Mixed ancestry and admixture in Kauai's feral chickens: invasion of domestic genes into ancient Red Junglefowl reserviors E. Gering, Martin Johnsson, P. Willis, T. Getty and Dominic Wright Linköping University Post Print N.B.: When citing this work, cite the original article. Original Publication: E. Gering, Martin Johnsson, P. Willis, T. Getty and Dominic Wright, Mixed ancestry and admixture in Kauai's feral chickens: invasion of domestic genes into ancient Red Junglefowl reserviors, 2015, Molecular Ecology, (24), 2112-2124. http://dx.doi.org/10.1111/mec.13096 Copyright: © 2015 The Authors. Molecular Ecology Published by John Wiley & Sons Ltd. This is an open access article under the terms of the Creative Commons Attribution NonCommercial-NoDerivs License http://eu.wiley.com/WileyCDA/ Postprint available at: Linköping University Electronic Press http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-117160 MolecularEcology(2015)24,2112–2124 doi:10.1111/mec.13096 INVASION GENETICS: THE BAKER AND STEBBINS LEGACY Mixed ancestry and admixture in Kauai’s feral chickens: invasion of domestic genes into ancient Red Junglefowl reservoirs E. GERING,* M. JOHNSSON,† P. WILLIS,‡ T. GETTY* andD. WRIGHT† *Kellogg BiologicalStation, Michigan StateUniversity, 3700EastGull LakeRoad, HickoryCorners, MI49060, USA, †IFM Biology, AVIAN Behavioural Genomicsand Physiology Group, Division ofZoology,AVIAN Behavioural Genomics and Physiology Group, Link€opingUniversity, S– 58183, Link€oping,Sweden,‡DepartmentofBiology, UniversityofVictoria, Cunningham202, 3800FinnertyRoad, Victoria,BC V8P5C2, Canada Abstract Amajorgoalofinvasiongeneticsistodeterminehowestablishmenthistoriesshapenon- native organisms’ genotypes and phenotypes. While domesticated species commonly escape cultivation to invade feral habitats, few studies have examined how this process shapes feral gene pools and traits. We collected genomic and phenotypic data from feral chickens(Gallusgallus)ontheHawaiianislandofKauaito(i)ascertaintheiroriginsand (ii) measure standing variation in feral genomes, morphology and behaviour. Mitochon- drialphylogenies(D-loop&wholeMtgenome)revealedtwodivergentcladeswithinour samples.TherarecladealsocontainssequencesfromRedJunglefowl(thedomesticchick- en’s progenitor) and ancient DNA sequences from Kauai that predate European contact. This lineage appears to have been dispersed into the east Pacific by ancient Polynesian colonists.ThemoreprevalentMtDNAcladeoccursworldwideandincludesdomesticated breedsdevelopedrecentlyinEuropethatarefarmedwithinHawaii.Wehypothesizethis lineage originates from recently feralized livestock and found supporting evidence for increased G. gallus density on Kauai within the last few decades. SNPs obtained from whole-genome sequencing were consistent with historic admixture between Kauai’s divergent (G. gallus) lineages. Additionally, analyses of plumage, skin colour and vocal- izations revealed that Kauai birds’ behaviours and morphologies overlap with those of domesticchickensandRedJunglefowl,suggestinghybridorigins.Together,ourdatasup- port the hypotheses that (i) Kauai’s feral G. gallus descend from recent invasion(s) of domesticchickensintoanancientRedJunglefowlreservoirand(ii)feralchickensexhibit greater phenotypic diversity than candidate source populations. These findings compli- catemanagementobjectives forPacificferalchickens,whilehighlightingthepotentialof thisandotherferalsystemsforevolutionarystudiesofinvasions. Keywords: conservation genetics, Gallius gallus,hybridization, invasivespecies Received17September 2014;revision received17December2014; accepted 20December 2014 migrating ancestors were accompanied by both acciden- Introduction tal stowaway species (Mack et al. 2000; Lockwood et al. Humans have dispersed over most of the Earth’s sur- 2005;Estoup&Guillemaud2010)anddomesticatedtaxa face, and we have not made these journeys alone. Our that were utilized for food, labour or companionship (Larson et al. 2007, 2012). Subsequent to anthropogenic dispersal, many domesticated species have escaped cul- Correspondence:EbenGering,Fax:+12696712165; E-mail:[email protected],Fax:+461313 tivation and colonized new habitats, a process termed 7568;E-mail:[email protected] feralization. It can be helpful to think about feralization ©2015TheAuthors.MolecularEcologyPublishedbyJohnWiley&SonsLtd. ThisisanopenaccessarticleunderthetermsoftheCreativeCommonsAttribution-NonCommercial-NoDerivsLicense, whichpermitsuseanddistributioninanymedium,providedtheoriginalworkisproperlycited,theuseisnon-commercialand