GENETIC DIVERSITY OF YELLOW-BILLED MAGPIES (PICA NUTALLI) BEFORE AND AFTER A WEST NILE VIRUS EPIDEMIC Open Access

Understanding effects of disease on wildlife populations is paramount to conservation and is a topic of great importance across many taxa and habitat types. Disease can result from habitat fragmentation and can create or eliminate population structure (van Riper et al. 2002; Trudeau et al. 2004; Skerratt et al. 2007). Diseases can also occur among different spatial scales, including local, regional, and global scales (Crosbie et al. 2008; Boyles and Willis 2010; Smith et al. 2017). This can be particularly problematic for species exposed to novel diseases, as in cases of emerging infectious diseases (Eggert et al. 2008; Boyles and Willis 2010). Diseases that spread across ecologic boundaries put endemic wildlife at risk (Daszak et al. 2000).

Diseases can lead to extinctions and cause cascading effects on ecosystem dynamics (Boyles and Willis 2010). Disease can also affect genetic diversity of populations, as measured by heterozygosity, allelic richness, patterns of inbreeding, and effective population size (Eggert et al. 2008; McKnight et al. 2017). For example, in the 4 yr after devil facial tumor disease struck Tasmania, genetic subdivision that did not exist prior to the disease developed in Tasmanian devils (Sarcophilus harrisii; Lachish et al. 2011). Despite a 70% decrease in Tasmanian devil population size, there was no evidence of a genetic bottleneck; however, there was an appreciable increase in inbreeding (Lachish et al. 2011). Similarly, after a severe mange outbreak, bobcats (Lynx rufus) in California, US exhibited lower heterozygosity and allelic richness, and higher inbreeding coefficients compared to the population before mange (Serieys et al. 2015). Finally, black-tailed prairie dog (Cynomys ludovicianus) colonies historically exposed to sylvatic plague had lower genetic diversity and allelic richness than did unexposed prairie dog colonies (Trudeau et al. 2004). The cost of lower genetic diversity is often a reduced ability to adapt to diseases and the environment (Spielman et al. 2004; Frankham 2005).

Severe population decline and exposure to disease does not necessarily result in changed genetic health, as measured by genetic diversity. For example, House Finch (Haemorhous mexicanus) populations affected by the Mycoplasma gallisepticum epidemic did not experience any significant changes in allelic richness or heterozygosity compared to before the epidemic (Hawley and Fleischer 2012). Hawaiian honeycreepers (Chlorodrepanis virens) in a low-elevation population were devastated by avian malaria (Foster et al. 2007). However, despite initial declines in population health, selection for pathogen resistance in this lowland population has allowed census numbers to rebound, and Hawaiian Honeycreepers presently have high genetic diversity (Eggert et al. 2008). The number of individuals that survive a disease epidemic, their genetic diversity, and the rate of population recovery, among other factors, can all affect the loss or maintenance of genetic diversity (McKnight et al. 2017). Thus, the outcomes of disease epidemics are not always predictable and must be assessed for each host system.

Birds are the primary hosts of West Nile virus (WNV) in the US and mosquitoes are the primary vectors (McLean 2002). West Nile spread across North America and reached California in 2003, with the first detected WNV-related bird death in California occurring in 2004 (McLean 2006). Compared to other families, corvids were predicted to be most highly impacted by WNV based on WNV laboratory exposure–mortality and seroprevalence experiments (LaDeau et al. 2007). North American corvids are highly susceptible to WNV and are particularly susceptible to the NY99-strain introduced to the eastern US in 1999 (McLean 2002; Komar et al. 2003). In 2001, 50% of the dead American Crows (Corvus brachyrhynchos) tested were WNV positive (McLean 2002). Additionally, within California, 84% of WNV bird deaths in 2004 were corvids. Despite the rapid devastation of some North American bird populations due to WNV (McLean 2002; Wheeler et al. 2009), there are no before–after studies that examined the effects of this epidemic on host genetic diversity.

The Yellow-billed Magpie (Pica nutalli) is of particular interest because the species is endemic to central California and suffered significant population decline after the arrival of WNV (Crosbie et al. 2008; Sauer et al. 2017). The range of the Yellow-billed Magpie extends from the Sacramento Valley to the San Joaquin Valley, and includes inhabitants in the Central Coastal ranges (Crosbie et al. 2014). As of 2006, the National Audubon Society reported the population size of the Yellow-billed Magpie to be about 180,000 (National Audubon Society 2007). More current estimates are highly variable and potentially unreliable, although there is evidence the population declined approximately 33% between 2005–16 (Crosbie et al. 2008, 2014; Sauer et al. 2017). In 2010, 66% of the dead Yellow-billed Magpies tested through the California Department of Public Health were WNV positive (Reisen et al. 2013; California Department of Public Health 2017) and in 2014, the International Union for Conservation of Nature listed the Yellow-billed Magpie as near-threatened (International Union for Conservation of Nature 2017). Because generalist pathogens, such as WNV, are shared and can be maintained among several species, they pose a great risk for extinction (Hudson et al. 2002). Although it is possible for populations to evolve disease resistance, that opportunity is decreased when populations are threatened, small, and vulnerable to other threats, as is the case for the Yellow-billed Magpie (Hudson et al. 2002).

