INTRODUCTION
Parasite prevalence and diversity patterns are highly influenced by environmental factors, such as climate and land cover, and their interplay with ecological factors such as host and parasite identity, interactions, routes of transmission, host spatial distribution and movement patterns and changes over time (Rogers and Randolf, Reference Rogers and Randolph2003; Patz et al. Reference Patz, Daszak, Tabor, Aguire, Pearl, Epstein, Wolfe, Kirkpatrick, Foufopoulos, Molyneux and Bradley2004; Jones et al. Reference Jones, Patel, Storeygard, Balk, Gittleman and Daszak2008). Due to this complex and dynamic net of interactions, early warning systems and management of zoonotic parasitic diseases require an exhaustive knowledge of zoonoses' origins and spatial patterns at relevant local and regional scales. Among parasites, endoparasites assume particular relevance since they have the potential to profoundly influence and regulate the structure and stability of natural communities (Hudson et al. Reference Hudson, Dobson and Newborn1998), especially those of their hosts. Regulation can be carried out by both micro- (e.g. Borrelia spp. – Murray et al. Reference Murray, Kapke, Evermann and Fuller1999) and macroparasites (e.g. helminths – Spratt, Reference Spratt1990).
Parasite richness and infection are inherently spatial processes, however, biogeographical approaches to the study of parasite richness, distribution and relationship with host populations at intermediate scales (e.g. nationwide) are still needed. The available studies at this scale are often restricted to a single host species (e.g. helminths and Eurasian badgers Meles meles Linnaeus, 1758 – Torres et al. Reference Torres, Miquel and Motjé2001, or red foxes Vulpes vulpes L., 1758 – Barbosa et al. Reference Barbosa, Segovia, Vargas, Torres, Real and Miquel2005).
The level of host colonization depends on the parasite infra-communities and their characteristics (life cycle, niche utilization, mean species diversity, number of high density species), host characteristics (digestive tract type and diet, probability and patterns of colonization by parasites) and the external environment (e.g. temperature and precipitation) (Pence, Reference Pence, Esch, Bush and Aho1990; Lindenfors et al. Reference Lindenfors, Nunn, Jones, Cunningham, Sechrest and Gittleman2007). Some studies have focused on the ecological processes responsible for differences in parasite community richness (i.e. number of parasite species) (e.g. Bush et al. Reference Bush, Aho and Kennedy1990; Lindenfors et al. Reference Lindenfors, Nunn, Jones, Cunningham, Sechrest and Gittleman2007). The analysis of more than 7000 carnivore-parasite species pairs described by Lindenfors et al. (Reference Lindenfors, Nunn, Jones, Cunningham, Sechrest and Gittleman2007) found that higher host body mass, population density, geographical range size, distance from the equator and home-range area influence positively parasite species richness. Further, other studies have pointed out the positive correlation of host population density (Nunn et al. Reference Nunn, Altizer, Jones and Sechrest2003), basal metabolic rate (Morand and Harvey, Reference Morand and Harvey2000), diet (Watve and Sukumar, Reference Watve and Sukumar1995) and longevity (Torres et al. Reference Torres, Miquel, Casanova, Ribas, Feliu and Morand2006), and the negative correlation of latitude (Nunn et al. Reference Nunn, Altizer, Sechrest and Cunningham2005), gregariousness and predatory pressure (i.e. trophic level) (Watve and Sukumar, Reference Watve and Sukumar1995) with parasite diversity. The above-mentioned species-specific studies focused on Iberian carnivores (Torres et al. Reference Torres, Miquel and Motjé2001; Barbosa et al. Reference Barbosa, Segovia, Vargas, Torres, Real and Miquel2005) highlighted the importance of the biogeographical region characteristics (namely the predator diet in the area, local patterns of temperature and mean rainfall, and animal and plant biocenosis structure – Torres et al. Reference Torres, Miquel and Motjé2001), distance to urban centres, predator densities, and water availability (e.g. lower soil permeability) on parasite diversity.
