Pervasive Effects of Aging on Gene Expression in Wild Wolves

Pauline CharruaudaggerDagger1 Rachel A Johnstondagger1 Daniel R Stahler2 Amanda Lea3 Noah Snyder-Mackler4

Douglas W Smith2 Bridgett M vonHoldt5 Steven W Cole67 Jenny Tung34 and Robert K Wayne1

1Department of Ecology and Evolutionary Biology University of California Los Angeles2Yellowstone Center for Resources National Park Service Yellowstone National Park3Department of Biology Duke University4Department of Evolutionary Anthropology Duke University5Department of Ecology and Evolutionary Biology Princeton University6Department of Medicine University of California Los Angeles7Cousins Center for Psychoneuroimmunology Semel Institute University of California Los AngelesdaggerThese authors contributed equally to this workDaggerPresent address Department of Ecology and Evolution University of Lausanne Switzerland

Corresponding author E-mail rwaynebiologyuclaedu

Associate editor Meredith Yeager

Abstract

Gene expression levels change as an individual ages and responds to environmental conditions With the exception ofhumans such patterns have principally been studied under controlled conditions overlooking the array of develop-mental and environmental influences that organisms encounter under conditions in which natural selection operatesWe used high-throughput RNA sequencing (RNA-Seq) of whole blood to assess the relative impacts of social status agedisease and sex on gene expression levels in a natural population of gray wolves (Canis lupus) Our findings suggest thatage is broadly associated with gene expression levels whereas other examined factors have minimal effects on geneexpression patterns Further our results reveal evolutionarily conserved signatures of senescence such as immunose-nescence and metabolic aging between wolves and humans despite major differences in life history and environmentThe effects of aging on gene expression levels in wolves exhibit conservation with humans but the more rapid expressiondifferences observed in aging wolves is evolutionarily appropriate given the speciesrsquo high level of extrinsic mortality dueto intraspecific aggression Some expression changes that occur with age can facilitate physical age-related changes thatmay enhance fitness in older wolves However the expression of these ancestral patterns of aging in descendant moderndogs living in highly modified domestic environments may be maladaptive and cause disease This work providesevolutionary insight into aging patterns observed in domestic dogs and demonstrates the applicability of studyingnatural populations to investigate the mechanisms of aging

Key words RNA-Seq transcriptome senescence immunosenescence Canis lupus canid

IntroductionIn contrast to genetic changes occurring across generationsalterations of gene expression allow immediate and acutelysensitive responses to changing conditions Gene expressioncan be affected by multiple factors such as age (Goring et al2007 Hong et al 2008 Lopez-Otın et al 2013 Rahman et al2013 Peters et al 2015) social stressors (Weaver et al 2006Cole et al 2007 McGowan et al 2009 Cole et al 2011 2012Tung et al 2012 Murphy et al 2013 Powell et al 2013) andinfectious disease status (Whitney et al 2003 Cobb et al 2005Ramilo et al 2007 Blankley et al 2014 Mejias and Ramilo2014) Aging is associated with a multitude of gene expressionchanges in blood including cellular senescence and dysfunc-tional immune responses (eg immunosenescence andinflammaging) (Boren and Gershwin 2004 Kirkwood 2005Baylis et al 2013) In contrast chronic stress due to adversesocial environments activates immune pathways and can

affect antiviral responses susceptibility to infectious agentswound healing abilities and individual fitness (Sapolsky 2005Weaver et al 2006 Miller et al 2009 Chen et al 2011 Coleet al 2011 2012 Tung et al 2012 Murphy et al 2013 Powellet al 2013)

Investigations of gene expression patterns in natural pop-ulations of nonmodel species can reveal which environmentalfactors are most influential on the regulation of specific mo-lecular pathways However with the exception of baboons(Runcie et al 2013 Tung et al 2015) the transcriptome-wideeffects of intrinsic and environmental variables on gene ex-pression in vertebrates have only been studied in humans andspecies in captivity Knowledge regarding the generality ofaging effects on gene expression patterns across species isfurther limited by short lifespan biases of model species andunintended artificial selection on aging rates (Ricklefs 2010)Thus despite a long history of research on the diverse ecology

Article

The Author 2016 Published by Oxford University Press on behalf of the Society for Molecular Biology and Evolution All rights reserved For permissions pleasee-mail journalspermissionsoupcom

Mol Biol Evol doi101093molbevmsw072 Advance Access publication April 15 2016 1

MBE Advance Access published May 13 2016 at D

uke University L

aw School on June 26 2016

httpmbeoxfordjournalsorg

Dow

nloaded from

social structures and life histories of wild animal populations(eg Wilson 2000) the impacts of aging and environmentalfactors on interindividual variation at a functional genomiclevel in natural animal populations remain poorly understood(Ricklefs 2010)

To address this knowledge gap we used RNA sequencing(RNA-Seq) to quantify genome-wide expression levels in wildgray wolves (Canis lupus) from Yellowstone National Park(YNP) USA Gray wolves serve as an excellent model forstudying the influences of age and social factors on geneexpression levels as wolves live in highly cooperative societiescharacterized by social hierarchies (Mech 1999 Packard 2003)Further the divergence time of approximately 80 My be-tween wolves and more extensively studied primate and mu-rine models (dos Reis et al 2012) affords the opportunity toassess the long-term evolutionary conservation of expressionpatterns We take advantage of the exceptional observationaldatabase available for the YNP wolf population in whichfactors potentially affecting gene expression such as socialstatus age and disease presence have been extensively doc-umented since the wolf reintroduction in 1995ndash1996(vonHoldt et al 2008 2010 Almberg et al 2009 MacNultySmith Vucetich et al 2009 Smith et al 2010 Almberg et al2012 Smith et al 2012 Stahler et al 2013 Cubaynes et al2014)

Wolf packs consist of an adult pair of breeding wolves(denoted alpha) their offspring (denoted subordinates)and additional adult wolves which may breed but are con-sidered subordinate to alphas (denoted beta) (Mech 19992003 Packard 2003) The breeding pair is considered domi-nant to other individuals and social hierarchy is most com-monly demonstrated by nonaggressive behavior andsubmissive postures resulting in unchanged or slightly lowerglucocorticoid levels in subordinates (Mech 1999 Packard2003 Smith et al 2012 Molnar et al 2015) This social systemis in contrast to hierarchies in which dominance is maintainedby displacements and threats toward lower ranking individ-uals such as in wild baboons and macaques which result inhigher levels of stress in low-ranking individuals (Sapolsky2004 2005) Given the widespread impacts that social stresscan have on individual health and the immune system asdetected in blood (Sapolsky 2005 Weaver et al 2006 Milleret al 2009 Chen et al 2011 Cole et al 2011 2012 Tung et al2012 Murphy et al 2013 Powell et al 2013) we aimed tocompare influences of social rank on gene expression levels inblood of wolves which could be collected with minimumhandling to those observed in primate social systems

Active disease infection also impacts gene expression var-iation (Whitney et al 2003 Cobb et al 2005 Ramilo et al2007 Blankley et al 2014 Mejias and Ramilo 2014) Sarcopticmange is an infectious disease frequently observed in YNPgray wolves induced by the mite Sarcoptes scabiei and anal-ogous to human scabies (Almberg et al 2012 2015) Sarcoptesmites are known to elicit gene expression changes in humansand trigger allergic reactions causing skin inflammation itch-ing and hair loss in wolves (Arlian et al 2004 2006 Almberget al 2012) Therefore we hypothesized that presence ofmange in YNP wolves results in expression signatures of

immunosuppression and inflammatory response (Arlianet al 2004 2006 Velavan and Ojurongbe 2011 Gregoriet al 2012)

Finally aging is also expected to be a factor affecting geneexpression in gray wolves We predict that rapid aging ob-served at the physical level in wolves (MacNulty SmithVucetich et al 2009 Stahler et al 2013 Cubaynes et al2014) would be mirrored by gene expression changes withage Although the molecular impact of age has been welldescribed in humans (Goring et al 2007 Hong et al 2008Lopez-Otın et al 2013 Rahman et al 2013 Peters et al 2015)the relative influence and specific pathways altered by age in awild population confronted with multiple environmentalchallenges is unknown In baboons the only wild species as-sessed to date the impact of age on gene expression wasfound to be modest and the molecular pathways alteredby age were not evaluated (Runcie et al 2013 Tung et al2015) Therefore to assess conservation of patterns across themammalian phylogeny and because impacts of aging ongene expression of specific pathways are virtually unknownfor wild populations we compared our age results to humans(Goring et al 2007 Hong et al 2008)

