Crumbling diversity: comparison of historical archived and contemporary natural populations indicate reduced genetic diversity and increasing genetic differentiation in the golden-cheeked warbler
Genetic viability of threatened and endangered species is of increasing concern with habitat loss and fragmentation. Valuable assessments of the genetic status of endangered species are difficult in most cases, where only single sample estimates are available. Using historical and contemporary samples, we assessed the impact of both historical and recent demographic changes on population genetics of the endangered golden-cheeked warbler, (Dendroica chrysoparia). Our study documents a steep decline in genetic diversity in an endangered species over a 100-year period, along with concurrent increase in genetic differentiation, and low contemporary effective sizes for all the populations we evaluated. While adding to the growing body of literature that describes the genetic impacts of habitat fragmentation, our study may also serve as an informative guide to future management of endangered species. Our study underlines the importance of long term population genetic monitoring in understanding the full extent of genetic changes in endangered species.
KeywordsGenetic diversity Fragmentation Effective population size Endangered birds Historical-contemporary samples
Habitat loss and fragmentation are among the greatest threats to biodiversity in the 21st century (Young and Clarke 2000). Species that are habitat specialists and require large home ranges are particularly sensitive to habitat loss and fragmentation and may experience increased genetic drift and reduced gene flow. While a number of studies have examined the role of fragmentation on genetic structure (Young et al. 1996; Martinez-Cruz et al. 2007; Howeth et al. 2008; Sato and Harada 2008; Yamamoto et al. 2004), these assessments are often restricted to contemporary patterns, when examination of temporal changes would provide considerably more information. Despite the burgeoning evidence for the implications of habitat fragmentation on contemporary populations (Rodriguez-Munoz et al. 2007; Segelbacher et al. 2008), it is difficult to tease out the magnitude or trajectory of demographic and genetic change accompanying fragmentation, so as to inform conservation management. This is usually compounded by the unavailability of pre-fragmentation or ancestral population samples as reference. Although typing specimens from museum archives is being embraced for various evolutionary and population genetic studies, its application has been relatively limited, owing to incomplete geographical representation or limited sampling. This is especially so for direct assessments of the effects of recent habitat fragmentation on contemporary population structure and genetic diversity. However, when applied, this approach has yielded valuable demographic or conservation relevant information in other species (Nichols et al. 2001).
Identifying temporal changes in genetic variation may be especially valuable in the case of endangered species, as such data might indicate whether the effective population size has been small enough to result in reduced genetic variation (and thus adaptive potential) and if populations have become recently isolated and thus subject to local extirpation. Temporal changes in genetic variance are also valuable for understanding the effective size (Ne) in relation to the census size (Nc) because of the disparity between the number of individuals in a population and actual genetic contributors to each generation (Nei and Tajima 1981; Crow and Denniston 1988; Frankham 1995a). Another advantage of a temporal analysis is in facilitating quantification of genetic change in populations, partly satisfying the proximate data demands in a conservation context.
In an earlier assessment of genetic structure in D. chrysoparia we found that few populations were significantly differentiated, with little evidence of genetic bottlenecks, and genetic diversity comparable to more abundant warbler species (Lindsay et al. 2008). As with similar studies, that investigation was based on contemporary samples alone. However, extrapolating from data based on a single time point limits the resolution of the full extent of recent and ongoing genetic changes. More complete assessments of genetic changes require assessing the temporal component of genetic change. Here we used samples taken from 2 to 3 time points spanning across a century, for three sites in central Texas (Fig. 1) to examine temporal change in genetic characteristics and estimate the effective population size (Ne).