nomodificationsoradaptationsaremade. FERAL CHICKEN INVASION 2113 as ‘domestication in reverse’, as it involves the removal resource, vectors of highly lethal pathogens and our of direct anthropogenic control over natural and sexual planet’s most abundant bird (for example, see http:// selection regimes (Price 1984). Thus, feral population www.uspoultry.org/economic_data/). In contrast, RJF persistence requires survival and reproduction within are poorly suited to commercial food production, are novel social and ecological contexts. While there has threatened or endangered in their native range and been extensive research into what facilitates or hinders merit stringent conservation effort (Peterson & Brisbin invasions of nondomesticated species (e.g. standing 1998). Thus, ascertaining the history of Kauai’s chickens genetic and phenotypic variation), the process of ferali- will have important implications for invasion biology, zation is less well understood. Progress in this area will cultural anthropology, and G. gallus conservation and help advance our basic understanding of biodiversifica- management. tion and can help mitigate feral species’ impacts on Our aim in this study was to characterize the demog- nativeecosystemsandcompetitors(Loope1999). raphy, genetics and phenotypes of G. gallus on Kauai Feral habitats can potentially exert strong selection on andtherebyelucidate theiroriginsandcapacity forevo- several components of fitness in the wild (e.g. mate lutionary responses to feral selection pressures. We acquisition, foraging success, predator avoidance and assessed population substructure and phylogeny using disease resistance). Evolutionary responses to these whole-genome sequencing (WGS) of modern samples selection pressures will depend upon both the genetic taken from disparate Kauai sampling localities. We then variability of feral populations and the genetic architec- determined relationships among sampled individuals’ tureoffitness-relatedtraits(Goodwin2007;Zoharyet al. nuclear and mitochondrial genomes, including previ- 2012). These properties of feral populations result from ously published data sets from (i) both ancient (pre- combined histories of domestication and feralization, European contact) and modern samples from the Paci- each of which can be complex (Verardi et al. 2006; Ste- fic, (ii) RJF and (iii) modern domestic chicken breeds. phens 2011; Feulner et al. 2013; McTavish et al. 2013; We also measured phenotypic traits that are known to Nussberger et al. 2013). By characterizing genetic and differ between domesticated chickens and RJF (rooster phenotypic variation in feral taxa, we can therefore vocalizations, leg colours and plumage; see Box 2) make progress towards the interrelated goals of under- among free-living G. gallus on Kauai. We used these standingferalpopulations’originsandascertainingtheir data together to determine (i) whether Kauai’s feral capacities to respond to current and future selection. To chickens are of mixed, Polynesian or European origin, date, very few studies have jointly examined genetic (ii) whether there is evidence of interbreeding between and phenotypic variation in feral species using modern feral lineages and (iii) whether co-ancestry and admix- tools(althoughseeHamptonet al.2004;Randi2008). ture on Kauai is associated with enhanced phenotypic Here, we examine genotypic and phenotypic varia- variation. tion in feral chickens (Gallus gallus) on the Hawaiian island of Kauai. The origins of these birds are presently Methods unclear; they have been alternatively described as either escaped farm pests (‘feral domestic fowl’) or as a ‘leg- Genetic and phenotypic data collection acy species’ introduced by Polynesian colonists (i.e. Red Junglefowl), the chicken’s closest free-living ancestor Chickens were donated from private individuals living (Eriksson et al. 2008; see Box 1). Broad-scale studies of on Kauai. We preserved blood in RNAlater and Pacific ‘chicken’ biogeography (using MtDNA markers) extracted DNA using salt extraction techniques (Aljana- have also drawn conflicting conclusions as to whether bi & Martinez 1997) in a Swedish laboratory (samples contemporary populations are of ancient origin (Polyne- imported under permit DRN 6.2.18-1361/13). A total of sian Red Junglefowl) or are descended from domestic 23 samples were thus procured from eight different breeds that originated more recently in Europe (Storey regions of the island (see Fig. 1A and Table S1, Sup- et al. 2012; Larson et al. 2014; Thomson et al. 2014). porting information). We also recorded vocalizations, These uncertainties complicate efforts to use G. gallus plumage and leg colours from an additional 21 individ- biogeography to reconstruct Polynesian expansion into uals at these and other nearby localities (Table S2, Sup- the Pacific and, possibly, South America (Storey et al. porting information). 