Ancestral Inbreeding Only Minimally Affects Inbreeding Depression in Mammalian Populations

Transcription

1 Ancestral Inbreeding Only Minimally Affects Inbreeding Depression in Mammalian Populations J. D. Ballou Inbreeding depression can be reduced, or purged, by selection against deleterious alleles. This prediction Is the basis of the recommendation that captive wildlife populations suffering from inbreeding depression be intentionally bred from healthy Inbred animals. Yet data on the effectiveness of purging Inbreeding depression are few. In this study I present and use two different regression models (an ancestral inbreeding model and a lethal recessive model) to test for the presence of purging effects in 25 captive mammalian populations. Fitness components examined were neonatal survival, survival from neonate to weaning, and litter size. In only one species was purging statistically significant. However, 15 of 17 species that showed Inbreeding depression exhibited a slight decline in Inbreeding depression In neonatal survival among descendants of inbred animals. These results show a small but highly significant trend of purging on neonatal survival. No trends In purging effects were observed in weaning survival or litter size. The effects were not likely to be strong enough to be of practical use in eliminating Inbreeding depression. From the Department of Zoological Research, National Zoological Park, Smithsonian Institution, Washington, D.C The data required lor pedigree analyses demand decades of meticulous record collection and management by many Individuals and Institutions worldwide. I thank all those Individuals and Institutions (listed in Table 1) for contributing their studbook databases to this study, and particularly at the National Zoo, I thank Judith Block, Barbara At wood, Frank Kohn, and Angela Marlow for their years of work In maintaining detailed animal records at'the National Zoo. This work benefited greatly Irom discussions with Jerry Wilkinson, Bob Lacy, Dick Frankham, Phil Hedrick, Un Chao, and Michele Dudash. The work was supported In part by the Friends of the National Zoo (FONZ). Journal of Heredity 1997^8: : /97/15.00 Inbreeding depression has been documented in numerous plant (Charlesworth and Charlesworth 1987) and animal species (Lacy et al. 1993; Rails et al. 1988; Thornhill 1993; Wright 1977). Two genetic mechanisms have been proposed as the cause of inbreeding depression. Both relate to the decrease in heterozygosity during the inbreeding process. The dominance hypothesis proposes that fitness depression results from the increased expression of deleterious recessive alleles (mutational load) during inbreeding, while the overdominance hypothesis proposes that depression is the result of declining heterozygosity among loci exhibiting heterozygote superiority (Charlesworth and Charlesworth 1987; Wright 1977). In general, dominance effects (the presence of deleterious recessive alleles) are thought to account for a large proportion of the inbreeding depression observed (Charlesworth and Charlesworth 1987; Lande and Barrowclough 1987; Lande and Schemske 1985; Morton et al. 1956; Simmons and Crow 1977; Wright 1977). Regardless of the genetic mechanism, fitness is expected to decline as inbreeding increases (Crow and Kimura 1970). In the presence of selection, however, inbreeding depression caused by deleterious alleles can in theory be reduced. Selection against deleterious recessive alleles is intensified under inbreeding because inbreeding increases the frequency of the recessive homozygous genotype (Crow and Kimura 1970), and in the absence of mutation, the population can be "purged" of its mutational load (Barrett and Charlesworth 1991; Charlesworth and Charlesworth 1987; Hedrick 1994; Lande and Schemske 1985). Fitness can increase and return to or even exceed that of the randomly breeding (noninbred) population (Hedrick 1994; Lande and Schemske 1985). Thus the level of inbreeding depression observed in a population is a function of the genetic mechanism of inbreeding depression as well as the mating system and demographic history of the population. For this reason the genetic basis for inbreeding depression is an important factor in the evolution of selling versus outcrossing mating systems (Charlesworth and Charlesworth 1987; Lande and Schemske 1985). It is also of interest to conservation biologists concerned with both the impact of inbreeding depression on population viability and the potential of purging a population of its genetic load through management intervention to reduce disasteriously high levels of inbreeding depression. In the captive population 169

2 of Speke's gazelle (Gazella spekei) inbreeding depression was reported to be reduced over two to three generations by maximizing retention of genetic diversity and selecting healthy, inbred animals as breeders. Inbred animals surviving to reproductive age are less likely to carry deleterious alleles than noninbred animals. As a result, Templeton and Read (1984) reported that inbreeding depression was lower In offspring bom to these selected inbred parents than it was prior to the selection program. Based on these results, this "purging" strategy has been recommended for use in other captive populations suffering from severe inbreeding depression (Ballou 1989; Foose et al. 1986; Rails and Ballou 1986; Templeton and Read 1983, 1984; Templeton et al. 1986). More recently, however, the Templeton and Read analysis was revisited by Willis and Wiese (1997) who concluded that purging had actually not occurred in the Speke's gazelle and that the results obtained by Templeton and Read were due to methods used in their (Templeton and Read's) statistical analyses. While the Speke's gazelle may not show significant purging effects, the question still remains as to the utility and feasibility of purging inbreeding depression from severally inbred populations using a strategy such as was used with the Speke's gazelle. There are very few data on the extent to which inbreeding can be reduced through such a "purging" strategy. If the strategy to purge inbreeding depression by selectively breeding inbred animals is successful in captive populations, then we might expect to see evidence of purging in many captive populations that have inbred ancestry. An inbred animal with inbred ancestry should be less susceptible to Inbreeding depression