ABSTRACT
The physiology of insects is directly influenced by environmental temperature, and thermal tolerance is therefore intrinsically linked to their thermal niche and distribution. Understanding the mechanisms that limit insect thermal tolerance is crucial to predicting biogeography and range shifts. Recent studies on locusts and flies suggest that the critical thermal minimum (CTmin) follows a loss of CNS function via a spreading depolarization. We hypothesized that other insect taxa share this phenomenon. Here, we investigate whether spreading depolarization events occur in butterflies exposed to cold. Supporting our hypothesis, we found that exposure to stressful cold induced spreading depolarization in all 12 species tested. This reinforces the idea that spreading depolarization is a common mechanism underlying the insect CTmin. Furthermore, our results highlight how CNS function is tuned to match the environment of a species. Further research into the physiology underlying spreading depolarization will likely elucidate key mechanisms determining insect thermal tolerance and ecology.
INTRODUCTION
Insects have limited ability to regulate their body temperature, and environmental temperature is therefore one of the main factors influencing physiological functions and, by extension, their behaviour and fitness (Heinrich, 2013). Consequently, exposure to stressful thermal conditions constrains insect performance, and the ability of insects to tolerate thermal extremes is tightly associated with their fundamental niche range limits (Andersen et al., 2015b; Kellermann et al., 2012; Kimura, 2004), a pattern that holds true for ectotherms in general (Addo-Bediako et al., 2000; Sunday et al., 2011). This pattern also applies to butterflies, where thermal tolerance traits have been used in attempts to model range limits and changes therein (Crozier, 2004; Dongmo et al., 2021; Keefe, 2023; Tremblay et al., 2021). Thus, gaining an integrative understanding of the physiological mechanisms limiting thermal tolerance and distribution of butterflies is critical if we want to improve conservation efforts for some of the most charismatic, yet vulnerable species (Pacifici et al., 2015; Parmesan et al., 1999; Thomas et al., 2004).
Several traits representing thermal tolerance limits can be measured in insects, the most common of which is the loss of movement at a temperature referred to as the critical thermal minimum or maximum (CTmin or CTmax; see Chown and Nicolson, 2004; Hazell and Bale, 2011; Lutterschmidt and Hutchison, 1997). At low temperature, the gradual loss of function has been separated into multiple distinct, yet mechanistically intertwined physiological events. Under the current consensus, CTmin refers to a gradual loss of coordinated movements quickly followed by paralysis as chill coma sets in (Andersen and Overgaard, 2019; MacMillan, 2019; Robertson et al., 2017). The physiological mechanism(s) underlying the CTmin and chill coma have been sought since cold-induced comas were first discovered in insects (MacMillan and Sinclair, 2011; Mellanby, 1939). The ability to sustain movement is intrinsically linked to neuromuscular signalling, and research has therefore been focused on the temperature sensitivity of neuronal and muscular physiology (Andersen et al., 2015a; Anderson and Mutchmor, 1968; Rodgers et al., 2010; Staszak and Mutchmor, 1973). Recent studies on fruit flies and locusts have found a strong correlation between the initial loss of coordination occurring at the CTmin temperature and a loss of central nervous system function caused by a spreading depolarization (Andersen et al., 2018; Andersen and Overgaard, 2019; Robertson et al., 2017). Specifically, the spreading depolarization event silences central, integrating neurons because of a rapid surge in the brain's extracellular K+ concentration (Robertson et al., 2020; Rodgers et al., 2010). The exact mechanisms underlying this event remain elusive, but multiple lines of evidence point towards impairment of ionoregulatory pumps and channels in perineurial glia that maintain the local ionic environment surrounding neurons (Andrew et al., 2022; Robertson et al., 2020, 2023).
