Multigenerational experimental simulation of climate change on an economically important insect pest

Abstract Long‐term multigenerational experimental simulations of climate change on insect pests of economically and socially important crops are crucial to anticipate challenges for feeding humanity in the not‐so‐far future. Mexican bean weevil Zabrotes subfasciatus, is a worldwide pest that attacks the common bean Phaseolus vulgaris seeds, in crops and storage. We designed a long term (i.e., over 10 generations), experimental simulation of climate change by increasing temperature and CO2 air concentration in controlled conditions according to model predictions for 2100. Higher temperature and CO2 concentrations favored pest's egg‐to‐adult development survival, even at high female fecundity. It also induced a reduction of fat storage and increase of protein content but did not alter body size. After 10 generations of simulation, genetic adaptation was detected for total lipid content only, however, other traits showed signs of such process. Future experimental designs and methods similar to ours, are key for studying long‐term effects of climate change through multigenerational experimental designs.


| INTRODUC TI ON
Current models of human activity and climate change predictions, according to the most pessimistic scenarios, foresee an atmospheric concentration of 1000 ppm of CO 2 and an associated global temperature increase of 6°C by 2100 (IPCC, 2007(IPCC, , 2014. Assuming that human population growth and food consumption follow its current trend (Bajželj et al., 2014), global crop production will require a 60% increase by mid-century to respond to food demands (Godfray et al., 2010). Notwithstanding, climate change will hinder this achievement in two ways: (a) crop production will encounter constraints due to plant productivity itself (Olesen et al., 2011;Waha et al., 2013), and (b) pest population dynamics and physiology will be altered (Bale et al., 2002;Harrington et al., 2001). While there is no easy way to deal with future plant productivity constrains, pest management can be adjusted (Estay et al., 2009(Estay et al., , 2014. Notwithstanding, we are still far away from understanding how pests will deal with realistic climate change scenarios (Bannerman & Roitberg, 2014;Gillespie et al., 2012;Haridas et al., 2016).
Typical variables associated to global change scenarios, namely increased temperature, and CO 2 , affect pest survival and/or fecundity. For example, development survival (measured from eggto-adult) of Heliothis viriplaca (Cui et al., 2018) and Thrips palmi (Yadav & Chang, 2014) increased with elevated temperature until reaching a certain threshold. Similarly, development survival, fecundity and parasitizing activity of Trichogramma buesi, increased with temperature until a maximal value was reached (Reznik et al., 2009).
Additionally, when the bean weevil Acanthoscelides obtectus grows at lower altitude, hence at higher temperature, fecundity increased, as well as egg hatching rate and ovarian production (Huignard & Biemont, 1978). These studies have also suggested that although an increase of temperature favors survival and fecundity, there is a temperature threshold that led trait expression start plateauing or collapsing, most likely because a maximum of metabolic rate was reached.
Resource allocation theory assumes that organisms have a limited amount of resources, which will be traded off among life history traits (Boggs, 2009;Deas & Hunter, 2014;Parker & Courtney, 1984;Pianka, 1981;Roff, 2002). For the case of females, a large proportion of their resources must be allocated to producing successful offspring. However, insect oviposition opportunities are often coerced to assign offspring to a limited patch of resources, such as seeds or insect hosts (Díaz-Fleischer & Aluja, 2003). Egg load, or the number of mature eggs a female is carrying (Ellers & Jervis, 2003;Harvey et al., 2001), is expected to shape the temporal (and spatial) variability in choices related to these oviposition resources. However, given that temperature is a major driver of insect lifespan, one expects oviposition strategy to covary with temperature. Such alteration will take place, for example, if death or loss of ability to reproduce is imminent (Sevenster et al., 1998). In this case, females will produce and lay eggs as soon as possible.
Selection and evolution of thermal reactions imply that environmental temperature and adult body size are linked in different geographical populations (i.e., Bergmann's rule). This relationship predicts that species living in colder conditions reach a larger adult size than species living in hotter climates (Bergmann, 1847).
Alternatively, the temperature-size rule stipulates that the plastic phenotypic response to increased temperatures can produce smaller insects by increasing developmental rate (Atkinson, 1994) as increased temperatures shorten insect life span (Papanikolaou et al., 2013). On the other hand, Bergmann's rule implies that environmental temperature and adult body size are linked in different geographical populations: species living in colder conditions reach a larger adult size than species living in hotter climates (Bergmann, 1847). In this regard, the same temperature-size rule, interestingly, stipulates that the plastic phenotypic response to increased temperatures can produce smaller insects by increasing developmental rate (Atkinson, 1994). Body size and temperature relationship rules have not been corroborated and are consequently not as straightforward as theory predicts (Angilletta & Dunham, 2003). To solve this, it has been suggested that a better approach might be to generate and test theories that are tailored specifically to organisms with similar behavior and physiology (Angilletta & Dunham, 2003;DeLucia et al., 2012). Indeed, thermal response is rather the expression of the coevolution of thermal reaction norms for growth rate and size at maturity than a simplistic response that focuses on one or two mechanisms influencing life history (Angilletta et al., 2004).
