Ecology and population genetics of the parasitoid Phobocampe confusa (Hymenoptera: Ichneumonidae) in relation to its hosts, Aglais species (Lepidoptera: Numphalidae)

The biology of parasitoids in natural ecosystems remain very poorly studied, while they are key species for their functioning. Here we focused on Phobocampe confusa, a vanessines specialist, responsible for high mortality rates in very emblematic butterfly species in Europe (genus Aglais). We studied its ecology and genetic structure in connection with those of its host butterflies in Sweden. To this aim, we gathered data from 428 P. confusa individuals reared from 6094 butterfly larvae (of A. urticae, A. io and in two occasions of Araschnia levana) collected over two years (2017 and 2018) and 19 sites distributed along a 500 km latitudinal gradient. We found that P. confusa is widely distributed along the latitudinal gradient. Its distribution is constrained over time by the phenology of its hosts. The large variation in climatic conditions between sampling years explains the decrease in phenological overlap between P. confusa and its hosts in 2018 and the 33.5% decrease in the number of butterfly larvae infected. At least in this study, P. confusa seems to favour A. urticae as host: while it parasitized nests of A. urticae and A. io equally, the proportion of larvae is significantly higher for A. urticae. At the landscape scale, P. confusa is almost exclusively found in vegetated open land and near deciduous forests, whereas artificial habitats are negatively correlated with the likelihood of a nest to be parasitized. The genetic analyses on 89 adult P. confusa and 87 adult A. urticae using COI and AFLP markers reveal a low genetic diversity in P. confusa and a lack of population genetic structure in both species, at the scale of our sampling. Further genetic studies using high-resolution genomics tools will be required to better understand the population genetic structure of P. confusa, its biotic interactions with its hosts, and ultimately the stability and the functioning of natural ecosystems.

dynamics of their hosts and have their own habitat requirements [4]. However, little empirical evidence exists to adequately inform these processes and our knowledge of the ecology of most parasitoids is often based on sparse data obtained from a few randomly captured specimens [4]. The lack of data sets the weather conditions, with larvae observed in the field from May to the end of August. Populations of A. io are univoltine in Sweden and their phenology is slightly more restricted than for A. urticae, with  Phenological synchrony between P. confusa and its hosts on all the samplings on the site j (eq. 3). The overlap index (OPH) is a parsimonious measure of the phenological overlap under the hypothesis that the parasitoid does not benefit from a surplus of resources [21]. The phenological overlap between species is calculated only when the two species, 145 namely P. confusa and each of its hosts, were sampled at a site within a given year.

Pattern of attack 148
We investigated differences in P. confusa attack rates on its two main host butterflies, A. urticae and

149
A. io, in two ways. First, we studied the proportion of butterfly nests parasitized by P. confusa. This 150 analysis was restricted to butterfly nests sampled within the temporal window of occurrence of P.

151
confusa (see Table S1) and at sites where P. confusa was observed at least once in the season (n = 359 152 nests). Second, we examined the proportion of larvae parasitized for each nest parasitized by P. confusa 153 (n = 145 nests). The proportion of butterfly nests parasitized and the proportion of larvae parasitized 154 by P. confusa per nest were modelled with a binomial error distribution.

155
We analysed variations in parasitism rates according to butterfly host, region, larval instar at 156 collection, the phenological overlap, the year and week of collection, and the total number of butterfly preliminary exploration of the data, we included a quadratic term for the sampling week and 159 phenological overlap. We also included the two-way interactions between the butterfly host and the 160 region, the year, the larval instar at collection, the total number of butterfly nests at sampling, and the 161 two-way interaction between region and year. Because few nests were collected at 1 st instar, we pooled 162 them with nests collected at 2 nd instar. Model selection followed a backward elimination procedure.

