Habitat fragmentation and vegetation structure impact gastrointestinal parasites of small mammalian hosts in Madagascar

Abstract Deleterious effects of habitat loss and fragmentation on biodiversity have been demonstrated in numerous taxa. Although parasites represent a large part of worldwide biodiversity, they are mostly neglected in this context. We investigated the effects of various anthropogenic environmental changes on gastrointestinal parasite infections in four small mammal hosts inhabiting two landscapes of fragmented dry forest in northwestern Madagascar. Coproscopical examinations were performed on 1,418 fecal samples from 903 individuals of two mouse lemur species, Microcebus murinus (n = 199) and M. ravelobensis (n = 421), and two rodent species, the native Eliurus myoxinus (n = 102) and the invasive Rattus rattus (n = 181). Overall, sixteen parasite morphotypes were detected and significant prevalence differences between host species regarding the most common five parasites may be explained by parasite–host specificity or host behavior, diet, and socioecology. Ten host‐ and habitat‐related ecological variables were evaluated by generalized linear mixed modeling for significant impacts on the prevalence of the most abundant gastrointestinal parasites and on gastrointestinal parasite species richness (GPSR). Forest maturation affected homoxenous parasites (direct life cycle) by increasing Lemuricola, but decreasing Enterobiinae gen. sp. prevalence, while habitat fragmentation and vegetation clearance negatively affected the prevalence of parasites with heterogenic environment (i.e., Strongyloides spp.) or heteroxenous (indirect cycle with intermediate host) cycles, and consequently reduced GPSR. Forest edges and forest degradation likely change abiotic conditions which may reduce habitat suitability for soil‐transmitted helminths or required intermediate hosts. The fragility of complex parasite life cycles suggests understudied and potentially severe effects of decreasing habitat quality by fragmentation and degradation on hidden ecological networks that involve parasites. Since parasites can provide indispensable ecological services and ensure stability of ecosystems by modulating animal population dynamics and nutrient pathways, our study underlines the importance of habitat quality and integrity as key aspects of conservation.


| INTRODUC TI ON
The natural world is highly impacted by human activities of various kinds (Jha & Bawa, 2006). Most importantly, land conversion leads to overall size reduction and an increasing degree of fragmentation of the remaining natural habitats, threatening biodiversity.
Parasites are essential components of ecosystems and act as regulators of host population dynamics and community structure (Dunne et al., 2013;Lafferty et al., 2006Lafferty et al., , 2007Marcogliese, 2004;Mouritsen & Poulin, 2005;Thomas et al., 1999). Whereas ecological impacts of habitat fragmentation on ectoparasites (e.g., mites, ticks) can be expected given their more direct exposure to external abiotic conditions (Bush et al., 2013;Carbayo et al., 2019;Kiene et al., 2020), the response of endoparasites to habitat fragmentation is less intuitive. However, depending on their life cycle, endoparasites can be exposed to direct environmental influences as free-living stages or in intermediate hosts (Simões et al., 2016).
The fragmented forests of Madagascar represent a highly suitable model region to investigate impacts of habitat fragmentation and degradation on parasites. The islands' forests are under particular pressure (Harper et al., 2007), since the human population of Madagascar has grown from around 4 million people in 1950 to almost 27 million people in 2019 (United Nations, 2019). By 2014, natural forest cover decreased to 56% of its size in 1953 (Vieilledent et al., 2018). In parallel, human impact on the remaining forests and the potential for ecological edge effects increased considerably, since 46% of the remaining forest areas are located closer than 100 m to a forest edge (Vieilledent et al., 2018).
Malagasy forest ecosystems harbor an extraordinary and unique species richness and are considered a worldwide hotspot for biodiversity (Goodman & Benstead, 2005;Raik, 2011). A few studies  Kiene et al. (2020) found significantly lower ectoparasite infestation rates in small mammalian hosts from smaller forest fragments and in proximity to the forest edge. While almost all ectoparasite types were affected, effects were particularly evident in temporary ectoparasites such as ticks and chigger mites. The authors argued that ectoparasite survival and reproduction of temporary parasite stages in edge environments are most likely reduced by unfavorable abiotic environmental conditions at the forest edge compared with the forest interior .

