Abstract

Zoo and aquarium breeding programs rely on accurate pedigrees to manage the genetics and demographics of captive populations. Breeding recommendations are often encumbered, however, by unknown parentage. If an individual has any amount of unknown ancestry, the relationships between that individual and all other individuals in a population are ambiguous, and breeding recommendations cannot be tailored to maximize genetic diversity and minimize inbreeding. In those situations, breeding program management might be improved by the incorporation of molecular data. We developed microsatellite markers for the parma wallaby (Macropus parma) and investigated how genetic data might be used to improve the management of the captive population. The parma wallaby is a small marsupial found in fragmented forests near the coast of New South Wales, Australia. Because the species is of conservation concern, the captive population in North America is managed by recurring breeding recommendations. The effectiveness of the population's management is hampered, however, because over half of the individuals have some amount of unknown ancestry. We used microsatellite data to resolve unknown parentage, described how molecular estimates of relatedness might inform future breeding recommendations, and used computer simulations to investigate how molecular estimates of relatedness among founders might contribute to the genetic management of the population. Our results indicated that microsatellite appraisals of parentage were useful with respect to clarifying pedigrees but that molecular assessments of founder relatedness provided very marginal benefits with regard to the preservation of genetic diversity and the avoidance of inbreeding.

Captive breeding programs sponsored by zoos and aquaria generally aim to maintain demographically stable, self-sustaining populations that retain genetic diversity and accumulate limited inbreeding (Foose and Ballou 1988; Hedrick and Miller 1992; Lacy 1994; Ballou and Lacy 1995). These goals are achieved through recurring breeding recommendations, which are based on each population's unique history and future needs. Pedigree analyses, which are the foundation of most captive breeding programs, are used to produce breeding recommendations intended to meet a population's demographic and genetic goals. The best breeding strategies for captive populations to meet genetic goals are those that minimize a population's average kinship (Ballou and Lacy 1995; Fernandez and Toro 1999; Sonesson and Meuwissen 2001), with the kinship (f) between 2 individuals being the probability that 2 alleles at a given locus, 1 randomly drawn from each individual, are identical by descent from a common ancestor (Falconer 1981). An individual's mean kinship (mk; Ballou and Lacy 1995) is a measure of an individual's genetic distinctiveness and is the average of f values between that individual and all living individuals in the population, including itself forumla. To minimize a population's average kinship, captive breeding programs breed genetically underrepresented individuals with low mks. Because captive breeding programs rely on pedigree analyses for developing breeding recommendations, captive population management is hampered if a population's pedigree is only partially known and neither f values nor mks can be calculated accurately. Algorithms for estimating f values and mks from only the known parts of incomplete pedigrees have been developed (Ballou and Lacy 1995), but breeding recommendations based on those estimates are considered suboptimal because the estimates are less accurate than equivalent measures of relationship calculated from complete pedigrees. Still, those algorithms currently provide the only option for conducting pedigree analyses on incomplete pedigrees, short of either assuming undocumented parents were unrelated population founders or excluding all animals with incompletely known ancestries from the analyses.

The captive parma wallaby (Macropus parma) population in North America is a typical example of a captive breeding program being managed by pedigree analyses. The population is currently managed through a Population Management Plan (PMP) by the Association of Zoos and Aquariums, and like most actively managed captive breeding programs, PMPs receive regular demographic and genetic management. Parma wallabies are small marsupials native to Australia, and the species is currently found in fragmented forest areas near the coast of New South Wales. Due to continuing habitat degradation and low census size, the parma wallaby is listed under various threat categories by the government of New South Wales (vulnerable; Parliament of New South Wales [1995]), the U.S. Fish and Wildlife Service (2005) (endangered; Endangered and Threatened Wildlife and Plants), and the IUCN (2007) (low risk: near threatened; IUCN Red List of Threatened Species). The parma wallaby was declared extinct in 1957 (Maynes 1974), but an introduced population of parma wallabies was subsequently discovered on Kawau Island, New Zealand, in 1965 (Ride 1970; Wodzicki and Flux 1971), and remnant populations of the species were eventually rediscovered in Australia in 1966 (Ride 1970; Wodzicki and Flux 1971). Although the species remains of conservation concern in its native Australia, the exotic parma wallabies on Kawau Island are considered invasive pests. The captive parma wallaby population in North America was founded in the late 1960s by animals originating from Kawau Island. Periodic imports to North America from the island have continued to occur, and to date, no population founders have been imported from Australia.

The pedigree of the captive parma wallaby population in North America is incomplete, and the relationships among its founders are unknown. As of January 2007, the PMP population included 157 animals, and over half of those individuals (90 total) had varying amounts of unknown ancestry. Some of the missing information is deep in the pedigree and stems from a time when zoological records were inconsistently maintained. Unknown parentage also has been perpetuated throughout the recent pedigree partly because the species is occasionally housed in groups of multiple males, making paternity in those situations uncertain. Additionally, in October 2003, the PMP population received 22 new potential founders (wild-caught animals that have not yet contributed descendants to the living, captive population; Lacy 1995) from Kawau Island. The population of parma wallabies on the island is currently of moderate size but was founded by only a few individuals. Because nothing is known about the genetics of the Kawau Island population and there was little information regarding how the 22 new animals were captured, PMP managers were concerned that some of the new founders might be closely related (e.g., parent–offspring pairs or full siblings). Captive breeding programs generally assume that founders are unrelated, but the newly acquired parma wallaby founders may well violate this assumption.

Molecular markers can contribute to captive breeding programs in a number of ways (Table 1), but genetic data can be particularly valuable for managing populations with incomplete pedigrees. One of the simplest ways genetic data can improve the management of captive populations is to resolve unknown parentage. Contemporary pedigrees can be reconstructed from genetic data when individuals with unknown genealogy and their potential parents are available for genetic testing. For living animals with unknown ancestry deep in their pedigrees (Figure 1), molecular markers can be used to estimate the pairwise relatedness between themselves and the rest of the population. However, the accuracy of molecular relatedness estimates suffers from high sampling variances (Lynch 1988; Lynch and Ritland 1999; Glaubitz et al. 2003) and the best methods for incorporating these values into captive breeding programs remain unclear.

Table 1

Types of contributions that molecular markers can make to captive breeding programs

Contribution to breeding program Species Citation 
Assessment of founder relationships California condor (Gymnogyps californianusGeyer et al. (1993) 
 Guam rail (Rallus owstoniHaig et al. (1994) 
 Micronesian kingfishers (Halcyon cinnamominaHaig et al. (1995) 
 Bearded vulture (Gypaetus barbatusGautschi et al. (2003) 
Pedigree reconstruction Lion-tailed macaques (Macaca silenusMorin and Ryder (1991) 
 Przewalski's horse (Equus ferus przewalskiiBowling et al. (2003) 
Subspecies identification Lion (Panthera leoO'Brien et al. (1987) 
 Chimpanzee (Pan troglodytesEly et al. (2005) 
Identification of geographic origin Galapagos tortoise (Geochelone nigraRussello et al. (2007) 
Quantification of wild genetic diversity captured Baird's tapir (Tapirus bairdiiNorton and Ashley (2004) 
 Iberian wolf (Canis lupus signatusRamirez et al. (2006) 
Identification of genetically valuable individuals Whooping crane (Grus americanaJones et al. (2002) 
 St Vincent parrot (Amazona guildingiiRussello and Amato (2004) 
Assessment of hybridization Lesser white-fronted goose (Anser erythropusRuokonen et al. (2007) 
Contribution to breeding program Species Citation 
Assessment of founder relationships California condor (Gymnogyps californianusGeyer et al. (1993) 
 Guam rail (Rallus owstoniHaig et al. (1994) 
 Micronesian kingfishers (Halcyon cinnamominaHaig et al. (1995) 
 Bearded vulture (Gypaetus barbatusGautschi et al. (2003) 
Pedigree reconstruction Lion-tailed macaques (Macaca silenusMorin and Ryder (1991) 
 Przewalski's horse (Equus ferus przewalskiiBowling et al. (2003) 
Subspecies identification Lion (Panthera leoO'Brien et al. (1987) 
 Chimpanzee (Pan troglodytesEly et al. (2005) 
Identification of geographic origin Galapagos tortoise (Geochelone nigraRussello et al. (2007) 
Quantification of wild genetic diversity captured Baird's tapir (Tapirus bairdiiNorton and Ashley (2004) 
 Iberian wolf (Canis lupus signatusRamirez et al. (2006) 
Identification of genetically valuable individuals Whooping crane (Grus americanaJones et al. (2002) 
 St Vincent parrot (Amazona guildingiiRussello and Amato (2004) 
Assessment of hybridization Lesser white-fronted goose (Anser erythropusRuokonen et al. (2007) 
Figure 1

An example of unknown parentage in the parma wallaby pedigree. Individual 719 was living in the population as of 01 January 2007. Squares in the pedigree represent sires, circles represent dams, individuals with “wild” parents were population founders, and individuals with “?” were born in captivity to undocumented parents. Historic gaps deep in a pedigree are difficult to resolve because genetic samples usually are not available for all individuals.

