Chapter 4. Studbook populations


The simulation experiments described and discussed in chapter 3 involve hypothetical populations that show no generation overlap. Mating structure in each generation has been modelled to either full-sib matings or according Maximal Avoidance of Inbreeding schemes [see Chapter 5]. These models will most likely not reflect the complexity of real populations as, for example, considerable generation overlap due to both definition of captive-born generations(14) and reproductive life-span [Princée, 1988] generally occurs. This means that populations can be composed of biological generation (age) groups with individuals which belong to different captive-born (genetic) generations. The ChromoFlow model can estimate genetic variation in populations at annual census dates [see Chapter 2]. Genetic drift in biological generations will be reflected in the results of such experiments. Furthermore, effects of trends in population growth and management decisions can be revealed with this type of simulations.

Genetic management of zoo populations, involving minimizing inbreeding and reducing genetic loss, was not introduced until the end of the 1970's [see for example Flesness, 1977]. This implies that genetic loss in zoo populations which were established before this period can be (unintentionally) large compared to populations which have been genetically managed from start. The zoo population of Przewalski's horse or Asian wild horse, Equus przewalskii, was established in 1900 (when population genetics was barely developed). An international studbook was not established until 1960 [Volf et seq., 1960] as a result. The zoo population of Przewalski's horses has been managed for 60 to 70 years without the aid of genetic management [Bouman, 1977; Flesness, 1977]. An international studbook for the zoo population of red pandas, Ailurus fulgens, was established in 1979 [Glatston et seq., 1980a]. Management of this studbook population began when the zoo population consisted of several wild-born animals and first generation descendants [Princée, 1988, 1989b].

The ChromoFlow model is used to evaluate genetic processes that have occurred in the studbook populations of Przewalki's horses and red pandas. General information on the status in the wild, general biology and status of the zoo population are described for each species in separate sections. Population specific problems such as incomplete pedigree data (Przewalski's horses) and possible low level of genetic variation in the wild population (red panda) are introduced in these sections.

The Przewalski's horse (Equus przewalskii)

Distribution, taxonomy and status in the wild

The Przewalski's horse, Equus Przewalskii, was discovered by Colonel Nikolai Michailovich Przewalskii in Mongolia in 1880, and described in 1881 [Polkajov, 1881]. This wild horse species is considered as the closest living relative of the domestic horse, Equus caballus. However, Przewalskii's horses do have a different number of chromosomes to domestic horses, of 2n=66 and 2n=64, respectively [Benirschke et al., 1965]. Rock engravings and paintings in caves in France, Italy and Spain, originating from 20,000 - 9,000 BC, suggest that its distribution once covered the whole Eurasian steppe belt. Since 1969 no sightings of Przewalki's horses have been documented and, therefore, this species is considered extinct in the wild. The rapid decline of this species since it was (re-)discovered in 1880 is most likely due to the fact that hunting and competition by livestock already had forced this species to live in less than optimal habitat in the peripheral zones of their original distribution. Thus this species was extra vulnerable to increased hunting pressure through the introduction of rifles and competition through intensified grazing of live-stock. See Bouman and Bouman [1994] for more details on the history of Przewalki's horses.

Although this wild horse is extinct in the wild, an ex situ population was established in 1900. On December 1994 this population consisted of almost 1300 horses (see Figure 13) all of which were living in zoos and semi-reserves. In 1992 the first programmes to reintroduce Przewalski's horses in Mongolia started [van Dierendonck and Wallis de Vries, 1996]. All Przewalski's horses living currently are the descendants of 13 founders (including a domestic Mongolian pony mare). Most of these founders were captured between 1899 and 1903. The last founder was captured in 1947 and was included in the ex situ population in 1957. Although Przewalski's horses and domestic horses can produce fertile offspring, no hybrid karyotypes (i.e. 2n=65) have been detected in the studbook population [see Ryder, 1994]. This implies that the 12 founders are most likely pure Przewalski's horses.