than an inbred animal with noninbred ancestors because surviving and reproducing inbred ancestors are less likely to be carriers of deleterious alleles (Templeton and Read 1984). If this is the case, then this also suggests that dominance effects are an important mechanism for inbreeding depression in these species. I present two different regression models to evaluate purging effects and investigate the mechanism of inbreeding depression. These models are then used to examine pedigrees of 25 populations of captive mammals for evidence that inbreeding depression has been purged or reduced through selection upon ancestry of inbred animals. More specifically, I am interested in determining if the level of inbreeding depression has been reduced by inbreeding among ancestors of inbred indi- f. =.125 LL"O Figure 1. Simple pedigree Illustrating the calculation of I, f, (ancestral Inbreeding coefficient), and LL values. viduals. Inbreeding depression is measured on three components of fitness: neonatal survival (survival to 7 days of age), survival from 7 days to age of weaning, and where appropriate, litter size. The two purging models are applied to each of these fitness components. Methods The two models used here are based on analyses of a population's pedigree and measure the extent to which selection upon inbred ancestors of inbred individuals modify the inbred individuals' susceptibility to inbreeding. Ancestral Inbreeding Model The extent of inbreeding among an individual's ancestors can be measured using the ancestral inbreeding coefficient (f a ). The value / is denned as the cumulative proportion of an individual's genome that has been previously exposed to inbreeding in its ancestors: f. = CO where f a Is the ancestral inbreeding coefficient for an individual, A is the Inbreeding coefficient and the subscripts s and d represent these values for the sire and dam of that individual, respectively. An individual's f a is then the proportion of its parent's genome that has been previously exposed to inbreeding (f a of the parent) plus the effect of the parent's inbreeding coefficient on the proportion that has not been previously exposed (1 Q, averaged across both parents; it ranges from 0 to 1. Calculation of f a values for a simple pedigree are shown in Figure 1. Typically, inbreeding depression effects are modeled by regressing some component of fitness against inbreeding coefficient: u = u o (2) where u is a measure of fitness, u 0 is mean fitness for noninbred animals, f is the inbreeding coefficient, and B, Is the slope (regression coefficient) of f regressed against fitness. The severity of inbreeding depression is determined by the magnitude and sign of B A When u = -log(survival), 2B, is a measure of the number of lethal equivalents per diplold genome in the population (Morton et al. 1956; Rails et al. 1988; Templeton and Read 1984). The model I used Includes the ancestral inbreeding coefficient as a modifier of the inbreeding depression effect, as well as effects for time [year of birth (YOB) to control for changes in husbandry over time] and maternal inbreeding [inbreeding coefficient of dam (Q]. Although the maternal inbreeding coefficient is a component of the ancestral inbreeding coefficient (Equation 1), it was included as a separate covariate because maternal inbreeding is often associated with poor offspring survival independent of the inbreeding coefficient of the offspring (Brewer et al. 1990; Rails et al. 1980). Furthermore, detrimental maternal effects can mask positive purging effects if they are not considered separately. The model used then becomes u = u 0 + B,KOB + p, A + B,//- a + B,/ rf (3) where u, u 0, B ft f, and f a are as described above in Equations 1 and 2; p, is the regression coefficient associated with year of birth (YOB); 6,, is the regression coefficient associated with maternal Inbreeding (Q; ff a is the interaction between inbreeding and ancestral inbreeding; and B, # is the regression coefficient associated with the interactive term. In this model, survival of noninbred animals is independent of ancestral inbreeding, but the inbreeding effect can be modified by the level of ancestral inbreeding (note that f o is entered in the equation only as an interaction with the inbreeding coefficient). This can be seen by expressing Equation 3 as if = u 0 + $,YOB + (p, + p,.4) f + B,/ ( (4) 170 The Journal oj Heredity 1997:88(3)

3 The coefficient associated with f is now a combination of the Inbreeding effect (PiD and the ancestral inbreeding effect 0,/J. Inbreeding depression is characterized by P, < 0. If there has been purging, then we predict that the coefficient p, # will be positive, thereby reducing the inbreeding effect. Lethal Recessive (LL) Model Slatls (1960) first proposed a model of purging based on the assumption that each founder of a pedigreed population carried a single, lethal recessive allele at a different locus. Using path analysis, he estimated the probability of lethal homozygosity (LL) as the probability that an individual was homozygous for any lethal allele under the assumption that none of the individual's ancestors could have been homozygous for any lethal allele. Thus LL measures autozygosity of genes that are being selected against. Slatis (1960) hypothesized that if inbreeding depression was due to the presence of lethal recessives in the population, then the relationship between survival and homozygoslty would be better predicted by the regression of LL on survival than by regressing the inbreeding coefficient on survival. Slatis (1960) applied this model to the captive population of European bison (Bison bonasus), and not very convincingly claimed that the LL model better fit the survival data than did the Inbreeding coefficient model. I used a similar approach to test the LL model against the inbreeding coefficient model as a means of examining pedigrees for the presence of a purging effect. As mentioned by Slatis (1960), calculating LL values is computationally complex in complicated pedigrees, and Slatis's original calculations and methods were only approximate. In addition, Slatis seems to have assumed that the probability of inheritance of alleles from the dam and sire were independent (which is not the case in Inbred populations), leading to errors in his calculations of some LL values. I therefore calculated LL values using Monte Carlo simulations. Each