To investigate natural variation in gene expression levelswe use a linear mixed model to simultaneously assess theeffects of social rank (alpha vs all subordinates) age diseasestatus (infected with mange vs not infected) and sex ongenome-wide expression in whole blood while controllingfor genetic relatedness As these factors can also influencecell-type abundance in blood (Linton and Dorshkind 2004Cole et al 2011 2012 Baylis et al 2013 Powell et al 2013) weperform transcript origin analysis (TOA) (Cole et al 2011) andtranscriptome representation analysis (TRA) (Powell et al2013) to infer the cellular origins of differentially expressedtranscripts Among the variables studied we find that geneexpression variation among wild wolves is primarily explainedby age with multiple age-related genes and processes con-served between wolves and humans In contrast we observe asurprising absence of expression differences with other vari-ables such as rank indicating that impacts of rank in primatemodels cannot be extended to the effects of social hierarchyin other vertebrate species

Results

Sequence DataTo assess transcriptome variation across gray wolves in theirnatural environment we quantified gene expression levelsfrom whole blood samples of 27 wolves using RNA-Seq Atotal of 126 billion 100 base-pair reads of sufficient quality(Phred scoregt 20) and length (gt 25 bp after trimming) weregenerated (mean of 4669 million reads per individual supplementary table S1 Supplementary Material online) On aver-age 875 of these reads aligned to the dog genome(CanFam3174) and up to 35 of the uniquely mapped readsaligned to annotated protein-coding regions (supplementarytable S1 Supplementary Material online) Of the 19856 pro-tein-coding genes annotated in the dog genome we obtained13558 genes (683) with sufficient read depth across

Charruau et al doi101093molbevmsw072 MBE

2

at Duke U

niversity Law

School on June 26 2016httpm

beoxfordjournalsorgD

ownloaded from

individuals (10 reads in at least 75 of cDNA libraries) to beincluded in expression analyses The number of detectedgenes in whole blood of wolves (Nfrac14 13558) was similar tothat found in lymphocytes of humans (Nfrac14 12758) (Goringet al 2007) and whole blood of baboons (Nfrac14 10409) (Tunget al 2015)

Association of Gene Expression Levels with Intrinsic andEnvironmental FactorsWe assessed the effects of intrinsic and environmental factorson transcriptome-wide gene expression using a significancecut-off of false discovery rate (FDR)frac14 02 given the limitedsample size Similar to other studies that evaluated effects ofsex on blood gene expression in mammals (eg Whitney et al2003 Tung et al 2015) we observed minimal expression dif-ferences between male and female wolves with only one genesignificantly associated with sex ENSCAFG00000030886 (folddifference [FD]frac14 008 Q valuefrac14 01 fig 1A and supplementary table S2 Supplementary Material online) This protein-coding gene belongs to the thymosin b4 family but has notbeen formally named Further despite the presence of phys-ical symptoms characteristic of mange in seven of the sam-pled wolves we did not identify any genes differentiallyexpressed in wolves affected by mange (fig 1B) Finally wepredicted that social rank would be broadly associated withgene expression levels comparable to effects of social condi-tions in humans and nonhuman primates (Cole et al 20072012 Tung et al 2012) However we did not identify anygenes in which expression levels were significantly associatedwith social rank (fig 1C)

In contrast controlling for rank sex and mange infectionstatus we detected 625 genes associated with age comprising214 genes up-regulated and 411 genes down-regulated withage with diverse functions (fig 1D and table 1 and supplementary table S2 Supplementary Material online) These age-re-lated gene expression changes may have been induced byintracellular changes in gene transcription or by age-relatedchanges in proportions of blood cell-types (Linton andDorshkind 2004 Baylis et al 2013) To evaluate potentialchanges in blood cell composition with age we performedTOA (table 2) which infers the cellular origin of transcripts(Cole et al 2011) in parallel with TRA (table 3) which assesseschanges in cell abundance based on genes previously identi-fied to have highly cell-type-specific expression (Powell et al2013) In accordance with observations of immunosenes-cence in humans genes down-regulated with age were largelyassociated with B cells (Plt 0001 table 2) which appear topartly be due to a lower prevalence of B cells in older wolves(TRA FD per yearfrac14 098 Pfrac14 0001 table 3) Further consis-tent with development of more pro-inflammatory pheno-types in older individuals (reviewed in Boren and Gershwin2004 Baylis et al 2013) genes up-regulated with age weresignificantly associated with innate immunity cells includingdendritic cells (Plt 0001 table 2) and natural killer cells(Pfrac14 0004 table 2) TRA indicated a slightly increased pro-portion of monocytes in older individuals (FD per yearfrac14 101Plt 0001 table 3)

To assess the conservation of aging processes across spe-cies we compared the 625 age-related genes in wolf leuko-cytes to a catalog of age-related genes in human lymphocytes(Goring et al 2007 Hong et al 2008) Wolf age-related geneswere significantly enriched with human age-related genes aswe found that of the 420 age-related wolf genes representedin the human study 255 (607) genes were associated withage in humans (hypergeometric test Pfrac14 0046) The majorityof these overlapping genes were associated with age in thesame direction (608 of overlapping genes Nfrac14 155 concor-dant genes v2frac1411863 dffrac14 1 Pfrac14 0001)

A diversity of genes and biological processes we found tobe associated with age in wolves (table 1) has previously beenassociated with aging in humans (Boren and Gershwin 2004Kirkwood 2005 Baylis et al 2013) For example we observedsignificant age-related down-regulation of ETS1 (FD peryearfrac14 0900 Q valuelt 0001 fig 1D and 2 and table 1 andsupplementary table S2 Supplementary Material online) atranscription factor involved in a variety of functions includ-ing regulation of the immune response (Garrett-Sinha 2013)and cellular senescence (Ohtani et al 2001) Similarly CDK4which is known to be inhibited by p16 a key protein activatedduring cellular senescence (Krishnamurthy et al 2004 Liuet al 2009 Lopez-Otın et al 2013) was down-regulatedwith age (FDfrac14 0909 Q valuefrac14 0137 fig 2 and table 1and supplementary table S2 Supplementary Material online)Additionally we observed significant age-related expres-sion differences in genes associated with telomere mainte-nance (eg DKC1 a component of the telomeraseholoenzyme FDfrac14 0948 Q valuefrac14 02 and NOP56 FDfrac14 0958 Q valuefrac14 0198 table 1 and supplementary table S2Supplementary Material online) (Poncet et al 2008) DNArepair and genomic stability (eg SIRT6 FDfrac14 0896 Q val-uefrac14 02) (Mostoslavsky et al 2006 Di Mauro and David2009) and epigenetic regulation (genes down-regulatedwith age were enriched for the gene ontology (GO) termsldquochromatin organizationrdquo Pfrac14 0005 and ldquohistone modifica-tionrdquo Pfrac14 0011 table 1 and supplementary table S3Supplementary Material online)

Paralleling intracellular senescence aging has been foundto dysregulate the propensity of the adaptive immune systemto respond to novel infectious disease (a phenomenonknown as ldquoimmunosenescencerdquo) (Baylis et al 2013 Lopez-Otın et al 2013) In agreement with previous work our resultssuggest reduced activity of B and T cells with age as interleu-kin and interleukin-related genes required for differentiationand proliferation of lymphocytes were down-regulated inolder wolves (IL7 IL21R IL27RA Q valueslt 02 fig 2 andtable 1 and supplementary table S2 SupplementaryMaterial online) In addition we observed age-relateddown-regulation of multiple MHC class II genes (DogLeukocyte Antigen genes DLA-DQA1 DLA-DRA DLA-88and DLA-DOB Q valueslt 02 fig 2 and table 1 and supplementary table S2 Supplementary Material online) whose ex-pression is likely limited to antigen-presenting cells (B cellsmonocytes) and activated T cells GO analyses further sup-port immunosenescence as 30 of terms enriched in thegenes down-regulated with age are explicitly related to