Our goal was to assess whether WNV correlated with changes to genetic diversity (i.e., heterozygosity, allelic richness, effective population size) and population structure of Yellow-billed Magpies. Specifically, we hypothesized that the extreme and rapid population decline in magpies caused a reduction in genetic diversity but did not change connectivity or population structure because of the mobility of the birds and colony-based mating systems (Birkhead 1991). To test these hypotheses, we used a before–during–after genetic data set, including genetic data from magpies sampled prior to, during, and several generations after the start of the California WNV epidemic.

The California Department of Public Health collected dead magpies from 2002–11 as part of its Dead Bird Survey (California Department of Public Health 2017). The California Department of Public Health submitted carcasses appearing to be less than 24 h old for WNV testing to the California Animal Health and Food Safety Laboratories in Davis and San Bernardino, California. We used muscle samples taken then for this study, including birds collected prior to WNV (2002–04; n=135), immediately after WNV arrived (2006–08; n=81), and several generations after WNV (2009–11; n=94). The birds were collected from the Sacramento and San Joaquin Valleys, as well as from the surrounding cities and counties within the magpie's range (Fig. 1).

We stored samples at –80 C until extraction. We extracted samples with QIAGEN DNEasy Blood and Tissue kits (Qiagen Inc., Germantown, Maryland, USA), with a modified protocol for Longmire solution extractions (Hull et al. 2007). We amplified DNA with the use of multiplex polymerase chain reaction (PCR) kits (Qiagen) in the following ratios per reaction: 6.25 µL of 2x QIAGEN multiplex PCR master mix, 1.25 µL of 5x Q-solution, 2.25 µL molecular water, 1.25 µL multiplex primer mix, and 2 µL DNA for a total reaction volume of 13 µL. The PCR conditions were 15 min at 95 C, 40 cycles each of 30 s at 94 C, 90 s at 57 C, 90 s at 72 C, and a final extension for 10 min at 72 C. We used 10 microsatellite loci developed for Yellow-billed Magpies (Ernest et al. 2008) or the Mariana Crow (Corvus kubaryi; Tarr and Fleischer 1998), including: 1B5D, 5A5F, 5A5G, A2, A106, A408, B213, C222, C405, and C424. We completed two PCR replicates per sample per multiplex, with three independent genotype callers.

We used GENEPOP 4.5.1 (Rousset 2008) to estimate frequency of null alleles across temporal periods and for each locus, with the use of the expectation maximization model. We calculated Hardy–Weinberg proportions with an exact probability test, because of the high number of alleles that some of the microsatellite loci contained (Allendorf et al. 2013), and we used a Bonferroni correction for multiple tests. Two loci not in Hardy–Weinberg proportions were dropped from further analyses. To assess if genetic structure was dictated by temporal sampling period, we used STRUCTURE 2.3.4 admixture models with time period as a model prior (i.e., LOCPRIOR; Pritchard et al. 2000). We set length of burn-in period to 100,000, with 500,000 Markov chain Monte Carlo iterations, and K, the number of assigned genetic clusters, varied from 1 to 10, with 10 replicates for each K. We used the Evanno et al. (2005) method to determine the most probable K. After population genetic structure was examined, we combined the results in CLUMPAK (Kopelman et al. 2015). To determine if there was finer-scale substructure at any time point, we ran each temporal period independently under the same conditions. To assess if genetic structure varied spatially, we used GENELAND 4.0 (Guillot et al. 2005). As spatial data were very coarse (city and county only), we set the uncertainty on coordinates to 1 km, with 1,000,000 iterations and 10 independent runs. The analysis did not include birds with missing spatial data, and we ran both a correlated and uncorrelated model.

We used FSTAT 2.9.3.2 (Goudet 1995) to measure allelic richness at each locus within each time point, as well as the inbreeding coefficient F_IS, observed heterozygosity, unbiased heterozygosity, and to detect null alleles. We used BOTTLENECK 1.2.02 to determine if a recent genetic bottleneck had occurred in any of the time periods (Cornuet and Luikart 1996; Piry et al. 1999). The two-phase model was applied, with 1,000 iterations, and allowed for 70% single step mutations, and 30% variance in the two-phase model. We used the Wilcoxon rank sign post hoc test because of our limited number of loci. We calculated the effective population size of each temporal sampling unit with the use of NeEstimator 2.01 via the linkage disequilibrium model (Waples and Do 2008), and assuming random mating with a critical value of 0.05 (Do et al. 2014). Generated results were likely coarse approximations of the true effective population size, because the assumption of nonoverlapping generations may have been violated by the data. We estimated population differentiation (F_ST) across time periods in GenAlEx. We also measured mean relatedness in GenAlEx with the use of the Ritland model for pairwise relatedness within each time period (Ritland 1996). Comparisons across time periods for varying statistics were run through χ² tests.

