ANIMAL GENETICS • ORIGINAL PAPER

Is black coat color in wolves of Iran an evidence of admixedancestry with dogs?

Rasoul Khosravi & Marzieh Asadi Aghbolaghi &Hamid Reza Rezaei & Elham Nourani &Mohammad Kaboli

Received: 12 August 2013 /Revised: 6 May 2014 /Accepted: 23 July 2014# Institute of Plant Genetics, Polish Academy of Sciences, Poznan 2014

Abstract Melanism is not considered a typical characteristicin wolves of Iran and dark wolves are believed to haveoriginated from crossbreeding with dogs. Such hybrid indi-viduals can be identified with the combined use of genetic andmorphological markers. We analyzed two black wolves usinga 544 base pairs (bp) fragment of the mtDNA control regionand 15 microsatellite loci in comparison with 28 dogs, 28wolves, and four known hybrids. The artificial neural net-works (ANNs) method was applied to microsatellite data toseparate genetically differentiated samples of wolves, dogs,and hybrids, and to determine the correct class for the blackspecimens. Individual assignments based on ANNs showedthat black samples were genetically closer to wolves. Also, inthe neighbor-joining network of mtDNA haplotypes, wolvesand dogs were separated, with the dark specimens located inthe wolf branch as two separate haplotypes. Furthermore, wecompared 20 craniometrical characters of the two black indi-viduals with 14 other wolves. The results showed thatcraniometrical measures of the two black wolves fall withinthe range of wolf skulls. We found no trace of recent hybrid-ization with free-ranging dogs in the two black wolves. Darkcoat color might be the result of a natural combination ofalleles in the coat-color-determining gene, mutation in the K

locus due to past hybridization with free-ranging dogs, or theeffect of ecological factors and adaption to habitat conditions.

Keywords Black wolf . Hybridization . mtDNA .

Microsatellite . Artificial neural networks

Introduction

One of the most documented variations among wolves thatoccupy different habitat types is color pattern. Mech (1970)described coat color in wolves ranging from white, buff,tawny, reddish, and gray to black, with gray being the mostcommon pelage (Apollonio et al. 2004). Anderson et al.(2009) showed that dark color in North American wolvesliving in forest habitats is the result of apparent selection forthe melanistic KB allele due to past hybridization with thedomestic dog. Also, many authors have suggested that anom-alous morphological traits in wolves, such as atypical colorpatterns, dewclaw, body proportions, or dental anomalies,might be reliable signs of hybridization with free-ranging dogs(Boitani 1992; Ciucci et al. 2003). By surveying three hybridsamples, Milenković et al. (2006) identified some atypicalmalformations, such as incompletely developed permanentteeth P1, spongy bony tissue in the foramen infraorbitale,semicircular lines of the hind part of the forehead in adultwolves, and atypical appearance of the sutura frontalis skulls,which were unusual for typical wolves. Andersone et al.(2002) found black coat color in a litter of seven mongrelpups in northern Latvia, whose individual genotypes showedthat most of the alleles were common with dogs. Randi andLucchini (2002) analyzed two wolves with black coat colorand found that one had mixed ancestry in the dog gene pool.On the other hand, Apollonio et al. (2004) showed that theoccurrence of the black coat color in wolves is not necessarilya result of interbreeding with free-ranging dogs and can be due

R. KhosraviDepartment of Environmental Sciences, Faculty of NaturalResources, Isfahan University of Technology, Isfahan, Iran

M. Asadi Aghbolaghi : E. Nourani :M. Kaboli (*)Department of Environmental Sciences, Faculty of NaturalResources, University of Tehran, Karaj, Irane-mail: mkaboli@ut.ac.ir

H. R. RezaeiDepartment of Environmental Sciences, Faculty of NaturalResources, Gorgan University of Agriculture and Natural Resources,Gorgan, Iran

J Appl GeneticsDOI 10.1007/s13353-014-0237-6

to the natural combination of wolf alleles in the gene thatdetermines coat color. The results of these studies showed thatcoat color in wolves is a complex mechanism affected by bothecological and genetic factors. Various studies showed thatecological features are more significant than geographic dif-ferences in determining the genetic and morphometric varia-tions in wolf populations (Geffen et al. 2004; Carmichael et al.2007; Musiani et al. 2007; Bennett 2010).

Wolves and free-ranging dogs are isokaryotypic, fully inter-fertile, and have been shown to mate in the wild as well as incaptivity (Wayne et al. 1995; Vila andWayne 1999). Therefore,being able to detect hybrid individuals is important from amanagement perspective (Vila et al. 2003). Polymorphic re-gions of the mitochondrial DNA (mtDNA) have successfullybeen applied to examine genetic relationships between popula-tions within and among closely related species (Tsuda et al.1997). SincemtDNAmarkers have shown a small rate of wolf–dog hybridization, the use of mtDNA alone cannot provide anyinformation about the introgression of hybrids of crosses be-tween a female dog and a male wolf in wolf populations.Recent studies involving nuclear markers have shown thathybridization occasionally occurs in the wild (Andersoneet al. 2002; Randi and Lucchini 2002). Therefore, the com-bined use of biparent and autosomal markers can be morehelpful in detecting hybrid individuals in wolf populations.

