Commonalities in Development of Pure Breeds and Population ...Sardinia experiences both isolating...

HIGHLIGHTED ARTICLE| INVESTIGATION

Commonalities in Development of Pure Breeds andPopulation Isolates Revealed in the Genome of the

Sardinian Fonnirsquos DogDayna L Dreger Brian W Davis Raffaella Coccodagger Sara Sechidagger Alessandro Di CerboDagger Heidi G Parker

Michele Pollisect Stefano P Marellisect Paola Crepaldisect and Elaine A Ostrander1

Cancer Genetics and Comparative Genomics Branch National Human Genome Research Institute National Institutes of HealthBethesda Maryland 20892 daggerDipartimento di Medicina Veterinaria Sezione Clinica Medica Universitagrave degli Studi di Sassari 07100

Italy DaggerDipartimento di Scienze Mediche Orali e Biotecnologiche Specializzazione Biochimica Clinica Universitagrave degli Studi GdrsquoAnnunzio Chieti Pescara 66100 Italy and sectDipartimento di Medicina Veterinaria Universitagrave degli Studi di Milano 20133 Italy

ABSTRACT The island inhabitants of Sardinia have long been a focus for studies of complex human traits due to their unique ancestralbackground and population isolation reflecting geographic and cultural restriction Population isolates share decreased genomic diversityincreased linkage disequilibrium and increased inbreeding coefficients In many regions dogs and humans have been exposed to the samenatural and artificial forces of environment growth and migration Distinct dog breeds have arisen through human-driven selection ofcharacteristics to meet an ideal standard of appearance and function The Fonnirsquos Dog an endemic dog population on Sardinia has not beensubjected to an intensive system of artificial selection but rather has developed alongside the human population of Sardinia influenced bygeographic isolation and unregulated selection based on its environmental adaptation and aptitude for owner-desired behaviors Through analysisof 28 dog breeds represented with whole-genome sequences from 13 dogs and 170000 genome-wide single nucleotide variants from155 dogs we have produced a genomic illustration of the Fonnirsquos Dog Genomic patterns confirm within-breed similarity while population anddemographic analyses provide spatial identity of Fonnirsquos Dog to other Mediterranean breeds Investigation of admixture and fixation indices revealsinsights into the involvement of Fonnirsquos Dogs in breed development throughout theMediterraneanWe describe how characteristics of populationisolates are reflected in dog breeds that have undergone artificial selection and are mirrored in the Fonnirsquos Dog through traditional isolating factorsthat affect human populations Lastly we show that the genetic history of Fonnirsquos Dog parallels demographic events in local human populations

KEYWORDS dog whole-genome sequence demography population structure

THE domestic dog has undergone intensive human-drivenselective pressure in an effort to develop canines of a

particular appearance or with a desired behavior pattern(Galibert et al 2011 Wayne and vonHoldt 2012) While con-siderable effort has gone into decoding the genetic basis ofmorphologic traits that vary within dog populations such asbody size or coat color the challenge of identifying genomicfeatures that distinguish breeds selected to support humansurvival is largely unmet (summarized in Boyko 2011

Rimbault and Ostrander 2012 Schoenebeck and Ostrander2014) Breeds specifically suited to perform such tasks includethose that excel at hunting herding and protection of prop-erty An appeal to aesthetics and functional design has alsoresulted in selection of dogs with specific and uniform mor-phology Since much of modern breed formation took place100ndash300 years ago (Fogel 1995) restrictive geography andburgeoning industry have played a part in influencing the dis-tribution of genetic variation among isolated populations

From a genetic viewpoint dog breeds are defined as pop-ulations with human-induced or geographic barriers to geneflow and fixed traits that will reliably reproduce selectedcharacteristics throughmultiple generations There are severalorganizations thatrecognizeandregisterdogbreedsOneofthelargest both in terms of breed recognition and geographicreach is the Federation Cynologique Internationale (FCI) an

Copyright copy 2016 by the Genetics Society of Americadoi 101534genetics116192427Manuscript received June 8 2016 accepted for publication August 10 2016published Early Online August 12 2016Available freely online through the author-supported open access option1Corresponding author National Human Genome Research Institute National Institutesof Health Bldg 50 Room 5351 50 South Dr Bethesda MD 20892 E-mail eostrandmailnihgov

Genetics Vol 204 737ndash755 October 2016 737

international organization encompassing91member countriesandrecognizing343breeds(httpwwwfcibeenPresentation-of-our-organisation-4html) The FCI however does not neces-sarily recognize regional varieties of dogs that exist worldwidewhich are often acknowledged at only a local or niche levelMany of these dog populations have often undergone selectionlargely for functional purposes and behaviors that are dis-tinctly important to the humans in the region and as suchthey represent isolated and unique genetic histories

The studyofhuman islandpopulationshasbeenofconsider-able interest to geneticists as such populations often exhibitdecreased genomic diversity (Sajantila et al 1996 Capocasaet al 2014 Di Gaetano et al 2014) an increase in linkagedisequilibrium (LD) (Bendjilali et al 2014) and increased in-breeding coefficients (Zhai et al 2016) These same popula-tion metrics have proven informative for the study of complextraits and heritable diseases particularly in the context ofrelatively young populations that have experienced rapidgrowth share a common environment and cultural practicesand frequently maintain detailed genealogical records(reviewed in Peltonen et al 2000) Populations can becomegenetically isolated through reproductive restraints imposedby cultural or religious practices language or industry Inthese instances societal restraints only influence the humangene pool rather than other cohabitant organisms Geo-graphic isolation however is an additional barrier that canaffect both the human inhabitants of a region as well as theendemic or introduced flora and fauna The Italian island ofSardinia experiences both isolating mechanisms providing aunique opportunity to investigate complex traits in a humanpopulation with low genetic diversity (Capocasa et al 2014 DiGaetano et al 2014) high frequency of private variants (Sidoreet al 2015) and unique genomic ancestry (Fiorito et al 2015)

The country of Italy is the site of origin for several in-ternationally recognized modern breeds such as the CaneCorso Maltese Neapolitan Mastiff Spinone Italiano BraccoItaliano and Cirneco dellrsquoEtna Italy also stakes claim to agreat number of regional varieties including the Fonnirsquos Dogand the Mastino Abruzzese both livestock guardians theCane Paratore a herding breed and the Levriero Meridio-nale a sighthound These breeds primarily exist as isolatessurrounding their historic regions of origin not havinggained popularity outside local breed enthusiasts

Among the most unique of the Italian varieties is the FonnirsquosDog also known as the Cane Fonnese Pastore Fonnese orSardinian Shepherd Dog It is a large livestock and propertyguardian breed originating from the region surrounding the cityof Fonni Present in rough- and smooth-coated varieties and invarying colors the unifying features of the breed include a char-acteristically intense facial expression and instinctive propensitytoward guarding behaviors andwariness of strangers Historicalaccounts portray dogs fitting this description residing in Fonniand the surrounding regions in the mid- to late-19th century(Tyndale 1849 Bresciani 1850 Cetti 1774 Edwardes 1889)

International breed recognition of the Fonnirsquos Dog is beingpursued by a dedicated group of breeders and enthusiasts

(httpwwwcanefonneseit) with a goal of preservingthe distinct heritage of this remarkable breed Preliminarystudies (Sechi et al 2016) have characterized the morpho-logical commonalities of the Fonnirsquos Dog showing that theyare consistent with features of a true-breeding populationStudies using a limited number of microsatellite markers sup-port this hypothesis (Sechi et al 2016)

In thisstudywehaveusedwhole-genomesequence(WGS)atan average coverage of 423 over 13 canids together with SNPanalysis of 173662 variants across 155 canids to characterizethe genomic architecture of Fonnirsquos Dogs in the context of geo-graphically and historically proximate dog breeds (Figure 1)We investigate the similarities between human population iso-lates and dog breeds and subsequently breed developmentwhen driven by human-defined regulation of selection or whenpredominantly based on behavioral aptitude for guardian func-tions in the absence of aesthetic preferenceWedemonstrate thepropensity for Fonnirsquos Dogs to display genomic characteristicsequivalent to those of established and acknowledged breedsproviding a dynamic account of how geographic isolation andbehavior-driven selection function to produce unique breedpopulations Finally we reveal parallelisms between the breedfoundation of the Fonnirsquos Dog and the human population de-mographics of Sardinia exposing an adroit consideration forinvestigations of population structure and the breadth of re-search application for dog breeds as population isolates

Materials and Methods

Sample acquisition and genotyping

Blood samples were obtained and sent to the National HumanGenomeResearch Institute (NHGRI) from six unrelated FonnirsquosDogs one each from the Sardinian regions of Fonni OzieriCagliari Sassari Porto Torres and Nuoro (Figure 1) Remain-ing dog breed samples were submitted by owners with a signedconsent in accordance with the NHGRI Animal Care and UseCommittee Genomic DNA was extracted from all blood sam-ples using a standard phenol chloroform method aliquotedand stored at 280 (Sambrook et al 1989) Extractions ofDNA from saliva collections were performed using the recom-mended Performagene protocol (DNA Genotek Ottawa ON)

A set of 135 dogs representing 19 breeds (Table 1) weregenotyped using the Illumina (San Diego CA) CanineHD SNPchip which has 173662 potentially informative markers Ge-notype callingwas conducted in IlluminaGenomeStudiowitha protocol specifying a 90 call rate and GenTrain scoreof04 Illumina Canine HD SNP genotypes for eight Maltesetwo Komondors two Spinone Italiano and eight Bouvier desFlandres were obtained from a previous publication (Haywardet al 2016) We thus analyzed data from 155 SNP-genotypeddogs representing 23 breeds