Importantly, the proximate phenotype for spreading depolarization (i.e. CTmin/max) is a widespread phenomenon in insects that is often used to capture discrepancies in thermal tolerance, predict species distributions, range expansions and sensitivity to global climate change (Kellermann et al., 2012; Lancaster, 2016; Weaving et al., 2022). However, so far, temperature-induced spreading depolarizations have only been demonstrated in fruit flies and locusts (Andersen et al., 2018; Jørgensen et al., 2020; Robertson et al., 2017; Spong et al., 2016); whether or not this neurophysiological limit is common to all insects is unclear. Indeed, the presence or absence of the spreading depolarization event might provide insight into the evolutionary relevance and potential adaptive nature of the event (Robertson et al., 2020). In the present study we sought to investigate whether the spreading depolarization phenomenon occurs in another insect order by monitoring the transperineurial potential during a temperature ramp in a range of tropical butterfly species. Chill coma phenotypes have been observed in butterflies, so we expected to see spreading depolarization at temperatures approximating those leading to coma in related species (Andersen et al., 2017; Keefe, 2023). Subsequently, we tested if this putative limit to organismal function (i.e. loss of central neural function) had any relation to species biogeography. For this, we built a phylogeny for our model system to correct for species relatedness, which further allowed us to investigate phylogenetic signals, which we expected to be lower than in, for example, drosophilids, as variation in the chill coma temperature of butterflies have been reported to be much lower.
MATERIALS AND METHODS
Animal husbandry
All butterflies (see Table 1) were bought and imported as pupae from a commercial supplier (LPS LLC Butterfly Import, Denver, CO, USA) with permission (permits P-2018-02344, P-2018-02345, and P-2020-02651), housed under common garden conditions in flight cages at ∼30°C until emergence and left to mature in their flight cage or in a greenhouse (28–30°C) for 1–3 days before being used in experiments.
Experimental procedure
On the day of the experiment, individual butterflies were gently moved from a holding cage to the lab and immobilized in wax on top of a custom-built thermoelectrically cooled plate. Once immobilized, a small cut was made in the head immediately above the antennae, and a glass microelectrode was inserted into the brain through the blood–brain barrier with a micromanipulator (Spong et al., 2016). A reference electrode (Ag/AgCl) was inserted into the hemolymph through a small hole cut in the abdomen. The glass microelectrodes were fashioned from 1 mm diameter borosilicate glass capillaries (1B100F-4, World Precision Instruments, Sarasota, FL, USA) pulled to a tip resistance of 5–7 MΩ in a Flaming-Brown P-1000 micropipette puller. Before use, glass electrodes were backfilled with 500 mmol l−1 KCl and placed in an electrode holder with an Ag/AgCl wire. Both glass and reference electrodes were connected to a Duo 773 two-channel intracellular/extracellular amplifier (World Precision Instruments, Sarasota, FL, USA), and the output was digitized with a PowerLab 4SP A/D converter (ADInstruments, Colorado Springs, CO, USA) and fed to a computer running Lab Chart 4 software (ADInstruments).
To approximate brain temperature, a type K thermocouple was placed immediately next to the head and at the same height as the tip of the glass microelectrode. Once ready, temperature was manually lowered from room temperature (∼22°C) by ∼1°C min−1 while the transperineurial potential (the electrical potential difference across the blood–brain barrier) and head temperature were continuously recorded. Loss of CNS function caused by a spreading depolarization was identified when an abrupt negative shift in the transperineurial potential (∼20–45 mV in 5–10 s; see Spong et al., 2016) was observed, and the temperature of this event was estimated as the temperature at the half amplitude of the negative shift in potential (see Fig. 1 for examples). After experiments, animals were dissected to determine their sex. Sample sizes are depicted in Fig. 2A.