The gray wolf (Canis lupus pallipes) is one of the mostimportant carnivores in Iran. This species has evolved tosurvive in a variety of habitats, from arid deserts to mountain-ous habitats and woodlands (Ziaie 2008). Diverse habitats,such as the Alborz and Zagros mountain ranges in the northand west, central deserts, the Caspian Sea, and the PersianGulf coasts, cause a lot of variation in morphological traits inwolves (Khosravi et al. 2013). The black phenotype is notconsidered a typical characteristic and has never been ob-served in the past. Nevertheless, recently, in western Iran,especially Hamadan and Zanjan provinces, few black individ-uals were observed. The presence of a black wolf could be apossible sign of crossbreeding with domestic dogs or a naturalcombination of wolf alleles (Apollonio et al. 2004).

This study presents an application of mtDNA, microsatel-lite, and morphological markers to describe genetically andmorphologically the occurrence of black wolves in westernIran. The genetic composition of two black specimens at 15microsatellite markers were compared with those expected inpure specimens and in hybrids. The artificial neural networks(ANNs) method was applied to microsatellite data for classi-fication and assignment. The ANNs are universal approxima-tions of functions and have been successfully used in variousfields (Ermis et al. 2007; Zangeneh et al. 2010; Azadeh et al.2008), but less in ecology and population genetics (Cornuetet al. 1996; Aurelle et al. 1999). An ANN is expected to becapable of classifying individuals in populations belonging tothe same subspecies which are relatively similar genetically.

Materials and methods

Tissue collection

A total of 30 tissue samples was collected from roadkill andillegally hunted wolves. Two of the specimens were complete-ly dark. One (W13) was an adult female shot in July 2010 inthe region of Bahar, Hamadan province in western Iran (34′46′ N; 48′ 35′ E). Another (W16) was a subadult male, shot inSeptember 2010 in Ghidar, Zanjan province (36′ 40′N, 48′ 30′E). Village and purebred dog tissue samples (28) were obtain-ed from private owners, roadkill, and feral individuals. More-over, samples were obtained for four wolf–dog hybrids (H38,H39, H40, H41) from local people in western Iran. Finally, 16wolf skulls (two black wolves, nine adults, and five sub-adults), from natural museums and private owners, wereexamined.

Genetic analyses

DNA extraction

Total DNA was extracted from tissues using the AccuPrep®Genomic DNA Extraction Kit (Bioneer, South Korea).

Microsatellite

Polymerase chain reaction (PCR) for 15 microsatellite loci(Andersone et al. 2002; Randi et al. 2000; Verardi et al. 2006)was performed in 12-μl volumes with the AccuPower® PCRPreMix kit (Bioneer, South Korea) using a Perkin Elmer 9600thermal cycler. PCR products were separated on 8 % poly-acrylamide gels, visualized with silver staining, andphotographed with the Molecular Imager® Gel Doc™ XRsystem. Images were analyzed by Gel-Pro Analyzer 6.3.

mtDNA sequencing

The 544 base pairs (bp) of the inner mtDNA control regionwas PCR amplified in all samples with the external primers L-pro (5′-CGTCAGTCTCACCATCAACCCCCAAAGC-3′)and H-Phe (5′-GGGAGACTCATCTAGGCATTTTCAGTG-3′) (Douzery and Randi 1997). The PCR mix (AccuPower®PCR PreMix kit, Bioneer) in the volume of 25 μl included20 ng DNA, 1U of Euro Top DNA polymerase, 1.5 μMMgCl2, and distilled water. The PCR was carried out on anApplied Biosystems thermocycler with an initial step of de-naturation at 95 °C for 10 min, followed by 35 cycles of 94 °Cfor 30 s, 58 °C for 30 s, and 72 °C for 60 s, and, finally, by afurther extension step of 72 °C for 10 min. Amplificationproducts were purified from low-melting agarose gel withthe AccuPrep kit (Bioneer). Double-strand cycle sequencingwas performed by the BigDye Terminator Cycle Sequencing

J Appl Genetics

Kit version 3.1 (Applied Biosystems) according to the manu-facturer’s instructions, using the external primer L-pro andinternal primer H-576 (5′-TTTGACTGCATTAGGGCCGCGACGG-3′) (Randi et al. 2000). The results of sequencingwere registered in GenBank (KC540917–KC540944)(Aghbolaghi et al. 2014).

