Genetic variation and differentiation in captive and wild zebra finches ...
Molecular Ecology (2007) 16, 4039–4050
doi: 10.1111/j.1365-294X.2007.03444.x
Genetic variation and differentiation in captive and wild zebra finches (Taeniopygia guttata) Blackwell Publishing Ltd
W O L F G A N G F O R S T M E I E R ,* G E R N O T S E G E L B A C H E R ,† J A K O B C . M U E L L E R * and BART KEMPENAERS* *Max Planck Institute for Ornithology, Postfach 1564, D-82305 Starnberg (Seewiesen), Germany, †Max Planck Institute for Ornithology, Vogelwarte Radolfzell, Schlossallee 2, D-78315 Radolfzell, Germany
Abstract The zebra finch (Taeniopygia guttata) is a small Australian grassland songbird that has been domesticated over the past two centuries. Because it is easy to breed in captivity, it has become a widely used study organism, especially in behavioural research. Most work has been conducted on domesticated populations maintained at numerous laboratories in Europe and North America. However, little is known about the extent to which, during the process of domestication, captive populations have gone through bottlenecks in population size, leading to inbred and potentially genetically differentiated study populations. This is an important issue, because (i) behavioural studies on captive populations might suffer from artefacts arising from high levels of inbreeding or lack of genetic variation in such populations, and (ii) it may hamper the comparability of research findings. To address this issue, we genotyped 1000 zebra finches from 18 captive and two wild populations at 10 highly variable microsatellite loci. We found that all captive populations have lost some of the genetic variability present in the wild, but there is no evidence that they have gone through a severe bottleneck, as the average captive population still showed a mean of 11.7 alleles per locus, compared to a mean of 19.3 alleles/locus for wild zebra finches. We found significant differentiation between the captive populations (FST = 0.062). Patterns of genetic similarity closely match geographical relationships, so the most pronounced differences occur between the three continents: Australia, North America, and Europe. By providing a tree of the genetic similarity of the different captive populations, we hope to contribute to a better understanding of variation in research findings obtained by different laboratories. Keywords: bottleneck, domestication, founder effect, genetic diversity, genetic drift, inbreeding. Received 1 April 2007; revision accepted 4 June 2007
Introduction The zebra finch, Taeniopygia guttata , is one of the most frequently studied bird species. It is a widely used model organism in studies of behaviour, evolution, ecology, neurobiology, endocrinology and many other fields (Zann 1996). This is signified by the fact that the zebra finch is the second bird species after the chicken for which the entire genome sequence is now becoming available. While there is some information on the behaviour and ecology of zebra finches in the wild (Zann 1996), most research is carried out using domesticated birds of the Australian subspecies Correspondence: W. Forstmeier, Fax: +49 8157 932400; E-mail: [email protected] © 2007 The Authors Journal compilation © 2007 Blackwell Publishing Ltd
Taeniopygia guttata castanotis. Many research laboratories maintain their own breeding stock, with birds typically originating from larger stocks held by aviculturists. The practice of obtaining birds for research from such variable sources brings about uncertainties regarding the genetic background of the study organisms. This is problematic because, during the process of domestication, zebra finches might have gone through repeated bottlenecks in population size, potentially leading to genetic differentiation between the various captive populations by means of genetic drift or differential selection pressures. Moreover, domesticated zebra finches are known to be larger than their wild ancestors (Zann 1996), a fact that has been ascribed to selective breeding for larger body size by aviculturists. These issues raise questions about the extent to which: (i) studies of
4040 W . F O R S T M E I E R E T A L . domesticated birds are distinct from the situation in the wild, and (ii) studies can be compared between the various captive stocks. There are several examples in the zebra finch literature of where findings from one research laboratory could not be replicated by another. In different studies, females showed different preferences with regard to male beak colour (see Collins & ten Cate 1996), but also with regard to colour and symmetry of artificial leg bands (see Jennions 1998). While mate preferences likely are affected by many other factors besides genetics, there are also examples of highly heritable traits that show different patterns in different laboratories. For instance, digit ratio, a morphological trait that is thought to partly reflect sex steroid levels during ontogeny, was sexually dimorphic and depended on the laying sequence within a clutch in a study population from Irvine, California (Burley & Foster 2004), but not in a population from Sheffield, UK, where the trait showed a heritability of 0.72 (Forstmeier 2005). The difference between the results regarding sexual dimorphism and laying-order effects was significant, and thus not due to the lack of statistical power. Unfortunately, there have been no rigorous studies exploring how much of the population differences in findings are due to genetic differences vs. differences in experimental techniques, bird husbandry or other sources of environmental variation. Whether or not such differences in findings can be attributed to genetic differentiation between the various populations depends on whether any genetic differentiation exists in the first place. According to Sossinka (1970), it seems likely that most of the European zebra finches go back to individuals that were brought from Australia approximately between 1870 and 1890, and that relatively few new imports reached Europe after the First World War. However, it is also known that some zebra finches that had been domesticated in Australia were exported to Europe in the 1950s (Sossinka 1970). Apart from such sporadic information, we know too little about the history of zebra finch domestication, about bottlenecks, about effective population sizes, and about gene flow between the captive populations to answer the question of genetic differentiation satisfactorily. We therefore decided to launch an initiative (starting at an international zebra finch workshop held at Bielefeld, Germany in March 2005; see Rutkowska 2005) to study the population genetics of wild and captive zebra finches. The present study is the outcome of this initiative. We compiled a list of laboratories maintaining a captive stock of zebra finches for behavioural research, and asked the owners to contribute DNA samples. All samples were genotyped at 10 highly polymorphic microsatellite loci (Forstmeier et al. 2007), which had not been preselected for their level of polymorphism in any particular population. We also obtained samples from two populations of wild zebra finches to serve as a comparison. We set out to address the following two main questions:
1 How severely inbred or bottlenecked are the various zebra finch laboratory populations as compared to the wild? 2 How strong is the genetic differentiation between the various captive populations? The first question is meant to evaluate the potential concern that behavioural studies on captive populations might suffer from artefacts arising from high levels of inbreeding or lack of genetic variation in such populations (see Fowler & Whitlock 1999; Miller et al. 1999; Keller & Waller 2002). The second question addresses whether genetic differentiation is large enough to potentially be responsible for some of the differences in findings between laboratories. Our study also includes data on body size to examine whether body size can yield similar or additional insights into the history of zebra finch domestication and the assumed selective breeding by aviculturists.
Materials and methods Study populations Table 1 shows a list of populations from which we obtained a sufficient number of samples (range 27–68 individuals per population). We deliberately excluded some populations that were known to be of the same origin and were expected to be very similar to the ones included. The list of populations includes two wild populations from Australia, as well as 18 captive populations from three continents: Australia (two populations), North America (six populations), and Europe (10 populations). The two wild populations were no. 1 from Alice Springs in Central Australia (23°44′S, 133°50′E), and no. 2 from northern Victoria (36°10′S, 145°20′E). The captive populations no. 3 and no. 4 are descendants of birds that were wild-caught at the same site as no. 2. Population no. 3 comprises first to fourth generation aviary-bred birds that are still maintained at Victoria, while population no. 4 contains fifth to sixth generation descendents of 12 pairs of first-generation birds from population no. 2 that had been exported to Bielefeld, Germany in 1992. Since then, this population has not been mixed with birds from other sources. Despite the current location in Germany, we still label population no. 4 as Australian. The North American populations are no. 5 from Vancouver, Canada, all bought from local pet shops and private breeders; no. 6 from Tempe, Arizona, with most of the birds coming from a breeder in San Jose, California, and a few from pet stores in Tempe; no. 7 from Irvine, California, all from local breeders (e.g. Magnolia Bird Farm); no. 8 from Davis, California, all descendents of approximately 80 individuals obtained in 1995 from a commercial breeder in California’s central valley; no. 9 from Cornell, New York, originating from various local sources; and no. 10 from Millbrook, New York. The 10 European populations © 2007 The Authors Journal compilation © 2007 Blackwell Publishing Ltd
© 2007 The Authors Journal compilation © 2007 Blackwell Publishing Ltd
Table 1 The 20 study populations named after their current location and origin. Origin indicates where most of the birds of a population came from immediately before the foundation of the population. The names of population owners and suppliers of DNA samples are listed, as well as the number of individuals sampled (N samples), the mean number of alleles per locus (N alleles), the minimum (Min) and maximum (Max) number of alleles per locus, the mean number of private (unique to a population) alleles per locus in each population (Private), allelic richness (AR), expected heterozygosity (HE), observed heterozygosity (HO), inbreeding coefficient (FIS). All data are based on 10 loci, except HE, HO, and FIS which are based on eight loci Population name
Current location
Origin
Continent
Supplier
N samples
N alleles
Min
Max
Private
AR
HE
HO
FIS
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Alice Springs (wild) Victoria (wild) Victoria Bielefeld-AUS Vancouver Tempe Irvine Davis Cornell Millbrook Cracow Paris Lund Groningen Seewiesen-NL Leiden Stirling Seewiesen-GB Bielefeld Bielefeld (white)
Australia Australia Australia Germany Canada Arizona California California New York New York Poland France Sweden The Netherlands Germany The Netherlands Great Britain Germany Germany Germany
Australia Australia Australia Australia Canada California California California New York New York Poland Czech Republ. Denmark The Netherlands The Netherlands The Netherlands Great Britain Great Britain Germany Germany
AUS AUS AUS AUS NAM NAM NAM NAM NAM NAM EUR EUR EUR EUR EUR EUR EUR EUR EUR EUR
Runciman Zann Zann Naguib & Witte Williams McGraw Burley Millam Adkins-Regan Nottebohm Rutkowska Alonso-Alvarez Sandell & Tobler von Engelhardt Gahr Riebel, Holveck & ten Cate Evans Forstmeier Bischof Bischof
48 27 50 50 50 41 68 52 50 55 48 52 52 59 53 50 52 61 50 32
24.7 18.5 18.6 10.4 16.8 16.1 16.1 11.4 16.1 13.8 13.2 16.9 16.0 13.9 13.1 12.4 12.7 9.5 8.5 6.4
20 15 15 9 15 13 12 9 12 9 9 11 11 9 10 9 9 7 7 4
30 23 25 13 23 22 19 15 21 19 17 24 22 19 18 21 19 12 12 9
1.8 0.4 1.0 0.1 0.3 0.0 0.0 0.2 0.2 0.1 0.1 0.2 0.0 0.1 0.1 0.1 0.0 0.2 0.2 0.4
20.21 18.31 15.46 9.59 14.30 14.22 13.51 10.39 14.18 11.99 11.71 14.09 13.59 11.91 11.67 11.10 11.03 8.52 7.85 6.31
0.94 0.93 0.91 0.84 0.89 0.90 0.89 0.85 0.90 0.88 0.88 0.88 0.88 0.87 0.86 0.86 0.86 0.82 0.79 0.79
0.93 0.93 0.91 0.83 0.86 0.83 0.88 0.86 0.85 0.85 0.87 0.86 0.84 0.87 0.83 0.85 0.81 0.78 0.79 0.80
0.009 0.004 – 0.003 0.016 0.031 0.076 0.018 – 0.004 0.051 0.039 0.010 0.034 0.049 – 0.015 0.044 0.013 0.051 0.049 0.005 – 0.015
AUS, Australia; NAM, North America; EUR, Europe.