International studbook

Breeding records have been maintained by institutions keeping Przewalski's horses since the ex situ population was established in 1900. However, the first pedigree data were not collected and published until 1959 [Mohr, 1959]. Since that time an International studbook has been published annually [Volf, 1960 et seq.; Volf et al., 1991]. Breeding management of Przewalki's horses initially did not involve the maintenance of genetic variation or the avoidance of inbreeding. Two different breeding lines, the 'Munich' and 'Prague' lines, respectively, were maintained in zoos with little gene flow [Ballou, 1994]. This subdivision was mainly due to the fact that institutions maintaining the 'Munich' line were reluctant to introduce horses from the 'Prague' line as this included a Mongolian domestic mare in the pedigree [Bouman, 1982]. Genetic management of the Przewalski's horse population was implemented by the end of the 1970's, before zoos had established regional management programmes [Bouman, 1977; Flesness, 1977; see also The role of zoos in conservation].

Selection and inbreeding may have resulted in expression of recessive traits such as 'fox-colour'. These traits were considered as originating from domestic horses (the domestic founder mare in particular) and therefore considered undesirable by various zoo managers [Princée et. al., 1990]. However, Princée [1990] showed that it is more likely that the trait 'fox-colour' is a single recessive allele which originates from pure founders and not from the domestic mare.

Incomplete pedigree data

Pedigree data of Przewalski's horses were considered as fairly complete [e.g. Ballou, 1994]. However, recent DNA fingerprinting studies showed that the parentage of 24 Przewalski's horses born in Askania Nova between 1979 and 1994 have to be considered (partly) unknown [Zimmerman, pers. comm.]. Assumptions regarding unknown parentages have to be made to enable genetic analyses of pedigrees. The following general strategies can be followed:

  1. Unknown parents are assumed to originate from the wild population.
  2. Select sires and dams, either randomly or according to a 'worst-case' scenario, from the pool of males living at the date of conception and females at the date of birth (of individuals with unknown parents). This "pool" can be defined as all males and females belonging to the reproductive age classes in the breeding group. Inter-birth intervals and average litter size can, additionally, be used to restrict the pool of possible dams.

The strategy to follow depends on the history of the studbook population. Strategy (A) can be a valid approach when historical data show that breeding success in the studbook population is low and imports from the wild occur on a regular basis. This strategy can even be restricted to specific breeding groups (for example, ex situ breeding locations which receive orphanages from the wild ).

Strategy (B) should be applied in cases where no imports from the wild (could) have occurred. The pool of possible parents is composed of individuals which are registered in a studbook. Demographic data to compute reproductive life-span and/or age-specific fertility rates, average litter size and inter-birth intervals are required to apply this strategy. Furthermore, dates of birth and death of all individuals that are (or have been) in the breeding group are required. Reconstruction of the pedigree in strategy can be based on random selection of sires and dams from the pool of possible parents for each individual with unknown parents ('random pool' scenario). Alternatively, selection of sires and dams can be based on 'worst-case' scenario's in which the same individuals are assumed to have sired individuals with unknown parents as long as they belong to the "pool". Inter-birth interval and average litter size need to be taken into account in both scenarios.

The strategies discussed above do not mutually exclude each other, but can be restricted to specific periods of time in the history of studbook populations or even to specific breeding groups. Since Przewalski's horses have not been imported from the wild since 1947, strategy (A) is not valid for solving unknown parentages in Askania Nova herds. Consequently, strategy (B) seems to be the most realistic approach. The effects of assumptions regarding unknown parentages in this population on expected genetic variation (and variances) in the living population as of 31 December 1994 are presented in this study.

The red or lesser panda (Ailurus fulgens)

Distribution, taxonomy and status in the wild

The red or lesser panda, Ailurus fulgens, inhabits slopes of the Himalayan region at heights between 1,500 and 4,000 m. Its main diet consists of bamboo [Yonzon and Hunter, 1989] while its digestive system resembles that of a carnivore [Bleijenberg and Nijboer, 1989]. The red or lesser panda is often considered as the smaller cousin of the giant panda, Ailuropoda melanoleuca, the symbol of the World Wildlife Fund. In the past taxonomists categorized both species in the procyonid subfamily Ailurinae. Modern taxonomists tend to place the bear-like giant panda in the family of bears (Ursidae), while the raccoon-like red panda is considered as the sole representative of the Ailurinae [Glatson, 1989].

