Frontiers of Biogeography Potential host number in cuckoo bees (Psithyrus subgen.) increases toward higher elevations

. In severe and variable conditions, specialized resource selection strategies should be less fre ‐ quent because extinction risks increase for species that depend on a single and unstable resource. Psithyrus ( Bombus subgenus Psithyrus ) are bumblebee parasites that usurp Bombus nests and display inter ‐ specific variation in the number of hosts they parasitize. Using a phylogenetic comparative frame ‐ work, we show that Psithyrus species at higher elevations display a higher number of hosts species com ‐ pared with species restricted to lower elevations. Species inhabiting high elevations also cover a larger temperature range, suggesting that species able to occur in colder conditions may benefit from recruit ‐ ment from populations occurring in warmer conditions. Our results provide evidence for an ‘altitudinal niche breadth hypothesis’ in parasitic species, showing a decrease in the parasites’ specialization along the elevational gradient, and also suggesting that Rapoport’s rule might apply to Psithyrus .


Introduction
Optimal resource selection is crucial for all organisms and directly impacts fitness (Parker and Stuart 1976). Among species, resource selection ranges from highly specialized (i.e., small range of resources consumed) to highly generalized (i.e., large range of resources consumed). Furthermore, different resource selection strategies may not generate identical costs and benefits in mild compared to severe environments. Indeed, resource specialization is expected to be more advantageous in stable environments, where fluctuations of resources' availability are low, and intense competition favours niche partitioning (MacArthur 1955). In contrast, more generalist species should be advantaged in severe environments where resources are fluctuating, since they do not depend on a restricted range of resources that may become temporarily unavailable (Currie 1991).
Generally, parasites obtain resources from a host in the form of nutrients (e.g., body fluids), and also habitats (e.g., nest). The latter is particularly widespread in social insects, because nest parasites benefit from the workforce of non-reproductive females after taking over the nest (Wcislo 1987). As host availability and population stability vary along environmental gradients (Brinck 1974), parasites should be more host specialized in more productive and stable environments, compared with severe conditions. Latitude has been traditionally associated with such an environmental gradient, inspiring the 'latitude niche breadth hypothesis' (MacArthur 1972). It may explain, for example, the increased host range exploited by cuckoo birds at higher latitude (Yom-Tov and Geffen 2005). By extending this rule to the geographical range of species, Rapoport suggested that species occupying, on average, higher latitudes should also display a larger latitudinal range (Rapoport 1975). Yet the use of latitude may not represent a stability gradient and is confounded by historical and biogeographical factors (Vázquez and Stevens 2004). For instance, it remains unclear whether the higher frequency of social parasites at higher latitude arises from ecological or historical constraints (Wcislo 1987). To keep away from such biases, one would need to study a geographically restricted zone which en-122 frontiers of biogeography 5.2, 2013 -© 2013 the authors; journal compilation © 2013 The International Biogeography Society ISSN 1948-6596 research letter Potential hosts number in cuckoo bees (Psithyrus subgenus) increases towards higher elevations compasses a wide range of environmental conditions.
Elevation clines developing from lowlands to the summits of mountain ranges show similar environmental gradients to that of latitude from the tropics to the poles. High-elevation environments form a series of marginal habitats often characterized by stronger fluctuations in physical factors, more pronounced temporal (daily and seasonal) variability, and also more stressful and stochastic climatic conditions, than in lowland environments (Brinck 1974). Such conditions probably affect species' population dynamics. Indeed, at high elevations, insect populations undergo heavy fluctuations and community structures are less stable (Gutiérrez and Menéndez 1998). This increased stochasticity may have important consequences for the life history of high-elevation insect parasites, but this is not yet known.
Psithyrus (cuckoo-bees) is a monophyletic group within the Bombus genus, whose species parasitize nests of other Bombus species (Cameron et al. 2007). The Psithyrus individual enters the host colony and either adopts the odour of the invaded nest or tries to take over the queen directly after the nest invasion (Goulson 2010). Then it lays eggs and uses the nonreproductive female workers of the invaded colony to raise its progeny (Goulson, 2010). Among Psithyrus species the number of hosts can vary (Rasmont 1988), but it is not yet known how this variation is distributed along environmental gradients and whether the number of hosts per species depends on the environmental conditions. Antonovics and Edwards (2011) showed that parasite -free zones exist at the limit of the range of the species. The advantage provided by being generalist (high number of potential hosts) or specialist (low number of potential hosts) may depend on the prevailing environmental conditions, varying along the altitudinal gradient. Following this rationale, we would thus expect to find higher proportions of generalist Psithyrus at high elevations, whereas at low elevations specialists should prevail.
Here we studied the link between host specificity and environmental conditions using the subgenus Psithyrus. We sampled 149 bumblebee communities along an elevational gradient in the western Swiss Alps. We coupled environmental niche information with data from the literature on host species and tested the following hypotheses: 1) Host species number decreases with temperature and increases with elevation. 2) Species occupying colder conditions also span a larger temperature range, in accordance with Rapoport's rule.

