Seasonal comparisons

Our study design involved the sampling of zebra over five seasons (two wet, three dry), resampling the same individuals as many times as possible over those five events to control for individual variation. We compared several immune and pathogen measures between seasons. To control for nutritional status and resource allocation, we measured total white blood cell (WBC) counts and hematocrits (HCT). Total WBC counts are a broad indicator of how much a host has invested in immune resources [39]. Hematocrit is a measure of percent of red blood cells per unit of blood, with a higher hematocrit often reflecting a higher nutritional state [40]. Eosinophils are white blood cells often important for eliminating helminths, are quite specific for macroparasite infections, and have been shown to be a good measure of Th2 responsiveness [11],[41]-[43]. Monocytes are white blood cells that are recruited and activated by interferon-gamma (IFN-γ) in Th1 immune reactions, and were used as a measurement of potential Th1 immune activity [11]. Immunoglobulin E (IgE) is an antibody isotype mediated by the Th2-associated cytokine interleukin 4 (IL-4) and important, and specific, in fighting against helminths [11]. Immunoglobulin G, subtype b (IgGb) is the most prevalent antibody isotype in equine serum [44]. While IgGb is important in the protective response against intracellular pathogens, suggesting a Th1-associated response, Hooper-McGrevy et al. [45] found that IgGb is likely a Th2-related antibody in horses. Antibodies against the protective antigen (PA) toxin component of anthrax have been shown to be essential for adaptive protection against anthrax [35],[36],[46]. As a bacterial infection, anthrax provokes primarily a Th1 type immune response [37],[47]. IFN-γ is a key Th1 cytokine that activates macrophages early in infection. IL-4 is an important Th2 cytokine that induces antibody isotype switching to IgE and augments recruitment of eosinophils in response to the presence of helminths [11],[48].

We first compared data from only the first sampling events for animals between the two season types. This ensured that we were only comparing unique individuals, and their population-level effects, between seasons. We then did pairwise comparisons across the same animals, between seasons, for first and second captures only to examine individual level effects; this allowed us to account for individual-level variation that otherwise might have obscured patterns at the population level. We found highly statistically significant differences between seasons for GI parasites and eosinophil counts for both individual and population-level comparisons (Figure 1A and D; Table 1). IgGb titers (Figure 1G) were nearly significantly different between seasons at a population level, and were significantly different between seasons for individual-level comparisons; thus, our paired comparisons aided in elucidating patterns otherwise obscured by individual variation. Similarly, ectoparasite counts were only significantly different between rain groups for individual-level comparisons (Figure 1B, E, and F; Table 1). In addition, we found that both WBC counts and HCT were significantly higher in the wet season for both population-level (p = 0.007, p = 0.003, respectively), and for individual-level comparisons (p = 0.001, p = 0.020, respectively).

We found no statistically significant differences when comparing IL-4 transcript ratios between seasons for all animals (Wilcoxon U = 232, N ₁ = 17, N ₂ = 36, p = 0.145; 4.6-fold increase in wet season), or for first or paired captures (Figure 1H; Table 1). However, we found a mean 15.2-fold increase in the mean transcript in the wet season in our population-level analysis. We found no statistically significant differences when comparing IFN-γ transcript ratios between seasons for all animals (Wilcoxon U = 245, N ₁ = 16, N ₂ = 37, p = 0.317; 2.4-fold increase in dry season), or for first or paired captures (Figure 1I; Table 1). Comparing mean ratios in the different seasons, however, we found a 3.0-fold increase in transcript in the dry season in our population-level analysis. Thus, we conclude that there is some evidence showing an increase in IL-4 production during wetter times, and for an increase in IFN-γ production during drier times but more data are needed.

Parasite coinfection models

In the best fitting PA model, as rainfall increased, positive titer prevalence increased, and as eosinophil count increased, titer prevalence increased (Table 2). The interaction term of rainfall with eosinophil count, however, was more highly statistically significant and negative, indicating that animals experiencing higher eosinophil counts were less likely to mount an anti-anthrax immune response than were animals with lower eosinophil counts during the wet (i.e. anthrax) season. Figure 2 illustrates that this negative relationship between eosinophils and anti-PA titer is strongly significant in the wet season (Figure 2A), whereas the significantly positive relationship between anti-PA titer and eosinophil count observed in the PA GEE model only holds for the dry season and at a less steep slope than that for the wet season relationships (Figure 2B).