Some studies investigated gastrointestinal parasite infections
of Malagasy vertebrates in relation to habitat disturbance and degradation and again with different results: Rakotoniaina et al. (2016) found no effect of habitat degradation on gastrointestinal parasites in gray mouse lemurs (Microcebus murinus), while the study of Raharivololona and Ganzhorn (2009) suggested that effects differ between parasite species, but the investigation was based on very small sample size. Studies on homoxenous pinworms of larger brown lemurs (Eulemur spp.), however, found higher prevalences in animals inhabiting secondary forests or previously logged habitats (Schwitzer et al., 2010;Winter et al., 2020), but the studies are lacking information on heteroxenous parasites. To our knowledge, no study has so far investigated the complex effects of habitat fragmentation on gastrointestinal parasites with different life cycle characteristics. study underlines the importance of habitat quality and integrity as key aspects of conservation. Rattus rattus: ~100 g), are known to occur in larger numbers even in forest fragments, and live in partial sympatry in the study region . All of them are nocturnal solitary foragers, exhibit a more or less arboreal lifestyle, and spend the day in protected sleeping sites. The four studied species, however, also differ in important aspects of their biology: the two Microcebus species and R. rattus are living in groups and feed on an omnivorous diet (Clark, 1982;Radespiel, 2000;Radespiel et al., 2006;Shiels & Pitt, 2014;Thorén et al., 2011;Weidt et al., 2004), while the western tuft-tailed rat has a solitary lifestyle and is categorized as frugivorous (Goodman, 2016;Randrianjafy et al., 2008). The rodents and M. murinus mainly use tree holes and dens for sleeping, each of them for longer stretches of time (Goodman, 2016;Münster, 2003;Radespiel et al., 2003), while M. ravelobensis is known to employ more open and ephemeral shelters such as self-built leaf nests and switch between sites more frequently (Radespiel et al., 2003;Thorén et al., 2010 In addition to host species, impacts of host sex and host population density can be anticipated. Male hosts are expected to exhibit higher gastrointestinal parasite prevalences compared with females, since specific male behavior is known to foster parasite infections (Altizer et al., 2003;Klein, 2004;Poirotte & Kappeler, 2019;Zuk & Mckean, 1996). Hosts from habitats with higher population density can be expected to show elevated gastrointestinal parasite infection rates, since increasing social interactions and density-associated higher parasite contamination of the environment might increase infection risk (Arneberg, 2002). In addition to these host-related factors, external environmental factors (e.g., forest size, proximity to the forest edge, vegetation structure, and human disturbance) can also be expected to impact infections, since gastrointestinal parasites spend periods of their life as free-living stages in the environment or rely on arthropod intermediate hosts. Thus, gastrointestinal parasites can be predicted to be negatively impacted by forest edges and habitat degradation, since lower humidity and UV radiation in edge and degraded habitats can be suspected to increase mortality of parasites and availability of arthropod intermediate hosts.
However, as gastrointestinal parasites should be overall more protected from external environmental influences than ectoparasites, we expected a weaker response than in the previous study on ectoparasites . Host-related factors, in contrast, are expected to explain most of the prevalence variation, although their investigation in a nonexperimental setting precludes full clarification of causality. 15°240S, 46°440E, 50 km northeast of Mahajanga; Figure 1b). Major differences between the two locations concern elevation, disturbance by human presence, the type of landscape separating forest fragments (= matrix), and the availability of surface water. Situated on a plateau at about 180 m above sea level (a.s.l.), the forest sites in ANK are surrounded by a homogeneous, dry grassland matrix dominated by Aristida barbicollis (Steffens & Lehman, 2016). Open water bodies are completely absent and expansion of forest vegetation into the matrix is mostly prevented by cattle herding and recurrent bushfires . The forest patches in MAR are located closer to the Mozambique Channel (4-15 km) at an elevation of 20-90 m a.s.l. and managed by the local municipality Mariarano, which is situated in the center of the area. Here, the fragmented dry deciduous forests are embedded in a rather heterogeneous matrix consisting of rice fields or of savannah-like grasslands with palm trees (Bismarckia nobilis) in varying densities. Ponds, streams, channels for field irrigation, and the Mariarano River provide some humidity throughout the year, and the riverine vegetation maintains a potential connection between some of the forest fragments. In general, a relatively cool dry season from May to October and a hot and humid rainy season from November to April ensure a highly seasonal climate in the entire region.

