Genetic variation analysis of the Bali street dog using microsatellites

Background Approximately 800,000 primarily feral dogs live on the small island of Bali. To analyze the genetic diversity in this population, forty samples were collected at random from dogs in the Denpasar, Bali region and tested using 31 polymorphic microsatellites. Australian dingoes and 28 American Kennel Club breeds were compared to the Bali Street Dog (BSD) for allelic diversity, heterozygosities, F-statistics, GST estimates, Nei's DA distance and phylogenetic relationships. Results The BSD proved to be the most heterogeneous, exhibiting 239 of the 366 total alleles observed across all groups and breeds and had an observed heterozygosity of 0.692. Thirteen private alleles were observed in the BSD with an additional three alleles observed only in the BSD and the Australian dingo. The BSD was related most closely to the Chow Chow with a FST of 0.088 and also with high bootstrap support to the Australian dingo and Akita in the phylogenetic analysis. Conclusions This preliminary study into the diversity and relationship of the BSD to other domestic and feral dog populations shows the BSD to be highly heterogeneous and related to populations of East Asian origin. These results indicate that a viable and diverse population of dogs existed on the island of Bali prior to its geographic isolation approximately 12,000 years ago and has been little influenced by domesticated European dogs since that time.


Background
Bali, a province of the Republic of Indonesia, is an island just 87 km from north to south and 142 km from east to west and home to more than 2.9 million people [1]. Approximately 800,000 stray dogs (Fig. 1) also live on the island based on a survey conducted by the Bali Street Dog Foundation (personal communication). Only a small percentage of these dogs live in homes or are provided routine veterinary care [2].
More than 90% of the residents of Bali are Hindu [3] with myth and ritual playing a vital part of daily life [1]. The dog is also an important part of Balinese life and mythology. A popular tale from the Mahabharata [4] describes King Yudisthira's journey to Heaven's Gate, and his love for a dog that befriended him on his arduous and tragic journey (Fig. 2). As a direct result of such mythology, BSDs are treated with a degree of reverence and are often provided ceremonial food offerings [2]. The deliberate killing of street dogs is not typically practiced, because Balinese people believe that all things should be allowed to die naturally [2]. These cultural mores have contributed to the current overpopulation of dogs on the island.
As a result of overpopulation, many BSDs suffer from chronic skin diseases, internal parasites, parvo-and distemper-virus infections, and malnutrition. In an effort to reduce the dog population and to care for their medical needs, the Bali Street Dog Foundation (Yayasan Yudisthira Swarga) was founded in 1998 [2]. They provide emergency care, treatment for skin disease and parasites, sterilization, public education on the plight of feral dogs, and improved veterinarian training. Twenty to 30 dogs are sterilized each day, with more than 9,000 dogs sterilized to date.
The BSD population is of interest for both its genetic diversity and historical relationships. It is also a population that has bred more or less randomly for thousands of years with limited genetic influx, due mainly to geographic barriers and a strict rabies control program in effect since 1926. The present study is concerned with the genetic diversity of this unique canine population and its relationship to other canine subpopulations in Asia and throughout the world. Data presented herein was derived from the DNA testing of 40 BSD samples from the Denpasar city region of Bali with 31 polymorphic microsatellite loci. The genetic diversity of the BSD was compared to that of the Australian dingo and 28 American Kennel Club (AKC) breeds.

Locus diversity
Analysis of locus diversity across all 30 subpopulations revealed that the number of observed alleles ranged from six to 20 with a total of 366 for all loci (Table 1). Overall heterozygosity of the loci was high, with an average of 0.779, and all but four loci having H T values greater than 0.700. Average H S was 0.577 for the 30 subpopulations, with all but three loci having H S values greater than 0.500. The H S and H T values were closest for C23.123 and farthest for C22.279 and C10.404. HWE analysis revealed that all but one locus had at least one population out of equilibrium for the 30 populations sampled. C01.424, C31.646 and CPH16 had 7 populations out of HWE and AHT130 did not have any populations with p values below 0.05. The level of locus diversity attributable to subpopulation structure was evaluated with two statistics -R ST and F ST . Both statistics gave similar average values at 0.230 and 0.236 respectively. However, R ST ranged from 0.098 to 0.486 while F ST ranged from 0.179 to 0.328. Figure 1 Typical Balinese street dogs. Their phenotypic appearance is similar to that described for randomly breeding feral dog subpopulations in other parts of the world.