In the best fitting ectoparasite model (Table 2), decreased rainfall significantly predicted higher ectoparasite counts, and having more GI parasites significantly predicted higher ectoparasite loads. Relationships for all three pathogen models are illustrated in Figure 3B.

Immune interaction models

Prior rainfall affected all immunity measures, with increased rainfall associated with increased eosinophils, monocytes, IgE, and IgGb titers (Figure 3C). GI parasite counts were significantly, negatively associated with eosinophil counts, as was observed in the GI parasite GEE model (Table 2). While GI parasite counts and ectoparasite counts were not statistically significant in the IgE model, they both showed a potentially negative association with IgE titer. IFN-γ was significantly negatively associated with animals that had both high eosinophil counts and GI parasite infection intensities (p = 0.000), and with animals that had both a high IgE titer and increased GI parasite infection intensities (p = 0.001). Rainfall and eosinophil counts both significantly positively predicted monocyte counts; however, the interaction between rainfall and eosinophils, was significantly negatively associated with monocyte counts, indicating a complex interplay between these three variables. IgE and IgGb antibody titers both significantly negatively predicted monocyte counts, and monocytes were significantly, negatively correlated with IgGb titers in the IgGb model. Relationships for immunity models are illustrated in Figure 3C.

Seasonal immune resource allocation

The seasonal patterns in these immune parameters may partly reflect seasonal nutritional differences. Higher HCT in the wet season likely reflects a higher nutritional state, especially in light of the fact that dehydration often results in an elevated HCT [40]; the fact that HCT in the wet season was still significantly higher than that of the more dehydrated animals in the dry season makes it likely that overall red blood cell numbers were even higher during periods of more rainfall. Higher WBC in the wet season may also reflect this higher plane of nutrition, as may the higher total numbers of eosinophils and monocytes, and higher antibody titers. However, several of these immune parameters had complex relationships with rainfall, pathogen markers, and other immune factors, and thus simple seasonality and nutritional effects cannot completely account for these patterns.

Several of the parasite and immune factors provide evidence that there may exist Th1-Th2 tradeoffs in zebra hosts. Demonstrating such trade-offs in a natural system is powerful in that it helps to ground truth and support similar work previously conducted in controlled laboratory settings. Our work in this natural system also allowed us to examine whether these trade-offs in nature can act in a temporal dimension, as we measured differential immune resource allocation in the wet versus the dry seasons.

As we expected from previous work [22], GI parasite infection intensities were significantly higher in wetter seasons (Figure 1A), likely due to increased survival of environmental life stages with increased moisture [49] and adult worms delaying egg production until it is environmentally advantageous to produce eggs [50]. Eosinophil counts were also significantly higher in the wetter seasons than in the drier ones, consistent with hosts, at a population level, fighting against new GI parasite infections during wetter conditions [51]-[53] (Figure 1D). Other work in ENP has found that zebra shed significantly more strongyle larvae during wetter seasons than in drier ones, lending further evidence of an overall, population-wide active immune response against new infections during the wet season [W.C. Turner, unpublished data]. Individually, we found a negative relationship between GI parasites and eosinophil counts (Figure 3A and B). Previous studies in wildlife have found a similar pattern, with a strongly negative association between GI helminth FECs and eosinophil counts, even when taking into account other host factors [16],[52],[54]. These findings suggest that individuals that are predisposed toward mounting stronger Th2 responses (i.e. higher eosinophil counts) are more resistant to parasite infections, and thus exhibit lower FECs.

While total monocyte counts were significantly correlated with increased rainfall, they were significantly, negatively correlated with a rainfall-eosinophil interaction effect (Table 2). This perhaps indicates that, while nutritional effects of the wet season may have increased host allocation to immune resources including monocytes, a trade-off between Th2-type eosinophil responses and Th1-type monocyte counts may be occurring. The fact that monocytes were significantly, negatively correlated with the Th2-type immune factors IgE and IgGb titers further supports the hypothesis of a trade-off between these different arms of immunity.