Microsatellite data analyses

Deviations from Hardy–Weinberg equilibrium (HWE) foreach locus per sample and linkage equilibrium (LE) betweenpairs of loci using the Markov chain method (Guo andThompson 1992) were computed separately for wolves anddogs using GENEPOP 4.1 (Raymond and Rousset 1995).Scoring errors, large allele dropouts, and null alleles werechecked using the program Micro-Checker version 2.2.3(Van Oosterhout et al. 2004). Significance levels were adjust-ed using the sequential Bonferroni method to apply multipletests on the same dataset (Rice 1989).

The classification of individuals of different populations isa prerequisite for the study of genetic interactions (Aurelleet al. 1999). The genetic structure and classification of thewolf and dog samples were investigated using ANNs.Khosravi et al. (2013) showed that, although there are numer-ous shared alleles between wolf and dog populations, micro-satellite markers are variable enough to separate the twospecies.

An ANN consists of interconnected identical simple pro-cessing units called neurons. Each neuron is connected withthe neighboring neurons by synapses. Each synapse can havea different level of weight of the connection (Heidari et al.2011). Each neuron integrates the signals received from theformer neurons and sends a new signal to the next ones. Thisnetwork of neurons and synapses stores the knowledge in a“distributed”manner: the information is coded as an electricalimpulse in the neurons and is stored by changing the weight(i.e., the conductivity) of the connections. A classical multi-layer feedforward network (MLFN) consists of three layers:an input layer, one or more hidden layers, and an output layer.An MLFN model that consists of a single hidden layer can beformulated as:

yk ¼ f 2 wk0 þX

j¼1

H

wkjf 1 wj0 þX

i¼1

I

wjixi

! !;

where xi is the input value to node i of the input layer,Hj is thehidden value to node j of the hidden layer, and yk is the outputat node k of the output layer (O). An input layer bias term I0=1with bias weightswj0 and an output layer bias termH0=1 withbias weights wk0 are included to permit adjustments of themean level at each stage (Heidari et al. 2011; Omid et al.2009).

In our study, incoming signals in the input layercorresponded to the code of samples based on 187 alleles.Each individual was scored as 0.0 (the allele was not ob-served), 0.5 (the individual was a heterozygote), or 1.0 (theindividual was a homozygote). The outgoing neuron of theoutput layer corresponded to the category where the studiedindividual was assigned by the network. Based on directobservation and morphological traits, the data were groupedinto three categories, including wolf, dog, and known hybrids.Therefore, the output layer consisted of three neurons. Theexpected scores for the three output neurons for the threepredefined classes were (1, 0, 0), (0, 1, 0), and (0, 0, 0),respectively. We run an ANN with these layers to train thenetwork. In this study, anMLFNwas trained based on an errorbackpropagation (BP) algorithm and gradient descent mo-mentum (GDM) for error minimization using NeuroSolutions5.07. We used a holdout procedure (Kohavi 1995) to test thevalidity of the network. Therefore, a dataset with knowncategories was divided into two parts. The first part was usedfor training the network and the second part was used fortesting. To test and evaluate the model, we used data that werenot used for learning. We used the holdout procedure, despitethe small dataset, because we were confident in thepreclassification of samples, the composition of samples waswell known, and there was no possibly of heterogeneity. Theperformance of the network was evaluated using the meansquared error (MSE), mean absolute error (MAE), and coef-ficient of determination (R2):

MSE ¼ 1

n

X

i¼1

n

YEstimated−YTarget

� �2R2

¼

X

i¼1

n

YEstimated−YTarget

� �2

X

i¼2

n

YEstimated−YTarget

� �2MAE ¼ 1

n

Xn

i¼1O1−Pij j;

where YTarget and YEstimated are the actual and estimated outputsignals for the test dataset and n is the number of test samples.When the best model was determined and the network wasverified as being well suited with no over-fitting of the learn-ing data, it was applied to unknown data for correct determi-nation (black samples).

After the best model was determined, each individualwas assigned to a category based on the scores in thethree output neurons. For example, individuals with anobserved score of one in a group (for example 0, 1, 0)could be considered as quite accurately classified. Howev-er, the interpretation of individuals with intermediatescores (0.5, for example) was not as easy as individualswith a score of zero to 0.1, which were considered to beincorrectly grouped.

J Appl Genetics

mtDNA data analyses

SeqScape 2.7 (Applied Biosystems) was used to reconcilechromatograms of complementary fragments and to alignsequences, using the ClustalW algorithm. The sequence datawere analyzed using the maximum composite likelihoodmod-el with the MEGA 5 program (Tamura et al. 2007). Haplo-types were obtained and FST were calculated in Arlequin 3.5(Excoffier et al. 2005). The phylogenetic tree, based on theobtained mtDNA sequences, was constructed in the MEGA 5program by the neighbor-joining (NJ) method. As anoutgroup, we used the corresponding control region sequenceof a jackal (AY289997).

