Adapting to change: Exploring the consequences of climate‐induced host plant shifts in two specialist Lepidoptera species

Abstract Asynchronous migration of insect herbivores and their host plants towards higher elevations following climate warming is expected to generate novel plant–insect interactions. While the disassociation of specialised interactions can challenge species' persistence, consequences for specialised low‐elevation insect herbivores encountering novel high‐elevation plants under climate change remain largely unknown. To explore the ability of two low‐elevation Lepidoptera species, Melitaea celadussa and Zygaena filipendulae, to undergo shifts from low‐ to high‐elevation host plants, we combined a translocation experiment performed at two elevations in the Swiss Alps with experiments conducted under controlled conditions. Specifically, we exposed M. celadussa and Z. filipendulae to current low‐ and congeneric high‐elevation host plants, to test how shifts in host plant use impact oviposition probability, number of eggs clutches laid, caterpillar feeding preference and growth, pupation rate and wing size. While our study shows that both M. celadussa and Z. filipendulae can oviposit and feed on novel high‐elevation host plants, we reveal strong preferences towards ovipositing and feeding on current low‐elevation host plants. In addition, shifts from current low‐ to novel high‐elevation host plants reduced pupation rates as well as wing size for M. celadussa, while caterpillar growth was unaffected by host plant identity for both species. Our study suggests that populations of M. celadussa and Z. filipendulae have the ability to undergo host plant shifts under climate change. However, these shifts may impact the ability of populations to respond to rapid climate change by altering developmental processes and morphology. Our study highlights the importance of considering altered biotic interactions when predicting consequences for natural populations facing novel abiotic and biotic environments.


| INTRODUC TI ON
Understanding how species respond to alterations in their abiotic and biotic environment is crucial to predict the ability of natural populations facing climate change to persist (IPBES, 2019).
Apart from coping with climate change in situ, species can track their optimal temperatures by migrating towards higher latitudes (Forister et al., 2010;Parmesan & Yohe, 2003;Walther et al., 2002) and altitudes (Chen et al., 2009;Parolo & Rossi, 2008;Pauli et al., 2012).However, the capacity and rate of migration towards climatically suitable habitats often differ between taxonomic groups (Urban et al., 2012).For example, limited dispersal capacity can cause migration rates of plants to lag behind changes in temperatures (Alexander et al., 2018;Ash et al., 2017;Corlett & Westcott, 2013), while ectothermic insects are more likely to keep up with climate change via rapid migration (Rödder et al., 2021;Vitasse et al., 2021).Asynchronous migration of plants and insects following climate change (Kerner et al., 2023;Vitasse et al., 2021) is thus expected to generate novel plant-insect interactions involving species whose ranges are currently non-overlapping (HilleRisLambers et al., 2013;Urban et al., 2012).Shifts in biotic interactions, such as those between plants and herbivores, pollinators and/or competitors, following altered abiotic conditions under climate change are described as indirect effects of climate change.
Although these indirect effects can play crucial roles in dictating species' responses to climate change (Alexander et al., 2015;Descombes, Kergunteuil, et al., 2020;Gilman et al., 2010), consequences of altered plant-insect interactions following asynchronous range shifts on species' performance and development remain understudied.
To understand how species will respond to climate change, it is necessary to expose natural populations to abiotic and biotic conditions they are expected to face in the future.Mountains serve as unique study systems allowing for simulations of realistic future scenarios of both warmer climates and novel biotic interactions (Nooten & Hughes, 2017;Tito et al., 2020).Natural temperature gradients of mountains allow for simulations of predicted climate warming by translocating species downhill (Tito et al., 2020).
Moreover, elevation gradients can be used for addressing the effect of novel plant-insect interactions formed because of asynchronous upslope migration under climate change.Previous studies simulating future scenarios of climate warming and altered interactions between plants and insects by transplanting plant and/or insect communities across elevation gradients have revealed that the establishment of low-elevation insects at high elevation plays an important role in shaping plant community responses to climate change (Descombes, Kergunteuil, et al., 2020;Descombes, Pitteloud, et al., 2020;Richman et al., 2020).For example, Descombes, Pitteloud, et al. (2020) showed that the introduction of low-elevation generalist insect herbivores at high elevation can alter biomass production, diversity and chemical properties of alpine plant communities.While the consequences of altered plant-insect interactions mainly have been explored by estimating responses in plant communities, insect herbivores are also expected to face challenges when encountering novel plant communities.Specifically, the absence of current host plants can threaten the persistence of insect herbivores by limiting their capacity to establish at higher elevations, unless they are able to shift their diet and consume new plant species (Hanspach et al., 2014;Merrill et al., 2008).
Host plant shifts can impact insect herbivores by altering vital rates associated with reproduction and growth, where interactions with a novel host plant may reduce oviposition success, offspring survival and size (Braschler & Hill, 2007;Näsvall et al., 2021;Parry & Goyer, 2004;Pelini et al., 2009).While generalist insect herbivores are expected to successfully establish novel plant communities (Berg et al., 2010;Gilman et al., 2010;Rödder et al., 2021), specialists might be less likely to do so (Andrew & Hughes, 2004;Warren et al., 2001;Yamamoto & Uchida, 2018).Specialised plantinsect herbivore interactions are shaped by co-evolutionary arms races (Becerra et al., 2009;Ehrlich & Raven, 1964), mediated by the reciprocal evolution of chemical defences produced by host plants and detoxification/sequestration mechanisms by insect herbivores.
Hence, long periods of co-evolution have optimised performance on one or a limited number of plant species, which may constrain the performance of other plant species (Futuyma & Moreno, 1988;Pelini et al., 2009).In this regard, successful establishment of novel species requires that adult can recognise the novel plants, mostly via chemical cues (Honda, 1995;Nishida, 2014) and that larvae can survive and grow on leaves containing (slightly) different secondary metabolites produced by these plants (Jeschke et al., 2017;Pelini et al., 2009Pelini et al., , 2010)).
Accordingly, theory suggests that specialised insect herbivores could recognise, survive and develop on other plant species if their phenotypes largely match those of the current host plants (Gassmann et al., 2006).It is thus expected that specialised lowelevation insect herbivores encountering alpine plant communities will choose to oviposit and feed on species closely related to their current low-elevation host plants (Moir et al., 2014), as these not only are likely to be morphologically but also chemically similar (Fahey et al., 2001;Farrell & Mitter, 1998;Griffin & Lin, 2000).
However, although they might be chemically similar, even small variations in chemistry can generate widespread impacts on the insect herbivores (Glassmire et al., 2016;Harrison et al., 2016), which may impact the likelihood of successful colonisation of the novel host plant.Therefore, to estimate the consequences for  (Radchuk et al., 2013).Therefore, apart from oviposition, we explored how shifts in host plant identity impact caterpillar preference and performance (estimated by growth; Knolhoff & Heckel, 2014), pupation rate and wing size, by conducting additional experiments exposing M. celadussa and Z. filipendulae to current low-and novel high-elevation host plants under controlled conditions.Finally, we performed targeted metabolomic analyses on plants to explore whether responses to shifts in host plant identity were associated with differences in concentrations of chemical defence compounds.