Ectoparasite counts were significantly higher in the drier seasons than in the wetter ones in paired samples but not in unique animal comparisons (Figure 1; Additional file 1: Table B1); this provides evidence that individual susceptibility, likely mediated by immune responses and/or immunomodulation, was at least partly responsible for seasonal differences in ectoparasite infection intensities. Previous studies have found that hard-bodied ticks in most systems prefer higher humidity and intermediate temperatures and are at their lowest numbers on hosts in very hot, dry seasons [55]. As ticks were actually less prevalent on ENP zebras during what should have been their preferred season, this lends more evidence that immunomodulatory conditions influenced tick infestation of hosts.

IgGb antibody titers were significantly greater in wetter seasons, but only for resampled animals, indicating that there are individual differences in immune allocation, immunomodulation, or both (Figure 1; Table 1). While IgGb is likely a Th2-type antibody subtype in equids [45], it is also the most prevalent antibody isotype in equine serum and is potentially involved in protection against other types of pathogens as well [44]. Thus, while changes in IgGb titers may in part reflect seasonal changes in GI parasite infection intensity in ENP zebras, they may also be affected by other factors such as seasonal changes in host nutrition or reproduction.

IL-4 is an instrumental component in protection against or the clearing of many helminth infections, with levels peaking soon following new parasite infections and falling a couple of months after infection attenuation [5],[11],[56]. While we had a small sample size and non-significant seasonal comparisons, we detected a 15-fold increase in mean IL-4 signal in resampled animals while these animals were experiencing their highest GI parasite infection intensities, consistent with previous findings [5],[56]. Similarly, we observed drops in mean IL-4 seen in the dry season as GI infection intensities waned. IFN-γ significantly decreases in the face of increased GI helminth infection intensity and increase after anthelminthic treatment [16]. Though our sample size was small and seasonal comparisons were nonsignificant, our approximately twofold decrease in mean IFN-γ signal in wet seasons when hosts were experiencing peak new GI helminth infections compared to in the relatively parasite-depleted dry season supports previous findings [15],[16]. In addition, the significantly negative association between IFN-γ signal and animals with high GI parasite infections in conjunction with either high eosinophil counts or IgE titers lends support for opposing patterns between this Th1-type cytokine and parasites and parasite related immune factors, as would be expected in a parasite driven Th2-type immune skewing.

Thus, the wet season eosinophil peak, combined with significantly higher IgGb antibody titers and preliminary evidence for increased IL-4 and decreased IFN-γ signaling at the same time provides evidence that host immune resources may be skewed toward a Th2-type response in the wet season. The addition of significant inverse relationships between monocytes and IgGb and IgE antibody titers, and between monocytes and eosinophils during times of higher rain, support this hypothesis. The timing of these immune increases in conjunction with the timing of GI helminths infections that are known to affect immunity in a Th2 positive manner suggests that these immune patterns are related to wet season new GI helminth infections. The significant drop in eosinophil counts and IgGb titers in the dry season, coupled with preliminary evidence of increased IFN-γ signaling at that time supports the hypothesis that immunomodulation toward a Th2 response by GI parasites may be attenuated when free-living GI parasite stages are less capable of survival and infecting new hosts.

Correlates of coinfection with immunomodulatory effects

Previous work in ENP has demonstrated that ungulates ingest significantly more soil in the wet season, likely leading to increased exposure to B. anthracis at that time [27]. In support of this, we found that anti-PA titer prevalence was significantly, positively related to rainfall in the PA GEE (Figure 3B), likely reflecting this increased exposure to B. anthracis during the wet season, with a subsequent immune response if hosts are exposed to sublethal doses [32]. As we previously found evidence that mean time for negative seroconversion in titer-positive animals is less than six months [32], we are confident that the rainfall-correlated immune signatures represent recent, seasonal B. anthracis exposure. This immune signature, coupled with the simultaneously significantly increased rates anthrax deaths supports the assertion that differential exposure to anthrax at least partly drives its seasonality in this system.