Two subspecies of red pandas are recognized: the Nepalese red panda, A. f. fulgens, which originates from Nepal, Sikkim, Northern Bhutan and Assam and the Styan's red panda, A. f. styani, which ranges from Northern Burma into the Yunan and Szechuan provinces of the People's Republic of China [Glatson, 1980b; IUCN, 1994b]. The validity of the subspecies status of A. f. fulgens and A. f. styani was questioned by Roberts [1982] who found no significant differences in results of craniomorphometric analyses. Gentz [1989] suggests that A. f. fulgens and A. f. styani may well be valid subspecies based on data from an electrophoretic study on 25 enzyme systems (see next section).

Information on the status of red pandas in the wild is not complete [IUCN, 1994b]. A two year ecological study of Yonzon and Hunter [1989, 1991] on red pandas in the Lantang National Park in the central Himalayan region in Nepal, however, may be indicative for its status. Nine specimens were observed in a study area of 35 km² within this reserve [Yonzon and Hunter, 1989]. Although the Lantang National Park covers some 1,710 km² of land, only 6% is covered by the preferred habitat of red pandas. Yonzon and Hunter [1991] assume that this specific nature reserve is inhabited by no more than 40 individuals. Furthermore, this small population may be sub-divided in four or more isolated demes. Growing human activities in the Lantang National park seriously threatens the habitat of red pandas [Yonzon and Hunter, 1991]. Information on the status of A. f. fulgens in other areas of its distribution is not available [IUCN, 1994b]. The status of A. f. styani may be deduced from the status of the giant panda. The range and habitat of both species overlap in Szechuan and Yunan. One could assume that wherever and whenever giant pandas suffer from habitat loss the red panda has to cope with same problems. Harvesting from the wild (live and dead) can also be considered as a major threat for, especially, A. f. styani. Relative large numbers of this subspecies have been - though legally - transferred from China to North American and Japanese zoos between 1985 and 1994 [Lu et al., 1993]. It is certain that not all of these individuals were captive-born [IUCN, 1994b].

The uncertain status of red pandas in the wild had its drawback on the level of global protection. This species was not protected as an Appendix I species under CITES until February 1995 [Glatston, 1994]. Although this protection status will not directly reduce threats which are due to habitat loss, it enables control of (legal) transfers of red pandas from habitat countries to zoos and ensures that only zoo-born animals are involved in such transfers.

International studbook

The IUDZG endorsed an International red or lesser panda studbook in 1979 [Glatston et seq., 1980a]. This studbook was from start not solely restricted to the registration of data on specimens in zoological gardens, but was actively used as a basis for population management [Glatston, 1980b]. It may be for this reason that the red panda studbook is one of the first computerized studbooks [Veeke and Glatston, 1980].

The living studbook population of red pandas on 1 January 1978 was mainly composed of A. f. fulgens (143 individuals) and a small number (10 individuals) of A. f. styani [Glatston, 1980a]. Therefore, the demographic analyses [Veeke and Glatson, 1980; Glatston and Roberts, 1988; Glatston, 1989; Glatston and Princée, 1993] and genetic analyses [Princée, 1988, 1989b] that were carried out in the past mainly refer to the studbook population of A. f. fulgens. The status of the zoo population of A. f. styani has drastically changed since initiation of the International Studbook. The eighth edition of the International Red Panda Studbook [Glatston et seq., 1980a] showed that A. f. styani outnumbered living A. f. fulgens with 363 to 258 as of 31 December 1993. This increase in numbers of A. f. styani in zoos was not only due to exports from China and births, but also by the fact that zoological gardens of the People's Republic of China joined the International Red Panda studbook after the "Red panda workshop" (Front Royal, Virginia) in June 1991 [Lu et al., 1993]. This means that data on red pandas in these zoos became available to the International Studbook for the first time. The number of A. f. styani in Chinese zoos was 135 individuals as of 31 December 1993 [Lu, 1994].

