Genome-wide analyses suggest parallel selection for universal traits may eclipse local environmental selection in a highly mobile carnivore

Genome-wide analyses sugges pa allel selec ion o uni e sal ai s may eclipse local en i onmen al selec ion in a highly mobile ca ni o e As id Vik S onen 1,2 , Bogumiła Je zd zejewska 2 , Cino Pe oldi 1,3 , Di e Demon is 4 , E o e Randi 1,5 , Magdalena Niedziałkowska 2 , Tomasz Bo owik 2 , Vadim E. Sido o ich 6 , Josip Kusak 7 , Ilpo Kojola 8 , Alexand os A. Ka amanlidis 9,10 , Janis Ozolins 11 , Vi alii Dumenko 12 & Sylwia D. Cza nomska 2 1 Sec ion o Biology and En i onmen al Science, Depa men o Chemis y and Bioscience, Aalbo g Uni e si y, F ed ik Baje s Vej 7H, DK-9220 Aalbo g Øs , Denma k 2 Mammal Resea ch Ins i u e, Polish Academy o Sciences, ul. Waszkiewicza 1, PL 17-230 Bialowieza, Poland 3 Aalbo g Zoo, Møllepa k ej 63, DK-9000 Aalbo g, Denma k 4 Depa men o Human Gene ics, Uni e si y o Aa hus, Wilhelm Meye s All e, DK-8000 Aa hus, Denma k 5 Labo a o io di Gene ica, ISPRA, ia C a Fo nace a 9, I-40064 Ozzano Emilia (BO), I aly 6 Ins i u e o Zoology, Scien i ic and P ac ical Cen e o Biological Resou ces, Na ional Academy o Science o Bela us, Akademicheskaya S 27, 220072 Minsk, Bela us 7 Depa men o Biology, Facul y o Ve e ina y Medicine, Uni e si y o Zag eb, Zag eb, C oa ia 8 Na u al Resou ces Ins i u e Finland, Box 16, FI-96500 Ro aniemi, Finland 9 ARCTUROS, Ci il Socie y o he P o ec ion and Managemen o Wildli e and he Na u al En i onmen , GR-53075 Ae os, G eece 10 Depa men o Ecology and Na u al Resou ces Managemen , No wegian Uni e si y o Li e Sciences, NO-1432  As, No way 11 La ian S a e Fo es Resea ch Ins i u e “Sila a”, R ıgas 111, LV-2169 Salaspils, La ia 12 Biosphe e Rese e Askania No a, F unze S . 13, Askania-No a, Chaplynka Dis ic , Khe son Region 75230, Uk aine Keywo ds CanineHD BeadChip mic oa ay, Canis lupus, en i onmen al selec ion, genome-wide associa ion s udy, single nucleo ide polymo phism, wol . Co espondence As id Vik S onen, Sec ion o Biology and En i onmen al Science, Depa men o Chemis y and Bioscience, Aalbo g Uni e si y, F ed ik Baje s Vej 7H, DK-9220 Aalbo g Øs , Denma k. Tel: +45 99403616; Fax: +45 96350558; E-mail: [email p o ec ed] Funding In o ma ion We g a e ully acknowledge unding om BIOCONSUS –Resea ch Po en ial in Conse a ion and Sus ainable Managemen o Biodi e si y (con ac no. 245737, FP7/ 2009-2014), he Mammal Resea ch Ins i u e o he Polish Academy o Sciences, he Polish Minis y o Science and Highe Educa ion (g an no. NN 303 418437) and BIOGEAST – Biodi e si y o Eas -Eu opean and Sibe ian la ge mammals on he le el o gene ic a ia ion o popula ions, 7 h F amewo k P og amme (con ac no. 247652). AVS ecei ed unding om he Danish Na u al Science Resea ch Council (pos doc o al g an 1337-00007). CP was suppo ed by he Abs ac Ecological and en i onmen al he e ogenei y can p oduce gene ic di e en ia ion in highly mobile species. Acco dingly, local adap a ion may be expec ed ac oss compa a i ely sho dis ances in he p esence o ma ked en i onmen al g adi- en s. Wi hin he Eu opean con inen , wol es (Canis lupus) exhibi dis inc no h–sou h popula ion di e en ia ion. We in es iga ed mo e han 67-K single nucleo ide polymo phism (SNP) loci o signa u es o local adap a ion in 59 un ela ed wol es om ou p e iously iden i ied popula ion clus e s (no hcen- al Eu ope n=32, Ca pa hian Moun ains n=7, Dina ic-Balkan n=9, Uk ai- nian S eppe n=11). Ou analyses combined iden i ica ion o ou lie loci wi h indings om genome-wide associa ion s udy o indi idual genomic p o iles and 12 en i onmen al a iables. We iden i ied 353 candida e SNP loci. We examined he SNP posi ion and neighbo ing megabase (1 Mb, one million bases) egions in he dog (C. lupus amilia is) genome o genes po en ially unde selec ion, including homologue genes in o he e eb a es. These egions included unc ional genes o , o example, empe a u e egula ion ha may indica e local adap a ion and genes con olling o unc ions uni e sally impo - an o wol es, including ol ac ion, hea ing, ision, and cogni i e unc ions. We also obse ed s ong ou lie s no associa ed wi h any o he in es iga ed a iables, which could sugges selec i e p essu es associa ed wi h o he unmea- su ed en i onmen al a iables and/o demog aphic ac o s. These pa e ns a e u he suppo ed by he examina ion o spa ial dis ibu ions o he SNPs asso- cia ed wi h uni e sally impo an ai s, which ypically show ma ked di e - ences in allele equencies among popula ion clus e s. Acco dingly, pa allel selec ion o ea u es impo an o all wol es may eclipse local en i onmen al selec ion and implies long- e m sepa a ion among popula ion clus e s. 4410 ª2015 The Au ho s. Ecology and E olu ion published by John Wiley & Sons L d. This is an open access a icle unde he e ms o he C ea i e Commons A ibu ion License, which pe mi s use, dis ibu ion and ep oduc ion in any medium, p o ided he o iginal wo k is p ope ly ci ed. Aalbo g Zoo Conse a ion Founda ion (AZCF), he Danish Na u al Science Resea ch Council (g an nos: 11-103926, 09-065999, 95095995), and he Ca lsbe g Founda ion (g an no. 2011-01-0059). Recei ed: 12 Augus 2015; Accep