Land-cover and climate factors contribute to the prevalence of the ectoparasitic fungus Laboulbenia formicarum in its invasive ant host

Understanding the distribution of parasites is crucial for biodiversity conservation. Here, we studied the distribution of the ectoparasitic fungus Laboulbenia formicarum in native and invasive Lasius ants in a 2000 km 2 area. We screened over 16,000 ant workers in 478 colonies of ﬁ ve different species. We found that Lab. formicarum was rare in native Lasius species but infected 58% of the colonies of the invasive species Las. neglectus . At landscape scale, Lab. formicarum presence could not be explained by geographic and genetic distances between Las. neglectus colonies but was associated with hotter and dryer climatic conditions and its prevalence in colonies increased with urbanization. Within infected colonies, fungal prevalence varied from 0 to 100 percent within meters and was negatively correlated with impervious ground cover. In a changing world, our ﬁ ndings emphasize the importance of land-use and climatic factors in shaping the distribution and prevalence of fungal parasites.

Ectoparasites are ideal models to study the factors shaping the distribution of parasites. They live on the external body envelope of other organisms and are thus particularly exposed to environmental conditions in addition to being easily detected (Hopla et al., 1994;De Kesel, 1996;Kołodziej-Soboci nska, 2019). Laboulbeniales (Ascomycota; Laboulbeniomycetes; Laboulbeniales) are one of the largest groups of ectoparasitic fungi, with about 2325 species described in 145 genera (Haelewaters et al, 2020(Haelewaters et al, , 2021Kirk, 2019;Reboleira et al., 2018;Rossi and Santamaría, 2012). They are obligate ectoparasites of arthropods and live attached to the cuticle of a wide variety of insects and of a few other taxa, including mites, harvestmen and myriapods Santamaria et al., 2017Santamaria et al., , 2020Seeman and Nahrung, 2000). Laboulbeniales form thalli that can cover the entire body of their hosts and may penetrate through the cuticle (Tragust et al., 2016). Transmission usually occurs via spores upon direct contact between conspecifics (e.g. during mating; De Kesel, 1995;Knell and Webberley, 2004). Laboulbeniales are commonly found in ant species (Hymenoptera; Formicidae; Santamaria and Espadaler, 2015). To date, six Laboulbeniales species are known to parasitize 43 ant species from ten genera (Santamaria and Espadaler, 2015). However, little is known of the factors determining the spatial distribution and prevalence of Laboulbeniales in ants (Haelewaters et al., 2015b;Szentiv anyi et al., 2019). Laboulbeniales are often assumed to have adapted to the ecological niche of their hosts (De Kesel, 1996), to thrive best in densely packed host populations (De Kesel, 1993) or to have an affinity for moist habitats (Santamaria and Espadaler, 2015;Mark o et al., 2016;Kołodziej-Soboci nska, 2019). Large-scale climatic variations affect the probability of infection of the ant Myrmica scabrinodis by the laboulbenian fungus Rickia wasmanni (Szentiv anyi et al., 2019), but it is not known whether landscape-and localscale environmental conditions, such as elevation and land cover type, affect the distribution or infection success of the Laboulbeniales that parasitize ants.
To understand what determines Laboulbeniales' spatial distribution at landscape-and local-scale, we studied the ectoparasitic fungus Laboulbenia formicarum, that parasitizes Lasius ants, including one of the most widespread invasive ant species in Europe, Lasius neglectus (originating from Asia Minor; Herraiz and Espadaler, 2007;Ugelvig et al., 2008;Blatrix et al., 2018). Over 300 introduced populations of this species have been detected so far in Europe (Gippet et al., 2017;Espadaler and Bernal, 2020), of which four are known to be infected by Laboulbenia formicarum (Herraiz and Espadaler, 2007;Espadaler et al., 2011). The native range of Lab. formicarum is still unknown and has been the subject of contrasting hypotheses. First, the fungus could have been introduced in Europe recently. A possible origin is North America, as suggested by the spatial and temporal distribution of records for the species in both continents (Espadaler and Santamaria, 2012). Another origin could be any part of Las. neglectus native range if both organisms were co-introduced in Europe, with Lab. formicarum being lost in most Las. neglectus colonies during the invasion process e reduced parazitation in introduced populations is indeed observed in many taxa (Torchin et al., 2003). Alternatively, Lab. formicarum may be native to the European ant fauna and might have found a new suitable host species in invasive Las. neglectus ants.
To first assess whether Lab. formicarum is common in native Lasius species, or only occurs in Las. neglectus, we screened 412 colonies from four native Lasius species and 66 colonies of the invasive species Lasius neglectus (Van Loon et al., 1990) sampled across the landscape of the middle Rhône valley in France (~2000 km 2 ; Figs. 1e3).
We then focused on the invasive ant Lasius neglectus. Because dispersal is crucial in shaping species spatial distribution (Clobert et al., 2012), we tested the importance of horizontal and vertical transmission in explaining Lab. formicarum presence across the Las. neglectus colonies occurring in our study landscape. In Las. neglectus, colonies can extend over several hectares and are composed of multiple nests connected by trails Ugelvig et al., 2008). Horizontal transmission occurs if Las. neglectus colonies transmit the fungus to each other via direct contact of their workers or reproductive individuals (males and females) or indirectly via vectors (e.g., commensals, other parasites). Geographically close colonies should thus be more likely to infect each other if horizontal transmission is an important driver in this host-parasite system. Lasius neglectus is dispersed throughout landscapes (and continents) via the transport of potted plants, soil or construction material (Ugelvig et al., 2008). Lab. formicarum spread could therefore occur vertically, when a portion of an infected colony is transported to a new location through human activities. Under this scenario, genetically close colonies should be more likely to be infected with Lab. formicarum than genetically distant colonies.
We then tested whether the presence and prevalence of Lab. formicarum were associated with environmental factors linked to climate (mean annual temperature and precipitation), land cover (vegetation cover, agriculture, urbanization) and topography (elevation and solar radiation). Finally, in 16 infected Lasius neglectus colonies, we tested whether within-colony spatial variation in Lab. formicarum prevalence was affected by local land cover types (open vegetation, forest, croplands, unsealed ways and impervious surfaces).