Z E B R A F I N C H P O P U L A T I O N G E N E T I C S 4041
ID
4042 W . F O R S T M E I E R E T A L . are no. 11 from Cracow, Poland, bought from various local breeders since 2000; no. 12 from Paris, France, most of which came from a breeder at Blois, SW France (who reported to get most of his birds from the Czech Republic), and the rest from Dijon, SE France; no. 13 from Lund, Sweden, going back to 160 birds obtained in 2001/2002 from various small breeders in Denmark; no. 14 from Groningen, the Netherlands, originating from about 100 individuals obtained from at least five different local breeders between 1998 and 2000; no. 15 from Seewiesen, Germany, all of which had been brought there from Amsterdam, the Netherlands; no. 16 from Leiden, the Netherlands; no. 17 from Stirling, UK, which have exchanged birds with Universities at Glasgow, St Andrews, Edinburgh, and Exeter; no. 18 from our own population at Seewiesen, Germany, going back to 231 individuals obtained in 2002 from a large colony held by T. R. Birkhead at Sheffield University, UK since 1985; no. 19 and no. 20 both from Bielefeld, Germany, with no. 19 partly going back to birds purchased by Sossinka throughout Germany and Switzerland in the 1960s (Sossinka 1970). All populations consist mostly or exclusively of birds carrying wild-type plumage, with the exception of population no. 20, which we particularly selected because it represents a pure strain of an autosomal recessive leucistic mutant known as ‘white’. We expected to find the lowest genetic diversity in this population, because colour mutant strains are usually created by back-crossing a novel mutant bird to its parents. However, that population was also smaller than all the others, consisting of only 32 individuals, all of which were included in this study (Table 1). None of these 18 captive populations seems to have a special history of either purposeful inbreeding to reduce genetic variability or of managed outbreeding to maximize the retention of genetic variability. Providers of samples were asked to randomly (irrespective of sex and relatedness) select 50 adult (> 100d of age) individuals from their entire population. Note that, when captive populations are sampled in this way, a variable number of closely related individuals will be included in each sample, whereas this is much less likely for our population samples from the wild (nos 1 and 2). Final sample sizes varied slightly between the populations (Table 1), mostly due to limitation of the number of birds available (population nos 1, 2, 20), and the availability of extra samples (particularly nos 7, 18) which allowed to keep the total sample size at 1000 individuals.
Genotyping As reported elsewhere (Forstmeier et al. 2007), we developed 12 microsatellite markers for the zebra finch using the DNA of a single female from our own population (no. 18). An extensive test for Mendelian inheritance showed that
two out of the 12 markers (Tgu11 and Tgu13) could not be genotyped reliably in our population (Forstmeier et al. 2007). Hence, in the present study we use the remaining 10 markers. All 1000 individuals were genotyped at each of the 10 loci closely following the methods described in Forstmeier et al. (2007). As described earlier (Forstmeier et al. 2007), two of the 10 loci (Tgu6 and Tgu7) showed evidence for the occurrence of null alleles, and this was again the case in the present data set: estimated null allele frequencies using cervus 3.0 (Kalinowski et al. 2006) were 0.26 and 0.12 for Tgu6 and Tgu7, respectively. Hence, we report deviations from Hardy–Weinberg equilibrium (HE, HO, and FIS; see below) based on the remaining eight loci. All other analyses presented are based on 10 loci, since null alleles do not seem to introduce a major bias (Dakin & Avise 2004). We repeated all analyses based on only eight loci, but all findings were qualitatively the same (data not shown).
Genetic analyses Allele frequencies and estimates of within-population diversity (observed number of alleles, heterozygosity) and between-population divergence (Weir & Cockerham’s (1984) analogue of Wright’s FST) were calculated using fstat 2.9.3 (Goudet 1995) and genetix (Belkir et al. 1996– 2004). Deviation from Hardy–Weinberg expectations for each locus in each population was estimated using the same program. Bayesian clustering software structure 2.1 (Pritchard et al. 2000) was used for assigning individuals to different clusters. We chose k = 3 clusters to represent the three continents, Australia, North America, and Europe. We used the ‘no admixture’ algorithm without any prior population information and used 50 000 runs as burn-in (i.e. simulation data that will be discarded) and 100 000 runs for each chain. Individual genotypes were ordinated in a multidimensional space by a factorial component analysis (FCA) performed with genetix. An unrooted tree showing population relationships was constructed using the Fitch–Margoliash method (Fitch & Margoliash 1967) in phylip (Felsenstein 1993) based on transformed pairwise FST values according to Reynolds et al. (1983).
Morphological data We asked the owners of the 20 zebra finch populations to contribute the following morphological data from any adult individuals of their population: flattened wing length, tarsus length, and body mass (see Appendix I). Tarsus length was measured by three different techniques, either with callipers from the bent foot to the end of the tibiotarsus bone (method 1), or with callipers from the bent foot to the rear edge of the tarsometatarsus, thus including the entire joint (method 2), or like method 2 but using a © 2007 The Authors Journal compilation © 2007 Blackwell Publishing Ltd
Z E B R A F I N C H P O P U L A T I O N G E N E T I C S 4043 Fig. 1 Average allelic richness at 10 microsatellite markers for the 20 study populations in relation to their FST values to the wild population at Alice Springs. Different symbol types refer to the three continents (crosses, Australia; open squares, North America; filled diamonds, Europe).
wing ruler (method 3). To assess the comparability of these methods, W.F. measured 10 birds of our population twice, using each of the three methods. From those data, we estimated that the larger measurements obtained by methods 2 and 3 can be adjusted to method 1 measurements by multiplication with 0.861 (± 0.0035 SE) and 0.849 (± 0.0024 SE), respectively. We used these correction factors, to compare the tarsus measurements from the various populations (Appendix I). Since most measurements were taken by different observers (see Appendix I), the results based on these data should be regarded with caution.
Results Loss of genetic variability in captive populations Ten out of the 20 populations (all captive) showed a significant F IS value (Table 1), indicating that they had been founded from two or more differentiated sources that have not yet been mixed completely, or that they were not bred in a panmictic manner. Allelic richness, i.e. the number of alleles per locus adjusted for the number of individuals sampled, varied substantially between the populations (Table 1). As expected, the two wild populations showed higher allelic richness (mean = 19.3) than the 18 captive populations (mean = 11.7; Mann–Whitney U-test: P = 0.011). Also as expected, the lowest number of alleles was found in the putatively inbred white mutant population no. 20 from Bielefeld (Table 1). The same results are found when focusing on expected heterozygosity as an indicator of genetic diversity (Table 1). We calculated pairwise FST values between all possible pairs of populations (see Appendix II). The overall genetic differentiation among all populations was highly significant © 2007 The Authors Journal compilation © 2007 Blackwell Publishing Ltd
(FST = 0.062). The degree to which captive populations have lost genetic variability (measured as allelic richness) was very closely reflected by their FST distance towards the most diverse wild population no. 1 from Alice Springs in Central Australia (Fig. 1). To better understand what the observed reduction in allelic richness in captive populations means in terms of the probability of losing a genetic polymorphism (e.g. a single nucleotide polymorphism), we did the following procedure. For every allele found in the Australian population nos 1–3 (N = 289 alleles across the 10 loci; N = 125 individuals), we calculated its allele frequency in Australia. The pooling of three populations was necessary to achieve a large-enough sample size, and is justified based on low pairwise FST values (Appendix II). We then plotted the percentage of the 17 captive populations currently held outside Australia that were not showing, and hence had ‘lost’, that allele (when sampling on average 51.5 individuals per population) against the initial allele frequency of the Australian populations. Figure 2 shows that these 17 captive populations primarily lost the alleles that are very rare in Australia, which has important implications for the maintenance of additive genetic variance (see Discussion). We fitted a hyperbolic regression line of the form f (x) = 1 – x/(ax – a + 1), where x is the initial allele frequency and a is a constant that determines the curvature. This function fulfils the necessary marginal criteria that the probability of loss is 1 if the initial allele frequency is 0, and that the probability of loss is 0 if the initial allele frequency is 1. The best fit (least squares method) was reached for a = 0.9696. Hence, while rare alleles (initial frequency = 1%) are absent in most populations (predicted probability of loss = 75%), more common alleles (frequencies, e.g. 10%, 20%, and 50%) are lost only rarely (risk of loss: 22%, 11%, and 3%, respectively).
4044 W . F O R S T M E I E R E T A L .
Fig. 3 Factorial component analysis of the genotypes of 1000 zebra finches from three different continents using genetix. Fig. 2 The proportion of 17 captive populations (nos 4–20) that have lost a particular allele in relation to the frequency of that allele in the three Australian populations nos 1–3. Dots represent the 289 alleles from the 10 different loci. Increasing symbol size refers to multiple data points (up to N = 22). A least squares regression line of the form y = 1 – x/(ax – a + 1) is shown to fit the observed data. For comparison, crosses illustrate the results of a Monte Carlo simulation of genetic drift. Each cross shows the proportion of 1000 populations that have lost an allele of a given starting frequency after 50 generations when the effective population size equals 200 individuals.
To see how the fitted function behaves compared to neutral expectations of loss in an ideal population of constant but limited size, we simulated genetic drift by a stochastic sampling process of gametes from the parental population in each generation using the software winpop 2.5 (Nuin 2005). The following settings turned out to best match the observed data: effective population size Ne = 200, number of generations t = 50. With these settings, an allele with a starting frequency of 0.0346 (i.e. the average observed allele frequency in Australia) has a probability of 57.51% to be lost (based on 10 000 Monte Carlo simulations), which closely fits the observed mean of 57.46%. A similarly good fit is obtained for other combinations where Ne/t = 4 (e.g. Ne = 100, t = 25, or Ne = 400, t = 100). Carrying out the simulation over the whole range of allele frequencies (Fig. 2), the simulated data fit the observed data fairly well (but less well than the fitted line).
Differentiation between populations All populations were significantly differentiated from each other (i.e. significant FST values; see Appendix II) with the following exceptions: (i) the two wild populations no. 1 and no. 2 were not significantly differentiated despite the fact that sampling locations were approximately 2000 km apart; (ii) the captive population no. 3 from Victoria was not different from the wild Victoria population no. 2 from
Table 2 Cluster analysis showing the percentage of individuals from each population assigned to one of three clusters. Bold print indicates which populations are from the same continent
ID
Population
Cluster 1 Australia
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Alice Springs (wild) Victoria (wild) Victoria Bielefeld-AUS Vancouver Tempe Irvine Davis Cornell Millbrook Cracow Paris Lund Groningen Seewiesen-NL Leiden Stirling Seewiesen-GB Bielefeld Bielefeld (white)
85% 82% 95% 95% 13% 12% 4% 3% 4% 5% 8% 17% 29% 3% 4% 4% 14% 2% 4% 2%
Cluster 2 North America
Cluster 3 Europe
10% 10% 2% 1% 43% 82% 89% 96% 93% 93% 8% 7% 4% 4% 3% 2% 4% 2% 0% 0%
5% 8% 3% 3% 44% 6% 7% 2% 3% 2% 83% 76% 67% 93% 93% 94% 82% 96% 95% 97%
which it had been obtained. Between the 18 captive populations, pairwise FST values varied between 0.014 and 0.165. We performed a factorial component analysis (FCA) of the genotypes of the 1000 individuals using genetix. Figure 3 shows the individual scores on the first two components. Populations from the three continents can be clearly distinguished. Table 2 shows the results from a cluster analysis using structure 2.1 where individual birds were allocated to one of three clusters. The three clusters that were identified by the program closely matched the three continents, © 2007 The Authors Journal compilation © 2007 Blackwell Publishing Ltd
Z E B R A F I N C H P O P U L A T I O N G E N E T I C S 4045 population genetics parameters such as genetic drift. This tree not only reflects the basic differentiation according to continents, but also illustrates the relatively high similarity of the two British, and the three Dutch populations, respectively. Absolute branch lengths closely reflect FST values, and hence in our case, the degree of loss of allelic richness (Fig. 1).
