Review of DNA-based census and effective population size estimators

The detection of reductions in effective population size (Ne) or census size (Nc) is essential for conservation. Recent developments allow wildlife researchers to obtain genetic material via non-invasive sampling techniques that may provide the large sample sizes necessary for precise estimates of Ne and Nc. Population genetic theory provides several methods to estimate Ne from allele frequency data: including temporal change in allele frequencies, gametic disequilibrium and heterozygote excess methods. Modification of capture–mark–recapture methods for use with multi-locus genotype data provides new means for estimating Nc. The combination of new DNA sampling techniques, polymerase chain reaction-based DNA markers and analytical methods may provide unprecedented power to detect reductions in Ne and Nc of endangered populations. However, these genetic methods are largely untested in the field. We review some relatively unexplored, but promising ways that multi-locus genetic data can be used to provide important genetic and demographic information and suggest avenues for further research in this area. All correspondence to: Michael K. Schwartz. E-mail: [email protected] Present Address: Gordon Luikart, Laboratoire de Biologie des Populations d’Altitude, CNRS UMR 5553, Université Joseph Fourier, BP 53, F-38041 Grenoble Cedex 9, France. PCR-based molecular markers Two advances in molecular genetics hold great promise for the application of genetic markers to the estimation of Ne or Nc of wildlife populations. These are: (1) the development of highly polymorphic DNA markers, and (2) the ability to amplify these markers with the polymerase chain reaction (PCR) from low-quality, lowquantity DNA samples. In particular, the use of microsatellites, a class of highly variable, single locus, genetic markers has proven valuable in assessing genetic variation at the population level (e.g. Bruford & Wayne, 1993). We focus on the use of these markers for Ne and Nc estimation, but also recognize that other nuclear, sexspecific, and mitochondrial DNA markers can yield important genetic and demographic information (Reed et al., 1997; Taberlet, Camarra et al., 1997). It has recently been shown that microsatellite genotypes can be determined using samples taken from hair, skin, saliva, feather, feces, or urine (for a review, see Morin & Woodruff, 1996). This means non-invasive sampling techniques, in which direct contact (visual or otherwise) between researchers and animals is avoided, may provide sufficient DNA for the genetic and demographic analyses outlined below. Perhaps one of the greatest advantages of non-invasive genetic sampling is that it will provide the large sample sizes needed to make more precise estimates of Ne or Nc. Additionally, these markers allow PCR amplification of DNA from museum specimens. For instance, Mundy et al. (1997) compared DNA from 30 loggerhead shrikes collected from San Clemente Island in 1915 (museum specimens) to those collected in the 1990s to quantify changes in genetic variation. However, when relying on microsatellite genotypes amplified from low-quality, low-quantity DNA it is important to acknowledge potential pitfalls, such as (1) allelic dropout (Gerloff et al., 1995; Pemberton et al., 1995; Taberlet, Griffin et al., 1996; Gagneux, Boesch & Woodroff, 1997) and (2) PCR generated ‘falsealleles’ (Taberlet, Camarra et al., 1997). To minimize these problems several laboratories have recommended repeating extractions (in the case of shed hair; Gagneux et al., 1997) or a ‘multiple-tubes approach’ where DNA is extracted and subsequently distributed among several tubes to be amplified separately by PCR (Taberlet et al., 1996, 1997). ESTIMATES OF Ne FROM GENETIC DATA Each of the estimators of Ne is based on the assumption that genetic drift increases when Ne decreases. Implicit with this assumption is that selection, mutation, population subdivision and migration do not change gene frequencies within a population. Microsatellite markers are less likely to be under selection than isozymes as they do not code for proteins. However, some microsatellite loci may be linked to regions of the genome under selection. Mutation is unlikely to be problematic unless the mutation rates are very high (i.e. >>10–3). Migration and substructure assumptions, on the other hand, are likely to be violated in many natural populations (for a thorough discussion, see Waples, 1989, 1991), but are less likely to be violated in small, endangered populations that are often isolated. Two of these estimators, linkage disequilibrium and heterozygote excess, actually assess Neb (the effective number of breeding adults). Neb is nearly equivalent to Ne in populations with non-overlapping generations. Ne can be estimated from Neb in populations with overlapping generations if generation length, age class specific birth and death rates, and the mating system of the species is well known (Waples, 1991; Jorde & Ryman, 1995; Scribner, Arntzen & Burke, 1997). One sample: gametic disequilibrium and heterozygote excess methods The first Ne estimation method we review is gametic disequilibrium (D), also known as linkage disequilibrium. This is simply the non-random association of alleles at different loci. D can be generated by many factors, including natural selection (Vrijenhoek, Pfeiler & Wetherington, 1992), hybridization or mixing of differentiated gene pools (Forbes & Allendorf, 1991), and genetic drift (Waples, 1991), and can occur between linked and unlinked loci. Here, we assume observed D is produced by drift in a small population among unlinked loci and focus on how this relationship can be exploited to estimate Ne. The validity of this assumption will depend upon factors such as the number of loci used, chromosome number and potential for hybridization, and should be carefully considered for each population. Just as genetic drift in a small population causes allele frequencies to change at individual loci, drift generates non-random associations among alleles at different loci. For example, with two genes, D is the difference between the observed frequency of co-occurence of alleles (A1, B2) at two loci (1 and 2) and that expected based on their allele frequencies (p1, q2), assuming random association of gametes (from Hill, 1981). D = freq (A1, B2) – p1q2 (1) As Ne decreases, drift plays an increasingly large role in determining allele frequencies, causing observed and expected frequencies of combinations of alleles from different loci to diverge and D to increase. D is closely related to the correlation, r, among alleles at different loci (r = D[(p1, q1, p2, q2]). The relationship between r and Ne has been determined by Hill (1981) and Waples (1991):