Animal trait variation at the within-individual level: erythrocyte size variation and malaria infection in a tropical lizard

Study site and species

To answer these questions our data consists of a subset of individuals from an ongoing long-term study on the dynamics of the Anolis–Plasmodium host–parasite system at El Verde Field Station, in the Luquillo Experimental Forest in Puerto Rico (18.3213° N, 65.8194° W, WGS 84; 357.9 m elev) (Schall, Pearson & Perkins, 2000; Otero et al., 2019). The surveys took place during the summer season of 2015 and the winter season of 2016. The Institutional Animal Care and Use Committee protocols from the University of Puerto Rico provided the permits for the handling of lizards, IACUC (01005-01-09-2015).

Out of the seven species of anoles inhabiting El Verde, we selected Anolis gundlachi as our study species because it is the member of the anole community with the highest prevalence of Plasmodium parasite infection in the site. Between 2015 and 2017 the probability of infections was 0.10–0.19 for males, and 0.06–0.12 for females (Otero et al., 2019). This anole is a medium-sized montane lizard, part of the trunk-ground ecomorph, with a snout-vent length (SVL) range of 42–72 mm (Reagan & Waide, 1996). This species can be infected by three Plasmodium species: P. azurophilum, P. floridense, and P. leucocytica. We analyzed infections by P. azurophilum because this is the most common blood parasite at the site, accounting for 60–80% of infections (Schall, Pearson & Perkins, 2000; Otero et al., 2019). Also, P. azurophilum is the only member of this Plasmodium community known to destroy lizard erythrocytes (Schall, 1996). Co-infections are infrequent. Still, we restricted our analysis to single infected individuals to ensure our results were not biased by potential interacting effects of multiple blood parasites (see below).

We captured anoles by hand or using the lasso technique while the lizards were perching on tree trunks, branches, or on the ground. The lizards were briefly kept in individual bags and moved to the laboratory where they were processed to identify sex, measure the snout-to-vent length (SVL) and weight, and obtain a blood sample via toe clip (Schall & Vogt, 1993). We used this blood sampling method because it provides enough blood to accurately diagnose infection by microscopy and PCR (see below). Also, toe clipping provides a permanent mark that prevents sampling of an individual more than once. Elevated levels of corticosterone have been shown to alter the hematopoietic cell composition and immune response in other animals; however, this method has less impact on corticosterone levels than implanted Passive Integrated Transponder (PIT) tags (Langkilde & Shine, 2006). All lizards were released back to the general area in which they were found within 24hrs.

Infection diagnostics

We diagnosed Plasmodium infection using microscopy and verifying through polymerase chain reaction (PCR). Combining both methods assures the most accurate classification of infection status. For microscopy diagnosis, we made thin-blood smears on glass slides. The thin blood smears were fixed with 100% methanol for a minute and stained with Giemsa at a pH of 7.5 for 50 min. We examined the blood smears at 1000×magnification using a Nikon Eclipse E2000 light microscope for 6–10 min (Otero et al., 2019). We classified the Plasmodium species detected by morphological traits (Telford, 2009). We considered a lizard infected if we detected through microscopy at least one cell invaded by P. azurophilum, and if PCR confirmed infection by positively amplifying Plasmodium. Similarly, we considered a lizard non-infected if we could not visually detect an invaded cell using microscopy and the PCR did not amplify Plasmodium. To ensure the most accurate classification of infection status, if there was a mismatch between the slide blood smear and the PCR result the individual was removed from the analysis.

Drops of lizard blood were spotted on Whatman filter paper and stored at −20 °C in sealed bags with silica gel for further PCR testing. DNA was extracted from blood-spot filters using the DNeasy Extraction kit with some modifications to the manufacturer’s instructions. To extract the blood from the filter paper, 180 µL of Buffer ATL and 20 µL Protease K was added. The samples were mixed and incubated at 56 °C overnight with shaking. The next day we added 200 µL of Buffer AL to the sample, mixed by thoroughly shaking for 15 s, and incubated at 56 °C for 15 min. Then, 200 µL ethanol was added to each sample. To elute DNA, we added 50 µL to 75 µL Buffer AE to each sample and incubated it for 5 min at room temperature. The extracted DNA(eluate) was stored at −20 °C. A partial fragment of the mitochondrial cytochrome b gene was amplified using nested PCR (nPCR). The nPCR was done with Taq PCR Master Mix kit (Qiagen cat 201445) under the conditions described by Perkins & Schall (2002) with minor modifications. Briefly, an outer reaction using primers DW2 F 5′- TAA TGC CTA GAC GTA TTC CTG ATT ATC CAG - 3′ DW4 R 5′- TGT TTG CTT GGG AGC TGT AAT CAT AAT GTG - 3′ and 2 µL of genomic DNA was subject to thermal cycling program included an initial denaturation step at 94 °C for 4 min followed by 35 cycles of 94 °C for 20 s, 60 °C for 30 s, and 72 °C for 1.5 min. The nested reactions were done with 1 µL of the previous PCR product and primers DW1 5′-TCA ACA ATG ACT TTA TTT GG-3′ and DW6 5′-GGG AGC TGT AAT CAT AAT GTG-3′ under initial denaturation step at 94 °C for 1 min followed by 40 cycles of 94 °C for 20 s, 50 °C for 20 s, and 72 °C for 1,5 min and then 72 °C for 7 min. Positive and negative reaction controls were included. Amplification products were resolved in a 2% agarose gel stained with gelStar™ (Lonza, cat # 50535) and visualized under ultraviolet light. The primers only amplified the malarial parasite DNA. PCR complemented diagnosis by helping avoid false negatives resulting from individuals with low parasitemia that may be missed by the slides.

