Genetic implications of bottleneck effects of differing severities on genetic diversity in naturally recovering populations: An example from Hawaiian coot and Hawaiian gallinule

Abstract The evolutionary trajectory of populations through time is influenced by the interplay of forces (biological, evolutionary, and anthropogenic) acting on the standing genetic variation. We used microsatellite and mitochondrial loci to examine the influence of population declines, of varying severity, on genetic diversity within two Hawaiian endemic waterbirds, the Hawaiian coot and Hawaiian gallinule, by comparing historical (samples collected in the late 1800s and early 1900s) and modern (collected in 2012–2013) populations. Population declines simultaneously experienced by Hawaiian coots and Hawaiian gallinules differentially shaped the evolutionary trajectory of these two populations. Within Hawaiian coot, large reductions (between −38.4% and −51.4%) in mitochondrial diversity were observed, although minimal differences were observed in the distribution of allelic and haplotypic frequencies between sampled time periods. Conversely, for Hawaiian gallinule, allelic frequencies were strongly differentiated between time periods, signatures of a genetic bottleneck were detected, and biases in means of the effective population size were observed at microsatellite loci. The strength of the decline appears to have had a greater influence on genetic diversity within Hawaiian gallinule than Hawaiian coot, coincident with the reduction in census size. These species exhibit similar life history characteristics and generation times; therefore, we hypothesize that differences in behavior and colonization history are likely playing a large role in how allelic and haplotypic frequencies are being shaped through time. Furthermore, differences in patterns of genetic diversity within Hawaiian coot and Hawaiian gallinule highlight the influence of demographic and evolutionary processes in shaping how species respond genetically to ecological stressors.