Body size as an indicator of domestication
Fig. 4 Unrooted dendrogram showing population relationships (Fitch-Margoliash tree based on Reynolds’ distances). Dotted lines show the separation of the three continents.
Australia, North America, and Europe. Most individuals of most populations from the same continent ended up in the same cluster, with the exception of birds from Vancouver (44% assigned to the cluster representing Europe). To illustrate the genetic similarity of populations, an unrooted tree based on pairwise Reynolds’ distances was constructed (Fig. 4). Reynolds’ distances are transformed F ST values that are linearly correlated with the relevant
Australian birds were the smallest (mean body mass ± SD = 12.1 ± 0.6 g; N = 4 populations), North American birds were intermediate (14.4 ± 0.6 g; N = 6 populations), and European birds were the largest (16.1 ± 1.5 g; N = 10 populations; anova: F2,17 = 18.8; P < 0.0001). Within-population variation in body mass (measured as SD) increased in the same order (F2,17 = 10.7; P < 0.001). Tarsus and wing length showed similar patterns (see Appendix I). Populations that have diverged more strongly from the Alice Springs population (measured as FST values) tended to be overall larger (Fig. 5). This trend, however, disappears when accounting for the nonindependence of data points (individual populations) between continents (general liner mixed model with continent as a random effect; F1,16 = 0.02, P = 0.90).
Discussion Using highly polymorphic microsatellite markers, we were able to detect the expected loss of genetic variability in captive as compared to wild zebra finch populations. The observed loss of alleles was comparable to what is expected to occur after 50 generations in an isolated population with an effective population size of 200 individuals. All captive populations still showed substantial genetic variability at all of the studied loci, suggesting that they never experienced a severe genetic bottleneck. We found
Fig. 5 Mean ± SD of body mass of the 20 study populations plotted against FST values to the wild population at Alice Springs. Different symbol types refer to the three continents (crosses, Australia; open squares, North America; filled diamonds, Europe).
© 2007 The Authors Journal compilation © 2007 Blackwell Publishing Ltd
4046 W . F O R S T M E I E R E T A L . that the probability that a captive population has lost (due to random drift if assuming selective neutrality of our markers) a genetic polymorphism that has a high allele frequency in the wild was relatively low. However, given that our markers showed an extremely high number of alleles in the wild, most of the alleles occurred at very low allele frequencies (i.e. 79% of the alleles < 0.05). This large number of rare alleles is practically impossible to maintain in a captive population of limited size. As a consequence, different populations have lost different alleles resulting in statistically significant differentiation between the various laboratory populations. Remarkably, no such differentiation was found between the two wild populations despite the large geographical distance between them, a finding that is in agreement with the species’ low site fidelity (Zann 1996). The presumably random loss of rare alleles in captive populations has created a signature that allows us to correctly assign the majority of individuals to one of the three continents from which we sampled: Australia, North America, or Europe.
Loss of variance due to bottlenecks Each of the 10 microsatellite loci studied showed an extraordinarily high number of alleles in the wild (mean of 24.7 alleles in 48 individuals from population no. 1; HO = 0.93). As far as we are aware, this is the highest allelic richness and observed heterozygosity in microsatellites described for any bird population so far (screening 117 primer notes published in Molecular Ecology and Molecular Ecology Notes up to May 2006). Assuming selective neutrality, such high microsatellite diversity could result from (i) high rates of mutation and/or (ii) large effective population size. Counter to the first explanation, we were unable to find even a single mutation event in 7368 offspring vs. parent comparisons in our captive population (Forstmeier et al. 2007), suggesting that mutation rates are on the lower end of what is known from other organisms (Ellegren 2000). In contrast, the explanation based on large population size is supported by the fact that we found no genetic differentiation between the two studied wild populations that are located approximately 2000 km apart, covering half of the range of the Australian subspecies. This enormous effective population size in the wild is contrasted by the limited number of birds that make up the captive populations. It seems obvious that the high number of alleles per locus that exist in this large wild population cannot be maintained in captivity. The marked reduction in allelic diversity (Fig. 1), however, primarily results from the loss of rare alleles (Fig. 2). Since the populationwide contribution of alleles to additive genetic variance increases with allele frequency (Lynch & Walsh 1998), we expect the captive populations to show only slightly reduced additive genetic variance as compared to the wild. For instance, a single nucleotide polymorphism at a
bi-allelic locus with allele frequencies of p = q = 0.50 contributes 25 times as much to additive genetic variance than a polymorphism with p = 0.01 and q = 0.99 (Lynch & Walsh 1998). While the latter polymorphism is likely to be lost in all but 25% of the captive populations, the more common polymorphism is still present in 97% of the captive populations (extrapolating from Fig. 2).
Comparison to other species In wild bird populations, bottlenecking events often occur when remote habitats like islands are colonized by a small number of individuals who thereby found a new population that is isolated from other populations (founder effect). A number of studies have looked at the relative genetic diversity at microsatellite loci in such bottlenecked populations vs. the mainland populations from which the founders originated (Tarr et al. 1998; Clegg et al. 2002; Pruett & Winker 2005; Hawley et al. 2006). The loss of genetic diversity in captive zebra finch populations (Fig. 1) seems well within the range of what has been observed in these studies on wild bird populations. This is especially the case when acknowledging that the markers we used show an unusually high number of alleles that are rare in the wild (Fig. 2), and that those are the most sensitive to bottlenecking events (Nei et al. 1975; Baker & Moeed 1987). Even the most strongly bottlenecked population of white zebra finches (population no. 20), which we specifically selected because we expected the greatest loss of genetic diversity in such colour mutant strains, still showed a substantial degree of polymorphism (HE = 0.79). For comparison, when microsatellite markers were designed for bird species that were known to have gone through a dramatic bottleneck, much lower levels of diversity have been found. Well-known examples are the Seychelles warbler (Acrocephalus sechellensis) once bottlenecked to 26 individuals (H E = 0.29; Richardson et al. 2000), and the crested ibis (Nipponia nippon) bottlenecked to four individuals (HE = 0.07; Ji et al. 2004). Similarly, domestic peafowl (Pavo cristatus) from various sources in the UK showed very low levels of genetic diversity (HE = 0.10; Hale et al. 2004), suggesting a much more severe bottleneck in the history of domestication as compared to the zebra finch. Note that data from wild peafowl are lacking for comparison, so we cannot make a really strong case here. Nevertheless, we list these examples to illustrate that severe bottlenecking events do indeed result in a dramatic loss of heterozygosity that goes far beyond what we see in the zebra finch.
Population differentiation Given the high variability of our markers, we were able to detect significant differentiation between the various captive populations. The magnitude of between-population © 2007 The Authors Journal compilation © 2007 Blackwell Publishing Ltd
Z E B R A F I N C H P O P U L A T I O N G E N E T I C S 4047 differences is closely mirrored by geographical distance when looking at the country of last origin rather than the current location of the population (see Table 1). Hence, with the exception of population no. 4, which has recently been brought from Australia to Germany, there seems to be fairly limited gene flow between the three continents. Given that most research on captive zebra finches is done in either Europe or North America, the relatively pronounced differentiation between these two continents should be kept in mind when comparing results between them. Additionally, the local variation should not be underestimated. For instance, the University of Bielefeld currently maintains three extremely differentiated zebra finch populations (see Appendix II). Hence, it is of great importance to be precise about the origin of the subjects used in any particular study.
Body size and history of domestication While body size may be an indicator of the intensity and/or duration of selection on this trait by aviculturists, it can definitely not serve as an indicator of genetic distance to the Australian source population (Fig. 5). For instance, the population ‘Bielefeld Victoria’ (no. 4) with a short history of domestication (five to six generations) has not undergone any selection for increased body size, but shows a comparable reduction in genetic diversity as other populations of large body size, which presumably have been domesticated for more than a century (Sossinka 1970). Studying the domestication of zebra finches, Sossinka (1970) pointed out the increased body size of domesticated stock. Interestingly, Sossinka’s domesticated zebra finches, which he obtained from various breeders throughout Germany but also Switzerland in the 1960s, were much smaller (mass 12–13 g, wing length 55–56 mm) than the vast majority of birds that we presently find in Europe. Only their direct descendants, which are still in Bielefeld (population no. 19), as well as birds from the neighbouring country Denmark (population no. 13) still show a comparably small body size compared to other European birds.
Conclusions and recommendations Many research laboratories throughout the world work on similar questions using different zebra finch populations as model organism and, for a variety of reasons, often come up with different findings (e.g. Collins & ten Cate 1996; Jennions 1998; Forstmeier 2005). One possible reason is the genetic differentiation between the various captive populations. The present study is the first to shed some light on the extent and geographical pattern of population differentiation. Our findings suggest that zebra finch researchers should be especially careful in their conclusions when birds originate from different continents. This is © 2007 The Authors Journal compilation © 2007 Blackwell Publishing Ltd
particularly important when the trait of interest shows a high heritability and is influenced by only a few loci with large effects. Knowing the degree of genetic differentiation, a promising approach for the future would be to study well-differentiated populations at the same location. Crosses between populations could be used to reveal the extent to which genetic differences are responsible for the variation in traits of interest, ultimately facilitating the mapping of quantitative trait loci. The present study is likely to revive a debate within the zebra finch research community about the best approach to deal with the genetic variation found within and between populations. More extreme debates have occupied mouse geneticists regarding the pros and cons of using genetically uniform and completely homozygous inbred mouse strains vs. genetically heterogeneous stocks (see Miller et al. 1999; Festing 1999; McClearn 1999). The one major issue of this debate, which is relevant here, is that decreasing genetic variability decreases the generalizability of findings to other populations, but increases the probability that the same finding can be replicated successfully within the same population. Hence, the more a given field of research has to struggle with nonrepeatability of findings between laboratories (e.g. Collins & ten Cate 1996; Jennions 1998), the more important becomes the question of repeatability within laboratories (e.g. Forstmeier 2005), and hence the more attractive becomes the use of relatively genetically uniform and well-defined study organisms. By well-defined, we mean populations that have lived for a long time under relatively constant environmental conditions, such that rapid changes in allele frequencies (selective sweeps) are relatively infrequent. This is in strong contrast to the attempt to work on a more natural study population by starting from a wild-caught stock. Such attempts give highly interesting insights into the actual process of domestication, i.e. how domestication shapes the captive population. However, the genetic composition of such populations is expected to change dramatically over the first few generations, providing a very unstable working ground making the replication of findings more difficult. Hence, the optimal approach clearly depends on the research goals: wild-caught birds are ideal for studying domestication, highly inbred strains are (or in the case of zebra finches would be) ideal for studying the genetics of traits in general (see Festing 1999). We thus conclude that ranking captive populations by allelic richness as in Fig. 1 is not equivalent to ranking them by quality. The value of a captive population is directly related to the amount of information that has been gathered for that population.
Acknowledgements This study would not have been possible without the help of all the colleagues listed in Table 1 who kindly provided DNA
4048 W . F O R S T M E I E R E T A L . samples of their birds, as well as the necessary background information. We are also grateful to Tim Birkhead, who initiated W.F. to zebra finch research and allowed us to take part of his population to Seewiesen. We are also very grateful to all those who helped by gathering the morphometric data (listed in Appendix I). We thank Melanie Schneider for genotyping the DNA samples. Kim Teltscher and Sylvia Kuhn provided supervision in the lab. Henryk Milewski helped fitting the regression line in Figure 2. This project was funded by the Max Planck Society and W.F. was supported by an Emmy-Noether Fellowship (DFG: FO 340/2).