Temperature has been described as a factor altering insect's body lipid and protein levels (Gligorescu et al., 2018;McCue et al., 2015).
One illustrating case is that of the beetle Ophraellla communa whose lipid and glycogen storages decrease and increase respectively when the insect was exposed to daily phasic high temperatures (Chen et al., 2019). Changes in metabolic rates leading to anatomical and physiological alterations are the most evident expected consequences of global warming on insects (González-Tokman et al., 2020;Sheridan & Bickford, 2011). However, combined effects of elevated temperature and carbon dioxide have been described to mitigate each other (Zvereva & Kozlov, 2006). Hence, insect body size and lipid reserves are expected to diminish (Atkinson, 1994) due to a higher metabolic rate as well as a higher total protein content produced by hydric stress and development time reduction (Papanikolaou et al., 2013). Simultaneously, fecundity is expected to increase (Huignard & Biemont, 1978), and larval development survival to decrease because of oviposition time compensation and lesser per-egg investment as females experiencing elevated temperature dispose of a shorter time window to lay eggs and harsher conditions are more likely to affect survival.
Insect responses to future climatic conditions are usually explored using the following approaches: assessment of current impacts of climatic changes based on accumulated data from the past (Andrew et al., 2013), bioassays testing climate drivers on a short-term scale (Dyer et al., 2013), field monitoring using a geographical gradient (Hodkinson, 2005;Read et al., 2014;Slatyer & Schoville, 2016), meta-analyses (Saban et al., 2019), and computer models predicting future scenarios (Estay et al., 2009;Northfield & Ives, 2013). Besides purely in silico models, most approaches tend to compile data to produce some predictions based on present or past conditions, which is relevant for extrapolations or climate change simulations on a short-term scale. Despite these reasons, only a handful of investigations have used experimental designs lasting longer than 3-5 generations or explored the impacts of climate change using multigenerational experimental designs. Most of these studies have concerned marine organisms and focused on a single climate driver such as temperature (Munday et al., 2017;Shama et al., 2016), acidification or water pCO 2 (Rodríguez-Romero et al., 2016). One exception to these studies where temperature and pCO 2 have been integrated is that with the marine polychaeta Ophryotrocha labronica (Gibbin et al., 2017).
As a matter of fact, the various climate change drivers tend to offset each other's effects (Gibbin et al., 2017;Kroeker et al., 2013).
Consequently, it seems reasonable to consider "climate change", that is, increased CO 2 concentration and temperature as a single factor.
The idea of a multigenerational selection experiment is to test the magnitude of rapid evolution. Hence, in order to discriminate whether a given phenotype is explained by plasticity or a genetic basis, a reciprocal transplant appears to be a powerful tool (Ågren & Schemske, 2012;Svensson et al., 2018). This technique was originally designed for detecting local adaptation between geographically distant populations or within a metapopulation pooling demes sharing geneflow (Blanquart et al., 2013;Kawecki & Ebert, 2004).
Interestingly, some recent studies used reciprocal transplants to measure the adaptive change in a multigenerational simulation of climate change on marine species (Gibbin et al., 2017;Rodríguez-Romero et al., 2016).
In this study, we investigated the impact of global change conditions on an insect pest's life history traits, physiology, and phenotypic plasticity. Our work is novel for the following reasons: (a) only few studies have been focused on pests' evolutionary responses to climate change; (b) we simultaneously manipulated the two driving factors of climate change; temperature and CO 2 concentration, and (c), a multigenerational approach is used. We used the Mexican bean weevil Zabrotes subfasciatus Boheman as a study subject and had the following specific aims: (i) to estimate and project the modulation of the insect´s fecundity and development survival, (ii) to measure the impact of 2100 predicted climatic conditions on body size and total protein and lipid content, (iii) and to detect whether 10+ generations settles genetic adaptation or whether phenotypical plasticity is solely responsible for any measured effect on insects. Hence, we expected insects' fecundity to increase (Huignard & Biemont, 1978), and larval development survival to decrease because of oviposition time compensation and lesser per-egg investment as females experiencing elevated temperature dispose of a shorter time window to lay eggs and harsher conditions are likelier to affect survival.
Simultaneously, we predicted that insect body size and lipid reserves would diminish (Atkinson, 1994) due to a higher metabolic rate as well as a higher total protein content produced by hydric stress and development time reduction (Papanikolaou et al., 2013). Finally, we also expected that genetic adaptations would be measurable by the end of the experiment and that more than phenotypic plasticity would be observed as over 10 generations have been described as more than sufficient to trigger adaptative responses (Christie et al., 2012;Laukkanen et al., 2018).