Pervasive Effects of Aging doi101093molbevmsw072 MBE

3

at Duke U

niversity Law

School on June 26 2016httpm

beoxfordjournalsorgD

ownloaded from

immune system processes (eg lymphocyte activation anti-gen receptor-mediated signaling pathway T- and B-cell acti-vation and proliferation and interleukin-2 productionsupplementary table S3 Supplementary Material online)

In contrast to the abundance of immunity-related GOterms enriched in the genes down-regulated with age inwolves GO terms enriched in genes up-regulated with agewere dominated by biological processes related to lipid

Rank

log2 (fold difference)

minuslo

g 10 (

Qminus

valu

e)

0

1

2

3

4

5

0- 2- 4 2 4

Mange

log2 (fold difference)

minuslo

g 10 (

Qminus

valu

e)

0

1

2

3

4

5

0- 2- 4 2 4

Age

log2 (fold difference)

minuslo

g 10 (

Qminus

valu

e)

ETS1

0

1

2

3

4

gt10000ENSCAFG00000024259

0-02-04 02 04

SexA B

C D

ENSCAFG00000030886

log2 (fold difference)

minuslo

g 10 (

Qminus

valu

e)

0

1

2

3

4

5

0- 2- 4 2 4

FIG 1 Difference in gene expression according to sex mange social rank and age in wolf blood (AndashD) Black dots represent genes for whichexpression is significantly associated with a variable whereas gray dots illustrate nonsignificant genes The significance threshold (Q valuefrac14 02) isdelimited by the horizontal line yfrac14log10(02) FD is expression of females relative to males (A) mange-affected wolves relative to nonaffectedindividuals (B) alpha wolves relative to subordinates (C) and difference per year of age (D) for 13558 genes expressed in wolf whole blood (D)Vertical lines depict the thresholds used in the TOA (table 3)

Table 1 Biological Processes and Genes Associated with Age in Gray Wolves

Enriched GO Terma GO P Value Gene Symbol Ensembl Gene ID Annual FD Q Valueb

Regulation of metabolic process 225E-5 ETS1 ENSCAFG00000010304 0900 lt0001CDK4 ENSCAFG00000000280 0910 0137

RNA metabolic process 191E-5 DKC1 ENSCAFG00000019620 0948 0175NOP56 ENSCAFG00000006651 0958 0198

Chromatin modification 297E-3 SIRT6 ENSCAFG00000019123 0896 0200HIRA ENSCAFG00000014619 0974 0137

Immune response 628E-6 DLA-DQA1 ENSCAFG00000000812 0945 0137DLA-DRA ENSCAFG00000000803 0950 0146DLA-88 ENSCAFG00000000487 0945 0180DLA-DOB ENSCAFG00000000819 0883 0175

Lymphocyte activation 155E-9 IL7 ENSCAFG00000008373 0907 0180IL27RA ENSCAFG00000016504 0919 0146CD5 ENSCAFG00000016332 0938 0157CXCR5 ENSCAFG00000012439 0898 0175

Lipid metabolic process 117E-2 LEP ENSCAFG00000001672 1078 0151LIAS ENSCAFG00000015991 1035 0183LIPH ENSCAFG00000013246 1113 0184

aA comprehensive gene list and full GO results are presented in supplementary tables S2ndashS4 Supplementary Material onlinebQ values of genes significantly associated with age after controlling for relatedness social rank mange presence and sex using linear mixed effects models

Charruau et al doi101093molbevmsw072 MBE

4

at Duke U

niversity Law

School on June 26 2016httpm

beoxfordjournalsorgD

ownloaded from

metabolism (supplementary table S4 SupplementaryMaterial online) and included leptin (LEP) lipoic acid synthe-tase and lipase member H (LIPH Q valueslt 02 fig 2 andtable 1 and supplementary table S2 Supplementary Materialonline) We also observed an age-related increase in expres-sion of insulin-like growth factor 1 receptor (IGF1R FDfrac14 1038Q valuefrac14 0188 fig 2 and supplementary table S2Supplementary Material online) which has been found toregulate lifespan (Holzenberger et al 2003)

DiscussionWe examined factors that we predicted to influence geneexpression in gray wolves based on gene expression patternsobserved in humans and other primates which included sexmange presence social rank and age The negligible effect ofsex on gene expression in blood is consistent with similarfindings in humans and macaques (Whitney et al 2003Tung et al 2015) However the lack of detectable differentiallyexpressed genes in mange-infected wolves was unexpectedgiven the presence of visible symptoms of mange such ashairless patches with skin lesions (Almberg et al 2015) Thisresult may partly be due to limited statistical power but mayalso be caused by the immunosuppressive properties that Sscabiei evoke in their hosts by impacting IL10 signaling (Arlianet al 2006) The possibility of altered IL10 signaling in wolveswith mange is supported by the presence of IL10RA in the topfive genes associated with mange in our data set althoughthis result is not significant at our Q-value threshold of 02

The absence of a significant effect of rank on gene expres-sion was also unexpected but likely reflects a less hierarchicallybased social system in wolves compared to many primatesspecies In wolves the term alpha denotes the breeding pair ofa pack (which are typically the parents of other pack mem-bers) and signifies little to no differences in stress levels re-source access or received aggression from other wolves(Mech 1999 Cubaynes et al 2014 Molnar et al 2015)Additionally wolf rank and age are positively correlated(Pearson correlation r frac14 0357 P frac14 0068) and may haveeffects on the same biological pathways (Snyder-Mackler et al2014) potentially limiting our statistical power to detect rankeffects independent of aging

Overall we found that age has a much broader impact ongene regulation than sex mange infection status or rank in asocial top-predator exposed to naturally occurring abioticand biotic challenges Wolves show signatures of senescenceconsistent with aging in model organisms and humans (Bayliset al 2013 Beirne et al 2014) which supports the hypothesisthat aging effects on cellular processes are evolutionarily con-served through mammal phylogeny (Palacios et al 2011Nussey et al 2013) Theory predicts that higher levels of ex-trinsic mortality should lead to an increased rate of senes-cence because a relatively larger amount of energy is expectedto be dedicated to reproduction than to somatic mainte-nance and repair (Kirkwood 1977 2005) The short time pe-riod over which age-related expression differences manifest inwolves (spanning a few years compared to decades in hu-mans) (Goring et al 2007) is evolutionarily appropriate giventhe high extrinsic mortality in this species due mainly tointerpack strife (Cubaynes et al 2014) and rapid physicalsenescence (MacNulty Smith Mech et al 2009 MacNultySmith Vucetich et al 2009 Stahler et al 2013)

Stress related to changes in rank physical injuries anddisease (Almberg et al 2015) may augment the expressionsignature of aging in wolves or result in expression differencesunique from that observed in humans Genes associated withage in wolves were significantly enriched for human age-re-lated genes yet we detected 165 age-related genes in wolvesthat were not found to change with age in humans (Goringet al 2007) Further 100 genes associated with age in bothwolves and humans showed opposite trends of expressionlevels with age between the two species (Goring et al 2007) Alongitudinal study and comparison to gene expression incaptive wolves would be valuable in decomposing the spe-cific factors that have long-lasting impacts on gene expressionin wolves

We identified age-related changes in the expression ofgenes that may be critical in the progression of the agingprocess such as up-regulation of IGF1R the primary receptorof the IGF1 protein product IGF1 signaling is thought to be anevolutionarily conserved regulator of the trade-off betweengrowth and survival as the IGF1 pathway promotes growthduring early development but can increase the rate of aginglater in life (Rollo 2002 Salminen and Kaarniranta 2010) Theexistence of a trade-off in IGF1 signaling is supported by stud-ies across domestic dog breeds in which plasma IGF1 proteinlevels across breeds positively correlate with body size but

Table 2 Cell Types Inferred to Mediate Age-Related Gene ExpressionDifferences

Cell Type P Value

FDgt 110a FD lt 090b

CD14thorn monocytes 0035 0998BDCA4thorn dendritic cells lt0001 0973CD56thorn NK cells 0004 0338CD4thorn T cells 0762 0424CD8thorn T cells 0460 0768CD19thorn B cells 0126 lt0001

