate Center Base Grant (ORIPODP51OD011132)

ReferencesAkbarian S Huang HS 2009 Epigenetic regulation in human brain-focus

on histone lysine methylation Biol Psychiatry 65198ndash203Babbitt CC Fedrigo O Pfefferle AD Boyle AP Horvath JE Furey TS Wray

GA 2010 Both noncoding and protein-coding RNAs contribute togene expression evolution in the primate brain Genome Biol Evol267ndash79

Balzer S Malde K Grohme MA Jonassen I 2013 Filtering duplicate readsfrom 454 pyrosequencing data Bioinformatics 29830ndash836

Bauernfeind AL Reyzer ML Caprioli RM Ely JJ Babbitt CC Wray GA HofPR Sherwood CC 2015 High spatial resolution proteomic compar-ison of the brain in humans and chimpanzees J Comp Neurol5232043ndash2061

Bernstein BE Stamatoyannopoulos JA Costello JF Ren B Milosavljevic AMeissner A Kellis M Marra MA Beaudet AL Ecker JR et al 2010The NIH Roadmap Epigenomics Mapping Consortium NatBiotechnol 281045ndash1048

Borck G Hog F Dentici ML Tan PL Sowada N Medeira A Gueneau LThiele H Kousi M Lepri F et al 2015 BRF1 mutations alter RNApolymerase III-dependent transcription and cause neurodevelop-mental anomalies Genome Res 25155ndash166

Brawand D Soumillon M Necsulea A Julien P Csardi G Harrigan PWeier M Liechti A Aximu-Petri A Kircher M et al 2011 The evo-lution of gene expression levels in mammalian organs Nature478343ndash348

Caceres M Lachuer J Zapala MA Redmond JC Kudo L Geschwind DHLockhart DJ Preuss TM Barlow C 2003 Elevated gene expressionlevels distinguish human from non-human primate brains Proc NatlAcad Sci U S A 10013030ndash13035

Chen L Qin ZS 2015 traseR an R package for performing trait-associated SNP enrichment analysis in genomic intervalsBioinformatics 151214ndash1216

Cheung I Shulha HP Jiang Y Matevossian A Wang J Weng Z AkbarianS 2010 Developmental regulation and individual differences of neu-ronal H3K4me3 epigenomes in the prefrontal cortex Proc Natl AcadSci U S A 1078824ndash8829

Daxinger L Whitelaw E 2012 Understanding transgenerational epige-netic inheritance via the gametes in mammals Nat Rev Genet13153ndash162

Dias BG Ressler KJ 2014 Parental olfactory experience influences be-havior and neural structure in subsequent generations Nat Neurosci1789ndash96

Dixon JR Selvaraj S Yue F Kim A Li Y Shen Y Hu M Liu JS Ren B 2012Topological domains in mammalian genomes identified by analysisof chromatin interactions Nature 485376ndash380

Djurovic S Gustafsson O Mattingsdal M Athanasiu L Bjella T Tesli MAgartz I Lorentzen S Melle I Morken G Andreassen OA 2010 Agenome-wide association study of bipolar disorder in Norwegianindividuals followed by replication in Icelandic sample J AffectDisord 126312ndash316

Ea V Baudement M-O Lesne A Forne T 2015 Contribution of topo-logical domains and loop formation to 3D chromatin organizationGene 6734ndash750

Enard W Fassbender A Model F Adorjan P Peuroaeuroabo S Olek A 2004Differences in DNA methylation patterns between humans andchimpanzees Curr Biol 14R148ndashR149

Falcon S Gentleman R 2007 Using GOstats to test gene lists for GOterm association Bioinformatics 23257ndash258

Farcas R Schneider E Frauenknecht K Kondova I Bontrop R Bohl JNavarro B Metzler M Zischler H Zechner U et al 2009 Differencesin DNA methylation patterns and expression of the CCRK gene inhuman and nonhuman primate cortices Mol Biol Evol261379ndash1389

Ferguson-Smith AC 2011 Genomic imprinting the emergence of anepigenetic paradigm Nat Rev Genet 12565ndash575

Gerstein MB Kundaje A Hariharan M Landt SG Yan KK Cheng C MuXJ Khurana E Rozowsky J Alexander R et al 2012 Architecture ofthe human regulatory network derived from ENCODE data Nature48991ndash100