Materials and methods
Locations and times of samples for temporal analyses
Sample locations, time points and sample sizes used for temporal analysis of changes in genetic diversity, population differentiation and estimation of the effective population size Ne
Sample size (N)
Archived samples and processing
We followed the guidelines of Bonin et al. (2004) in processing of historical and contemporary samples to minimize genotyping errors. Historical samples were obtained from four national ornithological collections, namely the National Museum of Natural History (NMNH), Washington, DC, USA, the Museum of Comparative Zoology (MCZ), Cambridge, MA, the Field Museum of Natural History (FMNH), Chicago, IL and the American Museum of Natural History (AMNH), New York, NY (sample IDs are provided in Appendix 2). A total of 73 specimens were sampled and from each archived specimen, a small piece (1 mm3) of skin was extracted from the toe-pads (Ellegren 1991) with a sterile razor blade and stored in a sterile microcentrifuge tube. DNA extraction was performed using a QiaAmp Microextraction kit (Qiagen, Valencia, CA), with initial overnight digest, followed by spin-column DNA isolation and final elution in 40 μl low-TE buffer. Samples were stored in low-TE buffer at −20°C to prevent degradation of potentially fragile, low-yield DNA extracts until genotyping. All archived sample processing was performed in a separate lab under a sterile hood.
Establishing reliability and estimating error rates
We first established a framework to quantify the reliability of our genotype dataset using the program RELIOTYPE (Miller et al. 2002). This program uses a maximum-likelihood approach for reducing genotyping errors and provides recommendations for replications at each locus to reach a ≥95% reliability level (the minimum acceptable level). Once individual loci-sample combinations reach or exceed the reliability threshold, the data can be considered to have a negligible rate of error. We initially determined consistent amplification if at least three out of four attempts resulted in the expected Polymerase Chain Reaction (PCR) product. We used MICROCHECKER (Van Oosterhout et al.2004) to check our data for genotyping errors, large-allele dropout and null alleles. For loci that passed these initial criteria, we analyzed the data in RELIOTYPE to determine if and how many replicates were necessary to meet a minimum reliability level of 95% (for each locus). Amplification of loci that did not meet the reliability threshold were replicated for the recommended number of times and analyzed again to confirm that they exceeded the 95% threshold. In the final dataset, we retained only loci with average reliability scores equaling or greater than 98%. We also used the program GIMLET (Valiére 2002) to estimate the error rate arising due to allelic dropout across all replicates.
Each sample was genotyped following amplification by PCR at each of nine microsatellite loci previously characterized for this species (Lindsay et al. 2008), followed by fragment analysis on an ABI 3130 Genetic Analyzer. PCR optimization and genotyping analyses were performed as reported in Lindsay et al. (2008). PCR on DNA from archived samples was optimized by changes to MgCl2 concentrations or with thermoprofile variations.
We collected allelic data following fragment analyses using GENESCAN software (version 3.1, Applied Biosystems, Foster City, CA). We performed exact tests for deviations from Hardy-Weinberg expectations on allele frequencies using the online tool GENEPOP ON THE WEB (Raymond and Rousset 1995). The program GENETIX (Belkhir et al. 2004) was used to estimate measures of unbiased expected heterozygosity (HEXP), and the program FSTAT (Goudet 2002) was used to estimate allelic richness (AR). AR controls for biases in allelic diversity arising from unequal sample sizes (Leberg 2002). Global diversity indices and other sample-by-locus statistics are provided in Appendix 1.
To quantify the changes in genetic diversity between the historical and contemporary samples, we used a paired t-test (PROC TTEST, Statistical Analysis Software, SAS Institute, Cary, NC), by locus, to compare HEXP and AR between time points.
Historical population genetic structure was characterized by estimating pairwise FST among the historical samples from Kerr, Bexar, and Travis populations in GENETIX (Belkhir et al. 2004). The significance of F-statistic estimates was assessed using a resampling approach (3,000 permutations) to account for differences in population sample sizes. Identical analyses were performed on the corresponding contemporary populations. To verify that obtained FST estimates are not an artifact of high intra-population diversity, we also compared differentiation using two other estimator, namely G’ST and DEST (Hedrick 2005; Jost 2008). We used a paired (by locus) t-test to determine if the degree of pairwise-differentiation was different between the historical and contemporary periods. Additionally, elevated linkage disequilibrium has been associated with genetic fragmentation (Zartman et al. 2006) and we tested whether the differentiation we observed also caused elevated linkage disequilibrium.
