Urban residency and leukocyte profiles in a traditionally migratory songbird

Abstract Many animals are shifting migrations in response to human activities. In particular, human-induced changes to climate and habitat (e.g., urbanization) likely facilitate animals becoming year-round residents. Because migration can be energetically expensive, shifts to sedentary behavior could minimize energetic demands incurred and any immunosuppressive effects. Residency in urban habitats could also provide abundant resources and allow sedentary animals to invest more in immunity. However, urban habitats could also expose sedentary animals to novel stressors that counter such benefits. To examine how recent shifts to residency affects physiology in ways that may shape infectious disease dynamics, we analyzed leukocyte profiles of two dark-eyed junco (Junco hyemalis) populations from southern California: the Laguna Mountain population, in which birds breed in high-elevation forests and migrate altitudinally, and the urban University of California San Diego population, which was likely established by overwintering migrants in the 1980s and has since become non-migratory. Over a two-year study of each population’s breeding season, we found no difference in the ratios of heterophils to lymphocytes between populations. However, urban residents had more leukocytes than birds from the altitudinal migrant population. A multivariate analysis suggested urban residents had fewer monocytes, but effect sizes were small. These results suggest no differences in energy demands or stressors between urban resident and altitudinal migrant populations during their breeding season. However, urban residency may confer immunological benefits through anthropogenic resources, which could have important consequences for disease dynamics..