The two sub-species of red pandas are managed as separate zoo populations, i.e. the long-term goal is to maintain 90 percent of the original genetic variation in each population. The genetic status of the zoo population of A. f. fulgens as of 31 December 1984 and 1986, respectively, were analysed using the GeneFlow model [Princée, 1988,1989b]. Inbreeding and genetic loss have already occurred in this population. The level of original gene diversity which can be expected to be retained in this population - including living wild-caught red pandas - was 96 percent as of 31 December 1986 [Princée, 1989b]. Since 1984 only two wild-born individuals of this subspecies have been imported by zoos [Glatson et seq., 1980a]. Considering the status of this subspecies in the wild (see previous section), the zoo population of A. f. fulgens needs be managed intensively to maintain sufficient levels of genetic variation without imports from the wild population.

Lu et al. [1993] analysed the zoo population of A. f. styani as on 31 December 1991. The captive-born generation groups in this population descend from 44 founders. The number of potential founders, i.e. living wild-born animals which have not reproduced, in this population is 150. Gene Drop simulations with the GENES software program [Lacy, 1994] indicate that 97,6 percent of the original "wild" gene diversity is retained in the captive-born generation. The potential gene diversity that can be retained, provided that the potential founders breed, is 99,7 percent of the original "wild" gene diversity. Considering the (potential) level of genetic variation in the A. f. styani zoo population, short term management strategies will primarily focus on improving breeding success [Glatston and Princée, 1993]. This study will, therefore, mainly refer to genetic management of the zoo population of A. f. fulgens.

Genetic variation in wild red panda populations

Results of simulation experiments in Chapter 3 show that assumptions regarding genetic composition may risk genetic variation being lower than expected. This implies that models that reflect the real genetic variation of a source population need to be used in simulation experiments involving real (e.g. studbook) populations. These models could be based on data from biochemical and molecular studies.

Gentz [1989] studied 25 enzyme systems (presumptive genetic loci) in A. f. fulgens (N=10) and A. f. styani (N=12). No heterozygous loci were observed and both sub-species were found to be polymorphic at a single locus: glucose-6-phosphate dehydrogenase (G6pd). Allele types A and C with frequencies of 0.1 and 0.9 respectively, were observed in A. f. fulgens and allele types A and B, with frequencies 0.333 and 0.666, respectively, were observed in A. f. styani. The fraction of polymorphic loci in each subspecies is 0.04. Values of average gene diversity over 25 loci in the studied groups are 0.0072 and 0.0178 for A. f. fulgens and A. f. styani, respectively.

The group of A. f. styani is presumed to consist of 12 unrelated wild-born individuals. Following equation 7, the expected original gene diversity in the wild population of this sub-species is 0.0185. Gene diversity in the wild population of A. f. fulgens is expected to be higher than determined by the data from Gentz [1989]. The studied group does not represent a random sample from a wild population as zoo-born animals that are related and/or inbred are included [see Glatston et seq., 1980a; Princée, 1988; Gentz, 1989]. The expected gene diversity in the wild population of A. f. fulgens can be extrapolated from values for gene diversity in the studied group as estimated through simulation experiments with ChromoFlow.

Genetic variation in red pandas can be considered low compared to average levels of polymorphism and gene diversity as obtained from protein electrophoresis in other mammals (see Table 1). However, the results from Gentz's electrophoretic study [1989] can be subject to sampling error. First, a relatively small number of individuals (N=10) is included in this study. This means that alleles which occur at lower frequencies in the wild population may not have been sampled (see also Figure 9). Second, as captive-born individuals are included in the electrophoretic study, alleles can have been lost due to genetic drift. Effects of both sampling individuals and genetic drift can be evaluated by ChromoFlow simulation experiments that assume different allele frequencies at the studied 25 loci in the wild population.

Material and methods

Pedigree data of Przewalski's horses have been extracted from the International studbook [Volf et al., 1991]. Additional pedigree data for the years 1990-1994 have been extracted from the EEP studbook of Przewalskii's horses [Zimmerman, pers. comm.]. Pedigree data of Nepalese red pandas have been extracted from the eight edition of the International Red or Lesser Panda Studbook [Glatston et seq., 1980a].

Experiments with the ChromoFlow simulation model are listed in table 13. These experiments are grouped as follows: estimation of genetic variation in the Przewalski's horse studbook population at census dates for reconstructed pedigrees (experiment 1); estimation of genetic variation in the wild red population from combined results of electrophoresis and pedigree simulations (experiment 2); expected genetic variation in molecular genetic studies for different levels of genetic variation in the wild red panda population (experiments 3 to 8); and assessment of the risk that actual genetic variation in the red panda studbook population is lower than expected for low initial levels of genetic variation (experiments 9 and 10).