| Recording of vegetation data
Vegetation density data were collected by counting seedlings (height: 1-100 cm), saplings (height: 101-250 cm), trees (height >250 cm), and lianas (diameter at breast height ≥2.5 cm) within plots of a size of 2 × 10 m according to Malcolm et al. (2016). Plots were installed in pairs, orthogonally directed from the forest transects, at a predefined set of distances to the forest edge (0,20,40,60,80,100,200,250, 300 m, then every 100 m). Thus, 16 pairs of vegetation plots were evaluated along a 1,000 m transect. In addition, the number of signs of disturbance by human presence per plot (number of cut trees, large holes in the ground as residuals from maciba (Dioscorea spp.) root harvesting as well as zebu scats) was recorded.
To condense data on vegetation structure and human disturbance, and to include these into the subsequent generalized linear mixed modeling, a principal component analysis (PCA) was performed. For each 100 m segment of a transect, average values of seedling, sapling, tree, liana, cut tree, maciba hole, and zebu scat counts across all vegetation plots were used as data points for the PCA. Resulting principal components (PC) were finally attributed to each host individual captured at trap positions within the respective segment. The PCA was performed using the R-command "prcomp()". The principal components (PC1, PC2) with Eigenvalues of >1 and a high explanatory power (PC1: 26.2%, PC2: 18.5%) were selected as predictor variables for the subsequent modeling. Factor loadings were utilized to interpret their effects (correlation coefficients; Table 1; File S1).
Lower densities of trees, seedlings and saplings, and higher numbers of cut trees and zebu scat were associated with increasing values of PC1 (Table 1). Consequently, PC1 is primarily associated with changes in general vegetation density. With increasing PC1, the vegetation opens up and forests are more frequently used by zebus, indicating increasing human impact ( Figure 2). We consequently interpret PC1 as factor illustrating "vegetation clearance". In contrast, PC2 is rather connected to changes along a gradient from secondary to primary forest vegetation ( Figure 2). It is linked to lower numbers of cut trees, seedlings, saplings and lianas, and higher numbers of maciba holes (Table 1). The decreasing number of cut trees and the reduced understory (less seedlings and saplings) along an increasing PC2 suggest the presence of larger old growing trees indicative of a mature primary forest. Although more maciba holes could indicate higher human impact, they also imply a greater density of maciba plants, which depend on a more pristine ecosystem (Andriamparany et al., 2015). For readability, we consequently interpret PC2 as factor illustrating "forest maturation". F I G U R E 1 Maps of two studied networks of fragmented dry deciduous forest, one in the western part of the Ankarafantsika National Park (a) and one in the Mariarano region (b). Modified after Andriatsitohaina et al. (2020)   were taken directly from the anus or collected from the traps, which were cleaned and disinfected prior to installation. Samples were directly preserved in 1.5 ml ethanol (90%-96%) and stored at 4°C.
For molecular verification, a minimum of one subsample of each parasite morphotype from each host species was selected for analysis. DNA was isolated from 5 to 50 eggs or larvae per sample with the NucleoSpin©Tissue kit (MACHEREY-NAGEL). Egg morphotypes with strong shells were homogenized using Precellys ® ceramic bead tubes (Bertin Instruments) prior to DNA isolation. The rDNA region spanning the internal transcribed spacer (ITS)1-5.8S-ITS2 sequence (hereafter referred to as ITS sequences) has been shown to be of excellent use for the taxonomic classification of nematodes (Blouin, 2002;Nabavi et al., 2014) and was therefore chosen for this study. Newly generated sequences were deposited in GenBank under accession nos. MW520838-MW520847, MW520852, and MW520853. Details of methods and the results of this molecular taxonomic evaluation are described in File S2 and were used to support parasite morphotype classification down to the family, genus, or species level, whenever possible. Those derived names are used throughout the manuscript. differences in parasite prevalence (data not shown), they were integrated as random factors into all models to control for possible and confounding spatiotemporal dynamics in the dataset. Using them as predictor variables was beyond the scope of this study and also not possible due to the heterogeneous sampling strategy across region, year, and month (i.e., the two regions were not sampled across all months in both years) which precluded their systematic analysis. However, we made sure that at any time we studied continuous forest sites and fragment sites in parallel to preclude systematic seasonal biases in the parasite dataset.