| Study system
The Mexican bean weevil Zabrotes subfasciatus Boheman (Coleoptera; Chrysomelidae; Bruchinae; Amblycerini) is a worldwide pest that affects crops and stored products of the common bean Phaseolus vulgaris L. This weevil is responsible for substantial agricultural damage, mostly in the New World as well as in Africa and Asia where the common bean is massively produced. The insect is sexually mature and ready for copulation immediately after emergence.
Indeed, females typically lay their eggs at the very beginning of their imago life, with a peak of oviposition reached within few days (Sperandio & Zucoloto, 2004). As a capital breeding animal, Zabrotes adults do not feed and instead use the resources accumulated during larval development (Teixeira et al., 2009;Teixeira & Zucoloto, 2002).
Consequently, females mature eggs from a limited amount of reserves and then stick them on the bean seed coat. The first instar larva will hatch and bore into the cotyledon where it will establish a larval chamber for its 30-50 days long juvenile life. A fully developed imago will emerge by cutting its way out of the seed coat.

| Insect collection and rearing
Wild Z. subfasciatus were obtained from Phaseolus lunatus (L.) seeds collected along the South Mexican Pacific coast (Figure 1). Four locations were selected based on their relative distance (more than 15 km apart from each other) and on the number of emerging beetles (more than 15 individuals per 10 gr of wild seeds): Las Salinas Furthermore, colonies were split another 2 times as the populations expanded and kept in 15 cm long side cubic glass jars each containing 2 kg of organic black bean seeds (Phaseolus vulgaris variety Negro Queretaro). Every 2 months, beans from all colonies were sieved, dead adults were discarded, as well as 500 g of infested seeds, and living adults were all randomly redistributed to all 20 jars with 500 g of fresh bean seeds. Despite this control of seeds, population size for all stages of our experiment cannot be calculated as beetles often hide inside the seeds.

| Experimental colonies settings and climate change simulation
Based on the IPCC model predictions (IPCC, 2007(IPCC, , 2014

from 2007
to 2014 (scenario A1F1 and more recently RCP8.5), an increase of global mean temperature of 6°C and a shift of atmospheric CO 2 concentration from 370 ppm to 1,000 ppm was selected. Since these values correspond to the worse scenario of the IPCC (2014), we decided to adopt them as the predictions from the 80's were fairly optimistic regarding the current climatic situation (Hansen et al., 1981).

Two incubators (Precision Model 818, Thermo Fisher Scientific Inc)
were used for the climate change simulation, a first as control (L/D: F I G U R E 1 From top to bottom: Field collection of wild Zabrotes subfaciatus in the Mexican South West pacific coast, laboratory colony creation and growth during 10 generations, long-term selection experiment with climate change simulation (2100) lines in the red box, and control lines (2017) in blue box, followed by the reciprocal experiment allowing the assessment of the genetic component of the multigenerational climate change simulation. The central black box described the technique permitting the isolation of less than one-day-old adult beetles for subsequent analysis and/ or bioassays. The left side chronogram resumes the different phase of the whole experimental process as well as the time range in terms of generations and sampling frequency

| Selection process monitoring
For 18 months, the colonies were monitored during the multigenerational exposure to simulated climate change and control conditions ( Figure 1). Every week, a volume of 60 cm 3 of seeds from every single jar was sieved to remove adult beetles and isolated in 5 × 5 × 4 cm plastic containers. 12 hr after, less than 1-day-old new imagoes emerged. This simple procedure of sieving-waiting-collecting provides younger than 1 day old, newly emerged adult beetles ( Figure 1) and was used repeatedly in this study. A maximum of 10 individuals per sieved samples were randomly collected in order not to affect the population dynamics of the colonies. Every sampled insect was frozen killed and kept at −20°C until the end of the experiment. Any replica jar that would fail to provide adult insects for 3 consecutive weeks was permanently discarded.