| Study system
As focal species, we selected two Lepidoptera species abundant and non-threatened in Switzerland (Wermeille et al., 2014); the Southern heath fritillary Melitaea celadussa Fruhstorfer, 1910, and six-spot burnet Zygaena filipendulae (Linnaeus, 1758).Melitaea celadussa is a butterfly ovipositing egg clutches of c. 50-100 eggs on species belonging to the Plantaginaceae family, mainly on Plantago lanceolata L. While P. lanceolata is considered to be the main hostplant, M. celadussa has also been recorded to oviposit on a few species belonging to the families Orobanchaceae and Asteraceae, but, to our knowledge, never on alpine Plantago species (Clarke, 2022;LSPN, 1987).Melitaea celadussa populations are found between 400 and 1400 m within the study region (Figure S1a).Zygaena filipendulae is a diurnal burnet moth exclusively laying egg clutches of c. 50-100 eggs on plants belonging to the genus Lotus (particularly on Lotus corniculatus L. in the Alps).Moreover, caterpillars of Z. filipendulae have been observed to feed on other Fabaceae species, but again, to our knowledge, never observed feeding on alpine Lotus species (Guenin, 2023;LSPN, 1999;Paolucci, 2013).
The elevational distribution of Z. filipendulae in the study region ranges mainly between 400 and 1600 m (85% of the observations are below 1560 m, Figure S1b).
For both species, we estimated their potential distribution following realistic scenarios of climate change (see Methods S1 section).Between 1985 and 2022, leading edges of elevation distributions (95th percentile, estimated following the methods used in Vitasse et al., 2021) have shifted towards higher elevations at a rate of 84.2 and 66.7 m/decade, for M. celadussa and Z. filipendulae, respectively (Figure 1a).In addition, predictions obtained from species distribution models (SDMs) using current occurrences of both Lepidoptera species in Switzerland suggest that leading edges of species' climatic niches will shift towards higher elevations by the end of the century following climate change.Specifically, an increase in global average temperatures by 1.1-2.6°C(Representative Concentration Pathways 4.5), as predicted by the end of the century (IPCC, 2014(IPCC, , 2023)), is expected to shift leading edges upwards by 227 and 494 m for M. celadussa and Z. filipendulae, respectively (Figures S2 and S3).
Under climate change scenario Representative Concentration Pathways 8.5, assuming that global average temperatures rise by 2.6-4.8°C by the end of the century (IPCC, 2014(IPCC, , 2023)), leading edges are predicted to shift by 544 and 868 m for M. celadussa and Z. filipendulae, respectively (Figure 1b and Figures S2   and S3).
The lowland plant species P. lanceolata and L. corniculatus were selected to represent current low-elevation host plants for M. celadussa and Z. filipendulae, respectively (Guenin, 2023;LSPN, 1987LSPN, , 1999)).As congeneric high-elevation species, we selected Plantago atrata Hoppe and Lotus alpinus (DC.)Ramond for M. celadussa and Z. filipendulae, respectively, as they are the closest, most abundant, high-elevation relatives to the selected low-elevation host plants.To date, M. celadussa and Z. filipendulae have never been observed feeding or ovipositing on these high-elevation plants.Both P. lanceolata and P. atrata are known to produce iridoid glycosides (IGs), acting not only as feeding deterrent compounds for several generalist herbivores (Biere et al., 2004;Rønsted et al., 2002) but also as feeding and oviposition stimulants for more specialist herbivores (Bowers, 1983;Nieminen et al., 2003;Pereyra & Bowers, 1988;Reudler Talsma et al., 2008).Lotus corniculatus and L. alpinus produce cyanogenic glycosides (CNGs), known not only to deter generalist herbivores via activation by β-glucosidase enzymes but also to protect specialists via sequestration (Zagrobelny et al., 2007).For more information on Lepidoptera and plant species used in this study, see Figures S1-S4 and Table S1.For sample sizes as well as the caterpillar instars used for experiments performed in field and under controlled conditions (see below), see Table S2.