Materials and methods
The study area is located in the western Swiss Alps and covers 700 km², ranging from 1000 m to 3210 m a.s.l. We selected the sampling sites using balanced stratified random sampling, restricted to open meadows and based on elevation, slope and aspect, thus allowing for accurate estimates of species' niche limits along the elevational gradient. From June to September 2009 and 2010, 149 sites of 50 m x 50 m were sampled ( Figure 1). We visited each site every three weeks during the most favourable hours for bumblebee activity (10:00-17:00) and in good weather conditions (i.e., little wind, sunny, high temperatures; number of field days = 146). At each site, we used a net to collect every Bombus and Psithyrus individual we could during a 45-minute search. Identification of specimens took place in the laboratory. The number of hosts for each Psithyrus species in the study area was obtained from the literature (Amiet 1996, Williams 2008) and ranged from one to five host species. These records constitute an accurate list of all host Bombus species that are known to be parasitized by each Psithyrus species and thus constitute the 'potential host range'. The 'realized host range', i.e., the field record of all the hosts used by Psithyrus species at each location, would be more informative. However, Bombus nests are cryptic and very difficult to find in the field, making such data almost impossible to record at the scale of our study zone.
As a measure of environmental severity, degree-days were calculated and interpolated using data from meteorological stations, with a digital elevation model at 100 m resolution (for more detail, see Zimmermann and Kienast 1999). Degree-days are the sum of degrees above 3°C for all days in the year. Decreasing temperature along the elevational gradient is also associated with increased precipitation and climatic variability. In the study area, degree-days are closely related to elevation (R² = 0.98, p < 0.001). The temperature variance increases with elevation (MeteoSwiss data for 187 recording stations in Switzerland, based on the variance of mean daily temperature, R² = 0.23, p < 0.001). As a measure of colder niche edge, we extracted, for each species, the degreedays values for the sites where it occurred and calculated the 20 th percentile of those values. We used the 20 th percentile instead of the minimum degree-days value to limit the noise of outlier occurrences. We also measured the variance of each species' degree-days values to represent its temperature range. We accounted for species relatedness by using a published phylogeny (Cameron et al. 2007). We related the species-specific host number, the 20 th percentile of degree-days and the variance in degree-days using phylogenetic Generalized Least Squares (pGLS), as implemented in the caper R package (Orme et al., 2011). The strength of phylogenetic dependency in our data was accounted for by estimating, using Maximum Likelihood, the λ branch transformation parameter. Other transformations assuming different evolutionary scenarios exist, but they require larger phylogenetic trees to be correctly estimated (R. Freckleton, personal communication). Finally, to show a potential change in the number of Bombus species that might affect the availability of hosts to Psithyrus, we related the number of Bombus species to degree-days using a Generalized Linear Model (GLM) with a quasi-Poisson error distribution.

Results
In the 149 inventoried sites we found eight of the nine Psithyrus species (total sampled individuals = 183) occurring in Switzerland and 20 non-parasitic Bombus species out of the 31 occurring in Switzerland (total sampled individuals = 1609) (see Table  1 and Figure 1, Pradervand et al. (2011) and Amiet (1996) for more information). We have lumped together three closely related Bombus species (B. terrestis, B. lucorum and B. cryptarum), at the subgenus level (Bombus sensu stricto), because of the high probability of wrong identification for this group, especially with old worker bees. The bumblebee species richness is higher at intermediate values of degree-days, corresponding to midelevations, and reaches its lowest value in the coldest conditions at high elevation (linear: t = 4.14, DF = 148, p < 0.001, quadratic: t = -3.91, p < 0.001; Figure 2; see Pellissier et al. 2012b for more information). The Psithyrus species richness shows a slight linear increase in species richness (linear: t = 2.56, DF = 74 p = 0.013 quadratic: t= -3.5*10 -7 , p > 0.05).
Overall, Psithyrus individuals are relatively rare, and, while they were found in 49% of our sites, the proportion of Psithyrus among bumblebee individuals collected was very low ( portions of healthy mountain populations; Goulson et al. 2008, Roulston andGoodell 2011). The pGLS regression showed a significant relationship between the species' colder range limit and the number of hosts: Psithyrus species tended to have more hosts in cold than in mild conditions (F = 11.54; DF = 2,6; p = 0.009; adjusted R² = 0.60; λ = 6*10 -5 ; Figure  3a). Because of the low value of λ we tested whether an Ornstein-Uhlenbeck model of evolution would better fit the data using the Geiger R package, but the Akaike Information Criterion difference was marginal (0.23). Finally, we also found a significant negative correlation between the species colder niche edge and the variance in temperature occupied by the species computed from occurrence data (F = 7.75; DF = 2,6; p = 0.022; adjusted R² = 0.49; Figure 3b). Psithyrus species occurring in colder conditions spanned a larger range of temperature.   Variance of degree-days for each species plotted against the colder niche edge. Species with a smaller temperature range tend to occur in warmer conditions (R² = 0.49, p = 0.022). The outlier corresponds to P. campestris, which occurs over the whole gradient but is found most in colder conditions. The sizes of the points in both graphs correspond to the number of sites occupied (see Table 1).   (Pellissier et al., 2012b), B. soroeensis is between the two clades. * Bombus sylvarum and Bombus pomorum are two species recorded in the study area but with distributions below 1000m.