Reconstruction of the incomplete pedigrees in the studbook population of Przewalski's horse is based on assumptions following the 'worst-case' scenario of strategy (B). The analyses package of the Zooresearch Studbook Management computer program [Princée, 1991; unpublished] have been used to compute census data, construct life-tables and to determine litter size from studbook data of Przewalski's horses. A genome model with 36 independent autosomal loci each with two alleles and equal frequencies is assumed in experiment 1. 10,000 simulation runs are used in this experiment.

Experiment 2 involves estimation of genetic variation in the group Nepalese red pandas which were included in study of Gentz [1989]. This group is referred to as the 'Gentz' sample. The expected gene diversity in the wild population (He[wild]) can be estimated using the following equation:

where Fraction He[CHROMOFLOW] is the fraction of original gene diversity that has been retained in the studied group as estimated in ChromoFlow simulations; and He[electrophoresis] is the gene diversity as computed from observed allele frequencies in electrophoretic studies. The value of He[CHROMOFLOW] for the 'Gentz' sample is estimated from results of experiment 3, which involves a genome model with 36 independent autosomal loci and two equally distributed allelic variants at each locus. The number of simulation runs used in this experiment is 100,000.

The third series of experiments (3 to 8) involves estimation of minimal gene diversity that can be expected in the 'Gentz' sample for different allele frequencies in the wild population. The number of loci as included in the study by Gentz [1989], i.e. r=25, and two alleles at each locus are assumed in these experiments. 25,000 simulation runs are used. This series of simulation experiments comprises two subsets. The first subset assumes initial frequencies of the first allele of 0.50, 0.25, 0.10 at each of the 25 loci for experiments 3 to 5, respectively. The second subset (6 to 8) assume levels of genetic variation in the wild population which more closely resemble the results of the study by Gentz [1989]. Allele frequencies at the first locus in these simulations resembles those as observed at the G6pd locus (i.e p1=0.1). Frequencies of 0.05, 0.025 and 0.01 for the first allele are assumed at the other 24 loci in experiments 4 to 6, respectively.

The minimal expected gene diversity He[min] is determined from the ChromoFlow distribution classes of values for gene diversity [see Chapter 2 for details]. The lower limit of the lowest class in this distribution that shows a frequency greater than 0.0 is considered as the minimal expected gene diversity. Furthermore, the probability that a value for gene diversity of 0.0072 ( p[He=0.0072] ) is observed in the 'Gentz' sample is determined from the ChromoFlow distribution classes for each simulation experiment in the series 4 to 9.

The fourth series of ChromoFlow simulation experiments (9 and 10) involve estimation of expected genetic variation in the living zoo population of red pandas at annual census dates (31 December) since 1978. A genome model assuming 36 independent loci with two alleles at each locus is assumed in these experiments. Experiment 9 assumes that allele frequencies are equal at all (36) loci. A frequency model assuming that p1=0,90 at all loci has been used in experiment 10 to determine the risk that genetic variation in the living zoo population of A. f. fulgens is lower than expected. 10,000 simulation runs are used in both experiments.

Results

Przewalski's horse

Demographic analyses show that Przewalki's mares predominantly produce one foal per breeding season (4 twin births out 2588 births) and that mares in the ex situ population can give birth annually. The reproductive life-span of Przewalski's horses ranges from age 2 to the mid-20s [see also Ballou, 1994]. These data have been used to construct a list of possible parents of Przewalski's horses with incomplete data on parentages according strategy (B). This list is included in Appendix B.

Figure 13 presents population size and mean values for gene diversity and observed heterozygosity at census dates since 1900 for the reconstructed pedigree. Two major historic trends can be observed in the studbook population of Przewalski's horses. The first trend refers to a period of slow growth from 24 horses in 1901 to 58 in 1959. The second trend refers to the period 1960 - 1994 during which the population increased to 1293 horses. Gene diversity and observed heterozygosity decline until the mid 1960's. Note that gene diversity decreases gradually, while the decline pattern of observed heterozygosity shows fluctuations (see Figure 13). Although gene diversity and observed heterozygosity have increased since 1965, both values as of 31 December 1994 were 84.5 and 79.6 percent of the original (wild) genetic variation, respectively.