However, the wet season increase in Th2-type immune resources observed in our system, in correlation with the timing of peak anthrax incidence, suggests that hosts may also experience seasonally increased anthrax susceptibility shortly following periods of peak GI helminth infection intensity due, at least in part, to immunomodulatory effects and immune tradeoffs. Th2-type skewing by helminths has been found to have significant consequences for hosts fighting against concurrent microparasitic infections [6],[57]. In support of this assertion, we also found indirect evidence (through eosinophils) that decreased anti-PA titers were correlated with increased GI parasite infection intensities, and vice versa (Figure 3B). Thus, as GI parasite numbers increase, hosts may be less likely to mount a humoral immune response against anthrax. The significant, negative relationship between the rainfall/eosinophils interaction term and anti-PA titer prevalence in the PA GEE model further supports the hypothesis that this anti-anthrax immune attenuation may be related to GI parasite immunomodulation even in the face of increased anthrax exposure. These anthrax-Th2 immunity relationships held only for the wet season when GI parasite infection intensities were highest (Figure 2A). Similar directional tradeoffs were observed in studies of concurrent GI helminth and tuberculosis infections in buffalo [16],[43].

The lack of a stronger, directly negative relationship between GI parasite infection intensity and anti-PA titer magnitude may be due to the ability of animals that ingest low levels of B. anthracis in the wet season to mount at least a minor Th1-type immune response, even in the face of strong Th2-skewing. Potian et al. [58] found that mice coinfected with GI helminths and lung tuberculosis bacteria were able to mount both Th2 and Th1 responses, respectively, to these pathogens, but only once the parasites colonized separate organs. As animals ingesting very high doses of B. anthracis spores are more likely to develop subacute fulminant infection and die [26], animals that have measurable anti-anthrax titers are even more likely to be individuals who experienced low infectious doses while mounting a measureable Th1 response against an established Th2-type skewing. Animals that experience high GI parasite infections and large anthrax doses are also likely at much greater risk for mortality, thus effectively removing very helminth-susceptible and coinfected animals from this study (similar to [43]) but still benefiting anthrax transmission [33]. As we were not able to sample anthrax carcasses for GI parasite and immune parameters, our results may be biased against accurately measuring those animals that succumbed to anthrax due to macroparasite Th2-type skewing.

While studies suggest that ectoparasites are more active in parasitizing hosts during milder and more humid times, ectoparasite infection intensities in our study were both higher on hosts in the dry season in rain group comparisons and were significantly, negatively associated with rainfall in GEE models (Figure 3). Interestingly, higher GI parasite infection intensities were significantly associated with higher ectoparasite counts, despite the opposite interaction of ectoparasites with rainfall (Figure 3). Several studies suggest that a Th1-type immune response is most effective against tick infestations and that ticks can suppress this response in favor of a less effective Th2-type skewing [59]-[61]. However, our study suggests that a Th2-type immune response is actually more effective in fighting against tick infestations, the more so because otherwise the combination of preferred environmental conditions coupled with a strong Th2 skewing would be expected to result in very high tick burdens in the wet season. As hard-bodied ticks are more capable of survival during dry seasons than are desiccation-prone GI helminth free-living stages [62], the increase in tick infestations in the dry season may represent ticks taking advantage of a decrease in GI parasite-driven Th2 immune pressure at this time. The directionality of these interactions, with GI parasites potentially driving a Th2-type immune response and indirectly affecting ectoparasites, is supported by the fact that rare major histocompatibility complex (MHC) alleles were found to be associated with increased GI parasite infection intensities, and common alleles with increased ectoparasites in ENP zebras [63]. This suggests that GI parasites exert stronger selective pressure at the gene locus in question than do ectoparasites, with GI parasites having higher fitness costs to hosts and likely driving host immune allocation in this system.

Study area and species

Etosha National Park (ENP) is a 22,915 km² fenced conservation area in northern Namibia, located between 30’S-19°30’S and 14°15’E-17°10’E (Additional file 1: Figure A1). Rainfall in ENP is highly seasonal: the rainy season lasts from November through April, with the greatest rainfall occurring during January and February [21],[64] (Figure 4). The only perennial water available to the park's wildlife is found in man-made boreholes, or in natural artesian or contact springs [64].

Plains zebra are one of the most abundant ungulate species in ENP, with a population of approximately 13,000 (95% CI rounded to nearest 100: 10,900-15,000) [Etosha Ecological Institute, EEI, unpublished data 2005].

Study design and animal capture

We first immobilized and sampled zebra on the plains within approximately 20 km of Okaukuejo, the area in which the majority of anthrax deaths are observed (Additional file 1: Figure A1). In subsequent seasons, zebras were sampled around Okaukuejo and 100 km to the east near Halali following their migration patterns. All animals were safely handled under animal handling protocol AUP R217-0509B (University of California, Berkeley). We originally sampled 45 animals in the wet season, intending to recapture them as many times as possible in five subsequent sampling seasons roughly six months apart (Additional file 1: Table A1) to both control for individual variation and to examine individual versus population level effects of seasonality on immunity and disease.