Typical Balinese street dogs
The Story of Yudisthira [4] Figure 2 The Story of Yudisthira [4].

Bali street dog diversity
Overall, the BSD was the most genetically diverse population surveyed here, displaying 239 total alleles out of the 366 seen in all 30 subpopulations or 65.3% of the total observed alleles ( To understand the loss of approximately 24 observed alleles after bootstrapping, the allele frequencies for the BSD at each locus was examined. While the BSD had the highest number of alleles, they also had the highest number of alleles with a frequency below 5% (67 out of 239, data not shown). F IS estimates were calculated to assess the level of inbreeding for each subpopulation ( Table 2). The BSD had the lowest value at 0.097 and the Australian dingo had the highest at 0.194 with the average of AKC breeds at 0.137.

Private alleles
Allele frequency analysis also revealed that 10 private alleles were observed in the BSD as well as three alleles shared only with the Australian dingo ( Table 3). The majority of private alleles in the BSD were below 5% in frequency with the exception of AHT121 where the private alleles had a combined frequency of 18.75%. The BSD and the Australian dingo also shared three alleles not seen in any AKC breed at locus C10.404 with a combined frequency of 16.25% in the BSD and 75% in the Australian dingo.     genome that are neutral to either environmental or human selection.

Genetic distance relationships
Further distance analysis was performed for all 31 loci between all 30 subpopulations using both Nei's DA distance and pairwise F ST estimates (Table 4) Genetic distance relationships amongst the five Asian subpopulations were further explored using neighbor-joining dendograms with four non-Asian subpopulations for comparison (Fig. 3). The BSD, Chow Chow, Australian dingo and Akita clustered together in 90% of the trees. The BSD, Chow Chow and Australian dingo further clustered in 87% of the trees. The BSD and Australian dingo maintained their relationship within the larger cluster in 84% of the trees. In the remainder of the tree, the Rhodesian Ridgeback, Greyhound, Airedale Terrier and Borzoi maintained a relationship in 51% of the trees and the Airedale Terrier/Borzoi cluster was seen in 63% of the trees. The Pug did not maintain a relationship with any other breed in this analysis, but was intermediate to the Asian and non-Asian subpopulations.

Population diversity
Microsatellites have been previously used to assess genetic diversity and relationships in feral dog subpopulations [6,7]. Kim et al. [6] found Given the size of the island of Bali, it is extraordinary that 800,000 feral dogs can thrive and maintain such high levels of genetic diversity. Of all the subpopulations surveyed here, the BSD has the highest number of observed alleles, the highest heterozygosity, the fewest number of loci out of Hardy-Weinberg equilibrium and the lowest F IS . Even after adjusting for sample size, the BSD maintains their status as the most heterogeneous population in the study. Unlike the Australian dingo which exhibits a much lower level of diversity, the BSD findings suggest either a large founding population on Bali and/or a consistent genetic influx since the geographic isolation of ~12,000 years ago. This data also supports that the BSD appears to approximate a randomly breeding population with little selection pressure.
When comparing the heterogeneity of the BSD to that observed within the AKC breeds some caveats should be addressed. One may initially expect long established, well-defined dog breeds to be much less heterogeneous than reported here. While some breeds do have a low H E , such as the Boxer with a H E of 0.320, breeds like the Jack Russell Terrier have a high H E of 0.713 and overall their H E is higher than that of the dingo. Of first note, the selection of the dogs that contribute to a breed composition mostly occurs prior to official breed recognition primarily by genetic drift due to geographic isolation and selection for particular working or physical characteristics. After official breed recognition future breeding choices are based primarily on the availability of sires and dams that approximate the breed standard. As a result, there is a founding population that proceeds to breed mostly by convenience. Also, many breed standards have changed considerably over the years resulting in retention of a certain level of diversity within each breed, some breeds retaining much more than others. Finally, dogs comprising the comparison AKC breeds were sampled from across the United States, removing any geographical bias of the genotypes observed and slightly elevating the heterozygosities.