Gifford CA Ziller MJ Gu H Trapnell C Donaghey J Tsankov A ShalekAK Kelley DR Shishkin AA Issner R et al 2013 Transcriptional andepigenetic dynamics during specification of human embryonic stemcells Cell 1531149ndash1163

Guintivano J Aryee MJ Kaminsky ZA 2013 A cell epigenotype specificmodel for the correction of brain cellular heterogeneity bias and itsapplication to age brain region and major depression Epigenetics8290ndash302

Hanna CW Pe~naherrera MS Saadeh H Andrews S McFadden DEKelsey G Robinson WP 2016 Pervasive polymorphic imprintedmethylation in the human placenta Genome Res 26756ndash767

Hansen KD Langmead B Irizarry RA 2012 BSmooth from whole ge-nome bisulfite sequencing reads to differentially methylated regionsGenome Biol 13R83

Heard E Martienssen RA 2014 Transgenerational epigenetic inheri-tance myths and mechanisms Cell 15795ndash109

Herculano-Houzel S 2014 The glianeuron ratio how it varies uniformlyacross brain structures and species and what that means for brainphysiology and evolution Glia 621377ndash1391

Hernando-Herraez I Heyn H Fernandez-Callejo M Vidal E Fernandez-Bellon H Prado-Martinez J Sharp AJ Esteller M Marques-Bonet T



2015 The interplay between DNA methylation and sequencedivergence in recent human evolution Nucleic Acids Res438204ndash8214

Hernando-Herraez I Prado-Martinez J Garg P Fernandez-Callejo MHeyn H Hvilsom C Navarro A Esteller M Sharp AJ Marques-Bonet T 2013 Dynamics of DNA methylation in recent humanand great Ape evolution PLoS Genet 9e1003763

Hon GC Rajagopal N Shen Y McCleary DF Yue F Dang MD Ren B2013 Epigenetic memory at embryonic enhancers identified in DNAmethylation maps from adult mouse tissues Nat Genet451198ndash1206

Houston I Peter CJ Mitchell A Straubhaar J Rogaev E Akbarian S 2013Epigenetics in the human brain Neuropsychopharmacology38183ndash197

Huh I Zeng J Park T Yi SV 2013 DNA methylation and transcriptionalnoise Epigenetics Chromatin 69

Jaffe AE Gao Y Deep-Soboslay A Tao R Hyde TM Weinberger DRKleinman JE 2016 Mapping DNA methylation across developmentgenotype and schizophrenia in the human frontal cortex NatNeurosci 1940ndash47

Jerison HJ 1973 Evolution of the Brain and Intelligence New YorkAcademic Press

Jin F Li Y Dixon JR Selvaraj S Ye Z Lee AY Yen C-A Schmitt ADEspinoza CA Ren B 2013 A high-resolution map of the three-dimensional chromatin interactome in human cells Nature503290ndash294

Kaas JH Preuss TM 2013 Human Brain Evolution In Squire L BergD Bloom FE du Lac S Ghosh A Spitzer NC editorsFundamental neuroscience 4th ed San Diego (CA) AcademicPress p 901ndash918

Keller TE Han P Yi SV 2016 Evolutionary transition of promoter andgene body DNA methylation across invertebrate-vertebrate bound-ary Mol Biol Evol 331019ndash1028

Kim SH Elango N Warden C Vigoda E Yi SV 2006 Heterogeneousgenomic molecular clocks in primates PLoS Genet 2e163

Konopka G Bomar JM Winden K Coppola G Jonsson ZO Gao F PengS Preuss TM Wohlschlegel JA Geschwind DH 2009 Human-specifictranscriptional regulation of CNS development genes by FOXP2Nature 462213ndash217

Konopka G Friedrich T Davis-Turak J Winden K Oldham MC Gao FChen L Wang GZ Luo R Preuss TM Geschwind DH 2012 Human-specific transcriptional networks in the brain Neuron 75601ndash617

Kozlenkov A Roussos P Timashpolsky A Barbu M Rudchenko SBibikova M Klotzle B Byne W Lyddon R Di Narzo AF et al 2014Differences in DNA methylation between human neuronal and glialcells are concentrated in enhancers and non-CpG sites Nucleic AcidsRes 42109ndash127

Krueger F Andrews SR 2011 Bismark a flexible aligner and methylationcaller for Bisulfite-Seq applications Bioinformatics 271571ndash1572