2007, 2012; Beavan 2014; Bryant 2014; Thomson et al. 2014). They also raise important questions about best Sequencing and variant calling practices for feral population management. Although chickens and Red Junglefowl (RJF) can interbreed, DNA samples were sequenced using the SOLiD 5500xl applied biologists regard the two lineages very differ- platform at Uppsala Genome Center, part of the ently. Domestic chickens are a globally critical food National Genomics Infrastructure, and were analysed ©2015TheAuthors.MolecularEcologyPublishedbyJohnWiley&SonsLtd. 2114 E. GERING ET AL. Box 1. Biogeographic history ofHawaiian Gallusgallus Potential sources of Kauai’s Gallus gallus Archaeological evidence indicates that chickens were first introduced to the Hawaiian island chain by AD1200 (including Kauai, see Fig. 1a) via human migration into the eastern Pacific (Wilmshurst et al. 2011; Thomson et al. 2014). Their sources were most likely Red Junglefowl (RJF) transported from the western Pacific by Polynesian set- tlers (Thomson et al. 2014). An additional 857 Pacific RJF were introduced to Kauai in 1939 in a state-sponsored effort to maintain game bird populations in the islands (Pyle & Pyle 2009). Therefore, it is possible that wild RJF persisted on these islands for over 1000 years, although this reservoir population may also be more recently derived(mostlikelyfromother Polynesian-dispersed sources inthe Pacific).Inthis manuscript,weconsiderG. gal- lus from both ancient and historic (1939) RJF introductions as ‘heritage’ animals because (i) both were dispersed from their native range without experiencing modern, artificial selection for food production, and (ii) modern and ancient samplesfrom Kauai share MtDNA genotypes(see Results); thus, ifRJFre-introductions contributed toferal gene pools, thenboth ancient andhistoric introductions originated from closely relatedsource populations. In the light of anecdotal claims from Kauai residents that contemporary G. gallus originated within the last few decades, it is alsopossible that RJF were extirpated from Hawaii and/or have been replaced byescaped domestics. In the recent past, multiple European-derived, modern breeds have been cultivated in Hawaii for food production and cockfighting (personal communication from Kauai residents to D. Wright and E. Gering; and online sales from Asagi hatchery, Oahu, see http://www.asagihatchery.com/). In the 1980s and 1990, Tropical storm Iwa and Hurri- cane Iniki destroyed many of the coops containing Kauai’s domestic birds, released their occupants into local for- ests and potentially spurred large-scale species invasions. Consistent with this possibility, our analysis of G. gallus point counts revealed marked increases in population densities during the last few decades (see Fig. 1). Nonethe- less, this expansion of domestic genes into Polynesian-derived reservoirs may have been preceded by earlier epi- sodes of introgression, as morphological analyses of five skins that were sampled on Kauai in 1895 also showed evidenceof‘genetic pollution’ from domesticated chickens(Peterson &Brisbin 2005). In summary, the gene pool of feral Kauai’s G. gallus may descend from ancient Polynesian RJF introductions, from historic (1930s)RJFre-introductions and/or fromdomestic chickens ofrecent Europeanorigins. History of Pacific G. gallus The domestication of the chicken is believed to have occurred up to 8000 years ago in China, South Asia and South-East Asia (West & Zhou 1988). Much more recently, domestic breeds have undergone a range of phenotypic and genotypic changes. Domestic breeds show a loss of nuclear genetic diversity (Muir et al. 2008) yet still exhibit a high degree of structure and variability in mitochondrial (Mt) sequences (Fumihito et al. 1996; Kanginakudru et al.2008; Silvaet al.2009; Thomsonet al.2014),with ninemajor Mtclusters identified