Red panda

ChromoFlow simulation experiment #2 with 36 loci and equal allele frequencies shows that the expected fraction of original (wild) gene diversity that is retained in the 'Gentz' sample is 0.868. If equation 19 is applied, the expected gene diversity in the wild population of this red panda subspecies is 0.0083.

Table 14 presents the minimal expected gene diversity (He [min]) in the 'Gentz' sample for the different allele frequencies that are assumed in the wild population (25 independent autosomal loci and 2 alleles per locus). This table shows that values for the minimal expected gene diversity are larger than the observed gene diversity of 0.0072 for allele frequencies in the range 0.05 to 0.5. Probabilities that gene diversity in the 'Gentz' sample is 0.0072 ( p[He= 0.0072] ) for different initial values of gene diversity (He[wild] ) are also presented in table 14. These results show probabilities of 0.006 and 0.072 for initial low levels of gene diversity of 0.054 and 0.026, respectively.

Figure 14 shows population size and fraction of original (wild) gene diversity and observed heterozygosity that is expected in the living studbook population of A. f. fulgens at census dates since 1978. Gene diversity declines more rapidly than observed heterozygosity between the years 1978 and 1985. The rate of decrease between 1985 and 1993, however, is lower than between the years 1978 and 1985. Observed heterozygosity slightly increases in 1987. The expected fractions of original gene diversity and observed heterozygosity as on 31 December 1993 are 0.953 and 0.974, respectively.

Census data for the studbook population of A. f. fulgens are presented in table 15. Population size has been increasing since 1986 due to an increase in the number of births which outnumbers the deaths. No imports of wild individuals have occurred since 1988.

Figure 15 shows effects of initial allele frequencies in ChromoFlow simulations on distribution of gene diversity in the living studbook population of A. f. fulgens as on 31 December 1993. The expected gene diversity belongs to class 19. Note that all values for gene diversity belong to this class for a model that assumes equal allele frequencies on 36 autosomal independent loci in the wild A. f. fulgens population.

The probability that gene diversity and observed heterozygosity are lower than 95 percent of the expected values (=mean values in simulations) of these genetic parameters at census dates are shown in figure 16. These data are based on results of ChromoFlow experiment 10. The probability that values of genetic variation are lower than 95 percent of the expected values increase per year (see Figure 16).The probability that gene diversity and observed heterozygosity are lower than expected as of 31 December 1993 are 0.302 and 0.312, respectively.

Discussion

Handling incomplete pedigree data

The assumptions which are made to solve unknown parentages of studbook individuals effect the pedigrees of their descendants in subsequent generations, and therefore have an impact on the results of genetic analyses. These effects depend on assumptions made regarding the origin of unknown parents and methods used to assign possible parents. The assumption that unknown parents originate from the wild implies that individuals with unknown parentages have to be considered as (potential) founders. This can result in an overestimation of the genetic variation retained in the studbook population and, depending on pedigree patterns, in underestimation of inbreeding coefficients [see also Ballou and Lacy, 1995].

The effects of the assumption that unknown parents originate from the studbook population depend on the method to assign parents from pools of possible parents. The 'worst-case' scenario has been applied in reconstruction of (a small part of) the pedigree of Przewalski's horses (see Appendix B) . This scenario considers all individuals (with unknown parents) which originate from the same breeding group to be related by assigning the same sire and dam (within the biological constraints as discussed on page 83). This means that inbreeding coefficients will be overestimated and genetic variation underestimated. The 'worst-case' scenario can be considered an appropriate method to minimize possible inbreeding. However, this scenario can result in increased genetic loss in future generations whenever individuals are paired according mean kinship values, the average relatedness of individuals to the rest of the population [Ballou and Lacy, 1995]. Mean kinship values of individuals which are assumed to be related, can be overestimates and therefore (wrongly) result in excluding these individuals from the breeding pool.