| Oviposition experiment
To explore how shifts in host plant identity and different climatic conditions influence the probability of oviposition and the number of egg clutches laid by M. celadussa and Z. filipendulae, we established two common garden experiments at two different elevations in the Swiss Alps in June 2022 (Figure 1c-f).The experiment included two sites, one situated at 1500 and one at 2150 m, selected specifically to ensure similarities in sun exposure, slope, and orientation (Table S3).
Average temperatures throughout the experiment (June-July) corresponded to 21 and 18.2°C for the low-and high-elevation sites, respectively (Table S3, Figure S5).at 1500 m, while P. atrata (n = 20) and L. alpinus (n = 30) were excavated near the 2150 m site (Table S4).Selected individuals grew at least 1 m apart from each other and were carefully excavated to prevent damage to below-ground structures.Only individuals with no or minimal herbivory damage were selected.Individuals were then planted into pots (20 cm diameter) with standard potting soil (Terreau Suisse, Ricoter, Aarberg, Switzerland), and reciprocally translocated to 1500 and 2150 m (Figure 1).High-elevation plants translocated to 1500 m and thus experienced an increase in temperature of 2.8°C on average, simulating climate warming.Although morphologically similar (Lauber et al., 2018), high-elevation individuals produced shorter stalks, but more leaves and stalks (see Methods S1 for trait measurements, Figure S6; Table S5).In addition, the leaf C:N ratio was higher for individuals of P. atrata compared to P. lanceolata (Figure S7; Table S6).Melitaea celadussa and Z. filipendulae were captured at multiple sites near the study area (between c. 400 and 1460 m; Table S5).In an optimal situation, Lepidoptera individuals would have been collected from multiple low-elevation sites at the same elevation, to avoid potential "home-elevation" effects.However, this was not possible in this study, as the number of populations of M. celadussa and Z. filipendulae in the study area is limited.Hence, to avoid imposing major effects on population size and to maximise the genetic variation, we sampled individuals from multiple populations at different elevations.Importantly, the aim of this study was to expose Lepidoptera species to future high-elevation host plants, with whom they never interacted before.Although absolute elevations from which Lepidoptera included in the field experiment originated varied, selected high-elevation plant species were absent from all collection sites as they started to appear at an elevation of c. 1750 m in the region where Lepidoptera were collected.Therefore, high-elevation plant species represent novel hosts for all Lepidoptera individuals included in the experiment.After their capture, Lepidoptera were stored in cylindrical plastic containers and sexed (see Methods S1).
Two females and one male of M. celadussa (total n = 60 individuals) and Z. filipendulae (total n = 90 individuals) were placed in each cage prepared as described above within 24 h following collection.
With this set-up, we simulated scenarios where M. celadussa and Z.To determine whether host plant identity and elevation influenced the oviposition preference of M. celadussa and Z. filipendulae, we fitted a generalised linear model (GLM) with binomial family distribution for each Lepidoptera species separately, where the probability of laying eggs was implemented as the response variable and the interactive effect between plant species and the elevation (high or low elevation) as explanatory variables.To test whether host plant identity and elevation influence oviposition success, we fitted GLMs with Poisson distribution for each Lepidoptera species separately, where the number of egg clutches was implemented as the response variable and elevation, plant species and the interactive effect between plant species and elevation as explanatory variables.Effects of host plant identity and elevation on oviposition preference and success were estimated by performing Chi 2 tests of fitted models using analyses of variance (ANOVA).Finally, as the focus of these analyses was imago preferences, cages where Lepidoptera did not oviposit were excluded from the statistical analyses (final sample size n = 5 for Z. filipendulae in each elevation and n = 10 and n = 6 for M. celadussa at low and high elevation, respectively).

| Effects of hostplant identity on the preference and performance of caterpillars
To investigate whether host plant identity impacts the performance and preference of caterpillars, pupation rate and adult wing size, eggs obtained from the field experiment were used to rear caterpillars under optimal climatic conditions (16 h of daylight, 27 and 22°C of average day and night temperature, respectively).