WGS alignment

Publically available WGS data were obtained from the Se-quence Read Archive (httpwwwncbinlmnihgovsra) for

738 D L Dreger et al

a total of 12 dogs representing 12 breeds (Table 1) WGSfrom one Fonnirsquos Dog was produced for this study by theNational Institutes of Health Intramural Sequencing Center

using the Illumina TruSeq DNA PCR-Free Protocol (Cat FC-121-3001) Data were aligned to the CanFam 31 reference genome(httpgenomeucsceducgi-binhgGatewaydb=canFam3)

Figure 1 Dog breeds used in molecular analyses (A) Geographic representation of the region of origin for the 23 Mediterranean breeds The (B)smooth-coated and (C) rough-coated varieties are both recognized in Fonnirsquos Dogs Breed abbreviations are listed in Table 1

Fonnirsquos Dog Mirrors Population Isolates 739

using Burrows-Wheeler Aligner 0710 MEM (Li and Durbin2009) SAMtools 0110 (Li et al 2009) was used for sortingand PicardTools 1119 (httpsgithubcombroadinstitutepicard) for screening for putative PCR duplicate reads Localrealignment was conducted with Genome Analysis Toolkit(GATK) 32-2 (DePristo et al 2011) and based on docu-mented and novel indels (Axelsson et al 2013) Training setsof dbSNP and Illumina CanineHD chip positions were usedfor base quality recalibration Single nucleotide variants werecalled with gVCF mode of HaplotypeCaller (Van der Auweraet al 2013) for each individual dog and again jointly acrossall sequenced dogs Variant quality score recalibration was con-ducted with GATK best practices and default parameters andthe initial alignment training sets Resultant jointly called vari-ant call formats (VCFs) were filtered for CpG islands gaps andrepeats based on CanFam 31 reference genome annotations(httpgenomeucsceducgi-binhgGatewaydb=canFam3)

The previously unpublished WGS from the Fonnirsquos Dog hasbeen uploaded to the Sequence Read Archive (httpwwwncbinlmnihgovsra) Variants were annotated with SNPeff(Cingolani et al 2012) based on CanFam 3176 (Lindblad-Toh et al 2005)

Genomic characterization of breeds

Inbreeding coefficients and homozygosity were calculatedfrom the SNP data using the ldquohetrdquo and ldquohomozygrdquo functionsof PLINK v107 (Purcell et al 2007) respectively Genome-wide length of homozygosity was measured using slidingwindows of 5 Mb allowing for one heterozygous call and amaximum of five missing genotypes per window Individualregions of homozygosity were identified from SNP genotypesusing PLINK v107 software and the homozyg function andrequired a minimum of 10 sequential SNPs with a heterozy-gous allowance of 1 Haplotypes were analyzed from regionsthat were shared by four of the six Fonnirsquos Dogs and smallerhomozygous regions of three or more sequential SNPs thatwere shared by all six Fonnirsquos Dogs were identified

Principle components analysis (PCA) was calculated withEigensoft v601(Patterson et al 2006 Price et al 2006) andPCA plots were drawn using Partek Genomics Suite v66(Partek St Louis MO) FastSTRUCTURE v10 (Raj et al2014) was used to identify population structure from SNPdata Neighbor-joining phylogeny was constructed with Phy-lip v3696 (Felsenstein 1989) using a distance matrix fromPLINK v107 (Purcell et al 2007) with 100 bootstrappedrepetitions and dendograms of the consensus tree weredrawn in FigTree v142 (Rambaut 2014) VCFtools v0114(Danecek et al 2011) was used to calculate Weir andCockerhamrsquos FST for every breed-to-breed pair Effectivepopulation size (Ne) of each breed with greater than four dogswas estimated through SNP-based LD analysis with SNeP(Barbato et al 2015) Private variants were called for eachindividual dog relative to the pool of 13 WGSs representing13 breeds using the ldquosingletonsrdquo filter of VCFtools v0114(Danecek et al 2011) These variants will be referred to asldquoindividualrdquo variants throughout to distinguish them asunique to single dogs though not necessarily indicative ofbreed-specific private variants

Admixture for every breed-to-breed combination wasassessed using the ldquothree_poprdquo function of TreeMix v112(Pickrell and Pritchard 2012 Pickrell et al 2012) and Admix-Tools v30 (Alexander et al 2009) using the SNP data andAnalysis of Next Generation Sequencing Data (ANGSD)v0911 (Korneliussen et al 2014) using the WGS data Zscores obtained from Pattersonrsquos D and f3 for introgressionsinvolving Fonnirsquos Dog were compared across computationalalgorithms and admixing events suggested by two or moreprograms were identified for further consideration A criticalZ value of |Z| = 3 was used to determine significance TheSNP data from the Mediterranean breeds was analyzed inTreeMix v112 to produce phylogenies with the most likelyintrogression events allowing for 1 through 10 15 and25 possible migrations

Table 1 Type and number of dogs used in analyses

Breed Abbreviation Processa No WGS source

Anatolian Shepherd ANAT SNP 6Azawakh Hound AZWK SNP 5Berger Picard BPIC WGS 1 SRR2016171Bouvier des Flandres BOUV SNP 8Cane Corso CANE WGS 1 SRR2747522

SNP 9Cane Paratore CPAT SNP 2Cirneco dellrsquoEtna CIRN SNP 5Fonnirsquos Dog FONN WGS 1 PRJNA318762

SNP 6Great Pyrenees GPYR WGS 1 SAMN03801670

SNP 10Ibizan Hound IBIZ SNP 10Istrian Shorthaired

HoundISHH WGS 1 SAMN02485584

Italian Greyhound ITGY WGS 1 SAMN03801673SNP 10

Komondor KOMO SNP 2Lagotto Romagnolo LAGO WGS 1 PRJNA318762Levriero Meridionale LVMD SNP 2Maltese MALT SNP 10Mastino Abruzzese MAAB SNP 2Neapolitan Mastiff NEAP SNP 6Pharaoh Hound PHAR SNP 2Portuguese Water

DogPTWD WGS 1 PRJNA318762

SNP 10Saluki SALU WGS 1 SAMN03801686

SNP 19Sloughi SLOU WGS 1 PRJNA318762

SNP 5Spanish Galgo GALG WGS 1 SAMN03168380Spanish Water Dog SPWD WGS 1 PRJEB7903Spinone Italiano SPIN SNP 2Standard Schnauzer SSNZ SNP 10Saint Bernard STBD WGS 1 PRJNA263947

SNP 10Volpino Italiano VPIN SNP 4

SRA accession numbers are included for all WGSa This column indicates individuals genotyped on the Illumina HD Canine SNP array(SNP) or WGS


Data availability

WGS was produced for this study from a single Fonnirsquos Dogsubmitted to the Sequence Read Archive (PRJNA318762) forrelease upon manuscript acceptance Previously publishedWGS obtained from the Sequence Read Archive is listed inTable 1 with corresponding accession numbers Genotypedata from the Illumina CanineHD array has been submittedto Gene Expression Omnibus (GSE83160)

Results

A representative population of dog breeds originatingin the Mediterranean

A selection of 28 dog breeds originating from regions sur-rounding the Mediterranean Sea was chosen for genomicpopulation analysis (Figure 1) Numbers of dogs per breedranged from1 (Istrian ShorthairedHound LagottoRomagnoloSpanishWater Dog SpanishGalgo) to 19 (Saluki) The FonnirsquosDog Cane Paratore Mastino Abruzzese Volpino ItalianoLevriero Meridionale and Cirneco dellrsquoEtna were sampledfrom Italy the Sloughi were collected from North Africaand the Azawakh fromMali The remaining breeds were pre-dominantly collected from the United States and are largelyAmerican Kennel Club-recognized breeds The entirety ofthese breeds is referred to herein as the ldquoMediterraneangrouprdquo while a subset consisting of those breeds with writ-ten or assumed history tracing to Italy (Cane Corso CaneParatore Cirneco dellrsquoEtna Fonnirsquos Dog Levriero Meridio-nale Maltese Mastino Abruzzese Neapolitan Mastiff Pha-raoh Hound Spinone Italiano and Volpino Italiano) (httpwwwenciitlibro-genealogicorazze-italiane httpwwwbordercolliemuseumorgBCCousinsEuropeWesternItalyhtml) are also classified specifically as the ldquoItalian grouprdquoWhile tracing its origin to Hungary previous analyses (un-published data) have repeatedly shown shared genetic iden-tity between the Komondor breed and those in the Italiangroup For the purposes of this study the Komondor has beenincluded in the Italian group All of the above are termedldquobreedsrdquo for the purposes of this study Note however thata subset of breeds (Mastino Abruzzese Cane Paratore andLevriero Meridionale) is not officially recognized by any for-mal registry while the remainder of the breeds have receivednational or international recognition

SNP-based measures of homozygosity forMediterranean breeds

Since decreased levels of genetic diversity can be indicative ofselective breeding founder effects or population bottlenecksand therefore breed formation molecular analyses includedmeasures of homozygosity and inbreeding coefficients for theMediterranean breeds The number of homozygous regionsfrom the Mediterranean breeds (Figure 2 Table 2) witha minimum length of 1 Mb ranged from a breed mean of12 (Mastino Abruzzese) to 114 (Saint Bernard) with a meanof 67 across all breeds The total length of homozygosity

calculated as the sum of the lengths for each of the previouslydetermined regions of homozygosity per dog and averagedacross each breed ranged from a breed mean of 1018 Mb(Mastino Abruzzese) to 7471 Mb (Pharaoh Hound) with amean of 4669 Mb across all dogs The Fonnirsquos Dog had abreed range of 8ndash49 homozygous regions and 596ndash7606Mb total homozygosity with within-breed means of 26 and3487 Mb respectively (Table 2)