References Baker AJ, Moeed A (1987) Rapid genetic differentiation and founder effect in colonizing populations of common mynas (Acridotheres tristis). Evolution, 41, 525–538. Belkhir K, Borsa P, Chikhi L et al. 1996–2004 GENETIX 4.05, logiciel sous Windows TM pour la génétique des populations. Laboratoire Génome, Populations, Interactions, CNRS UMR 5000, Université de Montpellier II, Montpellier (France). Burley NT, Foster VS (2004) Digit ratio varies with sex, egg order and strength of mate preference in zebra finches. Proceedings of the Royal Society of London. Series B, Biological Sciences, 271, 239– 244. Clegg SM, Degnan SM, Kikkawa J et al. (2002) Genetic consequences of sequential founder events by an island-colonizing bird. Proceedings of the National Academy of Sciences, USA, 99, 8127– 8132. Collins SA, ten Cate C (1996) Does beak colour affect female preference in zebra finches? Animal Behaviour, 52, 105–112. Dakin EE, Avise JC (2004) Microsatellite null alleles in parentage analysis. Heredity, 93, 504–509. Ellegren H (2000) Microsatellite mutations in the germline: implications for evolutionary inference. Trends in Genetics, 16, 551–558. Felsenstein J (1993) PHYLIP, Phylogeny Inference Package. Department of Genetics, University of Washington, Seattle, Washington. Festing MFW (1999) Warning: the use of heterogeneous mice may seriously damage your research. Neurobiology of Aging, 20, 237– 244. Fitch WM, Margoliash E (1967) Construction of phylogenetic trees. Science, 155, 279–284. Forstmeier W (2005) Quantitative genetics and behavioural correlates of digit ratio in the zebra finch. Proceedings of the Royal Society of London. Series B, Biological Sciences, 272, 2641–2649. Forstmeier W, Schielzeth H, Schneider M, Kempenaers B (2007) Development of polymorphic microsatellite markers for the zebra finch (Taeniopygia guttata). Molecular Ecology Notes (doi: 10.1111/j.1471-8286.2007.01762.x). Fowler K, Whitlock MC (1999) The distribution of phenotypic variance with inbreeding. Evolution, 53, 1143–1156. Goudet J (1995) Fstat version 1.2: a computer program to calculate F-statistics. Journal of Heredity, 86, 485 – 486. Hale ML, Petrie M, Wolff K (2004) Polymorphic microsatellite loci in peafowl (Pavo cristatus). Molecular Ecology Notes, 4, 528–530. Hawley DM, Hanley D, Dhondt AA, Lovette IJ (2006) Molecular evidence for a founder effect in invasive house finch (Carpodacus mexicanus) populations experiencing an emergent disease epidemic. Molecular Ecology, 15, 263–275.
Jennions MD (1998) The effect of leg band symmetry on female–male association in zebra finches. Animal Behavior, 55, 61–67. Ji YJ, Liu YD, Ding CQ, Zhang DX (2004) Eight polymorphic microsatellite loci for the critically endangered crested ibis, Nipponia nippon (Ciconiiformes: Threskiornithidae). Molecular Ecology Notes, 4, 615–617. Kalinowski ST, Taper ML, Marshall TC (2006) Revising how the computer program cervus accommodates genotyping error increases success in paternity assignment. Molecular Ecology, 16, 1099–1106. Keller LF, Waller DM (2002) Inbreeding effects in wild populations. Trends in Ecology & Evolution, 17, 230–241. Lynch M, Walsh B (1998) Genetics and Analysis of Quantitative Traits. Sinauer Associates, Sunderland, Massachusetts. McClearn GE (1999) Exotic mice as models for aging research: polemic and prospectus by R. Miller et al. Neurobiology of Aging, 20, 233–236. Miller RA, Austad S, Burke D et al. (1999) Exotic mice as models for aging research: polemic and prospectus. Neurobiology of Aging, 20, 217–231. Nei M, Maruyama T, Chakraborty R (1975) Bottleneck effect and genetic variability in populations. Evolution, 29, 1–10. Nuin PAS (2005) winpop 2.5: software for representing population genetics phenomena. Briefings in Bioinformatics, 6, 390–393. Pritchard JK, Stephens M, Donnelly P (2000) Inference of population structure from multi-locus genotype data. Genetics, 155, 945–959. Pruett CL, Winker K (2005) Northwestern song sparrow populations show genetic effects of sequential colonization. Molecular Ecology, 14, 1421–1434. Reynolds J, Weir BS, Cockerham CC (1983) Estimation for the coancestry coefficient: basis for a short-term genetic distance. Genetics, 105, 767–779. Richardson DS, Jury FL, Dawson DA et al. (2000) Fifty Seychelles warbler (Acrocephalus sechellensis) microsatellite loci polymorphic in Sylviidae species and their cross-species amplification in other passerine birds. Molecular Ecology, 9, 2226–2231. Rutkowska J (2005) Maternal effects in zebra finches — status quo and where we go. ISBE Newsletter, 17, 15. Sossinka R (1970) Domestikationserscheinungen beim Zebrafinken Taeniopygia guttata castanotis (Gould). Zoologischer Jahrbücher, 97, 455–524. Tarr CL, Conant S, Fleischer RC (1998) Founder events and variation at microsatellite loci in an insular passerine bird, the Laysan finch (Telespiza cantans). Molecular Ecology, 7, 719–731. Weir BS, Cockerham CC (1984) Estimating F-statistics for the analysis of population structure. Evolution, 38, 1358–1370. Zann R (1996) The Zebra Finch. Oxford University Press, New York.
Wolfgang Forstmeier is leader of a Junior Research Group working on the genetics of sexual behaviour in the zebra finch. Gernot Segelbacher is a research scientist interested in conservation genetics and the analysis of genetic variation in natural populations. Jakob Mueller is a research scientist focusing on population and evolutionary genetics as well as genetic mapping in model and non-model organisms. Bart Kempenaers is Director at the MPIO heading the Department of Behavioural Ecology and Evolutionary Genetics. His main interest is the study of sexual selection and the evolution of mating systems, parental care and promiscuity.
© 2007 The Authors Journal compilation © 2007 Blackwell Publishing Ltd
Z E B R A F I N C H P O P U L A T I O N G E N E T I C S 4049
Appendix I Morphometric data on zebra finch populations. Means, standard deviations, and number of individuals measured are given for wing length in mm, tarsus length in mm, and body mass in g, as well as the observer taking the measurements. For tarsus length three different methods were used. Measurements taken by methods 2 and 3 were adjusted to fit method 1 (see main text) Wing length ID
Population
Mean
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Alice Springs (wild) Victoria (wild) Victoria Bielefeld-AUS Vancouver Tempe Irvine Davis Cornell Millbrook Cracow Paris Lund Groningen Seewiesen-NL Leiden Stirling Seewiesen-GB Bielefeld Bielefeld (white)
53.70 54.95 54.15 56.03 59.80 55.78 56.83 57.79 55.92 55.58 56.77 59.84 58.40 59.96 59.11 58.40 58.47 55.75 58.35
SD
Tarsus length N
Mean
SD
Body mass N
1.20 1.45 1.23 1.37 1.12 2.11 2.36 1.52 2.32
50 551 92 30 40 16 20 81 12 20 30 140
14.10 14.30 13.60 13.99 16.63 14.05 16.72 16.47 16.38 16.78 14.35 14.68
0.60 0.55 0.66 0.53 0.52 1.15 0.79 0.46 0.68
50 551 92 30 271 28 20 82 29 20 30 140
1.90 1.03 1.74 1.94 1.57 0.92 1.62
131 14 48 30 1050 20 10
16.62 17.85 15.17 14.58 17.15 13.99 14.74
0.53 0.77 0.63 0.61 0.55 0.41 0.85
131 15 48 69 1049 20 10
© 2007 The Authors Journal compilation © 2007 Blackwell Publishing Ltd
Method
Adjusted
Mean
SD
N
Observer
1 1 1 1 2 1 3 3 3 3 1 1
14.10 14.30 13.60 13.99 14.31 14.05 14.18 13.97 13.90 14.23 14.35 14.68
2 3 1 1 3 1 1
14.30 15.15 15.17 14.58 14.55 13.99 14.74
11.30 12.55 11.92 12.43 15.24 14.65 13.68 13.77 14.71 14.32 15.15 15.43 14.38 16.80 18.58 16.84 16.91 17.30 13.71 16.21
1.10 0.87 0.81 1.23 1.67 1.48 0.78 1.38 1.97 1.32 1.46 1.95 1.97 2.10 2.15 2.35 2.21 1.99 1.19 1.77
50 551 92 30 271 28 20 83 29 20 30 140 168 82 16 48 69 1041 20 10
Runciman Zann Zann Naguib Williams McGraw Forstmeier Forstmeier Adkins-Regan Hertel Rutkowska Alonso-Alvarez Sandell von Engelhardt Forstmeier Holveck Roberts Forstmeier Naguib Naguib