Estimation of Ne
We used the pseudo-maximum likelihood estimator (Wang 2001; Wang and Whitlock 2003) as implemented in the program MLNE to estimate Ne (hereafter referred to as MLNe). MLNE provides estimates of the variance Ne across time-spans separating two or more sampling intervals (Wang and Whitlock 2003). We estimated Ne jointly with migration rate (m), where it is assumed that migrants originate from an infinitely large population. Estimation of m requires that a source population be specified from which gene flow may occur. For each of the three populations for which we estimated MLNe, we specified a source population comprised of samples combined from the remaining two populations. We used this estimator as it is the only available temporal estimator that considers populations open to migration, and as we were specifically interested in the temporal estimate of Ne. For comparison with the MLNe estimates, we also estimated the moment Ne (MtNe) based on equations 15 and 18 from Nei and Tajima (Nei and Tajima 1981), which assumes closed populations.
Estimation of Ne using either the moment or likelihood approach requires an assumption of discrete generations. As D. chrysoparia do not conform to the assumption of discrete generations, we used two different generation lengths (T = 1 and 2 years) to evaluate the effects that T may have on estimates of Ne. The majority of adult females in songbirds breed in their first year, whereas reports indicate that age of males at first breeding may be either one or two year (USFWS 1992). Considering this, either generation length may be appropriate, and we applied both in separate analyses in addition to the average of the two (1.5 year/generation). MLNe estimates obtained from T = 1 and T = 2 were different, but were qualitatively similar, with results for T = 1.5 being intermediate of the two. Therefore, only results from T = 1 are reported in the main text (Table 3). Estimates for T = 2 are presented in Table S1.
Current census population sizes (Nc) for the three sites are approximately 90 (Frels Jr 2006), 485 (Source: Camp Bullis, TX) and 1,000 (USFWS 2004) pairs, for Kerr, Bexar, and Travis, respectively (Lindsay et al. 2008). Census methodologies vary between sites and are subject to minor yearly variations, hence population sizes are not to be considered to be exact and are reported as 3-year averages.
Approximately 78% of PCR amplifications on archived samples amplified on the first attempt, compared to 98% for the contemporary samples. Consensus genotypes from historical samples were obtained following multiple replications (up to six replicates for different loci) based on the recommendations from the program RELIOTYPE. Only eight of the nine loci analyzed were amplified consistently and met the 98% reliability threshold that we had established. The eight polymorphic microsatellite loci (Lindsay et al. 2008) were scored in both historical and contemporary samples, with no evidence of null alleles. Error rates in historical sample genotypes were estimated to be 1.9% across loci. A total of 95 different alleles were detected in the historical samples, of which 92 were found in contemporary samples. We did not observe significant deviations from Hardy-Weinberg equilibria across loci.
Both AR and HEXP were significantly lower in contemporary samples compared to historical samples from the same sites, with average declines of 20 and 13%, respectively (Fig. 2). Though the two estimates showed similar declines, there was a greater proportional change in AR.
Pairwise differentiation (FST) between pairs of sites that were sampled for both historical and contemporary periods
Historical v/s contemporary comparison
Estimation of effective population sizes (Ne) using temporal sampling approaches and a generation length of 1 year
Historical sample comparisons
The historical-contemporary estimates of Ne for Bexar (1890–2005), Travis (1913–2005), and Kerr (1915–2005) were 49 (95% CI 24–212), 273 (95% CI 46–621), and 66 (95% CI 43–419), respectively (Table 3). As our estimates of FST indicate that rates of gene flow may not have been constant over the period separating the samples, we are not reporting estimates of m.
Understanding the genetic impacts of fragmentation is clearly of interest for conservation of species, and the increasing attempts to quantify these impacts in various taxa attest to the exigency of the threats faced by biodiversity (Alo and Turner 2005; Veit et al. 2005; Martinez-Cruz et al. 2007; Born et al. 2008; Jordan and Snell 2008; Kramer et al. 2008; Angelone and Holderegger 2009). Despite these efforts, the evidence for the genetic impacts of fragmentation is not unequivocal. Our study addresses some of the lingering questions by analyzing genetic changes from samples straddling a period of well documented habitat loss and fragmentation in central Texas. Our results follow similar reports of reduced genetic diversity, increased spatial structuring or low effective population sizes in various taxa, resulting from habitat fragmentation (Ellis et al. 2006; Chevolot et al. 2008; Dixo et al. 2009). However, our study is among the few that were able to report a decline by comparison with pre-fragmentation samples. Our results add to the important body of literature on the effects of fragmentation on conservation genetics and are potentially informative for future management of the golden-cheeked warbler.