Individual inbreeding coefficients had a maximumwithin-breed mean of 039 (Pharaoh Hound) and a minimum of003 (Mastino Abruzzese) with a mean of 022 across allbreeds (Figure 2 Table 2) The Fonnirsquos Dog ranged from 001to 030 and had a breed mean inbreeding coefficient of 013Thus the Fonnirsquos Dog presented homozygosity and inbreed-ing values within the range of equivalent values expressed byother Mediterranean dog breeds

Estimated Ne was calculated for 15 breeds for which fouror more dogs had SNP genotypes over a timeframe of 13ndash995 previous generations (Figure 3) The most recent Ne

values at 13 generations ago ranged from 71 (NeapolitanMastiff) to 303 (Saluki) with an across-breed mean of 117The Fonnirsquos Dog has a 13 generation Ne of 113 Each breedpresented a rapidly decreasing Ne with a breed-specific meanDNe ranging from 81 (Saint Bernard) to 176 (Azawahk) TheFonnirsquos Dog displayed a comparable mean DNe of 153

Individual and expected breed range of individualvariation in WGS

Wenext sought to compare the level ofunique variationacrossindividuals representative of theMediterraneandog breeds toassess the distinctiveness of each breed A substantially de-flated value of individual variation relative to the levels incomparablebreedswould imply increased similarity to oneormore of the other breeds Individual variants were identifiedfor eachdog relative to apool of 13Mediterranean-breeddogs(Table 1) The resulting variants were classified as either theheterozygous or homozygous state by VCFtools Variantsfound in the homozygous state in a single dog are consideredrepresentative of breed-specific variants with respect to the13 Mediterranean breeds However variants identified inthe heterozygous state in a single dog may represent breed-specific though not breed-fixed variants or variants uniqueto the individual The number of homozygous individualvariants ranged from 6776 (Spanish Water Dog) to 26623(Saluki) (Figure 2) The Saluki had the highest number oftotal individual variants (151426) and the Istrian Short-haired Hound had the lowest number (59838) The across-breed mean for all individual variants was 82056 with amean of 15203 for the homozygous variants (Table 3)The Fonnirsquos Dog had 86166 individual variants with 9224present in the homozygous state

Underlying population structure ofMediterranean breeds

Population structure was calculated using the SNP data over arangeof2 through20possiblegroupingsusingallMediterranean


Figure 2 Quantitation of genetic measures of (A) homozygosity and inbreeding from SNP-chip data representing within-breed means and (B) individualvariation from WGS of individual dogs


breeds and 2 through 12 groupings for the Italian breedsMaximum likelihood analyses identifiedK=15 (marginal likeli-hood = 2100 variance = 123 3 1024 DK14ndash15 = 2175 31023DK15ndash16=19931023) andK=3 (marginal likelihood=2105 variance = 186 3 1025 DK2ndash3 = 2382 3 1024DK3ndash4= 4853 1024) as the statistically appropriate number ofstructural groupings for Mediterranean and Italian breeds re-spectively (Figure 4) This analysis grouped the Mediterraneanbreeds as well as the Portuguese Water Dog Bouvier des Flan-dres Neapolitan Mastiff Great Pyrenees Ibizan Hound ItalianGreyhoundMaltese Saint Bernard and Standard Schnauzer asdistinct breed clusters The Saluki primarily formed a singlegrouping (orange in Figure 4) with some individuals demon-strating variable levels of a commonmulti-breed signature (darkpurple) presumably related to population substructure resultingfrom the region of sample collection The PharaohHound shared59 identitywith the IbizanHound (green) and8ndash12 identitywith each of Neapolitan Mastiff (dark blue) Portuguese WaterDog (pale blue) Italian Greyhound (pale purple) and SalukiThe common signature (dark purple) accounted for themajorityof the Volpino Italiano Komondor Spinone Italiano FonnirsquosDog Mastino Abruzzese Cane Paratore Levriero MeridionaleAzawakh and Sloughi The Neapolitan Mastiff structure repre-sented 4ndash7 of the Cane Paratore 2ndash5of theMastino Abruzz-ese up to 7 of the Fonnirsquos Dog and 2 of the SpinoneItaliano Komondor and Volpino Italiano The common multi-breed signature appeared in individual dogs of otherwise distinctbreeds at a rate of 13 (Standard Schnauzer) to 30 (MalteseandGreat Pyrenees)While the Anatolian Shepherd and Sloughiwere comprised of the Saluki (orange) signature at a level of

8ndash40 the remaining identity was that of the common multi-breed cluster

Within the Italianbreeds (K=3) theMaltese andNeapolitanMastiff identify as single clusters with the Cirneco dellrsquoEtnaCane Paratore Fonnirsquos Dog Komondor Levriero MeridionaleMastino Abruzzese Volpino Italiano Pharaoh Hound and Spi-none Italiano combined to produce the third cluster The CaneCorso was comprised of 27ndash73 of the Neapolitan Mastiff sig-nature with the remainder being the multi-breed cluster Atotal of 2 of the 10 Maltese demonstrated 7 or 27 identitywith the multi-breed grouping

Interbreed genome-wide FST values range from 00167(Fonnirsquos Dog 3 Mastino Abruzzese) to a maximum of02942 (Pharaoh Hound 3 Neapolitan Mastiff) (Figure 5)Values closer to zero highlight populations that are compar-atively more genetically homogeneous while values closer toone reflect populations with greater genetic divergence Inthis analysis the Fonnirsquos Dog is less diverged from the Mas-tino Abruzzese (FST = 00167) and the Cane Paratore (FST =00284) compared to the Saint Bernard (FST = 01669) andthe NeapolitanMastiff (FST = 01591) The distance betweenthe Fonnirsquos Dog and the NeapolitanMastiff is unexpected dueto the Italian heritage of both breeds

Genetic breed variation visualized through PCAand phylogeny

PCAwas conducted using two separate breed groupings theMediterranean and Italian groups Analysis of the first fourprincipal components (PCs) (PC1 = 554 P= 1333 10246PC2 = 489 P = 568 3 10242 PC3 = 470 P = 568 3

Table 2 Inbreeding and homozygosity metrics from SNP-chip analyses sorted by breed mean length of homozygosity

Breed No

Inbreeding coefficient Regions of homozygosity Length of homozygosity (Mb)

Mean Min Max Mean Min Max Mean Min Max

MAAB 2 003 002 005 1250 10 15 10179 6819 13538ANAT 6 008 005 015 1417 7 26 12400 38746 30120VPIN 4 006 001 016 2200 17 30 19443 3944 41317SLOU 5 012 002 019 3300 6 61 21391 1629 37194CPAT 2 009 003 015 3050 20 41 24339 7246 41432AZWK 5 018 013 036 3640 26 52 26891 9120 64437SALU 19 018 002 037 4532 4 83 31690 924 72928FONN 6 013 001 030 2617 8 49 34871 5955 76059MALT 10 020 010 034 6950 38 96 39753 20898 74885LVMD 2 023 019 027 6600 54 78 43595 41733 45456SPIN 2 022 020 024 5650 51 62 43680 38150 49211CANE 9 016 002 042 5067 16 75 44381 3200 107738CIRN 5 020 008 038 5360 34 85 45172 13327 90652KOMO 2 020 016 025 5550 53 58 46907 34357 59458SSNZ 10 025 018 039 8580 75 102 54317 37234 81958BOUV 8 026 014 047 7675 57 89 55460 25760 100320IBIZ 10 029 017 041 8900 81 103 55921 30230 85709ITGY 10 029 020 042 9110 69 104 60767 41160 88478PTWD 10 027 019 041 6930 57 91 61354 41786 92515STBD 10 029 023 033 11390 100 124 63682 48111 72925NEAP 6 030 026 036 10567 98 119 63751 50746 81894GPYR 10 035 019 050 10000 75 114 73575 34063 109651PHAR 2 035 034 035 10750 105 110 74711 72940 76481

Min minimum max maximum


10251 PC4 = 449 P= 2783 10262) of the Mediterraneanbreeds (Figure 6 A and B) showed that the Saint BernardItalian Greyhound and Great Pyrenees cluster as distinctbreeds independent of the others The Neapolitan Mastiffand Cane Corso identify as separate tightly clustered group-ings as did the Standard Schnauzer and Portuguese WaterDog The remaining Mediterranean breeds appeared to formloose breed clusters that were not readily distinguishablefrom one another The Italian group subset improved resolu-tion between those breeds (PC1 = 384 P = 241 3 10227PC2 = 271 P = 140 3 10223 PC3 = 203 P = 183 310213 PC4 = 165 P = 188 3 10205) (Figure 6 C andD) The Cane Corso and NeapolitanMastiff continued to formdistinct clusters in close proximity to each other The Malteseformed a clear cluster and the Pharaoh Hound and CirnecodellrsquoEtna grouped togetherHowever theCaneParatore FonnirsquosDog Komondor Levriero Meridionale Mastino AbruzzeseSpinone Italiano and Volpino Italiano did not resolve intodistinct breed clusters Further reduction of these regionalbreeds (Figure 6 E and F) resolved PC1ndash3 into breed-specificclusters though without statistically significant separation(PC1 = 151 P = 0510 PC2 = 146 P = 0319 PC3 =140 P=0185) The Spinone Italiano and LevrieroMeridionalewere most divergent from the others The Mastino AbruzzeseCane Paratore and Komondor breeds were distinct yet closelyrelated groups The Volpino Italiano and Fonnirsquos Dog while alsodistinct groupings were more diffuse