Lachance J Tishkoff SA 2013 Population Genomics of HumanAdaptation Annu Rev Ecol Evol Syst 44123ndash143

Lee CS Lee C Hu T Nguyen JM Zhang J Martin MV Vawter MP HuangEJ Chan JY 2011 Loss of nuclear factor E2-related factor 1 in thebrain leads to dysregulation of proteasome gene expression andneurodegeneration Proc Natl Acad Sci U S A 1088408ndash8413

Li H Durbin R 2009 Fast and accurate short read alignment withBurrows-Wheeler transform Bioinformatics 251754ndash1760

Lister R Pelizzola M Dowen RH Hawkins RD Hon G Tonti-Filippini JNery JR Lee L Ye Z Ngo QM et al 2009 Human DNA methylomesat base resolution show widespread epigenomic differences Nature462315ndash322

Liu X Somel M Tang L Yan Z Jiang X Guo S Yuan Y He L Oleksiak AZhang Y et al 2012 Extension of cortical synaptic developmentdistinguishes humans from chimpanzees and macaques GenomeRes 22611ndash622

Massari ME Murre C 2000 Helix-loop-helix proteins regulators of tran-scription in eucaryotic organisms Mol Cell Biol 20429ndash440

McKenna A Hanna M Banks E Sivachenko A Cibulskis K Kernytsky AGarimella K Altshuler D Gabriel S Daly M DePristo MA 2010 TheGenome Analysis Toolkit a MapReduce framework for analyzingnext-generation DNA sequencing data Genome Res 201297ndash1303

Mendizabal I Keller TE Zeng J Yi SV 2014 Epigenetics and evolutionIntegr Comp Biol 5431ndash42

Mendizabal I Yi SV 2016 Whole-genome bisulfite sequencing mapsfrom multiple human tissues reveal novel CpG islands associatedwith tissue-specific regulation Hum Mol Genet 2569ndash82

Nielsen R Hellmann I Hubisz MJ Bustamante C Clark AG 2007 Recentand ongoing selection in the human genome Nat Rev Genet8857ndash868

Nikolova YS Koenen KC Galea S Wang C-M Seney ML Sibille EWilliamson DE Hariri AR 2014 Beyond genotype serotonin trans-porter epigenetic modification predicts human brain function NatNeurosci 171153ndash1155

Pai AA Bell JT Marioni JC Pritchard JK Gilad Y 2011 A genome-widestudy of DNA methylation patterns and gene expressionlevels in multiple human and chimpanzee tissues PLoS Genet7e1001316

Preuss TM 2011 The human brain rewired and running hot Ann N YAcad Sci 1225 Suppl 1E182ndashE191

Preuss TM Caceres M Oldham MC Geschwind DH 2004 Human brainevolution insights from microarrays Nat Rev Genet 5850ndash860

Roadmap Epigenomics Consortium Kundaje A Meuleman W Ernst JBilenky M Yen A Heravi-Moussavi A Kheradpour P Zhang Z WangJ et al 2015 Integrative analysis of 111 reference human epige-nomes Nature 518317ndash330

Rogers J Gibbs RA 2014 Comparative primate genomics emergingpatterns of genome content and dynamics Nat Rev Genet15347ndash359

Schubeler D 2015 Function and information content of DNA methyl-ation Nature 517321ndash326

Sherwood CC Stimpson CD Raghanti MA Wildman DE Uddin MGrossman LI Goodman M Redmond JC Bonar CJ Erwin JM HofPR 2006 Evolution of increased glia-neuron ratios in the humanfrontal cortex Proc Natl Acad Sci U S A 10313606ndash13611

Shi L Lin Q Su B 2014 Human-specific hypomethylation of CENPJ a keybrain size regulator Mol Biol Evol 31594ndash604

Shulha HP Cheung I Whittle C Wang J Virgil D Lin CL Guo Y LessardA Akbarian S Weng Z 2012a Epigenetic signatures of autism tri-methylated H3K4 landscapes in prefrontal neurons Arch GenPsychiatry 69314ndash324

Shulha HP Crisci JL Reshetov D Tushir JS Cheung I Bharadwaj R ChouHJ Houston IB Peter CJ Mitchell AC et al 2012b Human-specifichistone methylation signatures at transcription start sites in prefron-tal neurons PLoS Biol 10e1001427