| Life history traits measurements
After 180 days of experimental simulation and using the same process of sieving and sampling as previously described, 1-day-old adult beetles were collected from each experimental jar. Males and females from the same experimental jar were randomly grouped by pairs before being deposited in 5 × 5 × 4 cm plastic boxes containing 10 bean seeds using the set.seed() function in R (R Development Core Team, 2013). Each pair was allowed to mate and lay eggs for 7 days. Each box was checked daily and any seed with eggs was removed and isolated after counting the eggs. Consequently, an emerging offspring individual could share siblings in the same seed, as well as in another seed (maximum 9 other seeds) oviposited by the same parents. Given this, replicates are nested by seed, parents and colony jar (c.f. statistical analysis). The larval development took place inside a modified 1.5 ml Eppendorf tube that was perforated 10 times with a Ø 0.7 mm needle to allow gas exchange with the controlled chamber's environment. This process was repeated every two weeks for each replica jar from day 200 to day 460. All tube-isolated seeds were checked daily for emerging adults, which were immediately collected, frozen killed and kept at −20°C for the subsequent measurements. This protocol allowed to record larval development time, fecundity (number of eggs per pair i.e. female) and survival (number of emerging offspring/number of eggs laid per pair), and prevented the young adults from using body-stored energy resources such as lipids before being collected.

| Reciprocal transplant experiment
After 400 days of experimental simulation, we performed a reciprocal transplant between both 2017 and 2100 chambers. Using the same protocol for measuring life history traits described above, four experimental combinations of insect and chamber were set.
However, to minimize maternal effects, F2 offspring were used to assess all measured traits. In other words, the sons and daughters of the insects emerging after the sieving process were used to provide the experimental individuals on which we performed all measurements. Our experimental groups were set as follows: (a) two control groups were arranged by introducing 2017 and 2100 insects into 2017 and 2100 chambers respectively ( Figure 1); and, (b) two experimental groups. For these, a 1 day old, freshly emerged pair of adult beetles from 2017 were offered 10 seeds to oviposit and were then introduced into the 2100 chamber, whereas 2100 insects were introduced into the 2017 chamber. Each pair could oviposit for 7 days and was discarded afterward. Daily, all seeds were examined for fresh eggs, if one or more was found, then the seed presenting the egg(s) would be isolated into an individual 1.5 ml Eppendorf tube and kept in the same chamber until emergence. The remaining eggs were kept in the box until the end of the 7-day oviposition phase. This process was repeated every two weeks until the end of the experiment (when the chambers were shut down, i.e. 240 days after the start of the reciprocal transplants). In the same manner as previously described, seeds were checked daily for emerging adults, which were immediately frozen killed and stored at −20°C for further analysis. For convention purposes and to simplify interpretation, the following terminology will be used: the 2017 chamber is referred to as "home" while the 2100 chamber is referred to as "away" for 2017 insects, and reciprocally for 2100 insects.

| Body size, weight, total protein, and lipid measurements
The length (from the anterior end of pronotum to the posterior end of pygidium) of each bruchid beetle was measured by digital photography and pixel-based measurement using Image J (Schneider et al., 2012) and weighted using a digital Cahn microbalance (Thermo Fisher Scientific Inc). Protein and lipid contents were assessed using a shortened version of a sequential colori- After a low-spin centrifugation of 180 g for 30 s, protein contents were measured using a simple Bradford essay (Bradford, 1976) having bovine serum albumin as standard. Absorbance at 595 nm was subsequently recorded with an Absorbance Reader ELx 800 spectrophotometer (BioTek, Inc.). Secondly, lipids were solubilized with 1,000 μl of chloroform-methanol solution (1:2 v/v) and their concentration was measured with the classic vanillin assay procedure (Van Handel, 1985). Triolein was used as standard, and absorbances were read at 540 nm using the same spectrophotometry equipment.