Similar phylogenetic relationships were also observedamong the same dogs (Figure 7) As with the PCA the clad-ograms indicated that the Cane Corso and Neapolitan Mas-tiff and the Standard Schnauzer and Portuguese Water Dogare closely related However the common structural group-

ing visualized using analysis with the STRUCTURE program(purple in Figure 4) encompassing the Cane Paratore Mas-tino Abruzzese Fonnirsquos Dog Spinone Italiano KomondorVolpino Italiano Saluki Anatolian Shepherd Sloughi Aza-wakh LevrieroMeridionale and Cirneco dellrsquoEtna separatedin the cladograms in a manner that best reflected the mor-phological or ancestral commonalities between breeds Forinstance the Cirneco dellrsquoEtna Pharaoh Hound and IbizanHound and the Sloughi Levriero Meridionale and Azawakhformed two monophyletic groups in putative sighthoundclades The Fonnirsquos Dog Mastino Abruzzese Cane Paratoreand Volpino Italiano were monophyletic and branched im-mediately outside of the PortugueseWater Dog and StandardSchnauzer breeds that reflect greater phenotypic similaritywith each other than with either the sighthounds or the mas-tiffs (Neapolitan Mastiff Cane Corso and Saint Bernard)

Admixture between Fonnirsquos Dogs andMediterranean breeds

We next focused on the Fonnirsquos Dog in terms of its historicalinvolvement in breed development throughout the Mediter-ranean Breed admixture with the Fonnirsquos Dog was calculatedusing three separate algorithms (Figure 8) Introgression ofthe Fonnirsquos Dog Portuguese Water Dog and Cane Paratorewas identified within the SNP data by the TreeMix three_popand AdmixTools software programs TreeMix predicted a1743 contribution of Fonnirsquos Dog to the ancestor of thePortuguese Water Dog and Cane Paratore (P = 000723)Pattersonrsquos D from AdmixTools replicated these findings forFonnirsquos Dog and PortugueseWater DogwithD(O PortugueseWater Dog Fonnirsquos Dog X) = 200139 to 201287Z=23737 to225274D(O Fonnirsquos Dog PortugueseWater

Figure 3 Ne estimates from SNP-basedcalculation of LD


Dog X) = 200127 to 201309 Z = 23078 to 225797and D(O Fonnirsquos Dog X PortugueseWater Dog) = 00100 to00767 Z = 3016 to 22189 Likewise AdmixTools reiter-ated the introgression of Fonnirsquos Dog and Cane Paratore withD(O Cane Paratore Fonnirsquos Dog X) =200104 to201317Z=23108 to226025D(O Fonnirsquos Dog Cane Paratore X) =200096 to201289Z=23055 to225805 andD(O FonnirsquosDog X Cane Paratore) = 00104ndash00811 Z = 3643ndash22974

A second instance of introgression was identified betweenFonnirsquos Dog and Komondor within the WGS dataset withANGSD-calculated significant D-statistics for D(O FonnirsquosDog Komondor Cane Corso) = 0039 Z = 4366 and D(O Komondor Fonnirsquos Dog Cane Corso) = 0031 Z =3920 and with nonsignificant values of D(O Cane CorsoKomondor Fonnirsquos Dog) = 0008 Z = 1028 This was con-firmed with AdmixTools which revealed a significant D fromSNP data D(O Fonnirsquos Dog Komondor X) = 200116 to201123 andD(O FonnirsquosDogX Komondor)=00116ndash01280Z= 3142ndash25720

Introgression between Fonnirsquos Dog and Saluki was alsoobserved using ANGSD D(O Fonnirsquos Dog Saluki Cane Corso) =0033 Z=3893D(O Saluki Fonnirsquos Dog Cane Corso) = 0040Z= 5075 and D(O Cane Corso Saluki Fonnirsquos Dog) =20007Z = 21060 These results were replicated with SNP data usingAdmixTools resulting in D-statistics of D(Saluki X Fonnirsquos DogY) = 0007ndash0131 Z= 3047ndash29352

Separately TreeMix predicted phylogeny trees for theMediterranean breeds with allowance for 1 through 10 15and25 introgressionevents The ln(likelihood) indicating theprobability of the suggested relationships explaining the ge-nomic data for the predicted phylogenies increased withadditional allowed introgressions from ln(likelihood)1 =159906 and ln(likelihood)25 = 188817 With the inclusionof 25 introgression events there was not yet any indication ofdecreasing ln(likelihood) and none of the proposed intro-gressions involved the Fonnirsquos Dog (Figure 8)

Regions of homozygosity represent putative regionsof selection

To identify regions of the genome potentially under selectionin the Fonnirsquos Dog we calculated SNP genotypes for each of

the six Fonnirsquos Dogs This revealed 258 total regions of ho-mozygosity (range per dog = 25ndash61) each of which spannedaminimum of 10 sequential SNP-chip variants Seven regionswere shared across four of the six dogs Analysis of SNP hap-lotypes across each shared region revealed 11 even shorterregions defined by three or more sequential SNPs that wereshared across all six Fonnirsquos Dogs (Table 4) Individual dogvariants extracted from the WGS sequence were filtered forthe 11 SNP-based regions of shared homozygosity resultingin 9 heterozygous and 1 homozygous variant within theselected regions Each of these variants was located in anoncoding region and annotated as a modifier by SNPeff(Cingolani et al 2012)

Discussion

Genetic investigation of population isolates can provide in-sights into inheritance of both rare and complex traits Inhumans for instance studies of Bedouin tribes have success-fully localized causal loci for single gene traits such as BardetndashBiedl Syndrome 3 (Farag and Teebi 1989 Sheffield et al1994 Chiang et al 2004) nonsyndromic hearing loss(Scott et al 1996) and infantile nephronophthisis (Haideret al 1998) Studies of the Finnish population by compari-son have been more useful for studies of complex traits re-vealing genes that increase susceptibility to various cancers(Nystroumlm-Lahti et al 1994 Kainu et al 2000 Sarantaus et al2000 Baffoe-Bonnie et al 2005 Hartikainen et al 2005Rokman et al 2005) Icelandic populations have also beenused successfully to map risk alleles associated with a varietyof complex traits and psychiatric conditions (Hicks et al2002 Thorgeirsson et al 2003 Karason et al 2005 Arasonet al 2010)

Studies of geographically or culturally restricted breedingin human populations have also informed our understandingof movement and growth of human populations and withthem genes of interest For instance characterization of pointmutations associated with b-thalassemia in Kurdistan Jewishpopulations have identified regional admixture and foundereffects specific to isolated Kurdish subpopulations (Rundet al 1991) Geographically remote populations can likewisereveal historical human migration or settlement events ex-emplified by the Pitcairn Island population which we nowknow has maternal Tahitian and paternal European lineages(Benton et al 2015) Within-country cultures that restrictmarriage outside the community have proven similarly infor-mative (Ginns et al 1998 Stone et al 1998 Hsueh et al2003 Seboun et al 2005 Simpson et al 2009 Hou et al2013 Georgi et al 2014 Kember et al 2015) The successof using population isolates for mapping of complex and raredisorders lies predominantly in the common features of suchisolates decreased genomic diversity (Sajantila et al 1996Capocasa et al 2014 Di Gaetano et al 2014) increased LD(Bendjilali et al 2014) and increased inbreeding values(Zhai et al 2016) Dog breeds reflect these same character-istics resulting in each breed forming a distinct population

Table 3 WGS individual variants for Mediterranean breed dogs

Dog Heterozygotes Homozygotes Total

Istrian Shorthaired Hound 49078 10760 59838Lagotto Romagnolo 53820 11733 65553Berger Picard 44080 25049 69129Italian Greyhound 53206 16827 70033Great Pyrenees 54289 22677 76966Standard Schnauzer 61347 16572 77919Spanish Galgo 61434 17335 78769Saint Bernard 65669 13884 79553Spanish Water Dog 75597 6776 82373Cane Corso 72221 12243 84464Portuguese Water Dog 76602 7939 84541Fonnirsquos Dog 76942 9224 86166Saluki 124803 26623 151426


Figure 4 STRUCTURE analysis of (A) Mediterranean and (B) Italian dog breeds Colors indicate separate STRUCTURE group signatures for individualdogs Maximum likelihood predicts groupings of K = 15 and K = 3 for Mediterranean and Italian breeds respectively


isolate formed through human intervention and selection(reviewed in Boyko 2011 Schoenebeck and Ostrander 2014)

Previous molecular analyses of the Fonnirsquos Dog of Sardiniaprovide strong evidence that this niche population meets the

formal definition of a breed (Sechi et al 2016) It appearsgenetically distinct from other breeds in the region as mea-sured by a small set of microsatellite-based polymorphicmarkers and both physical and behavioral traits are retained

Figure 5 Weir and Cockerhamrsquos weightedFST values based on breed-to-breed com-parisons of SNP genotypes (A) Scores of0 (yellow) indicate perfect identity scoresof 03 (blue) indicate the highest level ofdivergence observed in this set of breeds(B) Distribution of breed-specific FST values


from generation to generation (Sechi et al 2016) Histori-cally microsatellite-based studies have proven effective fordifferentiating between dog breeds (Parker et al 2004 Leroy

et al 2009 Mellanby et al 2013) identifying varieties thatinhabit a common insular region (Parra et al 2008Pribanova et al 2009 Suarez et al 2013) or identifying a