| Data analyses
Host density data were square-root-transformed, and forest size and distance to edge data were log-transformed to achieve normal distribution. Presence-absence data of parasite morphotypes are binomial by definition and the logit-link was therefore used in the respective models. Models concerning the GPSR were based on Poisson assumption and used log-link. Since some of the predictor variables could only be calculated for hosts from forest fragments, while others were available for the complete dataset, two sets of global models were built, one for the complete dataset and one for the fragment dataset (Table 2). Some numerical predictor variables were correlated with each other (Table 3) and were therefore never tested together in one global model. Therefore, three different global models were built for each dataset. As a consequence, six global models were built for each dependent variable (models A-C: complete dataset, models D-F: fragment dataset, Table 2).
The selection of the best model from a set of multiple candidate models (derived from each global model) was based on the Akaike information criterion (AIC) as described by Burnham and Anderson (2002) using the corrected AIC (AICc) to compensate for small sample sizes (Hurvich & Tsai, 1989). The best model with the highest statistical support and the ones with similarly low AICc values (∆AICc < 2) were considered for interpretation of content. To

| RE SULTS
of the best models with lowest AICc is provided in Table 5, while modeling results are summarized in the text for all submodels with ∆AICc < 2, and details of all submodels with ∆AICc < 2 are documented in File S4.
( Figure 3c). Due to low prevalence, ecological modeling could not be conducted.

F I G U R E 4
Total and host-specific prevalences (in %) of the 16 gastrointestinal parasite morphotypes and the proportion of noninfected individuals in the four host species 3.1.3 | Ascarididae Two parasite morphotypes were morphologically assigned to this family. Ascarid eggs 1 and 2 (Figure 3g,h) were detected only in the two mouse lemurs or M. ravelobensis alone (Figure 4). Again, hostspecific and total prevalences were low with 0.4% (4/903) and 0.2% (2/903), respectively. It is very likely that the two ascarid egg morphotypes belong to the same species; however, genetic allocation was impaired by insufficient egg quantity. Similarly, prevalences were too low for reliable ecological modeling.

| Trichosomoides
One egg morphotype, excreted only by R. rattus (13.8%; total prevalence: 2.8% [25/903]), was morphologically identified as Trichosomoides crassicauda (Figure 3i). The eggs of this bladder parasite are excreted with the urine and thus likely contaminated the feces in the traps. The low overall prevalence precluded ecological modeling.
spp. (Figure 3l) was shed by all host species. The total prevalence amounted to 40.5% (366/903), with highest infection rates in the two Microcebus species (53.7% and 45.7%) and lowest prevalence in E. myoxinus (8.8%; Figure 4, File S3). Genetic analyses revealed that eggs of this morphotype obtained from the two mouse lemur hosts most likely belong to the same species, while the two rodents harbored different species, which probably belong even to different genera.
GLMMs showed that prevalences were significantly lower in rodents than in mouse lemurs (all models; estimates rodents-mouse lemurs : The second egg type attributed to this superfamily, a Subuluroidea-like egg ( Figure 3m) with a total prevalence of 0.6% (5/903), was excreted by R. rattus and E. myoxinus only (Figure 4; File S3). Prevalences did not allow reliable ecological modeling.

| Gastrointestinal parasite species richness (GPSR)
Mouse lemurs and black rats had significantly higher GPSRs than