Morphometric analyses

Based on del Zorro Rojo (2005), Milenković et al. (2006),Milenkovic et al. (2010), and Khosravi et al. (2012), 18 cranialand two mandible characters of two gray wolf skulls weremeasured (Fig. 1). Craniometric measurements were takenwith a digital caliper with 0.01 mm accuracy. The cranial

characters of black wolves were compared to 14 other adultand subadult skulls.

Results

Genetic analysis

Microsatellite

Some of the microsatellite loci showed a deviation fromHWE. LE between pairs of loci after sequential Bonferronicorrection showed that all comparisons were at LE in bothsample groups (except for one comparison in dogs and two inwolves). The examination of genotyping errors using Micro-Checker revealed no evidence for large allele dropout orstutter band scoring at any of the 15 loci. A total of 187 alleleswere scored in dog, wolf, and known hybrid specimens. Allmicrosatellites were polymorphic, showing five (locusCPH22 in dogs) and 17 (CPH8 in wolves) alleles per locus,with an overall average of nine alleles per locus. The number

Fig. 1 The wolf cranium and mandible dimensions employed in thisstudy: 1, cranial length; 2, greatest length of the nasals; 3, least length ofthe nasals; 4, maximum zygomatic width; 5, cranial width; 6, postorbitalconstriction; 7, frontal breadth; 8, distance between holes in the undersocket; 9, rostrum width; 10, basal length; 11, maximum width of

occipital condyles; 12, least diameter of the auditory bulla; 13, greatestbreadth of the palatine; 14, carnasil length; 15, height of the upper canine;16, length of P2 toM2; 17, length of the cheektooth row; 18, length of theupper tooth row; 19, mandible length; 20, length of P1 to M3

J Appl Genetics

of private alleles varied between the two populations (34 inwolves and 14 in dogs).

Various MLFNs included different hidden layer neuronsand arrangements were trained to find the best model predic-tion for classifying wolf, dog, and hybrid samples. A total of15 configurations with different numbers of hidden layers(one or two), different numbers of neurons for each of thehidden layers (2–15 for one hidden layer and 2–12 for twohidden layers), and different interunit connection mechanismswere designed and tested. Therefore, ANNs with 187 inputsand three outputs have been trained to estimate the networkparameters. The results of the MLFN trained for differentnetworks showed that, among the trained networks, the(187-8-16-3)-MLFN, a network having 187 input variables,eight and 16 neurons in two hidden layers, and three outputneurons, resulted in the best-suited model classifying wolf,dog, and hybrid specimens. The coefficient of determination(R2) between the output of the ANN model (estimated) andthe actual (observed) value for the three outputs was 0.90,0.50, and 0.20, respectively. For this configuration, the MSEand MAE for class 1 were 0.04 and 0.16, for class 2 0.13 and0.25, and for class 3 0.14 and 0.23, respectively.

The percentage of correctly classified individuals by hold-out was 100 % in the global comparison between wolf anddog samples. Based on the best network, the scores ofW13 forthe three classes (dogs, wolves, and hybrids) were 0.05, 0.90,and 0.05, respectively. This result showed that W13 wascorrectly grouped in class 2 (wolf). The score of W16 forthe three classes was 0.20, 0.74, and 0.1, respectively. Basedon this score, W16 was also grouped in class 2. This findingshowed that the black wolves were genetically close to wolfsamples and individual assignments based on autosomalmicrosatellites also indicated that the two black wolves werelocated in the wolf cluster.

mtDNA

In this study, 544 bp of mtDNA control region sequences wereobtained for the wolf and dog samples. Overall, 25 haplotypeswere identified, including 12 in dogs, ten in wolves, and threeshared haplotypes between wolves and dogs. The phylogenet-ic tree, constructed as shown in Fig. 2, separated wolves anddogs into two distinct groups. The two black individuals (W13and W16) were positioned in the wolf haplotype.

Morphological analysis

At first glance, the subadult black male skull (W16) lookedlike a typical dog skull. We observed an added tooth in thelower tooth row and completely dark general shade of the fur,especially on the head. The rostrum width (36 mm) andgreatest breadth of the palatine (32.1 mm) were clearly smallerthan even the minimum values for subadult male wolves.

However, other craniometric characters such as the length ofthe M2, zygomatic width, and postorbital width fell in therange of wolf skull measurements, suggesting a close relation-ship with wolf samples.

Contrary to the subadult male skull, the appearance of theblack female (W13) specimen did not deviate from a typicalwolf and was not craniologically different from pure wolves inthe region. This specimen had a large skull with completelydark fur, especially on the head and sides. The maximumwidth of occipital condyles (47.9 mm) and the least diameterof the auditory bulla (27.9 mm) were slightly larger than themaximum values for adult female wolves. However, othercraniometric characters fell in the range of measurements forthis species (Table 1; Fig. 3).