| INTRODUCTION
Genetic diversity is the basis for which evolutionary forces act upon and ultimately shape the trajectory of variation within populations through time. Standing genetic variation within populations is influenced by a variety of factors: biological characteristics of the species (e.g., dispersal propensity, distribution, mating system, and generation time), human disturbance (e.g., habitat fragmentation, introduced predators, and hunting), and evolutionary forces acting although stochastic (e.g., genetic drift), deterministic (e.g., mutation), and adaptive (e.g., natural selection) processes (Amos & Hardwood, 1998). The strength at which evolutionary processes influence levels of genetic diversity is dependent, in part, on long-term effective population sizes.
Populations that are small and isolated, for example, may be predisposed to having low genetic diversity. Small populations are more likely to experience severe population fluctuations concomitant with environmental stochasticity (Melbourne & Hastings, 2008) which reduces long-term effective population sizes and ultimately the accumulation of genetic diversity (Habel & Zachos, 2013). None of these processes are the sole driver of change; it is the interaction between evolutionary and biological forces that affect the levels of diversity within populations and species over time.
Populations occupying islands are often characterized as having reduced levels of genetic diversity relative to continental counterparts, a pattern often attributed to founder effects associated with island colonization. Among newly founded species, the genetic legacy of the founder effect may still be observed as insufficient time has passed for the species/population to overcome reductions in genetic diversity accompanying the bottleneck postcolonization (Habel & Zachos, 2013).
Many island endemics, however, colonized their respective islands millions of years ago; therefore, a genetic signature of the original founding event is likely not present, and more recent processes (e.g., habitat loss) are driving observed levels of genetic variation (Stuessy, Takayama, López-Sepúlveda, & Crawford, 2014). Although the pattern of low genetic diversity of populations occupying islands may generally hold, processes (i.e., recent effects or legacy of colonization) promoting this pattern may differ among island taxa (Habel & Zachos, 2013).
The Hawaiian Archipelago is one of the most remote island groups in the world enabling it to function like a semiclosed system prior to human arrival (500 AD); the only vertebrates to colonize it were highly vagile species (birds and bats), and its diverse avifauna can be traced to as few as 20 colonizing species (James, 1991). Similar to many Oceanic islands, however, many of the endemic species of Hawaii have gone extinct or experienced severe population declines as a result of anthropogenic influences (e.g., habitat loss/modification, introduced predators, invasive species, and disease; Blackburn, Cassey, Duncan, Evans, & Gaston, 2004). Indeed, the main Hawaiian Islands have the highest number of threatened, threatened endemic, and critical endemic restricted-range species of the Pacific Endemic Bird Areas, many of which are likely close to extinction (BirdLife International, 2016a).
Because islands were likely founded by just a few individuals, due to the remoteness of the archipelago, populations may generally harbor low levels of genetic diversity relative to mainland congeners (even prior to recent declines); this would be especially true for species that are recent colonizers. Potentially low levels of genetic diversity coupled with small population sizes may make Hawaiian endemics more susceptible to deleterious effects of environmental and demographic fluctuations. Therefore, it is important to determine standing levels of genetic variation within species that occupy the Hawaiian Islands to evaluate the impact of the population declines on genetic diversity as well as to gain insight into the ability of Hawaiian endemics to respond to stochastic processes.
The Hawaiian coot (Fulica alai) and Hawaiian gallinule (Gallinula galeata sandvicensis) are endangered waterbirds endemic to the Hawaiian Archipelago ( Figure 1). Both species experienced severe population declines in the early 1900s attributed to wetland loss and modification, introduced plants and predators, disease, and altered hydrology (U.S. Fish and Wildlife Service, 2011). Prior to the 1900s, both species were common and distributed across the main Hawaiian Islands (U.S. Fish andWildlife Service, 2011), although no estimates (pre-1900s) (Underwood, Silbernagle, Nishimoto, & Uyehara, 2013), and it occupies its historical range. The Hawaiian gallinule's current distribution is restricted to the islands of Oahu and Kauai, and recent surveys recorded low numbers (~400 individuals range wide), although numbers are likely underestimated as Hawaiian gallinules are secretive (Underwood et al., 2013).
Hawaiian coot and Hawaiian gallinule are hypothesized to be relatively recent colonizers to the archipelago (Fleischer & McIntosh, 2001); however, based on fossil evidence, the Hawaiian coot may have occupied the islands for a longer duration than the gallinule. A Fulica sp. fossil was discovered at Ulupau Head, Oahu, a formation dated to 120,000 years before present (James, 1987), and on Kauai, F. alai bones were identified from stratigraphic units dating up to 6,760 years before present (Burney et al., 2001). Conversely, Gallinula chloropus (now G. galeata) bones have only been found in deposits postdating human colonization (~1,500 years before present; Barber's Point, Oahu; James, 1987), although the absence of detection does not necessarily equate to lack of presence.
Many factors influence the maintenance and recovery of genetic diversity as species go through population declines. Therefore, some species within an area may be more adversely affected, while others may be more resilient in retaining genetic diversity. Here, we examine the influence of a population decline on genetic diversity simultaneously experienced by two species, the Hawaiian coot and the Hawaiian gallinule, using microsatellite genotype and mitochondrial sequence data. Specifically, we aim to examine how these species have responded genetically to recent ecological stressors (e.g., habitat loss, predation, disease). Although these two species share many life history traits, a few potentially influential demographic differences exist, such as severity of decline, effective population sizes predecline, dispersal propensity, and longevity (Bannor & Kiviat, 2002;Pratt & Brisbin, 2002). Along with providing insights into the conservation genetics of Hawaiian waterbirds, the results of this study can inform conservation planning and management strategies, especially because both species are designated as "conservation reliant," such that they will require active management into perpetuity (Underwood et al., 2013).

| Samples
Specimens of Hawaiian coot (n = 14) and Hawaiian gallinule (n = 23) collected between 1893 and 1939 from Oahu, Hawaii, were obtained through museum tissue loans (Appendix 1). These samples represent the historical time period. Hawaiian coot (n = 34) and Hawaiian gallinule (n = 29) were trapped on James Campbell National Wildlife Refuge, Oahu, Hawaii, from 2012 to 2013. These samples represent the modern time period. Blood samples were collected from the brachial vein and stored in preservation buffer (Longmire et al., 1988).
Hereafter, we will refer to each sampled temporal time period as a unique "population."