| Caterpillar preference
To explore whether caterpillars prefer low-or high-elevation plant species, we placed caterpillars, being between the third and the sixth instars, in the centre of a Petri dish (9 cm × 1.4 cm) containing either whole leaves of both Lotus spp. or 20 × 5 mm leaf rectangles of both Plantago spp.High-elevation plants were collected at around 2150 m, while low-elevation plants were collected at 500 m (Table S4).Plants were then left to acclimatise for 7 days under climate-controlled conditions before leaf sampling.Leaf area was estimated using ImageJ (Schneider et al., 2012) before being placed equidistant from the centre of a Petri dish.After 5 or 3 h, depending on differences in feeding rate between M. celadussa and Z. filipendulae, the leaf area was estimated again, and the consumed leaf area was calculated by subtracting the initial from the final leaf area.To avoid biases due to potential preference towards the plant species on which caterpillars initially hatched and fed (before reaching third instar; Knolhoff & Heckel, 2014) To explore whether host plant identity and translocation treatments impact the performance of caterpillars, we fitted linear models for M. celadussa and Z. filipendulae separately, where growth rate (mg/day) was implemented as the response variable and the elevation where host plants were acclimatised, host plant identity and the interactive effect between host plant identity and elevation as explanatory variables.To obtain normally distributed residuals, growth rate was log-and square-root transformed for M. celadussa and Z. filipendulae respectively.Additionally, negative or null growth rates were removed prior to the analyses, as they indicate that caterpillars either did not feed or died.Effects of host plant identity and experimental site to which plants had been translocated on caterpillar performance were estimated by performing F-tests of fitted models using ANOVA.To explore pair-wise differences in growth rate between host plants and translocation treatments, we performed post hoc tests (pairwise comparisons with Tukey adjustment).

| Effects of hostplant identity on the pupation rate and wing size of M. celadussa
For M. celadussa, we further investigated whether host plant identity impacts pupation rate and adult wing size by performing additional experiments under controlled conditions.These experiments could not be performed for Z. filipendulae, as all caterpillars from this species entered diapause, and thus released to the site of origin of their parent or died.To estimate the effect of host plant identity on pupation rate and wing size for M. celadussa, we placed n = 10 and 9 replicates of three second instar caterpillars on living plants growing in butterfly-rearing cages on P. lanceolata and P. atrata respectively.All caterpillars used in this experiment hatched on the same day.As for the experiment performed to explore caterpillar preference, P. atrata individuals were collected around 2150 m, while P. lanceolata individuals were collected at 500 m (Table S4).
Plants were potted and left to acclimatise for 7 days under controlled conditions in the greenhouse before caterpillars were placed on plants.Caterpillars were placed on the same plant species on which they hatched.

| Pupation rate
To investigate whether host plant identity impacts the pupation rate for M. celadussa, we recorded whether caterpillars (total n = 30 and 27 for P. lanceolata and P. atrata, respectively) entered pupation or diapause.For details on how pupation/diapause was determined, see Methods S1.To test whether the identity of the host plants on which M. celadussa caterpillars fed impacted their probability of entering pupation (pupation rate), we fitted GLMs with a binomial family distribution with recordings of pupation (pupation = 1, diapause = 0) as the response variable and host plant identity as an explanatory variable.Effects of host plant identity on pupation rate were estimated by performing Chi 2 tests of fitted models using ANOVA.

| Wing size
To test whether host plant identity impacts wing size, we estimated the wing area of all individuals reaching adult stage (n = 17 and 11 individuals reared on P. lanceolata and P. atrata, respectively) of M. celadussa.Dead adults were dried at room temperature for c. 1 month before wings were removed.Wing area was estimated using ImageJ (Schneider et al., 2012).Perimeters of torn wings were corrected (n = 8) and when large proportions of a wing were damaged (n = 3), measurements of the opposite wing were used to estimate total wing area.To test whether the wing size of individuals reaching the adult stage after pupation is affected by host plant identity, we fitted linear models with square-root-transformed wing area as the response variable and host plant identity as an explanatory variable.
Square-root transformation of wing size was performed to obtain normally distributed residuals.Effects of host plant identity on wing area were estimated by performing F-tests of fitted models using ANOVA.

| Chemical analyses and data processing
To test whether the production of secondary metabolites related to anti-herbivore defences varies according to host plant identity and climatic conditions, we performed targeted secondary metabolites analyses on a total of n = 59 randomly selected plant individuals included in the field experiment (n = 7, 7, 7, 8 individuals growing at experimental sites situated at low elevation and n = 8, 8, 6, 8 individuals growing at experimental sites situated at high elevation for P. lanceolata, P. atrata, L. corniculatus and L. alpinus, respectively).
To test whether the production of chemical defence compounds was impacted by host plant identity and translocation treatments, we fitted separate linear models for each compound (catapol, aucubin, linamarin and lotaustralin), where log-transformed compound concentrations were implemented as response variables and the host plant identity, the experimental site to which plants were translocated, the interactive effect between host plant identity and experimental site to which plants were translocated as explanatory variables.Log transformations of compound concentrations were performed to obtain normally distributed residuals.Effects of host plant identity on secondary metabolites concentration were estimated by performing F-tests of fitted models using ANOVA.

| Oviposition preference and success
Host plant identity had no impact on oviposition probability for M. celadussa (Figure 2a; Table 1), while the probability of Z. filipendulae to oviposit on its current low-elevation host plant, L. corniculatus, was 2.25× higher on average compared to L. alpinus (Figure 2b; Table 1).2.3× and 4× more egg clutches were laid on current lowcompared to congeneric high-elevation host plants for M. celadussa (Figure 2c; Table 1) and Z. filipendulae (Figure 2d; Table 1) respectively.In contrast, elevation neither impacted oviposition probability nor the number of egg clutches laid (Table 1).