founder was assumed to carry one lethal recessive at a different locus. For each individual in the pedigree, the probability of receiving homozygous lethal alleles at any locus was estimated by simulating gene transmission (dropping genes, MacCluer et al. 1986) from the founders to the individual under the assumptions of random assortment and Mendellan segregation. During a simulation if any ancestor of the Individual received a homozygous lethal genotype at any locus, the genotype of its parents were resampled until a nonlethal genotype for that locus was obtained. Ten thousand simulations were conducted for each individual and LL was defined as the proportion of simulations in which the individual received a homozygous lethal genotype at any locus. LL values in a simple pedigree are shown in Figure 1. Depending on the structure of the pedigree, LL may be highly correlated with F. The value LL was used in the inbreeding depression regression model with the time effect (YOB): u = u o + fryob + p u (LL) (5) where f} u is the regression coefficient associated with LL To test for the presence of purging of lethal recessives, the fit of Equation 5 was compared to the fit of a simple inbreeding model (Equation 2) with the time effect (fi.yob) added: u = u o p,kob + p, f. (6) If inbreeding depression is caused by lethal recessives, then Equation 5 should have a better fit than Equation 6 (Slatis 1960). Data and Statistical Analyses The 25 populations of captive mammals analyzed are listed in Table 1. Values for f, f a, f d, and LL were calculated for each individual. The fitness components analyzed were survival to 7 days of age (neonatal survival), survival from 7 days to age of weaning, and litter size. For those species not producing litters, each individual was coded as either surviving to (coded 0) or dying before (coded 1) each of the survival ages. For those species producing litters, I controlled for nonindependence of within-lltter mortality by analyzing survival of litters rather than individuals. A litter was coded as surviving if average survivorship of litter mates was at or above the average survivorship in the population. Litter size at the time of weaning was also recorded and coded as either larger or smaller than average noninbred litter size. Animals with unknown birth or death dates or unknown ancestry were excluded, as were any animals born within weaning age of the cutoff date of the data. Multiple logistic regression was used to estimate the regression coefficients in Equations 3, 5, and 6 using the SAS LOGIS- TIC procedure (SAS 1991). In logistic regression, the fitness function takes the form u = 1 + (7) where x takes the form of the right side of Equations 3, 5, and 6. Coefficients are estimated using maximum likelihood and their statistical significance tested with likelihood ratio tests (Hosmer and Lemeshow 1989). Examination of collinearity among variables was conducted using the SAS REG procedure invoking the COLU- NOINT option (SAS 1991). Comparison of model fits (Equation 5 to Equation 6) was based on comparison of the Akaike Information Criterion (AIC) values for each model. Lower AIC values indicate better fit (SAS 1991). Trends across species were tested using the sign test. Change in Inbreeding Depression Due to Purging Inbreeding depression can be expressed as 8 = 1 w, /w o, where w, is the fitness of inbred animals (at some specific level of inbreeding) and w o is the fitness of noninbred animals (Lande and Schemske 1985). Fitness values (w n wj are determined from Equation 3 using the regression coefficients estimated from the data. Inbreeding depression in neonatal survival, weaning survival, and litter size was calculated for each taxon from the regression estimates in two ways: (1) 5 for inbreeding at the level of f = 0.25 using the estimated inbreeding effect fa and (2) 8' for inbreeding at the level of f = 0.25 including the estimated ancestral inbreeding effect (p,j applied to the mean value of f a for each taxon. The first method (8) is the inbreeding depression in individuals with noninbred ancestors (I.e., without purging), while the second (8') is the inbreeding effect after the opportunity for selection to purge deleterious alleles. The difference (8' - 8) Is the change in inbreeding depression at f = 0.25 due to ancestral inbreeding. To statistically determine if inbreeding depression has been eliminated by purging, I calculated a cumulative inbreeding effect (PJ, defined + &/ ). This is the inbreeding effect taking the ancestral inbreeding effect into consideration (see Equation 4). The p c and their variances were calculated for each species using the species' mean f a and the estimates of p, and p /# obtained from the logistic regression analysis. Variance of p c is calculated as were obwhere <T 2 (P,), (^(PJ, and (8) BalkXJ Ancestral Inbreeding in Mammalian Populations 171

4 Table 1. Source* and sample size* of data for taxa analyzed Taxon Elephant shrew* Elephantulus rufescens Golden lion tamarln* Leontopilhecus rosalia Golden-headed lion tamarln* Leonlopllhecus chrysomelas Black lion tamarln* Leonlopllhecus chrysopygus Goeldl's marmoset Callimico goeldii Brown lemur Lemur fulvus Greater galago Galago c crassicaudatus Melanotlc galago Galago c. argentatus Orang utan Pongo pygmaeus Kerodon* Kerodon rupestris Boris* Octodontomys gliroides Puna re* Cercomys cunicularus Maned wolf* Chrysocyon brachyurus Red panda* Ailurus fulgent Asiatic lion* Panthera leo persica Sumatran tiger* Panthera ligris sumatrae Przewalski's horse Equus przewalskii Pygmy hippopotamus Choeropsis iibenensis Muntjac Muntiacus reevesi Eld's deer Cervus eldi lhamin Gaur Bos gaurus European bison Bison bonasus Dorcas gazelle Gazella dorcas Speke's gazelle Gazella spekei Nllglri tahr Hemitragus hylocrius N 189 1, , , , , Average f>0(n) (57) 0.070(491) 0.250(34) (27) 0.100(264) (19) (90) 0.239(22) (70) (62) (22) (31) 0.196(137) (166) (122) (253) 0.216(1,856) 0.220(166) 0.162(94) (162) (360) (2,700) 0.195(111) (136) (151) Average /. > 0 (N) (22) 0.052(517) 0.189(15) (39) (220) (4) (61) 0.179(15) 0.116(7) 0.164(37) 0.142(15) 0.180(8) 0.159(119) (207) 0.237(101) 0.180(187) 0555(1,894) (259) 0.197(77) 0 178(122) 0.239(398) (2,776) (132) 0.176(100) (126) 1 Indicates analysis conducted on litters rather than Individuals. * N (or LL > 0 = yv for f > ft tained from the SAS LOGISTIC procedure invoking the COVOUT