Moreover, parasite adaptive evolution may also contribute to this diversity and it has been shown that there is a high degree of evolutionary association between parasites and hosts (Poulin, Reference Poulin1995). For instance, in the Iberian Peninsula the Lyme borreliosis group of tick-borne spirochetes presents various patterns and levels of host specialization, with some being associated with birds (e.g. Borrelia valaisiana and B. garinii) and others with rodents (e.g. B. afzelii) (Vitorino et al. Reference Vitorino, Margos, Feil, Collares-Pereira, Zé-Zé and Kurtenbach2008) and carnivores (e.g. B. burgdorferi) (Doby et al. Reference Doby, Betremieux, Barrat and Rolland1991).
External environmental factors alone may determine parasite survival and population dynamics even before the effect of host condition is evident. For example, several landscape-related factors, such as the presence of human infrastructures (e.g. dams and irrigation schemes), climatic (e.g. water temperature), physical (e.g. water velocity) and chemical (e.g. water pH) variables, have been identified as determining features in the transmission of Schistosoma species in Africa, since they constrain the distribution of an intermediate host (water snails) of those parasites (Brooker, Reference Brooker2007). Temperatures and moisture/precipitation also influence the development and survival of helminth infective stages (Kates, Reference Kates1965), and therefore may promote or restrain parasite transmission. Moreover, heterogeneity of environmental conditions has been suggested to affect parasite community composition by one of the few biogeographical studies of parasite richness (Torres et al. Reference Torres, Miquel, Casanova, Ribas, Feliu and Morand2006).
The above examples suggest that parasite richness may be driven by one of two alternative factors: (1) hosts bio-ecological features or (2) the external environment context in which parasite populations exist; or their combined effect. With the aim of contributing to the topic of multispecies analysis at intermediate geographical scales, we tested hypotheses on the importance of host and/or environmental factors upon endoparasite richness and distribution in the Iberian Peninsula, using carnivores as models. Carnivores are a good model to study endoparasite richness, since they often have large home ranges and distribution areas, encompassing different landscape units (Macdonald and Kays, Reference Macdonald, Kays and Nowak2005), they generally have a broad prey spectrum (Macdonald and Kays, Reference Macdonald, Kays and Nowak2005), which increases the probability of infection by parasites, and they are hosts to a wide variety of parasites, with high interspecific variation. Reviews by Campillo et al. (Reference Campillo, Ordóñez and Feo1994) and Torres et al. (Reference Torres, Miquel, Casanova, Ribas, Feliu and Morand2006), describing the hosts of Iberian helminths and protozoans, show that the number of endoparasites varies between 37 (for the red fox) or 25 (common genet, Genetta genetta L., 1758), and 3 (otter, Lutra lutra L., 1758) or 4 (stoat, Mustela erminea L., 1758). More specifically, we wanted to understand whether the richness of carnivore-infesting endoparasites was related to (1) the bio-ecological characteristics of the hosts (e.g. geographical range, body mass, longevity, diet, predatory pressure/trophic level, gregariousness), (2) the climate patterns of the region (e.g. temperature, precipitation – Kates, Reference Kates1965) or (3) a combination of both (Hybrid model). Given the complexity of hosts-parasites, parasites-parasites and parasites-environment interactions, which vary among host and parasite species, we hypothesize that, due to the high bioclimatic heterogeneity (EEA, 2008) and carnivore niche specializations (e.g. Iberian lynx Lynx pardinus Temminck, 1827 – Ferrer and Negro, Reference Ferrer and Negro2004) found within Iberia, parasite species richness will depend upon the interaction between climate and host characteristics.