| Statistical analysis
To test for an effect of climate change simulation on body size, protein content, lipid content as well as development time data over colonies age, we used linear mixed-effects model (LMM) via restricted maximum likelihood (REML with Satterthwaite-approximated degrees of freedom for the fixed effect), with replica jar as a random factor (Kuznetsova et al., 2017).
Regarding the life history trait approach, firstly, a GLMM (general linear mixed model) allowed us to fit by maximum likelihood (Laplace approximation) the variable development survival rate. Secondly, additional GLMM models fitted body size, protein content, lipid content, and development survival with fecundity and the number of generations, using replica jars as a random factor. The latter models were visualized as planes into a 3 days graphic representations when the models were significant.
Reciprocal transplants data were analyzed using a different approach. Primarily, for each measured trait, that is, body length, protein and lipid contents, time to emergence and fecundity, a linear mixed model including the original "home" treatment as factor for genotype (G) and the destination "away" treatment as environment factor (E) were designed in order to determine whether the variance can be attributed to genetic adaptation and assess how each trait evolved. These models were fitted by restricted maximum likelihood using Satterthwaite's method with jar replica, mother ID, and seed ID as random factors. Secondly, internal multiple comparisons were performed on least square means of the models. Regarding the predictor survival, a GLMM was performed as the data had a binomial distribution, and a Tukey (Contrasts) test permitted multiple comparisons of means in this specific case.

| Life history traits
Fecundity alone significantly explained development survival rate (z = −3.175, p = .001496), as well as when interacting with the treatment (z = 3.564, p = .000365). Indeed, development survival decreased as fecundity increased, however, the 2100 model seemed to maintain higher probability of completing development than the control group as fecundity increased (Figure 2a).
Overall development survival rate decreased over generations in the control chamber while the treatment showed a mild increase independently of fecundity ( Figure 2b). Nonetheless, when fecundity is considered, the control plane shows a stronger negative inclination as fecundity increases throughout the experiment. In other words, survival decreases strongly and significantly (Table 1)     of replica convergence show that variation of replicas variance was significant for body size and total protein but not for total lipids and the development survival rate versus fecundity data (Table 2).

| Reciprocal transplant experiment
All statistical values and details are listed in Table 3. For all bioassays performed in this study, 5.4 individuals in average depending on the fecundity of the mother, and an average of 2.7 seeds per mother. Body size content shows the same pattern of increment when insects are