All loci were polymorphic. Average of null alleles across temporal periods and loci was 0.15 (SE=0.05). Two loci, 1B6G and 2A5A, were significantly different after a Bonferroni correction from the null (α=0.05) under the Hardy–Weinberg exact probability test and were dropped from further analyses. Population differentiation (F_ST) among time points was low (mean=0.004±0.001), indicating little variance across time points and no substructure. Heterozygosity, inbreeding coefficients, and allelic richness remained consistent among time points, and each time point had genetic estimates similar to the combined data set (P>0.05; Table 1). Average relatedness was low in each time period and did not significantly differ among time periods (Table 2). The estimated effective population size for each temporal period with a lowest allele frequency used (≥0.05) is shown in Table 2.

We detected evidence for two bottlenecks based on heterozygosity excess, one in the 2002–04 time period (Wilcoxon signed-rank test; one-tailed: P=0.033), and a second supported bottleneck in the 2009–11 time point (P=0.026). Given the few generations between 2002 and 2011, this is likely the same bottleneck that we detected at two different time points, as opposed to two unique bottlenecks. There was no evidence of a bottleneck during the 2006–08 time point (P=0.12).

The Evanno method from STRUCTURE results supported K=2 populations; however, population assignment varied randomly across time periods, and is likely an artefact, because the Evanno method cannot determine if K=1 is the best model (Fig. 2). All 10 models in GENELAND converged on a single endemic population (i.e., K=1). Analyses of individual time points also failed to show any population substructure.

Our data indicate that Yellow-billed Magpies have low genetic diversity, and there was little change in the genetic diversity of Yellow-billed Magpies over the course of about 10 yr, before and since the arrival of WNV. Heterozygosity, allelic richness, and inbreeding coefficients from after the arrival of WNV were not different from genetic metrics from before the arrival of WNV. However, because of potential limitations in number of microsatellites, sample size, and distribution used in the study, we may not have had enough power to capture all subtle genetic changes adequately. It is possible that some genetic change resulted from the rapid and extreme population decline of Yellow-billed Magpies, but we may not have had enough statistical power to capture all genetic changes adequately. In addition, microsatellites we used are assumed to be neutral markers, and therefore we were unable to test for adaptation to the disease (Schlötterer 2000). This study does, however, provide a basis for Yellow-billed Magpie genetic diversity at three different time periods to monitor into the future.

We found one genetic cluster (genetic population) for Yellow-billed Magpies sampled in our study, and no substructure was found by treating each time period as a discrete data set. However, we were not able to collect many samples from areas other than Sacramento and Yolo counties in 2009–11. Therefore, we recommend that species range-wide sampling be done in future studies. If there is truly only one genetic population, there would be low opportunity for genetic rescue should the census size continue to diminish. However, current estimates for inbreeding and relatedness were low, and did not change over sampling period 2002–11. Although interspecific genetic diversity comparisons are difficult with microsatellites, because they are often species-specific markers and evolve uniquely in different species, the heterozygosity of Yellow-billed Magpies was low (H_e=0.64) in comparison to vulnerable corvids (International Union for Conservation of Nature 2017), including the Island Scrub Jay (Aphelocoma insularis; H_e=0.65; Delaney and Wayne 2005) and the Florida Scrub Jay (Aphelocoma coerulescens; H_e=0.67; Coulon et al. 2008). The common and ecologically stable Western Scrub Jay (Aphelocoma californica) has a much higher heterozygosity estimate (H_e=0.93; Delaney and Wayne 2005). Moreover, the average number of alleles was much less for Yellow-billed Magpies (5.6±0.4), than either Island Scrub Jays (6.71±2.1) or Florida Scrub Jays (9.6±1.01). Effective population sizes did not vary significantly across the time periods. We must caution, however, that our estimates of effective population size may be biased because our data do not have discrete generation sampling and sampling size is small. Dead magpies were collected opportunistically, and no pedigree or progeny information is available. It is expected that generations overlapped between our artificial time period bins, and that this would reduce the accuracy and power of any effective population size (Waples 2006). Additionally, genetic diversity may not decrease collinearly with population size if the population retains a genetically representative number of breeders (Zenger et al. 2003). Further, although genetic drift can increase the loss of rare alleles after a large population decline, fixation can only take place after several generations.