NOTEmdashNK natural killer FD fold difference per year of ageaTOA results of up-regulated genes (N frac14 137 genes)bTOA results of down-regulated genes (N frac14 111 genes)Significant (P lt 001) overrepresentation of genes predominantly expressed by aspecific cell type

Table 3 Leukocyte Prevalence Differences with Age in Wolf Blood

Cell Types Genes Z gt 6a FD P Valueb

CD14thorn monocytes 337 101 lt0001BDCA4thorn dendritic cells 250 100 0912CD56thorn NK cells 349 100 0043CD4thorn T cells 62 100 0606CD8thorn T cells 45 101 0220CD19thorn B cells 87 098 0001

NOTEmdashFD fold difference in cell prevalence per year of age NK natural killeraGenes diagnostic of a cell type are defined by average expression in a cell typegt 6standard deviations above the mean (Genes Z gt 6) (Powell et al 2013)bTRA results of cell-type differences with ageSignificant (P lt 001) difference in cell-type prevalence

Pervasive Effects of Aging doi101093molbevmsw072 MBE

5

at Duke U

niversity Law

School on June 26 2016httpm

beoxfordjournalsorgD

ownloaded from

negatively correlate with lifespan (Sutter et al 2007 Hoopeset al 2012) This mechanism has been suggested to be thereason that small dog breeds can have twice the lifespan oflarger breeds (Kraus et al 2013) The rapid development ofyearling wolves (MacNulty Smith Mech et al 2009MacNulty Smith Vucetich et al 2009) likely requires earlyIGF1R expression However the function of up-regulatedIGF1R we observe in old wolves is unknown and is potentiallynonadaptive IGF1R is down-regulated with age in humans(Goring et al 2007) which may be an important contributorto human longevity Consequently up-regulation of IGF1Rmay be a leading driver of the rapid senescence of wolvesand other short-lived species

We also observed increased expression of LEP in olderwolves concordant with increased leptin protein levels

observed in aging humans dogs and model species (Mobbs1998 Ishioka et al 2002 2007 Ma et al 2002) Leptin is pro-duced by adipocytes and reduces reflex appetite (Sanchez-Rodrıguez et al 2000) Its increase with age which is oftendisproportionately more than would follow from an increasein adipose tissue is thought to indicate a leptin-resistant statein elderly individuals and can contribute to obesity (Ahrenet al 1997 Li et al 1997 Sanchez-Rodrıguez et al 2000 Maet al 2002) Because the weight of wolves is stronglycorrelated with age (Pearson correlation r frac14 0768 P frac142967 105 supplementary table S1 SupplementaryMaterial online) we do not have the statistical power totest whether LEP expression increases with age more thanexpected by weight gain alone However the overall weightgain and increased expression of LEP with age in wolves

ETS1 CDK4

IL7 DLAminusDQA1

2 4 6 8

IGF1R

1 3 5 7 9

LEP

2 4 6 81 3 5 7 9

- 2- 1

01

Rel

ativ

e lo

g 2 (ex

pres

sion

)- 2

- 10

1 R = 0672

R = 0212

R = 0222

R = 0342

R = 0262

R = 0292

- 2- 1

01

- 2- 1

01

Rel

ativ

e lo

g 2 (ex

pres

sion

)R

elat

ive

log 2 (

expr

essi

on)

Age (year) Age (year)

FIG 2 Expression of a subset of genes significantly associated with age in gray wolves Relative expression represents the normalized log 2-transformed expression levels of wolves after controlling for relatedness using linear mixed effects models A comprehensive gene list is available insupplementary table S2 Supplementary Material online

Charruau et al doi101093molbevmsw072 MBE

6

at Duke U

niversity Law

School on June 26 2016httpm

beoxfordjournalsorgD

ownloaded from

supports the hypothesis that wild wolves undergo metabolicaging (MacNulty Smith Mech et al 2009)

Although we find age-related gene expression patternsshared by humans and wolves their impacts on health andfitness may be vastly different for each species For examplemetabolic aging and obesity are recognized as having negativehealth impacts in older humans (eg Alexander et al 2003) Incontrast increased weight in wolves enhances reproductivesuccess in females of all ages (Stahler et al 2013) and dispatch-ing ungulate prey in males and females (MacNulty SmithMech et al 2009) Further male wolves exhibit increasing ag-gressiveness with age which may facilitate dominance andbreeding opportunities within a pack and enhance successduring interpack territorial encounters (Cassidy KA MechLD MacNulty DR Stahler DR Smith DW personal communi-cation) Thus although senescence does occur in wolves someage-related changes may also allow older wolves to maintainfitness and consequently can be maintained by selection (egSchwarz et al 2016) In dogs which live in human-dominatedenvironments such age-related changes may be maladaptiveand result in morbidity Likewise gene expression patterns thatevolved under a hunter-gatherer lifestyle in humans may leadto health problems for individuals with a western lifestyle anddiet (Carrera-Bastos et al 2011 Farooqui 2015) Consequentlyfor both modern humans and their domestic companionssuch legacy effects may result in disease in aging individuals

Our results establish the ancestral baseline for gene expres-sion in dogs and show that transcriptome-wide gene expres-sion analysis applied to natural populations has the potentialto reveal novel functional and evolutionary insights into themechanisms and drivers of aging In general our study sug-gests that aging is the critical factor affecting gene expressionlevels in a wild wolf population and provides a new paradigmfor investigating the aging process in natural systems

Materials and Methods

Sampling and Phenotypic TraitsWhole blood was collected from 27 wild gray wolves in YNPUSA Sampling was conducted during winters 2011 2012 and2013 during annual wolf captures according to US nationalguidelines for handling animals and with all required permitsheld by the National Park Service (Smith et al 2012)Individual animal information is provided in supplementarytable S1 Supplementary Material online

Age and SexFor each of the 27 wolves sex was determined during handlingMales were coded as 0 and females as 1 Ages were estimatedfrom an assumed birth date of April 15 (mean whelp date instudy area) (Stahler et al 2013) and were continuously codedfor the linear models (supplementary table S1 SupplementaryMaterial online and supplementary fig S2 SupplementaryMaterial online) Age was determined by either capturingpups (Nfrac14 13) which are easy to distinguish due to bodyand tooth size or recapturing individuals that had known birthyears The known ages (years) of the 21 wolves included in theanalyses were 08 (Nfrac14 13) 28 (Nfrac14 1) 37 (Nfrac14 1) 38 (Nfrac14 3)

48 (Nfrac14 1) 58 (Nfrac14 1) 68 (Nfrac14 1) and 88 (Nfrac14 1) For the fiveindividuals included in the expression analyses that were notfirst identified as pups tooth wear was used to estimate the ageof adults (Gipson et al 2000) The estimated ages (years) of thefive wolves used in the analyses were 28 (Nfrac14 1) 4 (Nfrac14 1) 48(Nfrac14 1) 6 (Nfrac14 1) and 9 (Nfrac14 1) Estimated ages of the twoolder wolves (6 and 9 years) were corroborated by the fact thatthey had been monitored by NPS for multiple consecutiveyears (5 and 6 years of monitoring respectively)

Social StatusSocial rankings of the 27 individuals were determined frombehavioral observations and reproductive status and werecategorized specific to the time of sampling as alpha ornonalpha (supplementary table S1 Supplementary Materialonline) Individuals that were the only or primary breeders ina pack or behaviorally dominant to all pack members (ex-cluding their mates) were identified as alpha wolves In well-known and larger packs individuals that were subordinate tothe leading pair but dominant to other pack members wereranked as beta Individuals showing neither reproductive norantagonistic behavior were classified in the subordinate cat-egory (Mech 1999 Packard 2003 Smith et al 2012) Becauseage and social rank are highly correlated in YNP wolves(Pearson correlation rfrac14 0614 Pfrac14 655 105) we focusedon the distinction between alpha wolves (scored as 1) andnonalpha animals (scored as 0 betas and subordinates)which significantly reduced the magnitude of this correlation(Nfrac14 27 Plt 001 psych R package pairedr function to testfor a significant decrease in correlation) (Revelle 2015)