option (SAS 1991). Results Average LL>0" Source National Zoological Park Records 1993 Studbook 1993 Studbook (de Bols 1994) 1993 Studbook (Padua 1994) 1994 Studbook, Warneke, Brookfield Zoo Oregon Regional Primate Center, 1983 Oregon Regional Primate Center, 1983 Oregon Regional Primate Center, Studbook (Perkins 1994) National Zoological Park Records National Zoological Park Records National Zoological Park Records M. Rodden, National Zoo, and 1992 Studbook M. Roberts, National Zoo and 1993 Studbook (Glatston 1994) 1993 Studbook (Fouraker et al. 1993) S Christie, London Zoo and 1993 Studbook 0. Ryder, San Diego Zoo and 1992 Studbook (Voli 1991) 1993 Studbook (Tobler 1993) National Zoological Park Records 1993 Studbook (Wemmer 1993) D. Morris, Henry Doorly Zoo and 1991 Studbook (Kl&s 1992) 1988 Studbook (Pllarsld 1988) National Zoological Park Recordj 1988 Studbook (Read 1988) 1994 Studbook (Swengel 1994) Ancestral Inbreeding Model Levels of Inbreeding, ancestral inbreeding, and LL values varied greatly among species (Table 1). Average / for Inbred animals was 0.17, average f a for animals with f a > 0 was 0.17, and average LL was Analysis of the ancestral inbreeding effects could not be conducted in six populations because of limitations in the distribution of f a values. These populations were the golden-headed lion tamarin, black lion tamarin, brown lemur, melanotic galago, orang utan, and punare (Table 1). Furthermore, in four more populations (elephant shrew, golden lion tamarin, boris, and red panda), maternal inbreeding was sufficiently confounded with ancestral inbreeding (i.e., ancestral Inbreeding was limited to maternal inbreeding) that inclusion of both f d and f a in the model was not possible. In these cases, the model was run with only f a, recognizing that interpretation of maternal and ancestral effects were confounded. Inbreeding effects O/) f r neonatal survival ranged from to , and were significantly less than zero (indicating statistically significant inbreeding depression) in 7 of the 19 populations (Table 2; Figure 2). Within-species comparisons lacked statistical power since sample sizes for many species were small (Table 1). However, across-species inbreeding effects were less than zero in 17 of the 19 populations, indicating an overall trend consistent with inbreeding depression (P =.0004, sign test; Figure 2). These results are typical of and entirely consistent with those of other multitaxa studies on inbreeding depression (Lacy et al. 1993; Rails et al. 1988). Of more interest here are the effects of ancestral inbreeding on inbreeding depression. The ancestral inbreeding effect was greater than zero (indicating a reduction in inbreeding depression among inbred animals with inbred ancestors) in 15 of the 19 taxa (P =.01, sign test; Figure 2), but was statistically significant within only one species, the Sumatran tiger. If we consider only those taxa that show inbreeding depression O, < 0) (which is more appropriate than examining all taxa), the trend is even more apparent: 15 of the 17 ancestral inbreeding effects are greater than zero (P =.001, sign test). Thus ancestral inbreeding shows a highly significant trend toward reducing inbreeding depression, but the effect within any one species, in general, is weak. Maternal inbreeding had a significantly negative effect on neonatal survival in four populations (kerodon, Sumatran tiger, Dorcas gazelle, and Nilgiri tahr) and a significantly positive effect in European bison. Year-of-birth effects O^ were significantly greater than zero (survival improved over time) in the golden lion tamarin, and significantly less than zero in four species (elephant shrew, maned wolf, Przewalski's horse, and gaur). The majority of the mortality in most of the species analyzed occurred during the neonatal period, providing very little data for the analysis of purging effects on survival from 7 days to age of weaning. In 12 of the 19 species, survival during this period exceeded 90%; a table of these results is therefore not presented. In only 3 of the 19 species were inbreeding effects significant (Eld's deer, Goeldi's marmoset, and kerodon), but 14 of the 19 inbreeding effects were negative (P =.0318, sign test). 172 The Journal of Heredity 1997:88(3)

5 Table 2. Logistic regression coefficients for the neonatal survival regressions Time ft Inbreeding ft Ancestral Inbreeding ft, Maternal Inbreeding ft. Cumulative Inbreeding ft Species S' - S Elephant shrew Golden lion tamartn Goeldl's marmoset Greater galago Kerodon Boris Maned woll Red panda Asiatic lion Sumatran tiger PrzewalsU's horse Pygmy hippopotamus Muntjac Eld's deer Gaur European bison Dorcas gazelle Speke's gazelle Nllgiri tahr ** 0.030*** *** * * *** *** *** *** ** * *** * * * * * * *** *** "* *** *** ** *** * ** ' = P <.05; **=/»<.01; ** =/>< Neonatal Survival Figure 2. Comparison of neonatal survival In noninbred, Inbred (Inbred 1), and Inbred animals with Inbred ancestors (Inbred 2) In 19 mammalian species. Noninbred effects are based on median year of birth for each species with Inbreeding and ancestral Inbreeding effects set to zero. Inbred 1 are Inbreeding effects among individuals with no Inbred ancestors, and are based on median year of birth for each species and I = Inbred 2 are Inbreeding effects among Individuals with Inbred ancestors, and are based on median year of birth for each species, f = 0.25, and f. - mean /. value for each species. In all three cases, maternal Inbreeding effects are ignored. Species names marked with asterisks Indicate statistical significance of Inbreeding effects (*: P<.0S;":P<.01; ***:/><.001). Asterisk over the bars Indicates a statistically significant purging effect. In 11 of the 19 species, the ancestral inbreeding effects were In the predicted direction O*, > 0; P =.3238), and in the 14 species showing inbreeding effects consistent with inbreeding depression O, < 0), ancestral inbreeding effects were positive in 10 (P =.0898, sign test). These results are strongly affected by the paucity of mortality data between 7 days and weaning, so the remaining analyses of survival will focus solely on neonatal survival. For litter sizes, the inbreeding effect was significant only In the golden lion tamarln (Table 3), and in the predicted direction in six of the eight taxa analyzed (P =.1445, sign test). imates of purging effects on litter size were significant in the