Study area
The study area is a 2000 km 2 zone located in South-East France, in the city of Lyon and its surrounding suburban and rural areas. Lyon is the second largest French metropolitan area after Paris. The area is characterized by a temperate climate with Mediterranean influences. This area is heavily invaded by the ant Lasius neglectus (Gippet et al., 2017(Gippet et al., , 2018.

Datasets
We used two different datasets to study the spatial distribution of Lab. formicarum. The first dataset is a sampling of native and invasive Lasius ants throughout the study landscape (1248 locations; Methods section 2.2; Table 1, Figs. 2 and 3). The second dataset focuses on 16 colonies of Las. neglectus infected by Lab. formicarum. In each of these colonies, several nests or trails were sampled in order to assess local (i.e. intra-colonial) spatial variation in fungus prevalence (Methods section 2.3; Figs. 3 and 5).

Sampling of native and invasive Lasius ants
In the study landscape, a total of 1248 locations were sampled during spring and summer 2011, 2012 and 2013. Sampling locations consisted of haphazardly selected patches with vegetation, generally close to or along roadsides on public land. Sampling locations were separated by at least 200 m in dense urban areas and by at least 500 m in suburban, residential and rural areas. Sampling was done by directly searching ant nests and trails on the ground, trees and shrubs. Samples were collected by hand using custom entomological aspirators. Each time a trail or nest of Lasius ants was discovered, ants were sampled. We considered that each sample corresponded to a unique ant colony, except for Las. neglectus because in this species, all nests and trails occurring locally are interconnected and belong to the same colony. Thus, if different samples of Las. neglectus were collected in the same sampling location, they were pooled together for analyses. All samples were stored in 96% ethanol at À20 C. Ants were then identified to species level using morphological criteria (Seifert, 2007). Additional samples of Las. neglectus and native Lasius colonies were obtained from the local-scale ant sampling (see methods in section 2.3) and collated to this dataset.

Laboulbenia formicarum prevalence in native and invasive Lasius ants
Only samples containing at least 10 individuals were screened for the presence of Lab. formicarum. A total of 16,779 workers from 478 different colonies were screened for Lab. formicarum presence (Table 1), including 230 colonies of Las. niger, 118 of Las. alienus, 39 of Las. paralienus, 25 of Las. emarginatus and 66 of Las. neglectus (see Table 1 for details on the number of workers screened by colony). Workers were carefully examined under a stereomicroscope at 50Â magnification and were considered infected if at least one Lab. formicarum thallus was observed on a single ant's cuticle (Fig. 1).