Biogeography and climate data
Information on the geographical distribution of all 12 butterfly species used in the present study was extracted from the Global Biodiversity Information Facility (GBIF; see Table S1 for references to the species-specific datasets). However, we combined these observations with iNaturalist observations, as the GBIF had limited observations of two species (Heliconius atthis and Caligo memnon). Before obtaining climate data, the distribution datasets were cleaned up by (1) removing duplicate observations (two decimal accuracy in latitude and longitude) and (2) removing observations more than three standard deviations away from the mean latitude (Gotelli and Ellison, 2004). This resulted in a total of 18,030 observations (per species: N=45–4539; mean=1502). After clean-up, distributional data was converted into spatial data points using the SpatialPoints() function from the ‘sp’ package (https://CRAN.R-project.org/package=sp). Next, climate variables were downloaded from the WorldClim dataset (Fick and Hijmans, 2017) (http://worldclim.org/; accessed 30 September, 2021) in raster format at a resolution of 2.5 min (∼ 20 km2) and imported into R using the getData() function from the ‘raster’ package (https://CRAN.R-project.org/package=raster). Lastly, climate data was extracted from the spatial data points for each species using the extract() function (also from ‘raster’). Following extraction, mean and standard deviation of the climate variables was calculated for each species. For the purpose of investigating latitudinal distributions, we decided to use the absolute latitude to avoid distribution latitudes based on observations above and below the Equator (only for the sake of correlations; not for climate variable extractions) (Sunday et al., 2019; Sunday et al., 2011). Given the focus here was cold tolerance, we focused on the following biogeographical data: (1) mean latitude of observed distribution (degrees latitude from Equator); (2) mean altitude of the observed distribution; (3) annual mean temperature (in °C); (4) minimum temperature of the coldest month (in °C); (5) the mean temperature of the coldest quarter (in °C).
Statistical analysis
All statistical analyses were performed using R software (v. 4.2.2; https://www.r-project.org/). A two-way ANOVA was initially used to determine whether spreading depolarization temperature differed among species and sexes. No interaction was found between sex and species so this interaction was removed from the analysis. An effect of sex was found, but sex was not included as a factor in subsequent analyses because this would result in extremely low sample sizes (see Table S2) and because only females were sampled for one species (Siproeta stelenes).
To appropriately correlate spreading depolarization temperature to biogeographical data and climate variables, we wanted to assure that species relatedness would not confound our results (i.e. the close relationship between Heliconius species). Thus, we constructed a time-calibrated phylogeny with all 12 species in R using the ‘ape’ package (https://CRAN.R-project.org/package=ape; Paradis et al., 2004), based on the published nodes of divergence (Kawahara et al., 2023). Then, to correct our data for phylogeny, we calculated the phylogenetically independent contrasts (PICs) based on our constructed phylogeny using the pic() function from the ape package. Subsequently, PICs for spreading depolarization temperature were correlated to the biogeographical parameter PICs using linear regression through the origin (Garland et al., 1992). Phylogenetic signals (Bloomberg's K) were calculated using the phylosig() function from the ‘phytools’ package (https://CRAN.R-project.org/package=phytools; Revell, 2012). The threshold for statistical significance was 0.05 for all analyses, and values references below refer to mean±s.e.m. unless otherwise stated. Lastly, we constructed an additive linear model containing the PICs for all biogeographical parameters investigated here, and used the stepAIC() function from the ‘MASS’ package (https://CRAN.R-project.org/package=MASS) to run a stepwise model reduction based on the Akaike information criterion, to obtain the best predictive model for cold-induced spreading depolarization (output can be found in Table S3).
RESULTS AND DISCUSSION
Cold-induced spreading depolarization in a subset of tropical butterflies
A cold-induced spreading depolarization was observed in all animals tested (64/64) as observed for locusts and fruit flies previously (Rodgers et al., 2007; Spong et al., 2016): an abrupt negative shift in transperineurial potential of approximately 20–45 mV occurred over 5–10 s, which could be reversed when the animal was returned to a permissive temperature (example traces in Fig. 1). Spreading depolarization has also been observed in locusts and fruit flies when exposed to stressful heat and anoxia (Jørgensen et al., 2020; Robertson and Van Dusen, 2021; Rodgers et al., 2007) and it has been experimentally induced in cockroaches (Schofield, 1990). This phenomenon has been linked to the lower thermal limit for coordinated movements (i.e. the CTmin; see Andersen et al., 2018; Andersen and Overgaard, 2019; Robertson et al., 2017) and although we did not quantify behavioural CTmin, we did note that butterflies ceased moving at temperatures around those leading to spreading depolarization. Thus, the loss of CNS function by a spreading depolarization has now been observed in four distantly related insect orders, encompassing both holo- and hemimetabolous groups. This generality suggests that spreading depolarization is a neurophysiological phenomenon shared by all insects, and by extension, that the critical thermal minimum of all insects might be caused by spreading depolarization.