We did not observe a typical malformation, such as incom-pletely developed permanent teeth P1, spongy bony tissue inthe foramen infraorbitale, semicircular lines of the hind part ofthe forehead, and atypical appearance of the sutura frontalis, ineither of the black wolves. These findings showed that the twoskulls, both in appearance and in craniometric parameters, didnot deviate from the typical wolves; therefore, they cannot beidentified as hybrids.

Discussion

The existence of wolf–dog hybrids is rarely reported [seeAndersone et al. (2002), Randi and Lucchini (2002), Ciucciet al. (2003), and Verardi et al. (2006)]. While some traits likedewclaw, color pattern, and long tail can be considered as thesigns of hybridization between wolves and free-ranging dogs(Apollonio et al. 2004), wolf–dog morphological traits are notpredictable. Color pattern is one of the most commonly doc-umented variations among wolves. North American wolfpopulations, for instance, show different color patterns, rang-ing fromwhite in the Arctic regions to black coats observed inthe northwest USA (Brewster and Fritts 1995).

In Iran however, melanism is not considered a typicalcharacteristic. To understand whether two black wolves iden-tified in western Iran were hybrids or purebred, we analyzedthem both genetically and morphologically. ANNs providedimportant information about the genetic structure and classi-fication of wolf and dog populations. When samples weredefined well based on morphological traits and direct obser-vation, the results of ANN analyses showed that, when ap-plied to microsatellite data, neural networks give reliableresults. ANNs based on allele frequency and holdout proce-dures showed a clear distinction between wolf and dog pop-ulations. Based on the ANN results, individuals with interme-diate scores could be hybrids, and when such individuals arepresent in a population, the network is able to recognize them.The scores of two black specimens predicted by the best

J Appl Genetics

network grouped them into the wolf population with a highscore. ANNs showed that black wolves were genetically closeto wolf samples and individual assignments based on autoso-mal microsatellites showed that the two black wolvesbelonged to the wolf cluster.

Based on the results of the mtDNA analysis, althoughwolves and dogs share certain haplotypes, wolf and doggroups could be well discriminated. Some shared haplotypesbetween the two groups might be the result of interbreedingduring dog domestication or interspecific hybridization inprevious generations (Ardalan et al. 2011). Four individuals(H38, H39, H40, H41) identified as hybrids based on theresults of microsatellite analyses had three different haplo-types and belonged to the same haplogroup composed ofdog and wolf sequences. Two samples, H38 and H41, shareda common haplotype. The results of analyzing mtDNAmarkers, which are maternally inherited, points to the possi-bility that the hybrids were mothered by dogs and fathered bypurebred or hybrid wolves. Based on observations in the studyarea and interviews with local people, these four individuals(H38, H39, H40, H41) were the results of hybridization.Behavioral and genetic studies show that mating among var-ious genera of canid species is asymmetrical and differsamong species based on the direction and intensity of gene

flow (Vila and Wayne 1999). Since mtDNA markers deter-mine hybrid individuals, it is not possible to use them indetermining the exact intensity of the presence of hybridswithin wolf populations in nature (Vila and Wayne 1999).

The mtDNA results were in concordance with the results ofmicrosatellite analysis. The two dark wolves (W13 and W16)showed two separate haplotypes and were placed in twodistinct haplogroups. The results of mtDNA sequencing, amethod commonly used to trace maternal lineage in domesticand wild populations (Freeland 2005), showed that the blackcoat color does not indicate recent genetic flow betweenwolves and dogs.

Moreover, the morphometric results showed that unusualblack wolves, both in appearance and in craniometric param-eters, do not deviate from the typical wolf characters. Ourresults support evidence of no inevitable direct relationshipbetween the presence of coats darker than usual in wolves andrecent hybridization with dogs.

Our results support Randi and Lucchini (2002), who statedthat dark color in Italian wolves could have been fostered bythe past demographic decline and expansion after abottleneck. Apollonio et al. (2004) reported that 22 % ofobserved and 23 % of all dead wolves in a 3,300-km2 areawere completely black. Their analyses showed no evidence ofhybridization in ancestry and suggested that the occurrence ofthe black phenotype in this area may be derived from a naturalcombination of wolf alleles in coat-color-determining genes,and not necessarily from crossbreeding with the domestic dog.

There are many factors that determine coat color in dogs, amechanism which is quite complex (Sponenberg and Roth-schild 2001). Apollonio et al. (2004) suggested that it isunlikely that a single event of hybridization with dogs inrecent years in any case would produce a black wolf andthat black color in wolves is likely to result from acombination of dominant alleles. On the other hand,Anderson et al. (2009) compared the genes of wolves fromYellowstone National Park and the Canadian Arctic to thoseof domestic dogs and coyotes. They found that, in eachspecies, the black individuals had the same mutation, whichfirst arose around 45,000 years ago, and molecular analysisshowed that the oldest mutation happened in dogs, suggestingit originated in dogs and was then introduced to wolves andcoyotes through interspecific hybridization. Anderson et al.(2009) showed that the KB allele which codes for black coatcolor in wolves is more common in packs that inhabit foreststhan those occupying the tundra. These findings show thattraits selected in domesticated species can influence the mor-phologic diversity of their wild relatives.