Geographic and genetic distance between Lasius neglectus colonies
A matrix of geographic distances among all 66 Las. neglectus colonies was calculated using the 'dist' function from the stats package in R v.3.6.2 (R core team, 2019). To assess the genetic distance between colonies, a total of 793 workers from 33 Las. neglectus colonies (mean ± standard deviation: 24 ± 3.6 workers per colony; range: 14e36) were genotyped at 12 microsatellite markers (see supplementary Material and Methods and Table S1 for details). We calculated the genetic distances between Las. neglectus colonies as Fst/(1-Fst) using the 'pairwise.fst' function from the hierfstat package in R (Goudet and Jombart, 2015).

Landscape scale environmental factors
Seven climatic, land cover and topographical variables were compiled to test their association with Lab. formicarum presence and prevalence in Las. neglectus colonies: (i) mean annual temperature (in C;~1 km resolution), (ii) mean annual precipitation (in mm;~1 km resolution), (iii) elevation (25 m resolution), (iv) amount of solar radiation (in kWh.m À2 ) estimated from the elevation map (i.e. 25 m resolution) for June to August 2013 with ArcGIS 10.1 (default parameters in Spatial Analyst Tools), v) the proportion of vegetated land cover (2.5 m resolution), vi) the proportion of impervious land cover (2.5 m resolution) and vii) the proportion of cultivated land cover (vector data) (see Table 2 for more information on variables and their sources). For each variable (except climatic variables), average values were calculated in a 100 m zone around the centre of the sampling locations invaded by Las. neglectus. We computed the Euclidean environmental distance between locations invaded by Las. neglectus using the 'dist' function from the stats package in R.

Statistical analyses
We used the 'dist' function of the stats package in R to construct a binary infection status distance matrix between 66 Las. neglectus colonies. Pairs of colonies that were both infected by Lab. formicarum or both non-infected were assigned a distance of '0', and pairs of colonies with one infected and one non-infected colony were assigned a distance of '1' (following Gilbertson et al., 2016). A Mantel test with 10,000 permutation was then performed using the 'mantel.rtest' function (R package ade4; Dray and Dufour, 2007) to test whether Las. neglectus colonies with the same infection status were geographically closer to each other than expected from a random spatial distribution. A second Mantel test with 10,000 permutations between the infection and genetic distance matrices (N ¼ 33 colonies) was performed to test whether Las. neglectus colonies with the same infection status were genetically more similar. A third Mantel test between the infection and environmental distance matrices (N ¼ 66 colonies) was performed to determine whether Las. neglectus colonies with the same infection status occurred in more similar environmental conditions than random. To test if the infection of Las. neglectus colonies Lab. formicarum was associated with specific environmental conditions, we used a generalized linear model (GLM) with binomial link function (R package stats; N ¼ 66 colonies). Because the five environmental variables compiled were not independent from each other (especially land cover variables that are mutually exclusive), we summarized the five environmental variables into artificial uncorrelated variables using a Principal Component Analysis ('dudi.pca' function in R package ade4). We then used the axes of the PCA as explanative variable in the binomial GLM.
Finally, considering infected colonies only, we tested if prevalence, expressed as the proportion of infected workers in the colony, was associated with environmental conditions using a GLM with quasibinomial link function and weighted by the log number of workers screened (R package stats; N ¼ 38 colonies). For this GLM, we also summarized our five environmental variables using a PCA and used the PCA axes as explanatory variables. The coefficients of determination (Nagelkerke's pseudo-R 2 ) of the models were estimated using the function 'r2_nagelkerke' from the performance package in R (Ludecke et al., 2019).

Measurement and sampling of infected Lasius neglectus colonies
To study if and how Lab. formicarum prevalence varied locally, within the extent of infected Las. neglectus colonies, we measured the surface area occupied by 16 colonies (out of the 38 infected colonies detected in the landscape; see Fig. 3) and sampled workers from several nests and trails within each colony (see section 2.3.2 for details). Colonies measurements were performed during spring and summer 2012 and 2013 by teams of two to five persons, and ants were detected by searching for trails and nest entrances visually. Workers were sampled every 20e40 m depending on land access, and each sample was georeferenced precisely. Colony boundaries were defined when no more Las. neglectus were found in a 50 m radius from the last location where Las. neglectus were detected. Las. neglectus occurrences were mapped with ArcGIS 10.1 (ESRI, Environmental Systems Research Institute, Redlands, 2012).