Figure 1

An example of unknown parentage in the parma wallaby pedigree. Individual 719 was living in the population as of 01 January 2007. Squares in the pedigree represent sires, circles represent dams, individuals with “wild” parents were population founders, and individuals with “?” were born in captivity to undocumented parents. Historic gaps deep in a pedigree are difficult to resolve because genetic samples usually are not available for all individuals.

Most captive breeding programs assume that founders are unrelated. Research suggests that this assumption generally has little affect on a breeding program's ability to retain genetic diversity over the long term, but a marked increase in inbreeding in the early generations of captive breeding programs is observed if unidentified full siblings are present (Rudnick and Lacy 2008). Thus, the best use of molecular estimates of founder relatedness may be to identify first-order relatives (Rudnick and Lacy 2008). This information can be used to avoid breeding recommendations that would pair close relatives, reducing the amount of inbreeding that would accumulate in a population relative to assuming founders were unrelated.

We developed microsatellite markers for the parma wallaby and investigated how genetic data could improve the overall management of the captive population in North America. Our goals were to use molecular data to 1) resolve unknown parentage among contemporary breeders; 2) describe how molecular estimates of kinship could inform future breeding recommendations; and 3) use computer simulations to investigate how marker-based kinship estimates among founders might contribute to the genetic management of the population. Our expectations were that the incorporation of molecular data into the parma wallaby breeding program would improve genetic management over time by preserving more genetic variation and more effectively limiting inbreeding than might be expected otherwise.

Materials and Methods

Sample Collection and DNA Extraction

Blood samples were opportunistically collected during routine physical examinations of 71 captive parma wallabies. Tissue samples from an additional 5 individuals were collected during postmortem exams. All sampled individuals were part of the PMP population held and managed throughout North American zoos. A total of 20 individuals with some degree of unknown ancestry were sampled, as were 20 out of the 22 wild-caught individuals acquired in October 2003. As of 01 January 2007, the captive parma wallaby population was known to be descended from at least 37 founders imported from Kawau Island, New Zealand (exact numbers of founders were not available due to the high degree of unknown ancestry present in the pedigree). Four additional, potential founders were still living in the population but had yet to produce any offspring. A total of 20 established founders were sampled, as were all 4 prospective founders (Figure 2). Because the true number of population founders was unknown, it was difficult to determine how representative our samples were of the population's total founder variation.

Figure 2

Known and potential founders of the captive parma wallaby population, graphed in order of their representation in the living population on 01 January 2007. Black bars represent founders for which biological samples were collected. Individuals 38–41 were potential founders that had yet to contribute offspring to the population (all 4 of these individuals also were sampled).

Figure 2

Known and potential founders of the captive parma wallaby population, graphed in order of their representation in the living population on 01 January 2007. Black bars represent founders for which biological samples were collected. Individuals 38–41 were potential founders that had yet to contribute offspring to the population (all 4 of these individuals also were sampled).

A 1- to 2-cc aliquot of whole blood was collected from each individual in a standard liquid ethylenediaminetetraacetic acid (EDTA) collection tube. Samples were refrigerated or kept on ice for 24–36 h prior to DNA extraction. DNA was isolated from 300 μl of whole blood with the Wizard® Genomic DNA Purification Kit (Promega), following the manufacturer's instructions. Tissue samples collected during postmortem exams were placed in lysis buffer (100 mM Tris–HCl [pH 8.0], 100 mM EDTA, 10 mM NaCl, and 2% sodium dodecyl sulfate [SDS]) and stored at room temperature. DNA was isolated from 1- to 2-mm tissue samples that were incubated overnight at 55 °C in 700 μl of extraction buffer (86 mM NaCl, 43 mM Tris base, 21 mM EDTA, 80 mM Tris–HCl [pH 8.0], 0.01 mg proteinase K, 3% SDS, and 7 mM dithiothreitol). Protein was precipitated by adding 233 μl of 7.45 M ammonium acetate; extractions were centrifuged; and the supernatant was recovered. DNA was isolated from the supernatant with an isopropanol precipitation (Sambrook and Russell 2001), washed with 70% ethanol, and resuspended in 100 μl purified water.

Microsatellite Characterization and Genotyping

Microsatellites were characterized specifically for the parma wallaby by a modified version (Williams and DeWoody 2004) of the microsatellite enrichment protocols described by Hamilton et al. (1999) and Hauswaldt and Glenn (2003). Following microsatellite library construction, primers were initially designed for 10 microsatellite loci, and fluorescent chromatide rhodamine green 5′ deoxyuridine triphosphate was used as a label to screen those loci for allelic variation on an ABI377 automated sequencer (Applied Biosystems). Three of the candidate loci were ultimately genotyped for the study (MP03, MP04, and MP06; Table 2) because of their polymorphism and genotyping reliability.

Table 2

Characterization of 9 microsatellite loci in the parma wallaby

Locus Source or primer sequences (5′–3′) Ta (°C) Size (bp) No. of alleles HO HE 
Pa595 Spencer et al. (1995) 58 316–346 0.63 0.72 
Me14 Taylor and Cooper (1998) 65 156–176 0.52 0.49 
Me17 Taylor and Cooper (1998) 63 133–143 0.55 0.56 
Y175 Zenger et al. (2002) 58 275–291 0.52 0.55 
G19-1 Zenger and Cooper (2001) 65 169–201 0.77 0.77 
G31-1 Zenger and Cooper (2001) 62 118–122 0.42 0.52 
MP03 F: ccatgatttgactagcctgga 50 196–240 0.80 0.79 
R: tctacttcttttcctcaacttgaac 
MP04 F: tctctacaaaatagagatgtccgtgt 50 143–231 0.42 0.50 
R: agaacttctttgcatatttgacttt 
MP06 F: tgatagatcgattgaccgattg 50 235–271 0.76 0.75 
R: gaagccaggtatgccgatta 
Locus Source or primer sequences (5′–3′) Ta (°C) Size (bp) No. of alleles HO HE 
Pa595 Spencer et al. (1995) 58 316–346 0.63 0.72 
Me14 Taylor and Cooper (1998) 65 156–176 0.52 0.49 
Me17 Taylor and Cooper (1998) 63 133–143 0.55 0.56 
Y175 Zenger et al. (2002) 58 275–291 0.52 0.55 
G19-1 Zenger and Cooper (2001) 65 169–201 0.77 0.77 
G31-1 Zenger and Cooper (2001) 62 118–122 0.42 0.52 
MP03 F: ccatgatttgactagcctgga 50 196–240 0.80 0.79 
R: tctacttcttttcctcaacttgaac 
MP04 F: tctctacaaaatagagatgtccgtgt 50 143–231 0.42 0.50 
R: agaacttctttgcatatttgacttt 
MP06 F: tgatagatcgattgaccgattg 50 235–271 0.76 0.75 
R: gaagccaggtatgccgatta 

Ta, annealing temperature; HO, observed heterozygosity; HE, expected heterozygosity. The DNA sequences from which the MP primers were designed have been deposited in GenBank under the accession numbers EU851732 (MP03), EU851733 (MP04), and EU851734 (MP06).