Somel M Liu X Khaitovich P 2013 Human brain evolution tran-scripts metabolites and their regulators Nat Rev Neurosci14112ndash127

Teng L He B Wang J Tan K 2015 4DGenome a comprehensive data-base of chromatin interactions Bioinformatics 312560ndash2564

The 1000 Genomes Project Consortium 2015 A global reference forhuman genetic variation Nature 52668ndash74

The Chimpanzee Sequencing and Analysis Consortium 2005 Initial se-quence of the chimpanzee genome and comparison with the hu-man genome Nature 43769ndash87

Timp W Feinberg AP 2013 Cancer as a dysregulated epigenome allow-ing cellular growth advantage at the expense of the host Nat RevCancer 13497ndash510

Tobi EW Goeman JJ Monajemi R Gu H Putter H Zhang Y Slieker RCStok AP Thijssen PE Muller F et al 2014 DNA methylation signa-tures link prenatal famine exposure to growth and metabolism NatCommun 55592

Uddin M Wildman DE Liu G Xu W Johnson RM Hof PR Kapatos GGrossman LI Goodman M 2004 Sister grouping of chimpanzeesand humans as revealed by genome-wide phylogenetic analysis of



brain gene expression profiles Proc Natl Acad Sci U S A1012957ndash2962

Whalen S Truty RM Pollard KS 2016 Enhancer-promoter interactionsare encoded by complex genomic signatures on looping chromatinNat Genet 48488ndash496

Xi Y Li W 2009 BSMAP whole genome bisulfite sequence MAPpingprogram BMC Bioinformatics 10232

Yu G Wang L-G He Q-Y 2015 ChIPseeker an RBioconductor packagefor ChIP peak annotation comparison and visualizationBioinformatics 312382ndash2383

Zeng J Konopka G Hunt BG Preuss TM Geschwind D Yi SV 2012Divergent whole-genome methylation maps of human and

chimpanzee brains reveal epigenetic basis of human regulatory evo-lution Am J Hum Genet 91455ndash465

Zeng J Nagrajan HK Yi SV 2014 Fundamental diversity of human CpGislands at multiple biological levels Epigenetics 9483ndash491

Zhu J Adli M Zou JY Verstappen G Coyne M Zhang X Durham T MiriM Deshpande V De Jager PL et al 2013 Genome-wide chromatinstate transitions associated with developmental and environmentalcues Cell 152642ndash654

Ziller MJ Gu H Muller F Donaghey J Tsai LTY Kohlbacher O De JagerPL Rosen ED Bennett DA Bernstein BE et al 2013 Charting adynamic DNA methylation landscape of the human genomeNature 500477ndash481



evolution has modified the human brain at almost every levelof organization including changes in long-distance cortico-cortical connectivity histology and local organization (Preuss2011) as well as gene protein and metabolite expression(Brawand et al 2011 Konopka et al 2012 Liu et al 2012)Intriguingly increasing numbers of human brain studies arediscovering the critical importance of epigenetic mechanismsin regulatory functions development and manifestations ofneuropsychiatric diseases (Houston et al 2013 Nikolova et al2014) We thus ask do the human brain specific phenotypictransitions implicate changes at the epigenetic level If sohow did the epigenetic modifications affect evolutionaryinnovations

Recent studies have begun to elucidate epigenetic differ-ences between brains of humans and nonhuman primates(Enard et al 2004 Farcas et al 2009 Shi et al 2014 Shulhaet al 2012b Zeng et al 2012) In particular we have previouslycompared the whole genome DNA methylation maps (re-ferred to as ldquomethylomesrdquo) of human and chimpanzee brains(Zeng et al 2012) However our previous study focused ongene promoters missing putative differentially methylatedregions at un-annotated transcription start sites IndeedDNA methylation at gene bodies and gene deserts harborsstrong regulatory potential (Hon et al 2013 Huh et al 2013Mendizabal and Yi 2016) Zeng et al (2012) also lacked dataon outgroup species limiting the inference on human-specificmethylation changes Importantly to ascertain species-levelepigenetic divergence with high confidence it is critical tovalidate the observed epigenetic profiles in an independentdataset comprising a large number of brain samples whichwere not available in previous studies in this domainAddressing these concerns here we present an unbiased iden-tification and analysis of human brain specific DNA methyl-ation changes Our novel and robust research design allows usto infer lineage-specific epigenetic changes with high confi-dence Moreover using unbiased approaches we illustratethe significance of epigenetic changes occurring in noncodingregions of the human genome Comparative epigenomicanalyses can offer significant insight into understanding evo-lution of human-specific regulatory processes