| Gastrointestinal parasite infections in different host species
Significant prevalence differences between host species were shown for all five modeled parasite morphotypes as well as GPSR (Figures 4 and 6). A common feature of most oxyurids is egg deposition around the anus, facilitating autoinfection and intraspecies transmission during close contact, but impeding interspecies transmission (Baker, 2008;Hugot et al., 1999). Here, both mentioned morphotypes were detected in the two mouse lemur species. Additionally, the morphotypes were each found in one R. rattus individual. Since at least for the genus Lemuricola, host specificity for hosts of the family Cheirogaleidae is described (Hugot et al., 1999;Irwin & Raharison, 2009); these may, however, have resulted from gastrointestinal passage instead of patent infections or a confusion with Syphacia eggs.
Lemuricola prevalences differed significantly between the two mouse lemur species. The higher prevalences in M. murinus than in M. ravelobensis might result from different sleeping site preferences.
Wooden tree holes, as preferred by M. murinus (Ehresmann, 2000;Radespiel et al., 2003), are protected from weather influences and used over longer periods, promoting reinfections and parasite exchange between cosleepers up to several weeks or even months.
In contrast, M. ravelobensis often uses more open sites or self-built leaf nests (Radespiel et al., 2003;Thorén et al., 2010), which are more exposed to weather conditions, used less frequently and for shorter periods (Radespiel et al., 2003), possibly making spreading  , 2015), a striking feature of this genus is that larvae in the environment can either develop into infective third-stage larvae (homogenic development) or initiate a free-living generation (heterogenic development), whose progeny develop into infective larvae (Zhou et al., 2019). Hosts are typically infected percutaneously and, more rarely, orally during contact with the soil. A preliminary study did not show significant differences between M. murinus and M. ravelobensis in foraging on the ground (Radespiel et al., 2006). However, M. ravelobensis are known for an opportunistic choice of sleeping sites which can even have contact to the soil, for example, under a pile of leaves or in small holes in the ground (Radespiel et al., 2003).
Such sleeping habits, even if soil-contact sites are visited only occasionally, could lead to higher infection rates in this species, but this hypothesis requires further validation. 8.8% and 9.8%) while feeding on fruits, and the respective parasite species appear to have adapted to this native Malagasy rodent. Similarly, the spirurid P. muricola was found to have adapted to the vegetarian rodent Otomys tropicalis, the tropical vlei rat (Smales et al., 2009).
However, the prevalence of the Subuluroidea fam. gen. spp. egg morphotype also differed between the three omnivorous host species and was more than doubled in the two mouse lemurs (45.7% and 53.7%) compared with R. rattus (22.1%), while the spirurid egg 1 morphotype occurred much more frequently in R. rattus (49.7%) than in the two mouse lemur species (about 8% each). Such differences in Subuluroidea fam. gen. spp. prevalence between mouse lemurs and R. rattus may be attributed to the high phylogenetic distance between lemurs and rodents. In fact, it could be shown that Subuluroidea fam. gen. spp. that infected mouse lemurs and rodents belonged to different species. In addition, differences may also be explained by different food preferences.
In contrast, it remains unclear whether mouse lemurs and rodents carried different or the same spirurid species (all carried the spirurid egg 1 morphotype), since a broad variety of Spiruromorpha species produces eggs with similar morphology (Baker, 2008). The simultaneous presence of several spirurid species was, however, confirmed for R. rattus. One likely species was Protospirura muricola, known to occur primarily in rodents in Africa, Southeast Asia, and Central and South America (Smales et al., 2009), but capable of infecting also primates (Kouassi et al., 2015;Petrzelkova et al., 2006;Smales et al., 2009). P. muricola may hence also have parasitized the mouse lemurs and E. myoxinus in our study. Smales et al. (2009) reported that occurrences of P. muricola outside Africa could always be traced back to the cosmopolitan rodents R. rattus and R. norvegicus.
The significant prevalence differences between the invasive R. rattus and the native host species may then be attributed to R. rattus, which might have introduced and may continue to spread the parasite into the ecosystem serving as a reservoir. The second species detected in R. rattus, Gongylonema neoplasticum, was in contrast reported to be restricted to rats of the genus Rattus (Setsuda et al., 2018). This species may therefore have contributed to the spirurid eggs 1 in R.
rattus, but presumably not in the other host species.
Differences in parasite species richness between the four host species may at least partially result from differences in host so-