Introduction
Animal migration, the regular movement of individuals between breeding and overwintering sites to track seasonal changes in resources and habitats, is a widespread phenomenon [1]. Such behavior has important ecological implications for nutrient cycling, pollination and propagule dispersal, and species interactions [2,3]. However, animal migrations are changing globally in response to human activities [4,5]. Changes in climate and habitat can decrease migratory propensity or drive animals to eliminate migration altogether and form resident populations [6]. Such patterns have been especially well-documented in birds [7,8]. For example, white storks (Ciconia ciconia) have forgone fall migrations to Africa and now subsist on urban landfills in Spain throughout the year [9,10]. Similarly, urban habitats, by providing supplemental food in the form of bird feeders and possibly milder climates, have facilitated the loss of migration and adoption of sedentary behavior in songbirds such as the Eurasian blackcap (Sylvia atricapilla) and European blackbird (Turdus merula) [11,12]. In particular, this colonization of urban habitats in the traditional overwintering range appears to be a widespread phenomenon [13][14][15].
Because migration can be energetically expensive [16,17], shifts to sedentary behavior could minimize the energetic demands incurred from long-distance movements. Given that strenuous activities such as migration can force trade-offs with energetic investment in other physiological systems, such as suppressing components of the immune system [18,19], these transitions towards residency could in turn improve resistance or tolerance to pathogens and alter infectious disease dynamics [13,20]. Such effects could be especially amplified from residency in urban habitats, given the potential for greater food abundance from intentional (e.g., bird feeders) and unintentional (e.g., landfills) resources and therefore minimized foraging effort [21,22]. Environments with more resources could accordingly allow hosts to invest more resources in both immunity and other costly traits (e.g., reproduction) [23,24]. However, urban habitats could also expose residents to novel stressors through changes in intra-specific competition (e.g., elevated density and crowding) [25,26], exposure to novel parasites [27][28][29], artificial light at night and noise pollution [30,31], and toxicants [32,33]. Given such opposing effects, understanding how urbanized habitats affect the physiology of recent colonists is necessary to predict the consequences for wildlife health, conservation, and pathogen spillover [34][35][36]. Such work is especially needed given global increases in urban habitat and loss of migration [13,37].
To assess how urban residency shapes the physiology of traditionally migratory animals, we focus on two distinct populations of dark-eyed juncos (Junco hyemalis), a songbird species complex found across North America [38]. In southern California, the natural breeding range of juncos (J. h. thurberi) is confined to high-elevation forests (e.g., >1500 meters) [39]. However, in the early 1980s, a population of juncos became established in an urban habitat on the coastal campus of the University of California San Diego (UCSD) [40,41]. Population genetics and stable isotopes of hydrogen suggest this resident population was established by overwintering migrants from multiple mountains, including but not limited to the nearby Laguna Mountains (LM) [42,43] (Fig. 1). The milder climate and abundant food, water, and nesting resources in the urban environment may have facilitated the adoption of sedentary behavior, in contrast to the altitudinal migrations observed among LM juncos in winter [40,43,44]. Indeed, LM juncos still display more migratory restlessness than UCSD juncos [45]. Accordingly, the UCSD population has a breeding season more than twice the length of LM juncos and thus greater fecundity [40].
We capitalized on a longitudinal study comparing UCSD and LM juncos across their breeding seasons in 2006 and 2007 [46,47]. We first used differential leukocyte counts from blood smears to assess whether urban habituation minimizes or amplifies stressors. Stimulation of the hypothalamic-pituitary-adrenal axis can cause acute increases in plasma corticosterone (CORT) [48]; however, baseline levels are rapidly altered by capture [49]. However, elevated CORT stimulates an influx of heterophils from bone marrow to blood while distributing lymphocytes from blood to other tissue, increasing the ratio of heterophils to lymphocytes (i.e., HL ratios) [50]; these hematological shifts can remain unchanged during an acute stressor and only elevate hours to days after treatment [51]. Importantly, HL ratios shift in response to mild and chronic stressors [52]; for example, whereas Eurasian kestrel (Falco tinnunculus) nestlings elevated HL ratios and CORT in response to starvation, only HL ratios increased with less-severe stressors such as poor weather [53]. Elevated HL ratios from subtle, chronic stressors can remain elevated for long durations (e.g., a week or more) [53,54]. We also estimated total white blood cell (WBC) counts to indicate investment in cellular defense. High total WBC counts can suggest more robust defenses or an inflammatory response to infection [22,55]. Considering how HL ratios and total WBC counts both vary between urban resident and altitudinal migrant populations, alongside leukocyte composition more broadly (e.g., eosinophil counts can decline with stress but increase with parasitism, high monocyte counts can indicate infection [56,57]), can inform whether hematological patterns indicate stressors, disease, or altered defense [51].
Because UCSD juncos appear to benefit reproductively from urban residency [40], we predicted that HL ratios would be lower in urban residents compared to birds from the altitudinal migrant population [58]. Lower HL ratios could also be expected in urban residents given that attenuated plasma CORT responses have been observed in UCSD juncos [47] and other urban resident birds [59]. If the greater abundance of food, water, and nesting resources in urban habitats allows juncos to invest more resources in both immunity and reproduction, we would also predict UCSD juncos display greater total WBC counts compared to LM birds [23,24]. Alternatively, the additional stressors of urban residency (e.g., light pollution, crowding, disease) could have a greater influence on physiology than resource abundance or minimized energetics, which could result in urban residents showing higher total WBC counts (indicating infection) alongside elevated HL ratios (indicating infection or other stressors) than LM birds [33,60].

Junco sampling
In 2006 and 2007, we used mistnets and Potter traps to sample juncos in the respective breeding seasons of the UCSD population (February-August) and LM population (May-July) [46,47]. We fit each bird with an aluminum U.S. Fish & Wildlife Service leg band. We determined age class from wing plumage and secondarily from iris color [38]. Sex was determined by plumage coloration, wing chord length, and breeding morphology (i.e., cloacal protuberance or brood patch) [38]. We obtained ~100 μL of blood by pricking the brachial vein with a sterile 26G needle, followed by collection with heparinized capillary tubes (hematocrit was obtained for all birds sampled in 2006). Blood smears were prepared on glass slides, fixed with methanol, and stained with Wright-Giemsa [61]. All birds were bled shortly after capture (x̅ =17.18 minutes, SD=9.27) to minimize effects of capture stress on circulating leukocyte numbers [62]. All sampling protocols were approved by the Indiana University Institutional Animal Care and Use Committee (06-242) and conducted under scientific collecting permits issued by the U.S. Fish and Wildlife Service, the California Department of Fish and Game, and the U.S. Forest Service.