worldwide (Liuet al. 2006). MtDNA sequences from several ancient Hawaiian specimens fall solely within haplogroup D, a clade restricted to Asia–Pacific areas (Thomson et al. 2014; but see also Beavan 2014; Bryant 2014; Storey et al. 2007). In contrast, a small modern sample (n = 10) taken from the Koke’e region of Kauai was solely comprised of haplogroup E (Thomson et al. 2014). The E haplogroup currently occurs worldwide and, together with haplogroups A and B, is the source of European-derived domestic food production breeds (Liu et al. 2006). Both D and E clades have been found within modern Pacific samples, with the majority of samples outside of Hawaii being haplogroup D (Thom- sonet al.2014; see Fig. 1). using computational resources provided by the Uppsala LIFESCOPE GENOMIC ANALYSIS Software version 2.5.1. For Multidisciplinary Center for Advanced Computational mitochondrial (Mt) genome analyses, we used the con- Science (Lampa et al. 2013). Fragment reads of 75 bp sensus sequence generated by LIFESCOPE. For the nuclear were sequenced with one individual run per lane. In genome, we called variants as follows: first, we addition to the Kauai samples, we also sequenced two marked and removed duplicate reads with Picard RJF males, one in the same manner as the Kauai sam- (http://picard.sourceforge.net). We then performed ples and one to approximately 209 coverage with pair- local realignment around potential indels and base ended reads of 50 bp plus 35 bp. Reads were aligned to quality score recalibration with GATK, followed by the chicken reference genome version GALGAL4 with variant calling with the GATK Unified Genotyper ©2015TheAuthors.MolecularEcologyPublishedbyJohnWiley&SonsLtd. FERAL CHICKEN INVASION 2115 Box 2. Phenotypic signatures ofGallusgallus domestication Domestication has induced a multitude of heritable changes in G. gallus phenotypes, including changes in behavio- ural, reproductive and physiological traits (Wright et al. 2006, 2008, 2010; Johnsson et al. 2012, 2014). Perhaps some of the most striking alterations are in the plumage, with the classic red, black and green feather pattern of the RJF giving way to far more variable coloration in domestic and fancy chicken breeds. Broiler and layer birds (selected for meat and egg production, respectively) have been bred to display a range of coloration, although the vast majority of broiler breeds are white (of the Aviagen, Cobb and Grimaud breeds available, only the Rowan Ranger, Cobb Sasso and Hubbard Color breeds are brown or black, see www.aviagen.com, www.cobb-vantress.com, www.hubbardbreeders.com). Most commercial layer breeds are either white or reddish brown (e.g. the Hy-Line W36, CV22, Silver-Brown, Brown and White Leghorn breeds), while heritage breeds of layer chickens tend to exhi- bitfargreater plumagediversity (seewww.hpbaa.com). The genetics of plumage colour is fairly well understood in the chicken. For example, the major locus causing white coloration in the chicken is the Dominant White mutation, occurring at the PMEL17 gene (Kerje et al. 2004); other color mutations at MC1R are also known (Kerje et al. 2003). Yellow legs are another characteristic that distin- guished many domestic chickens from RJF (which are fixed for grey legs); the locus controlling this polymorphism has alsobeen previouslyidentified (Eriksson et al. 2008). The extensive variation in plumage and coloration introduced by domestication can be helpful in determining whether an RJF gene pool has been ‘contaminated’ by the introgression of domesticated alleles (e.g. Brisbin & Peterson 2007). However, captive intercross studies also show that it is difficult to infer the degree of introgression within individuals based on plumage or other phenotypic characters (Condon 2012). This is perhaps unsurprising, given that poultry breeders have long understood the inheritance of most G. gallus phenotypes (including plumage traits) tobe subject to epistasis. WhileRJFanddomesticchickensbearmanysimilaritiesinvocalrepertoires,theyarereportedtodifferconsistently in the length of the last syllable of the rooster crow (Collias 1987), a trait that is considered diagnostic of domestic vs. RJF ancestry (Miller 1978). Evidence of genetic effects on call phenotypes is further supported by enhanced call variation following hybridization between domestic G. gallus breeds (Marler et al. 1962). To our knowledge, the present study is the first to compare