Changes in diversity
Large losses of AR are characteristic of bottlenecks (Leberg 1992) and are of concern because random loss of alleles might reduce adaptive potential. Although the decline in heterozygosity was not to the same extent as allelic richness, these two indices change at different rates in response to demographic changes, suggesting that populations sizes may have been low for a sufficient period of time for inbreeding to accumulate (Nei et al. 1975; Frankham 1995c) in D. chrysoparia populations. Losses of AR are expected to occur rapidly following bottlenecks compared to changes in HEXP (Young et al. 1996). Recent population estimates of D. chrysoparia populations suggest that declines may have slowed, as a result of which declines in AR may have ceased, although it is possible that losses of heterozygosity are ongoing, given the slower rate at which heterozygosity responds to population size changes (Nei et al. 1975). Similar declines in genetic diversity have been documented in other species (Ellis et al. 2006; Lage and Kornfield 2006), but establishing a link with fragmentation has not always been straightforward. Perhaps the most striking finding of our study is its occurrence in combination with increased genetic differentiation.
We obtained slightly different results compared to Lindsay et al. (2008) partly as a result of increased sample size for the Kerr population (31 instead of 16) and using one fewer loci than in the previous study. Additionally, all samples were reanalyzed and scored on a different genotyping platform compared to Lindsay et al. (2008). It is also possible that the partitioning of variation would be different when using different numbers of populations in the analysis (a total seven populations were surveyed in that case). Our results suggest reduced gene flow in contemporary populations compared to historical times indicating that barriers to gene flow are recent phenomena. Additionally, our finding of elevated linkage disequilibrium, providing additional evidence for fragmentation. Historical accounts of habitat from central Texas at turn of the 20th century indicate that preferred D. chrysoparia habitat may have been more contiguous than today (Schmidly and Sansom 2002), and rapid urbanization of central Texas since the 1940s (including 225,000 km of road construction) is known to have greatly reduced the extent and quality of this songbird’s habitat (USFWS 1992). In a recent review on the genetics impacts of landscape alteration we report that changes in population structure appear to be more common in fragmented populations than losses of genetic diversity (Leberg et al. 2009). The finding of increased differentiation along with losses of diversity is one of the main findings of our study, and was achieved mainly by the availability of temporal samples.
Effective size estimates
We found relatively low estimates of Ne based on two different estimators, but they are difficult to interpret without accurate knowledge of census size. The relationship between Ne and Nc is much debated in literature (Frankham 1995b; Nunney 1995), but an Ne that is significantly lower than Nc indicates that a bottleneck may have occurred and may negatively impact genetic variation over time; thus this ratio is of obvious importance to species conservation (Nunney 2000). We did not estimate Ne/Nc ratios in D. chrysoparia as there is little information about historical census sizes, making such a comparison impossible. The relatively small estimates of Ne observed in our study may be a consequence of various demographic factors related to habitat loss and fragmentation, and may reflect poor survival in habitat edges (Lindsay et al. 2008), reduced nest success in degraded habitats (Peak 2007; Reidy et al. 2009), high levels of brood parasitism and nest predation (Sperry et al. 2008), or limitation of habitat on wintering grounds (Rappole et al. 2003). A follow-up temporal analysis of the managed populations will be required to assess if and how Ne is changing in response to management efforts.