Laboulbenia formicarum prevalence within infected Lasius neglectus colonies
Depending on the extent of Las. neglectus colonies, 5 to 50 nests (or trails) were sampled (mean ± s.d. ¼ 13 ± 10 samples by colony; total number of samples ¼ 219). Samples contained between 6 and 106 workers (mean ± s.d. ¼ 20 ± 11 workers by sample). A total of 4286 workers were screened. For each sample, workers were carefully examined under a stereomicroscope at 50Â magnification and were considered infected if at least one Lab. formicarum thallus was observed on an ant's cuticle (Fig. 1). These samples were also used in the landscape-scale analyses (pooled by colony).

Local-scale environmental factors
To assess variations in land cover within the extent of colonies, satellite images were obtained for the June 1, 2012 from Google Earth Pro v7.3.2.5776, saved individually, and 5 m radius circles around the sampling points (i.e. nest or trail) were drawn. The  proportions of the circles corresponding to four different land cover types (tree cover, open vegetation, impervious surface and unsealed ways) were measured using ImageJ v1.52 (Schneider et al., 2012) (see Table 2 for more information on variables).

Statistical analyses
The effect of local land cover on the prevalence of Lab. formicarum was tested using a general linear mixed model (GLMM) with a binomial link function and colony identity as random effect (R package lme4; Bates et al., 2015). Land cover variables were summarized using a PCA, with the PCA axes used as explanatory variables. We determined the best-fitting model using a backward model selection procedure based on sequential one-term deletions using Chi-square tests ('drop1' function in R package stats; only additive models were considered) and a significance threshold of 0.05. The coefficient of determination (Nakagawa's pseudo-R 2 ) of the model was estimated using the function 'r2_nakagawa' from the performance package in R.

Presence of Lab. formicarum in Lasius neglectus and native Lasius species
The ectoparasitic fungus Lab. formicarum was detected in 58% (38 of 66) of Las. neglectus colonies (Figs. 3) and 5% (11 of 230) of Las. niger colonies screened (Table 1). The fungus was not detected in any of the other three Lasius species (Fig. 2). On a total of 230 colonies of Las. niger screened, 134 were sampled during the landscape-scale survey (i.e., among the 1248 randomly selected sampling locations) and 96 were sampled during the measurement of Lasius neglectus colonies (i.e., these colonies adjacent to infected Las. neglectus colonies). Lab. formicarum prevalence was significantly different between these two sets of colonies: 1.5% (2 colonies infected on 134 screened) for the first one and 9.4% (9 colonies infected on 96 screened) for the second one (Chi-square test: c 2 ¼ 3.87, P ¼ 0.049).

Landscape-scale variations in
Laboulbenia formicarum presence and prevalence 3.1.2.1. Laboulbenia formicarum presence. There was no clear correlation between the infection status of Las. neglectus colonies and geographic distance (Mantel test: observed correlation ¼ 0.04, P ¼ 0.058) or genetic distance (Mantel test: observed correlation ¼ 0.05, P ¼ 0.12) between colonies. There was, however, a significant correlation between infection status and environmental distance (Mantel test: observed correlation ¼ 0.08, P ¼ 0.006), which indicates that ant colonies in similar environments were more likely to have the same infection status. A first PCA was performed with the seven landscape-scale environmental variables and all 66 Las. neglectus colonies. The first PCA axis explained 38.1% of the total variability and was associated with high mean annual temperature, low mean annual precipitation, low elevation and the absence of agricultural areas. The second PCA axis explained 26.7% of the total variability and was associated with high vegetation and low impervious cover (Fig. 4A). The probability of being infected by Lab. formicarum was positively correlated to the first axis of the PCA (Estimate ¼ 0.38 ± 0.17, z ¼ 2.2, P ¼ 0.027; R 2 ¼ 0.11; Fig. 4A), suggesting that ant colonies were more likely to be infected in areas characterized by high mean temperature, low mean precipitation, low elevation and low agricultural surfaces.