Results

Whole-Genome Discovery of DifferentiallyMethylated Regions in Human BrainsOur initial discovery data set consisted of whole genome bi-sulfite sequencing (WGBS) data from prefrontal cortices ofthree humans three chimpanzees (Zeng et al 2012) and twoadditional macaque methylomes from same brain region gen-erated in this study (supplementary table S1 SupplementaryMaterial online) We first identified differentially methylatedregions (DMRs) between human and chimpanzee brainWGBS data sets using BSmooth (Hansen et al 2012)BSmooth is specifically designed to utilize local averagingand information from biological replicates to identify DMRsfrom WGBS data (Hansen et al 2012) We applied additionalcriteria including a minimum DNA methylation difference of03 between humans and chimpanzees We used such a

stringent criterion compared to other differential DNA meth-ylation studies so that we can identify conservative candidateDMRs In addition using this criterion relieves concerns re-garding differential cell composition (see below) We addi-tionally imposed each DMR to harbor at least 10 mappedCpGs to avoid effects of outlier CpGs

The initial candidate DMRs following these proceduresincluded 278 regions (supplementary table S2Supplementary Material online for DMR coordinates) Weused WGBS data from two rhesus macaques (Macacamulatta) to estimate the polarity of humanndashchimpanzee dif-ferences (ie human-specific if DMR is present in human butlacking in chimpanzees and macaques) Using this parsimonyapproach we identified 85 human-specific and 102chimpanzee-specific DMRs (refer ldquoMaterial and Methodsrdquo)Interestingly there was a significant excess of hypo-methylated DMRs in the human lineage (Pfrac14 0005 v2 testsupplementary table S3 Supplementary Material online)Nevertheless given the current lack of knowledge on epige-netic evolutionary rates (refer ldquoDiscussionrdquo) these resultsshould be taken with caution

Validation of DMRs via Expanded Sampling and DeepSequencingPatterns of DNA methylation at some genomic positions arevariable among individuals within the same species In addi-tion the average read depth for our discovery dataset basedon whole-genome sequencing was low for some samples(ranged from 2 to 9 supplementary table S1Supplementary Material online) To examine the consistencyof the DMRs in the presence of these potential variabilitieswe analyzed DNA methylation patterns of selected candidateDMRs from expanded samples using deep sequencingSpecifically we performed targeted PCR and sequencing ofbisulfite converted genomic DNA (average sequencingdepthfrac14 365X supplementary table S4 SupplementaryMaterial online) for a subset of DMRs (38 out of the 278)in the prefrontal cortex of a wider set of individuals andspecies 23 humans 2 chimpanzees 1 Hoolock gibbon(H leuconedys) 7 rhesus macaques and 5 crab-eatingmacaques (M fascicularis)

The validation set is biased toward human-specific DMRsbecause of sample availability (ie chimpanzee samples arenot available in a large number) All validation DMRs showedsignificant humanndashchimpanzee differences (Plt 005 one-sided Wilcoxon test) indicating that our discovery strategysuccessfully identified differentially methylated regions be-tween the two species Moreover all but two validationDMRs showed consistent human brain specific DNA meth-ylation in the extended comparison of humanndashchimpanzeendashmacaques The majority of DMRs were human-specific evenwhen we included two additional primate species (Hoolockgibbon and crab-eating macaques) (supplementary table S5Supplementary Material online) In sum human-specificmethylation patterns of the majority of DMRs were consis-tent in a large number of individuals and in a larger panel ofprimates The consistency between the discovery and valida-tion sets is remarkable considering the age and sex differences


















Promoter (lt=1kb)

5 UTR

Exon

Intron

3 UTR

Downstream (lt3kb)

Distal Intergenic

30

13

27

6

14

4

32


































































































































































Promoter (lt=1kb)

5 UTR

Exon

Intron

3 UTR

Downstream (lt3kb)

Distal Intergenic

30

13

27

6

14

4

32


























































































































































Promoter (lt=1kb)

5 UTR

Exon

Intron

3 UTR

Downstream (lt3kb)

Distal Intergenic

30

13

27

6

14

4

32






















































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































Documents

1 Comparative Methylome Analyses Identify Epigenetic Regulatory