We obtained whole blood, serum, fecal, and ectoparasite samples from zebra over five seasons between 2008 and 2010 (Additional file 1: Table A2). We conducted a total of 154 zebra capture events, representing 69 unique individuals, with 20 sampled twice, 11 sampled three times, 12 sampled four times, and two sampled five times. We sampled nearly all females to control for sex effects. We sampled only adult animals, further determining age to half a year by combining tooth eruption observations, caliper measurements of upper incisors, and patterns of wear [65],[66].

We collected blood from peripheral veins for blood smears, and removed serum for antibody analyses. From blood smears we obtained eosinophil and monocyte counts (Additional file 1: Appendix A). From serum we obtained IgE and IgGb antibody titers, and anti-anthrax (anti-PA) antibody titers using enzyme-linked immunosorbent assays (ELISAs) (Additional file 1: Appendix A). We used an ELISA procedure with wildtype Bacillus anthracis PA as coating antigen to measure host anti-PA titers ([32]; Additional file 1: Appendix A). We stimulated whole blood using established protocols, and used these samples to measure IL-4 and IFN-γ cytokine RNA in T cells using RT-PCR (Additional file 1: Appendix A).

We collected feces by observing an individual prior to capture and collecting a homogenized sub-sample within ten minutes of defecation. For animals that were not observed defecating, we collected fecal samples when possible by inserting a gloved hand into the rectum. As capture events took place between 9:00 and 14:00, fecal samples were collected within this same time window, thereby controlling for potential differences in timing of fecal egg shedding [67].

We collected and counted all visible ectoparasites, regardless of life stage, on zebras during capture events. The vast majority of ticks observed were R. evertsi mimeticus, a tick species found throughout Namibia in wild equids and greater kudu [68]. These ticks parasitize hosts year-round, with more adults present from November to May and immature stages peaking from February to March and May to September [62].

Rainfall quantification

The original study design involved sampling based on calendar months of the typical wet/dry seasons. However, annual rainfall variation during the study period caused some individuals sampled during the nominal "wet season" months to be captured during dry conditions. In addition, previous studies in this system determined that gastrointestinal helminth infection intensity in zebras is significantly related to rainfall experienced over the two months prior to sampling [69]. Thus, we determined cumulative rainfall by adding up daily rainfall amounts (in mm) over the 60 days prior to each individual capture event, using local rainfall gauge data. Histograms of cumulative rainfall prior to each individual sampling event revealed a strongly bimodal distribution in the individually-linked rainfall, regardless of calendar-based seasons (Figure 4); we thus chose to group by quantitative rainfall groups. This made both ecological and biological sense: those animals sampled very late in the S3 "wet" season fell into the lower rainfall group as the dry season had begun (Additional file 1: Table A2); and, external moisture greatly influences the within-host development and egg-producing activity of gastrointestinal parasites [70],[71]. The rain group seasons were thus: "wet season," the high rainfall group (individuals that had experienced ≥ 200 mm rainfall two months prior to sampling); and "dry season," the low rainfall group (experienced ≤ 100 mm rainfall).

Gastrointestinal parasite species and quantification

The gastrointestinal nematodes examined in this host species were in the order Rhabditia, suborder Strongylida, primarily within the superfamily Strongyloidea, and family Strongylidae; this group contains both the "large strongyles" (spp. in the subfamily Strongylinae) and the "small strongyles" (spp. in the subfamily Cyathstominae) [72]. These parasites are oviparous and exhibit a direct life cycle with three, free-living larval stages. The first two molts to the infectious L3 stage occur over one to two weeks, and are highly susceptible to desiccation [23],[73]. L3 larvae also require a film of moisture to move [24]; thus, it is unsurprising that previous studies in ENP found a strongly seasonal pattern in zebra strongyle infection intensities, with hosts exhibiting greater new infections during the wet season than in the dry [22]. We evaluated fecal samples for strongyle eggs using a modified McMaster technique for fecal egg counts [74], a commonly used non-invasive method for quantifying parasitism [75] (Additional file 1: Appendix A).

Statistical analyses