MATERIALS AND METHODS
Data collection
We reviewed published studies in international peer-reviewed scientific journals focused on parasitology and zoology, describing parasite prevalence (e.g. abundance of individual parasite species) and richness in carnivores (Table 1). We restricted our search to the Iberian Peninsula (Fig. 1), and only studies describing the specific location of the data (e.g. latitude, longitude, valley, protected area, etc.) were considered. For each study, we gathered information on the number and identity of endoparasite species (Acanthocephala, Cestoda, Digenea, Nematoda, Trematoda, Protozoa and Mycobacterium spp.), host bio-ecological features, site location and climate characteristics of the region. Although relevant, information on parasite prevalence (i.e. proportion of individuals in a population that are infected) could not be considered because it was not mentioned in all the studies we reviewed. Therefore, only absence/presence data were recorded for each parasitic species.
Location of the reviewed studies used in our analysis.
Sources of data on carnivore parasites in Iberia
For each study we recorded the carnivore(s) species. In addition, we also reviewed the literature to gather information on the host species: weight and geographical range (Palomo and Gisbert, Reference Palomo and Gisbert2002), longevity in the wild (http://www.demogr.mpg.de/), trophic level (small, meso or top predator), type of diet (generalist, specialist), and life strategies (solitary, gregarious) (e.g. Palomo and Gisbert, Reference Palomo and Gisbert2002).
Study area location was described as the easting (x) and northing (y) coordinate in lat/long. When not explicitly reported, approximate location was estimated using the area's geographical description within a Geographical Information System (GIS). We also recorded sample sizes, which corresponded to the number of faeces, guts or individuals analysed in the reviewed studies.
Finally, we used a GIS in ArcGIS 9.3 (ESRI, California, USA) to collect climatic and biogeographical data for each study location. The mean annual temperature and mean annual precipitation for the sites were determined using data available at the WorldClim – Global Climate Data database (http://www.worldclim.org/). Data were downloaded in the format of global climate layers (climate grids) with a spatial resolution of 2·5arc min (∼5 km2). These layers represent the mean monthly temperature and mean total early precipitation values, restricted to the period 1950–2000 (see data details in Hijmans et al. Reference Hijmans, Cameron, Parra, Jones and Jarvis2005). Data regarding biogeographical regions were downloaded from the European Environment Agency database (EEA, 2008; http://www.eea.europa.eu/data-and-maps/data/biogeographical-regions-europe-2008) and overlapped with the studies' geographical locations (Table 2). Although 3 bioclimatic regions are found in the Iberian Peninsula (Alpine, Atlantic and Mediterranean), the studies analysed here fell in either the Atlantic or the Mediterranean regions. A single exception was 1 study in the Alpine region whose location neighboured the Atlantic region limits and that, therefore, was included in the latter region (also because the climatic characteristics of the Alpine and Atlantic regions are more alike than with the Mediterranean region).
Candidate variables assumed to influence parasitic burden of carnivores in Iberia (PIC – Phylogenetic Independent Contrasts)
Data analysis
Since the reviewed studies presented a high variation in sample size, and this parameter influences the number of species detected (Soetaert and Heip, Reference Soetaert and Heip1990; Poulin, Reference Poulin1998), we calculated the residual of parasite species richness from a linear regression on the studies sample size (Lindenfors et al. Reference Lindenfors, Nunn, Jones, Cunningham, Sechrest and Gittleman2007). The remaining used variables are listed in Table 2.
All the following analyses were performed using R software, version 2.11.0 (R Development Core Team 2008) and the specified packages. We assessed the data for spatial autocorrelation by calculating Moran's I index (“ape – Analyses of Phylogenetics and Evolution” package – Paradis et al. Reference Paradis, Claude and Strimmer2004). Spatial autocorrelation, which is intrinsic to most biological data, implies the violation of the assumption of independence, resulting in possible pseudoreplication (Legendre, Reference Legendre1993; Carl and Kühn, Reference Carl and Kühn2007). Due to the significant results, we opted to define geographical areas for each study, according to their proximity, and use this variable (Reg) as a random factor in the subsequent models. Data overdispersion was not tested since residuals showed a distribution not significantly different from normal (K-S=1·092, P=0·184).