Sumatran tiger, but there was no overall trend of the sort seen in the neonatal and weaning survival analysis (four of the eight purging effects are in the predicted direction; P =.6367, sign test). Maternal inbreeding had a significant negative impact on litter size only in the Asiatic Hon. Litter size was significantly negatively associated with time in the elephant shrew, boris, and maned wolf, but was positively associated with time in the golden lion tamarin. Change In Inbreeding Depression Dae to Purging For neonatal survival, changes in inbreeding depression (8' - 8) ranged from 0.03 (3% increase In inbreeding depression) to (23% decrease in inbreeding depression) with a median value of (Table 2; Figure 3). Inbreeding depression decreased in 15 of the 19 populations (P =.0096, sign test; Figure 3). Median value of the change in inbreeding for litter size was (Table 3), with only four of eight populations showing a decrease in inbreeding depression (P =.6367, sign test). imates of cumulative inbreeding effects (the combined effects of inbreeding and ancestral inbreeding) for neonatal survival and for litter size are shown in Tables 2 and 3, respectively. For neonatal survival, cumulative inbreeding depression is statistically significant in 9 species and in the direction of inbreeding depression (p c < 0) in 15 of the 19 species (P =.0096, sign test; Table 2). For litter size, the effect is significant for two species and less than zero in seven out of eight (P =.0352, sign test; Table 3). Comparison of Lethal Recessive and Inbreeding Models For many species, LL values were highly correlated with F; on the average the correlation coefficient of LL with F was 0.85, ranging from 0.53 to This resulted in highly similar A/C values for the two different models and low power to distinguish between them. The lethal recessive model (LL) fit the data better than the inbreeding model in only 14 of 25 taxa in the neonatal data (P >.10, sign test), and in 6 of 12 taxa in the litter size data (P =.6128; Table 4). Differences in fit were most apparent in the European bison (LL model fit better for neonatal survival data) and red panda (inbreeding model fit better for litter size). In general, however, the models were very close, and overall there was no trend indicating one model fit better than the other. BaDou Ancestral Intxeerfng in Mammalian Populations 173

6 Table 3. Logistic regression coefbcients for the ancestral inbreeding model applied to litter size data Species Time ft Inbreeding ft Ancestral Inbreeding ft* Maternal Inbreeding ft, Cumulative Inbreeding ft «- 5 Elephant shrew Golden lion tamarln Kerodon Boris Maned wolf Red panda Asiatic lion Sumatran tiger * 0.038*** * *** *** * ** *** * P <.05; ** = P <.01; ***=/><.001. Discussion Overall, purging effects, as measured by the ancestral inbreeding coefficient, were weak within any one species, however the overall trend in the sign of the purging effect on neonatal survival across species was highly significant. Purging appears to have a small but consistent effect on reducing inbreeding depression in neonatal survival (Table 2). Analysis of survival from day 7 to weaning failed to find any similar trend in purging effects across species, even though there was a trend in inbreeding depression. The inbreeding depression during this mortality period was not as strong as it was for neonatal survival. While the analysis was limited due to the small number of deaths that occurred during that age period, the results do suggest that the genetic effects are expressed more strongly earlier in life than later. Were the purging effects sufficiently large to have eliminated inbreeding depression? In no species did purging eliminate a statistically significant inbreeding effect for any of the three fitness components measured. Comparison of inbreeding effects with cumulative inbreeding effects showed that inbreeding depression -O '-6 Figure 3. Distribution across taxa of change In Inbreeding depression due to purging (5' - 6) In neonatal survival. Shaded bars Indicate populations In which Inbreeding depression was reduced. changed from statistical nonsignificance to significance at the P =.01 level in the Przewalskl's horse for neonatal survival (Table 2) and in the red panda for litter size (Table 3). In both cases, change in the significance of inbreeding effects was due to a reduction in the variance of f$ c rather then by a change in sign of the inbreeding effect. Ancestral Inbreeding was sufficient to change the sign of inbreeding effects from negative to positive in two species for neonatal survival (kerodon and Sumatran tiger; Table 2) and one species for litter size (Asiatic lion; Table 3). Despite these changes in sign, ancestral inbreeding does not appear to be a statistically significant factor in eliminating inbreeding depression even though it consistently has a minor effect across a wide variety of taxa (Figure 4). One explanation for the failure of ancestral inbreeding to significantly reduce inbreeding depression is that inbreeding depression Is due primarily to overdomlnance, In which case fitness is not expected to recover over prolonged Inbreeding (Charlesworth and Charlesworth 1987; Ziehe and Robards 1989). This seems unlikely. Studies on other species, summarized by Charlesworth and Charlesworth (1987), show that dominance effects, rather than overdominance, seem to account for a large part of the observed inbreeding depression (Charlesworth et al. 1990). While overdominance at individual loci is unlikely to account for the lack of purging, it is possible that inbreeding depression is maintained by associative overdominance. Associative overdominance is the buildup during Inbreeding of linkages of loci to overdominant loci or to loci containing deleterious recessive alleles in mutation/selection balance (Ohta 1971; Rumball et al. 1994). The overall effect of these linkage groups is the appearance of heterozygote advantage (overdominance), hence the failure of continued Inbreeding to purge the population of its genetic load. Strong associative overdominance can develop in highly inbred lineages (Charlesworth 1991) and have been observed in inbred lines of domestic chickens (Mina et al. 1991) and Drosophila (Rumball et al. 1994). Thus associative overdominance may account for the failure of purging in the more Inbred populations examined here. Whether or not these