We detected bottlenecks in the 2002–04 and 2009–11 time periods. Several factors may have contributed to the bottlenecks, including the natural history of magpies, extreme environmental conditions in California at the time, increasing human population, and changes in land use. Although there is anecdotal evidence of intentional poisoning of birds on agricultural properties in the 1960s, the LD50 (lethal dose for 50% of the test subjects) of 4-aminopyridine is 2.4 mg/kg (Schaefer et al. 1973). Reportedly, only a few birds would be exposed, and the resulting effects of loud screeching and seizures would alarm most of the birds to leave the site (Schaefer et al. 1973). Therefore, it remains anecdotal whether 4-aminopyridine or other agricultural “pest” eradication efforts caused any significant decline in the Yellow-billed Magpie population.

The growth of the human population and urban sprawl throughout the home range of magpies has increased in recent years. Sacramento County's population increased by 200,000 people from 2000–10 (State of California, Department of Finance 2017); half of the increase took place between 2000–04. The California Department of Finance predicts the human population of Sacramento County to increase from 1.5 million in 2017 to 1.6 million by 2020, and 1.7 million by 2025 (State of California 2017). Prior to 2002, the Palmer Drought Severity Index, which accounts for precipitation, temperature, and soil moisture to calculate the severity of a drought period, reached –3.8 (severe drought), and California experienced drought conditions from 2000 through 2005, and another drought from mid-2006 through the end of 2009 (National Climatic Data Center 2017). The effects of population decline, a highly pathogenic disease, severe drought, increasing human populations, and wildfires may have synergistically formed the bottlenecks detected from early in the study.

The prevalence of WNV in birds in California has been erratic over the past decade, and it is difficult to predict how WNV will continue to infect birds in the future (Sauer et al. 2017). Because of the high density of communal roosts among Yellow-billed Magpies, they may be at a greater risk of contracting WNV through atypical routes of transmission. The virus is most commonly transferred through mosquito–bird parasitism, but may also be transmitted via contact between conspecifics. Komar et al. (2003) found that infected Black-billed Magpies (Pica hudsonia) could infect cage mates by physical contact, and detected WNV in cloacal–fecal sheddings and oral exudates. Average distance between Yellow-billed Magpie roosts can be as close as 7 m, and several breeding pairs can live in one tree simultaneously during the mating season (Birkhead 1991). Crosbie et al. (2014) found the average density of magpies across the range of the species is 6.1/km². Additionally, 50–60% of young magpies have current home ranges that overlap with their natal ranges (Birkhead 1991). Low dispersal rates and high-density communal living could allow for persistence of the virus in the population, and eventually the loss of entire familial or location-specific genes. An important consideration for this study and the future genetic health of magpies is that WNV has only been present in California for a short amount of time. The generation time of Yellow-billed Magpies is reportedly 1–2 yr; however, young birds can reproduce as early as their first spring (Birkhead 1991). Typically, magpies will not produce replacement clutches should their offspring die, and thus only reproduce once per year. Only 3–5 generations likely occurred in the 10-yr period of this study, which may not be enough time to detect population structure or genetic changes (Lachish et al. 2011).

Future work aimed to monitor the population health of this endemic species should explore the whole genome of Yellow-billed Magpies and use genetic analyses that are finer-scale and include a broader array of microsatellites. Comparisons of future changes to the magpie population to our baseline genetic diversity assessment will provide a better indication of how this population will be affected after additional generations.

For their help with advice, sample collection, manuscript review, and analysis, we thank N. Anderson, A. Brault, R. Carney, S. Crosbie, L. Dalbeck, D. Gille, I. Holster, J. Hull, V. Kramer, L. Marcus, N. Pedersen, W. Reisen, T. Drazenovich, J. Well, and helpful personnel at the California Animal Health and Food Safety Laboratory, California Department of Public Health/Vector Borne Disease Section. We thank the California Department of Fish and Game (E. Loft, D. Steele, S. Torres), California Department of Health Services (V. Kramer, R. Carney, L. Marcus, K. Padgett), California Animal Health and Food Safety Laboratory (A. Ardans, I. Hoser, J. Jennings, L. Woods), University of California Davis Center for Vectorborne Diseases (V. Armijos, S. Wheeler, W. Reisen), the University of California Davis Wildlife Health Center (W. Boyce), and the University of California Davis Veterinary Genetics Laboratory (N. Pedersen, M.C.T. Penedo, T. Gilliland). For natural history and biology of Yellow-billed Magpies we gratefully thank S. Crosbie, L. Souza, M. Reynolds, W. Koenig, G. Bolen, W. Weathers, C. Trost, B. Berteaux, D. Bell, and the many Magpie Monitor volunteers. We thank the California Department of Fish and Wildlife; University of California Genetic Resources Conservation Program; Sacramento, Sequoia, and Yolo Audubon Societies; the Veterinary Genetics Laboratory at University of California Davis; and private donations for financial support.