Since bio-physical variables are often correlated (i.e. their variations assume similar patterns), it is important to test for multicollinearity between variables in order to exclude those highly correlated and be able to separate the effects of 2 (or more) variables on a response variable (Tabachnick and Fidell, Reference Tabachnick and Fidell1996). Thus, we tested the correlation between continuous variables using the Spearman's rank correlation coefficient (rs), the Kendall's Tau (T) for ordinal variables, and Cramer's V when at least one of the variables was nominal (Siegel and Castellan, Reference Siegle and Castellan1988). From those variable pairs that presented significant correlations (P<0·05), we selected the one with a higher correlation with the response variable.
To test whether host characteristics, environmental patterns or a combination of both influence carnivore endoparasite richness (as defined above) (see Table 2 for variable description), we used a General Linear Mixed Model (GLMM) with a Gaussian distribution and an identity link function (Zuur et al. Reference Zuur, Ieno, Walker, Saveliev and Smith2009). The use of a Gaussian distribution function was supported because the data had a tendency to normality. We created a set of 18 models, aggregated into 3 categories: (i) host model (host weight and trophic level), (ii) environmental model (temperature and biogeographical region), and (iii) hybrid models (combination of environmental and host features, including a full model with all variables). Furthermore, since several of the reviewed studies concerned the same species, this could be introducing a bias towards the causal characteristics of the more represented species. To deal with this bias and to account for the effects of host phylogenetic relationships we also included the host species (SP) as a random effect. We also controlled for the confounding effect of phylogeny by considering the phylogenetic independent contrasts (PICs) of the variables host weight and host longevity. Phylocom 4.1 (Webb et al. Reference Webb, Ackerly and Kembel2008) was used to compute PICs from a phylogeny of host sequences (Oliveira et al. Reference Oliveira, Castro, Godinho, Luikart and Alves2010) of the nuclear gene IRBP (interphotoreceptor retinoid-binding protein). The phylogeny was estimated using maximum likelihood in Treefinder (Jobb et al. Reference Jobb, von Haeseler and Strimmer2004). A phylogeny of host sequences (Fernandes et al. Reference Fernandes, Ginja, Pereira, Tenreiro, Bruford and Santos-Reis2008) of the mitochondrial gene cytb (cytochrome b) was also estimated, but the IRBP tree topology more faithfully represented the carnivore consensus phylogeny from Flynn et al. (Reference Flynn, Finarelli, Zehr, Hsu and Nedbal2005). The correlation between PICs for both variables calculated from the IRPB phylogeny was tested so that, if significant, the variable with the higher correlation with the dependent variable could be retained for modelling.
The GLMM was built using the ‘lme4 – Linear mixed-effects models using S4 classes’ package for R (Bates and Maechler, Reference Bates and Maechler2010).
We used an information criterion model selection procedure to detect which was the most parsimonious model to explain the measured variability in carnivore endoparasite standardized richness. We used an extension of the Akaike's information criterion for small samples (AICc), as this index includes a bias-adjustment for small sample sizes. AICc should be used if the ratio between the number of cases (n) and number of parameters (K) is less than 40 (Burnham and Anderson, Reference Burnham and Anderson2002, Reference Burnham and Anderson2004). The best models were assumed to be those with the lowest AICc value. We also calculated the difference in AICc between each pair of models (ΔAICc), which corresponds to the difference between the AICc of the ith model and the minimum AICc value. This metric allows determinion of how good the ith model is as an approximation to the expected best model. We considered best models to be those with ΔAICc <2 (Burnham and Anderson, Reference Burnham and Anderson2002, Reference Burnham and Anderson2004). Finally, we also calculated the Akaike weights (w i) to obtain each model's probability of being the best model for the data (Burnham and Anderson, Reference Burnham and Anderson2004).