Mange Infection StatusThe severity of mange infection in each wolf was assessed basedon the presence of physical symptoms (hairlessness andscratching lesions) at the time of capture and scored accordingto the percentage of body surface affected (0 no physical symp-toms 1lt 5 of the body affected by lesions 2 6ndash50 and3gt 50) (Pence et al 1983 Almberg et al 2012) As only oneindividual presented severe symptoms of mange infection weclassified mange as a binary variable (0 no symptom 1 visiblesymptoms supplementary table S1 Supplementary Materialonline) Continuous coding of mange severity led to qualita-tively identical null findings (results not shown)

RNA Extraction Library Preparation and SequencingWhole blood was preserved in PAXgene Blood RNA tubes(PreAnalytiX Qiagen Hombrechtikon Switzerland) and storedat80 C Total RNA extraction was performed following themanufacturerrsquos instructions of PAXgene Blood RNA kit(PreAnalytiX Qiagen Hombrechtikon Switzerland) RNA qual-ity was assessed with an Agilent 2100 Bioanalyzer (AgilentTechnologies Santa Clara CA) All samples had a RNA integritynumber (RIN)gt 7 Total RNA was treated with the Globin-Zero kit (Epicentre Illumina Madison WI) and purified with amodified Qiagen RNeasy MinElute (Qiagen Inc Valencia CA)cleanup procedure or ethanol precipitation (supplementarytable S1 Supplementary Material online) cDNA librarieswere synthesized using a strand-specific kit from Epicentre

Pervasive Effects of Aging doi101093molbevmsw072 MBE

7

at Duke U

niversity Law

School on June 26 2016httpm

beoxfordjournalsorgD

ownloaded from

(ScriptSeq v2 Library Prep kit Illumina Madison WI) withScriptSeq Index PCR primers (Epicentre Illumina MadisonWI) cDNA libraries were quantified with the KAPA SYBRFast qPCR library quantification kit (Kapa Biosystems IncWilmington MA) and pooled at five to six samples per laneSingle-end 100 bp sequencing was performed on IlluminarsquosHiSeq2000 platform (supplementary table S1 SupplementaryMaterial online) at the Broad Stem Cell Research Centerrsquos bio-sequencing core UC Los Angeles CA (BSCRC) and the VincentJ Coates Genomics Sequencing Laboratory (GSL) UC BerkeleyCA The data have been deposited in NCBIrsquos Gene ExpressionOmnibus (Edgar et al 2002) and are accessible through GEOseries accession number GSE80440 (httpwwwncbinlmnihgovgeoqueryacccgiaccfrac14GSE80440)

Mapping and Expression QuantificationQuality filtering and adaptor trimming was performed usingTRIM GALORE v031 (wwwbioinformaticsbabrahamacukprojectstrim_galore) A moderate coverage (25x) genome as-sembly of the gray wolf is available (Freedman et al 2014)However this assembly is based on mapping reads to thedomestic dog Therefore given the availability of the well-annotated dog genome (Canis l familiaris EnsemblCanFam3174) (Lindblad-Toh et al 2005 Hoeppner et al2014) as well as the short divergence time between domesticdogs and wolves (lt 29 Ka) (Thalmann et al 2013 Freedmanet al 2014 Skoglund et al 2015 Fan et al 2016) we used thedomestic dog genome as a reference Splice-aware mappingto the domestic dog genome (Ensembl CanFam3174 un-masked) was implemented with TopHat2 v2010(Engstrom et al 2013 Kim et al 2013) We allowed up tothree mismatches a total of 3-bp length of gaps and an editdistance no more than 3 bp between the reads and the ref-erence sequence to account for the divergence between thedomestic dog and the gray wolf Only reads that uniquelymapped to the reference were kept in subsequent analysesOnly uniquely mapped reads were kept in subsequent anal-yses The expression level of each gene was quantified usingthe union mode of the python script htseq-count (HTSeq 061) (Anders et al 2015) Individual expression profiles werefiltered to only include protein-coding genes with at leastten reads in at least 75 of cDNA libraries (Nfrac14 13558 pro-tein-coding genes) Genome assembly and GTF files of thedomestic dog genome (CanFam3174) used as reference inthis study can be retrieved at httpdec2013archiveensemblorgCanis_familiarisInfoIndex

Data PreprocessingTo identify outlier samples we used the adjacency function inthe WGCNA R package (Langfelder and Horvath 2008) tobuild a Euclidean distance-based sample network from thelog 2-transformed read counts Samples were designated asoutliers if their standardized connectivity deviated by morethan three standard deviations from the mean (Horvath2011) After removing two individual outliers from the data(supplementary table S1 Supplementary Material online) weused conditional quantile normalization (cqn package in R)to normalize for sequencing depth (using trimmed mean of

M values) (Robinson et al 2010) GC-content and genelength (Hansen et al 2012) averaged from the transcriptsin the canFam3174 gtf file We used principal componentanalysis of the normalized log 2-transformed expression datato assess general impacts of technical phenotypic or envi-ronmental variables on gene expression variance We identi-fied significant effects of technical factors on overall geneexpression (supplementary fig S1A Supplementary Materialonline) including effects of sequencing core (correlatedwith PC1 [Pearson correlation rfrac14 059 Pfrac14 1892 103]and PC2 [Pearson correlation rfrac140519 Pfrac14 7809 103] of the normalized gene expression data) library prep-aration procedure (column purifications vs ethanol precipi-tation correlated with PC1 [Pearson correlation rfrac14 0791Pfrac14 2557 106]) barcode bleed (Kircher et al 2012) ofRNA-Seq from other samples sequenced in the same lane(Johnston R unpublished data March 2015 correlated withPC2 [Pearson correlation rfrac140566 Pfrac14 3178 103] andPC6 [Pearson correlation rfrac140486 Pfrac14 1387 102])and year of sampling (correlated with PC1 [Pearson correla-tion rfrac14 0623 Pfrac14 886 104]) To control for these effectswe regressed out sequencing core library preparation proce-dure and sequencing lane composition for each gene prior tothe main analyses (supplementary fig S1B SupplementaryMaterial online and supplementary table S1 SupplementaryMaterial online) Year of sampling which significantly corre-lated with library preparation procedure (Pearson correlationrfrac14 0721 Pfrac14 4853 105) was not significantly associatedwith PC1-12 after regressing out the effect of library prepara-tion (supplementary fig S1B Supplementary Material online)For the subset of samples in which time of day of collectionwas available (supplementary table S1 SupplementaryMaterial online) we did not detect a significant effect oftime of day on overall gene expression (supplementary figS1 Supplementary Material online)

Linear Mixed Effects ModelsFor each gene we modeled the fixed effects of age sex socialrank and mange infection status on the residuals of the nor-malized log 2-transformed gene expression levels of the 25samples using linear mixed models in gemma v094 (Zhouand Stephens 2012) as done previously (Tung et al 2012) Tofit a random effect controlling for kinship we calculated pair-wise relatedness between individuals using the triadic maxi-mum likelihood approach in COANCESTRY (Wang 2011) basedon genotypes obtained from 24 microsatellite loci (supplementary table S5 Supplementary Material online) Allele fre-quencies for the 24 markers were estimated using a compre-hensive data set of 371 YNP gray wolves (vonHoldt et al 20082010 vonHoldt BM unpublished data)

All significance levels for the linear models were ad-justed for multiple hypothesis testing using the Q valuemethod in R (Storey and Tibshirani 2003) with the nulldistribution constructed based on 100 random permuta-tions Because of the difficulty of obtaining samples fromanimals in the wild for RNA analyses our sample sizelimited the statistical power of the linear model analysisTherefore we set a relatively liberal FDR threshold of 02

Charruau et al doi101093molbevmsw072 MBE

8

at Duke U

niversity Law

School on June 26 2016httpm

beoxfordjournalsorgD

ownloaded from

to identify candidate genes associated with rank age sexand mange The FDR threshold of 02 corresponded to aminimum detectably significant FD of approximately22 for age and approximately 78 for sex (supplementary table S2 Supplementary Material online) Q values ofall genes meeting the 02 FDR threshold are presented insupplementary table S2 Supplementary Material online