Figure 6 PCA of SNP-chip data for (A and B) Mediterranean breeds (PC1 = 554 P = 133 3 10246 PC2 = 489 P = 568 3 10242 PC3 = 470 P =568 3 10251 PC4 = 449 P = 278 3 10262) (C and D) all Italian breeds (PC1 = 384 P = 241 3 10227 PC2 = 271 P = 140 3 10223 PC3 = 203P = 1833 10213 PC4 = 165 P = 1883 10205) and (E and F) regional Italian breeds (PC1 = 151 P = 051 PC2 = 146 P = 032 PC3 = 140 P = 019PC4 = 140 P = 004) Breed clusters are denoted with ellipses of 2 SD Breed abbreviations are listed in Table 1


subset of the component breeds that contributed to individ-ual mixed breed dogs (Parker et al 2004) However sincemicrosatellites are multi-allelic and highly mutable they areuseful largely for family studies and not necessarily popula-

tion studies providing information that reflects decades andnot hundreds or thousands of years When indels and copynumber variants are also considered it is clear that caninegenetic studies are best served using dense SNP-chip analyses

Figure 7 SNP-based neighbor-joining cladograms for (A) Mediterranean and (B) Italian breeds Clade coloration reflects the designations assigned byFastSTRUCTURE (Figure 4) Branch numbers indicate the bootstrapping values from 100 repetitions the nine Cane Corso dogs did not form a singleclade however the bootstrap values for their phylogenetic locations relative to each other are each 100 Breed abbreviations are listed in Table 1


Figure 8 (A) Admixture predictions by three computational algorithms TreeMix three_pop analysis (green arrow) identified Fonnirsquos Dog contribution tothe Portuguese Water DogCane Paratore clade at 1743with P = 000723 ANGSD (red dashed line) calculated significant values of D(O Fonnirsquos DogKomondor Cane Corso) = 0039 Z = 4366 D(O Komondor Fonnirsquos Dog Cane Corso) = 0031 Z = 3920 D(O Fonnirsquos Dog Saluki Cane Corso) =0033 Z = 3893 D(O Saluki Fonnirsquos Dog Cane Corso) = 0040 Z = 5075 AdmixTools (blue dashed line) supports the findings of TreeMix withsignificant D-statistic values for Fonnirsquos Dog and Portuguese Water Dog of D(O Portuguese Water Dog Fonnirsquos Dog X) = 200139 to 201287


combined with WGS Such methods offer the ability to usebioinformatic tools developed for problems such as popula-tion structure a common issue in dog breeds and one forwhich it is nearly impossible to correct for using microsatel-lites thus providing a degree of confidence and precision nototherwise available (Ostrander et al 1993 Francisco et al1996)

The Fonnirsquos Dog represents an unusual case study in that itmaintains a standard appearance and well-defined behaviorin the absence of an organized human-directed breedingprogram Such programs are the hallmark of most modernbreeds and have been key to both the development andmain-tenance of purebred dogs (American Kennel Club 2006) Wechose therefore to further investigate the Fonnirsquos Dog asa means to understand how an isolated population can bedeveloped through selection based solely on functionalaptitude without dilution from surrounding mating popula-tions to meet environmental demands (Figure 1) We pro-posed to do this by assessing the historical development ofthe breed and the composition of modern canine genomestructure using SNP and WGS analyses

Accounts of the Fonnirsquos Dog from the late 1880s describethem as ldquo a very large and shaggy haired race half-mastiffhalf-bloodhound rdquo (Tyndale 1849) and a combinationof ldquo the greyhound with a big dog where the oppositedimensions of the greyhound and the mastiff cancel eachother outrdquo (Cetti 1774) While there is the expected level ofhusbandry by breed enthusiasts selection of dogs for breed-ing purposes is based predominantly on functional abilityand there has only been a very limited organized attempt

to direct the breed or to select on the basis of appearanceThe modern standard description of the breed emphasizesthe fundamental hostility of Fonnirsquos Dogs toward strangersnatural tendencies toward territoriality and loyalty only to itschargesmuch as did early descriptions of the breed (Cetti 1774)(httpwwwcanedifonniitwebenthe-official-standard)Thus within the context of dog breeds existing in Italy130 years ago the Fonnirsquos Dog had a noteworthy althoughlimited appeal that it retains today

To accomplish our goals we used data from our laboratorythat tested correlations between SNP-derived homozygosityvalues SNP- WGS- and pedigree-derived inbreeding coeffi-cients and effective numbers of early breed ancestors in-clusive of several breeds (unpublished data) This datarevealed that increases in either the number or length ofhomozygous regions within a breed as measured fromSNP-chip data significantly correlate with an increase ininbreeding coefficients calculated from SNP and WGS dataand also correlate with a decrease in the effective number ofbreed ancestors as observed by pedigree analysis (unpub-lished data) Applying those principles to the present studywe found that the sum length of homozygous stretches in thegenome of Fonnirsquos Dogs the number of regions of homozy-gosity and the inbreeding values for the Fonnirsquos Dog wereintermediate to those defining the remaining Mediterraneanbreeds (Figure 2) The Mediterranean breeds with lower ho-mozygosity values than the Fonnirsquos Dog suggestive of a morecomplex ancestral history include the Saluki Azawakh CaneParatore Sloughi Volpino Italiano Anatolian Shepherd andMastino Abruzzese The relative ranking of the Fonnirsquos Dog in

Z = 23737 to 225274 D(O Fonnirsquos Dog Portuguese Water Dog X) = 200127 to 201309 Z = 23078 to 225797 D(O Fonnirsquos Dog XPortuguese Water Dog) = 00100ndash00767 Z = 3016ndash22189 Cane Paratore of D(O Cane Paratore Fonnirsquos Dog X) = 200104 to 201317Z = 23108 to 226025 D(O Fonnirsquos Dog Cane Paratore X) = 200096 to 201289 Z = 23055 to 225805 D(O Fonnirsquos Dog X Cane Paratore) =00104ndash00811 Z = 3643ndash22974 Komondor of D(O Fonnirsquos Dog Komondor X) =200116 to201123 D(O Fonnirsquos Dog X Komondor) = 00116ndash01280 Z = 3142ndash25720 and Saluki of D(Saluki X Fonnirsquos Dog Y) = 0007ndash0131 Z = 3047ndash29352 TreeMix predicted phylogenies with (B) 1 or(C)25 allowed introgression events and corresponding standard error residuals for (D) 1 and (E) 25 introgressions

Table 4 Regions of homozygosity shared by six Fonnirsquos Dogs as identified from SNP-chip analysis

Region of homozygosity Length (bp) Personal variantsa Variant impactb Gene affected

chr62942512-2998885 56373 Nonechr7551194-675969 124775 Nonechr7733453-914233 180780 chr7742610 Modifier LGR6chr7928930-988934 60004 chr7955472 Modifier ELF3chr71027580-1298049 207469 chr71179044 Modifier NAV1

chr71190728 Modifier NAV1chr71221667 Modifier NAV1

chr81069127-1108694 39567 Nonechr13991157-1022654 31497 Nonechr131115146-1156529 41383 chr131131868 Modifier VPS13B

chr131157079 Modifier VPS13Bchr131272226-1291574 19348 chr131290464 Modifier VPS13Bchr131571048-1655073 84025 chr131572305 Modifier VPS13Bchr222578384-2618437 40053 chr132584594 Modifier FNDC3A

Chr chromosomea Individual variants from Fonnirsquos Dog WGS that are within the homozygous regionsb Variant impact and affected genes annotated from CanFam3176


terms of homozygosity and inbreeding coefficients confirmspreliminary data (Sechi et al 2016) categorizing FonnirsquosDog as a distinctive dog breed compared to a dataset of dogsfrom the Mediterranean region Additionally the rate of DNe

demonstrates that the Fonnirsquos Dog obtained its breed statusat a rate of population consolidation equivalent to breedsregulated by intensive selection systems

We produced WGS from a single Fonnirsquos Dog sampledfrom the Cagliari region of Sardinia achieving a mean depthof 363 Assessment of WGS-derived individual variants inthat dog relative to a pool of 12 additional Mediterraneandogs was undertaken to obtain a measure of individualvariation for each dog compared to representatives of distinctbreeds from the same geographic region Specific to theFonnirsquos Dog were 86166 variants of which 9224 were in theheterozygous state and 76942 were in the homozygous state(Table 3) The Saluki displayed the greatest number of indi-vidual variants (151426) with 65260 more variants thanthe dog with the second-highest number of individual vari-ants the Fonnirsquos Dog However excluding the Saluki the12 remaining dogs with WGS presented numbers of individ-ual variation within a range of 26328 Therefore the level ofindividual variation in the Saluki is substantially greater thanany of the other breeds Populations that have experiencedintensive natural or artificial selective pressures would beexpected to have a lower number of individual variants in theheterozygous state than dogs from less well-defined breedsConversely individual variants are more indicative of breed-specific though not necessarily breed-fixed privatization(Szpiech et al 2013 Marsden et al 2016)

Combined consideration of breed-specific measures ofhomozygosity inbreeding Ne and individual variation fur-ther confirms our designation of dog breeds as populationisolates While the precise degrees of population homogene-ity vary between other breeds the Fonnirsquos Dog has not dem-onstrated any substantial deviation from this definition ofbreed relative to the other Mediterranean breeds We nextaddressed the observation that contrary to most modernbreeds the Fonnirsquos Dog has attained this breed status withoutthe highly structured framework of selection toward a de-fined ideal