populations are sufficiently inbred to have developed such associations will have to be examined further. The lack of significant purging effects within species and the overall trend across species is consistent with the hypothesis that inbreeding depression Is not due entirely to lethal alleles, but is more likely due to less deleterious alleles or a combination of detrimental and lethal alleles. The LL model fit the data as equally well as the F model (Table 4), further suggesting that inbreeding depression Is due to a combination of both lethal and detrimental alleles, as is the case with Drosophila (Simmons and Crow 1977). However, the high correlation between Fand LL, and thus the similarity in the models, makes it difficult to distinguish between the relative contribution of lethal versus less deleterious alleles in these species. In only one species, the European bison, was there a suggestion that the LL model fit better than the inbreeding model. This may be a case where lethals were contributing to inbreeding depression, as has been previously suggested (Slatis 1960). In the species analyzed in this study, if inbreeding depression was due entirely to lethal recessive alleles, purging is expected to be rapid (Hedrick 1994) and possibly to have been detected here. Changes in the level of inbreeding depression (8' - 8) varied widely among species (Figure 3) and are not inconsistent with the diversity of inbreeding effects observed in other multi-taxa studies of inbreeding depression (Brewer et al. 1990; Lacy et al. 1993; Rails et al. 1988). 174 The Journal of Heredity 1997:88(3)

7 Table 4. Comparison of estimates and fits (A/C value*) (or LL and F Inbreeding models Species Neonatal survival Utter size F model LL model F model LL model imate offt AIC» imate of ft* A1C imate of A A1C imate of ft. AIC Elephant shrew Golden lion tamarln *** 1, * Golden-headed lion tamarln Black lion tamarin Goeldi's marmoset *** 1, * Brown lemur " * Greater galago Melanotlc galago Orangutan " * Kerodon Boris Punare ManedwoU Red panda Asiatic lion Sumatran tiger Przewalsld's horse *** 1, * Pygmy hippopotamus *** * Muntjac Eld's deer *** * Gaur " * European bison * 1, * Dorcas gazelle * * Speke's gazelle " * Nilglritahr * , , , , " * " J * * * *=/><.05; " = P <.01; *** = P <.001. * AIC = Akalke Information Criterion; + indicates best fitting model (lower AIC value). The variation in results among species could be due to a number of factors. An Important consideration is the degree of inbreeding in the population prior to establishing the captive population (hlstorical inbreeding due to either small population size or an inbreeding mating system). Purging of deleterious alleles may already have occurred in populations derived from previously inbred sources. A Inbreeding Coefficient 0.4 Figure 4. The overall relationship (median across all species) between Inbreeding coefficient and neonatal survival In animals without Inbred ancestory (dotted line) and animals with Inbred ancestors (solid line). Both lines are based on the median estimates (across species) of regression coefficients for year of birth (8, = ) and Inbreeding effect (B, ). The median year of birth (1980) was used as the year-of-blrth parameter. For the line (or animals with Inbred ancestory, the median ancestral Inbreeding coefficient (I. = 0.079) was used with the median estimate of the regression coefficient B<. (6.321). Maternal inbreeding effects are not Included. number of studies have shown that populations derived from inbred sources exhibit less inbreeding depression than populations from outbred sources (hyacinth, Barrett and Charlesworth 1991; house flies, Bryant et al. 1990; Mimulus, Dole and Rltland 1993; Clarkia tembloriensis, Holtsford and Ellstrand 1990; mice, Lorenc 1980; Japanese quail, MacNeil et al. 1984; Peromyscus, Ribble and Miller 1992). Among the taxa analyzed here, the Asiatic lion is one species with a known history of small population size and inbreeding (Wildt et al. 1987). Asiatic lions also exhibit extremely low levels of genetic diversity and high percentages of abnormal sperm (Wlldt et al. 1987). The results found here are consistent with a history of small population size: inbreeding depression was nonsignificant for all three fitness measures, and ancestral inbreeding failed to decrease inbreeding depression in either neonatal survival (Table 2) or weaning survival. For litter size, however, ancestral inbreeding did decrease inbreeding effects (Table 3). On the other hand, numerous other studies show that a prior history of inbreeding (due to historically small population size or mating system) is often not successful in completely eliminating inbreeding depression, although as mentioned above, inbreeding depression Ballou Ancestral Inbreeding in Mammalian Populations 175

8 might be reduced (Agren and Schemske 1993; Brewer et al. 1990; Charlesworth et al. 1990; Charlesworth and Charlesworth 1987; Dole and Ritland 1993; Frankham et al. 1993; Lorenc 1980; Wright et al., in press). Among the more notable cases are the Pere David's deer (Elaphurus davidianus) which is known to have gone through a severe bottleneck; the cheetah (Acinonyx jubatus), which genetic data suggest may have gone through a bottleneck (O'Brien et al. 1985; but see Hedrick 1996); and analyzed here, the European bison, which descended to only a few individuals. These species still show statistically significant levels of inbreeding depression (Foose and Foose 1983; Hedrick 1987). Other factors affecting the ability to purge are the severity of selection (e.g., lethal versus mildly deleterious recessive alleles), the duration of selection, the number of lethal equivalents, the level of Inbreeding, and the rate at which inbreeding accumulates in the population (Hedrick 1994). Hedrick (1994), using stochastic simulation models, showed that purging was most successful when levels of selection were high, as might be the case when inbreeding depression is due to lethal recessives. Ehoibu et al. (1989) found that in equally inbred lines of Drosophila melanogaster, Inbreeding depression was lower in lines with slow rates of inbreeding than in lines with more rapid rates of inbreeding, presumably because of the greater opportunity (i.e., more generations) for selection to act. However, maize inbred to f = 0.5 using full-sib and half-sib matings did not show any difference in