Laboulbenia formicarum prevalence.
A second PCA was performed with the same seven landscape-scale environmental variables, but with the 38 infected Las. neglectus colonies only. The first PCA axis explained 34.2% of the total variability and was associated with high mean annual temperature, low mean annual precipitation, low elevation and a small proportion of agricultural areas. The proportion of infected workers was negatively associated with the first PCA axis, although not significantly (Estimate ¼ À0.36 ± 0.19, z ¼ À1.9, P ¼ 0.06). The second PCA axis explained 27.9% of the total variability and was associated with high impervious and low vegetation covers (Fig. 4B). The proportion of infected workers was positively associated with the second PCA axis (Estimate ¼ 0.58 ± 0.17, z ¼ 3.4, P ¼ 0.002; R 2 ¼ 0.55; Fig. 4B), indicating that the prevalence of the fungus was positively associated with urbanization.

Local-scale analyses
A PCA was performed with the five local environmental variables and all 219 Las. neglectus nests (or trails) sampled across the 16 infected colonies measured. The first PCA axis explained 43.5% of the total variability and opposed high tree cover to open areas (i.e. impervious and open vegetation). The second PCA axis explained 31.8% of the total variability and opposed open vegetation and impervious surfaces (Fig. S1). In infected colonies, the proportion of infected workers in nests could vary from 0 to 100% within a few meters (Fig. 5) and was negatively associated with the second axis of the PCA (Estimate ¼ À0.26 ± 0.04, z ¼ À7.1, P < 0.0001) suggesting that prevalence was negatively associated with ground imperviousness (Fig. S1). However, the proportion of variance explained by this variable was very low (marginal R 2 ¼ 0.012) as most of the explained variation was linked to colony identity (i.e. the random factor of the mixed model; conditional R 2 ¼ 0.55).