| Effects of host population density, sex, and body condition on gastrointestinal parasites
Increasing host population density can lead to increasing parasite infection risk either indirectly by increased environmental contamination with infective stages or directly by increased interactions with conspecifics and transmission (e.g., oxyurid eggs deposited at the anus) during social contacts (Altizer et al., 2003;Chapman et al., 2006;Stringer & Linklater, 2015). However, our study did not reveal an impact of host population density on any detected gastro- becoming too large (Radespiel et al., 2001;Weidt et al., 2004). As matings are confined to a very limited time period of the year, most infections with oxyurids will take place between animals sleeping in close body contact or performing allogrooming with familiar individuals of the same social group (Eichmueller et al., 2013;Thorén et al., 2016). Under these conditions, infection risk may become mostly independent from population density, but may vary rather between different sleeping groups (Nunn et al., 2011). Regarding the soil-transmitted Strongyloides spp., developing host immunity may account for the lacking effect of population density on increased parasite infection risk. For example, experimental infections of mice with as few as six S. ratti larvae produced marked resistance to reinfections by reducing larvae excretion during challenge infection by 97% (Dawkins & Grove, 1982).
In our modeling approach, we also considered two individual host traits generally assumed to impact the parasite infection risk, host sex, and body condition. An effect of host sex was only detected in Lemuricola sp., for which some models suggested higher infection rates for male hosts than for females. Sex differences in parasite prevalence of mouse lemurs have already been demonstrated in several studies, mostly on ectoparasites. In Microcebus rufus, males were shown to be exclusively responsible for sucking louse transmission (Zohdy et al., 2012) and also exhibited higher sucking louse and nematode prevalences (Rafalinirina et al., 2007). In our study population, male M. murinus and M. ravelobensis were also found to have higher sucking louse prevalences than females . Higher parasite prevalences in males are typically explained by different sex hormone profiles, which can either stimulate (estrogens in females) or depress (androgens in males) immunity (Klein, 2004;Schalk & Forbes, 1997;Zuk & Mckean, 1996). However, the relationship between testosterone and immunity is certainly more complex, as for example positive effects of testosterone have also been reported (Ezenwa et al., 2012) and effects also seem to depend on the parasite type (Fuxjager et al., 2011). Sex differences in parasite infection risk are also often attributed to behavioral differences. Due to more extensive ranging patterns (Greenwood, 1980;Lawson Handley & Perrin, 2007;Radespiel, 2000) and more frequent "risk-taking" behaviors, males are more likely to have social encounters, which may raise pathogen transmission risk (Kraus et al., 2008;Poirotte & Kappeler, 2019;Soliman et al., 2001;Zuk & Mckean, 1996). However, these mechanisms may probably act stronger on ectoparasite than on helminth infections (Schalk & Forbes, 1997).
In our study, the spirurid egg 1 morphotype prevalence as well as GPSR showed a positive relationship between body condition and parasite infections. In general, the host's body condition is subject to seasonal and ontogenetic plasticity. This implies that body condition may influence parasite infections, but parasite infections in turn may also influence body condition. The impact of parasitism on body condition is usually attributed to the competition for nutrients between host and parasite, but also to tissue damage resulting in organ malfunction and protein loss (Holmes, 1985). A dwindling food intake due to loss of appetite has also been associated with parasite infections (Arneberg et al., 1995;Fox, 1997;Ghai et al., 2015). Conversely, hosts in better body condition should rather be able to mobilize resources for the defense against pathogens than hosts in a worse situation (Bonneaud et al., 2003;Martin et al., 2003;Ujvari & Madsen, 2006). Both explanations would result in lower parasitism in animals with better body condition. The results of our study, however, correspond rather to the findings of Rafalinirina et al. (2007), who demonstrated that M. rufus in better body condition exhibited higher gastrointestinal parasite and ectoparasite prevalences. The ability of good quality hosts to sustain higher parasite loads was suggested by the author to explain these results. This "well-fed host hypothesis" should, however, be especially applicable to ectoparasites (Christe et al., 2003;Hawlena et al., 2005). Due to an active infestation compared with the more passive infection mode by ingesting gastrointestinal parasites, ectoparasites could be attracted to hosts in good body condition (Christe et al., 2007). Instead, explanations for gastrointestinal parasites rather involve that such hosts might have a higher feeding capacity, increasing the probability of oral pathogen intake. In addition, older individuals may accumulate parasites over time if no protective immunity develops (Bellay et al., 2020).