| Laboratory techniques
DNA from historical samples was extracted using a phenol-chloroform protocol and recovered with a Microcon centrifugal filter (Millipore, Massachusetts). Genomic DNA from contemporary samples was extracted using a "salting out" procedure described by Medrano, Aasen, and Sharrow (1990). Genomic DNA concentrations were quantified using fluorometry and diluted to 50 ng/ml working solutions. Genotype data were collected at 13 loci for Hawaiian coot (Fal02, Fal04, Fal08, Fal10, Fal12, Fal14, Fal16, Fal19, Gch03, Gch07, Gch12, Gch14, Sonsthagen, Wilson, & Underwood, 2014;and KiRa10, Brackett, Maley, Brumfield, & McRae, 2013) and 10 loci for Hawaiian gallinule (Fal08, Fal10, Fal12, Fal14, Fal19, Gch06, Gch12, Gch13, Gch17, and Gch19, Sonsthagen et al., 2014). Polymerase chain reaction (PCR) amplifications and thermocycler conditions followed Talbot et al. (2011). DNA extractions and PCR reactions for museum specimens were performed in a designated low-copy laboratory to reduce the possibility of contamination from modern samples. DNA extractions and PCR reactions contained sample blanks, which were included through the data collection process. Sample blanks did not yield any product. Because DNA extracted from museum specimens may be of low quality and quantity resulting in genotyping errors (i.e., allelic dropout, and null alleles), all DNA samples from museum specimens were amplified in duplicate at all microsatellite loci, and individuals that were homozygous at loci were amplified in triplicate. In addition, 10% of the samples from James Campbell National Wildlife Refuge were extracted, amplified, and genotyped in duplicate for quality control. No inconsistencies in genotype scores were observed between replicates. Microsatellite genotype data are accessioned at the USGS Alaska Science Center data repository (https:// doi.org/10.5066/f74q7sxc).  (2007). For quality control purposes, we extracted, amplified, and sequenced 10% of the samples in duplicate. Data generated during this study are available at doi.org/10.5066/F74Q7SXC, and sequences are accessioned in GenBank (>200 bp; MF673896-MF673904).

| Analysis of genetic diversity
We calculated allelic richness, observed and expected heterozygosities, Hardy-Weinberg equilibrium (HWE), and linkage disequilibrium at the microsatellite loci in FSTAT version 2.9.3 (Goudet, 1995). Tests for null alleles and allelic dropout were implemented in MicroChecker (Van Oosterhout, Hutchinson, Wills, & Shipley, 2004). MtDNA sequences were trimmed to the same fragment length between historical and modern samples. Haplotype (h) and nucleotide (π) diversity were calculated at mtDNA loci in ARLEQUIN version 2.0 (Schneider, Roessli, & Excoffier, 2000). Fu's F S (Fu, 1997) and Tajima's D (Tajima, 1989) were calculated to test the hypothesis of selective neutrality for mtDNA loci and implemented in ARLEQUIN. We applied critical significance values of 5%, which requires a p-value of below .02 for Fu's F S (Fu, 1997). An unrooted haplotype network for mtDNA loci was constructed in NETWORK version 4.613 (Fluxus Technology Ltd., 2015) using the reduced median method (Bandelt, Forster, Sykes, & Richards, 1995), to illustrate possible reticulations in the gene tree because of homoplasy or recombination.

| Analysis of genetic structure
The degree of population genetic structure within Hawaiian coot and Hawaiian gallinule sampled pre-and postdecline was assessed by calculating F ST and Φ ST for microsatellite and sequence data, respectively, in ARLEQUIN, adjusting for multiple comparisons using Bonferroni correction (α = 0.05). Tamura-Nei nucleotide substitution model (Tamura & Nei, 1993) was used to calculate Φ ST . Because samples were assayed over temporal scales and sizes varied among populations, population differentiation based on χ 2 distributions of alleles and haplotypes was also determined using GENEPOP 3.1 (Raymond & Rousett, 1995).
We used the Bayesian clustering program, STRUCTURE 2.3.2 (Hubisz, Falush, Stephens, & Pritchard, 2009;Pritchard, Stephens, & Donnelly, 2000), to assign individuals to clusters based on their microsatellite allelic frequencies and infer the occurrence of genetic structure without a priori knowledge of putative temporal period.
Data were analyzed using an admixture model assuming correlated frequencies and sample location information as a prior with a 50,000 burn-in period, 500,000 Markov chain Monte Carlo iterations, and number of possible populations (K) ranging from one to five; the analysis was repeated 10 times to ensure consistency across runs.
We followed the method of Evanno, Regnaut, and Goudet (2005) to determine the most likely number of clusters given the data.