Leukocyte profiles
Smears were analyzed from a subset of all sampled birds (n=75; Fig. S1) by two individuals who received identical training, including a phase where both observers analyzed a subset of the same slides to validate correspondence in cell identification and enumeration. More samples were analyzed from 2006 (n=51) than 2007 (n=24), although sampling effort was consistent between LM (n=36) and UCSD (n=39). Greater variation in sampling date stems from the longer breeding season of the UCSD population [40]. For differential leukocytes, we recorded the percentage of heterophils, lymphocytes, monocytes, eosinophils, and basophils by counting at least 50 WBCs at 1000X magnification (oil immersion) [61]. By recording total leukocytes detected during this screening and the number of fields scored, we estimated the total WBCs per 10,000 red blood cells based on the approximation of 100 red blood cells per field. This method is common in ecological studies [22,55,63] but is less precise than hemocytometer-derived total WBC counts [64]. This semi-quantitative approach also assumes that the total red blood cell count is constant across birds. To account for this variation, we multiplied the total WBC counts by hematocrit (i.e., adjusted total WBC count) [65]; however, hematocrit was only available for 2006 birds.
We also used blood smears to screen all individuals for the presence of intracellular and extracellular blood parasites [61]. We scanned each blood smear at 200X for the presence of microfilariae and haemosporidian parasites (e.g., Plasmodium spp. and Haemoproteus spp.).

Statistical analysis of leukocyte data
Differential WBC data are compositional; changes in the proportion of one leukocyte type correspond to changes in the other four cell types [66]. Because analyses of compositional data deal with logratio transformations [67], we treated zeros as values below the detection limit and replaced them with 50% of the bounds for detection (i.e., 0.5%) [68]. We next calculated HL ratios and log 10transformed both this outcome and the total WBC counts; these two physiological measures were not strongly associated with one another (ρ=0.11, p=0.33).
As our data included recaptured juncos (n=10 individuals), we used generalized linear mixed models (GLMMs) with Gaussian errors and a random effect of band number to analyze HL ratios and total WBC counts. We mainly tested the hypothesis that hematological outcomes varied between the LM and UCSD population; however, we also adjusted for factors such as sex (e.g., leukocyte counts are often female-biased [69]), age (e.g., immune senescence [70]), and time (e.g., leukocytes can vary within seasons and across years [71]). We also considered interactions between population and each of these factors, because the effect of population on physiology could also be modified by individual or temporal variation. Because including all these predictors would dramatically overfit any one GLMM given our sample size (n=75), we used a model comparison approach [72]. We considered a base GLMM with population, sex, age, year, and Julian date alongside models that included each interaction between population and these individual and temporal predictors; this produced five a priori GLMMs. We used the lme4 package in R to fit GLMMs with maximum likelihood, and we used MuMIn to compare models with Akaike information criterion corrected for small sample sizes (AICc) [73][74][75]. We considered models within two AICc to be competitive [72]. We then refit GLMMs with restricted maximum likelihood to quantify marginal and conditional R2 (R2m and R2c) [76].
Variation in total red blood cell counts could introduce variation into our estimated total WBC count. Because hematocrit data were only available for juncos sampled in 2006 (n=51), we repeated this model comparison approach with the log 10 -transformed adjusted total WBC count and removed year as a predictor. To assess whether hematocrit itself varied between LM and UCSD juncos, we logit-transformed this response variable and fit our most parsimonious model.
We lastly used a multivariate approach to analyze leukocyte composition in relation to population. We used a permutational multivariate analysis of variance (PERMANOVA) with the vegan package to quantify the proportion of variance explained by population, sex, age, year, and Julian date [77]. The PERMANOVA used Euclidean distances on the isometric logratio (ILR)-transformed relative abundances of all leukocytes [78]. We visualized differences using a robust principal components analysis (PCA) with the robCompositions package [79,80]. Such methods appropriately analyze compositional data (given dependencies among proportions) and handle outliers; however, results can still be vulnerable to inconsistencies in cell counts (e.g., inaccurate smear detection of monocytes [81]). To facilitate interpretation of the PERMANOVA, we also logit-transformed each leukocyte type, fit our parsimonious GLMM, and adjusted the p values.