GO AnalysisTo assess which biological functions were significantly en-riched in the genes affected by our variables of interest (de-fined as genes associated at Q valuelt 02) we used GPROFILER(Reimand et al 2007) based on the Canis familiaris gene an-notation (Ensembl 79) Queries were composed of genes up-or down-regulated with each variable of interest and thebackground set included all genes tested in our analysis(Nfrac14 13558) We set the minimal overlap between the GOterm and the query as well as the minimal number of geneswithin a functional category to 3 All GO analyses were cor-rected for multiple testing using the BenjaminindashHochbergFDR method (Benjamini and Hochberg 1995)

Differential White Blood Cell CountsTo assess the impacts of age rank mange and sex on therelative abundance of lymphocytes neutrophils monocytesbasophils and eosinophils blood smears were prepared foreleven of the 27 individuals during the collection of PAXgene-preserved blood (supplementary table S6 SupplementaryMaterial online) Subsequently the histological slides werefixed and stained according to standard procedures (ie fix-ation using the methanol and staining according to theWrightndashGiemsa method) White blood cell (WBC) differen-tials were determined by quantifying the abundance oflymphocytes neutrophils monocytes basophils andeosinophils relative to a total number of white cells fixed atNtotal WBCfrac14 100 WBC counts were calculated as the averageof two replicated slides For each WBC type counts werecorrelated in a linear model with each explanatory variable(age sex rank or mange) No significant result was obtainedTherefore we assessed effects of age on cell population withTRA (Powell et al 2013) which is more sensitive to smalldifferences in cell prevalence than blood smear cell countsbecause TRA assesses gene expression of multiple cellmarkers

TOA and TRATOA was used to assess the extent to which differentiallyexpressed genes were predominately characteristic of oneor more subtypes of circulating leukocytes (Cole et al2011) Differential expression for TOA was defined by theFD with differentially expressed genes defined bylt090andgt110 FD per year of age Thresholds of FD were selectedto achieve the largest biological effect size possible while stillyielding an input list with at least 50 differentially expressedgenes TRA was used to assess the prevalence of cell-type-specific RNAs (Powell et al 2013) identified a priori based onhighly cell-type-diagnostic transcripts in humans (Cole et al2011 Powell et al 2013)

Supplementary MaterialSupplementary figures S1 and S2 and tables S1ndashS6 are avail-able at Molecular Biology and Evolution online (httpwwwmbeoxfordjournalsorg)

AcknowledgmentsWe would like to acknowledge Emily Almberg and ErinStahler for their contribution to data collection We arethankful to Devaughn Fraser for her significant help in theblood smear cell counts and to Ryan Harrigan for his com-ments on the manuscript This work was supported by theNational Science Foundation (grant numbers DEB 1257716 toRKW and BMV and DEB 1245373 to DRS and DWS) theNational Institutes of Health (grant numbers P30 AG017265to SWC S10 RR029668 and S10 RR027303 to work with theVincent J Coates GSL at UC Berkeley) the Austrian Academyof Sciences (to PC) the Max Kade Foundation (to PC) theHoward Hughes Medical Institute (to RAJ) the NationalPark Service (to DRS and DWS) the Yellowstone ParkFoundation (to DRS and DWS) the Tapeats Fund (toDRS and DWS) the Perkin-Prothro Foundation (to DRSand DWS) and an anonymous donor (to DRS and DWS)

ReferencesAhren B Mansson S Gingerich RL Havel PJ 1997 Regulation of plasma

leptin in mice influence of age high-fat diet and fasting Regul IntegrComp Physiol 273R113ndashR120

Alexander CM Landsman PB Teutsch SM Haffner SM 2003 NCEP-defined metabolic syndrome diabetes and prevalence of coronaryheart disease among NHANES III participants age 50 years and olderDiabetes 521210ndash1214

Almberg ES Cross PC Dobson AP Smith DW Hudson PJ 2012 Parasiteinvasion following host reintroduction a case study of Yellowstonersquoswolves Philos Trans R Soc Lond Ser B Biol Sci 3672840ndash2851

Almberg ES Cross PC Dobson AP Smith DW Metz MC Stahler DRHudson PJ 2015 Social living mitigates the costs of a chronic illnessin a cooperative carnivore Ecol Lett 18660ndash667

Almberg ES Mech LD Smith DW Sheldon JW Crabtree RL 2009 Aserological survey of infectious disease in Yellowstone NationalParkrsquos canid community PLoS One 4e7042

Anders S Pyl PT Huber W 2015 HTSeq a Python framework to workwith high-throughput sequencing data Bioinformatics 31166ndash169

Arlian LG Morgan MS Neal JS 2004 Extracts of scabies mites(Sarcoptidae Sarcoptes scabiei) modulate cytokine expression byhuman peripheral blood mononuclear cells and dendritic cells JMed Entomol 4169ndash73

Arlian LG Morgan MS Paul CC 2006 Evidence that scabies mites (AcariSarcoptidae) influence production of interleukin-10 and the func-tion of T-regulatory cells (Tr1) in humans J Med Entomol43283ndash287

Baylis D Bartlett D Patel H Roberts H 2013 Understanding how we ageinsights into inflammaging Longev Healthspan 21ndash8

Beirne C Delahay R Hares M Young A 2014 Age-related declines anddisease-associated variation in immune cell telomere length in a wildmammal PLoS One 9e108964

Benjamini Y Hochberg Y 1995 Controlling the false discovery rate apractical and powerful approach to multiple testing J R Stat Soc SerB 57289ndash300

Blankley S Berry MPR Graham CM Bloom CI Lipman M OrsquoGarra A2014 The application of transcriptional blood signatures to enhanceour understanding of the host response to infection the example oftuberculosis Philos Trans R Soc Lond Ser B Biol Sci 36920130427

Pervasive Effects of Aging doi101093molbevmsw072 MBE

9

at Duke U

niversity Law

School on June 26 2016httpm

beoxfordjournalsorgD

ownloaded from

Boren E Gershwin ME 2004 Inflamm-aging autoimmunity and theimmune-risk phenotype Autoimmun Rev 3401ndash406

Carrera-Bastos P Fontes-Villalba M OrsquoKeefe JH Lindeberg S Cordain L2011 The western diet and lifestyle and diseases of civilization ResRep Clin Cardiol 215ndash35

Chen E Miller GE Kobor M Cole SW 2011 Maternal warmth buffers theeffects of low early-life socioeconomic status on pro-inflammatorysignaling in adulthood Mol Psychiatry 16729ndash737

Cobb JP Mindrinos MN Miller-Graziano C Calvano SE Baker HV XiaoW Laudanski K Brownstein BH Elson CM Hayden DL et al 2005Application of genome-wide expression analysis to human healthand disease Proc Natl Acad Sci U S A 1024801ndash4806

Cole SW Conti G Arevalo JMG Ruggiero AM Heckman JJ Suomi SJ 2012Transcriptional modulation of the developing immune system byearly life social adversity Proc Natl Acad Sci U S A 10920578ndash20583

Cole SW Hawkley LC Arevalo JMG Cacioppo JT 2011 Transcript originanalysis identifies antigen-presenting cells as primary targets of so-cially regulated gene expression in leukocytes Proc Natl Acad Sci U SA 1083080ndash3085

Cole SW Hawkley LC Arevalo JM Sung CY Rose RM Cacioppo JT2007 Social regulation of gene expression in human leukocytesGenome Biol 8R189

Cubaynes S MacNulty DR Stahler DR Quimby KA Smith DW CoulsonT 2014 Density-dependent intraspecific aggression regulates sur-vival in northern Yellowstone wolves (Canis lupus) J Anim Ecol831344ndash1356

Di Mauro T David G 2009 Chromatin modifications the driving force ofsenescence and aging Aging 1182ndash190

dos Reis M Inoue J Hasegawa M Asher RJ Donoghue PCJ Yang Z 2012Phylogenomic datasets provide both precision and accuracy in es-timating the timescale of placental mammal phylogeny Proc R SocLond Ser B Biol Sci 2793491ndash3500

Edgar R Domrachev M Lash AE 2002 Gene Expression Omnibus NCBIgene expression and hybridization array data repository NucleicAcids Res 30207ndash210