ed: 18 Augus 2015 Ecology and E olu ion 2015 5(19): 4410–4425 doi: 10.1002/ece3.1695 In oduc ion Local adap a ion may be p edic ed in a eas wi h limi ed in lux o no el genes, which can in e up selec ion o local en i onmen al condi ions, o in egions o high gene low coun e ed by s ong selec i e p essu es (Sla kin 1987 and e e ences he ein). An al e na e explana ion is selec- i e dispe sal wi h geno ypes p eadap ed o he local en i- onmen –o na al habi a -biased dispe sal (Da is and S amps 2004; Nosil e al. 2005; Edelaa e al. 2008) –a p ocess ha may help explain local adap a ion in highly mobile o ganisms wi h b oad geog aphic dis ibu ions. Ecological and en i onmen al di e en ia ion can cause popula ion gene ic s uc u e in highly mobile species (Da is and S amps 2004; Nosil e al. 2005), whe eby dis- pe se s selec habi a condi ions o which hey ha e na al expe ience and a e be e able o su i e and ep oduce. Acco dingly, gene ic di e gence may be expec ed ac oss compa a i ely sho geog aphic dis ances in he p esence o ma ked en i onmen al g adien s i gene ic d i is no o e - whelming he selec i e o ces. Long- e m esponses o selec ion in a ini e popula ion a e also in luenced by ac- o s dependen on he e ec i e popula ion size and popula- ion s uc u e (De Souza e al. 2000; Pe oldi e al. 2007). Al hough long-dis ance gene low occu s su icien ly o en o p oduce gene ic homogenei y o e a wide geo- g aphic ange (Sla kin 1985), new indings imply ha eco- logical and en i onmen al a ia ion can esul in gene ic di e en ia ion ac oss axa including wide- anging e es- ial and ma ine species. Examples include ish such as he - ing (Clupea ha engus, And  e e al. 2011), hake (Me luccius me luccius, Milano e al. 2014), and Bal ic Sea s ickleback (Gas e os eus aculea us, DeFa e i e al. 2013); sea u les ( e iewed in Bowen and Ka l 2007); and mammals includ- ing o ca (O cinus o ca, Hoelzel e al. 2007), couga (Puma concolo , McRae e al. 2005), lynx (Lynx canadensis, Rue- ness e al. 2003), and coyo e (Canis la ans, Sacks e al. 2004, 2005). The unde s anding o local adap a ion he e- o e has implica ions ac oss he axonomic ange including wild species and domes ic animals (e.g., Pa ise e al. 2009). Whe eas ca ni o es a e highly mobile, hey can exhibi ma ked popula ion gene ic s uc u e ha may ha e impo an e olu iona y implica ions. P e e ence o na al habi a s is p oposed o explain popula ion s uc u e in one o he mos mobile and widely dis ibu ed species o la ge ca ni o es, he g ay wol (Canis lupus, Ca michael e al. 2001; Weckwo h e al. 2005, 2010, 2011; Pilo e al. 2006, 2012; Musiani e al. 2007; Mu~ noz-Fuen es e al. 2009; S onen e al. 2014). Eu opean wol es ha e been a ec ed by human-induced landscape changes ha esul ed in small and o en isola ed popula ions (Linnell e al. 2008) in pa due o o e ha es ing (Randi 2011). Popula ions such as hose o he I alian and Ibe ian peninsulas ha e been subjec o a subs an ial amoun o gene ic d i due o low e ec i e popula ion size and demog aphic s ochas ici y (Lucchini e al. 2004; Fabb i e al. 2007; S onen e al. 2013; Pilo e al. 2014a). The gene ic di e gence be ween popula ions o wide- anging species has been in luenced by biogeog aphic p o- cesses such as glacia ions, and ecoloniza ion om glacial e ugia may help explain di e en ia ion be ween neigh- bo ing popula ions o wide- anging ca ni o es (e.g., Manel e al. 2004 and e e ences he ein). Wol es appea o ha e been common in he Eu asian La e Pleis ocene aunal complex and may ha e been dis ibu ed h ough- ou Eu ope du ing his ime (Kahlke 1999). The occu - ence o cold-adap ed p ey species such as eindee (Rangi e a andus) and mammo h (Mammu hus p imige- nius) (Kahlke 1999; Somme and Nadachowski 2006) in sou he n and cen al Eu ope du ing he las glacial maxi- mum sugges s ha a wol eco ype adap ed o a c ic con- di ions migh ha e been widely dis ibu ed. Wol es may ha e been p esen in cen al Eu ope du ing he Pleni-Gla- cial epoch (ci ca 75–15,000 BC) wi h dynamic ange changes du ing he Holocene o di e en eco ypes adap ed o condi ions such as a c ic und a, o es , and humid clima es (Somme and Benecke 2005). The spa io- empo al ex en o selec ion in wol es may be highly complex, and he ela i e in luence o local en i onmen al selec ion since he las glacial maximum e sus indepen- ª2015 The Au ho s. Ecology and E olu ion published by John Wiley & Sons L d. 4411 A. V. S onen e al.Genome-wide Analyses o Selec ion in Wol es den selec ion in p e iously sepa a ed popula ions is no well unde s ood. Fo simplici y, we hence o h e e o “ancien ” selec ion as ha ha ing occu ed p io o he las glacial maximum and “ ecen ” as ha ing aken place a e wa d. The Eu opean con inen encompasses impo an en i- onmen al a ia ion. The di e se geog aphy wi h (pa - ially) eas –wes -o ien ed moun ain chains (Alps, Ca pa hians) and he Medi e anean and Bal ic Seas migh exe mo e complex spa ial in luence on popula ion s uc u e and gene low han ha obse ed in No h Ame ica wi h well-sepa a ed coas al and con inen al cli- ma es (e.g., Ge en e al. 2004). Eu opean wol es showed clea popula ion gene ic s uc u e when e alua ed o e 67,000 (hence o h 67 