vocalizations from numerous chicken breeds and from individuals sampled in multiple (worldwide) localities. It therefore offers new insights into the relative roles of genes and environments in G. gallus vocalizations. Our results confirm a significant difference between calls recorded from chickens and RJF (see Results). Thus, plumage colour, skin colour and vocalizations of Kauai birds comprise three genetically con- trolled traits that canbe compared withG. gallus’ancestral (RJF)and derived(domesticated) states. (DePristo et al. 2011). Finally, we took a random subset used, along with others from eastern Polynesia). We of markers from each chromosome (1-28 and Z) for constructed two mid-point-rooted control region phy- use in the PCA and STRUCTURE analyses (see details logenies: one including only D and E haplogroup below). sequences from Hawaii and one including all D and E haplogroup sequences from Pacific islands. Sequences were aligned with CLUSTALW version 2.1.0 (Larkin et al. Mitochondrial DNA phylogeny 2007) and the phylogeny was constructed with MRBAYES For the whole Mt genome phylogeny, we aligned our version 3.2.3 (Ronquist & Huelsenbeck 2003; Altekar 23 Mt sequences with 61 whole mitochondrial genomes et al. 2004) using the general time reversible model available in GenBank (using haplogroup designations allowing for a proportion of invariable sites and a established by Miao et al. (2013); see their Supplemen- gamma distribution for the evolutionary rates at the tary data set 3). We also included the Mt sequence of other sites. We ran two metropolis-coupled Markov the chicken reference genome (also RJF) and used the chain Monte Carlo simulations for one million itera- duck Mt genome (BGI duck version 1.0) as an out- tions, saving every tenth iteration and discarding the group. For the Mt control region phylogenies, we first 25 000 samples as burn-in. The estimated sample aligned our samples with the Mt sequences from size was above 100 and the potential scale reduction Hawaii chickens collated by Thomson et al. (2014), see factor was close to 1.0 for all parameters, suggesting their Supplementary data set 6 (10 modern Hawaiian convergence. Trees were drawn with FIGTREE version samples and seven ancient Hawaiian samples were 1.4.0(http://tree.bio.ed.ac.uk/software/figtree/). ©2015TheAuthors.MolecularEcologyPublishedbyJohnWiley&SonsLtd. 2116 E. GERING ET AL. (A) (B) (C) 5 1 r u o H ty- 10 r a p r e p 5 s d r Bi 0 1980 1990 2000 2010 Fig.1 (A)MapofKauaishowingMtDNAhaplogroupfrequenciesfromsamplinglocalitiesinthewestern,centralandnorthernareas of the island (details provided in Table S1, Supporting information). (B) Data from modern and ancient MtDNA sequences show a recent increase in frequency of clade E, which is associated with domestic chickens of European origin, which are now farmed worldwide. Data shown consist of western and eastern Polynesian samples taken from Thomson etal. (2014) and Dancause etal. (2011).*Ehaplogroupsamplesthataredisputedaspotentialcontamination(seetext),#Hawaiisamplesfromthecurrentstudyonly. (C)DatafromChristmasbirdcountsin Kapa’aandWaimea(Kauai).Increaseddensitiesof feralG.galluscoincidedwithtwomajor stormevents(indicatedbydashedlines)thatdamagedislandinfrastructureandmayhavefacilitatedtheferalizationofescapedlive- stock. algorithm (suitable for genomewide data sets) to assess Population structure analysis Kauai population genetics in the context of worldwide Population structure was analysed using three different G. gallus biogeography. We included the 60k chicken approaches.First,we usedprincipal componentanalysis SNP chip genotypes published by Wragg et al. (2012) (PCA) (Wu et al. 2011). The genotype data consisted of for comparison. Specifically, we extracted the 13 412 2900 single nucleotide and indel variants spanning the SNP markers that were typed both on the 60k chip and chicken genome, where we selected 100 markers each in our sequencing data, that did not contain private from chromosomes 1-28 and Z. Principal component allelesandthat could bemapped to theGALGAL4 version analysis was performed in the R statistical computation of the chicken reference genome in Ensembl version 76. environment(RCoreTeam2012)usingtheprcompfunc- The Japanese Totenko chickens formed an extreme iso- tion. The first and second principal components lated group in a preliminary principal component explained 5% and 4% of the variance, respectively. Indi- analysis (data not shown) and were excluded from fur- vidual scores on the first and second principal compo- ther analysis, as well as the White Star chickens of nents were displayed in scatterplots created with the unknown geographical origin. We fit models with K ggplot2packageinR(seeggplot2.org).Oftheabove-men- ranging from 1 to 6 and chose a value of K = 3 based tioned2900variants,weusedthe1042thatwerebiallelic on fivefold cross-validation. Cross-validation means andhadcompletegenotypesforall23samplesforprinci- partitioning the data, in this case into fifths, and repeat- palcomponentanalysiswithintheKauaipopulation. ing the analysis one time without each subset. Each Next, we analysed the same variants using the Bayes- time, the excluded subset is predicted based on the ian clustering approach implemented in the program fitted model. The model with K = 3 had the lowest STRUCTURE (Pritchard et al. 2000). We used a model run cross-validation error. We also performed principal burn-in procedure of 100k replicates, followed by 100k component analysis on this data set. For principal com- MCMC simulations, repeating parameters for 20 runs at ponent analysis of the worldwide chicken data set, we each value of K (K = 1 through K = 5). We extracted used the 3860 markers that had complete data. Four assignment proportions for the best-supported value of individuals were excluded from our ADMIXTURE, STRUC- K (following guidelines from the author; see ‘Results’) TUREandPCA analysesastheywere known siblings. using STRUCTURE HARVESTER (Earl 2012) and plotted these results using the package pophelper in R (available at Vocalization and colour traits https://github.com/royfrancis/pophelper). Finally, we used ADMIXTURE software (Alexander et al. Recordings of crowing roosters were made in the 2009; Alexander & Lange 2011), which fits the same field using a Pocketrak 2G recorder (Yamaha). We modelasSTRUCTUREbutwithafastermaximum-likelihood simultaneously collected field observations and digital ©2015TheAuthors.MolecularEcologyPublishedbyJohnWiley&SonsLtd. FERAL CHICKEN INVASION 2117 photographstodetermineindividuals’plumagefeatures variation on our analyses. Vocalizations were quantified andlegcolours.Themostcommonplumagephenotypes using RAVEN PRO software (Cornell). We limited analyses observed approximated ‘classic’ RJF types for both to durations of previously described call syllables (Col- males and females (see Fig. 2A–C). We also observed lias 1987) because (i) inspections of sonograms sug- multiple individuals with plumage phenotypes that are gested that the recording equipment and conditions not observed in RJF, including several individuals with might confound call frequency analyses and (ii) call fre- whitemarkingflecksandasmallernumberofindividu- quencies, unlike syllable durations, are known to be als with other plumage patterns (e.g. black or mostly influenced by hormones and the social environment white;Fig. 2D–F).To compare call traits amongindivid- (Leonard & Horn 1995). We discarded measurements uals that differed in morphology, we categorized indi- whenever recording quality and/or call properties viduals with non-RJF phenotypes (alternative plumage made it difficult to isolate syllable onset/offset, which and/or yellow legs) as ‘Kauai chickens’ and those with was occasionally the case for the third syllable of calls. classicRJFmorphologyas‘KauaiRJF’. Thiswasnotanissuefortheacousticallydistinctivelast Recordings of localities were selected to span the syllable (see TableS2,Supporting information). same major regions of the island as genetic analyses (Fig. S1, Supporting information). We supplemented Results field-collected vocalizations with published measure- ments and recordings collected from public databases Mt DNA Phylogenies (Table S2, Supporting information). We deliberately selectedbothRJFandchickensfromarangeoflocalities Whole Mt genome analyses revealed a total of 3 ‘D’ ha- worldwide to reduce any contribution of environmental plogroupindividualsand20‘E’haplogroupindividuals. (A) (B) Fig.2 Sample G.gallus phenotypes from Kauai. Panes A–C depict the standard Red Junglefowl (RJF) plumage. Panes D–F illustrate white coloration (D, E, F) and yellow legs (D, E), two genetically regulated traits that do not occur in nativeRJF. (C) (D) (E) (F) ©2015TheAuthors.MolecularEcologyPublishedbyJohnWiley&SonsLtd. 