Further studies are required to determine the consequences of using 'worst-case' scenarios in handling unknown parentages on genetic management of studbook populations. These studies should also involve the 'random pool' scenario (see page 83). This requires implementation of algorithms in simulation models such as ChromoFlow and GENES [Chapter 2; Lacy, 1994), which randomly assign possible parents in each run. Results of these simulations provide averages and ranges for genetic variation and mean kinship values. It is important to determine whether both scenarios can be combined in genetic management. This means that minimizing inbreeding should be based on 'worst-case' scenarios and breeding recommendations on average mean kinship values (of individuals).

The previous discussion is based on the underlying assumption that studbooks include all individuals in the ex situ population. Given the historical attention by zoological gardens and other organizations this assumption may be valid for this specific studbook population. However, studbooks do not necessarily cover the entire history nor include all individuals which are or have been in ex situ populations. In-house registration of zoo animals has started to increase over the last ten to fifteen years [Flesness and Mace, 1988; Flesness et al., 1995]. This means that complete overviews of the captive history of a population is not always available. Furthermore, adequate in-house registration of studbook species has not yet been implemented by all (zoological) institutions and private owners, worldwide.

Incomplete studbook overviews can limit the possibility (and the validity) of reconstructing pedigrees from pools of possible parents. This introduces true unknown parents into the pedigree. In this, two types of unknown parents can be distinguished: (1) those of individuals which are born within the studbook population and; (2) those of individuals which originate from external populations (other than the original wild population) for which no data are available. The strategy to reconstruct pedigrees from pools of possible parents can be applied to handle the first type of unknown parents. In this, it is assumed that at least all individuals which have reproduced are included in the studbook.

Strategies to handle the second type of unknown parents depend on the level of exchanges (migration) between studbook populations and external populations. Assumptions could be based on 'worst-case' scenarios where individuals imported from external populations are considered siblings, born to a single hypothetical founder pair. However, strategies will become more complex in those cases where multiple exports, involving different founder lineages, from the studbook population to external populations have occurred. This means that individuals which are imported from external populations can be related to those already in the studbook population. However, the level of relatedness (mean kinship) of these individuals is unknown.

It is disputable whether strategies to handle incomplete pedigree data as discussed above should be applied whenever studbooks include a relatively large proportion of individuals with unknown parents. For example, the 'random pool' scenario can result in the same effect as if such populations were considered panmictic. Further studies are required to determine which population model is appropriate for different proportions of unknown parentages. Furthermore, metapopulation management can involve combined management of studbook populations and external - either in situ or ex situ - populations for which minimal data may be available. Assuming that individuals originating from these external populations are related to each other, implies that the metapopulation concept is irrelevant to the genetic management of the studbook population. Chapters 5 and 6 of the present study involve genetic management of small (in situ and ex situ) populations on the basis of limited (pedigree) data.

Molecular genetics and studbook data

The Nepalese red pandas (A. f. fulgens) which were studied by Gentz [1989] is assumed to represent 86.8 percent of the gene diversity in the wild population. The expected gene diversity in the wild population can be extrapolated from results of simulations and protein electrophoresis of the 'Gentz' sample (see page 88) using equation 19. However, protein electrophoretic studies are subject to sampling errors as discussed on page 87 [see also Chapter 3]

.

Results of the ChromoFlow experiments 3 to 8 indicate that expected gene diversity in red pandas could be as low as that observed in protein electrophoresis [Gentz, 1989]. Higher values of gene diversity than those observed in protein electrophoresis would be expected if gene diversity in the wild population ranges from 0.18 to 0.50 (Table 14). The second subset of experiments resemble models in which rare alleles are assumed at the 24 loci that were shown to be monomorphic in the electrophoretic study. Results of these experiments show that gene diversity in the 'Gentz' sample would still expected to be higher than indicated by electrophoretic data for these rare allele models. The probability that a gene diversity of 0.0072 was observed in the 'Gentz' sample is 0.006 and 0.0072 for models which assume allele frequencies of 0.025 and 0.010, respectively (Table 14). These results indicate that the 24 loci which were shown to be monomorphic in study are either monomorphic in the wild population or have allele frequencies lower than 0.01.