Parma wallaby samples were genotyped at 6 additional microsatellite loci (Table 2): 1 locus derived from an allied rock wallaby (Petrogale assimilis) library (Pa595; Spencer et al. 1995), 1 locus from a yellow-footed rock wallaby (Petrogale xanthopus) library (Y175; Zenger et al. 2002), 2 loci from a tammar wallaby (Macropus eugenii) library (Me14 and Me17; Taylor and Cooper 1998), and 2 loci from an eastern gray kangaroo (Macropus giganteus) library (G19-1 and G31-1; Zenger and Cooper 2001). Polymerase chain reactions (PCRs) for all loci were performed in a final volume of 10 μl and contained 0.5 units Taq, 1× Thermopol buffer, and 0.2 mM each deoxynucleoside triphosphate. PCRs for MP03, MP04, and MP06 contained 0.3 μM forward and reverse primers, whereas PCRs for the remaining loci contained 0.75 μM forward and reverse primers. Annealing temperatures for all loci are listed in Table 2. Previously described thermal profiles were used for PA595, Me14, Me17, G19-1, and G31-1 (Spencer et al. 1995; Taylor and Cooper 1998; Zenger and Cooper 2001), with the exception that the 6 “touchdown” cycles for G19-1 and G31-1 were excluded. The thermal profile for Y175 included an initial denaturation step of 94 °C for 2 min 30 s, followed by 30 cycles of 94 °C for 35 s, 58 °C for 60 s, and 72 °C for 30 s, with a final extension step at 72 °C for 10 min. The thermal profile for MP03, MP04, and MP06 included an initial denaturation step of 94 °C for 2 min, followed by 30 cycles of 94 °C for 30 s, 50 °C for 30 s, and 72 °C for 30 s, with a final extension step at 72 °C for 5 min. All forward primers were fluorescently labeled, and PCR products were electrophoresed in 4.75% denaturing polyacrylamide gels using an ABI377 automated sequencer (Applied Biosystems). Alleles were sized with respect to an internal size standard using the GENESCAN and GENOTYPER softwares (Applied Biosystems).

Estimating Basic Genetic Parameters

The software program MICROSATELLITE ANALYZER (Version 4.05; Dieringer and Schlotterer 2003) was used to calculate the number of alleles, allele frequencies, and both the observed and expected heterozygosities for each microsatellite locus used in the study. The software program GENEPOP (Version 4.0; Rousset 2008) was used to test for linkage disequilibrium between pairs of loci and to test for deviations from Hardy–Weinberg equilibrium. Population-level allele frequencies were used to calculate multilocus parental exclusion probabilities (Selvin 1980).

Parentage Analyses and Relatedness Estimation

Among the parma wallabies sampled for this study, there were 13 individuals for which at least 1 parent (as identified from studbook records) also was sampled. Both the sire and dam were sampled for 5 individuals, only the sire was sampled for 3 individuals, and only the dam was sampled for 5 individuals. Genotypic profiles of the offspring and parents were compared to determine if they were consistent with the presumptive parentage. Samples also were collected from a group of 2 male and 4 female parma wallabies that were housed together. Parentage records for 7 offspring produced by the group were unavailable, so parentage was determined by exclusion (Selvin 1980). Finally, exclusion also was used to resolve the maternity of 1 additional individual that was unrelated to the previously described group.

A relatedness estimator was used to calculate relatedness coefficients between all pairs of sampled individuals. Although captive breeding programs generally use kinships (f values) to quantify relationships, molecular data are usually used to calculate relatedness. The relatedness between 2 individuals is the probability that, at a given locus, an allele sampled at random from 1 individual is present also in the other individual, due to identity by descent from a common ancestor. Furthermore, in noninbred, diploid populations, relatedness is equal to 2f. We used estimates of relatedness to describe how individuals with unknown ancestry might be related to other sampled individuals and to determine if some of the new founders were close relatives. Because the performance of different relatedness estimators varies depending on the numbers and allele frequency distributions of the microsatellite loci employed (Van de Casteele et al. 2001; Russello and Amato 2004), computer simulations were used to estimate sampling variances for 3 estimators across a range of known relationship categories. The 3 relatedness estimators tested were rxyQG (Queller and Goodnight 1989), rxyLR (Lynch and Ritland 1999), and rxyWang (Wang 2002). A simulation written in the C programming language was used to create microsatellite genotypes for 1000 pairs of individuals from each of the following relationship categories: unrelated, parent–offspring, full siblings, and half siblings. Genotypes for unrelated individuals were assigned based on the observed allele frequency distributions from the captive parma wallaby population. Pairs of individuals with known relationships were created by sampling alleles from unrelated individuals in a manner consistent with the desired relationship category. The 3 relatedness estimators were used to calculate relatedness coefficients for all pairs of individuals within a given relationship category and then means and variances of the relatedness coefficients were calculated for each estimator. The relatedness estimator that displayed the smallest sampling variance across the greatest number of relationship categories was selected to calculate relatedness. The software program SPAGEDI (Version 1.2; Hardy and Vekemans 2002) was used to calculate relatedness coefficients for both the simulated data set and the empirical data set.

Investigating Founder Relationships

We used a series of computer simulations to investigate how molecular estimates of relatedness among the newly acquired parma wallaby founders might be used in the genetic management of the captive population. Breeding recommendations for PMP populations, as well as many other types of zoo breeding programs, are based on mks; animals with low mks are generally preferentially bred because they are genetically underrepresented. Population founders are usually assumed to be unrelated, and their f values, both with other founders and with all other individuals in the population, are consequently zero. Molecular estimates of pairwise relatedness can provide information about the true relationships among founders. However, because empirical estimates of relatedness suffer from high sampling variances (Blouin et al. 1996), the use of the values themselves in mk calculations can result in inaccurate estimates of relationship. Instead, we tested 2 methods for using molecular relatedness coefficients to identify first-order relatives (i.e., parent–offspring pairs or full siblings). The methods were tested through computer simulations that investigated the long-term genetic impacts of designating some pairs of founders as first-order relatives, with f values assigned to be 0.25, rather than assuming all founders to be unrelated.

The first method we tested for identifying first-order relatives used a graph of the pairwise relatedness coefficients estimated for the sampled founders. All relatedness coefficients were graphed in order of increasing value, and we looked for a natural break in the distribution that could indicate first-order relatives. This approach for identifying first-order relatives was more qualitative than quantitative, but we felt it was reasonable because the relatedness coefficients of first-order relatives should cluster around a higher mean than the relatedness coefficients of more distant relatives. The value around which first-order relatedness coefficients should cluster is not always obvious a priori; thus, this method allowed close relatives to be identified without prior genetic knowledge of the population or preliminary data analyses.

The second method we tested for identifying first-order relatives used the distributions of relatedness generated from the simulated pairs previously described to choose a cutoff value that would produce an acceptable compromise between the misclassification of first-order relatives as unrelated pairs and the misclassification of unrelated pairs as first-order relatives (see Blouin et al. 1996, which described a similar approach). One approach to balancing these misclassification rates would be to equalize the probability of classifying unrelated pairs as first-order relatives and the probability of classifying first-order relatives as unrelated pairs. However, if there are many more unrelated pairs than first-order relatives present, which is likely the case within groups of captive population founders, such a cutoff will lead to many more pairs of unrelated individuals being incorrectly identified as first-order relatives than first-order relatives being incorrectly identified as unrelated. We chose a cutoff of r = 0.45 to identify first-order relatives because this value failed to identify only 36% of full siblings and 40% of parent–offspring pairs as first-order relatives while incorrectly identifying only 7% of unrelated pairs and 22% of half siblings as first-order relatives. Pairs of founders in the parma wallaby population with relatedness coefficients higher than the cutoff were identified as first-order relatives.