2118 E. GERING ET AL. ‘E’ haplogroup sequences from Kauai clustered with haplogroup individuals we sampled on Kauai were those from domestic chickens of recent European origin highly similar to sequences from archaeological speci- (notably, various commercial line chickens; see Fig. 3). mens thatpredate European contact with Hawaii.These In contrast, the three ‘D’ haplogroup sequences were sequences were obtained from multiple, geographically affiliated with sequences from wild fowl collected in distant Kauai localities (Kapa’a and Haena). Thus, the Manila, Bali, India and Myanmar and from domestic ‘ancient-like’ clade D is present on the island today, is chickens sampled in China. Within Hawaii, the co- not restricted to one regionand islikely to be recovered occurrence of divergent Mt clades (with different, from additionallocalities upon further sampling. although overlapping, geographical ranges) suggests the possibility ofmultiple geographical origins. Coloration and vocalizations Phylogenetic analyses of the Mt control region (CR) allowed us to compare ancient and contempo- Most birds on Kauai exhibited the RJF plumage pheno- rary sequences from the Hawaiian Islands (Fig. S2, Sup- type, but in several cases, we observed moderate porting information). The CR sequences of three ‘D’ amountsofwhiteand/orbrownfeathersand/oryellow Fig.3 BayesianwholeMtDNAgenomephylogenyforbirdsfromKauaiinrelationtodomesticchickensandRJF.Subtreesrepresent- ing haplogroups other than D and E have been collapsed. Posterior probabilities (expressed as percentages) are indicated at nodes. Kauaisamplesarehighlightedwithredbrackets. ©2015TheAuthors.MolecularEcologyPublishedbyJohnWiley&SonsLtd. FERAL CHICKEN INVASION 2119 legs (see Fig. 2). These traits indicate the presence of Fig. 5A). When plotted among genotypes from world- domestication-specific alleles and thus further support wide chicken breeds, genotypes from Kauai exhibited mixed and/or admixed ancestry for the island’s feral considerable variation, particularly given the island’s G. gallus. As previously described in G. gallus (e.g. diminutive size in relation to other sampling areas. As Miller 1978), individual roosters on Kauai produced a group, the Kauai sample was largely nested among highly repeatable crows in the field (results not shown). those from both RJF and European G. gallus domesticus ANOVA tests of differences in crowing traits among indi- (see Fig. 5A). It is therefore difficult to determine from viduals found no differences infirst and second syllable nuclear data alone whether Kauai birds stem from RJF, durations (first syllable F =1.98, P = 0.14, second syl- European or hybrid origin. In contrast, Kauai genotypes 3,32 lable F =0.671; P = 0.58). In contrast, we found highly were easily distinguished from those of Africa, Asia, 3,32 significant differences in durations of third (F =6.074; North America andallbutone SouthAmerican breed. 3,31 P = 0.0022) and fourth syllables of rooster crows Our ADMIXTURE analysis supported three origins for (F =10.85; P = 1.97e-05). Post hoc (Tukey HSD) tests worldwide G. gallus genotypes (see Fig. 5B). Kauai 3,43 revealed a differencebetweencallsofdomestic chickens genotypes were most similar to RJF (with a mostly ‘red’ and all other groups in the third syllable (Fig. 4). The origin in Fig. 5B, in addition to some ‘blue’ and a small fourth syllable duration found significant differences in amount of ‘green’) and distinct from most other G. gal- each pairwise comparison, with the exceptions of lus genotypes. In an exception to this pattern, 4–6 of the chicken 9 Kauai chicken and RJF 9 Kauai RJF. In other European samples showed the same population compo- words (i) fourth syllable durations differed between nentsastheseRJFindividuals.STRUCTUREanalysesofKa- domestic chickens and RJF, and (ii) feral G. gallus with uai genotypes suggested k = 2 founder populations, chicken-like coloration produced chicken-like calls, indicative of multiple origins (see Fig. 5C). This finding whereas feral G. gallus with RJF-like coloration pro- adds support for a joint European and RJF ancestry for duced RJF-like calls (Fig. 4). Finally, a Levene’s test these birds, with the caveat that European genotypes indicated that the different Kauai phenotypes (Kauai were highly variable. Thus, further samples from both RJF and Kauai chicken) had significantly more variation RJF and Europe will be highly informative, particularly than the RJF and domestic groups analysed for fourth for elucidating whether