To determine the foundational components of the FonnirsquosDog several analyses were undertaken Genotype data from150000 SNPs were processed through the programSTRUCTURE as well as used in PCA neighbor-joining phy-logeny and FST calculations Each associated algorithmmeasures distinct population characteristics model-basedclustering is calculated from the program STRUCTURE pop-ulation differentiation from FST definition of variable corre-lation from PCA and visualization of genetic distances vianeighbor-joining phylogenyWe found that in each of the fouranalyses the Cane Paratore Mastino Abruzzese Komondorand Volpino Italiano were most similar as well as closest ingenomic structure to the Fonnirsquos Dog The Sloughi Anato-lian Shepherd Cane Corso and Saluki were also identified bySTRUCTURE and FST as sharing ancestral similarity with the

Fonnirsquos Dog Like the Fonnirsquos Dog the Cane Paratore MastinoAbruzzese Volpino Italiano and Cane Corso are breeds withhistorical origins in Italy As is suggested by the STRUCTUREoutput (Figure 4) a common genetic signature is sharedacross the regional Italian breeds (Cane Corso Cane Para-tore Fonnirsquos Dog Mastino Abruzzese Spinone Italiano Vol-pino Italiano Levriero Meridionale and Cirneco dellrsquoEtna)together with a group of breeds from the South or East Med-iterranean region (Komondor Saluki Anatolian ShepherdSloughi and Azawakh) Phylogeny and PCA aid in the reso-lution of this genetic commonality PCs distinguish breedswithin the STRUCTURE-derived common cluster isolatingthe Cane Paratore Fonnirsquos Dog Komondor Levriero Meridio-nale Mastino Abruzzese Spinone Italiano and Volpino Ital-iano as distinct from the Cirneco dellrsquoEtna Cane CorsoSaluki Sloughi Azawakh and Anatolian Shepherd (Figure6) Phylogenetic inference for the Italian breeds further sep-arates these breeds identifying a monophyletic clade consist-ing of Fonnirsquos Dog Mastino Abruzzese and Cane Paratorewhere each breed still maintains its individuality (Figure 7)These conclusions should be interpreted with caution as asmall number of dogs were analyzed for certain breeds It isunlikely however that the addition of more dogs would sig-nificantly restructure the PCAs as present separation ofbreeds is already well defined

Our analyses also show that the early claims of bothTyndale (1849) and Cetti (1774) who hypothesized thatthe Fonnirsquos Dog was created through a combination of breedsthat resembled sighthounds and molossers is surprisinglyaccurate Instead of a greyhound and a mastiff howeverknowledge of breed origins implicates a Saluki-like coursinghound and a Komondor-like molossoid livestock guardianIndeed even the more geographically available sighthound(Pharaoh Hound) and molosser (Neapolitan Mastiff) breedsfrom Italy are demoted as plausible ancestors through FSTanalysis in favor of the eastern breed-type equivalents Ex-amination of breed introgression further supports the inter-twining of the Fonnirsquos Dog Komondor and Saluki

Our analysis also suggests a directional admixture of FonnirsquosDog into an ancestral clade consisting of the PortugueseWater Dog and Cane Paratore Thus Fonnirsquos Dog continues tobe present in the genetic profiles of other Mediterraneanbreeds while other breeds do not appear to have contributedin recent time to the Fonnirsquos Dog This is unexpected as giventhe availability of other island breeds for mating we wouldexpect that at the genetic level Fonnirsquos Dog would be littlemore than a mongrel which is clearly not the case Poten-tially this relates to the prosperity of early human populationson Sardinia which were at least somewhat dependent onthe Fonnirsquos Dog Thus the desire to maintain this successfulcomposite breed as it first existed was and continues to bestrong By extension in the absence of strong selection formorphological traits in a breed that retains the critical fea-tures of its ancestors even minor selection of behavioral traitsis presumably sufficient to retain homozygosity in genomicregions critical for performance While such regions still


require fine mapping our homozygosity analyses demon-strate that such regions exist A total of 11 genomic regionsof putative selection integral for the Fonnirsquos Dog have beenidentified (Table 4) We hypothesize that future fine map-ping of these regions will reveal genes necessary for charac-teristic behaviors of the Fonnirsquos Dog

The association of the Fonnirsquos Dog with sighthound andmolossoid breeds from the Eastern and Southern Mediterra-nean as defined by the present study is particularly intrigu-ing in light of genetic analyses of modern day humansAnalysis of 300 Italian people against a background of1272 European Middle Eastern and North African individ-uals revealed that the population of Sardinia shows greatestgenetic identity with populations from Hungary EgyptIsrael and Jordan (Fiorito et al 2015) The populations fromthe peninsular regions of Italy however share greater simi-larity with European cultures to the west and north (Fioritoet al 2015) Conversely human populations across Sardiniashow no population stratification based on linguistic orgeographic regions as measured by FST STRUCTURE orPCA (Di Gaetano et al 2014) This implies that Sardiniawas primarily populated by Eastern Mediterranean andNorth African peoples who theoretically brought dogs fromtheir homelands thus populating the island of Sardinia withthe makings of what would in time become the Fonnirsquos DogOnce inhabited the island maintained a distinctive genetichomogeneity although geographic and linguistic barriers ledto the development of various well-defined subpopulationson the island The patterns of human population develop-ment on Sardinia are thus reflected in the Fonnirsquos Dog pri-marily when considering its genetic similarity to breedsoriginating in the Middle East and North Africa (KomondorAnatolian Shepherd Sloughi Saluki and Azawakh) and itscommonalities with other breeds subsequently developed inItaly (Cane Corso Cane Paratore Mastino Abruzzese Spi-none Italiano Volpino Italiano Levriero Meridionale andCirneco dellrsquoEtna)

Through application of GWS and SNP analyses we havedemonstrated that 28 Mediterranean dog breeds present thenecessary characteristics of population isolates the FonnirsquosDog has obtained equivalent characteristics through distinc-tive behaviorally-focused selective forces and the genomicinfrastructure of the Fonnirsquos Dogmirrors the ancestral humandemographic of its vicinity In addition the componentbreeds of the original Fonnirsquos Dog are reminiscent of thevarious peoples that populated the island highlighting thevalue of studying other niche populations particularly inareas of the world where humans and dogs have developedin parallel to garner skills necessary for survival

Acknowledgments

The authors thank all the dog owners and their dogs forproviding DNA samples Special thanks to Sir Terence ClarkMauricio Lima and Robert Gennari for obtaining DNAsamples from regional dog breeds from Italy North Africa

and the Middle East DLD EAO BWD and HGPacknowledge support from the Intramural Program of theInternational Human Genome Research Institute This studyis part of an Italian study ldquoDetection of morphological ge-netic and behavioral characteristics of Fonnirsquos Dog aimed atselection and official recognition of this breed by ENCIrdquofunded in part by misura 313 del Programma OperativoRegione (P O R) Sardegna 2006-2008 (Fondi Europep diSviluppo Regionale (FESR))

Literature Cited

Alexander D H J Novembre and K Lange 2009 Fast model-based estimation of ancestry in unrelated individuals GenomeRes 19 1655ndash1664

American Kennel Club 2006 The Complete Dog Book BallantineBooks New York

Arason A H Gunnarsson G Johannesdottir K Jonasson P OBendahl et al 2010 Genome-wide search for breast cancerlinkage in large Icelandic non-BRCA12 families Breast CancerRes 12 R50

Axelsson E A Ratnakumar M L Arendt K Maqbool M T Web-ster et al 2013 The genomic signature of dog domesticationreveals adaptation to a starch-rich diet Nature 495 360ndash364

Baffoe-Bonnie A B J R Smith D A Stephan J Schleutker J DCarpten et al 2005 A major locus for hereditary prostate can-cer in Finland localization by linkage disequilibrium of a hap-lotype in the HPCX region Hum Genet 117 307ndash316

Barbato M P Orozco-terWengel M Tapio and M W Bruford2015 SNeP a tool to estimate trends in recent effective pop-ulation size trajectories using genome-wide SNP data FrontGenet 6 109

Bendjilali N W C Hsueh Q He D C Willcox C M Nievergeltet al 2014 Who are the Okinawans Ancestry genome diver-sity and implications for the genetic study of human longevityfrom a geographically isolated population J Gerontol A BiolSci Med Sci 69 1474ndash1484

Benton M C S Stuart C Bellis D Macartney-Coxson D Eccleset al 2015 lsquoMutiny on the Bountyrsquo the genetic history of Nor-folk Island reveals extreme gender-biased admixture InvestigGenet 6 11

Boyko A R 2011 The domestic dog manrsquos best friend in thegenomic era Genome Biol 12 216

Bresciani A 1850 Dei costumi dellrsquoisola di Sardegna comparaticogli antichissimi popoli orientali per Antonio Bresciani Uffiziodella Civilita Cattolica Naples Italy

Capocasa M P Anagnostou V Bachis C Battaggia S Bertonciniet al 2014 Linguistic geographic and genetic isolation a col-laborative study of Italian populations J Anthropol Sci 92201ndash231

Cetti F 1774 I Quadrupedi di Sardegna Kessinger PublishingLLC

Chiang A P D Nishimura C Searby K Elbedour R Carmi et al2004 Comparative genomic analysis identifies an ADP-ribosylationfactor-like gene as the cause of Bardet-Biedl syndrome (BBS3) AmJ Hum Genet 75 475ndash484

