Increased immune marker variance in a population of invasive birds

Study species

The Egyptian goose is native to Africa and was introduced to Europe in the twentieth century²⁷. Its native population is distributed on the sub-Saharan African continent and is one of the most common and wide spread African waterfowl species. Egg laying occurs throughout the whole year with a peak between late winter and early summer²⁸.

The neozootic population invades Europe eastwards starting from the Netherlands, where they were introduced as ornamental species to parks²⁷. It is now one of the most spreading neozootic bird species in Europe²⁴. From the 1980s Egyptian geese also invade Germany where its population size increased rapidly^29,30. The Egyptian goose is a resident (non-migratory / short distance migratory), monogamous, territorial bird species occurring as neozootic species in a variety of water habitats (e.g. streams, rivers, ponds, lakes,) in Europe³¹.

Sampling

Parasite prevalence and immunity of Egyptian geese from a native population in Namibia were investigated and compared to those of a currently spreading invasive population of the same species in Germany. In both regions, geese were sampled during ringing procedures (live trapped) or dissected after general pest control hunting (necropsy). Blood samples for immunological assays and serology exclusively originate from live trapped individuals whereas macro-parasite investigation was performed during necropsy. Micro-parasite investigation was performed in birds from both groups. Therefore, the resulting datasets are analysed separately (see method section: Statistical analysis) but a potential interplay between the different parasite prevalences and immune results is evaluated in the discussion.

Live trapping

Twenty-one Egyptian geese (9 male, 12 female) were live trapped in Namibia (22.35° S, 17.05° E) (native range) in February 2016. Additionally, data from a subset of 110 adult Egyptian geese from Germany (65 male, 45 female) investigated by Prüter et al.³² were included. German geese were sampled in the Rhine and Mosel areas (50.4° N, 7.6° E) (invasive range) in 2015 (n = 78) and 2016 (n = 32) in different months (supplementary data Table S1). Sex and reproductive status were recorded. Reproductive status was defined as breeding (e.g. guiding gosling, showing territorial behavior with a partner, having an egg-laying active cloaca) or non-breeding (e.g. not fulfilling criteria of breeding and/or being part of a non-family-flock). Numbers of breeding vs. non-breeding individuals can be found in Table S1. All Namibian birds were likely non-breeding individuals, which were sampled at a cattle feedlot were thousands of birds fly in to feed on the corn provided to the cattle. Blood was drawn from the vena metatarsalia plantaris superficialis using needles with a diameter of 0.06 mm for males and 0.04 mm for females. A fresh blood smear was prepared at capture and air dried. Blood samples were kept at 4–8 °C, centrifuged and sera were frozen in liquid nitrogen within eight hours after blood draw. Pharyngeal swabs were collected using sterile cotton swabs. Once field work finished, samples were transported to the Leibniz Institute for Zoo and Wildlife Research, Berlin, Germany and sera, blood clot and pharyngeal swabs were kept frozen at – 80 °C. Sampling in Germany was authorized by the Landesuntersuchungsamt Rheinland-Pfalz (G 15-20-005) and Landesamt für Natur, Umwelt und Verbraucherschutz Nordrhein-Westfalen (LANUV) (84-08.04.2015.A266). Permission to collect samples in Namibia was granted to GM and HK by the Ministry of Environment and Tourism (MET). Permission to export sample material from Namibia was granted by a MET export permit (No. 107513), and samples were transported to Germany in compliance with the Nagoya Protocol on Access and Benefit-sharing. All experimental procedures described in the materials and methods section were approved by the Internal Committee for Ethics and Animal Welfare of the Leibniz Institute for Zoo and Wildlife Research (permit #2014-11-03). All experiments were carried out in accordance with the approved guidelines of the Leibniz Institute for Zoo and Wildlife Research.

Necropsy

Additionally to live trapping, twenty-six free ranging Egyptian geese (17 male, 9 female) hunted during the autumn/winter season 2014/2015 and 2015/2016 in the North and West of Germany and twenty-seven Egyptian geese (11 male, 16 female), which were shot in February 2016 during regular pest control in Central Namibia were dissected. One of twenty-seven Namibian birds was live trapped and sampled before death and is thus included in both groups (live trapped and necropsy). Geese from Germany were kept frozen at – 20 °C after hunting until further analysis. Namibian geese were dissected immediately post mortem. During necropsy, ectoparasites, intestinal helminths and nasal leeches were collected. Additionally, pharyngeal swabs were taken for molecular analyses.

Parasitological and microbiological analysis

Both macro-parasites (ectoparasites, nasal leeches (Euhirundidae), intestinal helminths) and selected micro-parasites (blood parasites (Haematozoa), bacteria, viruses) of Egyptian geese from the two populations were characterized for community composition and prevalence (methods see Table S2). Hereafter we use the term “parasites” combining macro- and micro-parasites and only explicitly distinguish between the type of parasites when differences can be expected and/or occur.