one Kauai subpopulation truly syllable duration (F =3.2, P = 0.03), although not for corresponds to RJF according to nuclear genetic mark- 3 43 third syllable duration. ers. At k = 2, STRUCTURE also revealed (i) the presence of To test whether vocalizations of birds on Kauai were admixed individuals in Kauai and (ii) that Mt and predicted by location, birds were classified into three nuclear genotypes do not cosegregate (i.e. Mt clades ‘E’ regions (north, east and west), with third and fourth and ‘D’ are not associated with divergent nuclear back- syllable duration then tested between pairs of Kauai grounds). locations (see Fig. S3, Supporting information). Tukey Finally, it should also be noted that geographical sep- HSD tests revealed no differences with either third syl- aration could play a role in the population structure lable (smallest P > 0.9) or fourth syllable (smallest based on nuclear genetic markers, with individuals P > 0.29). from central regions appearing to be more variable and distinct from those from northern and western regions. Therefore, in summary, the nuclear genetic markers, Population genetic structure although potentially indicating a hybridized population PCA analysis of nuclear genotypes from Kauai revealed between European and RJF chickens, cannot rule out a variable, yet continuous, single population (see other possibilities. Fig.4 Durations of third and fourth syl- lables of rooster crows sampled in the field (Kauai) and mined from public databasesandliterature(worldwide).For sampling details, see Table S2 (Support- inginformation). ©2015TheAuthors.MolecularEcologyPublishedbyJohnWiley&SonsLtd. 2120 E. GERING ET AL. (A) Fig.5 (A) PCA plot of genetic data showing PC1 vs. PC2 for samples from Kauaiinrelationtovariousotherchicken breeds(takenfromWraggetal.2012).(B) ADMIXTURE plot showing probable ances- try of Kauai samples in relation to other chicken breeds (using data from Wragg etal. 2012). (C) STRUCTURE plot indicating assignment proportions for individuals sampled on Kauai. *Individuals with D- clade mitochondrial sequences. RJF = Red Junglefowl sequences sampled from a captive population (see Table S1, Sup- portinginformation). (B) (C) Kauai’s recent past (Fig. 1C). Among Kauai residents, Discussion this change is typically attributed to the damage of island infrastructure following tropical storms Iwa and Evidence of mixed ancestry in Kauai’s feral G. gallus Iniki, which potentially released farm birds into local We discovered several intriguing patterns of genetic forests. Alternatively (or additionally), increased tourist variation within Kauai’s feral chickens. First, our analy- activity since the 1970s may have contributed to the fer- ses of whole Mt genomes revealed that two divergent alization of Kauai domestics by providing habitat, food Mt lineages co-occur on the island (clades ‘E’ and ‘D’). or other key resources to escaped animals (Pyle & Pyle The E haplogroup includes sequences found in modern 2009). Further study is needed to ascertain the contribu- European breeds that are cultivated worldwide for tions of these biotic, abiotic and anthropogenic facilita- food. In contrast, the D haplogroup is overwhelmingly tors of invasion and to assess their potential role(s) in restricted to Asia and the Pacific and (based on ancient the recent expansionofclade E. DNA sequences) was already present on Kauai nearly 1000 years ago. Ancient and modern sequences from Evidence of admixture from pacific G. gallus other Pacific Islands suggest that this lineage was dis- persed by Polynesian settlers long before European This is the first study to jointly examine Mt and nuclear exploration. Thus, clade D either persisted on Kauai genotypes from Pacific feral chickens. Our nuclear into the present day or was subsequently repopulated (PCA) analyses reveal that some genotypes found in from aclosely related sourcepopulation. Kauai are distinct from other populations, although Within Kauai, the historic displacement of clade D by they are similar to European samples (see Fig. 3 and clade E may have accompanied the feralization of Fig. S2, Supporting information). ADMIXTURE analyses of domestic animals, a possibility that is supported by nuclear data indicated mixed ancestry of Kauai individ- evidence of a rapid increase in G. gallus density within uals, which share source populations with candidate ©2015TheAuthors.MolecularEcologyPublishedbyJohnWiley&SonsLtd.