Environmental factors are more important than geographicdistance in determining the genetic and morphometric varia-tions in wolf populations (Bennett 2010). Environmental gra-dients such as vegetation type and vegetation cover in ahabitat or type of prey available can influence the genetic

Fig. 2 Phylogenetic tree of mtDNA haplotypes. The dark specimens(W13 and W16) and the four probably hybrid individuals (H38, H39,H40, H41) are shown in ellipses and rectangles, respectively

J Appl Genetics

makeup of local wolf populations. Musiani et al. (2007) foundthat 93 % of wolves from tundra populations exhibited lightcoloration, whereas only 38 % of boreal coniferous forestwolves had this type of coloration. These findings showedthat genetic and phenotypic differentiations among wolvescan be caused by prey–habitat specialization rather than dis-tance or topographic barriers. Pilot et al. (2006) examined theeffect of ecological factors on the genetic structure of Euro-pean gray wolves. They found that the genetic differentiationamong local populations was correlated with climate, habitattype, and wolf diet composition. This result indicates thatecological processes may strongly influence the amount ofgene flow among populations. Carmichael et al. (2007) stated

that the genetic structure in wolves correlates strongly withhabitat type, and is probably determined by natal habitat-biased dispersal.

Although this study showed that the two black wolves didnot have a sign of hybridization in first or second past gener-ations, this result is not accurate enough to exclude the possi-bility of more ancient hybridization. Anderson et al. (2009)found no evidence of hybridization in black wolves using over48,000 single nucleotide polymorphisms (SNPs) and sug-gested that hybridization occurred at least hundreds of yearsago.

Wolves in Iran occupy a wide range of habitats and areabsent only in the central deserts and Dasht-e-Lut. Variation in

Table 1 Minimum and maximum values (mm) of 20 cranial and dental characters in ten purebred wolves and two black wolf specimens (female fromBahar in Hamadan province, male from Ghidar in Zanjan province, both in western Iran)

Measurements Subadult male (5) Female (9) W13 W16

Min Max Min Max Male Female

1 Cranial length 100.64 130.24 121.22 140.24 110.9 139.36

2 Greatest length of the nasals 71.32 82.22 82.98 100.46 77.2 89.32

3 Least length of the nasals 64.3 72.48 72 88.9 65.9 77.56

4 Maximum zygomatic width 95.1 128 118.7 138.24 110.5 136.36

5 Cranial width 66.28 75.22 69.54 86.55 72.72 78.9

6 Postorbital constriction 33.42 43.94 40.7 51.24 38.26 48.98

7 Frontal breadth 49.9 61.98 54.62 70.34 55.36 66.88

8 Distance between holes in the under socket 38.46 47.46 45.78 55.86 45.04 49.5

9 Rostrum width 39.08 43.64 39.22 55.86 36 48.46

10 Basal length 161.54 190.12 210.86 240.24 187.96 220.1

11 Maximum width of occipital condyles 35 40.22 39 47.34 36.12 47.9

12 Least diameter of the auditory bulla 20.22 24.9 21.98 27.0 22.4 27.9

13 Greatest breadth of the palatine 32.6 41.38 36.4 42 32.1 41.74

14 Carnasil length 20.1 23.4 23.4 26.5 22.6 25.5

15 Height of the upper canine 19.14 22.2 18.64 25.3 22.64 25.2

16 Length of P2 to M2 54.72 70.14 58.36 75.24 64.1 72.76

17 Length of the cheektooth row 65 78.74 74.68 85.12 73.16 80.12

18 Length of the upper tooth row 79.4 96.2 91.62 110.2 88.34 101.84

19 Mandible length 143.72 177.86 182.56 225.16 154.56 182.18

20 Length of P1 to M3 71.66 87.44 78.58 96.26 78.3 82.82

Fig. 3 Lateral view of the skull of two black wolves. Subadult male (W16; a) and adult female (W13; b)

J Appl Genetics

color patterns might be due to the great diversity of habitatswhich is caused by the existence of two large water bodies inthe north and south and the vast mountain ranges expanding inthe north and west. In western Iran, due to the existence of theZagros mountain range, variation in color can most probablybe resulted from ecological traits.