Cingolani P and A Platts L le Wang M Coon T Nguyen et al2012 A program for annotating and predicting the effects ofsingle nucleotide polymorphisms SnpEff SNPs in the genomeof Drosophila melanogaster strain w1118 iso-2 iso-3 Fly (Aus-tin) 6 80ndash92

Danecek P A Auton G Abecasis C A Albers E Banks et al2011 The variant call format and VCFtools Bioinformatics 272156ndash2158


DePristo M A E Banks R Poplin K V Garimella J R Maguireet al 2011 A framework for variation discovery and genotyp-ing using next-generation DNA sequencing data Nat Genet 43491ndash498

Di Gaetano C G Fiorito M F Ortu F Rosa S Guarrera et al2014 Sardinians genetic background explained by runs of ho-mozygosity and genomic regions under positive selection PLoSOne 9 e91237

Edwardes C 1889 Sardinia and the Sardes Richard Bentley andSon London

Farag T I and A S Teebi 1989 High incidence of Bardet Biedlsyndrome among the Bedouin Clin Genet 36 463ndash464

Felsenstein J 1989 PHYLIP Phylogeny Inference PackageCladistics 5 164ndash166

Fiorito G C Di Gaetano S Guarrera F Rosa M W Feldmanet al 2015 The Italian genome reflects the history of Europeand the Mediterranean basin Eur J Hum Genet 24 1056ndash1062

Fogel B 1995 The Encyclopedia of the Dog Dorling KindersleyNew York

Francisco L V A A Langston C S Mellersh C L Neal and E AOstrander 1996 A class of highly polymorphic tetranucleotiderepeats for canine genetic mapping Mamm Genome 7 359ndash362

Galibert F P Quignon C Hitte and C Andre 2011 Towardunderstanding dog evolutionary and domestication history CR Biol 334 190ndash196

Georgi B D Craig R L Kember W Liu I Lindquist et al2014 Genomic view of bipolar disorder revealed by whole ge-nome sequencing in a genetic isolate PLoS Genet 10 e1004229

Ginns E I P St Jean R A Philibert M Galdzicka P Dam-schroder-Williams et al 1998 A genome-wide search for chro-mosomal loci linked to mental health wellness in relatives athigh risk for bipolar affective disorder among the Old OrderAmish Proc Natl Acad Sci USA 95 15531ndash15536

Haider N B R Carmi H Shalev V C Sheffield and D Landau1998 A Bedouin kindred with infantile nephronophthisis dem-onstrates linkage to chromosome 9 by homozygosity mappingAm J Hum Genet 63 1404ndash1410

Hartikainen J M H Tuhkanen V Kataja A M Dunning AAntoniou et al 2005 An autosome-wide scan for linkage dis-equilibrium-based association in sporadic breast cancer cases ineastern Finland three candidate regions found Cancer Epide-miol Biomarkers Prev 14 75ndash80

Hayward J J M G Castelhano K C Oliveira E Corey C Balkmanet al 2016 Complex disease and phenotype mapping in thedomestic dog Nat Commun 7 10460

Hicks A A H Petursson T Jonsson H Stefansson H SJohannsdottir et al 2002 A susceptibility gene for late-onsetidiopathic Parkinsonrsquos disease Ann Neurol 52 549ndash555

Hou L G Faraci D T Chen L Kassem T G Schulze et al2013 Amish revisited next-generation sequencing studies ofpsychiatric disorders among the Plain people Trends Genet 29412ndash418

Hsueh W C P L St Jean B D Mitchell T I Pollin W CKnowler et al 2003 Genome-wide and fine-mapping linkagestudies of type 2 diabetes and glucose traits in the Old OrderAmish evidence for a new diabetes locus on chromosome14q11 and confirmation of a locus on chromosome 1q21-q24Diabetes 52 550ndash557

Kainu T S H Juo R Desper A A Schaffer E Gillanders et al2000 Somatic deletions in hereditary breast cancers implicate13q21 as a putative novel breast cancer susceptibility locusProc Natl Acad Sci USA 97 9603ndash9608

Karason A J E Gudjonsson H H Jonsson V B Hauksson E HRunarsdottir et al 2005 Genetics of psoriasis in Iceland evi-dence for linkage of subphenotypes to distinct Loci J InvestDermatol 124 1177ndash1185

Kember R L B Georgi J E Bailey-Wilson D Stambolian S MPaul et al 2015 Copy number variants encompassing Mende-lian disease genes in a large multigenerational family segregat-ing bipolar disorder BMC Genet 16 27

Korneliussen T S A Albrechtsen and R Nielsen 2014 ANGSDAnalysis of Next Generation Sequencing Data BMC Bioinfor-matics 15 356

Leroy G E Verrier J C Meriaux and X Rognon 2009 Geneticdiversity of dog breeds between-breed diversity breed assigna-tion and conservation approaches Anim Genet 40 333ndash343

Li H and R Durbin 2009 Fast and accurate short read align-ment with Burrows-Wheeler transform Bioinformatics 251754ndash1760

Li H B Handsaker A Wysoker T Fennell J Ruan et al2009 The sequence alignmentmap format and SAMtools Bi-oinformatics 25 2078ndash2079

Lindblad-Toh K C M Wade T S Mikkelsen E K Karlsson D BJaffe et al 2005 Genome sequence comparative analysisand haplotype structure of the domestic dog Nature 438803ndash819

Marsden C D D Ortega-Del Vecchyo D P OrsquoBrien J F TaylorO Ramirez et al 2016 Bottlenecks and selective sweeps dur-ing domestication have increased deleterious genetic variationin dogs Proc Natl Acad Sci USA 113 152ndash157

Mellanby R J R Ogden D N Clements A T French A G Gowet al 2013 Population structure and genetic heterogeneity inpopular dog breeds in the UK Vet J 196 92ndash97

Nystroumlm-Lahti M P Sistonen J P Mecklin L Pylkkanen L AAaltonen et al 1994 Close linkage to chromosome 3p andconservation of ancestral founding haplotype in hereditary non-polyposis colorectal cancer families Proc Natl Acad Sci USA91 6054ndash6058

Ostrander E A G F Sprague Jr and J Rine 1993 Identificationand characterization of dinucleotide repeat (CA)n markers forgenetic mapping in dog Genomics 16 207ndash213

Parker H G L V Kim N B Sutter S Carlson T D Lorentzenet al 2004 Genetic structure of the purebred domestic dogScience 304 1160ndash1164

Parra D S Mendez J Canon and S Dunner 2008 Geneticdifferentiation in pointing dog breeds inferred from microsatel-lites and mitochondrial DNA sequence Anim Genet 39 1ndash7

Patterson N A L Price and D Reich 2006 Population structureand eigenanalysis PLoS Genet 2 e190

Peltonen L A Palotie and K Lange 2000 Use of populationisolates for mapping complex traits Nat Rev Genet 1 182ndash190

Pickrell J K and J K Pritchard 2012 Inference of populationsplits and mixtures from genome-wide allele frequency dataPLoS Genet 8 e1002967

Pickrell J K N Patterson C Barbieri F Berthold L Gerlachet al 2012 The genetic prehistory of southern Africa NatCommun 3 1143

Pribanova M P Horak D Schroffelova T Urban R Bechynovaet al 2009 Analysis of genetic variability in the Czech Dachs-hund population using microsatellite markers J Anim BreedGenet 126 311ndash318

Price A L N J Patterson R M Plenge M E Weinblatt N AShadick et al 2006 Principal components analysis corrects forstratification in genome-wide association studies Nat Genet38 904ndash909

Purcell S B Neale K Todd-Brown L Thomas M A Ferreiraet al 2007 PLINK a tool set for whole-genome associationand population-based linkage analyses Am J Hum Genet81 559ndash575

Raj A M Stephens and J K Pritchard 2014 fastSTRUCTUREvariational inference of population structure in large SNP datasets Genetics 197 573ndash589


Rambaut A 2014 FigTree version July 2014 Edinburgh httphttptreebioedacuksoftwarefigtree

Rimbault M and E A Ostrander 2012 So many doggone traitsmapping genetics of multiple phenotypes in the domestic dogHum Mol Genet 21 R52ndashR57

Rokman A A B Baffoe-Bonnie E Gillanders H Fredriksson VAutio et al 2005 Hereditary prostate cancer in Finland fine-mapping validates 3p26 as a major predisposition locus HumGenet 116 43ndash50

Rund D T Cohen D Filon C E Dowling T C Warren et al1991 Evolution of a genetic disease in an ethnic isolate beta-thalassemia in the Jews of Kurdistan Proc Natl Acad Sci USA88 310ndash314

Sajantila A A H Salem P Savolainen K Bauer C Gierig et al1996 Paternal and maternal DNA lineages reveal a bottleneckin the founding of the Finnish population Proc Natl Acad SciUSA 93 12035ndash12039

Sambrook J E F Fritsch and T Maniatis 1989 Molecular clon-ing A laboratory manual Cold Spring Harbor Laboratory PressCold Spring Harbor NY

Sarantaus L P Huusko H Eerola V Launonen P Vehmanenet al 2000 Multiple founder effects and geographical cluster-ing of BRCA1 and BRCA2 families in Finland Eur J HumGenet 8 757ndash763

Schoenebeck J J and E A Ostrander 2014 Insights into mor-phology and disease from the dog genome project Annu RevCell Dev Biol 30 535ndash560

Scott D A R Carmi K Elbedour S Yosefsberg E M Stone et al1996 An autosomal recessive nonsyndromic-hearing-loss locusidentified by DNA pooling using two inbred Bedouin kindredsAm J Hum Genet 59 385ndash391