4050 W . F O R S T M E I E R E T A L .
Appendix II Pairwise FST values between zebra finch populations. In this table all FST values > 0.01 are significant at P < 0.05
© 2007 The Authors Journal compilation © 2007 Blackwell Publishing Ltd
Alice Springs Victoria (wild) Victoria Bielefeld-AUS Vancouver Tempe Irvine Davis Cornell Millbrook Cracow Paris Lund Groningen Seewiesen-NL Leiden Stirling Seewiesen-GB Bielefeld Bielefeld (white)
Alice Victoria Springs (wild) Victoria Bielefeld-AUS Vancouver Tempe
Irvine
Davis
Cornell Millbrook Cracow Paris
Lund
Bielefeld Groningen Seewiesen-NL Leiden Stirling Seewiesen-GB Bielefeld (white)
0.00573 0.01878 0.05573 0.03044 0.02802 0.03306 0.04882 0.02666 0.04343 0.03847 0.02878 0.03056 0.04694 0.04117 0.04901 0.04662 0.07325 0.09342 0.09808
0.04271 0.01980 0.02949 0.05352 0.04163 0.05594 0.04971 0.05380 0.06741 0.06148 0.08328 0.09716 0.11664
0.03782 0.06117 0.06666 0.06194 0.07888 0.07336 0.08063 0.09148 0.08763 0.09917 0.13399 0.14715
0.02940 0.05030 0.04323 0.05050 0.05454 0.05435 0.06354 0.06815 0.09168 0.10956 0.10404
0.03936 0.03226 0.03890 0.04200 0.07282 0.10824 0.09731
0.01999 0.02473 0.05215 0.08372 0.08169 0.10504
0.00436 0.06067 0.02704 0.02317 0.02618 0.05066 0.02272 0.03574 0.04120 0.03027 0.03268 0.04395 0.03969 0.04986 0.04255 0.07387 0.08678 0.09587
0.06412 0.04238 0.03751 0.04226 0.06580 0.04036 0.05171 0.05102 0.03868 0.04239 0.05249 0.04775 0.05248 0.05636 0.08693 0.10109 0.10092
0.07833 0.07402 0.07370 0.08702 0.08219 0.09196 0.08572 0.07116 0.08261 0.09187 0.08418 0.08949 0.08012 0.09687 0.12976 0.16526
0.03739 0.03782 0.06277 0.03641 0.04133 0.04192 0.03393 0.03305 0.03650 0.04011 0.04811 0.04146 0.07395 0.09723 0.09681
0.02063 0.04373 0.02547 0.02984 0.04822 0.03135 0.04380 0.04201 0.04001 0.05341 0.05729 0.07243 0.09459 0.10659
0.06236 0.04916 0.05410 0.05692 0.06088 0.07490 0.07190 0.10655 0.10629 0.11299
0.03205 0.04602 0.04275 0.04410 0.04506 0.05072 0.07899 0.09307 0.09507
0.02291 0.01611 0.01413 0.02796 0.04456 0.07297 0.08709 0.09449
0.02277 0.04688 0.07599 0.09147 0.08987
0.04812 0.07279 0.05911 0.09874 0.11114 0.12324 0.10404 0.10429 0.13925
0.13135
doi: 10.1111/j.1365-294X.2007.03444.x
Genetic variation and differentiation in captive and wild zebra finches (Taeniopygia guttata) Blackwell Publishing Ltd
W O L F G A N G F O R S T M E I E R ,* G E R N O T S E G E L B A C H E R ,† J A K O B C . M U E L L E R * and BART KEMPENAERS* *Max Planck Institute for Ornithology, Postfach 1564, D-82305 Starnberg (Seewiesen), Germany, †Max Planck Institute for Ornithology, Vogelwarte Radolfzell, Schlossallee 2, D-78315 Radolfzell, Germany
Abstract The zebra finch (Taeniopygia guttata) is a small Australian grassland songbird that has been domesticated over the past two centuries. Because it is easy to breed in captivity, it has become a widely used study organism, especially in behavioural research. Most work has been conducted on domesticated populations maintained at numerous laboratories in Europe and North America. However, little is known about the extent to which, during the process of domestication, captive populations have gone through bottlenecks in population size, leading to inbred and potentially genetically differentiated study populations. This is an important issue, because (i) behavioural studies on captive populations might suffer from artefacts arising from high levels of inbreeding or lack of genetic variation in such populations, and (ii) it may hamper the comparability of research findings. To address this issue, we genotyped 1000 zebra finches from 18 captive and two wild populations at 10 highly variable microsatellite loci. We found that all captive populations have lost some of the genetic variability present in the wild, but there is no evidence that they have gone through a severe bottleneck, as the average captive population still showed a mean of 11.7 alleles per locus, compared to a mean of 19.3 alleles/locus for wild zebra finches. We found significant differentiation between the captive populations (FST = 0.062). Patterns of genetic similarity closely match geographical relationships, so the most pronounced differences occur between the three continents: Australia, North America, and Europe. By providing a tree of the genetic similarity of the different captive populations, we hope to contribute to a better understanding of variation in research findings obtained by different laboratories. Keywords: bottleneck, domestication, founder effect, genetic diversity, genetic drift, inbreeding. Received 1 April 2007; revision accepted 4 June 2007
Introduction The zebra finch, Taeniopygia guttata , is one of the most frequently studied bird species. It is a widely used model organism in studies of behaviour, evolution, ecology, neurobiology, endocrinology and many other fields (Zann 1996). This is signified by the fact that the zebra finch is the second bird species after the chicken for which the entire genome sequence is now becoming available. While there is some information on the behaviour and ecology of zebra finches in the wild (Zann 1996), most research is carried out using domesticated birds of the Australian subspecies Correspondence: W. Forstmeier, Fax: +49 8157 932400; E-mail: [email protected] © 2007 The Authors Journal compilation © 2007 Blackwell Publishing Ltd
Taeniopygia guttata castanotis. Many research laboratories maintain their own breeding stock, with birds typically originating from larger stocks held by aviculturists. The practice of obtaining birds for research from such variable sources brings about uncertainties regarding the genetic background of the study organisms. This is problematic because, during the process of domestication, zebra finches might have gone through repeated bottlenecks in population size, potentially leading to genetic differentiation between the various captive populations by means of genetic drift or differential selection pressures. Moreover, domesticated zebra finches are known to be larger than their wild ancestors (Zann 1996), a fact that has been ascribed to selective breeding for larger body size by aviculturists. These issues raise questions about the extent to which: (i) studies of
4040 W . F O R S T M E I E R E T A L . domesticated birds are distinct from the situation in the wild, and (ii) studies can be compared between the various captive stocks. There are several examples in the zebra finch literature of where findings from one research laboratory could not be replicated by another. In different studies, females showed different preferences with regard to male beak colour (see Collins & ten Cate 1996), but also with regard to colour and symmetry of artificial leg bands (see Jennions 1998). While mate preferences likely are affected by many other factors besides genetics, there are also examples of highly heritable traits that show different patterns in different laboratories. For instance, digit ratio, a morphological trait that is thought to partly reflect sex steroid levels during ontogeny, was sexually dimorphic and depended on the laying sequence within a clutch in a study population from Irvine, California (Burley & Foster 2004), but not in a population from Sheffield, UK, where the trait showed a heritability of 0.72 (Forstmeier 2005). The difference between the results regarding sexual dimorphism and laying-order effects was significant, and thus not due to the lack of statistical power. Unfortunately, there have been no rigorous studies exploring how much of the population differences in findings are due to genetic differences vs. differences in experimental techniques, bird husbandry or other sources of environmental variation. Whether or not such differences in findings can be attributed to genetic differentiation between the various populations depends on whether any genetic differentiation exists in the first place. According to Sossinka (1970), it seems likely that most of the European zebra finches go back to individuals that were brought from Australia approximately between 1870 and 1890, and that relatively few new imports reached Europe after the First World War. However, it is also known that some zebra finches that had been domesticated in Australia were exported to Europe in the 1950s (Sossinka 1970). Apart from such sporadic information, we know too little about the history of zebra finch domestication, about bottlenecks, about effective population sizes, and about gene flow between the captive populations to answer the question of genetic differentiation satisfactorily. We therefore decided to launch an initiative (starting at an international zebra finch workshop held at Bielefeld, Germany in March 2005; see Rutkowska 2005) to study the population genetics of wild and captive zebra finches. The present study is the outcome of this initiative. We compiled a list of laboratories maintaining a captive stock of zebra finches for behavioural research, and asked the owners to contribute DNA samples. All samples were genotyped at 10 highly polymorphic microsatellite loci (Forstmeier et al. 2007), which had not been preselected for their level of polymorphism in any particular population. We also obtained samples from two populations of wild zebra finches to serve as a comparison. We set out to address the following two main questions:
1 How severely inbred or bottlenecked are the various zebra finch laboratory populations as compared to the wild? 2 How strong is the genetic differentiation between the various captive populations? The first question is meant to evaluate the potential concern that behavioural studies on captive populations might suffer from artefacts arising from high levels of inbreeding or lack of genetic variation in such populations (see Fowler & Whitlock 1999; Miller et al. 1999; Keller & Waller 2002). The second question addresses whether genetic differentiation is large enough to potentially be responsible for some of the differences in findings between laboratories. Our study also includes data on body size to examine whether body size can yield similar or additional insights into the history of zebra finch domestication and the assumed selective breeding by aviculturists.
Materials and methods Study populations Table 1 shows a list of populations from which we obtained a sufficient number of samples (range 27–68 individuals per population). We deliberately excluded some populations that were known to be of the same origin and were expected to be very similar to the ones included. The list of populations includes two wild populations from Australia, as well as 18 captive populations from three continents: Australia (two populations), North America (six populations), and Europe (10 populations). The two wild populations were no. 1 from Alice Springs in Central Australia (23°44′S, 133°50′E), and no. 2 from northern Victoria (36°10′S, 145°20′E). The captive populations no. 3 and no. 4 are descendants of birds that were wild-caught at the same site as no. 2. Population no. 3 comprises first to fourth generation aviary-bred birds that are still maintained at Victoria, while population no. 4 contains fifth to sixth generation descendents of 12 pairs of first-generation birds from population no. 2 that had been exported to Bielefeld, Germany in 1992. Since then, this population has not been mixed with birds from other sources. Despite the current location in Germany, we still label population no. 4 as Australian. The North American populations are no. 5 from Vancouver, Canada, all bought from local pet shops and private breeders; no. 6 from Tempe, Arizona, with most of the birds coming from a breeder in San Jose, California, and a few from pet stores in Tempe; no. 7 from Irvine, California, all from local breeders (e.g. Magnolia Bird Farm); no. 8 from Davis, California, all descendents of approximately 80 individuals obtained in 1995 from a commercial breeder in California’s central valley; no. 9 from Cornell, New York, originating from various local sources; and no. 10 from Millbrook, New York. The 10 European populations © 2007 The Authors Journal compilation © 2007 Blackwell Publishing Ltd
© 2007 The Authors Journal compilation © 2007 Blackwell Publishing Ltd
Table 1 The 20 study populations named after their current location and origin. Origin indicates where most of the birds of a population came from immediately before the foundation of the population. The names of population owners and suppliers of DNA samples are listed, as well as the number of individuals sampled (N samples), the mean number of alleles per locus (N alleles), the minimum (Min) and maximum (Max) number of alleles per locus, the mean number of private (unique to a population) alleles per locus in each population (Private), allelic richness (AR), expected heterozygosity (HE), observed heterozygosity (HO), inbreeding coefficient (FIS). All data are based on 10 loci, except HE, HO, and FIS which are based on eight loci Population name
Current location
Origin
Continent
Supplier
N samples
N alleles
Min
Max
Private
AR
HE
HO
FIS
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Alice Springs (wild) Victoria (wild) Victoria Bielefeld-AUS Vancouver Tempe Irvine Davis Cornell Millbrook Cracow Paris Lund Groningen Seewiesen-NL Leiden Stirling Seewiesen-GB Bielefeld Bielefeld (white)
Australia Australia Australia Germany Canada Arizona California California New York New York Poland France Sweden The Netherlands Germany The Netherlands Great Britain Germany Germany Germany
Australia Australia Australia Australia Canada California California California New York New York Poland Czech Republ. Denmark The Netherlands The Netherlands The Netherlands Great Britain Great Britain Germany Germany
AUS AUS AUS AUS NAM NAM NAM NAM NAM NAM EUR EUR EUR EUR EUR EUR EUR EUR EUR EUR
Runciman Zann Zann Naguib & Witte Williams McGraw Burley Millam Adkins-Regan Nottebohm Rutkowska Alonso-Alvarez Sandell & Tobler von Engelhardt Gahr Riebel, Holveck & ten Cate Evans Forstmeier Bischof Bischof
48 27 50 50 50 41 68 52 50 55 48 52 52 59 53 50 52 61 50 32
24.7 18.5 18.6 10.4 16.8 16.1 16.1 11.4 16.1 13.8 13.2 16.9 16.0 13.9 13.1 12.4 12.7 9.5 8.5 6.4
20 15 15 9 15 13 12 9 12 9 9 11 11 9 10 9 9 7 7 4
30 23 25 13 23 22 19 15 21 19 17 24 22 19 18 21 19 12 12 9
1.8 0.4 1.0 0.1 0.3 0.0 0.0 0.2 0.2 0.1 0.1 0.2 0.0 0.1 0.1 0.1 0.0 0.2 0.2 0.4
20.21 18.31 15.46 9.59 14.30 14.22 13.51 10.39 14.18 11.99 11.71 14.09 13.59 11.91 11.67 11.10 11.03 8.52 7.85 6.31
0.94 0.93 0.91 0.84 0.89 0.90 0.89 0.85 0.90 0.88 0.88 0.88 0.88 0.87 0.86 0.86 0.86 0.82 0.79 0.79
0.93 0.93 0.91 0.83 0.86 0.83 0.88 0.86 0.85 0.85 0.87 0.86 0.84 0.87 0.83 0.85 0.81 0.78 0.79 0.80
0.009 0.004 – 0.003 0.016 0.031 0.076 0.018 – 0.004 0.051 0.039 0.010 0.034 0.049 – 0.015 0.044 0.013 0.051 0.049 0.005 – 0.015
AUS, Australia; NAM, North America; EUR, Europe.