| Caterpillar preference
When caterpillars could choose, they consumed larger leaf areas of low-compared to high-elevation host plants (Table 1).Caterpillars of M. celadussa consumed 21% more leaf area on P. lanceolata compared to P. atrata (Figure 2e; Table 1), while leaf consumption was 29% higher on L. corniculatus compared to L. alpinus for Z. filipendulae caterpillars (Figure 2f; Table 1).

| Caterpillar performance
Host plant identity had no impact on caterpillar growth rate (Figure 2g,h; Table 1).In contrast, the elevation where food plants were translocated impacted the caterpillar growth rate for M. celadussa, where the growth rate was 34% lower for caterpillars feeding on P. atrata growing under warmer temperatures at low elevation compared to caterpillars feeding on P. atrata growing under cooler temperatures at high elevation.Inversely, for Z. filipendulae, the growth rate was 1.9× higher for caterpillars feeding on L. alpinus growing at low elevation compared to caterpillars feeding on L. alpinus growing at high elevation (Figure 2g, Table 1).

| Pupation rate and wing size for M. celadussa
The pupation rate was 4.4× higher for M. celadussa caterpillars feeding on P. lanceolata compared to those feeding on P. atrata (Figure 2i; Table 1).Moreover, host plant identity also impacted wing size for M. celadussa, which was 13% smaller for individuals reared on P. atrata compared to those reared on P. lanceolata (Figure 2j; Table 1).

| Secondary metabolites of plants
Relative concentrations of catapol were 4.6× higher in P. lanceolata compared to P. atrata, while no differences in relative aucubin concentrations were detected (Figure 3a,b, Table S7).Absolute concentrations of linamarin and lotaustralin were 144× and 98× (Figure 3c,d, Table S7) higher in L. corniculatus compared to L. alpinus.
Elevation where plants were translocated had no impact on any of the compounds analysed here (Table S7).

| Preference towards ovipositing and feeding on current low-elevation host plants
Assessing whether specialised low-elevation insect herbivores can establish novel host plants at higher elevations is crucial to understanding their chances to persist under future scenarios of climate change.Here, we revealed that M. celadussa and Z. filipendulae can oviposit and feed on the congeneric alpine plant species, suggesting that shifts from current low-to novel high-elevation host plants under climate change are possible.However, as hypothesised, we found that both species prefer to oviposit and lay more egg clutches, and their caterpillars prefer to feed on current low-rather than novel high-elevation host plants.Chemical cues play important roles in dictating the discrimination between current and potentially novel host plants ( Gamberale-Stille et al., 2014;Ikeura et al., 2010;Ozaki et al., 2011).Accordingly, previous studies have shown that host plant preference for Z. filipendulae (Zagrobelny et al., 2007(Zagrobelny et al., , 2014)), as well as congeneric species closely related to M. celadussa (Nieminen et al., 2003;Reudler Talsma et al., 2008), is driven by the presence and concentration of toxic secondary metabolites.For example, feeding on current low-elevation host plants may be preferred as caterpillars hatching and feeding on more chemically potent plants can sequester higher concentrations of secondary metabolites, toxic for predators and parasitoids (Biere et al., 2004;Bradley et al., 2018;Nieminen et al., 2003).Consequently, preference towards ovipositing and feeding on low-elevation host plants was not inhibited by higher levels of chemical defences.Instead, possessing higher concentrations of chemical defence compounds could be associated with strengthened resistance against natural F I G U R E 2 Oviposition probability (a and b), oviposition success (c and d), caterpillar preference (e and f), caterpillar growth rate (g and h), pupation rate (i) and wing area (j) across host plant identities (a-j) and elevation (g and h) for Melitaea celadussa and Zygaena filipendulae.Note that for a-d, effects of elevation are not illustrated as experimental site (1500/2150 m) neither impacted oviposition probability (a and b) nor the number of egg clutches (c and d; Table 1).For effects of elevation, see Figure S8 and Table 1.Colours indicate host plant identities where Plantago lanceolata and Plantago atrata are illustrated in dark and light green, respectively, and Lotus corniculatus and Lotus alpinus in dark and light blue respectively.For boxplots (c-h, j), bold, horizontal lines show medians, error bars illustrate distributions of the first to fourth quantile, and points represent outliers.For (a and b, i), points indicate means and error bars illustrate ±1 standard error.Asterisks (a-f, i and j) and letters (g and h) indicate significant differences (p < .05),which were calculated using F-test (g and h, j) or Chi 2 test (a-f, i).For details on significant effects of host plant identity and elevation on oviposition probability, number of egg clutches, consumed leaf area, caterpillar growth rate, pupation rate and wing area, see Table 1.
enemies via sequestration.Hence, higher concentrations of these molecules could indeed function as oviposition and feeding stimuli for the Lepidoptera species studied here (Bowers, 1983;Pereyra & Bowers, 1988).Future work should be devoted to elucidating the consequences of shifts in host plant identity on the ability of insect herbivores to cope with natural enemies across different elevations (Bruno et al., 2023).
While previous studies show that oviposition probability and success of Lepidoptera species increase under warmer temperatures (Berger et al., 2008;Saastamoinen & Hanski, 2008), in our system, climatic conditions had no impact on oviposition.Specifically, whether M. celadussa and Z. filipendulae interacted with low-and high-elevation host plants under warmer climates at low elevation or cooler climates at high elevation neither impacted the preference  et al., 2015;Descombes, Pitteloud, et al., 2020;Nomoto & Alexander, 2021), highlighting the importance of considering the effects of altered species interactions to reliably predict consequences for natural populations facing climate change.