Discussion
We screened over 16,500 individual Lasius ants from 478 colonies to detect the ectoparasitic fungus Lab. formicarum and understand how local and landscape-scale environmental conditions affect its distribution. The fungus was present but uncommon in colonies of Lasius niger, absent in four other native Lasius species, and common in the nests of the invasive ant Las. neglectus. At the scale of the landscape, the presence of Lab. formicarum in Las. neglectus colonies was positively associated with low elevation, the absence of agriculture and dry and warm environments. Its prevalence in infected colonies was positively associated with urbanization. The prevalence of the fungus also varied spatially at the scale of the colony and was negatively linked to impervious surfaces. We did not detect Lab. formicarum in colonies of Lasius emarginatus, Las. alienus and Las. paralienus and the fungus was present in only 5% of the sampled colonies of Lasius niger (11 colonies infected among 230 screened). However, nine out of these eleven infected colonies were near infected Las. neglectus colonies. The prevalence of Lab. formicarum in Las. niger is six times higher when the species occurs near infected Las. neglectus (9.4% versus 1.5% when randomly sampled in the landscape). Laboratory experiments have shown that infected Las. neglectus can transmit Lab. formicarum to Las. niger (Tragust et al., 2015). Our findings suggest that cross-species transmission occurs between these two species in natural settings and that Las. neglectus might constitute a reservoir for Lab. formicarum to spill over the native species Las. niger.
Fifty-eight percent of Las. neglectus colonies were infected by the fungus (38 out of 66 colonies, Table 1). This was higher than expected from the literature, as Lab. formicarum had only been reported in four colonies of Las. neglectus in Europe, despite extensive sampling and monitoring (Herraiz and Espadaler, 2007;Espadaler et al., 2011;Espadaler and Bernal, 2020). The most extensive study to date screened nearly 5000 workers from 21 Hungarian Las. neglectus colonies without detecting Lab. formicarum . These results are consistent with the hypothesis of a recent introduction of Lab. formicarum in Western Europe (Espadaler and Santamaria, 2003). However, we cannot exclude that Lab. formicarum may be a native and widespread, albeit not abundant parasite of European ants. Among the randomly sampled Las. niger colonies, 1.5% were infected by Lab. Formicarum, which is in line with infection rates found in two native ant-parasitic Laboulbeniales in Europe (B athori et al., 2014, 2015). Lasius niger is an extremely abundant ant species (Gippet et al., 2017) and a suitable host to Lab. formicarum. Laboulbenia formicarum could thus be a native parasite that regularly jumps from native ant species to invasive Las. neglectus colonies. This scenario was described in the Laboulbeniales fungus Hesperomyces virescens, a parasite that occurs at low prevalence in native ladybirds but that is common in the invasive ladybird Harmonia axydris (Ceryngier and Twardowska, 2013). Similarly, Las. neglectus may be a natural host for Lab. formicarum, both of them possibly cointroduced in some areas across Europe. Establishing the genetic profiles of North American and European populations of Lab. formicarum might help understand the origin and colonization history of the fungus (Haelewaters et al., 2015a).
We found no clear evidence that the geographic proximity between Las. neglectus colonies was associated with infection status. We expected, under a horizontal transmission scenario, that geographically closer colonies would have more similar infection status. For example, sexual transmission could occur if spores or thalli are dispersed by reproductive ant individuals, although young reproductive female and male ants do not appear to bear Laboulbeniales thalli (Haelewaters et al., 2015b). Sexual transmission is also unlikely in Las. neglectus, because this species rarely or never performs nuptial flights: females seem to mate with males from the same colony  although we witnessed males taking off from an infected colony, see Fig. 1). Cross-infection between spatially close Las. neglectus supercolonies cannot be ruled out in the very rare cases where separate colonies are not kilometres apart (Fig. 3); only one such instance is known to us, where two genetically distinct colonies are separated by a broad boulevard. Similarly, it has also been suggested that horizontal transmission may occur via ant-associated 'myrmecophilous' invertebrates (Santamaria and Espadaler, 2015), but again, Las. neglectus colonies are too distant to make such events likely. Relationship (mean ± 95% CI) between environmental conditions (PCA axis 1) and the probability of Las. neglectus colonies to be infected by Lab. formicarum (N ¼ 66 colonies, observed values are 0 or 100% but slightly shifted for visual purpose). As illustrated by the bottom-right circle of correlations, the first PCA axis represents a gradient of temperature and precipitation. (B) Relationship (mean ± 95% CI) between environmental conditions (PCA axis 2) and Lab. formicarum prevalence in infected Las. neglectus colonies (N ¼ 38 colonies). As illustrated by the bottom-right circle of correlations, the second PCA axis represents a gradient of urbanization as it is negatively correlated with the proportion of vegetated areas and positively correlated to the proportion of impervious surfaces (e.g., buildings, roads).
We expected vertical transmission to explain the distribution of Lab. formicarum, but there was no evidence that genetic proximity between Las. neglectus colonies was associated with infection status. This surprising result suggests the vertical transmission of Lab. formicarum is not systematic. Vertical transmission may be uncommon if human-mediated dispersal is detrimental to Lab. formicarum (Gippet et al., 2019). Humans may also propagate uninfected portions of infected colonies because the intra-colonial prevalence of Lab. formicarum is extremely variable (Fig. 5). Finally, Lab. formicarum may disappear over time if the environmental conditions at the place of introduction are not favourable (Las. neglectus colonies kept in laboratory conditions lose the fungus in a few months; S. Tragust, unpublished data). Altogether, these results question the importance of both horizontal and vertical transmission and suggest that environmental limitation is a stronger determinant of Lab. formicarum distribution.
Our landscape-scale analysis showed Lab. formicarum presence was associated with warmer and dryer climatic conditions in low elevation areas. It has been hypothesized that humidity should favour Rickia wasmanni, a fungus that parasitizes Myrmica ants, because these ants live in moist environments (Santamaria and Espadaler, 2015). However, Haelewaters et al. (2015b) found no such trends across three distinct habitats, Mark o et al. (2016) found no difference in fungal prevalence within Myrmica colonies from dry and humid sites and Szentiv anyi et al. (2019) found that Rickia wasmanni was more common in colder and dryer areas (Szentiv anyi et al., 2019). In addition, the most recently described ant-associated Laboulbeniales, Rickia lenoirii, was described from ants in the genus Messor, which live in dry or arid habitats (Santamaria and Espadaler, 2015). Together with the literature, our results suggest ant-associated Laboulbeniales prefer warm and dry climates.
Fungal prevalence was also negatively associated with agriculture (Fig. 4A, first PCA axis). High concentrations of fungicides are commonly found in the soil and water surrounding crops (Zubrod et al., 2019). Laboulbeniales, including species associated with ants, are sensitive to fungicides (Gemeno et al., 2004;but see Pech and Heneberg, 2015), and the contamination of agricultural areas by fungicides leaching into the environment may explain why Las. neglectus colonies located near crops were less infected by Lab. formicarum.
When focusing only on infected Las. neglectus colonies (N ¼ 38) we found that urbanization was positively associated with fungal prevalence (Fig. 4B). A similar association between urbanization and Laboulbeniales prevalence was reported in Hesperomyces virescens parasitizing native ladybirds in the UK (Welch et al., 2001). It was suggested that urbanization increased the overlapping time of successive ladybug generations, increasing the probability that new fungus-free cohorts would mate with older infected individuals (Welch et al., 2001;Knell and Webberley, 2004). This mechanism could not explain our observations in ants because new workers are produced all year long (often with pulses of production in spring and fall and a diapause in winter; H€ olldobler and Wilson, 1990). Environmental changes associated with urbanization, like increased heat or pollution (Grimm et al., 2008), may benefit the fungus, either directly by changing local environmental conditions or indirectly by altering Las. neglectus immunity or behaviour (Youngsteadt et al., 2015).
Finally, we found that the prevalence of Lab. formicarum was highly variable within colonies (from 0 to 100% within meters). It was negatively associated with impervious surfaces such as roads and buildings (Fig. S1). This correlation was weak, and differed from landscape-scale analysis, suggesting a scale-dependent relationship between Laboulbeniales prevalence and environmental conditions. The prevalence of ant-associated Laboulbeniales may also vary with time. At the individual level, the number of thalli of the Laboulbeniales Rickia wasmanni increases with the age of its hosts (i.e. Myrmica scabrinodis ants; B athori et al., 2018). The high variability in numbers of Laboulbeniales thalli we observed within ant colonies may result from a heterogeneous spatial distribution of age cohorts within the nest. Such spatial age structures may originate in large ant colonies because ants gather and move their brood to optimize development (generally hot and dry places for pupae; H€ olldobler and Wilson, 1990). The prevalence of Lab. formicarum also increases with time in Las. neglectus colonies (Tragust et al., 2015). Temporal fluctuations in the ectoparasite prevalence may complexify the relationship between Laboulbeniales and their hosts. Repeated sampling may be needed to further our understanding of ant-Laboulbeniales interactions (Haelewaters et al., 2015b).
Overall, our results show that environmental conditions and land use play an important role in shaping the distribution of antassociated Laboulbeniales. Improving our understanding of this role might help predict current and future distribution of fungal parasites in a changing world. This knowledge will be crucial to protect endangered or important flora and fauna from threatening fungal parasites, and to control pests and invasive species.