| Edge effects, fragmentation responses, and the impact of vegetation parameters on gastrointestinal parasites
Six environmental factors were used to infer effects of habitat fragmentation and structure on gastrointestinal parasites ( Figure 6).
While four factors (forest category, forest size, distance to the forest edge, percentage of edge habitat) are directly related to habitat fragmentation, the two others provide estimates of vegetation structure and human disturbance (i.e., vegetation clearance, forest maturation). The homoxenous oxyurids Enterobiinae gen. sp. and Lemuricola spp. were neither affected by habitat fragmentation parameters or edge effects nor vegetation clearance, which may be explained by their egg deposition at the host's anus (Taffs, 1976), providing a stable microenvironment for parasite development and survival despite altering macroenvironmental conditions. Nevertheless, the significant impact of forest maturation on prevalence of two oxyurids (Enterobiinae gen. sp., Lemuricola sp.) suggests that they are not entirely independent from environmental conditions. Intraspecies transmission occurs by oral uptake of infective eggs during interindividual contacts, mostly at highly frequented sheltered sleeping sites (Baker, 2008;Irwin & Raharison, 2009 (Chen et al., 1992;Laurance, 1991;Murcia, 1995;Nelson & Halpern, 2005). In the study region, the northwestern Malagasy dry forests, it has been demonstrated that tree stem density was lower in proximity to the edge compared with the forest interior (Malcolm et al., 2016). Such differences in vegetation structure may in turn lead to different levels of protection from weather influences like solar radiation or wind (Foggo et al., 2001). Resulting differences in environmental buffering effects can cause differences in microclimates at the edge compared with the forest interior, which may constrain or facilitate species survival or reproduction (Gehlhausen et al., 2000).
Here, the Strongyloides morphotype proved to be most vulnerable to environmental conditions as it was impacted by most habitatrelated factors indicative for forest fragmentation and degradation.
This parasite genus showed significantly higher prevalences in continuous than in fragmented forests, and parasite infection risk increased with increasing forest size. Intriguingly, there was also a positive impact of the distance to forest edge and a negative effect of vegetation clearance on parasite infection risk which both suggest strong negative ecological edge effects on Strongyloides spp.
These impacts are most likely related to the complex life cycle of Strongyloides spp., which are homoxenous parasites but can undergo heterogenic development with a free-living generation in the environment (Baker, 2008;Eberhardt et al., 2007;Viney, 1999;Viney & Lok, 2015). Adverse influences of forest edges and vegetation clearance can be assumed for both homogenic and heterogenic cycles in terms of direct (e.g., ultraviolet damage or desiccation of developmental stages) and indirect (e.g., insufficient soil moisture) negative impacts on survival or reproduction, respectively.
Compared with Strongyloides spp., the heteroxenous Subuluroidea fam. gen. spp. and spirurid egg 1 morphotypes spend large parts of their life cycles outside the definitive hosts in an arthropod intermediate host (Baker, 2008;Irwin & Raharison, 2009 shown for some other arthropods, namely ectoparasites of the same host species (e.g., ticks or different mites, Kiene et al., 2020). The authors related these findings, among others, to abiotic factors, as ultraviolet radiation or humidity differ between the edge and interior of a forest (Kapos, 1989;Kiene et al., 2020;Murcia, 1995

| CON CLUS IONS
This study revealed a variety of impacts on gastrointestinal parasite infections that were partly host-and partly environment-related.
Thus, the initial expectation of gastrointestinal endoparasites being less susceptible to environmental changes than previously studied ectoparasites was not supported. Whereas homoxenous are suggested to provide vital ecosystem services, for example, by stabilizing a high host species diversity through controlling effects on common or invasive species which may otherwise outcompete rarer native species (Lafferty, 2003;Mouritsen & Poulin, 2005). In that sense, a high parasite diversity can be regarded as a sign of a healthy ecosystem (Hudson et al., 2006). In conclusion, this study shows that habitat fragmentation in northwestern Madagascar has negative effects on the native gastrointestinal parasite communities of native small mammals and even the invasive R. rattus.
Further research will be needed to clarify the underlying causal effects, to evaluate the host-parasite networks in fragile fragmented environments in more depth, starting by clarifying the taxonomy and the specific life cycles of the different parasite species. Overall, the results demonstrate that forest fragmentation should not only be regarded as a threat to the diverse suite of host species that inhabit such habitats Steffens & Lehman, 2018) but also to their suite of often highly adapted and coevolved parasites. Future conservation planning should take these complex evolutionary relationships and habitat requirements of native parasites into account, since they may be even more vulnerable than their hosts.

CO N FLI C T O F I NTE R E S T
No conflict of interest has been declared by the authors.

DATA AVA I L A B I L I T Y S TAT E M E N T
The modeling data are publicly accessible in the Zenodo repository