Spreading depolarization events are characterized by a collapse of ionoregulatory capacity, resulting in a rapid surge in the extracellular K+ concentration within the central nervous system. The processes involved in ion balance regulation in the insect central nervous system are numerous, but a key transporter is the Na+/K+-ATPase (Treherne and Schofield, 1981), which was recently shown to be involved in modulating the temperature leading to loss of neural function (Andersen et al., 2022; Robertson and Moyes, 2022). Interestingly, some butterfly species (e.g. milkweed butterflies, Danainae) possess Na+/K+-ATPases with an unusual resistance to inhibitors (e.g. ouabain, see Petschenka et al., 2013), which are often used to induce spreading depolarization experimentally (Andersen et al., 2022; Andrew et al., 2022). Similarly, butterflies generally possess an ‘unconventional’ ion homeostasis consisting of a relatively high concentrations of K+ and low concentrations of Na+ in the hemolymph (Andersen et al., 2017), of which the former is also a common way to induce spreading depolarization (Schofield, 1990; Smith et al., 2006). Future research on the curious ionoregulatory homeostasis of butterflies may provide key insights into to the mechanism underlying spreading depolarization, and their establishment as a new model of spreading depolarization represents a first step toward this goal.
Correlations between cold-induced loss of neural function and biogeography
Temperatures leading to cold-induced spreading depolarization were species specific (F11,51=2.0, P=0.044) and ranged from 2.96±0.15°C in Papilio memnon to 5.32±0.73°C in Parides eurimedes (Fig. 2A). Despite these differences, the phylogenetic signal was low (K=0.156), indicating that variation was spread randomly across species. This is similar to what has been reported for ant species previously (Willot et al., 2023), but opposite to the situation observed in drosophilid fruit flies, where the signal is moderate to strong (Kellermann et al., 2012; MacMillan et al., 2015). In terms of the variation in butterfly cold tolerance, previous studies have found conflicting evidence: Goller and Esch (1990) failed to find variation in the lower thermal limit (chill coma onset temperature) of a range of wild-caught lepidopterans collected at the same site, whereas Andersen et al. (2017) found clear differences in behavioural chill coma temperatures between domesticated (Bombyx mori), neotropical (Heliconius cydno) and continent-spanning (Manduca sexta) species. In our common garden experiment the differences were clear, yet small compared with those from Drosophila, for example (Andersen et al., 2018; Kellermann et al., 2012). Thus, while there is interspecific variation in the lower thermal limit to CNS function, species relatedness appears to play a relatively minor role in shaping the observed differences.
We also found an effect of sex on the temperature leading to spreading depolarization (F1,51=7.8, P=0.008) such that females generally experienced the event at lower temperatures than males. Sex-specific differences in the critical thermal minimum have been noted previously (e.g. mosquitos; Jass et al., 2019) but will not be discussed here.