K) single nucleo ide polymo phism (SNP) ma ke s (S onen e al. 2013), bu i emains unclea whe he adap a ion o a ious en i onmen al con- di ions migh help explain he obse ed popula ion clus- e s. Al hough some le el o gene ic s uc u e seems o ha e been es ablished p io o he las glacial maximum (Pilo e al. 2010), wol es likely had a con inuous ange h ough he Holocene wi h popula ion agmen a ion and habi a loss p ima ily occu ing in he pas ew cen u ies (Pilo e al. 2014a). Whe eas gene ic d i has a ec ed Eu opean wol es o e he pas hund ed yea s, his p ocess seems o ha e been less p onounced in eas –cen al Eu - ope whe e popula ions ha e emained ela i ely well con- nec ed (S onen e al. 2013; Pilo e al. 2014a). Gene ic d i , popula ion demog aphic his o y and o he neu al p ocesses could be majo in luences on allele equencies and dis ibu ions whe e selec ion is weak (Coop e al. 2009). We none heless expec gene ic d i o ha e an o e all in luence ac oss he en i e genome whe eas selec- ion is p edic ed o ac only on ce ain genes. Addi ion- ally, we expec he co ela ion be ween neu al molecula di e si y and non-neu al gene ic a ia ion o be weak in s able popula ions, and o dec ease u he when popula- ions expand o decline in size (Pe oldi e al. 2007). Ou s udy aimed o de e mine whe he popula ion s uc u e associa ed wi h unc ional gene ic a ia ion in Eu opean wol es 1) is consis en wi h p e iously obse ed (and assumed p edominan ly “neu al”) gene ic s uc u e and 2) appea s be e explained by ancien selec ion o com- mon ai s occu ing in pa allel in sepa a e popula ions, o by ecen selec ion based on local en i onmen al con- di ions. Ma e ials and Me hods Samples DNA ex ac ion and geno yping We examined wol p o iles om 10 coun ies ac oss Eu - ope, geno yped wi h he CanineHD BeadChip mic oa ay wi h 170,000 SNP loci om Illumina (Illumina, Inc., San Diego, CA) as desc ibed in S onen e al. (2013). The ea - lie s udy included I alian wol es, bu owing o hei highly di e gen s a us (S onen e al. 2013; Pilo e al. 2014a) and he possibili y ha s ong gene ic d i be ween I alian and o he Eu opean wol es migh con- ound signals o selec ion, we excluded all I alian indi id- uals om he analyses. Mo eo e , we emo ed ou lie p o iles om o he coun ies including pu a i e wol –dog hyb ids, which esul ed in a sample o n=113 wol es. Subsequen ly, we used PLINK (Pu cell e al. 2007) o iden i y pai s o wol es wi h an iden i y-by-descen (IBD, o PI_HAT) sco e o ≥0.1 and emo ed one indi idual pe pai (some wol es had alues abo e he h eshold o mul iple pai wise compa isons) o limi he po en ially con ounding e ec o c yp ic ela edness (see, e.g., Smi h e al. 2010) on possible signals o selec ion. The sc eening esul ed in a sample o n=59 Eu opean wol es om ou popula ion clus e s (no hcen al Eu ope n=32, Ca pa hian Moun ains n=7, Dina ic-Balkan n=9, Uk ainian S eppe n=11, Fig. 1) p e iously iden i ied by S onen e al. (2013). S a is ical analyses o gene ic s uc u e We pe o med analyses in wo s ages. We i s combined genome-wide associa ion s udy (GWAS, e.g., Smi h e al. 2010) o geno ype–en i onmen associa ions in PLINK wi h a complimen a y app oach using BayeScan (Foll and Gaggio i 2008) o de ec ing ou lie loci wi hou consid- e ing en i onmen al da a. We pe o med GWAS wi h 99,551 SNPs quali y-con olled and il e ed o mino allele equency and geno yping call a e (PLINK se ings: ma 0.01, geno 0.02) as desc ibed in S onen e al. (2013). Fo he GWAS, we included all da a o e ain as much in o ma ion as possible o indi iduals and SNP loci asso- cia ed wi h en i onmen al ac o s. Subsequen ly, we used a 67-K e sion o he da a p uned o linkage disequilib- ium as desc ibed in S onen e al. (2013) o pe o m he BayeScan analyses ca ied ou pe popula ion (clus e ). The esul ing candida e SNPs we e e alua ed wi h he spa- ial analysis me hod (SAM) implemen ed in he p og am Ma SAM 2 Be a (Joos e al. 2007, 2008) because analysis in ol ing housands o loci was no p ac ically easible in Ma SAM 2 Be a. Howe e , he SAM app oach is de el- oped o analyses o geno ype–en i onmen associa ions in wild o domes ic species (see, e.g., Pa ise e al. 2009) and he e o e well-sui ed o he pu pose o ou s udy. We pe o med GWAS in PLINK using he linea eg ession op ion, whe eby each indi idual was assessed based on 12 en i onmen al a iables (Table 1). En i on- men al a iables we e es ed o de ia ions om no mal dis ibu ion in PAST (Hamme e al. 2001), and we log- 4412 ª2015 The Au ho s. Ecology and E olu ion published by John Wiley & Sons L d. Genome-wide Analyses o Selec ion in Wol es A. V. S onen e al. ans o med alues o which he p obabili y plo co ela- ion coe icien (PPCC) was <0.8. As a esul , PPCC o all a iables excep wo (log al i ude =0.83, log biome =0.86) was >0.93. We examined co ela ion among en i onmen al a iables in PAST using he es op ion Kendall’s au o nonpa ame ic da a, which is a ecommended op ion o da a se s wi h many ied anks (Legend e and Legend e 1998). We adjus ed o mul iple es ing using he Bon e oni co ec ion and ca ego ized ela ionships be ween pai s o a iables as highly (>0.6), mode a ely (0.3–0.6), o no co ela ed (<0.3). A p io i exclusion