Authors contribution
JMWG, TC, ST and BK designed the study. JMWG, TC, and BK morphologically identified the ants. AD performed DNA extraction and genotyping and BK processed microsatellite raw data. FW, MH, TC and JMWG screened the ants for Lab. formicarum. JMWG, JG and BK measured Las. neglectus colonies in the field. JMWG and TC processed the data. JMWG performed statistical analyses. JMWG, TC, ST, NM and BK wrote the first draft of the manuscript and all coauthors participated in improving the subsequent versions.

Availability of data and materials
Datasets supporting the conclusions of this article are included in Supplementary Material and Methods, Table S1, Fig. S1 and supplementary datasets (available at: https://github.com/JGippet/ Datasets- Gippet-et-al.-2021-Fungal-Ecology).

Declaration of competing interest
The authors declare that they have no competing interests.

Acknowledgments
ST, JMWG, BK would like to thank the German Academic Exchange Service (DAAD, project number: 57445489) and the French Programme Hubert Curien (PHC Procope 42518QL) for funding cooperation between University of Halle and University of Lyon. The study was funded by the D epartement de l'Is ere. It was also supported by the French National Research Agency (ANR) through the LABEX IMU (ANR-10-LABX-0088) of Universit e de Lyon, within the program Investissements d'Avenir (ANR-11-IDEX-0007). We warmly thank St ephanie Mermet and all the great interns who participated in sampling and counting ants. We thank the two anonymous reviewers that helped improve the manuscript.