Simulations were used to test the impact of using different sets of founder relationships on the long-term genetic management of the PMP population. The 2 f matrices generated from the methods of identifying first-order relatives were the “modified” f matrices to be tested. The f matrix of the captive parma wallaby population living on 01 January 2007, which assumed all founders to be unrelated, was considered to be the “current” f matrix. Because there was no way to ascertain if the 2 methods we used for identifying first-order relatives correctly identified the relationships of the parma wallaby founders, we designed simulations to quantify 1) the genetic benefits of correctly identifying first-order relatives among the founders assuming such relationships existed and 2) the genetic costs, or detriments, of incorrectly identifying pairs of founders as first-order relatives when pairs were actually unrelated. To quantify these costs and benefits, it was necessary for simulations to simultaneously track 2 f matrices (see Supplementary Material for a detailed description of the simulation). One matrix represented the “true” relationships in a simulated population, and this matrix was used to determine the loss of genetic diversity and accumulated inbreeding in the population. The other matrix represented the relationships that were being tested as the basis for breeding program management. Both f matrices were maintained throughout the simulation, but breeding pairs were selected based on mks calculated from the f matrix being tested. Thus, whereas population management was driven by the relationships being tested, the relationships assumed to represent the true relationships among individuals for a given simulation scenario were being tracked and quantified to measure genetic costs and benefits. To fully quantify the possible genetic benefits and detriments of using molecular data to identify first-order relatives, 4 scenarios were considered for each of the modified f matrices (Figure 3): 1) The current f matrix for the population was used for both the true relationships and the tested relationships; this generated the genetic outcomes if none of the parma wallaby founders were first-order relatives. 2) The current f matrix was used for the true relationships and a modified f matrix was used for the tested relationships; this generated the genetic outcomes if molecular data identified some pairs of founders as first-order relatives when all founders were actually unrelated. 3) A modified f matrix was used for the true relationships and the current f matrix was used for the tested relationships; this generated the genetic outcomes if molecular data failed to identify first-order relatives that were actually present among the founders. 4) A modified f matrix was used for both the true relationships and the tested relationships; this generated the genetic outcomes if molecular data correctly identified the first-order relatives that were actually present among the founders. The genetic outcomes of scenarios 1 and 2 were compared to quantify the detriment to incorrectly identifying first-order relatives among the parma wallaby founders when none were present, whereas the genetic outcomes of scenarios 3 and 4 were compared to quantify the benefit to correctly identifying first-order relatives when some of the parma wallaby founders were indeed related.

Figure 3

The 4 simulation scenarios used to test the impact of using different sets of founder relationships on the long-term genetic management of the parma wallaby PMP population. Each scenario used 2 f matrices, one represented the true relationships in the population and the other represented the relationships being tested. The “current” f matrix was the parma wallaby population living on 01 January 2007. The “modified” f matrix was generated through 1 of 2 methods used to identify first-order relatives.

Figure 3

The 4 simulation scenarios used to test the impact of using different sets of founder relationships on the long-term genetic management of the parma wallaby PMP population. Each scenario used 2 f matrices, one represented the true relationships in the population and the other represented the relationships being tested. The “current” f matrix was the parma wallaby population living on 01 January 2007. The “modified” f matrix was generated through 1 of 2 methods used to identify first-order relatives.

Because unknown ancestry was present in the parma wallaby pedigree, an additional adjustment was necessary to facilitate the comparison of variables across the 4 described scenarios. When unknown ancestry is present, f values can only be calculated from the known part of the pedigree according to the equations described by Ballou and Lacy (1995). These estimates of f are inaccurate, however, and can be either larger or smaller than the true values that would be calculated if the pedigrees were completely known. Because the accuracies of f values calculated in this manner are dependent on the specific distribution of unknown ancestries in a pedigree, variables that are measured as a function of f cannot be directly compared across simulation scenarios. Thus, to ensure that results were comparable across the 4 described scenarios, only the initial f matrices used to start simulations were calculated according to the equations described by Ballou and Lacy (1995). These initial f matrices were then assumed to represent the actual relationships in the population, all individuals were assumed to have 100% known ancestry at the beginning of the simulations, and all further f values were calculated according to the standard methods described by Falconer (1981).

Simulations were designed to model standard captive breeding programs (Wiese and Willis 1996) and were parameterized in accordance with parma wallaby biology. Although parma wallabies are not monogamous in the wild, we modeled monogamous breeding pairs to simulate a captive breeding program's ability to regulate breeding pairs in captivity. Simulations ran on a yearly timestep, and the general yearly process of the simulations was as follows: breeding pairs were selected by a method designed to minimize mk, offspring were produced by the selected pairs, all individuals were aged 1 year, and a proportion of individuals were removed from the population to simulate stochastic mortality. Specific simulation details and parameterizations are presented in the Supplementary Material. In addition to minimizing mk, breeding pair selection also avoided close inbreeding by rejecting any breeding pair that exhibited an f greater than the current average f in the population. An offspring's inbreeding coefficient (F) is equal to the f of its parents (Falconer 1981); thus, limiting breeding pairs based on f slowed the accumulation of inbreeding in the population.

For each simulation, genetic variation and inbreeding were evaluated on a per timestep basis. Inbreeding was measured as the average inbreeding coefficient (forumla), and genetic variation was measured as proportional gene diversity (GD). GD was calculated as 1 − forumla, where forumlawas the average mk in the population forumla; Ballou and Lacy 1995). For a modified f matrix, the impacts of using the given founder relationships were assessed by comparing the GD and forumla observed for each of the 4 previously described simulation scenarios. Due to the stochastic nature of the simulations, each scenario was run 1000 times and results were averaged over all iterations. For each timestep, 95% confidence intervals for average GD and forumlavalues were calculated across all iterations.

Results

The 9 microsatellites assayed for this study averaged 5.3 alleles per locus, and the mean observed heterozygosity across loci was 0.60 (Table 2). All exact tests for Hardy–Weinberg equilibrium were nonsignificant at the 0.05 significance level, except the test for locus MP06 (P = 0.0011). Tests for linkage disequilibrium revealed that genotypes at loci MP04 and Pa595 were linked (P < 0.05), even after a sequential Bonferroni correction was applied for multiple comparisons (Rice 1989). Loci MP04 and MP06 ultimately were removed from the study so that the analyzed suite of microsatellite loci met standard assumptions. The 7 remaining microsatellites exhibited a combined parental exclusion probability of 0.97 when 1 parent was unknown and 0.86 when both parents were unknown.

The basic microsatellite parameters reported were calculated from the entire set of sampled parma wallabies. Parameters calculated from only wild-caught individuals were similar to those reported for the entire set of samples, and preliminary analyses indicated that the results of the study were equivalent regardless of which set of samples (wild-caught individuals vs. all individuals) the microsatellite allele frequencies used to estimate relatedness were calculated from. All microsatellite loci used in the study were in Hardy–Weinberg equilibrium, indicating that inbreeding should not have impacted relatedness estimation.

Parentage Analyses and Relatedness Estimation

Genetic parentage was investigated for 13 offspring and compared with studbook records. Of the 8 sires and 10 dams considered, genotypic profiles suggested that the identities of 2 dams were incorrectly recorded. Both offspring with suspect parentage were born within the same month to a group of wallabies that included 3 adult females, all of which were sampled for this study. The presumptive dam of the first offspring was excluded as a parent of that individual at 3 out of 7 microsatellite loci. Of the 2 other possible dams, 1 was excluded as a parent at 4 microsatellite loci. The last remaining dam could not be excluded at any of the 7 microsatellite loci used in the study; thus, that individual was considered to be the true dam of the first offspring with suspect parentage. The dam of the second offspring with suspect parentage was listed as the dam that was genetically assigned to the first offspring. Because both offspring were born within the same month, one of the other females in the group must have been the dam of the second offspring. Neither of those females could be excluded as a parent of the second offspring at any microsatellite loci; thus, the true parentage of that offspring remains unclear. Genotypic profiles also were used to conclusively determine both the sires and dams for 7 offspring with no parentage records. All 7 offspring were produced by a single group of parma wallabies that included 2 males and 4 females. Six of the 7 offspring were sired by 1 male, whereas only one of the offspring was sired by the alternate male. Finally, the maternity of 1 additional individual also was investigated. Studbook records indicated that there were 2 possible dams for the individual, but only 1 of the 2 possibilities was sampled for this study. The female that was sampled could be excluded as the dam at 3 microsatellite loci; thus, the unsampled female was assumed to be the individual's dam.

Sampling variances were calculated for 3 relatedness estimators across 4 categories of relationship (unrelated, UR; parent–offspring, PO; full siblings, FS; and half siblings, HS). The estimator that displayed the smallest sampling variance across the greatest number of relationship categories was rxyWang (Table 3), so that estimator was selected to calculate relatedness among the sampled parma wallabies. Over all relationship categories, sampling variances ranged from 0.02 to 0.08 across all estimators, and the smallest sampling variances were observed for parent–offspring pairs.