o co ela ed a iables (e.g., July, Janua y, and annual empe a u e) migh miss impo an in o ma ion, and we chose o e ain all a iables and epo hei ex en o co ela ion (Table S1). We e alua ed he inclu- sion o 2–15 co a ia es ob ained om mul idimensional scaling o he da a in PLINK o accoun o popula ion s a i ica ion (F eedman e al. 2004 and e e ences he ein; S onen e al. 2013) and pe o med GWAS wi h six co a ia es, he lowes numbe o co a ia es o which he genome-in la ion ac o was <1.05 o all a iables. GWAS es s we e pe o med o he mino allele o each locus, and we implemen ed Bon e oni co ec ions o mul iple es ing (P<0.05). We included all en i onmen al a i- ables o he inal analysis in Ma SAM, as his app oach is based on logis ic eg ession and does no equi e no mal dis ibu ions. Subsequen ly, we pe o med simula ions in BayeScan (Foll and Gaggio i 2008) o he 67-K SNPs o iden i y ou lie loci. We es ed a ious le els o p io (10, 100, Uk ainian S eppe Dina ic-Balkan Ca pa hianNo hcen al 1 2 3 4 5 6 7 1 2 3 4 5 6 7 1 2 3 4 5 6 7 5G oups Sample loca ions 1 2 3 4 5 6 7 0 500 km Figu e 1. S udy a ea and loca ions o 59 wol es used in analyses o single nucleo ide polymo phisms (SNPs). Spa ial in e pola ion o ou SNPs wi h geno ypes speci ic o di e en popula ion clus e s is shown as examples. The la ge no hcen al Eu opean clus e was di ided in o g oups 1–4 o in es iga ion o possible egional pa e ns (geno ype 223AA), g oup 5 is he Ca pa hian Moun ains (342GA), g oup 6 is he Uk ainian S eppe (236AG), and g oup 7 is Dina ic-Balkan (214AA). SNP allele equencies among samples in each clus e we e classi ied as <25% (whi e), 25–49% (ligh g ay), 50–75% (medium g ay), and >75% (da k g ay). SNP iden i ica ions a e p o ided in Table S2. ª2015 The Au ho s. Ecology and E olu ion published by John Wiley & Sons L d. 4413 A. V. S onen e al.Genome-wide Analyses o Selec ion in Wol es and 1000) as he chosen alue ep esen s a ade-o be ween alse posi i es and he abili y o de ec possible ou lie s (Foll 2012). Because he loci iden i ied as ou lie s we e highly consis en among uns, we e ained he esul s o p io o 10 and epo loci o which he log10(PO) al- ues we e >0.5 as ecommended in he p og am guidelines (Foll 2012). We an analyses including all ou popula ion clus e s (labeled 4P) and hen pe o med compa isons be ween each pai o clus e s (no hcen al Eu ope =N, Ca pa hian Moun ains =C, Dina ic-Balkan =B, Uk ai- nian S eppe =U). Posi i e alues o he pa ame e alpha (alpha >0) indica e di e gen selec ion, whe eas nega i e alues (alpha <0) sugges balancing selec ion. Fo he second s age o analysis, we e alua ed GWAS and BayeScan candida e loci wi h Ma SAM. The p og am pe o ms es s o logis ic eg ession o each SNP geno- ype ( o which he e a e no mally h ee: AA, AB, BB) and he en i onmen al a iable in ques ion. The p og am implemen s wo sepa a e es s, a likelihood a io (hence- o h G) es and a Wald-Be a es (Joos e al. 2007), o de e mine whe he a pa icula geno ype is associa ed wi h a gi en en i onmen al a iable. The p og am epo s bo h es esul s, as well as a cumula i e es . The cumula- i e es is signi ican when bo h Wald and G- es s ejec he null hypo hesis ha he model wi h he obse ed a i- able does no explain he obse ed geno ype dis ibu ion be e han a model wi h a cons an only (Joos e al. 2007). The p og am implemen s he Bon e oni co ec- ion o mul iple es s, and we chose a P-le el o 0.05. Values o wo ca ego ical a iables, ecozone and biome, we e en e ed as numbe s using he “independen ” design (Joos and Kalbe ma en 2010). We e alua ed spa ial pa e ns h oughou he s udy a ea by plo ing allele equency dis ibu ions o all candida e loci. The la ge no hcen al clus e was di ided in o ou g oups based on geog aphic p oximi y o sample loca ions o e alua e he possibili y o local pa e ns. In e pola ion maps o esul s ha displayed geog aphic pa e ns we e p epa ed wi h A cGIS 10.2 (ESRI 2013). Samples we e in e pola ed in o con inuous su aces wi h he in e se dis ance weigh ed (IDW) me hod. The in e pola ion was conduc ed o ou SNPs wi h geno ypes speci ic o di e en popula ion clus e s. Geno ype equencies we e s o ed in bina y o ma (1 – p esen , 0 –no p esen ), and he mean alue was cal- cula ed o each clus e . Fo he no hcen al clus e , we used he ou abo e-men ioned g oups o assess he possi- ble p esence o local pa e ns. In e se dis ance weigh ed in e pola ion was pe o med wi h de aul pa ame e s. Sin- gle nucleo ide polymo phism allele p obabili y was classi- ied in o ou g oups (low –less han 25%, mode a e –25– 50%, high –50–75%, and e y high –mo e han 75%). Subsequen ly, we used he 67-K SNPs in Genepop (Rousse 2008) o calcula e F ST alues o each locus. We hen employed Hie Fs a (Goude 2005) o ob ain pai - wise F ST alues wi h 95% con idence in e als be ween popula ion clus e s o he 353 candida e loci iden i ied in GWAS and BayeScan. We examined hese popula ion clus e s by p incipal componen analyses (PCA) wi h he adegene package (Jomba 2008) in R 2.14.2 (R De elop- men Co e Team 2012). Resul s We de ec ed 178 ou lie SNPs in BayeScan and 175 SNPs wi h pu a i e associa ion wi h en i onmen al a iables by Table 1. En i onmen al a iables o genome-wide associa ion s udy o Eu opean wol es (n=59) wi h 67-K single nucleo ide polymo - phism (SNP) loci. Va iable Label Uni Da a sou ce Longi ude long Decimal deg ees Sample coo dina es La i ude la Decimal deg ees Sample coo dina es Human popula ion densi y popd Numbe o people/km 2 1) Mean annual empe a u e ann Deg ees Celsius 2) Mean Janua y empe a u e jan Deg ees Celsius 2) Mean July empe a u e jul Deg ees Celsius 2) Annual p ecipi a ion p ed mm 2) Road densi y oad km oad/100 km 2 3) Al i ude al Me e s abo e sea le el 4) Snow co e dep h snow cm 5) Ecosys em code ecoc Numbe (o dinal) 6) Biome code bioc Numbe (o dinal) 6) 1) h p://epp.eu os a .ec.eu opa.eu/po al/page/po al/eu os a /home/, Ma ch 2012. 