Z E B R A F I N C H P O P U L A T I O N G E N E T I C S 4041
ID
4042 W . F O R S T M E I E R E T A L . are no. 11 from Cracow, Poland, bought from various local breeders since 2000; no. 12 from Paris, France, most of which came from a breeder at Blois, SW France (who reported to get most of his birds from the Czech Republic), and the rest from Dijon, SE France; no. 13 from Lund, Sweden, going back to 160 birds obtained in 2001/2002 from various small breeders in Denmark; no. 14 from Groningen, the Netherlands, originating from about 100 individuals obtained from at least five different local breeders between 1998 and 2000; no. 15 from Seewiesen, Germany, all of which had been brought there from Amsterdam, the Netherlands; no. 16 from Leiden, the Netherlands; no. 17 from Stirling, UK, which have exchanged birds with Universities at Glasgow, St Andrews, Edinburgh, and Exeter; no. 18 from our own population at Seewiesen, Germany, going back to 231 individuals obtained in 2002 from a large colony held by T. R. Birkhead at Sheffield University, UK since 1985; no. 19 and no. 20 both from Bielefeld, Germany, with no. 19 partly going back to birds purchased by Sossinka throughout Germany and Switzerland in the 1960s (Sossinka 1970). All populations consist mostly or exclusively of birds carrying wild-type plumage, with the exception of population no. 20, which we particularly selected because it represents a pure strain of an autosomal recessive leucistic mutant known as ‘white’. We expected to find the lowest genetic diversity in this population, because colour mutant strains are usually created by back-crossing a novel mutant bird to its parents. However, that population was also smaller than all the others, consisting of only 32 individuals, all of which were included in this study (Table 1). None of these 18 captive populations seems to have a special history of either purposeful inbreeding to reduce genetic variability or of managed outbreeding to maximize the retention of genetic variability. Providers of samples were asked to randomly (irrespective of sex and relatedness) select 50 adult (> 100d of age) individuals from their entire population. Note that, when captive populations are sampled in this way, a variable number of closely related individuals will be included in each sample, whereas this is much less likely for our population samples from the wild (nos 1 and 2). Final sample sizes varied slightly between the populations (Table 1), mostly due to limitation of the number of birds available (population nos 1, 2, 20), and the availability of extra samples (particularly nos 7, 18) which allowed to keep the total sample size at 1000 individuals.
Genotyping As reported elsewhere (Forstmeier et al. 2007), we developed 12 microsatellite markers for the zebra finch using the DNA of a single female from our own population (no. 18). An extensive test for Mendelian inheritance showed that
two out of the 12 markers (Tgu11 and Tgu13) could not be genotyped reliably in our population (Forstmeier et al. 2007). Hence, in the present study we use the remaining 10 markers. All 1000 individuals were genotyped at each of the 10 loci closely following the methods described in Forstmeier et al. (2007). As described earlier (Forstmeier et al. 2007), two of the 10 loci (Tgu6 and Tgu7) showed evidence for the occurrence of null alleles, and this was again the case in the present data set: estimated null allele frequencies using cervus 3.0 (Kalinowski et al. 2006) were 0.26 and 0.12 for Tgu6 and Tgu7, respectively. Hence, we report deviations from Hardy–Weinberg equilibrium (HE, HO, and FIS; see below) based on the remaining eight loci. All other analyses presented are based on 10 loci, since null alleles do not seem to introduce a major bias (Dakin & Avise 2004). We repeated all analyses based on only eight loci, but all findings were qualitatively the same (data not shown).
Genetic analyses Allele frequencies and estimates of within-population diversity (observed number of alleles, heterozygosity) and between-population divergence (Weir & Cockerham’s (1984) analogue of Wright’s FST) were calculated using fstat 2.9.3 (Goudet 1995) and genetix (Belkir et al. 1996– 2004). Deviation from Hardy–Weinberg expectations for each locus in each population was estimated using the same program. Bayesian clustering software structure 2.1 (Pritchard et al. 2000) was used for assigning individuals to different clusters. We chose k = 3 clusters to represent the three continents, Australia, North America, and Europe. We used the ‘no admixture’ algorithm without any prior population information and used 50 000 runs as burn-in (i.e. simulation data that will be discarded) and 100 000 runs for each chain. Individual genotypes were ordinated in a multidimensional space by a factorial component analysis (FCA) performed with genetix. An unrooted tree showing population relationships was constructed using the Fitch–Margoliash method (Fitch & Margoliash 1967) in phylip (Felsenstein 1993) based on transformed pairwise FST values according to Reynolds et al. (1983).
Morphological data We asked the owners of the 20 zebra finch populations to contribute the following morphological data from any adult individuals of their population: flattened wing length, tarsus length, and body mass (see Appendix I). Tarsus length was measured by three different techniques, either with callipers from the bent foot to the end of the tibiotarsus bone (method 1), or with callipers from the bent foot to the rear edge of the tarsometatarsus, thus including the entire joint (method 2), or like method 2 but using a © 2007 The Authors Journal compilation © 2007 Blackwell Publishing Ltd
Z E B R A F I N C H P O P U L A T I O N G E N E T I C S 4043 Fig. 1 Average allelic richness at 10 microsatellite markers for the 20 study populations in relation to their FST values to the wild population at Alice Springs. Different symbol types refer to the three continents (crosses, Australia; open squares, North America; filled diamonds, Europe).
wing ruler (method 3). To assess the comparability of these methods, W.F. measured 10 birds of our population twice, using each of the three methods. From those data, we estimated that the larger measurements obtained by methods 2 and 3 can be adjusted to method 1 measurements by multiplication with 0.861 (± 0.0035 SE) and 0.849 (± 0.0024 SE), respectively. We used these correction factors, to compare the tarsus measurements from the various populations (Appendix I). Since most measurements were taken by different observers (see Appendix I), the results based on these data should be regarded with caution.
Results Loss of genetic variability in captive populations Ten out of the 20 populations (all captive) showed a significant F IS value (Table 1), indicating that they had been founded from two or more differentiated sources that have not yet been mixed completely, or that they were not bred in a panmictic manner. Allelic richness, i.e. the number of alleles per locus adjusted for the number of individuals sampled, varied substantially between the populations (Table 1). As expected, the two wild populations showed higher allelic richness (mean = 19.3) than the 18 captive populations (mean = 11.7; Mann–Whitney U-test: P = 0.011). Also as expected, the lowest number of alleles was found in the putatively inbred white mutant population no. 20 from Bielefeld (Table 1). The same results are found when focusing on expected heterozygosity as an indicator of genetic diversity (Table 1). We calculated pairwise FST values between all possible pairs of populations (see Appendix II). The overall genetic differentiation among all populations was highly significant © 2007 The Authors Journal compilation © 2007 Blackwell Publishing Ltd
(FST = 0.062). The degree to which captive populations have lost genetic variability (measured as allelic richness) was very closely reflected by their FST distance towards the most diverse wild population no. 1 from Alice Springs in Central Australia (Fig. 1). To better understand what the observed reduction in allelic richness in captive populations means in terms of the probability of losing a genetic polymorphism (e.g. a single nucleotide polymorphism), we did the following procedure. For every allele found in the Australian population nos 1–3 (N = 289 alleles across the 10 loci; N = 125 individuals), we calculated its allele frequency in Australia. The pooling of three populations was necessary to achieve a large-enough sample size, and is justified based on low pairwise FST values (Appendix II). We then plotted the percentage of the 17 captive populations currently held outside Australia that were not showing, and hence had ‘lost’, that allele (when sampling on average 51.5 individuals per population) against the initial allele frequency of the Australian populations. Figure 2 shows that these 17 captive populations primarily lost the alleles that are very rare in Australia, which has important implications for the maintenance of additive genetic variance (see Discussion). We fitted a hyperbolic regression line of the form f (x) = 1 – x/(ax – a + 1), where x is the initial allele frequency and a is a constant that determines the curvature. This function fulfils the necessary marginal criteria that the probability of loss is 1 if the initial allele frequency is 0, and that the probability of loss is 0 if the initial allele frequency is 1. The best fit (least squares method) was reached for a = 0.9696. Hence, while rare alleles (initial frequency = 1%) are absent in most populations (predicted probability of loss = 75%), more common alleles (frequencies, e.g. 10%, 20%, and 50%) are lost only rarely (risk of loss: 22%, 11%, and 3%, respectively).
4044 W . F O R S T M E I E R E T A L .
Fig. 3 Factorial component analysis of the genotypes of 1000 zebra finches from three different continents using genetix. Fig. 2 The proportion of 17 captive populations (nos 4–20) that have lost a particular allele in relation to the frequency of that allele in the three Australian populations nos 1–3. Dots represent the 289 alleles from the 10 different loci. Increasing symbol size refers to multiple data points (up to N = 22). A least squares regression line of the form y = 1 – x/(ax – a + 1) is shown to fit the observed data. For comparison, crosses illustrate the results of a Monte Carlo simulation of genetic drift. Each cross shows the proportion of 1000 populations that have lost an allele of a given starting frequency after 50 generations when the effective population size equals 200 individuals.
To see how the fitted function behaves compared to neutral expectations of loss in an ideal population of constant but limited size, we simulated genetic drift by a stochastic sampling process of gametes from the parental population in each generation using the software winpop 2.5 (Nuin 2005). The following settings turned out to best match the observed data: effective population size Ne = 200, number of generations t = 50. With these settings, an allele with a starting frequency of 0.0346 (i.e. the average observed allele frequency in Australia) has a probability of 57.51% to be lost (based on 10 000 Monte Carlo simulations), which closely fits the observed mean of 57.46%. A similarly good fit is obtained for other combinations where Ne/t = 4 (e.g. Ne = 100, t = 25, or Ne = 400, t = 100). Carrying out the simulation over the whole range of allele frequencies (Fig. 2), the simulated data fit the observed data fairly well (but less well than the fitted line).
Differentiation between populations All populations were significantly differentiated from each other (i.e. significant FST values; see Appendix II) with the following exceptions: (i) the two wild populations no. 1 and no. 2 were not significantly differentiated despite the fact that sampling locations were approximately 2000 km apart; (ii) the captive population no. 3 from Victoria was not different from the wild Victoria population no. 2 from
Table 2 Cluster analysis showing the percentage of individuals from each population assigned to one of three clusters. Bold print indicates which populations are from the same continent
ID
Population
Cluster 1 Australia
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Alice Springs (wild) Victoria (wild) Victoria Bielefeld-AUS Vancouver Tempe Irvine Davis Cornell Millbrook Cracow Paris Lund Groningen Seewiesen-NL Leiden Stirling Seewiesen-GB Bielefeld Bielefeld (white)
85% 82% 95% 95% 13% 12% 4% 3% 4% 5% 8% 17% 29% 3% 4% 4% 14% 2% 4% 2%
Cluster 2 North America
Cluster 3 Europe
10% 10% 2% 1% 43% 82% 89% 96% 93% 93% 8% 7% 4% 4% 3% 2% 4% 2% 0% 0%
5% 8% 3% 3% 44% 6% 7% 2% 3% 2% 83% 76% 67% 93% 93% 94% 82% 96% 95% 97%
which it had been obtained. Between the 18 captive populations, pairwise FST values varied between 0.014 and 0.165. We performed a factorial component analysis (FCA) of the genotypes of the 1000 individuals using genetix. Figure 3 shows the individual scores on the first two components. Populations from the three continents can be clearly distinguished. Table 2 shows the results from a cluster analysis using structure 2.1 where individual birds were allocated to one of three clusters. The three clusters that were identified by the program closely matched the three continents, © 2007 The Authors Journal compilation © 2007 Blackwell Publishing Ltd
Z E B R A F I N C H P O P U L A T I O N G E N E T I C S 4045 population genetics parameters such as genetic drift. This tree not only reflects the basic differentiation according to continents, but also illustrates the relatively high similarity of the two British, and the three Dutch populations, respectively. Absolute branch lengths closely reflect FST values, and hence in our case, the degree of loss of allelic richness (Fig. 1).
Body size as an indicator of domestication
Fig. 4 Unrooted dendrogram showing population relationships (Fitch-Margoliash tree based on Reynolds’ distances). Dotted lines show the separation of the three continents.