Engstrom P Steijger T Sipos B Grant G Kahles A Reuroatsch G Goldman NHubbard T Harrow J Guigo R et al 2013 Systematic evaluation ofspliced alignment programs for RNA-seq data Nat Methods101185ndash1191

Fan Z Silva P Gronau I Wang S Armero AS Schweizer RM Ramirez OPollinger J Galaverni M Ortega Del-Vecchyo D et al 2016Worldwide patterns of genomic variation and admixture in graywolves Genome Res 26163ndash173

Farooqui AA 2015 Effect of long term consumption of high calorie diet andcalorie restriction on human health High calorie diet and the humanbrain Cham (Switzerland) Springer International Publishing p 1ndash28

Freedman AH Schweizer RM Gronau I Han E Ortega-Del Vecchyo DSilva PM Galaverni M Fan Z Marx P Lorente-Galdos B et al 2014Genome sequencing highlights genes under selection and the dy-namic early history of dogs PLoS Genet 10e1004016

Garrett-Sinha L 2013 Review of Ets1 structure function and roles inimmunity Cell Mol Life Sci 703375ndash3390

Gipson PS Ballard WB Nowak RM Mech LD 2000 Accuracy and pre-cision of estimating age of gray wolves by tooth wear J WildlManage 64752ndash758

Goring HHH Curran JE Johnson MP Dyer TD Charlesworth J Cole SAJowett JBM Abraham LJ Rainwater DL Comuzzie AG et al 2007Discovery of expression QTLs using large-scale transcriptional pro-filing in human lymphocytes Nat Genet 391208ndash1216

Gregori S Goudy KS Roncarolo MG 2012 The cellular and molecularmechanisms of immuno-suppression by human type 1 regulatory Tcells Front Immunol 31ndash12

Hansen KD Irizarry RA Wu Z 2012 Removing technical variability inRNA-seq data using conditional quantile normalization Biostatistics13204ndash216

Hoeppner MP Lundquist A Pirun M Meadows JRS Zamani N JohnsonJ Sundstrom G Cook A FitzGerald MG Swofford R et al 2014 Animproved canine genome and a comprehensive catalogue of codinggenes and non-coding transcripts PLoS One 9e91172

Holzenberger M Dupont J Ducos B Leneuve P Geloen A Even PCCervera P Le Bouc Y 2003 IGF-1 receptor regulates lifespan andresistance to oxidative stress in mice Nature 421182ndash187

Hong M-G Myers AJ Magnusson PKE Prince JA 2008 Transcriptome-wide assessment of human brain and lymphocyte senescence PLoSOne 3e3024

Hoopes B Rimbault M Liebers D Ostrander E Sutter N 2012 Theinsulin-like growth factor 1 receptor (IGF1R) contributes to reducedsize in dogs Mamm Genome 23780ndash790

Horvath S 2011 Integrated weighted correlation network analysis ofmouse liver gene expression data Weighted network analysis appli-cations in genomics and systems biology New York Springer Bookp 421

Ishioka K Hosoya K Kitagawa H Shibata H Honjoh T Kimura KSaito M 2007 Plasma leptin concentration in dogs effects ofbody condition score age gender and breeds Res Vet Sci8211ndash15

Ishioka K Soliman MM Sagawa M Nakadomo F Shibata H Honjoh THashimoto A Kitamura H Kimura K Saito M 2002 Experimentaland clinical studies on plasma leptin in obese dogs J Vet Med Sci64349ndash353

Kim D Pertea G Trapnell C Pimentel H Kelley R Salzberg SL 2013TopHat2 accurate alignment of transcriptomes in the presence ofinsertions deletions and gene fusions Genome Biol 14R36

Kircher M Sawyer S Meyer M 2012 Double indexing overcomes inac-curacies in multiplex sequencing on the Illumina platform NucleicAcids Res 40e3

Kirkwood TBL 1977 Evolution of ageing Nature 270301ndash304Kirkwood TBL 2005 Understanding the odd science of aging Cell

120437ndash447Kraus C Pavard S Promislow DEL 2013 The sizendashlife span trade-off

decomposed why large dogs die young Am Nat 181492ndash505Krishnamurthy J Torrice C Ramsey MR Kovalev GI Al-Regaiey K Su L

Sharpless NE 2004 Ink4aArf expression is a biomarker of aging JClin Invest 1141299ndash1307

Langfelder P Horvath S 2008 WGCNA an R package for weightedcorrelation network analysis BMC Bioinformatics 9(1) 13

Li H Matheny M Nicolson M Turner N Scarpace PJ 1997 Leptin geneexpression increases with age independent of increasing adiposity inrats Diabetes 462035ndash2039

Lindblad-Toh K Wade CM Mikkelsen TS Karlsson EK Jaffe DBKMClamp MCJ Kulbokas EJ 3rd Zody MCME Xie XBM Wayne RKet al 2005 Genome sequence comparative analysis and haplotypestructure of the domestic dog Nature 438803ndash819

Linton PJ Dorshkind K 2004 Age-related changes in lymphocyte devel-opment and function Nat Immunol 5133ndash139

Liu Y Sanoff HK Cho H Burd CE Torrice C Ibrahim JG Thomas NESharpless NE 2009 Expression of p16INK4a in peripheral blood T-cells is a biomarker of human aging Aging Cell 8439ndash448

Lopez-Otın C Blasco MA Partridge L Serrano M Kroemer G 2013 Thehallmarks of aging Cell 1531194ndash1217

Ma XH Muzumdar R Yang XM Gabriely I Berger R Barzilai N 2002Aging is associated with resistance to effects of leptin on fat distri-bution and insulin action J Gerontol A Biol Sci Med Sci57B225ndashB231

MacNulty DR Smith DW Mech LD Eberly LE 2009 Body size andpredatory performance in wolves is bigger better J Anim Ecol78532ndash539

MacNulty DR Smith DW Vucetich JA Mech LD Stahler DR Packer C2009 Predatory senescence in ageing wolves Ecol Lett 121347ndash1356

McGowan P Sasaki A DrsquoAlessio AC Dymov S Labonte B Szyf M TureckiG Meaney MJ 2009 Epigenetic regulation of the glucocorticoidreceptor in human brain associates with childhood abuse NatNeurosci 12342ndash348

Mech LD 1999 Alpha status dominance and division of labor in wolfpacks Can J Zool 771196ndash1203

Mech LD 2003 Wolf social ecology In Mech LD Boitani L editorsWolvesmdashbehavior ecology and conservation Chicago (IL) TheUniversity of Chicago Press p 1ndash34

Charruau et al doi101093molbevmsw072 MBE

10

at Duke U

niversity Law

School on June 26 2016httpm

beoxfordjournalsorgD

ownloaded from

Mejias A Ramilo O 2014 Transcriptional profiling in infectious diseasesready for prime time J Infect 68S94ndashS99

Miller GE Chen E Fok AK Walker HA Lim A Nicholls EF Cole SWKobor M 2009 Low early-life social class leaves a biological residuemanifested by decreased glucocorticoid and increase proinflamma-tory signaling Proc Natl Acad Sci U S A 10614716ndash14721

Mobbs CV 1998 Leptin and aging In Mobbs CV Hof PR editorsFunctional endocrinology of aging Basel (Switzerland) Karger p228ndash240

Molnar B Fattebert J Palme R Ciucci P Betschart B Smith DW Diehl P-A 2015 Environmental and intrinsic correlates of stress in free-rang-ing wolves PLoS One 10e0137378

Mostoslavsky R Chua KF Lombard DB Pang WW Fischer MR Gellon LLiu P Mostoslavsky G Franco S Murphy MM et al 2006 Genomicinstability and aging-like phenotype in the absence of mammalianSIRT6 Cell 124315ndash329

Murphy MLM Slavich GM Rohleder N Miller GE 2013 Targeted rejec-tion triggers differential pro- and anti-inflammatory gene expressionin adolescents as a function of social status Clin Psychol Sci 130ndash40

Nussey DH Froy H Lemaitre J-F Gaillard J-M Austad SN 2013Senescence in natural populations of animals widespread evidenceand its implications for bio-gerontology Ageing Res Rev 12214ndash225