RESULTS
We reviewed 20 studies, corresponding to a total of 34 different study areas, widely distributed within the Iberian Peninsula (Fig. 1 and Table 1). These sites showed a non-significant spatial autocorrelation (Moran I=0·076, P=0·236). Being conservative, and although no significant autocorrelation was detected, we included the study location as a random effect to account for possible bias associated with this factor in the models.
We detected that several explanatory variables presented multicollinearity. From the original set of candidate variables that could be influencing endoparasite richness, only 5 were used for the subsequent modelling procedure: host weight, PIC host weight, mean annual temperature, trophic level and biogeographical region (Table 2).
The most parsimonious model to describe the richness of endoparasites in carnivore species was a hybrid model where biogeographical region and carnivore trophic level were included (Tables 3 and 4). This model had the lowest ΔAICc, highest support (highest w i, reaching almost 0·74), and showed a negative effect of biogeographical region 2 (Mediterranean) and of host trophic level (top predator) on residuals of parasite species richness (i.e. these 2 variables reduce the residuals and thus influence positively parasite diversity in hosts).
Summary of fitted models Information Criteria (AICc – Akaike's Information Criterion; ΔAICc – difference to the lowest AICc value; wi – Akaike weights), that corresponded to the 3 tested hypotheses: endoparasite infection depends upon hosts, environment or a combination of both
Estimated coefficients and standard error for the variables of the best model
DISCUSSION
Our study revealed that the richness of endoparasites in Iberia is higher in the Mediterranean biogeographical region, and when host species correspond to top predators (e.g. wolf, Canis lupus, and Iberian lynx). The distribution and success of endoparasite infection is usually associated with the environment, heterogeneity in parasite burden among individuals, age and genetic specificities of hosts, and polyparasitism (Hotez et al. Reference Hotez, Brindley, Bethony, King, Pearce and Jacobson2008).
Among the environmental promoters of endoparasite colonization success, climate seems to be crucial (Brooker, Reference Brooker2007). The Mediterranean biogeographical region, which showed a positive effect on carnivores' endoparasite richness, is characterized by moderate and mild temperatures, except in the summer when it is hot and dry (Blondel and Aronson, Reference Blondel and Aronson1999; Condé and Richard, Reference Condé and Richard2002). In this region, unpredictable diurnal temperatures and wind fluctuations occur, and infrequent floods and prolonged droughts are also characteristic (Condé and Richard, Reference Condé and Richard2002).
Most of the detected endoparasites were helminths (Acantocephala, Cestoda, Digenea, Nematoda and Trematoda), and several environmental factors have been identified as necessary conditions for the development and survival of these parasites outside the final hosts: moderate temperatures, adequate moisture, and protection from freezing, desiccation and direct sunlight (Kates, Reference Kates1965). Even some of the Protozoa detected in the reviewed studies have dispersal benefits from mild environments (Schustera and Visvesvara, Reference Schustera and Visvesvara2004), such as those found in the Mediterranean region of the Iberian Peninsula throughout most of the year. Genera like Isospora, Sarcocystis or Cryptosporidium, whose oocysts are present in the faecal matter prior to infecting vectors (e.g. insects – Graczyk et al. Reference Graczyk, Knight and Tamang2005), are sensitive to freezing and desiccation, although freezing is more limiting to survival (Robertson et al. Reference Robertson, Campbell and Smith1992).
Thus, most endoparasites, prior to infecting their hosts, survive better in environments with mild temperatures and some humidity (e.g. Christensen et al. Reference Christensen, Frandsen and Roushdy1980). Mediterranean regions of Iberia may suffer severe summer droughts, but most of the year the climate is mild, which may favour parasite survival outside hosts (in contrast with the Atlantic and Alpine regions, in which winters are rigorous, with temperatures often dropping below 0°C).