Erythrocyte image capturing and measurement

We photographed 10 microscope optical fields for each lizard’s thin blood smear (Fig. 1). Optical fields were selected at random. Fields in which less than 10 mature erythrocytes were measurable were discarded. We measured only mature erythrocytes in each picture and disregarded other hematocytes because we wanted to estimate variation in erythrocyte size independent of stage. We took the pictures at 40x power using a Nikon DS-Fi2 camera head and a Nikon DS-U3 microscope camera controller. This setting provided high-quality pictures that allowed precise measurements of the erythrocyte area. Since measuring erythrocyte size in clumps of cells is challenging and may result in inaccurate measures, we considered cells to be measurable if there was at least 0.070 µm of space between cells, and if they were not damaged during the smear making process. We also discriminated between infected and non-infected erythrocytes. We discarded 19 Plasmodium-invaded cells (out of a total of 50,241 measured erythrocytes) from the analysis because erythrocyte sickling rate has been observed to differ between Plasmodium stages (Roth et al., 1978). By restricting our analysis on mature erythrocytes that were not invaded by the parasite we segregated the processes allowing us to assess the global effects of Plasmodium on WIV independent of local effects.

We measured the area of each erythrocyte in each optical field with the image processing software Image J (Schneider, Rasband & Eliceiri, 2012). We set the scale to 50.0 µm and the image type to 8-bit grayscale to allow segmentation by pixel color intensity, also known as image thresholding. The image threshold was manually adjusted for each image due to differences in the sharpness, quality of staining, and brightness of images. We selected the optimum threshold that allowed us to minimize measurement error by decreasing the amount of whitespace or non-erythrocyte area inside the measurement polygon. We stored measurements of erythrocyte per picture automatically on individual spreadsheets with their unique ID as the file name. We also took lower resolution pictures of the selected erythrocytes per picture which had the order in which erythrocytes were measured. This allowed us to keep a permanent record of which measurement belonged to which erythrocyte for verification purposes while maintaining our original pictures intact to ensure study reproducibility.

Data analyses

We conducted a prospective power analysis to estimate an appropriate sample size to test our hypotheses with statistical power >0.8. We analyzed pilot data (n = 21) to estimate the variance and mean coefficient of variation of the population. To estimate the effect size, we used Cohen’s d defined as, $d=\frac{{\bar{x}}_{inf}-{\bar{x}}_{ninf}}{S{D}_{pooled}}$, where ${\bar{x}}_{inf}=12.59$ (infected CV) and ${\bar{x}}_{ninf}$ =12.02 (non-infected CV). We set a significance level of α = 0.05 and a power of 0.8 as it is conventionally done in studies of ecology and evolution.

We quantified within-individual erythrocyte size variation of A. gundlachi using the coefficient of variation (CV), such that:

${\displaystyle C{V}_i=\frac{s_i}{\overline{x_i}}\times 100,}$ where ${\bar{x}}_i$ represents the sample mean and s_i the sample standard deviation of the size of mature erythrocytes of lizard i. The coefficient of variation (CV) is a measure of relative variation allowing comparison among lizards while controlling for mean erythrocyte size (Donnelly & Kramer, 1999). This measure is commonly used in similar studies to quantify within-individual variation (Biernaskie, Cartar & Hurly, 2002; Herrera, Medrano & Bazaga, 2015; Langkilde & Shine, 2006).

We modeled lizard body condition index (BCI) as a function of infection status, and sex and season as controlling variables using a multiple linear regression. Although reproduction data would be ideal to test for the negative consequences of infection on individual fitness, collecting breeding data is challenging and often infeasible in many natural systems including anoles (Jakob, Marshall & Uetz, 1996). Here we use BCI as a proxy for fitness as the next best alternative. We calculated BCI as the residuals of the linear regression of log10 weight (g) explained by log10 lizard SVL (mm) (Cox & Calsbeek, 2015). Positive residuals suggest a better body condition relative to the population average (Schall, Pearson & Perkins, 2000). Male A.gundlachi BCI has been shown to be higher than females and BCI also responds to seasonal variation in resource availability (Schall, Pearson & Perkins, 2000). Previous studies validate the use of BCI to represent the energetic state of animals (Ardia, 2005; Schulte-Hostedde et al., 2005). Still, a study found that BCI is weakly related to survival in Anolis sagrei (Cox & Calsbeek, 2015) and a meta-analysis found small, negative average effect sizes between BCI and parasite infection that was stronger in laboratory settings (Sánchez et al., 2018). In an experimental setting, mass/SVL was found to be the most accurate estimator for the mass of chemically extracted fat for a reptile among five compared estimators (Sion, Watson & Bouskila, 2021). Therefore, while BCI is the best approximation to fitness given our data constraints we are careful interpreting the results given its mixed support as a proxy for fitness (Wilder, Raubenheimer & Simpson, 2016).

To test for a relationship between P. azurophilum infection and WIV in erythrocyte size, we modeled infection status as a function of CV. Previous studies show that larger males in the summer season have a higher probability of getting infected (Otero et al., 2019). Therefore, we included sex, season, and body size(SVL) as controlling covaraites. We modeled this relationship using a generalized linear model with a Bernoulli distribution and logit link function.

To test if individuals with higher WIV have improved fitness, we modeled BCI as a function of CV in erythrocyte size with sex and season as controlling covaraites using a multiple linear regression. These controlling covariates were included because previous studies in this system found that body condition is higher in males and during the summer season where food resources for lizards are more abundant (Schall, Pearson & Perkins, 2000).