Ohtani N Zebedee Z Huot TJG Stinson JA Sugimoto M Ohashi YSharrocks AD Peters G Hara E 2001 Opposing effects of Ets and Idproteins on p16INK4a expression during cellular senescence Nature4091067ndash1070

Packard JM 2003 Wolf behavior reproductive social and intelligent InMech LD Boitani L editors Wolvesmdashbehavior ecology and conser-vation Chicago (IL) and London The University of Chicago Press p35ndash65

Palacios MG Winkler DW Klasing KC Hasselquist D Vleck CM 2011Consequences of immune system aging in nature a study ofimmunosenescence costs in free-living Tree Swallows Ecology92952ndash966

Pence DB Windberg LA Pence BC Sprowls R 1983 The epizootiologyand pathology of sarcoptic mange in coyotes Canis latrans fromSouth Texas J Parasitol 691100ndash1115

Peters MJ Joehanes R Pilling LC Schurmann C Conneely KN Powell JReinmaa E Sutphin GL Zhernakova A Schramm K et al 2015 Thetranscriptional landscape of age in human peripheral blood NatCommun 68570

Poncet D Belleville A trsquoKint de Roodenbeke C Roborel de Climens ABen Simon E Merle-Beral H Callet-Bauchu E Salles G Sabatier LDelic J et al 2008 Changes in the expression of telomere mainte-nance genes suggest global telomere dysfunction in B-chronic lym-phocytic leukemia Blood 1112388ndash2391

Powell ND Sloan EK Bailey MT Arevalo JMG Miller GE Chen E KoborMS Reader BF Sheridan JF Cole SW 2013 Social stress up-regulatesinflammatory gene expression in the leukocyte transcriptome viabeta-adrenergic induction of myelopoiesis Proc Natl Acad Sci U S A11016574ndash16579

Rahman AA Karim AN Hamid ANA Harun R Wan Ngah WZ 2013Senescence-related changes in gene expression of peripheral bloodmononuclear cells from octononagenarians compared to their off-spring Oxid Med Cell Longev 201314

Ramilo O Allman W Chung W Mejias A Ardura M Glaser CWittkowski KM Piqueras B Banchereau J Palucka AK et al 2007Gene expression patterns in blood leukocytes discriminate patientswith acute infections Blood 1092066ndash2077

Reimand J Kull M Peterson H Hansen J Vilo J 2007 g profilermdasha web-based toolset for functional profiling of gene lists from large-scaleexperiments Nucleic Acids Res 35W193ndashW200

Revelle W 2015 Psych procedures for personality and psychologicalresearch Evanston (IL) Northwestern University

Ricklefs RE 2010 Life-history connections to rates of aging in terrestrialvertebrates Proc Natl Acad Sci U S A 10710314ndash10319

Robinson MD McCarthy DJ Smyth GK 2010 edgeR a Bioconductorpackage for differential expression analysis of digital gene expressiondata Bioinformatics 26139ndash140

Rollo CD 2002 Growth negatively impacts the life span of mammalsEvol Dev 455ndash61

Runcie DE Wiedmann RT Archie EA Altmann J Wray GA Alberts SCTung J 2013 Social environment influences the relationship be-tween genotype and gene expression in wild baboons PhilosTrans R Soc Lond Ser B Biol Sci 36820120345

Salminen A Kaarniranta K 2010 InsulinIGF-1 paradox of aging regu-lation via AKTIKKNF-jB signaling Cell Signal 22573ndash577

Sanchez-Rodrıguez M Garcıa-Sanchez A Retana-Ugalde R Mendoza-Nu~nez VcM 2000 Serum leptin levels and blood pressure in theoverweight elderly Arch Med Res 31425ndash428

Sapolsky RM 2004 Social status and health in humans and other ani-mals Annu Rev Anthropol 33393ndash418

Sapolsky RM 2005 The influence of social hierarchy on primate healthScience 308648ndash652

Schwarz F Springer SA Altheide TK Varki NM Gagneux P Varki A2016 Human-specific derived alleles of CD33 and other genes pro-tect against postreproductive cognitive decline Proc Natl Acad SciU S A 11374ndash79

Skoglund P Ersmark E Palkopoulou E Dalen L 2015 Ancient wolf ge-nome reveals an early divergence of domestic dog ancestors andadmixture into high-latitude breeds Curr Biol 251515ndash1519

Smith DW Stahler DR Albers E McIntyre R Metz M Irving J RaymondR Anton C Cassidy-Quimby K Bowersock N 2010 Yellowstonewolf project Annual Report Yellowstone National Park (WY)National Park Service

Smith DW Stahler DR Stahler E Metz M Quimby K McIntyre R Ruhl CMartin H Kindermann R Bowersock N et al 2012 Yellowstone wolfproject Annual report Yellowstone National Park (WY) NationalPark Service

Snyder-Mackler N Somel M Tung J 2014 Shared signatures of socialstress and aging in peripheral blood mononuclear cell gene expres-sion profiles Aging Cell 13954ndash957

Stahler DR MacNulty DR Wayne RK vonHoldt B Smith DW 2013 Theadaptive value of morphological behavioural and life-history traits inreproductive female wolves J Anim Ecol 82222ndash234

Storey JD Tibshirani R 2003 Statistical significance for genomewidestudies Proc Natl Acad Sci U S A 1009440ndash9445

Sutter NB Bustamante CD Chase K Gray MM Zhao K Zhu LPadhukasahasram B Karlins E Davis S Jones PG et al 2007 A singleIGF1 allele is a major derterminant of small size in dogs Science316112ndash115

Thalmann O Shapiro B Cui P Schuenemann VJ Sawyer SK GreenfieldDL Germonpre MB Sablin MV Lopez-Giraldez F Domingo-RouraX et al 2013 Complete mitochondrial genomes of ancientcanids suggest a European origin of domestic dogs Science342871ndash874

Tung J Barreiro LB Johnson ZP Hansen KD Michopoulos V Toufexis DMichelini K Wilson ME Gilad Y 2012 Social environment is asso-ciated with gene regulatory variation in the rhesus macaque im-mune system Proc Natl Acad Sci U S A 1096490ndash6495

Tung J Zhou X Alberts SC Stephens M Gilad Y 2015 The geneticarchitecture of gene expression levels in wild baboons eLife 4e04729

Velavan T Ojurongbe O 2011 Regulatory T cells and parasites J BiomedBiotechnol 20118

vonHoldt BM Stahler DR Smith DW Earl DA Pollinger JP Wayne RK2008 The genealogy and genetic viablity of reintroduced Yellostonegrey wolves Mol Ecol 17252ndash274

vonHoldt BM Stahler DR Bangs EE Smith DW Jimenez MD Mack CMNiemeyer CC Pollinger JP Wayne RK 2010 A novel assessment ofpopulation structure and gene flow in grey wolf populations of theNorthern Rocky Mountains of the United States Mol Ecol194412ndash4427

Wang J 2011 COANCESTRY a program for simulating estimating andanalysing relatedness and inbreeding coefficients Mol Ecol Resour11141ndash145

Weaver ICG Meaney MJ Szyf M 2006 Maternal care effects on hippo-campal transcriptome and anxiety-mediated behaviors in the

Pervasive Effects of Aging doi101093molbevmsw072 MBE

11

at Duke U

niversity Law

School on June 26 2016httpm

beoxfordjournalsorgD

ownloaded from

offspring that are reversible in adulthood Proc Natl Acad Sci U S A1033480ndash3485

Whitney AR Diehn M Popper SJ Alizadeh AA Boldrick JC Relman DABrown PO 2003 Individuality and variation in gene expression pat-terns in human blood Proc Natl Acad Sci U S A 1001896ndash1901

Wilson EO 2000 Sociobiology the new synthesis Cambridgeand London The Belknap Press of Harvard UniversityPress

Zhou X Stephens M 2012 Genome-wide efficient mixed-model analysisfor association studies Nat Genet 44821ndash824

Charruau et al doi101093molbevmsw072 MBE

12

at Duke U

niversity Law

School on June 26 2016httpm

beoxfordjournalsorgD

ownloaded from