Discussion
Host specialization is known to vary between Psithyrus species (Rasmont 1988), but it was unknown whether, as for latitudinal gradients, a specialist-to-generalist progression could be found along an elevation cline. In this study we combined environmental data on Psithyrus species with their evolutionary history and found that, indeed, species occurring at higher elevation are more generalist than their low-elevation counterparts. The trend was observed despite the decrease of bumblebee species richness at high elevation ( Figure 2). Bumblebee species richness follows a hump-shaped pattern typical of a midaltitudinal diversity peak (Pellissier et al. 2012b). Also, we provide evidence of increasing niche breadth with altitude, which parallels the results found for latitude (Krasnov et al. 2008). The fact that generalist Psithyrus occurring in the coldest environment also occupy a larger temperature range suggests that Rapoport's rule might also be valid for altitude (Stevens 1992). Generalist species may be more prevalent in colder environments than specialists because they can compensate for the variation in host populations by parasitizing more species and thus increase the chance of finding a host. Cold and unstable conditions may disadvantage species that are bound to a particular host species. Indeed, it is known that host populations fluctuate more in colder environments (Brinck 1974), making the resources less reliable for specialist parasites. In addition, the length of the vegetation growing season is reduced at high elevations, which constrains the foraging period and thus the time to complete the life cycle (Goulson et al. 2005). This forces Psithyrus species emerging at high elevations to quickly find a host to raise their progeny before the end of the nesting season. Being able to colonize any nest encountered is thus likely to be an advantage. Interestingly, generalist species occurring in cold conditions, such as P. rupestris and P. campestris, lack the specific cuticular carbohydrates that are used as recognition chemicals (Martin et al. 2010), leaving open the possibility of invading a larger number of host species' nests and reducing the search time needed. In contrast, specialist species that mimic the odours of a particular host species probably show a double limitation to their distribution. First, they are associated with host species that are mostly absent in cold environments (Pradervand et al. 2011), and second, environmental constraints of unstable environments may prevent the persistence of populations.
Generalist Psithyrus species occurring in the coldest environmental conditions also occupy warmer sites. Among their potential hosts they thus have a range of different species occurring at different elevations. This result parallels the much -discussed Rapoport's rule, which suggests that latitudinal ranges of plants and animals are generally smaller at lower than higher latitudes (Rapoport 1975, Stevens 1992. We found that generalist species occupy a wider temperature range than more specialized species and are therefore likely to encounter more potential host species to parasitize. Paralleling our results, Sanders (2002) found that the elevational range sizes of ants from Colorado, Nevada and Utah widen with increasing elevation. Such patterns following Rapoport's rule may be explained by a rescue effect: unstable populations occurring at high elevations might benefit from constant recruitment from stable populations occurring at lower elevations (Stevens 1992).
In this study, we sampled and analysed the distribution of adult specimens under the assumption that they were caught close to the nest they parasitize. To remove this potential source of bias, one would need to sample the Psithyrus directly in the parasitized nest, which is very difficult because they are cryptic and buried deep in the soil. Moreover, invasion can only be confirmed when the offspring leaves the nest because parasites sometimes use the nest as periodic shelter (Rasmont 1988).
In conclusion, we showed that Psithyrus species occurring at high elevation tend to parasitize more host species (Figure 3a), suggesting an environmental filtering effect that allows only generalist species to survive in those conditions. Our results not only provide direct support consistent with the altitudinal niche breath hypothesis, 127 frontiers of biogeography 5.2, 2013 -© 2013 the authors; journal compilation © 2013 The International Biogeography Society but also provide echoing evidence consistent with Rapoport's rule as applied to an altitudinal gradient. Interestingly, generalist species able to occur at higher elevation have a more basal position in the phylogeny, suggesting that the specialized species may have evolved from generalist ancestors. However, this requires further research using a larger Psithyrus species sample because incomplete taxon sampling could bias ancestral state reconstruction (Litsios & Salamin 2012). Recent studies are starting to reveal a common pattern for the prevalence of generalist species in more severe conditions at high elevations in trophic and mutualistic interactions, in insects and plants (Pellissier et al. 2012a). Further investigations of other species' interactions and niche breadths along altitudinal gradients are thus required to assess the universality across biotic interactions of the altitudinal niche breadth hypothesis.