In conclusion, the results of this study showed that blackcoat in wolves in western Iran is not necessarily the result ofgenetic pollution by domestic genes in recent generations andcould be caused by habitat variation, local adaptation, andnatural combination of wolf alleles and introgression of theKB allele from dogs into wolves due to past hybridization withfree-ranging dogs. Considering the limited sampling, howev-er, the obtained results should be interpreted and generalizedwith caution. Further investigations (SNP data) are needed forassessing more ancient hybridization between wolf and dogpopulations in Iran.

Acknowledgments This research was supported financially by the IranDepartment of Environment, Hamedan Provincial Office. We thank AliShaabani, Vahid Nouri, and Shahabaddin Montazami for their help. Theauthors also thank the anonymous referees for their valuable commentson an earlier version of this article.

References

Aghbolaghi MA, Rezaei HR, Scandura M, Kaboli M (2014) Low geneflow between Iranian Grey Wolves (Canis lupus) and dogs docu-mented using uniparental genetic markers. Zool Middle East 60:95–106

Anderson TM, vonHoldt BM, Candille SI, Musiani M, Greco C, StahlerDR, Smith DW, Padhukasahasram B, Randi E, Leonard JA,Bustamante CD, Ostrander EA, Tang H, Wayne RK, Barsh GS(2009) Molecular and evolutionary history of melanism in NorthAmerican gray wolves. Science 323:1339–1343

Andersone Ž, Lucchini V, Ozoliņš J (2002) Hybridisation betweenwolves and dogs in Latvia as documented using mitochondrial andmicrosatellite DNA markers. Mamm Biol Z Säugetierkd 67:79–90

ApollonioM,Mattioli L, ScanduraM (2004) Occurrence of black wolvesin the Northern Apennines, Italy. Acta Theriol 49:281–285

Ardalan A, Kluetsch CFC, Zhang A-B, Erdogan M, Uhlén M,Houshmand M, Tepeli C, Ashtiani SRM, Savolainen P (2011)Comprehensive study of mtDNA among Southwest Asian dogscontradicts independent domestication of wolf, but implies dog–wolf hybridization. Ecol Evol 1:373–385

Aurelle D, Lek S, Giraudel J-L, Berrebi P (1999) Microsatellites andartificial neural networks: tools for the discrimination between nat-ural and hatchery brown trout (Salmo trutta, L.) in Atlantic popula-tions. Ecol Model 120:313–324

Azadeh A, Ghaderi SF, Sohrabkhani S (2008) A simulated-based neuralnetwork algorithm for forecasting electrical energy consumption inIran. Energ Policy 36:2637–2644

Bennett K (2010) Genetic conservation of the Grey Wolf. http://kendellbennett.writersresidence.com/system/attachments/files/4895/original/wolf_conservation_paper.pdf

Boitani L (1992) Wolf research and conservation in Italy. Biol Conserv61:125–132

Brewster W, Fritts S (1995) Taxonomy and genetics of the gray wolf inwestern North America: a review. In: Carbyn LN, Fritts SH, Seip

DR (eds) Ecology and conservation of wolves in a changing world.Canadian Circumpolar Institute, University of Alberta, Edmonton,pp 353–374

Carmichael LE, Krizan J, Nagy JA, Fuglei E, Dumond M, Johnson D,Veitch A, Berteaux D, Strobeck C (2007) Historical and ecologicaldeterminants of genetic structure in arctic canids. Mol Ecol 16:3466–3483

Ciucci P, Lucchini V, Boitani L, Randi E (2003) Dewclaws in wolves asevidence of admixed ancestry with dogs. Can J Zool 81:2077–2081

Cornuet J-M, Aulagnier S, Lek S, Franck S, Solignac M (1996)Classifying individuals among infra-specific taxa using microsatel-lite data and neural networks. C R Acad Sci III 319:1167–1177

del Zorro Rojo EM (2005) Morphometric examination of red fox (Vulpesvulpes) from the Van-Yoncatepe necropolis in eastern Anatolia. Int JMorphol 23:253–260

Douzery E, Randi E (1997) The mitochondrial control region ofCervidae: evolutionary patterns and phylogenetic content. MolBiol Evol 14:1154–1166

Ermis K, Midilli A, Dincer I, Rosen MA (2007) Artificial neural networkanalysis of world green energy use. Energ Policy 35:1731–1743

Excoffier L, Laval G, Schneider S (2005) Arlequin (version 3.0): anintegrated software package for population genetics data analysis.Evol Bioinformatics Online 1:47

Freeland JR (2005) Molecular ecology. John Wiley & Sons, ChichesterGeffen ELI, AndersonMJ,Wayne RK (2004) Climate and habitat barriers to

dispersal in the highly mobile grey wolf. Mol Ecol 13:2481–2490Guo SW, Thompson EA (1992) Performing the exact test of Hardy–

Weinberg proportion for multiple alleles. Biometrics 48:361–372Heidari MD, Omid M, Akram A (2011) Application of artificial neural

network for modeling benefit to cost ratio of broiler farms in tropicalregions of Iran. Res J Appl Sci Eng Technol 3:546–552