| Host plant identity has no impact on caterpillar performance
Contrary to general expectations that feeding on novel host species affects the performance of insects (Ali & Agrawal, 2012;Pelini et al., 2009), we found that caterpillar growth was independent of host plant identity.A potential explanation for this finding is that differences in plant quality were insufficient to be detected within the timeframe of our experiment, as we only monitored caterpillar growth during two instars.Possibly, growth occurring during earlier or later instars may better capture differences in caterpillar growth rates, while monitoring growth during only two developmental stages as in this study may be insufficient to detect the effects of host plant identity on caterpillar growth rates.
Growth rate for caterpillars was lower for M. celadussa while higher for Z. filipendulae when feeding on high-elevation plants

| Novel interactions with high-elevation host plants reduce pupation rate and wing size for M. celadussa
While our study suggests that specialised low-elevation Lepidoptera can colonise and develop on high-elevation plants, we found that the pupation rate for individuals of M. celadussa was 4.4× higher if hatching and feeding on its current low-elevation host plant P. lanceolata instead of the congeneric high-elevation plant P. atrata.
In other words, those caterpillars feeding on P. atrata were more likely to enter diapause at the fourth instar, instead of continuing to grow and entering pupation.Shifts towards entering diapause rather than pupation when exposed to high-elevation host plants could be explained by a sub-optimal development due to a less nutritious diet provided by P. atrata compared to P. lanceolata (Hunter & McNeil, 1997;Takagi & Miyashita, 2008).Accordingly, lower C:N ratios (i.e. more nitrogen available) in P. lanceolata than in P. atrata might have favoured pupation for individuals feeding on the former rather than in the latter species (Loranger et al., 2012).Contrary to rearing performed under controlled conditions, M. celadussa feeding on P. lanceolata under natural conditions enters diapause in the fall and only pupates in the spring (LSPN, 1987).When switching to the high-elevation P. atrata plants, caterpillars of this species might benefit from a higher diapause rate, which could be more suitable  S7.
in alpine environments where growing seasons are shorter.Thus, a higher diapause rate may facilitate the ability of M. celadussa to undergo rapid adaptation in response to changing climates (Buckley & Bridle, 2014).
Previous studies show that Lepidoptera caterpillars experiencing abiotic stress and/or resource limitation develop into adults with smaller wings (Sweeney et al., 1986;Talloen et al., 2009).
Apart from abiotic factors, our results indicate shifts in host plant identity also could act to alter wing size, as individuals of M. celadussa developing on high-elevation host plants had smaller wings when feeding on P. atrata instead of P. lanceolata.As for pupation rate, limited resources due to lower nutrient contents in leaves of high-elevation plants (Hunter & McNeil, 1997;Takagi & Miyashita, 2008) could explain observed declines in wing size for individuals feeding on P. atrata instead of P. As wing size is positively correlated with mobility and flying time in Lepidoptera (Pöyry et al., 2009), M. celadussa may thus face challenges finding host plants as well as mates due to reduced flight capacity when established at high elevations and forced to oviposit and feed on P. atrata.