inbreeding depression (Falconer and Mackay 1995). Likewise there was no significant difference in extinction rates in lines of Drosophila inbred to the same level using fullsib and double first-cousin matings (Frankham 1995). The variation in results here might then be due to differences in the degree of inbreeding or other factors within the population. Level of inbreeding (as measured by average /), degree of inbreeding depression (B^), tne opportunity to detect purging (as measured by average f a of inbred animals), and selection (as measured by the ratio of inbred to noninbred survival rates) varied greatly among species. To determine which factors might most affect purging, I used a stepwise multiple regression (PROC REG with STEPWI option; SAS 1991). The following variables for each taxa were included in the multiple regression as independent variables, with Inbreeding Effect (Pf) Figure 5. Plot of estimated Inbreeding effect O,) against change In Inbreeding depression due to ancestral Inbreeding (5' - 8). Regression line is (6' - 8) = $ r 1 =.334; P =.03. the standardized ancestral inbreeding effect (BJ used as the dependent variable: total sample size, average f a for individuals with f > 0, average f, overall mortality rate (as an index on absolute selection), ratio of inbred to non-inbred survival rates (an index of selection operating on inbred animals relative to noninbred), and the standardized estimate of the inbreeding depression effect (B,). To control for the range of B, and B /# estimates across species, B, and B, # were standardized by multiplying them by the standard deviations of the f and f a values, respectively, within each species. This resulted in standardized B, and B, a values equivalent to those that would have been obtained if the logistic regression had been conducted on standardized f and f a values in the first place. This analysis was not conducted on litter size data because of the limited number of species producing litters. The only significant predictor of the purging effect in neonatal survival was the standardized B, (P =.030; Figure 5) the degree of inbreeding depression. The larger the inbreeding effect, the larger the purging effect. Since the inbreeding effect Is a function of the number of lethal equivalents in a population (Morton et al. 1956), these results suggest that purging is most effective in populations with the highest number of lethal equivalents, as was previously suggested using simulation modeling (Hedrick 1994). Surprisingly, the level of inbreeding and the degree of ancestral inbreeding in the taxa were not correlated with the observed purging effect. Speke's gazelle is often used to illustrate the effectiveness of reducing inbreeding depression by purging populations of their lethal or deleterious genes (Templeton and Read 1983, 1984). More recently however, Willis and Wiese (1997), in a reanalysis of Templeton and Read's data, found that the reported reduction in inbreeding depression may have been due to the sample size correction factor applied to the data rather than to purging per se. In the reanalysis of the Speke's gazelle data here, purging effects were shown to have only minimally and nonsignificantly reduced inbreeding depression (see cumulative inbreeding effects, Table 2). In fact, inbreeding depression in this species was the highest of any of the species analyzed here. Nevertheless, regardless of the success of purging in the Speke's gazelle program, the general issue of using purging to eliminate inbreeding depression is still one that needs to be addressed because It has frequently been recommended as a potential strategy for captive breeding programs under certain circumstances (Rails and Ballou 1986; Simberloff 1988; Templeton et al. 1986). The results presented here suggest that although the purging that occurs naturally in small inbreeding populations may have a slight impact on reducing inbreeding depression (Figure 4), it is not sufficient to eliminate inbreeding depression. Eliminat- 176 The Journal of Heredity (3)

9 Ing inbreeding depression is likely to require a rapid rate of inbreeding and high levels of selection (Hedrick 1994). This will almost certainly incur some risk to the population during the purging period and will likely result in long-term detrimental effects. If much of the inbreeding depression is due to deleterious recessive alleles, the chance of fixing deleterious alleles during the inbreeding process is high and the probability of population extinction increased (Barrett and Charlesworth 1987; Hedrick 1994; Mills and Smouse 1994). Furthermore, a program of intensive purging genetically alters the population, adapting it more rapidly to both its physical captive environment and its inbred genetic environment (Templeton and Read 1983), neither of which is desirable for species of conservation interest (Arnold 1995). Because of these concerns, strategies to purge inbreeding depression in species being bred for conservation purposes are ill advised. Captive breeding programs should adopt strategies that maintain genetic variation in large populations, while avoiding inbreeding to the extent possible, because these strategies not only minimize selection to the captive environment, but also mitigate existing inbreeding depression (Lacy et al. 1995). Reference* Agren J and Schemske DW, Outcrossing rate and Inbreeding depression In two annual monoecious herbs, Begonia hirsuta and B semiouata Evolution 47: Arnold S, Monitoring quantitative genetic variation and evolution In captive populations. In: Population management for survival and recovery (Ballou JD, Gllpln M, and Foose TJ, eds). New York: Columbia University Press. Ballou JD, Inbreeding and outbreedlng depression In the captive propagation of black-looted ferrets. In: Conservation biology and the black-footed ferret (Seal US, Thome ET, Bogan MA, and Anderson SH, eds). New Haven, Connecticut: Yale University Press. Barrett SCH and Charlesworth B, Effects of a change In the level of inbreeding on the genetic load. Nature Brewer BA, Lacy RC, Foster ML, and Alaks G, Inbreeding depression In Insular and central populations of Pemmyscus mice. J Heredity 81: Bryant EH, Meffert LM, and McCommas SA, Fitness rebound in serially bottlenecked populations of the housefly.am Nat Charlesworth B, Charlesworth D, and Morgan MT, Genetic loads and estimates of mutation rates In highly