Locus diversity
The average allelic diversity of the loci used in the present study was 11.8 alleles per locus, versus 7.75 in the Kim et al work [6]. However, the average number of alleles observed is 4.6 among the subpopulations in the present study and the average H T is 0.577. The average values for the 11 subpopulations surveyed in the Kim et al [6] work were 4.34 and 0.547, respectively. The higher total allelic diversity in the present study is likely due to the fact that nearly three times more subpopulations were studied. R ST and F ST values were nearly identical across all subpopulations and all loci, indicating that approximately 23% of the differences observed in allele frequencies can be attributed to differences between subpopulations. F ST provides an unbiased estimate of genetic drift between subpopulations by comparing alleles identical by state. R ST takes advantage of the stepwise mutation model, which assumes that mutations most often occur as whole repeat unit losses or gains from the original allele size. As a result, the number of mutations provides an estimate of time from divergence. It is interesting, therefore, to compare R ST and F ST values by locus. Eighteen of the 31 loci studied have an R ST to F ST ratio greater than 1.1 (Table 1) indicating that the populations have been separated for a sufficient amount of time for mutations to impact genetic structure. An interesting exception is observed at CPH16 where the ratio is 0.420. CPH16 may have a mutation pattern where both stepwise additions and subtractions occur at equal and high frequency. Of note, the average pairwise R ST value between the BSD and each of the 29 comparison subpopulations is 0.056 at locus CPH16. The highest R ST to F ST ratio occurs at locus CPH03 with a value of 1.724. Interestingly, the BSD and the Australian dingo have a pairwise R ST value of 0.017 at CPH03, whereas the average value of the BSD compared to the other 28 subpopulations has a value of 0.254. The distance between the BSD and the Australian dingo at CPH03 may support that those two populations were isolated most recently from each other relative to the other 28 subpopulations.

Bali street dog origin
The origin of the people of Bali is clouded by myth and a scarcity of archeological findings. Therefore, the origin of the dog on Bali is also speculative. Nonetheless, a hypothesis can be formed based on known human and dog histories. Current evidence points to an early migration of humans from Africa through Indonesia and into Australia approximately 60,000 to 70,000 years ago [8,9]. Recent excavations have also revealed that there was a great expansion into Indonesia from China between 4,000 and 5,000 years ago that could have contributed to a population pre-existing on Bali [1]. Supportive evidence that Indonesia was populated prior to 5,000 years ago is a higher degree of heterogeneity in the Indonesian population than seen in the North Asian population, suggesting that the Indonesia was populated earlier than regions to the North [10]. The "Slow Boat Model" for the peopling of Polynesia also suggests a prolonged mixing of Southeast Asians with Indonesians, which predated migration to the East [11]. In short, Indonesia appears to be a human genetic melting pot with genetic influences over tens of thousands of years.
The dog on the island of Bali may also be a parallel "canine genetic melting pot." While the domestication date of the dog is in much dispute [12], approximately 14,000 years ago is accepted as a late date. During the earliest human migrations through Indonesia however, it is highly possible that wolf packs or feral dogs traveled the same routes, establishing a feral population on Bali in the process. Even if humans were not capable of taming the dog at that time, dogs could still have benefited from close proximity to humans. Figure 4 shows a superimposition of the proposed geographic origin for five Asian and four non-Asian dog subpopulations presented herein and the major theorized human migration routes. It is noteworthy that the BSD, Chow Chow and Australian dingo, related breeds by genetic analysis, all share one proposed human migration route.
If a feral dog population was established on the island of Bali more than 14,000 years ago, then that population became isolated approximately 10,000 years ago when the sea levels drastically rose, submerging the land bridges of the Indonesia archipelago [13]. Geographic isolation was unlikely to have been absolute; genetic diversity of the BSD was invariably enhanced at various times by the influx of new dogs. At the time humans migrated to Indo-nesia from China, dogs were known to be domesticated and undoubtedly accompanied people as companions [17]. Mitochondrial DNA sequencing evidence suggests that the dingo was introduced into Australia about that time from the Indonesian archipelago [15,8,9]. Bali's documented history of repeated war and trade spanning the last 2,000 years [1, 16,17] represents actions that are often associated with the introduction of new animals. Indeed, a somewhat free movement of dogs probably occurred up to 1926, when the import of dogs to Bali was greatly curtailed as a means to prevent the introduction of rabies trast to the Australian dingo population, which appears to have undergone a severe population bottleneck or founder effect based on microsatellite alleles and mtDNA [18], the BSD population maintains a high level of genetic variation. There is no evidence for a genetic bottleneck or small founding population for the BSD.
The relatedness of the BSD to the Australian dingo and the Chow Chow is evidenced by common unique alleles and allele frequencies despite the very different levels of genetic diversity between the subpopulations. According to the hypothesis presented herein, one could imagine that feral dog subpopulations were established throughout Indonesia with much mixing until ~12,000 years ago. At that time, each population became closed with little influx of new genetic material until humans migrated south from Asia between 4,000 and 5,000 years ago. The degree of influx since that period would have been influenced by the frequency of trade and conflict, factors determined by accessibility, available natural resources, and political structure. The island of Bali is historically a less visited island than it's neighbor Java and therefore the indigenous dog population would have been subjected to less influence.