2) Hijmans, R.J., S.E. Came on, J.L. Pa a, P.G. Jones and A. Ja is (2005). Ve y high esolu ion in e pola ed clima e su aces o global land a eas. In e na ional Jou nal o Clima ology 25: 1965–1978. (Wo ldClim p ojec da a). 3) ESRI Da a & Maps (2008). Redlands, CA: En i onmen al Sys ems Resea ch Ins i u e [CD-ROM]. 4) U.S. Geological Su ey (2004), EROS Da a Cen e Dis ibu ed Ac i e A chi e Cen e (EDC DAAC), Global Digi al Ele a ion Model (GTOPO30), Redlands, Cali o nia, USA. (GTOPO30 da abase). 5) A onin, A.N., S.L. G eene, N.I. Dzyubenko, A.N. F olo (2008) In e - ac i e Ag icul u al Ecological A las o Russia and Neighbo ing Coun- ies. Economic Plan s and hei Diseases, Pes s and Weeds. A ailable a : h p://www.ag oa las. u. 6) Olson, D. M., E. Dine s ein (2002). The Global 200: P io i y eco e- gions o global conse a ion. (PDF ile) Annals o he Missou i Bo ani- cal Ga den 89:125–126. A ailable a : h p://www.wo ldwildli e. o g/science/da a/ e eco.c m. (WWF da abase). 4414 ª2015 The Au ho s. Ecology and E olu ion published by John Wiley & Sons L d. Genome-wide Analyses o Selec ion in Wol es A. V. S onen e al. GWAS. The e was no o e lap be ween he loci epo ed by each me hod. One hund ed and se en y- i e o 178 SNPs (98%) iden i ied by BayeScan had F ST alues ≥0.15 (as es ima ed by Genepop ac oss he 59 wol es and 353 loci), which may be conside ed as a high (Balloux and Lugon-Moulin 2002), whe eas o GWAS he numbe o SNPs wi h F ST alues ≥0.15 was 21 o 175 (12%). Mean F ST alue o BayeScan loci was 0.305 ( ange 0.118– 0.571), and o GWAS, i was 0.085 ( ange 0.000–0.432). All BayeScan esul s had a posi i e alpha alue, sugges ing di ec ional a he han balancing selec ion. O e 66% o he BayeScan loci had a high loading (he e de ined as ≥[0.01]) on one o wo o he h ee PC axes in a PCA o he Eu opean wol popula ion wi h 67-K loci (S onen e al. 2013 Fig. 2B and C) and hus made an ob ious con ibu ion o popula ion s uc u e. GWAS loci showed no such pa e n. O he 353 SNPs, geno ypes in 117 (46 om GWAS and 71 om BayeScan) we e iden i ied as associa ed wi h en i onmen al a iables by SAM. All cases in which geno- ypes we e signi ican ly associa ed wi h he a iable “biome code” (bioc) we e iden i ied by he Wald es . No o he geno ype–en i onmen associa ion was ound by he Wald es , and esul s o all o he a iables we e iden i ied by he G- es . GWAS esul s a ec ed by linkage (n=99) a e ma ked in Table S2. Wi h he excep ion o i e SNPs (iden i ied in Table S4), he ollowing esul s include only loci una ec ed by linkage. We examined each SNP and one megabase (Mb; one million bases) on ei he side (he ea e lanking egions) in he UCSC dog genome b owse (h p://genome.ucs- c.edu/cgi-bin/hgT acks) and he NCBI Map Viewe (h p://www.ncbi.nlm.nih.go /p ojec s/map iew/) o iden- i y genes o genomic egions known o assumed o be 34 7 0 0.4 0.6 1.0 0.8 0.2 12345 0 0.4 0.6 1.0 0.8 0.2 1234567 12345 0 0.4 0.6 1.0 0.8 0.2 12345 12345 12 5667 67 67 0 0.4 0.6 1.0 0.8 0.2 1234567 0 0.4 0.6 1.0 0.8 0.2 12345 67 67 No hcen al Ca pa hian 0 0.4 0.6 1.0 0.8 0.2 Uk ainian S eppe 0 0.4 0.6 1.0 0.8 0.2 Dina ic-Balkan 0 0.4 0.6 1.0 0.8 0.2 Region F equency o geno ypes F equency o geno ypes Figu e 2. Spa ial dis ibu ions o Eu opean wol single nucleo ide polymo phism (SNP) loci/ geno ypes ypical o single popula ion clus e s. The g aphs show equencies o loci/geno ypes di e en ia ing among wol es in no hcen al (g oups 1–4), Ca pa hian (5), Uk ainian S eppe (6), and Dina ic-Balkan clus e s (g oup 7). Numbe s on x-axis a e wol g oups 1–7 (see Fig. 1). Le panels: loci/geno ypes wi h high equencies in a gi en clus e . Righ panels: loci/geno ypes wi h low equencies in a gi en clus e . SNP loci and geno ypes a e lis ed in Table S4. ª2015 The Au ho s. Ecology and E olu ion published by John Wiley & Sons L d. 4415 A. V. S onen e al.Genome-wide Analyses o Selec ion in Wol es Table 2. Func ional genes nea single nucleo ide polymo phisms (SNP) iden i ied as ou lie loci and/o associa ed wi h en i onmen al a iables based on a s udy o 59 wol es in ou Eu opean popula ion clus e s. En i onmen al a iables a e gi en in Table 1. Full locus iden i ica ion om he Illumina CanineHD BeadChip is p o ided in Table S2. Func ion summa y is based on e e ences om he NCBI da abase (h p:// www.ncbi.nlm.nih.go /gene). Ch and SNP numbe 1 BayeScan log10(PO) 2 BayeScan FDR 3 SAM esul 4 F ST5 Gene(s) Func ion summa y TEMPERATURE Ch 9_143 0.904 (4P) 1.134 (BU) 0.058 (4P) 0.034 (BU) –0.327 RPTOR The mogenesis Ch 9_148 1.217 (4P) 0.027 (4P) jul (AA) 0.332 TRPV1/TRPV3 The mo egula ion Ch 25_269 0.771 (BC) 0.069 (BC) bioc (AA) 0.197 TRPM8 The mosensa ion (cold senso ) METABOLISM Ch 5_85 –– bioc (GA) 0.022 SGIP1 Fa mass, ood in ake Ch 5_85 –– bioc (GA) 0.022 LEPR Fa me abolism Ch 5_100 0.860 (CU) 0.079 (CU) bioc (AC) 0.260 TK2 m DNA syn hesis Ch 9_151 –– bioc (AA,CC) 0.236 CRAT Ene gy homeos asis, a me abolism Ch 9_151 –– bioc (AA,CC) 0.236 DNM1 Exe cise-induced collapse Ch 15_188 1.089 (4P) 0.034 (4P) bioc (GG) 0.230 NPYR1 Vasocons ic ion in exe cising skele al muscle Ch 18_208 0.725 (4P) 1.118 (NC) 0.080 (4P) 0.064 (NC) bioc (AG,GG) 0.202 CPT1A m DNA memb ane, lipid me abolism Ch 26_280 0.813 (NU) 0.068 (NU) bioc (AA), jul (GG) 0.255 SLC5A1 Ca bohyd a e diges ion/abso p ion. Ch 32_326 –– bioc (CG) 0.033 SCD5 Ene gy me abolism PHYSICAL DEVELOPMENT Ch 3_23 0.841 (4P) 1.341 (BC) 0.067 (4P) 0.028 (BC) –0.342 IGFI1R Reduced size (dogs) Ch 4_46 0.889 (4P) 1.423 (NB) 0.059 (4P) 0.017 (NB) la , al (AA) 0.497 ZFR RNA egula ion Ch 13_169 1.207 (4P) 1.214 (CU) 0.028 (4P) 0.042 (CU) –0.260 RSPO2 Dog coa colo Ch 13_175 1.329 (CU) 0.038 (CU) –0.283 KIT Dog coa pa e ns (spo ed Weima ane ) Ch 15_183 0.511 (NB) 0.098 (NB) –0.231 ATP2B1 In acellula calcium homeos asis; ascula smoo h muscle cells; possibly Chagas disease (Ame ican ypanosomiasis) Ch 15_187 1.168 (4P) 1.648 (NU) 0.030 (4P) 0.015 (NU) –0.332 FNIP2 Hypomyelina ion in he b ain; spinal co d de ec s (Weima ane dogs) Ch 18_208 0.725 (4P) 1.118 (NC) 0.080 (4P) 0.064 (NC) bioc (AG,GG) 0.202 FGF4 Bone mo phogenesis Ch 19_210 1.206 (NU) 0.033 (NU) –0.303 DARS Hypomyelina ion (b ain, spinal co d) Ch 21_217 0.591 (4P) 1.475 (NB) 0.672 (BU) 0.096 (4P) 0.014 (NB) 0.096 (BU) –0.288 PPFIBP2 Neu al synapse de elopmen Ch 21_222 0.629 (NU) 0.111 (NU) bioc (AG,GG) 0.231 HPS5 He mansky–Pudlak synd ome (oculocu aneous albinism, pla ele abno mali y) Ch 21_223 0.818 (NU) 0.061 (NU) la , ann (AA) 6 0.395 NAV2 Neu on g ow h and egene a ion Ch 21_225 0.804 (NU) 0.074 (NU) –0.226 ANO3 Dominan c anioce ical dys onia (sus ained muscle con ac ions – epe i i e mo emen s o abno mal pos u es); eczema, as hma Ch 23_236 0.908 (4P) 0.794 (NU) 0.052 (4P) 0.085 (NU) p ec (AG,GG) 0.361 AGTR1 Angio ensin II (blood p essu e and olume) Ch 23_236 0.908 (4P) 0.794 (NU) 0.052 (4P) 0.085 (NU) p ec (AG,GG) 0.361 HPS3 He mansky–Pudlak synd ome (oculocu aneous albinism, pla ele abno mali y) Ch 23_236 0.908 (4P) 0.794 (NU) 0.052 (4P) 0.085 (NU) p ec (AG,GG) 0.361 CP Ace uloplasminemia (i on accumula ion and issue damage) Ch 24_245 0.537 (4P) 0.106 (4P) long (AA) 0.340 BMP7 Bone g ow h Ch 24_246 0.579 (NB) 0.079 (NB) bioc (GG, GA, AA) 0.338 COL9A3 Collagen (dwa ism, ocula de ec s) 4416 ª2015 The Au ho s. Ecology and E olu ion published by John Wiley & Sons L d. Genome-wide Analyses o Selec ion in Wol es A. V. S onen e al. o unc ional impo ance (hence o h e e ed o as unc ional genes). Thi y- wo key unc ional genes (o g oups o genes) nea SNP loci iden i ied as ou lie s (n=27) and/o associa ed wi h en i onmen al a iables (n=22) a e lis ed in Table 2 and di ided in o g oups based on unc ion: empe a u e (n=3), me abolism (n=9), and physical de elopmen (n=20). One SNP was associa ed wi h a iables (la i ude, annual empe a- u e) ound o be co ela ed (Table 2; Table S1). Comple e nomencla u e and iden i ica ion o SNP loci a e p o ided in Table S2. Fu he mo e, we obse ed SNPs nea key unc ional genes associa ed wi h ea u es o which we do no ha e en i onmen al da a o ha appea impo an o all wol es ac oss hei ange. We ha e highligh ed n=12 SNPs associa ed wi h disease and pa asi es, n=16 o senso y unc ions, and n=9 o b ain and cogni ion (Table S3). Fou o hese SNPs we e associa ed wi h co ela ed a iables (Tables S1 and S3). Ou esul s exhibi ed clea spa ial pa e ns in one (Figs. 1, 2) o –less equen ly – wo popula ion clus- e s (Fig. 3), including SNPs nea genes o unc ions belie ed o be impo an o wol es ac oss hei ange. Ce ain o hese SNPs showed dis inc geog aphic dis i- bu ions o geno ypes (Table S4). Fo example, geno ypes a ied be ween he Ca pa hian Moun ains and he Uk ainian S eppe/Dina ic-Balkan clus e s o SNP Ch 13_175 loca ed nea he gene KIT (dog coa pa - e n). We selec ed one ep esen a i e geno ype o each clus e o in e pola ion in o con inuous su ace (Fig. 1). We hen no ed SNPs nea genes o impo an unc ions ha showed no ob ious spa ial pa e ns (Table S5). Se - e al SNPs we e also iden i ied as s ong ou lie s in BayeScan bu no associa ed wi h any o he 12 en i on- men al a iables, no we e he e any unc ional genes epo ed in he 1-Mb lanking egions. These esul s migh ne e heless be o in e es o u u e in es iga ion (Table S6). Pai wise F ST alues be ween he ou popula ion clus- e s, wi h he ull sample o 113 indi iduals and all 353 loci, showed he highes alue o Ca pa hian Moun ains – Uk ainian S eppe –and he lowes alue o Ca pa hian Moun ains –no hcen al Eu ope (Table 3). Pai wise F ST alues o he sample o 59 indi iduals we e simila ly high and gene ally consis en wi h he la ge sample, al hough he highes alue was be ween no hcen al