| D ISCUSS I ON
In general, our results suggest that climate change alters life history strategies. For example, higher temperature and CO 2 concentrations favored egg-to-adult development survival of Z. subfasciatus.
Despite this, one would expect a higher mortality due to lesser peregg investment which remains true even at high female fecundity ( Figure 2a). Indeed, when the component of evolutionary time is added to the model (Figure 2b), developmental survival maintains itself across generations independently of fecundity. However, it is necessary to mention that the fecundity decrease observed in the control group may be the response of laboratory selection for increased fecundity but also resulting in decreased survival.
As Z. subfasciatus development survival appears to increase, even in the case of high maternal fecundity, several potential explanations can be put forward. First, egg quality is greater in 2100 conditions, meaning that the ovipositing females are capable to provide more viable eggs despite the cost of laying more eggs. Second, since Z. subfasciatus is a capital breeder, the physiological assignment of resources during larval development is shifted from egg number to egg quality. Third, eggs and larvae simply develop better in the conditions we simulated as this species has a relatively wide temperature tolerance but a thermal optimum of 27-30°C (Sperandio & Zucoloto, 2009). In other words, our data suggest that augmented temperature and CO 2 have the potential to lead to an increased fitness as the females seem to change their ovipositing strategy by laying fewer eggs and the developing larvae show a higher probability to reach the imago stage. Before moving on to the next section, it is important to mention that formally estimating the absence of drift is difficult as we had no total knowledge of the genetic diversity of the founder individuals used to start each colonies.
Insect body size for the 2100 simulation did not diminish throughout the experiment which is contrary to what we predicted.
Rather, there was a tendency of a reduction in body size of both control and treatment groups which is likely an effect of the artificial environment. In this regard, distinct climate change drivers tend to offset each other's effects (Gibbin et al., 2017;Kroeker et al., 2013), also, body size is usually poorly described by general rules (Angilletta & Dunham, 2003;DeLucia et al., 2012) Given that our study includes CO 2 , is interesting yet not surprising to observe no body size difference between the control group and the 2100 treatment. We predict that using identical experimental designs involving either increased temperature or CO 2 would, however, provide different outcomes.
Worth mentioning is the fact that body size and temperature relationship rules are being repeatedly broken and are consequently not as straightforward as theory predicts. To fully clarify these body size responses, a better approach might be to generate and test theories using organisms with similar behavior and physiology (Angilletta & Dunham, 2003;DeLucia et al., 2012).
Protein content increased and lipid content decreased as predicted for 2100 insects. This may be a consequence of mechanisms for coping with dehydration and a different assignment of energy reserves and metabolic water. Indeed, increased metabolic flow due to higher temperature and enhanced uncoupling of mitochondrial respiration from oxidative phosphorylation, allow for greater production of metabolic water in insects in "dry environments" (Jindra & Sehnal, 1990). A similar pattern has been previously observed in a short-term experiment using Acanthoscelides obtectus, a bruchid beetle species close to Z. subfasciatus, when temperature increased from 20 to 30°C (Sönmez & Gülel, 2008), but no relevant explanation was found regarding the context of our multigenerational simulation. Consequently, as triglycerides yield almost two times more metabolic water than glycogen (Arrese & Soulages, 2010), fat storage would have been compromised in a dryer environment during larval development. Hence, the pupation process might have further reduced the fat storage through metabolic water extraction as the insect stops feeding and experiences costly transformations in dryer air conditions. This is coherent with findings in other insects. For example, the tsetse fly Glossina spp. uses lipid storage during pupation for water balance control depending on the ambient humidity and temperature. Moreover, fat consumption increases with the temperature while the pupal period reduces (Bursell, 1958(Bursell, , 1960Kleynhans & Terblanche, 2009). Regarding the clear positive slope of protein content in 2100 group, one explanation is that it shows desiccation and increasing tolerance to desiccation. As the experiment progressed, it is likely that individuals from the 2100 group achieved a greater tolerance to warmer and dryer conditions.
Possibly these animals afforded to be functional with less water in their environment and body as the lipid content stabilizes over time.
This is in agreement with studies in Drosophila melanogaster which were selected for increased desiccation resistance (Telonis-Scott et al., 2006), While studies on heat shock proteins and other temperature stress-related processes are common (Adamo, 2012;Sørensen & Loeschcke, 2007;Sørensen et al., 2005;Wang et al., 2014), very scarce data are available on total protein and lipid contents. In a similar way, numerous studies have provided evidence of, for example, the indirect role of CO 2 through plant tissue alteration (Cornelissen, 2011;Knepp et al., 2005;Murray et al., 2013;Xu et al., 2019) or behavioral approaches on hematophagous insects using CO 2 as a pointer (Guerenstein & Hildebrand, 2008;Jones, 2013;Lazzari et al., 2013;McPhatter & Gerry, 2017). Despite this, literature on direct effects of CO 2 is scarce, aside from studies testing extreme cases such as over 20% of CO 2 (Mitcham et al., 2006), as it has been considered that direct metabolic effects of CO 2 on herbivorous insects are insignificant when the effect on the plant is removed (Coviella & Trumble, 1999). However, laboratory studies using CO 2 anesthesia on Drosophila melanogaster show that metabolic changes persist 14 hr after acute CO 2 exposure (Colinet & Renault, 