Conclusions
This study into the diversity and relationship of the BSD to other domestic and feral dog populations shows the BSD to be highly diverse and related to populations of East Asian origin. These results indicate that a viable and diverse population of dogs existed on the island of Bali prior to its geographic isolation approximately 12,000 years ago and has been little influenced by domesticated European dogs since that time. It would be of interest to study feral subpopulations on other islands in the archipelago to determine if the same level of diversity is observed elsewhere, or if the situation on Bali is truly unique. Y-chromosome, mitochondrial and MHC marker typing on the BSD, as well as feral dogs from other regions, would help to determine if indeed dogs followed the same migration routes as their likely human companions.

Animal selection
BSDs were randomly captured and taken to a BSD Foundation field clinic for treatment or sterilization and simultaneously sampled for DNA collection with buccal swabs. Familial relationships of the BSDs sampled could not be easily determined; therefore the sample population was doubled (40 vs. 19-20 samples) over that of other study groups. Blood samples from the Australian dingo were taken from captive animals in Australia. Australian dingoes were known to be unrelated by at least one generation. Dogs from 28 American Kennel Club (AKC) breeds, equally representing the AKC group designations, were sampled with buccal swabs for a previous study [19]. Twenty dogs were tested for each breed, with the exception of two breeds (Doberman Pinscher and the Border Collie) that comprised 19 individuals. The

Marker selection
Thirty-one of the 100 microsatellites multiplexed into 12 PCRs by the Veterinary Genetics Laboratory [20] had been previously used to evaluate the Australian dingo samples (unpublished data). For comparison purposes, those same 31 microsatellites were selected for use in the present study. All markers but one (PEZ02) were mapped on either the 1999 canine genetic linkage map [21] or the Radiation hybrid map [22]. Loci selected for study represented 25 of the 38 autosomes of the dog, with five autosomes represented by two loci. The average distance for the markers on chromosomes CFA06, CFA11, CFA20 and CFA23 is 23.5 cM and 23.4 Mb between AHT139 and RVC1 on CFA15. As a result, only 25 loci are known to be unlinked. PEZ02 has not been mapped and may be linked to a marker in the study.
Forward primers were synthesized and dye labeled with either Fam, Hex or Vic, or Tamra or Ned (Applied Biosystems, Inc. (ABI), Foster City, CA). Reverse primers were synthesized by Operon (Alameda, CA). Primer sequences and concentrations for all markers are available as Additional file 1.