During necropsy, wing and breast feathers were macroscopically checked for the presence of ectoparasites. The upper beak was cut open and macroscopically investigated for the presence of nasal leeches. Intestinal helminths were extracted from the intestine of the birds and were determined to the family level based on morphology³³. Additionally, blood smears of all live-trapped animals were investigated for the presence of blood parasites during white blood cell counts³⁴.

To compare with previously determined bacterial prevalence of adult German Egyptian geese³² (Table 2), the Namibian birds were screened for Mycoplasma spp. and Riemerella (R.) anatipestifer using conventional 16S rRNA-based PCR assays as described by Prüter et al.³². To verify the specificity of the Mycoplasma PCR assay, products with a clear band were further investigated by sequence analysis, following the procedure described by Prüter et al.³². Only samples with a clear sequencing result were designated positive.

The seroprevalence of antibodies (Ab) against Influenza A virus (IAV), Avian avulavirus 1 (AAvV-1) and West Nile virus (WNV) were determined³². For the detection of Abs against IAV, a commercial competitive enzyme linked immunosorbent assay (ELISA) was used following the manufacturer instructions (ID.vet, Grabels, France, Influenza A Antibody competition, FLUACA ver 0917DE). A commercial competitive ELISA for detection of Abs against AAvV-1 (Avian paramyxovirus 1; syn. Newcastle disease virus) was used according to the manufacturer protocol (ID.vet, Grabels, France, Newcastle Disease Competition, NDVC ver 0913 DE). Commercial competitive ELISA for Abs against Flaviviridae including WNV were applied following the manufacture protocol (ID.vet, West Nile Competition, WNC ver 1014-1P DE).

Immunological assays

Several eco-immunological tests were used to quantify both the cellular and humoral parts of the acquired and innate immune responses of Egyptian geese³⁵. Most of the methods are species-non-specific and have been used in a wide variety of free-living avian species, including different waterfowl^36,37,38. We quantified the amounts of different humoral (natural antibodies, complement, lysozyme and haptoglobin) and cellular (monocytes, heterophils, eosinophils and basophils) effectors of innate immunity. For adaptive immunity we measured the total immunoglobulin Y (IgY) concentration and the number of lymphocytes³⁶. Sample sizes (n) for each assay were dependent on the total amount of serum available from each individual and therefore differ among the tests (Table 1).

Immunoglobulin Y

Total IgY, the avian equivalent to mammalian IgG, was measured using a sensitive ELISA with commercial anti-chicken antibodies^38,40. Ninety six-well high-binding ELISA plates (82.1581.200, Sarstedt) were coated with 100 µl of diluted serum sample (2 samples per bird 1:16,000 diluted in carbonate–bicarbonate buffer) and incubated first for 1 h at 37 °C and then overnight at 4 °C. After incubation, the plates were washed with a 200 µl solution of phosphate buffer saline and PBS–Tween, before 100 µl of a solution of 1% gelatine in PBS–Tween was added. Plates were then incubated at 37 °C for 1 h, washed with PBS–Tween and 100 µl of polyclonal rabbit anti-chicken IgY conjugated with peroxidase (A-9046, Sigma) at 1:250 (v/v) was added. Following 2 h incubation at 37 °C, the plates were washed again with PBS–Tween three times. After washing, 100 µl of revealing solution [peroxide diluted 1:1000 in ABTS (2,20-azino-bis- (3-ethylbenzthiazoline-6-sulphonic acid))] was added, and the plates were incubated for 1 h at 37 °C. The final absorbance was measured at 405 nm using a photometric microplate reader (μQuant Microplate Spectrophotometer, Biotek) and subsequently defined as total serum IgY levels⁴¹.

Lysozyme

To measure lysozyme concentration in serum, we used the lysoplate assay³⁷: 25 μl serum were inoculated in the test holes of a 1% Noble agar gel (A5431, Sigma) containing 50 mg/100 ml lyophilized Micrococcus lysodeikticus (M3770, Sigma), a bacteria which is particularly sensitive to lysozyme concentration. Crystalline hen egg white lysozyme (L6876, Sigma) (concentration: 1, 1.25, 2.5, 5, 6.25, 10, 12.5, 20 and 25 µg/ml) was used to prepare a standard curve for each plate. Plates were incubated at room temperature (25–27 °C) for 20 h. During this period, as a result of bacterial lysis, a clear zone developed in the area of the gel surrounding the sample inoculation site. The diameters of the cleared zones are proportional to the log of the lysozyme concentration. This area was measured three times digitally using the software ImageJ (version 1.48, http://imagej.nih.gov/ij/) and the mean was converted to a semi-logarithmic plot into hen egg lysozyme equivalents (HEL equivalents, expressed in μg/mL) according to the standard curve⁴².