| CON CLUS ION
Our study reveals that host plant shifts following upward migration of specialised low-elevation insect herbivores under climate change are possible, although such shifts could impact their morphology and performance.These changes could alter population dynamics and affect the ability of populations to respond to rapid climate change.
Despite the challenges, our findings emphasise the potential for these species to adapt over multiple generations, highlighting the importance of long-term evolutionary studies to predict species' responses to climate change.
Yet, to confirm our results, several methodological considerations, caveats, and future perspectives could be brought forward: first, performing realistic simulations of climate change in species' natural environments is challenging, often involving limitations in sample size.For instance, the sample size for the oviposition experiment was smaller for Z. filipendulae compared to M. celadussa, as egg clutches were detected in only a third of the experimental cages.A possible explanation for the low oviposition rate of Z. filipendulae could be that some collected females had already laid their eggs before capture.
Second, responses to climate change are likely to be species specific.
Here, we explored responses to host plant shifts in only two species, which limits the ability to draw general conclusions about how spe- Chemical ecology, Community ecology, Global change ecology specialised low-elevation insect herbivores encountering highelevation plants due to climate warming, it is essential to study both adult responses and the ability of larvae to adapt to novel host plants.We here explored how shifts from current low-to novel highelevation host plants impact the ability of two specialist Lepidoptera species, Melitaea celadussa (Nymphalidae) and Zygaena filipendulae (Zygaenidae), to establish at higher elevations.Specifically, we asked: (1) Can M. celadussa and Z. filipendulae oviposit and feed on novel high-elevation host plants?(2) How do shifts from current low-to novel high-elevation host plants impact caterpillar growth, pupation rate and wing size?(3) Are high-and low-elevation congeneric plants phenotypically and chemically different?Overall, we hypothesised M. celadussa and Z. filipendulae prefer to oviposit and feed on current low-elevation host plants, while we expected reduced chemical defence in high-elevation plants to contribute to enhanced performance of caterpillars feeding on high-compared to low-elevation plants.To address our questions, we combined a field reciprocal transplant experiment established at two elevations with experiments performed under laboratory conditions.By exposing M. celadussa and Z. filipendulae to current low-and novel high-elevation host plants at both high-and low-elevation under field conditions, we explored how shifts in host plant identity under different climates impact oviposition probability and success (estimated as the number of egg clutches) of M. celadussa and Z. filipendulae.As for other taxonomic groups (Dahlke et al., 2020; Doak & Morris, 2010), responses to climate change in Lepidoptera can vary between life stages Individuals of P. lanceolata (n = 20) and L. corniculatus (n = 30) were excavated in the surroundings of the experimental site situated F I G U R E 1 Migration rates of Melitaea celadussa and Zygaena filipendulae (a and b) and the experimental design for field experiment (c-f).Both M. celadussa and Z. filipendulae are currently migrating towards higher elevations and are expected to migrate at a similar rate over the coming decades.(a) Illustrates the current shift in elevation of our focal Lepidoptera species in Switzerland (y-axis) of both the 95th percentile and the median of the elevation distributions (x-axis).Points illustrate the average shift in elevation for 37 years (1985-2022) in Switzerland, and error bars indicate 95% confidence intervals.(b) Illustrates the shift in elevation of the climatic niches of the Lepidoptera by comparing their current (1981-2010) and future (RCP 8.5, 2085, i.e. average of 2070-2100) climatic niches.The plots represent the density of suitable environments in Switzerland (x-axis) according to the elevation (y-axis).To simulate a realistic migration range, two common gardens were set up in the Western Swiss Alps at 1500 m (c) and 2150 m (d), thus representing an elevation shift of c. 650 m.Each common garden included cages containing two females and one male (e and f) of either M. celadussa or Z. filipendulae originating from low elevation together with one congeneric pair of Plantago spp.(P.lanceolata-P.atrata; n = 9 cages/site) or Lotus ssp.(L.corniculatus-L.alpinus; n = 15 cages/ site) respectively.By quantifying oviposition probability and success of Lepidoptera exposed to current low-and novel high-elevation host plants under warmer (1500 m) and cooler (2150 m) climates, we investigated the effects of host plant identity and climate on oviposition.

For
the oviposition choice experiment, one individual of the low-elevation plant species was placed together with an individual of the congeneric high-elevation species in cages (n = 2 individuals/cage) designed for butterfly rearing (50 × 50 × 70 cm, BugDorm-2S120, MegaView Science Co., Ltd, Taichung, Taiwan), generating a total of n = 11 and n = 15 congeneric pairs for Plantago spp.and Lotus spp., respectively, at low elevation and n = 9 and n = 15 congeneric pairs for Plantago spp.and Lotus spp., respectively, at high elevation.The bottoms of the cages were covered by a nylon mesh to prevent underground disturbance from local arthropod communities.Plants were allowed to acclimatise after translocation for 7 days before M. celadussa and Z. filipendulae adults were added into cages.
filipendulae can encounter current low-and potential novel highelevation host plants simultaneously under warmer (1500 m) and cooler (2150 m) climates.Three days after releasing M. celadussa and Z. filipendulae individuals in the cages, the presence/absence and the number of egg clutches laid on each plant were recorded.Surviving M. celadussa and Z. filipendulae individuals were released at the sites from which they were captured.The field experiment was performed during June-July since not all Lepidoptera could be captured simultaneously for filling all the replicates per treatment.At the end of the experiment, all plants (including egg clutches) were brought to the University of Neuchâtel and used for experiments performed under controlled conditions and for chemical analyses of plants (see below).
, preference was tested for M. celadussa caterpillars initially reared on either P. lanceolata (n = 19) or P. atrata (n = 20).Similarly, for Z. filipendulae, the experiment was performed on caterpillars initially reared on either L. corniculatus(n = 20)    or L. alpinus (n = 20).To estimate the effects of host plant identity on caterpillar preference, we accounted for the potential effects of initial leaf size on the leaf area consumed by caterpillars of M. celadussa and Z. filipendulae by extracting residuals of a linear model including log-transformed leaf area consumed as response variable and log-transformed initial leaf size as explanatory variable.Log transformations of the leaf area consumed and initial leaf area were performed to obtain normally distributed residuals.To test whether host plant identity impacts the preference of caterpillars, we fitted linear mixed effects (LME) models for each Lepidoptera species separately, where residuals of linear models described above were implemented as the response variable, host plant identity as the explanatory variable and host plant identity on which caterpillars had fed on before the start of the preference experiment as a random factor.Effects of host plant identity on caterpillar preference were estimated by performing Chi 2 tests of fitted models using ANOVA.2.3.2 | Caterpillar performanceTo test whether host plant identity and translocation of plants to different elevations (i.e.different climates) impact caterpillar performance, we exposed caterpillars (second or third instar) to lowand high-elevation host plants that were used at both elevations in the field experiment (Figure1c-f).After estimating initial weight, one caterpillar was placed in the centre of a Petri dish (9 × 1.4 cm) containing one leaf for Plantago spp., or one stem with several leaves for Lotus spp.For M. celadussa, caterpillars were placed into separate Petri dishes containing leaves originating from P. lanceolata individuals translocated at the experimental site situated at low (n = 42) and high elevation (n = 43), as well as P. atrata individuals, translocated to the experimental site situated at low (n = 46) and high elevation (n = 36).Similarly, for Z. filipendulae, caterpillars were exposed to leaves originating from L. corniculatus individuals translocated to the experimental site situated at low (n = 30) and high elevation (n = 27) as well as L. alpinus individuals translocated to the experimental site situated at low (n = 20) or high elevation(n = 30).Cut parts of leaves were wrapped in moist cotton and aluminium to keep leaves fresh(Descombes et al., 2017).After 4 days, caterpillars were weighed and the growth rate was calculated by subtracting the final from the initial weight.Differences in weight were divided by the number of days to estimate growth rate (mg/ day).
translocated to low elevation compared to plants remaining at high elevation.These results could indicate that a rapid increase in temperatures following the translocation of high-elevation plants to lower elevations may have altered the chemistry of plants and thus changed the palatability of P. atrata and of L. alpinus (González-Teuber et al., 2023).While the production of secondary metabolites in high-elevation plants remained constant across experimental sites, the translocation across elevations may have altered other phytochemical properties, as well as nutrient contents, of plants, which were not assessed here.For example, Descombes, Kergunteuil, et al. (2020) recorded a shift in the production of 76 untargeted metabolites following the translocation of P. atrata towards lower elevations.Apart from performing targeted metabolomic analyses, future studies should aim to explore responses using untargeted metabolomics as well as in primary metabolites to better understand how warming per se can change the entire metabolome of host plants and, further, how these potential shifts impact altered plant-insect herbivore interactions.