Sample preparation and PCR conditions
BSD and AKC breed DNA was derived from buccal cells harvested from the inside of the cheek with nylon bristle cytology brushes (Medical Packaging Corp., Camarillo, CA). Samples were collected by owners or field volunteers and submitted directly to the laboratory. DNA was extracted by heating a single swab for 10 min at 95°C in 400 µl 50 mM NaOH and then neutralized with 140 µl 1 M Tris-HCl, pH 8.0. Australian dingo DNA was extracted from blood using a standard sodium hydroxide digest.
A 2 µl aliquot of extract was used in each PCR which equates to approximately 50 ng DNA. All markers and DNAs were amplified with a PCR reagent mix of 1X PCR buffer (ABI), 4.17 mM MgCl 2 , 200 µM of each dNTP (Hoffmann-La Roche Inc, Nutley, NJ), 0.6 unit AmpliTaq (ABI), and 2% DMSO (Sigma) then covered with 15 ul Chill-out™ Liquid Wax (MJ Research, Inc., Waltham, MA) to prevent evaporation. One of five thermal cycler programs was used for each primer mix ranging from 56° to 64° degrees for the annealing temperature. All PCR work was done in polycarbonate 96-well v-bottom microtiter plates (USA Scientific, Ocala, FL) on MJ Research PTC-100 thermal cyclers (MJ Research, Inc., Waltham, MA). Protocols are also available in Additional 1.

Gel electrophoresis conditions and DNA fragment analysis
One µl aliquots of PCR product were mixed with 2 µl Fluorescent Ladder (CXR) 60-400 (Promega 400) or Internal Lane Standard 600 (Promega 600) (Promega, Madison, WI) fluorescent size standard, denatured on MJ Research PTC-100 thermal cyclers for three minutes at 95°C, then held at 5°C or placed on ice for at least one minute before gel loading. Two µl aliquots were then loaded onto a 6% denaturing polyacrylamide gel and run on an ABI 377 Automated Sequencer using ABI 10" × 7 1/8" short plates (12 cm). Gels were run at 1.10 kV (constant) voltage, 60.0 mA current, 200 W power, 51°C and 40.0 mW (constant) laser power for up to 2 hours when using Promega 400, and up to 3 hours using Promega 600. DNA fragment analysis was performed with in-house designed STRand software [23], which replaces ABI Genotyper and Genescan software. This data was then transferred to an in-house database compatible with the STRand software.

Statistical analysis
Allelic diversity and observed heterozygosities (H O ) were determined by direct counting for each of the 30 subpopulations. Hardy-Weinberg equilibrium (HWE) tests were performed using Genepop version 3.4 [24]. Pairwise F ST estimates and subpopulation expected heterozygosities (H E ) for the 30 breeds or dog groups were performed using Genepop version 3.4 [24]. F IS estimates (inbreeding coefficient of each subpopulation) for each allele following Weir and Cockerham [25] were calculated using Genepop version 3.4 and are presented as averages across all loci.
Gene diversity or total population heterozygosity (H T ) and its associated parameters, H S (average heterozygosity among subpopulations) and G ST (coefficient of genetic differentiation), were calculated across all loci using the public domain software, DISPAN [26]. Two additional measures of variance, F ST [25] and R ST [27,28] were calculated using Genepop version 3.4. A pairwise genetic distance matrix using Nei's DA distance was also created using DISPAN with bootstrapping. Genotype data for all populations is available in Additional file 2.

Phylogenetic tree construction
Allele frequencies were used to compute a matrix of genetic distances [29], which were then used to construct a phylogenetic tree of relationships among 5 Asian and 4 non-Asian dog subpopulations. Takezaki's [30] POPTREE program was used to create a neighbor joining tree using DA distances with 1000 bootstrap replications. The output of POPTREE was then converted to the New Hampshire format for editing in the stand alone program TREEVIEW version 1.6.6 [31] and bootstrap values were added.