Analysis of total WBC counts
Juncos at our southern California sites ranged from 0.12 to 1.85 WBCs per field of view, and estimated total WBCs were more variable than HL ratios (x̅ =0.67, SD=0.37). We also detected no outliers (G=2.74, U=0. 90, p=0.38). The simplest   Fig. S2).

Compositional leukocyte profiles
Juncos from both populations were dominated by lymphocytes and heterophils ( Table 2). The first two PCs explained 88% of variance in leukocyte composition, with the former indicating negative eosinophil loadings and the latter indicating negative monocyte loadings (Table S1). The PERMANOVA showed that leukocyte composition model was again the most parsimonious (wi=0.38), but here the fixed effects explained 21% of the hematological variation (Table 1B). Estimated total WBC counts varied by population (χ 2 =5.73, p=0.02) and sex (χ 2 =4.96, p=0.03) and were higher in urban residents (β=0. 15, t=2.40) and in females (β=0.13, t=2.23; Fig. 3A). When we adjusted total WBC counts by hematocrit, the fixed effects from the simplest model (wi=0.38) explained 15% of hematological variation (Table 1C). However, the adjusted total WBC counts were no longer associated with population (χ 2 =2.29, p=0.13), although the effect direction was similar (β=0.09; Fig. 3B) and the prior sex difference remained (β=0.12, p=0.04). When analyzing   Table S2). Our post-hoc analysis showed this difference was driven by UCSD juncos having relatively fewer monocytes than LM juncos (Fig. 4, Table 2).

Blood parasites
Across our 75 individual juncos, we detected only Haemoproteus spp. in blood smears. This parasite was present in only two individuals, one from each population (LM: 1/36; UCSD: 1/39). We note that microscopy is not sensitive enough to conclude all other birds were uninfected [82].