Eu ope and Dina ic-Balkan and he lowes was o no hcen al Eu - ope –Uk ainian S eppe (Table S7). P incipal componen analyses o all 113 wol es showed di e en ia ion among all popula ion clus e s (Fig. 4). Al hough no hcen al Eu ope and Ca pa hian Moun ain indi iduals o e lapped on he 1s axis, hey we e clea ly dis inc on he 3 d axis. The 1s axis e lec s no h–sou h di e en ia ion in Eu o- pean wol es, whe eas he 2nd axis gene ally (al hough he e is some spa ial o e lap be ween no hcen al Eu ope and Uk ainian S eppe) indica es eas –wes s uc u e. When compa ed o he o he h ee clus e s, he indi idual p o iles om no hcen al Eu ope appea highly concen- a ed ela i e o hei spa ial dis ibu ion (Figs. 1, 4). Discussion Ou esul s iden i ied genes po en ially in luencing local adap a ion o empe a u e, me abolism, physical de elop- men , and disease/immune sys em unc ions in Eu opean wol es. Howe e , he impo ance o SNPs associa ed wi h genes o pu a i e local adap a ions appea s o e shadowed by indings linked o ai s o uni e sal impo ance, including hea ing, ision, ol ac ion, and cogni i e unc- ions. This sugges s ha ancien , concu en , and possibly pa allel selec ion may ha e played a mo e p ominen ole han ecen local adap a ion in s uc u ing unc ional Table 2. Con inued. Ch and SNP numbe 1 BayeScan log10(PO) 2 BayeScan FDR 3 SAM esul 4 F ST5 Gene(s) Func ion summa y Ch 26_281 1.547 (4P) 0.779 (NU) 0.012 (4P) 0.089 (NU) bioc (GG) 0.352 ADORA2A Ca diac hy hm and ci cula ion, blood low, immune unc ion, pain egula ion, sleep Ch 28_299 1.1880 (NB) 0.008 (NB) la (GG) 0.290 SPRCS3 Cen al ne ous sys em de elopmen Ch 31_317 1.062 (BC) 0.043 (BC) bioc (GG) 0.315 ADAMTS1 O gan mo phology and unc ion 1 Full SNP iden i ica ion gi en in Table S2. 2 Pai wise compa isons o : B –Balkan-Dina ic; C –Ca pa hian Moun ains.; U –Uk ainian S eppe; N –no hcen al Eu ope. 4P: ac oss all ou clus- e s. 3 False disco e y a e h eshold (q- alue). 4 En i onmen al a iables iden i ied by he spa ial analysis me hod (SAM) as signi ican ly associa ed wi h one o mo e geno ypes. SAM inco po a es wo sepa a e es s: he Wald and he likelihood a io (G) es (Joos e al. 2007). The a iable “bioc” was iden i ied by he Wald es ; all o he a iables by he G- es . No esul was iden i ied in bo h. 5 F ST calcula ed ac oss all 353 loci o all popula ion clus e s. 6 Co ela ions be ween (some) a iables. See Table S1 wi h esul s o all a iable combina ions. ª2015 The Au ho s. Ecology and E olu ion published by John Wiley & Sons L d. 4417 A. V. S onen e al.Genome-wide Analyses o Selec ion in Wol es gene ic a ia ion in wol es h oughou ou s udy a ea. Concu en selec ion o ubiqui ous ai s in sepa a e popula ions may ha e been di e gen o pa allel. How- e e , o ai s such as hea ing and ision a ajec o y o pa allel selec ion appea s mos likely. Ou esul s none heless sugges local adap a ion may play a ole. Al hough he wol is a highly mobile species, i has been epo ed o exhibi popula ion s uc u e co e- sponding wi h en i onmen al he e ogenei y in Eu ope (Pilo e al. 2006, 2012) and No h Ame ica (Ge en e al. 2004; Musiani e al. 2007; Mu~ noz-Fuen es e al. 2009; S onen e al. 2014). Ou indings indica e ha a iables such as empe a u e and habi a may in luence local adap a ion, which appea s consis en wi h ea lie esul s om he s udy a ea (Pilo e al. 2006). A SNP lanking wo genes epo ed o in luence empe a u e egula ion (TRPV1/TRPV3) was associa ed wi h July empe a u e. Because wol es a e long-dis ance pu suing (as opposed o ambush) p eda o s, physiological mechanisms o p e en o e hea ing could ep esen impo an selec i e ac o s. The possibili y o local en i onmen al selec ion o em- pe a u e egula ion me i s u he in es iga ion, pa icu- la ly in ligh o wa ming ea h su ace empe a u es and changes in he deg ee o a iabili y o empe a u e and o he clima ic ac o s. Local adap a ion can occu i indi iduals a e mo e likely o su i e and ep oduce wi hin hei na al habi a s (Da is and S amps 2004; Nosil e al. 2005; Edelaa e al. 2008), which subsequen ly a ec s popula ion gene ic s uc u e. The wol popula ion clus e s examined in his s udy (S o- nen e al. 2013) a e exposed o ma kedly di e en clima ic ac o s such as empe a u e and p ecipi a ion. No hcen al Eu opean and Ca pa hian wol es a e no usually subjec o e y ho wea he bu expe ience cold (including subze o) empe a u es ex ensi e pa s o he yea , whe eas he oppo- si e is ypically ue o he Balkan-Dina ic wol es o sou h- e n Eu ope. Uk ainian S eppe wol es, in con as , may expe ience bo h ho summe s and cold win e s. Al hough specula i e, he capaci y o empe a u e egula ion migh play a pa icula ly impo an ole o wol es in he s eppe. The no hcen al and Ca pa hian en i onmen s ha e much in common wi h ega d o clima e. The di e ences in day leng h be ween he wo a eas likely in luence o he p o- cesses o ecological impo ance such as plan pho ope iods, and di e ences in day leng h ha e been epo ed o a ec he beha io o A c ic mammals such as S alba d eindee (R. . pla y hynchus) ( an Oo e al. 2005). The clea s uc- u ing seen be ween Ca pa hian and no hcen al Eu opean wol es, which e lec s he di ision in o wo majo phyloge- ne ic