Inbred plant populations. Nature 347: Charlesworth D, The apparent selection on neutral marker loci in partially Inbreeding populations. Genet Re* 57: Charlesworth D and Charlesworth B, Inbreeding depression and Its evolutionary consequences. Annu Rev Ecol Syst 18: Crow JF and Kimura M, An Introduction to population genetic theory. New York: Harper and Row. de Bols, H International golden-headed lion tamarin studbook. Antwerp, Netherlands: Antwerp Zoological Gardens. Dole J and Rltland K, Inbreeding depression In two Mimulus taza measured by multigeneratlonal changes in the Inbreeding coefficient. Evolution 47: Ehiobu NG, Goddard ME, and Taylor JF, Effect of rate of Inbreeding on Inbreeding depression In Drosophila melanogaster. Theor Appl Genet 77: Falconer DS and Mackay TFC, Introduction to quantitative genetics. Harlow, Essex, England: Longman Group. Foose TJ and Foose E, Demographic and genetic status and management. In: The biology and management of an extinct species: Pere David's deer (Beck BB and Wemmer C, eds). Park Ridge, New Jersey: Noyes Publishers. Foose TJ, Lande R, Flesness NR, Rabb G, and Read B, Propagation plans. Zoo biology 5: Fouraker M, McMillan G, and Drummer L, Panthera leo persica Asiatic lion International studbook. Fort Worth, Texas: Fort Worth Zoological Association. Frankham R, Inbreeding and extinction: a threshold effect. Conserv Blol 9: Frankham R, Smith GJ, and Briscoe DA, Effects on heterozygoslty and reproductive fitness of inbreeding with and without selection on fitness in Dmsophila melanogaster. Theor Appl Genet 86: Glatston AR, International red panda studbook. Rotterdam, Netherlands: Rotterdam Zoological Gardens. Hedrick PW, 1987 Genetic bottlenecks. Science 23:963. Hedrick PW, Purging Inbreeding depression and the probability of extinction: full-sib mating. Heredity 73: Hedrick PW, Bottleneck or metapopulation In cheetah Conserv Blol 10: Holtsford TP and Ellstrand NC, 1990 Inbreeding effects In Ctarkia tembtoriensis (Onagraceae) populations with different natural outcrosslng rates. Evolution 44: Hosmer DW and Lemeshow S, Applied logistic regression. New York: John Wiley & Sons. Kids H, International studbook of the gaur. Berlin, Germany: Zoologlscher Garten Berlin. Lacy R, Ballou JD, Starfield A, Thompson EA, and Thomas A, Pedigree analyses. In: Population management for survival and recovery (Ballou JD, Gllpln M, and Foose TJ, eds) New York. Columbia University Press. Lacy RC, Petrlc A, and Wameke M, Inbreeding and outbreedlng In captive populations of wild animal species. In: The natural history of Inbreeding and outbreedlng (Thomhlll NW, ed). Chicago, Illinois: University of Chicago Press. Lande R and Barrowclough G, Effective population size, genetic variation, and their use In population management. In: Viable populations for conservation (Soule ME, ed). Cambridge: Cambridge University Press. Lande R and Schemske DW, The evolution of self fertilization and inbreeding depression in plants. I. Genetic models. Evolution 39: Latta R and Rltland K, The relationship between Inbreeding depression and prior Inbreeding among populations of four Mimulus taxa. Evolution 48* Lorenc E, Analysis of lertlllty in inbred lines of mice derived from populations differing In their genetic load. Zeierzeta Lab 17:3-16. MacCluer JW, VandeBerg JL, Read B, and Ryder OA, Pedigree analysis by computer simulation. Zoo Blol 5: MacNell MD, Kress DD, Flower AE, and Blackwell RL, Effects of mating system In Japanese quail. 2. Genetic parameters, response and correlated response to selection. Theor Appl Genet 67: Mills LS and Smouse PE, Demographic consequences of Inbreeding in remnant populations. Am Nat 144:412^31. Mina NS, Sheldon BL, Yoo BH, and Frankham R, Heterozygoslty at protein loci In Inbred and outbred lines of chicken. Poult Sci 70: Morton NE, Crow JF, and Muller HJ, An estimate of the mutatlonal damage In man from data on consanguineous marriages. Proc Natl Acad Scl USA 42: O'Brien SJ, Roelke ME, Marker L, Newman A, Wlnkler CA, Meltzer D, Colly L, Evermann JF, Bush M, and Wlldt DE, Genetic basis for species vulnerability in the cheetah. Science Ohta T, Associative overdomlnance caused by linked detrimental mutations. Genet Res 18: Padua C, International black lion tamarin studbook. Sao Paulo, Brazil: Sao Paulo Zoological Gardens. Perkins L, International studbook for the orangutan. Atlanta, Georgia; Zoo Atlanta. Pilarski HWK, European bison pedigree book. Warsaw, Poland: State Publications for Agriculture and Forestry. Rails K and Ballou JD, Captive breeding programs for populations with a small number of founders. Trends Ecol Evol Rails K, Ballou JD, and Templeton AR, imates of lethal equivalents and the cost of Inbreeding in mammals. Conserv Blol 2: Rails K, Brugger K, and Gllck A, 1980 Deleterious effects of Inbreeding in a herd of captive Drocas gazelle. Int Zoo Yearbk 20: Read B, Spekes gazelle studbook. St. Louis, Missouri. St. Louis Zoological Gardens. Ribble DO and Miller JS, Inbreeding effects among Inbred and outbred laboratory colonies of Peromysaa maniculalus. Can J Zool 70: Rumball W, Franklin IR, Frankham R, and Sheldon BL, Decline In heterozygosity under full-sib and double first-cousin Inbreeding In Dmsophila melanogaster. Genetics 136: SAS, SAS/STAT user's guide, version Cary, North Carolina: SAS Institute. Simberloff D, The contribution of population and community biology to conservation science. Annu Rev Ecol Syst 19: Simmons MJ and Crow JF, Mutations affecting fitness In Dmsophila populations Annu Rev Genet 11: Slatis HM, An analysis of Inbreeding In the European bison. Genetics 45: Swengel, FB Nilgiri tahr regional studbook Minneapolis, Minnesota: Minnesota Zoological Gardens. Templeton AR, Hemmer H, Mace G, Seal US, Shields WM, and Woodruff DS, Local adaptation, coadaptation and population boundaries. Zoo Blol 5: Templeton AR and Read B, The eliminations of inbreeding depression in a captive herd to Speke's gazelle. In; Genetics and conservation (Schonewald-Cox CM, Chambers SM, MacBryde B, and Thomas L, eds). Menlo Park, California; Ben amin/cummlngs. Templeton AR and Read B, Factors eliminating inbreeding depression In a captive herd of Speke's gazelle. Zoo Blol 3: Thomhlll NW, The natural history of Inbreeding and outbreedlng. Chicago, Illinois: University of Chicago Press. Tobler K, International studbook for the pygmy Ballou Ancestral Inbreeding in Mammalian Populations 177