Table 3

Mean relatedness coefficients and variances for 3 relatedness estimators, based on allele frequencies from the captive North American parma wallaby population

 Relationship
 
 UR HS FS PO 
rxyQG 0.0167 (0.0704) 0.2510 (0.0668) 0.5065 (0.0573) 0.5030 (0.0304) 
rxyLR 0.0159 (0.0480)a 0.2509 (0.0674) 0.5052 (0.0643) 0.4952 (0.0437) 
rxyWang 0.0199 (0.0784) 0.2425 (0.0654)a 0.5063 (0.0527)a 0.4998 (0.0198)a 
 Relationship
 
 UR HS FS PO 
rxyQG 0.0167 (0.0704) 0.2510 (0.0668) 0.5065 (0.0573) 0.5030 (0.0304) 
rxyLR 0.0159 (0.0480)a 0.2509 (0.0674) 0.5052 (0.0643) 0.4952 (0.0437) 
rxyWang 0.0199 (0.0784) 0.2425 (0.0654)a 0.5063 (0.0527)a 0.4998 (0.0198)a 

Values were based on 1000 simulated pairs from each of the following relationship categories: unrelated (UR), half siblings (HS), full siblings (FS), and parent–offspring (PO).

a

The smallest sampling variances per relationship category

Investigating Founder Relationships

We tested the impact of using molecular relatedness coefficients to identify first-order relatives among founders on the long-term genetic management of the parma wallaby PMP population. The first method we tested for identifying first-order relatives used a graph of the 190 pairwise relatedness coefficients estimated for the 20 sampled founders (Figure 4). In a noninbred diploid population, first-order relatives should have a relatedness coefficient of 0.5. We looked for a break in the ordered distribution of relatedness coefficients either approximately equal to or greater than this value and ultimately selected the 4 highest pairwise relatedness coefficients to represent first-order relatives (Figure 4). The 4 selected pairs of individuals were assigned f values of 0.25, and the genetic impacts of using the resulting modified f matrix were quantified through computer simulations. Figure 5 illustrates the rates of change observed in GD and forumla across 100 years as compared with random mating.

Figure 4

Pairwise relatedness coefficients estimated for the 20 sampled founders, graphed in order of increasing value. A dashed horizontal line is drawn at 0.5, which is the relatedness coefficient of first-order relatives in a noninbred diploid population. Values above the dashed 0.5 relatedness line (4 values in total) were considered to be closely related.

Figure 4

Pairwise relatedness coefficients estimated for the 20 sampled founders, graphed in order of increasing value. A dashed horizontal line is drawn at 0.5, which is the relatedness coefficient of first-order relatives in a noninbred diploid population. Values above the dashed 0.5 relatedness line (4 values in total) were considered to be closely related.

Figure 5

Trends in (a) GD and (b) forumla for simulation scenarios that used a distribution of pairwise relatedness coefficients to identify first-order relatives among parma wallaby founders. The differences between the 4 tested scenarios were indistinguishable at this scale, thus data are presented for only a single scenario: the current f matrix for the parma wallaby PMP was used for both the true founder relationships and the tested founder relationships (solid line). For comparison, the outcome under random mating also is included (dotted line). Figures 5 and 6 illustrate the final 10 timesteps of these graphs, where differences between the 4 tested scenarios are visible.

Figure 5

Trends in (a) GD and (b) forumla for simulation scenarios that used a distribution of pairwise relatedness coefficients to identify first-order relatives among parma wallaby founders. The differences between the 4 tested scenarios were indistinguishable at this scale, thus data are presented for only a single scenario: the current f matrix for the parma wallaby PMP was used for both the true founder relationships and the tested founder relationships (solid line). For comparison, the outcome under random mating also is included (dotted line). Figures 5 and 6 illustrate the final 10 timesteps of these graphs, where differences between the 4 tested scenarios are visible.

When results were graphed across all 100 years of the simulations, differences between the scenarios could not be discerned on Figure 5. Therefore, the results from only the last 10 timesteps were graphed to better display the comparisons between scenarios 1 and 2 (Figure 6) and scenarios 3 and 4 (Figure 7). Table 4 shows the final values at the end of 100 years for GD and forumla from all scenarios. If the 4 selected founder pairs we designated as first-order relatives were actually unrelated (the contrast between scenarios 1 and 2), both the loss of GD and the increase in forumla were approximately 0.0004. If the 4 selected founder pairs were indeed related and we designated them as such (the contrast between scenarios 3 and 4), both the increase in GD and the reduction in forumla were approximately 0.0009. Thus, when the distribution of pairwise relatedness coefficients was used to identify first-order relatives, the possible benefits of using molecular data to correctly assign founder relationships were slightly higher than the possible detriments to incorrectly identifying founders as first-order relatives when they were actually unrelated.

Table 4

Final GD and mean inbreeding (F) values after 100 simulation timesteps for the 4 scenarios that were designed to quantify the impacts of using different sets of founder relationships on the long-term genetic management of a captive parma wallaby population

 Distribution break
 
Specified cutoff
 
Scenario GD F GD F 
1: No first-order relatives 0.9334 0.0529 0.9334 0.0529 
2: Incorrect identification of first-order relatives 0.9330 0.0533 0.9325 0.0538 
3: Failure to detect first-order relatives 0.9310 0.0533 0.9291 0.0572 
4: Correct identification of first-order relatives 0.9319 0.0544 0.9311 0.0552 
 Distribution break
 
Specified cutoff
 
Scenario GD F GD F 
1: No first-order relatives 0.9334 0.0529 0.9334 0.0529 
2: Incorrect identification of first-order relatives 0.9330 0.0533 0.9325 0.0538 
3: Failure to detect first-order relatives 0.9310 0.0533 0.9291 0.0572 
4: Correct identification of first-order relatives 0.9319 0.0544 0.9311 0.0552 

First-order relatives were identified among the founders by either an observed break in the distribution of all empirically generated pairwise kinships or a specified kinship cutoff value. In all cases, the 95% confidence intervals around mean GD and F at the end of the 100 timesteps were approximately 0.0001.

Figure 6

Graphs of (a) GD and (b) forumla that demonstrate the impact of designating selected founder pairs as first-order relatives when all founders were actually unrelated. Data are presented for the final 10 timesteps of the following simulation scenarios: the current f matrix for the parma wallaby PMP was used for both the true founder relationships and the tested founder relationships (circle); the current f matrix for the parma wallaby PMP was used for the true founder relationships and a modified f matrix was used for the tested founder relationships (diamond). The distribution of pairwise relatedness coefficients for the newly acquired parma wallaby founders was used to identify first-order relatives for the modified f matrix.

Figure 6

Graphs of (a) GD and (b) forumla that demonstrate the impact of designating selected founder pairs as first-order relatives when all founders were actually unrelated. Data are presented for the final 10 timesteps of the following simulation scenarios: the current f matrix for the parma wallaby PMP was used for both the true founder relationships and the tested founder relationships (circle); the current f matrix for the parma wallaby PMP was used for the true founder relationships and a modified f matrix was used for the tested founder relationships (diamond). The distribution of pairwise relatedness coefficients for the newly acquired parma wallaby founders was used to identify first-order relatives for the modified f matrix.

Figure 7

Graphs of (a) GD and (b) forumla that demonstrate the impact of correctly designating selected founder pairs as first-order relatives rather than assuming all founders to be unrelated. Data are presented for the final 10 timesteps of the following simulation scenarios: a modified f matrix was used for both the true founder relationships and the tested founder relationships (triangle); a modified f matrix was used for the true founder relationships and the current f matrix for the parma wallaby PMP was used for the tested founder relationships (square). The distribution of pairwise relatedness coefficients for the newly acquired parma wallaby founders was used to identify first-order relatives for the modified f matrix.

Figure 7

Graphs of (a) GD and (b) forumla that demonstrate the impact of correctly designating selected founder pairs as first-order relatives rather than assuming all founders to be unrelated. Data are presented for the final 10 timesteps of the following simulation scenarios: a modified f matrix was used for both the true founder relationships and the tested founder relationships (triangle); a modified f matrix was used for the true founder relationships and the current f matrix for the parma wallaby PMP was used for the tested founder relationships (square). The distribution of pairwise relatedness coefficients for the newly acquired parma wallaby founders was used to identify first-order relatives for the modified f matrix.

For the second method of identifying first-order relatives, we used the distributions of relatedness values generated from simulated relationship pairs to choose a relatedness cutoff value that indicated first-order relatives. We selected relatedness coefficients of 0.45 and higher to represent first-order relatives, as simulated data suggested that this cutoff would correctly identify more than 60% of first-order relatives while misclassifying only 7% of unrelated pairs as related. Seven of the pairwise relatedness coefficients among the parma wallaby founders were higher than 0.45, and the 7 selected pairs of individuals were assigned f values of 0.25. The genetic impacts of using the resulting modified f matrix were similar to those described for the first method of identifying first-order relatives (full results not shown, but final values are given in Table 4). Using this method for modifying the f matrix to account for assumed founder relationships, if we designated the 7 selected founder pairs as first-order relatives when they were actually unrelated (the contrast between scenarios 1 and 2), both the loss of GD and the increase in forumla was approximately 0.0009. If we correctly designated the 7 selected founder pairs as first-order relatives when they were indeed related (the contrast between scenarios 3 and 4), both the gain of GD and the decrease in forumla was approximately 0.0020. In summary, like the results observed for the first method of identifying first-order relatives, the possible benefits to using molecular data to identify closely related founders were higher than the possible detriments.

