Haemosporidian Infection Risk Variation Across an Urban Gradient in a Songbird

Abstract

Urbanization is a significant source of inter- and intra-city environmental variation and is associated with declining avian population sizes, with a shift towards more homogeneous communities that consist of large populations of the same species. However, whether this shift extends to urban disease ecology and related parasite communities requires further examination. By comparing the diversity of two related parasite genera (largely host-generalist Plasmodium and largely host-specialist Haemoproteus) and infection status of dark-eyed juncos (Junco hyemalis) across an urbanization gradient in California, we can determine how broad urban-associated land use changes and localized habitat composition correlate with pathogen communities. Additionally, by examining vector abundance responses, we can begin to assess broader impacts on urban disease transmission and ecology. We report fewer birds were infected with Haemoproteus in urban habitats, with a larger presence of host-generalist lineages, suggesting urbanization increases homogenization of host-specialist pathogens. Unsurprisingly, the largely host-generalist pathogen, Plasmodium, showed no correlation with urbanization, but infections increased with rainfall. Local habitat characteristics had limited effects on Plasmodium infection status, but biotic characteristics, including wing chord length and human presence, were associated with Plasmodium infections. Lastly, Culex tarsalis, an important vector for Plasmodium and zoonotic pathogens, was the only vector to also increase in abundance in response to rainfall. Our results show that broad land use changes associated with urbanization decrease avian parasite biodiversity and highlight localized abiotic and biotic habitat characteristics that may reduce infection prevalence.

Keywords: avian infectious disease ecology; dark‐eyed juncos; urban ecology; vector ecology.

FIGURE 1

Map of the study sites and the corresponding infection prevalence of Haemoproteus (black) and Plasmodium (gray) at each site are shown. Mixed infections were not observed and were therefore not included. Sample sites include Angeles National Forest (ANF, n = 55), Santa Monica Mountains (SMM, n = 18), Los Angeles parks (LA, n = 57), Occidental College (OCC, n = 20), San Francisco State University (SFSU, n = 6), University of California, Los Angeles (UCLA, n = 262), University of California, Santa Barbara (UCSB, n = 35), and University of California, San Diego (UCSD, n = 38).

FIGURE 2

Infection prevalence across study sites. The mean infection prevalence for “non‐urban” (green) and “urban” (gray) sites of both Plasmodium and Haemoproteus are shown. Bars represent 95% confidence intervals. For Haemoproteus, letters (A and B) represent significance: Same letters indicate no significant differences, and different letters indicate statistically significant differences.

FIGURE 3

Generalized linear regression with 95% confidence intervals (shaded light gray) of parasite–host specificity (S_TD) along an urbanization gradient (Built‐Up Index) for (a) Plasmodium and (b) Haemoproteus. Scatter plot points are jittered along y‐axis for each integer‐value. Low S_TD indicates parasites have only been reported in closely related species. Color differences, “non‐urban” sites in green and “urban” sites in black, are provided for improved visualization but were not included as variables in model.

FIGURE 4

Plotted generalized linear mixed effect model and 95% confidence intervals (shaded light gray) for (a) Plasmodium and (b) Haemoproteus. Points are jittered along the y‐axis. Color differences, “non‐urban” sites in green and “urban” sites in black, are provided for improved visualization but were not included as variables in model.

FIGURE 5

Random Forest classification of local habitat characteristics within 50 m of a capture site and host morphological characteristics as predictors of presence/absence of Plasmodium infections. Variable importance was calculated via the corrected Gini impurity index. Significance was set at the absolute value of the lowest Gini impurity index value (dotted red line). Variables that exceeded this threshold were considered significant.

FIGURE 6

Classification tree of Plasmodium‐infected second‐year dark‐eyed juncos. Variables were set on the basis of significance from Random Forest classification. Individuals were classified as infected (1) or not infected (0) using a threshold set to > 50% chance. At each final leaf, the proportion of samples that met the decision criteria and were not infected is shown on the bottom left, and the proportion that were infected in Plasmodium are shown on the bottom right. The table count and trash count refers to absolute values, “Wing” refers to wing chord length (mm).

FIGURE A1

Haemoproteus (dashed line), Plasmodium (dotted line), and combined haemosporidian infection prevalence (solid line) per site from 2021 to 2023. Sample sites include Angeles National Forest (ANF, n = 55), Santa Monica Mountains (SMM, n = 18), Los Angeles parks (LA, n = 57), Occidental College (OCC, n = 20), University of California, Los Angeles (UCLA, n = 262), University of California, Santa Barbara (UCSB, n = 35), and University of California, San Diego (UCSD, n = 38). San Francisco (0% infection prevalence, n = 6) was only sampled once (2023) and therefore not included in this figure.

FIGURE A2

Cladogram of the haemosporidian lineages. Only branches with > 80% posterior probability based on Bayesian inferences are labeled. Reference sequences are shown in black with associated GenBank accession numbers in parentheses. * denotes lineages found in our study. Sequences in orange represent lineages exclusively found in urban habitats, green represents lineages exclusively found in non‐urban habitats, and purple represents lineages found across both habitat types.

FIGURE A3

Generalized linear (mixed) models of the abundance of each vector species from 2021 to 2023 in response to the degree of urbanization (Built‐Up Index) with higher values indicating more urban habitat. The gray dots represent the log of the abundance. The red lines show the predicted trendline based on the appropriate model. The samples were sorted by species when possible and grouped by genus when species‐level identification was not confirmed.

FIGURE A4

Generalized linear (mixed) models of abundance for each species of vector from 2021 to 2023 in response to cumulative monthly precipitation averaged 50 m around each trap site. The blue dots represent the log of abundance. The red lines show the predicted trendline based on the appropriate model. The samples were sorted by species when possible and grouped by genus when species‐level identification was not confirmed.