163
Model diagnostics were assessed using the R package DHARMa [22].

165
We examined how habitat heterogeneity and fragmentation influenced the distribution of P.

166
confusa. Using the models selected in the analyses of the proportion of butterfly nest parasitized and 167 the proportion of larvae parasitized by P. confusa per nest (see above), we estimated the additional 168 variance explained when including land cover variables. In the analyses, absences were informed by 169 including data on butterfly nests collected at sites where P. confusa was not observed (n = 31), but that 170 were sampled during its period of activity (Table S1). Land cover heterogeneity was modelled as the 171 percentage of arable land, vegetated open land (e.g. field, meadow, grassland, offering easy running), deciduous forests, and artificial surfaces (buildings and roads) within the vicinity of the nests sampled.
Habitat fragmentation was estimated from the total length of the edges measured between habitat types 174 in the landscape surrounding each sampled nest. Land use heterogeneity and fragmentation were 175 extracted from a land cover map produced at 10 m resolution by Naturvårdsverket

176
(https://www.naturvardsverket.se/). To assess the effect of land cover on the propensity and intensity 177 of parasitism, we computed each metric within buffers of increasing radius (10,20,30,40,70,100,200 178 and 500 meters) around each sampled nest. All metrics were calculated with the R packages sf [23] and 2.6.
Genetic structure of P. confusa and of A. urticae

185
The genetic structure of Swedish P. confusa and A. urticae were studied using two types of molecular markers, a fragment of the cytochrome c oxidase subunit (COI) mitochondrial gene and expensive, which make them suitable to study non-model species such as the ones examined here.

192
Comparative studies have also shown that the genetic diversity found by SSRs and AFLPs are 193 comparable, as the distribution over the entire nuclear genome of the latter counterbalances the

246
Phobocampe confusa was observed throughout the southern and northern regions of Sweden in both 247 years, but its abundance in our samples varied between hosts, sites, counties and years (  Figure 2).

274
We also observe a significant decrease in the phenological overlap between P. confusa and its hosts 275 in 2018 compared to 2017 (estimate = -0.144 ± 0.063, t = -2.31, p = 0.024, Figure 2a, Table 2). This probably 276 reflects the considerable difference in temperature profiles between the two sampling years ( Figure S1).

277
In fact, if we replace the year variable by the corresponding cumulative growing degree-days above

Pattern of attack 290
The proportion of butterfly nests parasitized by P. confusa significantly varies with the larval instar at collection and shows a concave relationship with the phenological overlap (Table 3, Figure 2b). The

292
proportion of butterfly nests parasitized by P. confusa increases with increasing phenological overlap 293 and is higher for larval nests collected at the 3th and 4 th instar than for larvae collected at the 1 st and 2 nd 294 instar and 5 th instar (Figure 2b). While the proportion of butterfly nests parasitized by P. confusa do not 295 vary between the butterfly hosts, the proportion of larvae parasitized by P. confusa per nest is higher 296 for A. urticae nests than for A. io nests (estimate = 0.41 ± 0.18, z = 2.27, p = 0.024, Figure 2c). The proportion 297 of larvae parasitized by P. confusa also varies significantly between sampling years and this effect is 298 specific to region. While in the northern region, the proportion of larvae parasitized by P. confusa per 299 nest decreases between 2017 and 2018, the opposite is observed in the southern region (estimate = 1.04 the larval instar at collection (Table 3)        respectively. We obtained a total of 243 polymorphic AFLPs fragments with a very low genotyping 344 error rate (< 1%). We do not observe a significant difference in gene diversity between regions (Table   time (see SM 2). We did not find any difference between native species in the probability of a nest to 374 be parasitized; however the intensity of parasitism, taken as the proportion of larvae parasitized per 375 nest, differs between species and is significantly higher for A. urticae than for A. io. This result suggests 376 that, at least in this study, P. confusa seems to favour A. urticae as host.

377
The large between-year variation in climatic profile highlights the potential impact of warming 378 on our study system. Climate change is a challenge for ectothermic species such as parasitoids and would be insightful in that respect.

401
We found that butterfly nests and, therefore P. confusa, preferentially occur in habitat

429
To date, no genetic data have been made available for P. confusa. Here, we show that the COI 430 genetic diversity is extremely low in this species, at least within the geographical scale of our study.

431
We found only five different haplotypes which diverged by no more than 3 mutational steps ( Figure   432 4