Discussion

The resolution of unknown parentage is one of the primary ways molecular data can be used to improve captive population management. Unknown parentage plagues many contemporary pedigrees, especially in species housed in groups with multiple males. This problem is exemplified in parma wallabies; 57% of individuals living in the PMP population as of January 2007 had varying amounts of unknown parentage, and 9 of those individuals had either one or both parents listed as unknown. We successfully resolved parentage for 7 of those 9 individuals, as well as for 1 additional individual that died prior to 2007 (DNA samples were not available for the 2 remaining individuals with unknown parentage). Although a considerable amount of unknown ancestry persists deeper in the parma wallaby pedigree, the resolution of the majority of unknown parentage in the contemporary pedigree still improves the accuracy of mk calculations and, thus, the genetic management of the parma wallaby PMP. For species held in groups with multiple males, unknown paternity will nearly always continue to be perpetually generated. Furthermore, as demonstrated by the pedigree records investigated here, maternity is not always known with certainty. We demonstrated that molecular markers can help resolve unknown parentage in captive pedigrees, which increases the effectiveness of genetic management by improving the accuracy of mk calculations used for breeding recommendations.

Studbook records indicated that a total of 18 known parent–offspring pairs were included in the group of parma wallabies sampled for this study. Microsatellite data for 16 of these pairs (89%) were consistent with the recorded pedigree, indicating that parentage records for this population were generally accurate. Still, the identities of 2 presumptive dams (11%) appear to have been incorrectly recorded. Further analyses indicated that both individuals with misassigned dams were born within the same month to a group of parma wallabies that were being housed together, which means that the inaccuracies were not independent of each other. There are several explanations for how these errors could have arisen. One possibility is that the dam identification numbers were incorrectly recorded in the parma wallaby studbook. If this was the case, the parentage records should be changed to accurately reflect the true pedigree. A second explanation, however, is that the blood samples used for the parentage analyses were incorrectly labeled at the time of sample collection; both offspring with misassigned maternity and 2 of their 3 possible dams were sampled at the same location on the same date. In this case, parentage records should not be adjusted. The collection of additional blood samples is necessary to either confirm or discount the possibility that the original samples were incorrectly labeled, and, at this time, the parentage of the 2 offspring with misassigned maternity should be considered to be provisional.

In the absence of a complete pedigree, molecular estimates of relatedness can provide information about the relationships that exist among individuals in a population. We tested 2 methods of using molecular relatedness coefficients to identify first-order relatives among parma wallaby founders and quantified the possible genetic benefits and detriments to using that information for managing the parma wallaby PMP. Both methods exhibited similar trends and results. The benefit to correctly identifying first-order relatives among the founders was small; after 100 years, the potential gain in GD and the reduction in the accumulation of forumla was 0.0009 or 0.0020, depending on the method used to identify first-order relatives. The detriment to identifying some founder pairs as first-order relatives when all founders were actually unrelated also was small; after 100 years, the potential gain in GD and the reduction in the accumulation of forumla was 0.0004 or 0.0009, again depending on the method used to identify first-order relatives. As expected, the method that identified the largest number of relatives exhibited both the largest potential benefit and the largest potential detriment. For both methods, the potential genetic benefits of using molecular data to identify related founders were greater than the potential costs (in terms of GD and forumla). However, the results suggest that incorporating molecular estimates of founder relatedness into the genetic management of the parma wallaby PMP would be neither particularly beneficial nor detrimental over the next 100 years of management.

The 2 methods that we tested for identifying first-order relatives among founders identified 4 and 7 possible first-order relatives, respectively. The identification of 4 or 7 pairs of presumptive first-order relatives among the founders is plausible given the history of this population and may be typical of other managed populations with some uncertainty caused by missing records or uncertain parentage due to group housing. We used the population of 157 parma wallabies living on 01 January 2007 to start our simulations. Thus, when founders were identified as first-order relatives in the initial simulation timestep, we modified f values for only 4 or 7 pairs of founders. Because those founders had already produced some offspring in the population, the new founder relationships also were carried through the pedigree to modify the f values of those founder's descendants: f values were modified for the same 30 descendants when both 4 and 7 pairs of founders were identified, but the specific f modifications associated with each set of assumed founder relationships produced different mks among those descendants (Figure 8). Collectively, these modifications were simply too minor in the context of the total population to affect significant change in genetic management. Our previous work on founder relationships, which demonstrated a measurable benefit to identifying any full siblings present among founders, was based solely on simulated populations (Rudnick and Lacy 2008). Furthermore, although a few studies have suggested that molecular estimates of relatedness might be incorporated into captive breeding programs (e.g., Jones et al. 2002; Russello and Amato 2004), none have quantified the potential impact. Our work on the parma wallaby indicates that additional work on other captive populations may be necessary to determine if molecular data should be incorporated into their genetic management.

Figure 8

Changes in individual mks due to identifying first-order relatives among the captive parma wallaby founders. mks for a total of 38 individuals are graphed for 3 different sets of founder relationships: founders were unrelated (graphic), 4 pairs of first-order relatives were present among the founders (▪), and 7 pairs of first-order relatives were present among the founders (graphic). The 38 individuals for which data are presented are the founders (n = 8) and their descendants (n = 30) living on 01 January 2007 that experienced a change in mk when first-order relatives were present among the founders.

Figure 8

Changes in individual mks due to identifying first-order relatives among the captive parma wallaby founders. mks for a total of 38 individuals are graphed for 3 different sets of founder relationships: founders were unrelated (graphic), 4 pairs of first-order relatives were present among the founders (▪), and 7 pairs of first-order relatives were present among the founders (graphic). The 38 individuals for which data are presented are the founders (n = 8) and their descendants (n = 30) living on 01 January 2007 that experienced a change in mk when first-order relatives were present among the founders.

In addition to quantifying the impacts of using molecular estimates of founder relatedness to genetically manage the parma wallaby PMP, we also investigated how molecular relatedness coefficients might be incorporated into the parma wallaby breeding program for individuals with unknown ancestry. Microsatellite data can theoretically be used to discriminate among multiple relationship categories within a population, but approximately 40 independently segregating loci are required to differentiate full siblings from half siblings with greater than 90% reliability (Blouin et al. 1996). Furthermore, individuals from more distant relationship categories are even harder to classify. Rather than using inaccurate estimates of relatedness for mk calculations, we suggest identifying a series of relatedness ranges that represent different degrees of relationships. Then, if a pair of individuals with ambiguous ancestry has a relatedness coefficient that falls within a given range, the pair's f could be assigned a predetermined value intended to represent that relationship category (e.g., full siblings, half siblings, unrelated, etc.). For a given suite of microsatellites, means and sampling variances calculated for simulated pairs of individuals that belong to specific relationship categories could be used to select both appropriate relatedness ranges and the f values that those ranges represent. Here, we tested this approach for the parma wallaby. Because our microsatellites provided little resolution between simulated relationship categories (Table 3), we chose only a single relatedness range for evaluation. We selected relatedness coefficients equal to or greater than 0.45 to indicate “close relatives” because simulated data indicated that this value should identify a majority of parent–offspring and full sibling pairs as first-order relatives while keeping the probability of misclassifying unrelated pairs as first-order relatives low. We then selected 0.25 to replace the f values of all pairs with relatedness coefficients equal to or greater than the cutoff.