Although most life cycles of endoparasites take place inside hosts, which act as a protective barrier against deleterious environmental conditions (e.g. associated with latitude – Willig et al. Reference Willig, Kaufman and Stevens2003), some of the most crucial phases associated with dispersal or host changing take place outside hosts' bodies (Piekarski, Reference Piekarski1962; Graczyk et al. Reference Graczyk, Knight and Tamang2005). In such occasions parasites are sensitive to adverse environmental factors and benefit from mild conditions (e.g. soil-transmitted helminths are constrained by surface temperature, soil type and rainfall – Hotez, Reference Hotez, Brindley, Bethony, King, Pearce and Jacobson2008). Therefore, carnivores living under Iberian Mediterranean climate, with high temperatures in the summer and mild winters, are more prone to have a higher richness of parasites than their counterparts inhabiting Atlantic-type areas.
Our results also indicate that the highest level of endoparasite richness is at the top predator level. This is likely because these species have larger home ranges (Nowak, Reference Nowak2005) and thus are likely to encounter more opportunities of infestation and a higher number of endoparasite species than carnivores with smaller home ranges. Moreover, endoparasite infection has been reported to have cascading effects throughout the food web, due to bottom-up (number of species of parasites that top-predators encounter) and/or top-down (the area that the top predator covers and its interactions with other species) effects. It has been shown that there is an increase in the richness of endoparasites in upper levels of the trophic chain (Chen et al. Reference Chen, Liu, Davis, Jordán, Hwang and Shao2008). Evidence exists that parasites frequently adapt to host predation by parasitizing the predator (Dobson et al. Reference Dobson, Lafferty, Kuris, Pascual and Dunne2006). For example, many of the endoparasites infect hosts by ingestion, and therefore host diet and trophic level should influence, at least partially, the number of species to which a host will be exposed (Poulin, Reference Poulin1995). This mechanism is particularly important in heteroxenic parasites (i.e. with complex life cycles), whose stages occupy several different positions in the trophic web, leading to complex long-loops in interactions (Dobson et al. Reference Dobson, Lafferty, Kuris, Pascual and Dunne2006).
Our results, however, do not match previous findings on Iberian carnivore endoparasite richness. The helminthological study of Torres et al. (Reference Torres, Miquel, Casanova, Ribas, Feliu and Morand2006) showed that the number of parasite species in each carnivore species was positively affected by host body mass, density and distribution and, marginally, by host longevity. Although we tested similar ecological and environmental parameters, we believe that the differences found may be related to the parasite taxa included in each study. In our study we also considered Protozoa and Mycobacterium, both having substantially different requirements from those of helminths (Baker, Reference Baker1969). Also at variance with the results of Torres et al. (Reference Torres, Miquel, Casanova, Ribas, Feliu and Morand2006), the review by Watve and Sukumar (Reference Watve and Sukumar1995) showed that host body weight, home range, population density, gregariousness, and diet did not affect parasite loads. This lack of consistency in results may suggest that parasite species richness is more affected by historical (e.g. phylogenetic determinants – Bush et al. Reference Bush, Aho and Kennedy1990) and regional processes (including climate) than by host ecological processes. However, the factors that drive parasite richness may vary regionally, since several parameters may interact upon this community, with the relative contribution of each one depending on their interplay (Poulin, Reference Poulin2004).
Finally, it is important to mention that our review is limited to already published research and some of the carnivore species inhabiting Iberia were hence not considered in the analysis (e.g. stoats, polecats Mustela putorius L., 1758, stone martens Martes foina Erxleben, 1777). How the inclusion of these mesocarnivores would affect the results is unknown, but the present study has the value of highlighting that the factors that influence parasite richness vary, and that this variation may be dependent on the host and parasite groups considered, as well as on the environmental characteristics of the study areas.
ACKNOWLEDGEMENTS
The study was funded by the Fundação para a Ciência e a Tecnologia and Fundo Social Europeu (III Quadro Comunitário de Apoio) (SFRH/BPD/35842/2007). C.F. acknowledges support from the Fundação para a Ciência e a Tecnologia (FCT-MCTES, Portugal) through the Ciência 2007 Research Fellowship C2007-UL-342-CBA1.