To test whether temperatures leading to cold-induced spreading depolarization correlate with relevant biogeography we constructed a phylogeny of the species investigated to correct for relatedness. The phylogenetic signals were generally low (K<0.2, see Fig. 2) but were nevertheless accounted for in subsequent correlations. Doing so, we found significant correlations between the spreading depolarization temperature and (1) the mean distribution latitude (PPIC=0.037, R2PIC=0.367; Fig. 2B), and (2) the lowest temperature of the coldest month (PPIC=0.029, R2PIC=0.394; Fig. 2E) when tested in isolation. No relationship was found with the mean distribution altitude (Fig. 2C), the annual mean temperature (Fig. 2D) or the mean temperature of the coldest quarter (Fig. 2F). Interestingly, the best predictive model (based on Akaike information criterion (AIC) scores did not contain latitude, and instead consisted of the lowest temperature of the coldest month and the mean distribution altitude (P<0.001 for both). Strong relationships between cold tolerance measures and insect biogeography have been reported for several insect groups including, but not limited to, ants (Willot et al., 2023), fruit flies (Andersen et al., 2015b; Kellermann et al., 2012; Kimura, 2004) and seed bugs (Käfer et al., 2020), and meta-analyses have confirmed this trend on a global scale (Addo-Bediako et al., 2000; Sunday et al., 2019). Similarly, altitude has been demonstrated as a predictor for thermal tolerance in butterflies (Karl et al., 2008; Montejo-Kovacevich et al., 2020). Nonetheless, this type of analysis does not account for microclimate variability and selection (Duffy et al., 2015; Rebaudo et al., 2016), nor does it account for the ability of most adult lepidopterans to behaviourally thermoregulate (Heinrich, 2013), their dependence on appropriate host species (Quinn et al., 1998; Slove and Janz, 2011), ontogeny (Bowler and Terblanche, 2008), overwintering life stage (Bale, 2002) or phenotypic plasticity (Weaving et al., 2022). Furthermore, numerous butterfly species are known for their seasonal migration (Chowdhury et al., 2021), adding further complexity to modelling seasonal range limits. Finally, it should also be noted that the temperatures leading to the cold-induced spreading depolarization are 11.5±1.5°C (mean±s.d.; range=9.1–13.3°C) below the average lowest temperature of the coldest month, meaning that the butterflies are unlikely to experience them in nature. By extension cold-induced loss of central nervous function is unlikely to set the limits to their poleward or altitudinal distribution; instead, this may be limited by other abiotic or biotic conditions that prevail outside their native tropical habitats (e.g. host availability). Nonetheless, low temperatures may still limit distribution via other mechanisms (e.g. degree-days; see Keefe, 2023). Thus, to fully appreciate the factors determining butterfly distribution and insect distribution in general, more complex models encompassing a wider range of species, species from varying climates, as well as a suite of abiotic and biotic factors, are needed. Lastly, gaining an integrative understanding of mechanisms underlying other critical or subcritical limits to butterfly fitness and performance is likely to provide valuable predictive power in a world with a changing climate (Harvey et al., 2023; Parmesan et al., 1999).
Acknowledgements
We would like to extend our utmost gratitude to Ed Bruggink who took care of the butterflies during their development. Butterfly images are used with permission from LPS LLC Butterfly Import, which we greatly appreciate.
Footnotes
Author contributions
Conceptualization: M.K.A., H.A.M.; Methodology: M.K.A., Q.W.; Validation: M.K.A.; Formal analysis: M.K.A., Q.W.; Investigation: M.K.A.; Data curation: M.K.A.; Writing - original draft: M.K.A.; Writing - review & editing: M.K.A., Q.W., H.A.M.; Visualization: M.K.A.; Supervision: M.K.A., H.A.M.; Project administration: M.K.A., H.A.M.; Funding acquisition: M.K.A., H.A.M.
Funding
This research was funded by a Carlsberg Foundation Internationalization Fellowship (CF18-0940 and CF19-0472) to M.K.A., a European Union grant (Marie Skłodowska-Curie Fellowship no. 101029380) to Q.W., and a Discovery Grant from the Natural Sciences and Engineering Research Council of Canada (RGPIN-2018-05322) to H.A.M. Equipment used in this study was purchased with support from the Canadian Foundation for Innovation.
Data availability
All relevant data can be found within the article and its supplementary information.
References
Competing interests
The authors declare no competing or financial interests.