Parma wallabies with studbook numbers 617 and 623 both have unknown ancestry that originated before the value of accurate pedigree records had been recognized. The amount of an individual's ancestry that is known is the proportion of its genome that can be traced back to known founders, and for an individual, this value can range from 0 to 1 (Ballou and Lacy 1995). The ancestry of parma wallaby 617 is 0.0 known because none of the individual's ancestors can be traced back to wild founders, and the ancestry of parma wallaby 623 is 0.09 known because the individual's pedigree can be traced back to 4 wild founders and 9 ancestors with unknown parents. Because nothing is known about the ancestry of 617 and very little is known about the ancestry of 623, a pedigree cannot be used to determine how those 2 individuals are related and their pairwise f is undefined. Our molecular estimate of relatedness for the pair, however, was 0.4948. Thus, in accordance with the methods we have described, the pairwise f should be changed from undefined to 0.25. This change in f would raise both mk values, making each individual less genetically valuable. These changes in mk are desirable if the individuals are truly closely related because captive breeding programs preferentially breed genetically underrepresented individuals (i.e., individuals with low mks) to equalize founders contributions and retain high levels of genetic diversity.

Although the method we have described for incorporating molecular estimates of relatedness into captive breeding programs is quite plausible, additional research on both simulated and real populations is needed to fully quantify the impacts this, or a similar method, would have on a population's long-term genetics. Willis (2001) demonstrated that overestimating kinship among individuals with unknown ancestry was generally more detrimental to genetic diversity than underestimating kinship to an equivalent degree. Molecular estimates of relatedness can either overrepresent or underrepresent kinship, and the direction of error for a specific estimate is unknowable. Thus, any change in the f matrix used to manage a captive breeding program should be made with caution and be consistent with any ancillary information that may be available. For example, Gautschi et al. (2003) combined molecular relatedness data, mitochondrial DNA sequence data, and studbook information to investigate founder relationships for a captive bearded vulture (Gypaetus barbatus) population. Likelihood tests using the molecular relatedness data indicated that 40 pairs of founders were first-order relatives. However, when all available information was considered together, it was determined that 33 of those pairs (82.5%) could not be closely related. For the parma wallabies, the limited amount of pedigree information that is available for 617 and 623 does not preclude a close relationship; both the sire of 617 and the sire of 623 were born to unknown parents around the same time at the same location. Thus, if their sires were closely related, 617 and 623 would be closely related as well. In a case such as this, where both molecular and pedigree data are compatible with the supposition that 2 individuals are closely related, the f for the individuals could be modified to reflect this relationship.

Data on parma wallabies 617 and 788 demonstrate the risks of using modest molecular data sets for identifying relatives. The pair's relatedness coefficient was estimated as 0.4746, which was greater than the 0.45 cutoff we selected for identifying close relatives. Parma wallaby 788 was one of the new founders acquired from the wild in 2003. Although the grandparents of 617 are unknown, studbook records indicate that both the sire and dam of 617 were born to parents within the captive population around 1989. Thus, it seems highly unlikely that 617 and 788 could be closely related. These results exemplify the need for large suites of molecular markers to accurately estimate genetic parameters such as heterozygosity or relatedness (Blouin et al. 1996; Glaubitz et al. 2003; DeWoody and DeWoody 2005).

Conclusions

We used the captive parma wallaby population in North America to investigate the methods and prospects of incorporating molecular data into zoo and aquarium breeding programs. We evaluated the efficacy of molecular markers in resolving captive pedigrees and quantified the impact of the improved pedigrees on the standard genetic goals of captive breeding programs. Specifically, we described a suite of microsatellites sufficient for resolving contemporary unknown parentage in the captive parma wallaby population and used those microsatellites to 1) successfully resolve parentage for 8 individuals and 2) identify 2 dams that may have been incorrectly recorded in the parma wallaby studbook. Although previous research suggested molecular estimates of relatedness may be useful for identifying close relatives among population founders, results from our computer simulations suggested the incorporation of this information into the genetic management of the parma wallaby PMP provides little benefit. This result is likely due to both the small number of related founders we identified and the relatively small impact of those initial relationships on kinships in the descendant population. Thus, the benefit of using molecular data to ascertain founder relationships would be greater for populations that have a higher proportion of related founders or more descendants from those founders that were interrelated. Extensive analysis of such pedigrees would be necessary to determine how rapidly the benefits of incorporating molecular data on founder relationships accrued as the amount of kinship among founders increased. Finally, we used the parma wallaby to demonstrate a plausible method for incorporating molecular relatedness coefficients into captive breeding programs that have unknown ancestry deep in their pedigrees. This method can be applied to a broad range of captive breeding programs, but additional research on both simulated and real populations is needed to fully quantify long-term genetic impacts.

Supplementary Material

Supplementary material can be found at http://www.jhered.oxfordjournals.org/.

Funding

Marsupial and Monotreme Taxon Advisory Group for the Association of Zoos and Aquariums; Donations from holding institutions to the Parma Wallaby Population Management Plan; Conservation Project grant from the Institute of Museum and Library Services to J.A.I.

We thank the following institutions for kindly providing parma wallaby samples: Bermuda Zoo, Brevard Zoo, Cape May Zoo, Cleveland Metroparks Zoo, Disney's Animal Kingdom, Fort Worth Zoo, Happy Hollow Zoo, Henry Doorly Zoo, Miami Metrozoo, North Carolina Zoo, Oklahoma City Zoo, Riverbanks Zoo, Roger Williams Park Zoo, San Antonio Zoo, San Diego Zoo, and Sunset Zoo. We also thank Guha Dharmarajan for his help in constructing the parma wallaby microsatellite library and Joseph Busch for his help in the final stages of microsatellite data collection. The DeWoody laboratory group provided many insightful comments on a previous draft of this paper.

References

Ballou
JD
Lacy
RC
Ballou
JD
Gilpin
M
Foose
TJ
Identifying genetically important individuals for management of genetic diversity in pedigreed populations
Population management for survival and recovery
 , 
1995
New York
Columbia Press
(pg. 
76
-
111
)
Blouin
MS
Parsons
M
Lacille
V
Lotz
S
Use of microsatellite loci to classify individuals by relatedness
Mol Ecol
 , 
1996
, vol. 
5
 (pg. 
393
-
401
)
Bowling
AT
Zimmermann
W
Ryder
O
Penado
C
Peto
S
Chemnick
L
Yasinetskaya
N
Zharkikh
T
Genetic variation in Przewalski's horses, with special focus on the last wild caught mare, 231 Orlitza III
Cytogenet Genome Res
 , 
2003
, vol. 
101
 (pg. 
226
-
234
)
DeWoody
YD
DeWoody
JA
On the estimation of genome-wide heterozygosity using molecular markers
J Hered
 , 
2005
, vol. 
96
 (pg. 
85
-
88
)
Dieringer
D
Schlotterer
C
MICROSATELLITE ANALYSER (MSA): a platform independent analysis tool for large microsatellite data sets
Mol Ecol Notes
 , 
2003
, vol. 
3
 (pg. 
167
-
169
)
Ely
JJ
Dye
B
Frels
WI
Fritz
J
Gagneux
P
Khun
HH
Switzer
WM
Lee
DR
Subspecies composition and founder contribution of the captive U.S. chimpanzee (Pan troglodytes) population
Am J Primatol
 , 
2005
, vol. 
67
 (pg. 
223
-
241
)
Falconer
DS
Introduction to quantitative genetics
 , 
1981
2nd ed
New York
Longman
Fernandez
J
Toro
MA
The use of mathematical programming to control inbreeding in selection schemes
J Anim Breed Genet
 , 
1999
, vol. 
116
 (pg. 
447
-
466
)
Foose
TJ
Ballou
JD
Population management: theory and practice
Int Zoo Yearb
 , 
1988
, vol. 
27
 (pg. 
26
-
41
)
Gautschi
B
Jacob
G
Negro
JJ
Godoy
JA
Muller
JP
Schmid
B
Analysis of relatedness and determination of the source of founders in the captive bearded vulture, Gypaetus barbatus, population
Conserv Genet
 , 
2003
, vol. 
4
 (pg. 
479
-
490
)
Geyer
CJ
Ryder
OA
Chemnick
LG
Thompson
EA
Analysis of relatedness in the California condors, from DNA fingerprinting
Mol Biol Evol
 , 
1993
, vol. 
10
 (pg. 
571
-
589
)
Glaubitz
JC
Rhodes
EO
Jr
DeWoody
JA
Prospects for inferring pairwise relationships with single nucleotide polymorphisms
Mol Ecol
 , 
2003
, vol. 
12
 (pg. 
1039
-
1047
)
Haig
SM
Ballou
JD
Casna
NJ
Identification of kin structure among Guam rail founders: a comparison of pedigrees and DNA profiles
Mol Ecol
 , 
1994
, vol. 
3
 (pg. 
109
-
119
)
Haig
SM
Ballou
JD
Casna
NJ
Genetic identification of kin in Micronesian kingfishers
J. Hered
 , 
1995
, vol. 
86
 (pg. 
423
-
431
)
Hamilton
MB
Pincus
EL
Di Fiore
A
Fleischer
RC
Universal linker and ligation procedures for construction of genomic DNA libraries enriched for microsatellites
Biotechniques
 , 
1999
, vol. 
27
 (pg. 
500
-
507
)
Hardy
OJ
Vekemans
X
SPAGeDi: a versatile computer program to analyze spatial genetic structure at the individual or population levels
Mol Ecol Notes
 , 
2002
, vol. 
2
 (pg. 
618
-
620
)
Hauswaldt
SJ
Glenn
TC
Microsatellite DNA loci from the diamondback terrapin (Malaclemys terrapin)
Mol Ecol Notes
 , 
2003
, vol. 
3
 (pg. 
174
-
176
)
Hedrick
PW
Miller
P
Sandlund
OT
Hindar
K
Brown
AHD
Conservation genetics: theory and management of captive populations
Conservation of biodiversity for sustainable development
 , 
1992
Oslo (Norway)
Scandinavian University Press
(pg. 
70
-
87
)
IUCN
2007 IUCN Red List of Threatened Species
 , 
2007
 