Discussion
As many animals are decreasing their migratory propensity or are abandoning migration in response to urbanization [11][12][13][14][15], understanding how urban residency is associated with physiology is necessary to predict consequences for wildlife health, conservation, and pathogen spillover [34][35][36]. By focusing on the recent colonization of the UCSD campus by migratory juncos and comparing UCSD birds to juncos from a nearby altitudinal migrant population (LM), we show that urban residents only weakly differ in their HL ratios. However, a multivariate analysis suggested that monocyte counts may be lower in urban residents and that UCSD birds may have higher estimated total WBC counts. Urban residents and altitudinal migrants thus do not differ in hematological proxies of stress, and the observed hematological differences could instead indicate urban variation in either parasite exposure of immunological investment. One of our predictions was that urban resident juncos could have lower HL ratios than juncos from the LM population, as the adoption of sedentary behavior could reduce energy expenditures associated with seasonal altitudinal migrations [16,17]. The abundant food, water, and nesting resources in the UCSD environment could amplify this effect through minimizing additional energy demands associated with foraging and nest building [13,22]. However, HL ratios varied only weakly between populations and tended to be slightly higher in the UCSD population. This pattern agrees with prior work in which female juncos from these populations also did not differ in baseline plasma CORT [47]. Because HL ratios can detect the long-term impact of more mild stressors [52,53], our results suggest migrants that become resident are not more stressed and have similar energetic demands, at least during their breeding season. One parsimonious explanation for this pattern could be that the relatively short altitudinal migrations of LM juncos do not carry sufficient energetic costs to alter HL ratios. Comparisons of leukocyte profiles between UCSD juncos and mountain populations across their broad range in California (e.g., [41,42]) would be informative, because overwintering migrants in UCSD are likely from breeding populations across a wide latitudinal gradient [43]. We also sampled urban residents and juncos from the altitudinal migrant population during their breeding season, when animals are also faced with other reproductive stressors. Following the need for more studies across the annual cycle more broadly [83], sampling urban residents alongside migrants closer to the migration itself (e.g., prior to breeding in the high-elevation forest habitats and recently after arrival in the UCSD habitat) would help quantify the physiological costs of bird migration.
Although we observed no difference in HL ratios between LM and UCSD juncos, urban residents had higher estimated total WBC counts. We caution that the relationship between population and total WBC counts was weaker when we accounted for hematocrit; however, hematocrit data were only available from one year, and thus it is unclear whether this weaker association stems from temporal or hematocrit variation. Hematocrit was higher in LM juncos likely from increased metabolic requirements of altitudinal migration or high elevation [84,85]. As total leukocyte counts show inconsistent relationships with stressors [51], higher estimated total WBC counts of urban residents instead suggests stronger constitutive immunity or a greater likelihood of acute infections [51]. The former could arise if urban residents, through having greater energy reserves from more access to food, water, and nesting resources, could maintain a higher leukocyte reserve in preparation for fighting infections [22]. Alternatively, increased risk of parasite exposure in urban sedentary populations offers an alternative explanation. Shifts from migratory to resident behavior have been associated with greater parasite exposure owing to the loss of ecological mechanisms that reduce transmission [86,87]. In the junco system, a resident subspecies in the eastern United States (J. h. carolinensis) maintains higher prevalence of haemosporidians than migrants (J. h. hyemalis) during seasonal sympatry [88]. However, our microscopy screening for haemosporidians was inconclusive; only Haemoproteus was detected at low prevalence and was not sufficiently sensitive to detect lowintensity infections [82]. We instead used a multivariate analysis of leukocyte composition to provide additional insights [51]. Our PCA loaded strongly by monocytes, for which elevated relative counts were higher in LM juncos ( Table 2) and could indicate high bacterial infection risks and inflammation [51,61]. However, we note that population alone only explained a relatively small fraction of variance in leukocyte composition. We note that avian pox and ectoparasites are more prevalent in UCSD juncos [89]. Yet additional data on parasite prevalence or even diversity, such as that offered by advances in metagenomic approaches [90,91], would be necessary to assess whether such immunological variation between populations stems from parasite risk or urban resources.
Higher estimated total leukocyte counts of urban residents could signal the reduced energetic demands of non-migratory individuals and improved resources in urban habitats [22]. However, other metrics of constitutive immunity, such as microbicidal or haptoglobin activity, could better test how urban residency influences host defense [24]. Such work warrants further investigation, because potentially stronger immunity of urban residents could carry important implications for infectious disease dynamics. Prior mathematical models have examined the consequences of urban residency when migrants and residents are sympatric in the breeding season [92]. Further consideration of how susceptibility varies between urban residents and migrants during the breeding season, either in sympatry or allopatry, may further alter disease dynamics. For example, immunological benefits in urban residents could reduce or prevent transmission by homogenizing resource allocation to resistance [93,94]. However, if stronger immunity of urban residents translates into greater tolerance of infection, adoption of sedentary behavior could increase prevalence and make urban residents a source for infecting vectors [92].
As an increasing number of traditionally migratory species reduce migratory propensity or fully adopt sedentary behaviors in urbanized habitats, understanding whether such novel environments represent a net stressor or benefit to hosts is important to ultimately predict how infectious diseases and the health of migratory populations will respond to environmental change [12,13,15,92]. By using leukocyte profiles to assess physiological differences between recently established urban residents and a nearby altitudinal migrant population, we show that these two populations do not differ in their HL ratios but may differ in their estimated total WBC counts. Future work studying a broader range of urban resident and mountain populations would help capture a wider gradient of migratory distances, and sampling such populations more broadly across their annual cycles (including but not limited to closer to the timing of migration) would help quantify the physiological consequences of long-distance movements. Further assessments of parasite diversity and immunology will also be necessary to disentangle if urban residency is associated with greater exposure or stronger defense. Incorporating potential immunological effects of urban habitats into new modeling frameworks could have significant effects on infectious disease dynamics not only within residents but also for seasonally sympatric migrants.
for Science and Education through an interagency agreement between the U.S. Department of Energy and the Office of the Director of National Intelligence.

Competing interests:
We declare no competing interests.