clades o wol es (Pilo e al. 2010; Cza nomska e al. 2013), could, a leas in pa , also be caused by habi a ag- men a ion and human landscape de elopmen (Huck e al. 2011). The di e gen p o iles o Uk ainian S eppe wol es may o some ex en be a esul o immig a ion om ou side he s udy a ea. The Uk ainian pa o ou s udy a ea could be ecei ing immig an s om he s eppe o o es - s eppe egions a he eas and no h, and simila immi- g a ion om eas e n and no he n egions may occu in he wes e n Russian pa o ou s udy a ea (Pilo 2005). F ST alues o 67-K loci we e lowes be ween no hcen al Eu ope and Uk ainian S eppe wol es (S onen e al. 2013). D i is expec ed o in luence he en i e genome and selec ion o ac only on ce ain loci, and he F ST al- ues o he 353 loci be ween no hcen al Eu ope and he Uk ainian S eppe sugges di e si ying selec ion migh play a ole in inc easing di e gence be ween wol es om hese egions. Spa io- empo al esolu ion o he selec i e o ces ha may ha e p oduced he cu en pa e ns is challeng- ing because o limi ed a ailable da a om he eas e n pa o ou s udy a ea and beyond – o ou s udy and in gen- e al. P ey and habi a ha e been epo ed as impo an a i- ables in ea lie in es iga ions wi h (p esumed) neu al ma ke s (Ge en e al. 2004; Musiani e al. 2007; Mu~ noz- Fuen es e al. 2009; S onen e al. 2014). This includes indings om ou s udy a ea (Pilo e al. 2006, 2012). Impo an ly, nei he ecosys em no biome may be he app op ia e scale a which o examine he local pa e ns o selec ion in species such as wol es; ecosys em may be oo na ow, whe eas biome could be oo b oad. O he ea u es o he local en i onmen , such as he size and beha io o a ailable p ey, may be mo e in o ma i e o elucida ing he pa e ns o selec ion (Benson e al. 2012; Monzon e al. 2014). We did no ha e p ey da a o ou s udy a ea, bu ea lie in es iga ions wi hin Eu ope (Je zd zejewski e al. 2012; Pilo e al. 2012, 2014a) acco d wi h new da a om No h Ame ica (Benson e al. 2012; Monzon e al. 2014) in sugges ing ha he in luence o die me i s u he a en ion. None o he BayeScan and GWAS candida e loci o e - lapped, al hough a numbe o SAM esul s we e suppo ed by ou lie de ec ion as well as gene–en i onmen associa- ions. Ea lie s udies ha e epo ed simila lack o o e lap be ween BayeScan and o he es s (e.g., Na um and Hess 2011). Se e al po en ially impo an d i e s o selec ion a e no included in ou s udy (e.g., die , disease, and pa asi es), which migh help explain why a numbe o ou lie loci in BayeScan we e no iden i ied in gene–en i onmen es s. Howe e , we would expec loci de ec ed by en i onmen al selec ion o be iden i ied by BayeScan, and i is unce ain why his did no occu . Possibly, me hodical di e ences may play a ole. Fo example, Bayesian me hods imple- men ed in BayeScan di e om ha o signi icance es ing in classical s a is ics (Foll 2012). Ano he ac o could be ou small sample sizes om some popula ion clus e s. We 4418 ª2015 The Au ho s. Ecology and E olu ion published by John Wiley & Sons L d. Genome-wide Analyses o Selec ion in Wol es A. V. S onen e al. Wabakken, P., H. Sand, I. Kojola, B. Zimme mann, J. M. A nemo, H. C. Pede sen, e al. 2007. Mul is age, long- ange na al dispe sal by a global posi ioning sys em-colla ed Scandina ian wol . J. Wildl. Manage. 71:1631–1634. Weckwo h, B. V., S. Talbo , G. K. Sage, D. K. Pe son, and J. Cook. 2005. A signal o independen coas al and con inen al his o ies among No h Ame ican wol es. Mol. Ecol. 14:917–931. Weckwo h, B. V., S. L. Talbo , and J. A. Cook. 2010. Phylogeog aphy o wol es (Canis lupus) in he Paci ic No hwes . J. Mammal. 91:363–375. Weckwo h, B. V., N. G. Dawson, S. L. Talbo , M. J. Flamme, and J. A. Cook. 2011. Going coas al: sha ed e olu iona y his o y be ween coas al B i ish Columbia and Sou heas Alaska wol es (Canis lupus). PLoS One 6:e19582. Suppo ing In o ma ion Addi ional Suppo ing In o ma ion may be ound in he online e sion o his a icle: Table S1. Co ela ion be ween en i onmen al a iables (de ailed in Table 1). Table S2. Comple e iden i ica ion o single nucleo ide polymo phism (SNP) loci on he Illumina CanineHD BeadChip (170K SNPs) wi h in o ma ion om he MAP- ile in PLINK. Table S3. Summa y o majo unc ional genes nea single nucleo ide polymo phism (SNP) loci iden i ied as ou lie loci and/o associa ed wi h en i onmen al a iables based on a s udy o 59 wol es in ou Eu opean popula ion clus e s. Table S4. Func ional genes whe e geno ype equencies show spa ial pa e ns be ween popula ion clus e s (e.g., No hcen al agains he o he h ee, o No hcen al and Uk ainian S eppe agains o he s). Table S5. Func ional genes wi hou ob ious spa ial pa - e ns. Table S6. SNP loci iden i ied as ou lie s by BayeScan bu no associa ed wi h en i onmen al a iables included in his s udy. Table S7. Pai wise F ST - alues wi h 95% con idence in e - als o n=59 wol es in ou popula ion clus e , ac oss n=353 SNP loci epo ed as ou lie s (BayeScan) o asso- cia ed wi h en i onmen al a iables (GWAS in PLINK), calcula ed in Hie Fs a wi h boo s ap esampling (n=1000). ª2015 The Au ho s. Ecology and E olu ion published by John Wiley & Sons L d. 4425 A. V. S onen e al.Genome-wide Analyses o Selec ion in Wol es