F
Relative concentrations of catapol and aucubin (a and b) in Plantago lanceolata and Plantago atrata and absolute concentrations of linamarin and lotaustralin (c and d) in Lotus corniculatus and Lotus alpinus.Bold, horizontal lines show medians, error bars illustrate distributions of the first to fourth quantiles, and points represent outliers.Asterisks indicate significant differences (p < .05)which were calculated using F-tests.Note that effects of elevation are not illustrated here, as translocation at experimental sites (1500/2150 m) had no impact on chemical defence compounds.For effects of elevation on chemical defence compounds, see Figure S8 and Table cialised insects are affected by climate change.Future studies should include a broader range of taxa and larger sample sizes to better understand the consequences for insect herbivores facing novel environments under climate change.Third, in our study, we examined the consequences for M. celadussa and Z. filipendulae experiencing a shift from their main host plants at low elevations to their closest highelevation relatives.Although P. lanceolata and L. corniculatus are the main host plants, neither species is strictly monophagous.As argued in the introduction, it is likely that M. celadussa and Z. filipendulae would prefer to establish on species that are closely related, as well as morphologically and chemically similar to their current host plants.However, it is possible that they may encounter other suitable host plant species following upward migration.These potential alternative host plants may increase the chances for these butterflies to successfully establish and persist at high elevations.Performing experiments to explore the ability of M. celadussa and Z. filipendulae to establish other potential host plants is important to better predict the impact of climate change on the persistence of these species.Fourth, the complex nature of climate change also involves alterations in various biotic interactions, such as plant-plant competition, which may interact with plant-insect dynamics to shape natural communities.Our controlled experiments simulated extreme climate change scenarios, exposing the two insect species to either current low-or novel highelevation host plants.Under realistic climate change scenarios, lowelevation insects are initially expected to encounter both low-and high-elevation host plants when migrating upslope.Our study suggests that low-elevation insect herbivores are likely to cause more damage to low-elevation plants by feeding and ovipositing more frequently on current low-elevation host plants.This could potentially counterbalance the negative impacts of highly competitive lowelevation plants on high-elevation species(Alexander et al., 2015), initially mitigating their effects(Kempel et al., 2015).Further research examining the combined impacts of multiple abiotic and biotic factors is crucial to understanding the consequences for species facing climate change.Overall, these considerations and caveats underscore the need for more comprehensive research to improve our understanding of how insect herbivores will adapt to novel environments under climate change.Future studies should estimate longer-term evolutionary consequences by conducting experiments across multiple generations to predict species' responses to climate change more accurately.AUTH O R CO NTR I B UTI O N S Baptiste Bovay: Conceptualization (equal); data curation (equal); formal analysis (lead); investigation (equal); methodology (lead); visualization (equal); writing -original draft (equal); writing -review and editing (equal).Patrice Descombes: Conceptualization (lead); formal analysis (equal); investigation (equal); methodology (equal); supervision (equal); writing -review and editing (equal).Yannick Chittaro: Conceptualization (equal); methodology (equal); writing -review and editing (equal).Gaétan Glauser: Formal analysis (equal); writing -review and editing (equal).Hanna Nomoto: Conceptualization (equal); data curation (equal); formal analysis (equal); investigation (equal); methodology (equal); supervision (lead); validation (equal); visualization (equal); writing -original draft (equal); writing -review and editing (equal).Sergio Rasmann: Conceptualization (lead); formal analysis (equal); funding acquisition (equal); investigation (equal); methodology (equal); project administration (equal); supervision (lead); visualization (equal); writing -original draft (equal); writingreview and editing (equal).