Australia, North America, and Europe. Most individuals of most populations from the same continent ended up in the same cluster, with the exception of birds from Vancouver (44% assigned to the cluster representing Europe). To illustrate the genetic similarity of populations, an unrooted tree based on pairwise Reynolds’ distances was constructed (Fig. 4). Reynolds’ distances are transformed F ST values that are linearly correlated with the relevant
Australian birds were the smallest (mean body mass ± SD = 12.1 ± 0.6 g; N = 4 populations), North American birds were intermediate (14.4 ± 0.6 g; N = 6 populations), and European birds were the largest (16.1 ± 1.5 g; N = 10 populations; anova: F2,17 = 18.8; P < 0.0001). Within-population variation in body mass (measured as SD) increased in the same order (F2,17 = 10.7; P < 0.001). Tarsus and wing length showed similar patterns (see Appendix I). Populations that have diverged more strongly from the Alice Springs population (measured as FST values) tended to be overall larger (Fig. 5). This trend, however, disappears when accounting for the nonindependence of data points (individual populations) between continents (general liner mixed model with continent as a random effect; F1,16 = 0.02, P = 0.90).
Discussion Using highly polymorphic microsatellite markers, we were able to detect the expected loss of genetic variability in captive as compared to wild zebra finch populations. The observed loss of alleles was comparable to what is expected to occur after 50 generations in an isolated population with an effective population size of 200 individuals. All captive populations still showed substantial genetic variability at all of the studied loci, suggesting that they never experienced a severe genetic bottleneck. We found
Fig. 5 Mean ± SD of body mass of the 20 study populations plotted against FST values to the wild population at Alice Springs. Different symbol types refer to the three continents (crosses, Australia; open squares, North America; filled diamonds, Europe).
© 2007 The Authors Journal compilation © 2007 Blackwell Publishing Ltd
4046 W . F O R S T M E I E R E T A L . that the probability that a captive population has lost (due to random drift if assuming selective neutrality of our markers) a genetic polymorphism that has a high allele frequency in the wild was relatively low. However, given that our markers showed an extremely high number of alleles in the wild, most of the alleles occurred at very low allele frequencies (i.e. 79% of the alleles < 0.05). This large number of rare alleles is practically impossible to maintain in a captive population of limited size. As a consequence, different populations have lost different alleles resulting in statistically significant differentiation between the various laboratory populations. Remarkably, no such differentiation was found between the two wild populations despite the large geographical distance between them, a finding that is in agreement with the species’ low site fidelity (Zann 1996). The presumably random loss of rare alleles in captive populations has created a signature that allows us to correctly assign the majority of individuals to one of the three continents from which we sampled: Australia, North America, or Europe.
Loss of variance due to bottlenecks Each of the 10 microsatellite loci studied showed an extraordinarily high number of alleles in the wild (mean of 24.7 alleles in 48 individuals from population no. 1; HO = 0.93). As far as we are aware, this is the highest allelic richness and observed heterozygosity in microsatellites described for any bird population so far (screening 117 primer notes published in Molecular Ecology and Molecular Ecology Notes up to May 2006). Assuming selective neutrality, such high microsatellite diversity could result from (i) high rates of mutation and/or (ii) large effective population size. Counter to the first explanation, we were unable to find even a single mutation event in 7368 offspring vs. parent comparisons in our captive population (Forstmeier et al. 2007), suggesting that mutation rates are on the lower end of what is known from other organisms (Ellegren 2000). In contrast, the explanation based on large population size is supported by the fact that we found no genetic differentiation between the two studied wild populations that are located approximately 2000 km apart, covering half of the range of the Australian subspecies. This enormous effective population size in the wild is contrasted by the limited number of birds that make up the captive populations. It seems obvious that the high number of alleles per locus that exist in this large wild population cannot be maintained in captivity. The marked reduction in allelic diversity (Fig. 1), however, primarily results from the loss of rare alleles (Fig. 2). Since the populationwide contribution of alleles to additive genetic variance increases with allele frequency (Lynch & Walsh 1998), we expect the captive populations to show only slightly reduced additive genetic variance as compared to the wild. For instance, a single nucleotide polymorphism at a
bi-allelic locus with allele frequencies of p = q = 0.50 contributes 25 times as much to additive genetic variance than a polymorphism with p = 0.01 and q = 0.99 (Lynch & Walsh 1998). While the latter polymorphism is likely to be lost in all but 25% of the captive populations, the more common polymorphism is still present in 97% of the captive populations (extrapolating from Fig. 2).
Comparison to other species In wild bird populations, bottlenecking events often occur when remote habitats like islands are colonized by a small number of individuals who thereby found a new population that is isolated from other populations (founder effect). A number of studies have looked at the relative genetic diversity at microsatellite loci in such bottlenecked populations vs. the mainland populations from which the founders originated (Tarr et al. 1998; Clegg et al. 2002; Pruett & Winker 2005; Hawley et al. 2006). The loss of genetic diversity in captive zebra finch populations (Fig. 1) seems well within the range of what has been observed in these studies on wild bird populations. This is especially the case when acknowledging that the markers we used show an unusually high number of alleles that are rare in the wild (Fig. 2), and that those are the most sensitive to bottlenecking events (Nei et al. 1975; Baker & Moeed 1987). Even the most strongly bottlenecked population of white zebra finches (population no. 20), which we specifically selected because we expected the greatest loss of genetic diversity in such colour mutant strains, still showed a substantial degree of polymorphism (HE = 0.79). For comparison, when microsatellite markers were designed for bird species that were known to have gone through a dramatic bottleneck, much lower levels of diversity have been found. Well-known examples are the Seychelles warbler (Acrocephalus sechellensis) once bottlenecked to 26 individuals (H E = 0.29; Richardson et al. 2000), and the crested ibis (Nipponia nippon) bottlenecked to four individuals (HE = 0.07; Ji et al. 2004). Similarly, domestic peafowl (Pavo cristatus) from various sources in the UK showed very low levels of genetic diversity (HE = 0.10; Hale et al. 2004), suggesting a much more severe bottleneck in the history of domestication as compared to the zebra finch. Note that data from wild peafowl are lacking for comparison, so we cannot make a really strong case here. Nevertheless, we list these examples to illustrate that severe bottlenecking events do indeed result in a dramatic loss of heterozygosity that goes far beyond what we see in the zebra finch.
Population differentiation Given the high variability of our markers, we were able to detect significant differentiation between the various captive populations. The magnitude of between-population © 2007 The Authors Journal compilation © 2007 Blackwell Publishing Ltd
Z E B R A F I N C H P O P U L A T I O N G E N E T I C S 4047 differences is closely mirrored by geographical distance when looking at the country of last origin rather than the current location of the population (see Table 1). Hence, with the exception of population no. 4, which has recently been brought from Australia to Germany, there seems to be fairly limited gene flow between the three continents. Given that most research on captive zebra finches is done in either Europe or North America, the relatively pronounced differentiation between these two continents should be kept in mind when comparing results between them. Additionally, the local variation should not be underestimated. For instance, the University of Bielefeld currently maintains three extremely differentiated zebra finch populations (see Appendix II). Hence, it is of great importance to be precise about the origin of the subjects used in any particular study.
Body size and history of domestication While body size may be an indicator of the intensity and/or duration of selection on this trait by aviculturists, it can definitely not serve as an indicator of genetic distance to the Australian source population (Fig. 5). For instance, the population ‘Bielefeld Victoria’ (no. 4) with a short history of domestication (five to six generations) has not undergone any selection for increased body size, but shows a comparable reduction in genetic diversity as other populations of large body size, which presumably have been domesticated for more than a century (Sossinka 1970). Studying the domestication of zebra finches, Sossinka (1970) pointed out the increased body size of domesticated stock. Interestingly, Sossinka’s domesticated zebra finches, which he obtained from various breeders throughout Germany but also Switzerland in the 1960s, were much smaller (mass 12–13 g, wing length 55–56 mm) than the vast majority of birds that we presently find in Europe. Only their direct descendants, which are still in Bielefeld (population no. 19), as well as birds from the neighbouring country Denmark (population no. 13) still show a comparably small body size compared to other European birds.
Conclusions and recommendations Many research laboratories throughout the world work on similar questions using different zebra finch populations as model organism and, for a variety of reasons, often come up with different findings (e.g. Collins & ten Cate 1996; Jennions 1998; Forstmeier 2005). One possible reason is the genetic differentiation between the various captive populations. The present study is the first to shed some light on the extent and geographical pattern of population differentiation. Our findings suggest that zebra finch researchers should be especially careful in their conclusions when birds originate from different continents. This is © 2007 The Authors Journal compilation © 2007 Blackwell Publishing Ltd
particularly important when the trait of interest shows a high heritability and is influenced by only a few loci with large effects. Knowing the degree of genetic differentiation, a promising approach for the future would be to study well-differentiated populations at the same location. Crosses between populations could be used to reveal the extent to which genetic differences are responsible for the variation in traits of interest, ultimately facilitating the mapping of quantitative trait loci. The present study is likely to revive a debate within the zebra finch research community about the best approach to deal with the genetic variation found within and between populations. More extreme debates have occupied mouse geneticists regarding the pros and cons of using genetically uniform and completely homozygous inbred mouse strains vs. genetically heterogeneous stocks (see Miller et al. 1999; Festing 1999; McClearn 1999). The one major issue of this debate, which is relevant here, is that decreasing genetic variability decreases the generalizability of findings to other populations, but increases the probability that the same finding can be replicated successfully within the same population. Hence, the more a given field of research has to struggle with nonrepeatability of findings between laboratories (e.g. Collins & ten Cate 1996; Jennions 1998), the more important becomes the question of repeatability within laboratories (e.g. Forstmeier 2005), and hence the more attractive becomes the use of relatively genetically uniform and well-defined study organisms. By well-defined, we mean populations that have lived for a long time under relatively constant environmental conditions, such that rapid changes in allele frequencies (selective sweeps) are relatively infrequent. This is in strong contrast to the attempt to work on a more natural study population by starting from a wild-caught stock. Such attempts give highly interesting insights into the actual process of domestication, i.e. how domestication shapes the captive population. However, the genetic composition of such populations is expected to change dramatically over the first few generations, providing a very unstable working ground making the replication of findings more difficult. Hence, the optimal approach clearly depends on the research goals: wild-caught birds are ideal for studying domestication, highly inbred strains are (or in the case of zebra finches would be) ideal for studying the genetics of traits in general (see Festing 1999). We thus conclude that ranking captive populations by allelic richness as in Fig. 1 is not equivalent to ranking them by quality. The value of a captive population is directly related to the amount of information that has been gathered for that population.
Acknowledgements This study would not have been possible without the help of all the colleagues listed in Table 1 who kindly provided DNA
4048 W . F O R S T M E I E R E T A L . samples of their birds, as well as the necessary background information. We are also grateful to Tim Birkhead, who initiated W.F. to zebra finch research and allowed us to take part of his population to Seewiesen. We are also very grateful to all those who helped by gathering the morphometric data (listed in Appendix I). We thank Melanie Schneider for genotyping the DNA samples. Kim Teltscher and Sylvia Kuhn provided supervision in the lab. Henryk Milewski helped fitting the regression line in Figure 2. This project was funded by the Max Planck Society and W.F. was supported by an Emmy-Noether Fellowship (DFG: FO 340/2).