2012;Nilson et al., 2006). This fact should be considered to understand that a The multigenerational simulation of climate change conditions provided a clear contrasting pattern between anatomical and physiological data. As suggested by literature (Angilletta & Dunham, 2003), body size and temperature were not tightly associated by the temperature-size rule (Atkinson, 1994) as no difference was observed between 2017 and 2100. This would fit with a recent metanalysis that explored the species-specificity of the temperature-size rule (Klok & Harrison, 2013). This study indicated that large interspecific variation is either explained by strong interactions with nutrition, or selection based on microclimatic or seasonal variation not captured in classic macro-environmental variables. Indeed, the clear patterns of protein content increase and lipid decrease (relatively to control) imply that metabolic changes are undergoing while the overall exoskeleton size remained unchanged. Interpreting the body size and total protein data must be done with the knowledge that not all replicas converged toward the same outcome as they did for the other measured traits. A small initial population size could have generated such a phenomenon, but some traits converged while others did not, suggesting a greater starting effective population for future experimental designs.
Besides the anticipated faster growth in elevated temperature, an interesting pattern of reduced variance was observed in the development time data. This pattern is actually coherent with previous bioassays (Marinho et al., 2015) and models (Régnière et al., 2012) addressing the impact of temperature on this parameter. Indeed, development time variance tends to shrink at warmer temperatures only if a metabolic and/or physiological threshold is not reached or verged upon (Régnière, 1987). A narrower temporal phenology might reduce variability in ecosystems and agrosystems and is also prone to desynchronize plants-herbivores-predators. Another aspect that should be considered for further investigation is that development time can be offset by temperature fluctuation range (Xing et al., 2015) and a greater climatic, hence, temperature stochasticity is expected with global warming predictions (IPCC, 2007(IPCC, , 2014. Therefore, longer development time and larger variances are expected under a more realistic climate variability simulation. The overall pattern leads to hypothesize that a pest insect such as Z. subfasciatus could in fact be advantaged when facing elevated temperature and CO 2 levels. Bean seeds providing a micro-environment, the growing larvae are virtually affected by a highly similar number of factors in the field and in a laboratory colony, aside from, of course, parasitoid attacks which are common predators in this system (Schneider & Córdoba-Aguilar, 2019a). Given that the young beetles spend over one month encapsulated into their seed, larvae are protected from most physical factors such as atmospheric and mechanical damages. However, Chacidoids and Braconids parasitoid wasps would obviously be able to attack the protected larvae inside the bruchid larval chamber and then constrain population growth (Aebi et al., 2008;Schneider & Córdoba-Aguilar, 2019b).
Our reciprocal transplant experiment produced interesting patterns of phenotypic variation. However, genetic adaptation can be attributed to the variance of the total lipid content only, despite our  (Zvereva & Kozlov, 2006 as an organism is adapted to 2100 conditions. Nonetheless, it is well known that different organs, pathways, and genes evolve at different speeds (Gillespie, 1986;Wilke, 2004;Zhang & Townsend, 2009) Plasticity is usually expected to enable organisms to cope with fast-changing environments (Gienapp et al., 2008). However, even though plasticity mostly occurs within a generation, it has been reported that the conditions experienced by one generation could interact with the conditions experienced by the subsequent generations (Donelson et al., 2018). This phenomenon is known as transgenerational plasticity (TGP), which is likely to take place in a reciprocal transplant experiment as logistics do not fully discard maternal and paternal effects as well as epigenetic transmission (Donelson et al., 2018;Shama et al., 2016). In fact, we cannot discard maternal effects due to the ovipositing behavior (several eggs per seed and on several seeds) of Z. subfasciatus, we should discuss our results consistently with this fact. Hence, we cannot exclude TGP as the outcome of our study. It is true that Zabrotes, being a worldwide spread multivoltine pest, should have a great potential for phenotypic plasticity (Aebi et al., 2008;Alvarez et al., 2006;Cuny et al., 2017). Consequently, the experiment should be performed for a greater number of generations. Ideally a similar design could be easily implemented using faster developing insects such as Drosophila spp. or sister genders causing economic damage such as Rhagoletis. Another approach would be to use full genome transcriptomics, as mentioned previously, or even more modern tools such as "Evolve and resequence" (Schlötterer et al., 2015) to pinpoint which genes or gene cluster regulations are affected by increased temperature and CO 2 .

| CON CLUS ION
The role of life history traits plasticity and evolution has been overlooked in climate change ecology (Donelson et al., 2018;Lancaster et al., 2017). This fact is probably explained by the lack of multigenerational experimental data. Our study provides such data and helps guiding the way to more realistic predictions in climate change biology. Moreover, it informs that elevated temperature and CO 2 together affect the physiology, life history traits and the evolutionary direction of a laboratory raised colony of Zabrotes subfaciatus. So far, it seems that this pest will deal with climate change by adjusting mainly survival and physiological traits. Future research should look at whether such changes imply higher costs for plant productivity and thus risks for food security.

ACK N OWLED G M ENT
We

CO N FLI C T S O F I NTE R E S T S
None declared.

DATA AVA I L A B I L I T Y S TAT E M E N T
Multigenerational experimental simulation data & reciprocal experiment data: Dryad https://doi.org/10.5061/dryad.h44j0 zph5.