Haemolysis–haemagglutination assay

The levels of the natural antibodies and complement were assessed by using a haemolysis–haemagglutination assay as described by⁴³ adjusted to the limited volume of serum. After pipetting 15 μl of serum into the first two columns of a U-shaped 96-well microtitre plate, 15 μl sterile PBS were added to columns 2–12. The content of the second column wells was serially diluted (1:2) until the 11th column, resulting in a dilution series for each sample from 1/1 to 1/1024. The last column of the plate was used as negative controls, containing PBS only. Fifteen μl of 1% rabbit red blood cells (supplied as 50% whole blood, 50% Alsever’s solution, Envigo) suspension was added to all wells and incubated at 37 °C for 90 min. After incubation, in order to increase the visualisation of agglutination, the plates were tilted at a 45° angle at room temperature. Agglutination and lysis, which reflect the activity of the natural antibodies and the interaction between these antibodies and complement^43,44, was recorded after 20 and 90 min, respectively. Haemagglutination is characterised by the appearance of clumped red blood cells, as a result of antibodies binding multiple antigens, while during haemolysis, the red blood cells are destroyed. Titres of the natural antibodies and complement were given as the log2 of the reciprocal of the highest dilution of serum showing positive haemagglutination or lysis, respectively^43,45.

Haptoglobin

We measured haptoglobin concentrations with a commercial kit (TP801, Tri-Delta Diagnostics, Inc.) following the instructions of the manufacturer. Haptoglobin concentrations (mg/ml) in undiluted serum samples were calculated according to the standard curve on each plate³⁶.

White blood cell counts

To count leucocytes, blood smears were prepared, air-dried and stained using Giemsa- and May-Grünwald staining. Smears were examined at 1,000 fold magnification with oil immersion and the relative number and types of leucocytes were assessed by counting 100 leucocytes. The number of white blood cells of different types was expressed per 10⁴ erythrocytes⁴⁵.

Statistical analyses

Parasite prevalence

To investigate potential differences in the prevalence of parasites between native and invasive Egyptian geese, we used Fisher’s exact tests because relatively low sample sizes of dissected animals did not allow to perform multivariate analysis.

Immunity

The means and variances of the different immune effectors were compared between the invasive and native Egyptian geese populations. To this end, we used an extension of commonly applied linear models. Linear models assume that the response variable y is a function of a linear combination of n predictor variables x with coefficients c₀,..,c_n and an error ε:

where ε is the so-called residual variance which captures all the variation in the response variable that is not explained by the predictors. In linear models ε is assumed to be normally and independently distributed around zero. An additional usual assumption is that the variance of this distribution is a constant ({upsigma }_{0}), i.e.:

which corresponds to the assumption of normally distributed residuals with homogeneous variance.

Thus, the estimated effects of the predictors c₁,..,c_n only describe changes in the mean of the response variable, but not in the variance around that mean.

In our analysis, models were used in which the variance was allowed to be a linear function of some predictor variables z (which might be the same or different from the predictors x of the mean in Eq. 1), i.e.:

Thus, we were able to estimate simultaneously the effect and respective p-values of predictors x upon the variation in the mean of the response variable (Eq. 1) and also the effects and respective p-values of predictors z upon the residual variation around that mean (Eq. 3).

In order to appropriately capture all the potential variation in the response variables we used linear mixed-effects models (LMMs), which in addition to fixed effect predictors in Eq. (1) also included a random effect as a predictor of the mean. However, for enhanced clarity random effects were omitted in the equations above.

Different immune effectors were used as response variables (Table 1). As predictors for the mean sex (male vs. female), reproductive status (breeding vs. non-breeding) and invasion status (native vs. invasive) were included as fixed effects and month of sampling as a random effect. In this way, we control for potential confounding effects of breeding status between the two populations. As predictors for the variance, we included invasion status (native vs. invasive), sex and reproductive status, which allowed us to test our prediction that the variance in immune effects is higher among invasive individuals compared to native individuals.

Some of the immune effectors were transformed (see tables supplementary data S3, S4, S5) to ensure normality of residuals. For haptoglobin we were not able to perform a transformation that ensured normality, because of the high proportion of values below the detection threshold. To account for this, we performed a generalized linear mixed model (GLMM) with a binominal error distribution and with a binary response variable (haptoglobin being either above or below the detection threshold). In this model, it was necessary to constrain the error variance to a fixed value. Thus, for haptoglobin we were only able to test for a change in mean but not for a change in variance. In addition to analysing total leucocytes, different leucocyte subtypes were analysed separately.

The LMMs and GLMMs were implemented using the R package glmmTMB version 0.2.0⁴⁶. To test for differences in residual variance we used the option disformula in the function glmmTMB. Potential collinearity of predictors was tested by calculating variance inflation factors using the R package car version 2.1-6⁴⁷. All statistical analyses were performed using R version 3.3.2⁴⁸.