Low levels of genetic variation in the original (wild) population result in high probabilities that genetic variation in studbook populations are lower than expected, i.e lower than average values in simulation experiments [See Chapter 3]. This means that the low genetic variation in the wild Nepalese red panda population should be taken into account in breeding management. The fraction of wild gene diversity and observed heterozygosity in this studbook population was expected to be 0.953 and 0.974, respectively, as on December 1993 (Figure 14). Figure 15 shows that gene diversity in this population can be lower than the average value (class 20) for a model that assumes low initial allele frequencies (experiment 10; Table 13). Probabilities that actual gene diversity and observed heterozygosity for this model are lower than 95 percent of the average values are 0.302 and 0.312, respectively, on December 1993 (Figure 16).

The maintenance of at least 90 percent of the original (wild) gene diversity for a period of 100 to 200 years is an important objective of breeding programs [Soulé et al., 1986; see also The role of zoos in conservation]. Although the average percentage of original gene diversity in the studbook population of Nepalese red pandas is 95.3 percent, a high probability (0.302) exists that actual gene diversity is lower than 90.5 percent (i.e. 0.95 x 95.3). This implies that it will be difficult to maintain gene diversity above the '90 percent criterion' in this studbook population.

This study shows the importance of collecting empirical data on genetic variation in wild populations to assess the risk that genetic variation is overestimated by simulation models. Vice-versa, simulation experiments on the 'Gentz' sample illustrates that models such as ChromoFlow can be used to estimate genetic variation in wild populations through extrapolation of results from molecular studies on descendants of wild-born animals. Furthermore, sampling errors in the results of such studies can be determined by using different assumptions on initial genetic variation in the source (wild) population. The use of ChromoFlow requires, however, that (complete) pedigree data are available. Modern zoological institutions maintain computerized animal registration data that include pedigree information [Flesness and Mace, 1988; Flesness et al., 1995]. This means that individuals from zoo populations can be used - regardless the complexity of their pedigree - in studies on genetic variation of the original wild population. As more and more species become endangered [see Chapter 1] it may not be feasible to retrieve blood and tissue samples from wild individuals; zoo populations can provide alternatives.

Population trends

The management of the studbook population of Przewalski's horses did not involve the purposeful avoidance of inbreeding and maintenance of genetic variation until the end of the 1970's [Bouman, 1977]. The consequences of this management on genetic variation in the studbook population can be observed in figure 13. Gene diversity had already dropped below the '90 percent criterion' in 1950. Due to inbreeding observed heterozygosity declined more rapidly than gene diversity. However, the increase in observed heterozygosity during the period 1940 to 1950 indicates that exchanges of individuals between breeding herds, resulting in non-inbred offspring, must have occurred in that period. A similar pattern can be observed in the period 1967 to 1994 (Figure 13). Observed heterozygosity increases from 72.6 to 79.6 percent in this period. This indicates a population management that aims at minimizing further inbreeding in the population.

Gene diversity also increases during the period 1967 to 1994 ( from 82.1 to 84.5 percent; see Figure 13). Since no individuals were imported from the wild, this slight increase must be the result of the combined effects of rapid population growth and implementation of genetic management. For example, increasing reproduction of individuals which represent rare founder lineages, can result in restoring original allele frequencies in the population. However, restoring gene diversity and observed heterozygosity are limited (Figure 13) since the fixation of alleles at some loci already occurred in earlier generations.

The zoo population of Nepalese red pandas is managed following avoidance of inbreeding and maintenance of 90 percent of the wild gene diversity [see Glatston et seq., 1980a; Princée, 1988,1989b; Glatston and Princée, 1993]. Observed heterozygosity in the studbook population is higher than gene diversity (Figure 14). This shows the effect of a management strategy aimed at minimizing inbreeding. The slight increase in observed heterozygosity in 1988 is the result of reducing inbreeding in regional populations by exchanging individuals between regions [see Princée, 1989b]. Gene diversity in this population decreased since 1978 from 99 to 95.3 percent of the wild gene diversity. This decrease mainly occurred between 1978 and 1985 (i.e. a loss of 3.1 percent) and has been reduced to 0.6 percent between 1985 and 1993 (see Figure 14). Reduction in genetic loss is the result of both population growth since 1985 (see Figure 14 and Table 15) and implementation of genetic management involving planned matings [Glatston et seq., 1980a].