| Analysis of effective population size
Contemporary effective population size (Ne) was estimated with NeESTIMATOR v2 (Do et al., 2014), using the linkage disequilibrium-based and the molecular coancestry methods based on the microsatellite data. Molecular coancestry method examines the level of allele sharing among individuals; conversely, the linkage disequilibrium method tests for nonrandom associations formed among alleles at different loci that occur when Ne is low and genetic drift influences allelic frequencies (Luikart, Ryman, Tallmon, Schwartz, & Allendorf, 2010). Microsatellite data were analyzed using the randommating model, with 95% confidence limits determined by jackknifing over loci. We evaluated the effects of low-frequency alleles on Ne estimates by excluding rare alleles (Pcrit). Variance in Ne estimates across a range of Pcrit values is suggestive of a history of gene flow and/or the presence of first-generation dispersers, whereas stable Ne estimates across a range of Pcrit values are indicative of isolated populations (Waples & England, 2011). We estimated Ne with Pcrit values for lowest allele frequency observed ranging from 0.02 and 0.10 and without a frequency restriction.

| Analysis of population demography
Evidence for fluctuations in historical population demography was evaluated using BOTTLENECK 1.2.02 (Cornuet & Luikart, 1996;Piry, Luikart, & Cornuet, 1999). Fluctuations in population size inferred from microsatellite data were assessed using a Wilcoxon sign-rank test using 5,000 permutations under three models: infinite allele model (IAM), stepwise mutation model (SMM), and two-phased model of mutation (TPM; parameters: 70% SMM, variance 9). Heterozygote deficiency relative to the number of alleles indicates recent population growth, whereas heterozygote excess indicates a recent population bottleneck (Cornuet & Luikart, 1996). It is important to note that BOTTLENECK compares heterozygote deficiency and excess relative to genetic diversity, not to HWE expectation (Cornuet & Luikart, 1996). Results were adjusted for multiple comparison using Bonferroni correction (α = 0.05).

| Genetic diversity
Null alleles and allelic dropout were not detected for any loci assayed for historical and contemporary Hawaiian gallinule populations and contemporary Hawaiian coot. Within the Hawaiian coot historical population, three loci (Fal10, Gch12, and KiRa10) had evidence of null alleles. Reamplification of these loci yielded consistent genotypes, and all loci and populations were in HWE and linkage equilibrium. Therefore, we retained all loci in subsequent analyses. Indices of genetic diversity based on microsatellite loci were similar between time periods for both species (95% confidence limits overlapped), although fewer private alleles were observed within the contemporary Hawaiian gallinule (Table 1).
Within the mtDNA control region sequences, five haplotypes characterized by four variable sites were observed for Hawaiian coot, and two haplotypes characterized by a single variable site were observed for Hawaiian gallinule (Figure 2). Greater variation at ND2 was observed for Hawaiian gallinule; three haplotypes were characterized by two variable sites, whereas only two haplotypes were observed within Hawaiian coot ( Figure 2). All haplotypes were retained when sequences were trimmed to a common length. A reduction in genetic diversity was observed at mtDNA for both species, although differences in indices of genetic diversity are greater for Hawaiian coot than Hawaiian gallinule (Table 1). Fewer haplotypes were observed within the mtDNA control region for the contemporary populations of Hawaiian coot and Hawaiian gallinule; further haplotype and nucleotide diversity were lower in the contemporary populations (Table 1); however, one of the two haplotypes observed in Hawaiian gallinule historical population was only found in a single individual (Figure 2).