Seboun E A Lemainque and C E Jackson 2005 Amish brittlehair syndrome gene maps to 7p141 Am J Med Genet A 134290ndash294

Sechi S M Polli S Marelli A Talenti P Crepaldi et al2016 Fonnirsquos Dog morphological and genetic characteristicsfor a breed standard definition Ital J Anim Sci (in press)

Sheffield V C R Carmi A Kwitek-Black T Rokhlina D Nishi-mura et al 1994 Identification of a Bardet-Biedl syndrome

locus on chromosome 3 and evaluation of an efficient approachto homozygosity mapping Hum Mol Genet 3 1331ndash1335

Sidore C F Busonero A Maschio E Porcu S Naitza et al2015 Genome sequencing elucidates Sardinian genetic archi-tecture and augments association analyses for lipid and bloodinflammatory markers Nat Genet 47 1272ndash1281

Simpson M A H E Cross L Cross M Helmuth and A H Crosby2009 Lethal cystic kidney disease in Amish neonates associ-ated with homozygous nonsense mutation of NPHP3 AmJ Kidney Dis 53 790ndash795

Stone D L R Agarwala A A Schaffer J L Weber D Vaske et al1998 Genetic and physical mapping of the McKusick-Kaufmansyndrome Hum Mol Genet 7 475ndash481

Suarez N M E Betancor R Fregel and J Pestano 2013 Geneticcharacterization at the mitochondrial and nuclear DNA levels offive Canary Island dog breeds Anim Genet 44 432ndash441

Szpiech Z A J Xu T J Pemberton W Peng S Zollner et al2013 Long runs of homozygosity are enriched for deleteriousvariation Am J Hum Genet 93 90ndash102

Thorgeirsson T E H Oskarsson N Desnica J P Kostic J GStefansson et al 2003 Anxiety with panic disorder linked tochromosome 9q in Iceland Am J Hum Genet 72 1221ndash1230

Tyndale J W 1849 The Island of Sardinia including pictures ofthe manners and customs of the Sardinians and notes on theantiquities and modern objects of interest in the island to whichis added some account of the House of Savoy Richard BentleyNew Burlington Street London

Van der Auwera G A M O Carneiro C Hartl R Poplin G delAngel et al 2013 From FastQ data to high confidence variantcalls the Genome Analysis Toolkit best practices pipeline CurrProtoc Bioinformatics 43 111011ndash111033

Wayne R K and B M vonHoldt 2012 Evolutionary genomics ofdog domestication Mamm Genome 23 3ndash18

Zhai G J Zhou M O Woods J S Green P Parfrey et al2016 Genetic structure of the Newfoundland and Labradorpopulation founder effects modulate variability Eur J HumGenet 24 1063ndash1070

Communicating editor R Nielsen


international organization encompassing91member countriesandrecognizing343breeds(httpwwwfcibeenPresentation-of-our-organisation-4html) The FCI however does not neces-sarily recognize regional varieties of dogs that exist worldwidewhich are often acknowledged at only a local or niche levelMany of these dog populations have often undergone selectionlargely for functional purposes and behaviors that are dis-tinctly important to the humans in the region and as suchthey represent isolated and unique genetic histories

The studyofhuman islandpopulationshasbeenofconsider-able interest to geneticists as such populations often exhibitdecreased genomic diversity (Sajantila et al 1996 Capocasaet al 2014 Di Gaetano et al 2014) an increase in linkagedisequilibrium (LD) (Bendjilali et al 2014) and increased in-breeding coefficients (Zhai et al 2016) These same popula-tion metrics have proven informative for the study of complextraits and heritable diseases particularly in the context ofrelatively young populations that have experienced rapidgrowth share a common environment and cultural practicesand frequently maintain detailed genealogical records(reviewed in Peltonen et al 2000) Populations can becomegenetically isolated through reproductive restraints imposedby cultural or religious practices language or industry Inthese instances societal restraints only influence the humangene pool rather than other cohabitant organisms Geo-graphic isolation however is an additional barrier that canaffect both the human inhabitants of a region as well as theendemic or introduced flora and fauna The Italian island ofSardinia experiences both isolating mechanisms providing aunique opportunity to investigate complex traits in a humanpopulation with low genetic diversity (Capocasa et al 2014 DiGaetano et al 2014) high frequency of private variants (Sidoreet al 2015) and unique genomic ancestry (Fiorito et al 2015)

The country of Italy is the site of origin for several in-ternationally recognized modern breeds such as the CaneCorso Maltese Neapolitan Mastiff Spinone Italiano BraccoItaliano and Cirneco dellrsquoEtna Italy also stakes claim to agreat number of regional varieties including the Fonnirsquos Dogand the Mastino Abruzzese both livestock guardians theCane Paratore a herding breed and the Levriero Meridio-nale a sighthound These breeds primarily exist as isolatessurrounding their historic regions of origin not havinggained popularity outside local breed enthusiasts

Among the most unique of the Italian varieties is the FonnirsquosDog also known as the Cane Fonnese Pastore Fonnese orSardinian Shepherd Dog It is a large livestock and propertyguardian breed originating from the region surrounding the cityof Fonni Present in rough- and smooth-coated varieties and invarying colors the unifying features of the breed include a char-acteristically intense facial expression and instinctive propensitytoward guarding behaviors andwariness of strangers Historicalaccounts portray dogs fitting this description residing in Fonniand the surrounding regions in the mid- to late-19th century(Tyndale 1849 Bresciani 1850 Cetti 1774 Edwardes 1889)

International breed recognition of the Fonnirsquos Dog is beingpursued by a dedicated group of breeders and enthusiasts

(httpwwwcanefonneseit) with a goal of preservingthe distinct heritage of this remarkable breed Preliminarystudies (Sechi et al 2016) have characterized the morpho-logical commonalities of the Fonnirsquos Dog showing that theyare consistent with features of a true-breeding populationStudies using a limited number of microsatellite markers sup-port this hypothesis (Sechi et al 2016)

In thisstudywehaveusedwhole-genomesequence(WGS)atan average coverage of 423 over 13 canids together with SNPanalysis of 173662 variants across 155 canids to characterizethe genomic architecture of Fonnirsquos Dogs in the context of geo-graphically and historically proximate dog breeds (Figure 1)We investigate the similarities between human population iso-lates and dog breeds and subsequently breed developmentwhen driven by human-defined regulation of selection or whenpredominantly based on behavioral aptitude for guardian func-tions in the absence of aesthetic preferenceWedemonstrate thepropensity for Fonnirsquos Dogs to display genomic characteristicsequivalent to those of established and acknowledged breedsproviding a dynamic account of how geographic isolation andbehavior-driven selection function to produce unique breedpopulations Finally we reveal parallelisms between the breedfoundation of the Fonnirsquos Dog and the human population de-mographics of Sardinia exposing an adroit consideration forinvestigations of population structure and the breadth of re-search application for dog breeds as population isolates

Materials and Methods

Sample acquisition and genotyping

Blood samples were obtained and sent to the National HumanGenomeResearch Institute (NHGRI) from six unrelated FonnirsquosDogs one each from the Sardinian regions of Fonni OzieriCagliari Sassari Porto Torres and Nuoro (Figure 1) Remain-ing dog breed samples were submitted by owners with a signedconsent in accordance with the NHGRI Animal Care and UseCommittee Genomic DNA was extracted from all blood sam-ples using a standard phenol chloroform method aliquotedand stored at 280 (Sambrook et al 1989) Extractions ofDNA from saliva collections were performed using the recom-mended Performagene protocol (DNA Genotek Ottawa ON)

A set of 135 dogs representing 19 breeds (Table 1) weregenotyped using the Illumina (San Diego CA) CanineHD SNPchip which has 173662 potentially informative markers Ge-notype callingwas conducted in IlluminaGenomeStudiowitha protocol specifying a 90 call rate and GenTrain scoreof04 Illumina Canine HD SNP genotypes for eight Maltesetwo Komondors two Spinone Italiano and eight Bouvier desFlandres were obtained from a previous publication (Haywardet al 2016) We thus analyzed data from 155 SNP-genotypeddogs representing 23 breeds

WGS alignment

Publically available WGS data were obtained from the Se-quence Read Archive (httpwwwncbinlmnihgovsra) for




























Data availability


Results























Breed No




















Discussion


















































Acknowledgments



Literature Cited












































































































Data availability


Results























Breed No




















Discussion


















































Acknowledgments



Literature Cited








































































































Data availability


Results























Breed No




















Discussion


















































Acknowledgments



Literature Cited


















































































Data availability


Results























Breed No




















Discussion


















































Acknowledgments



Literature Cited



























































































Breed No




















Discussion


















































Acknowledgments



Literature Cited

























































































Breed No




















Discussion


















































Acknowledgments



Literature Cited
































































































Discussion


















































Acknowledgments



Literature Cited

























































































Discussion


















































Acknowledgments



Literature Cited





























































































































Acknowledgments



Literature Cited



























































































































Acknowledgments



Literature Cited






















































































































Acknowledgments



Literature Cited


















































































































Acknowledgments



Literature Cited














































































































Acknowledgments



Literature Cited












































































































Acknowledgments



Literature Cited






























































































Acknowledgments



Literature Cited





















































































Acknowledgments



Literature Cited












































































































































































Date post:	20-Sep-2020
Category:	Documents
Upload:	others
View:	0 times
Download:	0 times

Commonalities in Development of Pure Breeds and Population ...Sardinia experiences both isolating...

Documents