[Internet]. Available from: URL http://www.iucnredlist.org/
Jones
KL
Glenn
TC
Lacy
RC
Pierce
JR
Unruh
N
Mirande
CM
Chavez-Ramirez
F
Refining the whooping crane studbook by incorporating microsatellite DNA and leg-banding analyses
Conserv Biol
 , 
2002
, vol. 
16
 (pg. 
789
-
799
)
Lacy
RC
Bowles
ML
Whelan
CJ
Managing genetic diversity in captive populations of animals
Restoration and recovery of endangered plants and animals
 , 
1994
Cambridge (UK)
Cambridge University Press
(pg. 
63
-
89
)
Lacy
RC
Clarification of genetic terms and their use in the management of captive populations
Zoo Biol
 , 
1995
, vol. 
14
 (pg. 
565
-
577
)
Lynch
M
Estimation of relatedness by DNA fingerprinting
Mol Biol Evol
 , 
1988
, vol. 
5
 (pg. 
584
-
599
)
Lynch
M
Ritland
K
Estimation of pairwise relatedness with molecular markers
Genetics
 , 
1999
, vol. 
152
 (pg. 
1753
-
1766
)
Maynes
GM
Occurrence and field recognition of Macropus parma
Aust Zool
 , 
1974
, vol. 
18
 (pg. 
72
-
87
)
Morin
PA
Ryder
OA
Founder contribution and pedigree inference in a captive breeding colony of lion-tailed macaques, using mitochondrial DNA and DNA fingerprint analyses
Zoo Biol
 , 
1991
, vol. 
10
 (pg. 
341
-
352
)
Norton
JE
Ashley
MV
Genetic variability and population differentiation in captive Baird's tapirs (Tapirus bairdii)
Zoo Biol
 , 
2004
, vol. 
23
 (pg. 
521
-
531
)
O'Brien
SJ
Joslin
P
Smith
GL
3rd
Wolfe
R
Schaffer
N
Heath
E
Ott-Joslin
J
Rawal
PP
Bhattacharjee
KK
Martenson
JS
Evidence for African origins of founders of the Asiatic lion species survival plan
Zoo Biol
 , 
1987
, vol. 
6
 (pg. 
99
-
116
)
Parliament of New South Wales
Threatened Species Conservation Act, 1995
 , 
1995
New South Wales (Australia)
Parliament of New South Wales
Queller
DC
Goodnight
KF
Estimating relatedness using genetic markers
Evolution
 , 
1989
, vol. 
43
 (pg. 
258
-
275
)
Ramirez
O
Altet
L
Ensenat
C
Vila
C
Sanchez
A
Ruiz
A
Genetic assessment of the Iberian wolf Canis lupus signatus captive breeding program
Conserv Genet
 , 
2006
, vol. 
7
 (pg. 
861
-
878
)
Rice
WR
Analyzing tables of statistical tests
Evolution
 , 
1989
, vol. 
43
 (pg. 
223
-
225
)
Ride
WDL
A guide to the native mammals of Australia
 , 
1970
New York
Oxford University Press
Rousset
F
GENEPOP’007: a complete re-implementation of the GENEPOP software for Windows and Linux
Mol Ecol Notes
 , 
2008
, vol. 
8
 (pg. 
103
-
106
)
Rudnick
JA
Lacy
RC
The impact of assumptions about founder relationships on the effectiveness of captive breeding strategies
Conserv Genet
 , 
2008
, vol. 
9
 (pg. 
1439
-
1450
)
Ruokonen
M
Andersson
A
Tegelstrom
H
Using historical captive stocks in conservation. The case of the lesser white-fronted goose
Conserv Genet
 , 
2007
, vol. 
8
 (pg. 
197
-
207
)
Russello
A
Amato
G
Ex situ population management in the absence of pedigree information
Mol Ecol
 , 
2004
, vol. 
13
 (pg. 
2829
-
2840
)
Russello
MA
Hyseni
C
Gibbs
JP
Cruz
S
Marquez
C
Tapia
W
Velensky
P
Powell
JR
Caccone
A
Lineage identification of Galapagos tortoises in captivity worldwide
Anim Conserv
 , 
2007
, vol. 
10
 (pg. 
304
-
311
)
Sambrook
J
Russell
DW
Molecular cloning: a laboratory manual
 , 
2001
3rd ed
New York
Cold Spring Harbor Laboratory Press
Selvin
S
Probability of nonpaternity determined by multiple allele codominant systems
Am J Hum Genet
 , 
1980
, vol. 
32
 (pg. 
276
-
278
)
Sonesson
AK
Meuwissen
THE
Minimization of rate of inbreeding for small populations with overlapping generations
Genet Res Camb
 , 
2001
, vol. 
77
 (pg. 
285
-
292
)
Spencer
PBS
Odorico
DM
Jones
SJ
March
HD
Miller
DJ
Highly variable microsatellites in isolated colonies of the rock-wallaby (Petrogale assimilis)
Mol Ecol
 , 
1995
, vol. 
4
 (pg. 
523
-
525
)
Taylor
AC
Cooper
DW
A set of tammar wallaby (Macropus eugenii) microsatellites tested for genetic linkage
Mol Ecol
 , 
1998
, vol. 
7
 (pg. 
925
-
931
)
U.S. Fish and Wildlife Service
Endangered and threatened wildlife and plants 50 CFR 17.11 and 17.12
 , 
2005
U.S. Fish and Wildlife Service
Van de Casteele
T
Galbusera
T
Matthysen
E
A comparison of microsatellite-based pairwise relatedness estimators
Mol Ecol
 , 
2001
, vol. 
10
 (pg. 
1539
-
1549
)
Wang
JL
An estimator for pairwise relatedness using molecular markers
Genetics
 , 
2002
, vol. 
160
 (pg. 
1203
-
1215
)
Wiese
RJ
Willis
W
Genetic, demographic, and biological considerations in making Species Survival Plan© breeding recommendations
Anim Keepers Forum
 , 
1996
, vol. 
23
 (pg. 
82
-
86
)
Williams
RN
DeWoody
JA
Fluorescent dUTP helps characterize 10 novel tetranucleotide microsatellites from an enriched salamander (Ambystoma texanum) genomic library
Mol Ecol Notes
 , 
2004
, vol. 
4
 (pg. 
17
-
19
)
Willis
K
Unpedigreed populations and worst-case scenarios
Zoo Biol
 , 
2001
, vol. 
20
 (pg. 
305
-
314
)
Wodzicki
K
Flux
JEC
The parma wallaby and its future
Oryx
 , 
1971
, vol. 
11
 (pg. 
40
-
47
)
Zenger
KR
Cooper
DW
A set of highly polymorphic microsatellite markers developed for the eastern grey kangaroo (Macropus giganteus)
Mol Ecol Notes
 , 
2001
, vol. 
1
 (pg. 
98
-
100
)
Zenger
KR
McKenzie
LM
Cooper
DW
The first comprehensive genetic linkage map of a marsupial: the tammar wallaby (Macropus eugenii)
Genetics
 , 
2002
, vol. 
162
 (pg. 
321
-
330
)

Author notes

Corresponding Editor: Robert C. Fleischer