References Baker AJ, Moeed A (1987) Rapid genetic differentiation and founder effect in colonizing populations of common mynas (Acridotheres tristis). Evolution, 41, 525–538. Belkhir K, Borsa P, Chikhi L et al. 1996–2004 GENETIX 4.05, logiciel sous Windows TM pour la génétique des populations. Laboratoire Génome, Populations, Interactions, CNRS UMR 5000, Université de Montpellier II, Montpellier (France). Burley NT, Foster VS (2004) Digit ratio varies with sex, egg order and strength of mate preference in zebra finches. Proceedings of the Royal Society of London. Series B, Biological Sciences, 271, 239– 244. Clegg SM, Degnan SM, Kikkawa J et al. (2002) Genetic consequences of sequential founder events by an island-colonizing bird. Proceedings of the National Academy of Sciences, USA, 99, 8127– 8132. Collins SA, ten Cate C (1996) Does beak colour affect female preference in zebra finches? Animal Behaviour, 52, 105–112. Dakin EE, Avise JC (2004) Microsatellite null alleles in parentage analysis. Heredity, 93, 504–509. Ellegren H (2000) Microsatellite mutations in the germline: implications for evolutionary inference. Trends in Genetics, 16, 551–558. Felsenstein J (1993) PHYLIP, Phylogeny Inference Package. Department of Genetics, University of Washington, Seattle, Washington. Festing MFW (1999) Warning: the use of heterogeneous mice may seriously damage your research. Neurobiology of Aging, 20, 237– 244. Fitch WM, Margoliash E (1967) Construction of phylogenetic trees. Science, 155, 279–284. Forstmeier W (2005) Quantitative genetics and behavioural correlates of digit ratio in the zebra finch. Proceedings of the Royal Society of London. Series B, Biological Sciences, 272, 2641–2649. Forstmeier W, Schielzeth H, Schneider M, Kempenaers B (2007) Development of polymorphic microsatellite markers for the zebra finch (Taeniopygia guttata). Molecular Ecology Notes (doi: 10.1111/j.1471-8286.2007.01762.x). Fowler K, Whitlock MC (1999) The distribution of phenotypic variance with inbreeding. Evolution, 53, 1143–1156. Goudet J (1995) Fstat version 1.2: a computer program to calculate F-statistics. Journal of Heredity, 86, 485 – 486. Hale ML, Petrie M, Wolff K (2004) Polymorphic microsatellite loci in peafowl (Pavo cristatus). Molecular Ecology Notes, 4, 528–530. Hawley DM, Hanley D, Dhondt AA, Lovette IJ (2006) Molecular evidence for a founder effect in invasive house finch (Carpodacus mexicanus) populations experiencing an emergent disease epidemic. Molecular Ecology, 15, 263–275.
Jennions MD (1998) The effect of leg band symmetry on female–male association in zebra finches. Animal Behavior, 55, 61–67. Ji YJ, Liu YD, Ding CQ, Zhang DX (2004) Eight polymorphic microsatellite loci for the critically endangered crested ibis, Nipponia nippon (Ciconiiformes: Threskiornithidae). Molecular Ecology Notes, 4, 615–617. Kalinowski ST, Taper ML, Marshall TC (2006) Revising how the computer program cervus accommodates genotyping error increases success in paternity assignment. Molecular Ecology, 16, 1099–1106. Keller LF, Waller DM (2002) Inbreeding effects in wild populations. Trends in Ecology & Evolution, 17, 230–241. Lynch M, Walsh B (1998) Genetics and Analysis of Quantitative Traits. Sinauer Associates, Sunderland, Massachusetts. McClearn GE (1999) Exotic mice as models for aging research: polemic and prospectus by R. Miller et al. Neurobiology of Aging, 20, 233–236. Miller RA, Austad S, Burke D et al. (1999) Exotic mice as models for aging research: polemic and prospectus. Neurobiology of Aging, 20, 217–231. Nei M, Maruyama T, Chakraborty R (1975) Bottleneck effect and genetic variability in populations. Evolution, 29, 1–10. Nuin PAS (2005) winpop 2.5: software for representing population genetics phenomena. Briefings in Bioinformatics, 6, 390–393. Pritchard JK, Stephens M, Donnelly P (2000) Inference of population structure from multi-locus genotype data. Genetics, 155, 945–959. Pruett CL, Winker K (2005) Northwestern song sparrow populations show genetic effects of sequential colonization. Molecular Ecology, 14, 1421–1434. Reynolds J, Weir BS, Cockerham CC (1983) Estimation for the coancestry coefficient: basis for a short-term genetic distance. Genetics, 105, 767–779. Richardson DS, Jury FL, Dawson DA et al. (2000) Fifty Seychelles warbler (Acrocephalus sechellensis) microsatellite loci polymorphic in Sylviidae species and their cross-species amplification in other passerine birds. Molecular Ecology, 9, 2226–2231. Rutkowska J (2005) Maternal effects in zebra finches — status quo and where we go. ISBE Newsletter, 17, 15. Sossinka R (1970) Domestikationserscheinungen beim Zebrafinken Taeniopygia guttata castanotis (Gould). Zoologischer Jahrbücher, 97, 455–524. Tarr CL, Conant S, Fleischer RC (1998) Founder events and variation at microsatellite loci in an insular passerine bird, the Laysan finch (Telespiza cantans). Molecular Ecology, 7, 719–731. Weir BS, Cockerham CC (1984) Estimating F-statistics for the analysis of population structure. Evolution, 38, 1358–1370. Zann R (1996) The Zebra Finch. Oxford University Press, New York.
Wolfgang Forstmeier is leader of a Junior Research Group working on the genetics of sexual behaviour in the zebra finch. Gernot Segelbacher is a research scientist interested in conservation genetics and the analysis of genetic variation in natural populations. Jakob Mueller is a research scientist focusing on population and evolutionary genetics as well as genetic mapping in model and non-model organisms. Bart Kempenaers is Director at the MPIO heading the Department of Behavioural Ecology and Evolutionary Genetics. His main interest is the study of sexual selection and the evolution of mating systems, parental care and promiscuity.
© 2007 The Authors Journal compilation © 2007 Blackwell Publishing Ltd
Z E B R A F I N C H P O P U L A T I O N G E N E T I C S 4049
Appendix I Morphometric data on zebra finch populations. Means, standard deviations, and number of individuals measured are given for wing length in mm, tarsus length in mm, and body mass in g, as well as the observer taking the measurements. For tarsus length three different methods were used. Measurements taken by methods 2 and 3 were adjusted to fit method 1 (see main text) Wing length ID
Population
Mean
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Alice Springs (wild) Victoria (wild) Victoria Bielefeld-AUS Vancouver Tempe Irvine Davis Cornell Millbrook Cracow Paris Lund Groningen Seewiesen-NL Leiden Stirling Seewiesen-GB Bielefeld Bielefeld (white)
53.70 54.95 54.15 56.03 59.80 55.78 56.83 57.79 55.92 55.58 56.77 59.84 58.40 59.96 59.11 58.40 58.47 55.75 58.35
SD
Tarsus length N
Mean
SD
Body mass N
1.20 1.45 1.23 1.37 1.12 2.11 2.36 1.52 2.32
50 551 92 30 40 16 20 81 12 20 30 140
14.10 14.30 13.60 13.99 16.63 14.05 16.72 16.47 16.38 16.78 14.35 14.68
0.60 0.55 0.66 0.53 0.52 1.15 0.79 0.46 0.68
50 551 92 30 271 28 20 82 29 20 30 140
1.90 1.03 1.74 1.94 1.57 0.92 1.62
131 14 48 30 1050 20 10
16.62 17.85 15.17 14.58 17.15 13.99 14.74
0.53 0.77 0.63 0.61 0.55 0.41 0.85
131 15 48 69 1049 20 10
© 2007 The Authors Journal compilation © 2007 Blackwell Publishing Ltd
Method
Adjusted
Mean
SD
N
Observer
1 1 1 1 2 1 3 3 3 3 1 1
14.10 14.30 13.60 13.99 14.31 14.05 14.18 13.97 13.90 14.23 14.35 14.68
2 3 1 1 3 1 1
14.30 15.15 15.17 14.58 14.55 13.99 14.74
11.30 12.55 11.92 12.43 15.24 14.65 13.68 13.77 14.71 14.32 15.15 15.43 14.38 16.80 18.58 16.84 16.91 17.30 13.71 16.21
1.10 0.87 0.81 1.23 1.67 1.48 0.78 1.38 1.97 1.32 1.46 1.95 1.97 2.10 2.15 2.35 2.21 1.99 1.19 1.77
50 551 92 30 271 28 20 83 29 20 30 140 168 82 16 48 69 1041 20 10
Runciman Zann Zann Naguib Williams McGraw Forstmeier Forstmeier Adkins-Regan Hertel Rutkowska Alonso-Alvarez Sandell von Engelhardt Forstmeier Holveck Roberts Forstmeier Naguib Naguib
4050 W . F O R S T M E I E R E T A L .
Appendix II Pairwise FST values between zebra finch populations. In this table all FST values > 0.01 are significant at P < 0.05
© 2007 The Authors Journal compilation © 2007 Blackwell Publishing Ltd
Alice Springs Victoria (wild) Victoria Bielefeld-AUS Vancouver Tempe Irvine Davis Cornell Millbrook Cracow Paris Lund Groningen Seewiesen-NL Leiden Stirling Seewiesen-GB Bielefeld Bielefeld (white)
Alice Victoria Springs (wild) Victoria Bielefeld-AUS Vancouver Tempe
Irvine
Davis
Cornell Millbrook Cracow Paris
Lund
Bielefeld Groningen Seewiesen-NL Leiden Stirling Seewiesen-GB Bielefeld (white)
0.00573 0.01878 0.05573 0.03044 0.02802 0.03306 0.04882 0.02666 0.04343 0.03847 0.02878 0.03056 0.04694 0.04117 0.04901 0.04662 0.07325 0.09342 0.09808
0.04271 0.01980 0.02949 0.05352 0.04163 0.05594 0.04971 0.05380 0.06741 0.06148 0.08328 0.09716 0.11664
0.03782 0.06117 0.06666 0.06194 0.07888 0.07336 0.08063 0.09148 0.08763 0.09917 0.13399 0.14715
0.02940 0.05030 0.04323 0.05050 0.05454 0.05435 0.06354 0.06815 0.09168 0.10956 0.10404
0.03936 0.03226 0.03890 0.04200 0.07282 0.10824 0.09731
0.01999 0.02473 0.05215 0.08372 0.08169 0.10504
0.00436 0.06067 0.02704 0.02317 0.02618 0.05066 0.02272 0.03574 0.04120 0.03027 0.03268 0.04395 0.03969 0.04986 0.04255 0.07387 0.08678 0.09587
0.06412 0.04238 0.03751 0.04226 0.06580 0.04036 0.05171 0.05102 0.03868 0.04239 0.05249 0.04775 0.05248 0.05636 0.08693 0.10109 0.10092
0.07833 0.07402 0.07370 0.08702 0.08219 0.09196 0.08572 0.07116 0.08261 0.09187 0.08418 0.08949 0.08012 0.09687 0.12976 0.16526
0.03739 0.03782 0.06277 0.03641 0.04133 0.04192 0.03393 0.03305 0.03650 0.04011 0.04811 0.04146 0.07395 0.09723 0.09681
0.02063 0.04373 0.02547 0.02984 0.04822 0.03135 0.04380 0.04201 0.04001 0.05341 0.05729 0.07243 0.09459 0.10659
0.06236 0.04916 0.05410 0.05692 0.06088 0.07490 0.07190 0.10655 0.10629 0.11299
0.03205 0.04602 0.04275 0.04410 0.04506 0.05072 0.07899 0.09307 0.09507
0.02291 0.01611 0.01413 0.02796 0.04456 0.07297 0.08709 0.09449
0.02277 0.04688 0.07599 0.09147 0.08987
0.04812 0.07279 0.05911 0.09874 0.11114 0.12324 0.10404 0.10429 0.13925
0.13135