Alternatively, D. chrysoparia populations may not have been very large historically. Although forested areas are presumed to have been more contiguous in historical times than at present, it is debatable whether suitable habitat was extensive (given the highly specialized habitat requirements). In a scenario where D. chrysoparia would have historically existed at relatively small effective sizes, continuing fragmentation might significantly impact genetic diversity. The observed steep increase in genetic differentiation indicates that fragmentation becomes critical for populations that are not large to begin with, and suffer further declines due to habitat loss. This suggests that gene flow may be much more sensitive to habitat changes than to within population demographic events (Leberg et al. 2009). This apparent loss of connectedness among populations is expected when habitat becomes fragmented and remnant populations become isolated on patches of preserved habitat—the current situation with D. chrysoparia.
Our findings of a significant decline in genetic diversity, and increases in genetic differentiation are an important addition to the data on fragmentation effects on vertebrate species. Given the high degree of habitat specialization of D. chrysoparia and the slow recovery of habitat to climax oak-juniper communities, our results elicit additional concern for the species’ long-term viability. Considering that our earlier study on contemporary populations failed to detect many of the genetic changes that have occurred in the last century, this study emphasizes the importance of temporal sampling for characterizing genetic trends in endangered species.
We thank the AMNH, New York, NY, MCZ Harvard University, Cambridge, MA, FMNH, Chicago, IL, and NMNH, Washington DC for tissue samples. We thank Ft. Hood, Kerr WMA, Balcones Canyonlands NWR, TPWD, K. Barr, C. Goates, J. Hernandez, S. Pathikonda, and L. Butler for access to field sites, and help with field sampling. We thank J. Neigel, S. Mopper and D. Johnson and anonymous reviewers for comments on an earlier version of this manuscript. This study was funded by U.S. Department of Defense under the Section 6.1 Basic Research Program and U.S. Army 6.2 Threatened and Endangered Species Program to RFL and PLL.
- Belkhir K, Borsa P, Chiki L, Raufaste N, Bonhomme F (2004) Genetix 4.05. logiciel sous windows TM pour la genetique des populations. Laboratoire genome, populations, interactions, CNRS UMR 5000. Universite de Montpellier II, MontpellierGoogle Scholar
- Born C, Hardy OJ, Chevallier MH, Ossari S, Atteke C, Wickings J, Hossaert-Mckey M (2008) Small-scale spatial genetic structure in the Central African rainforest tree species Aucoumea klaineana: a stepwise approach to infer the impact of limited gene dispersal, population history and habitat fragmentation. Mol Ecol 17:2041–2050PubMedCrossRefGoogle Scholar
- Frels D Jr (2006) Performance report. Kerr Wildlife Management Area, Hunt, p 8Google Scholar
- Goudet J (2002) FSTAT 184.108.40.206, a program to estimate and test gene diversities and fixation indices. http://www.unil.ch/izea/softwares/fstat.html. Accessed 25 April 2010
- Leberg PL, Athrey GNR, Barr KR, Lindsay D, Lance R (2009) Implications of landscape alteration for the conservation of genetic diversity of endangered species. In: DeWoody JA, Bickham JW, Michler C, Nichols K, Rhodes OE, Woeste K (eds) Molecular insights into natural resource conservation and management. Cambridge University Press, Cambridge, p 392Google Scholar
- Raymond M, Rousset F (1995) GENEPOP (version 1.2): population genetics software for exact tests and ecumenicism. J Hered 86:248–249Google Scholar
- Schmidly DJ, Sansom A (2002) Texas natural history: a century of change, 1st edn. Texas Tech University, LubbockGoogle Scholar
- USFWS (1992) Golden-cheeked warbler (Dendroica chrysoparia) recovery plan. U.S. Fish and Wildlife, Albuqurque, p 97Google Scholar
- USFWS (2004) Biological opinion, consultation number: 2-12-05-F-021. U.S. Fish and Wildlife Service, PennsylvaniaGoogle Scholar
- Valiére N (2002) GIMLET: a computer program for analysing genetic individual identification data. Mol Ecol Notes 2:377–379Google Scholar
- Young A, Clarke GM (2000) Conclusions and future directions: what do we know about the genetic and demographic effects of habitat fragmentation and where do we go from here? In: Young A, Clarke GM (eds) Genetics, demography and viability of fragmented populations, 1st edn. Cambridge University Press, Cambridge, pp 361–366CrossRefGoogle Scholar