Khosravi R, Kaboli M, Imani J, Nourani E (2012) Morphometric varia-tions of the skull in the GrayWolf (Canis lupus) in Iran. Acta Theriol57:361–369

Khosravi R, Rezaei HR, Kaboli M (2013) Detecting hybridization be-tween Iranian wild wolf (Canis lupus pallipes) and free-rangingdomestic dog (Canis familiaris) by analysis of microsatellitemarkers. Zool Sci 30:27–34

Kohavi R (1995) A study of cross-validation and bootstrap for accuracyestimation and model selection. IJCAI 14:1137–1145

Mech LD (1970) The wolf: the ecology and behavior of an endangeredspecies. American Museum of Natural History, New York

Milenković M, Habijan-Mikeš V, Matić R (2006) Cases of spontaneousinterbreeding of wolf and domestic dog in the region of SoutheastBanat. Arch Biol Sci 58:225–231

Milenkovic M, Šipetic VJ, Blagojevic J, Tatovic S, Vujoševic M (2010)Skull variation in Dinaric-Balkan and Carpathian gray wolf popu-lations revealed by geometric morphometric approaches. J Mammal91:376–386

Musiani M, Leonard JA, Cluff H, Gates CC, Mariani S, Paquet PC, VilàC, Wayne RK (2007) Differentiation of tundra/taiga and borealconiferous forest wolves: genetics, coat colour and association withmigratory caribou. Mol Ecol 16:4149–4170

Omid M, Baharlooei A, Ahmadi H (2009) Modeling drying kinetics ofpistachio nuts with multilayer feed-forward neural network. DryTechnol 27:1069–1077

Pilot M, Jedrzejewski W, Branicki W, Sidorovich VE, Jedrzejewska B,Stachura K, Funk SM (2006) Ecological factors influence populationgenetic structure of European grey wolves. Mol Ecol 15:4533–4553

Randi E, Lucchini V (2002) Detecting rare introgression of domestic doggenes into wild wolf (Canis lupus) populations by Bayesian admix-ture analyses of microsatellite variation. Conserv Genet 3:29–43

Randi E, Lucchini V, Christensen MF, Mucci N, Funk SM, Dolf G,Loeschcke V (2000) Mitochondrial DNA variability in Italian andEast European wolves: detecting the consequences of small popu-lation size and hybridization. Conserv Biol 14:464–473

J Appl Genetics

Raymond M, Rousset F (1995) An exact test for population differentia-tion. Evolution 49:1280–1283

Rice WR (1989) Analyzing tables of statistical tests. Evolution 43:223–225

Sponenberg D, Rothschild MF (2001) Genetics of coat colour and hairtexture. In: Ruvinsky A, Sampson J (eds) The genetics of the dog.CABI Publishing, New York, pp 61–85

Tamura K, Dudley J, Nei M, Kumar S (2007) MEGA4: MolecularEvolutionary Genetics Analysis (MEGA) software version 4.0.Mol Biol Evol 24:1596–1599

Tsuda K, Kikkawa Y, Yonekawa H, Tanabe Y (1997) Extensive inter-breeding occurred among multiple matriarchal ancestors during thedomestication of dogs: evidence from inter- and intraspecies poly-morphisms in the D-loop region of mitochondrial DNA betweendogs and wolves. Genes Genet Syst 72:229–238

Van Oosterhout C, Hutchinson WF, Wills DPM, Shipley P (2004)MICRO‐CHECKER: software for identifying and correctinggenotyping errors in microsatellite data. Mol Ecol Notes 4:535–538

Verardi A, Lucchini V, Randi E (2006) Detecting introgressive hybridizationbetween free‐ranging domestic dogs and wild wolves (Canis lupus) byadmixture linkage disequilibrium analysis. Mol Ecol 15:2845–2855

Vila C, Wayne RK (1999) Hybridization between wolves and dogs.Conserv Biol 13:195–198

Vila C, Sundqvist A-K, Flagstad Ø, Seddon J, Kojola I, Casulli A, SandH, Wabakken P, Ellegren H (2003) Rescue of a severelybottlenecked wolf (Canis lupus) population by a single immigrant.Proc Biol Sci 270:91–97

Wayne RK, Lehman N, Fuller TK (1995) Conservation genetics of thegray wolf. In: Carbyn L, Fritts S, Seip D (eds) Ecology and conser-vation of wolves in a changing world. Canadian CircumpolarInstitute, Occasional Publication No. 35, Edmonton, pp 399–407

Zangeneh M, Omid M, Akram A (2010) Assessment of machineryenergy ratio in potato production by means of artificial neuralnetwork. Afr J Agric Res 5:993–998

Ziaie H (2008) A field guide to mammals of Iran, 2nd edn. Iran WildlifeCenter, Tehran

J Appl Genetics