Indices of genetic diversity appear similar between timescales for
Hawaiian gallinule based on ND2, although similar reductions in variation were observed for Hawaiian coot (Table 1). Tests of selective neutrality were not significant (Table 1).

| Genetic structure
Patterns of temporal variation at allelic and haplotypic frequencies varied between species. A strong signature of genetic structure was observed within Hawaiian gallinule as frequency differences among microsatellite loci were detected across all three metrics (F ST , R ST , and χ 2 ; Table 2) and was also uncovered in STRUCTURE (K = 2, ΔK = 162.4, r = .44; Figure 3). Conversely, only a weak signature of genetic structure was detected in Hawaiian coot based on microsatellite data; differences in allelic distributions (χ 2 ) were observed; however, no structure was detected based on F ST and R ST (Table 2) nor was genetic partitioning uncovered in STRUCTURE (K = 1, LnP|K = −1548.9; K = 2, LnP|K = −1567.5). Frequency differences were detected in mtDNA control region haplotypic data based on F ST and χ 2 distributions between historical and contemporary Hawaiian coot populations ( control region within Hawaiian gallinule between time periods, as only two haplotypes were observed and one haplotype represented by a single individual (Figure 2).  (Table 1).

| Population demography
Fluctuations in population demography were observed within the contemporary Hawaiian gallinule timeframe; significant heterozygote excess was detected at all mutation models (IAM, SMM, and TPM; p < .001) suggestive of population decline. The other timeframes were in mutation drift equilibrium. often reduced to just a few individuals (e.g., Bouzat, Lewin, & Paige, 1998;Groombridge, Jones, Bruford, & Nichols, 2000). However, long-lived species can reduce the effective time spent experiencing a bottleneck thus enabling the retention of genetic diversity over short timescales (e.g., Hailer et al., 2006;Johnson, Gilbert, Virani, Asim, & Mindell, 2008;White et al., 2014). It is important to note, our ability to detect a recent bottleneck is dependent on the level of genetic diversity pre-and postdecline as ancient reductions in genetic diversity may mask recent declines (e.g., Dussex, Rawlence, & Robertson, 2015). While the strength of the decline appears to be greater within Hawaiian gallinule than Hawaiian coot, coincident with the reduction in census size, these species exhibit similar life history characteristics and generation times (7 and 6 years, respectively; BirdLife International 2016b,c). Therefore, we hypothesize that differences in behavior and colonization history are likely T A B L E 2 Estimates of genetic differentiation (F ST , R ST , χ 2 , and Φ ST ) calculated from 13 and 10 microsatellite loci, respectively, 140 bp of mtDNA control region, and 172 bp of mtDNA ND2 between Hawaiian coot and Hawaiian gallinule sampled at two time periods. Significant comparisons (α = 0.05) are in bold text and marked with an asterisk Microsatellites mtDNA control region mtDNA ND2 Regardless, reduction in gene flow among remnant patches will lead to loss of genetic diversity through drift, as drift is inversely related to the effective population size (1/2Ne; Frankham, Ballou, & Briscoe, 2010).

| DISCUSSION
Movement of individuals among islands may provide an avenue for dispersal and in turn greater retention of genetic diversity through increased effective population size thereby reducing the effects of drift (Frankham et al., 2010;Jangjoo, Matter, Roland, & Keyghobadi, 2016).
Within the more sedentary Hawaiian gallinule, population recovery was likely the result of recruitment within remnant wetland patches, and therefore, the survivors of the decline changed the trajectory of neutral genetic variation as evidenced by significant structuring ( neously. Yet despite these similarities, allelic frequencies differed markedly pre-and postdecline within Hawaiian gallinule, and only minimal differences were observed within Hawaiian coot. These findings illustrate the importance of considering how demographic (e.g., dispersal) and evolutionary processes (e.g., bottleneck effects) influence population recovery when planning and implementing conservation programs, especially for endemic species that likely harbor reduced levels of genetic variation relative to their mainland counterparts. In light of these results, Hawaiian gallinule may be at increased risk of genetic endangerment relative to its ecological equivalent, the Hawaiian coot. Although indices of genetic diversity are not reduced (confidence intervals overlap), allelic frequencies have changed indicating genetic drift is strongly influencing genetic variation within the Hawaiian gallinule on Oahu. Continued persistence in isolation, coupled with low genetic diversity, could result in the continued reduction of genetic diversity through genetic drift making the Hawaiian gallinule more vulnerable to stochastic processes and ultimately extirpation.