Metabolic cost of freeze–thaw and source of CO₂ production in the freeze-tolerant cricket Gryllus veletis

Insects overwintering in temperate regions may encounter temperatures low enough to freeze their body fluids. Although freezing is typically harmful to biological systems, some insects are freeze tolerant and survive internal ice formation (Denlinger and Lee, 2010). The processes of cooling, freezing and thawing each impose challenges (Toxopeus and Sinclair, 2018). Cooling disrupts ion and water balance (MacMillan et al., 2012b), reduces metabolic rate (Marshall and Sinclair, 2012) and can damage membranes (Drobnis et al., 1993; Hazel, 1995). During freezing, growing (presumably extracellular) ice crystals exclude solutes, increasing hemolymph osmotic pressure and dehydrating cells (Toxopeus and Sinclair, 2018). Ice can also structurally damage tissues (Des Marteaux et al., 2018; Storey and Storey, 1988) or promote the accumulation of harmful metabolites (Storey and Storey, 1985b). While thawing, an imbalance between oxygen supply and demand can cause oxidative damage (Doelling et al., 2014). In response to these stressors, freeze-tolerant insects may control the quality and distribution of ice crystals (Zachariassen and Kristiansen, 2000), suppress metabolism (Toxopeus et al., 2019a), clear harmful metabolites and minimize oxidative damage (Lopez-Martinez et al., 2008), stabilize macromolecules (Crowe et al., 1987), and facilitate anaerobic metabolism (Storey and Storey, 1985b). The capacity of insects to recover physiological processes, repair damage and manage metabolic costs while thawing may also contribute to freeze tolerance (Toxopeus and Sinclair, 2018). The metabolic costs associated with all these processes have not been quantified in freeze-tolerant insects.

Freezing appears to be a metabolically passive process in insects. When freeze-tolerant frogs and oligochaetes freeze, they initiate active metabolic and transcriptional responses and mobilize glucose as a cryoprotectant (Costanzo et al., 1993; Pedersen and Holmstrup, 2003; Storey, 2004; Storey and Storey, 1985a). In frogs, these responses appear to have a metabolic signature both before and during freezing (Sinclair et al., 2013). By contrast, insects appear to mobilize few (Marshall and Sinclair, 2011; Teets et al., 2011) or no (Williams and Lee, 2011) additional cryoprotectants in response to freezing, although we are not aware of fine-scale explorations of molecular responses during the freezing processes in insects. However, freeze-tolerant caterpillars of both Pringleophaga marioni (Sinclair et al., 2004) and Pyrrharctia isabella (Marshall and Sinclair, 2011; B.J.S., unpublished observations) release a burst of CO₂ when they freeze. There are at least three possible explanations for this CO₂ release. First, the response to freezing might be energetically expensive, requiring aerobic metabolism and consequent CO₂ production by mitochondria. This CO₂ release would not necessarily need to be a physiologically coordinated response in insects; for example, an open tracheal system allows aerobic metabolism to persist, and even transiently increase, past organismal death (e.g. above the critical thermal maximum; Heinrich et al., 2017). Second, ice expansion in the hemocoel might collapse the tracheae and expel tracheal CO₂ (Sinclair et al., 2004). However, synchrotron X-ray visualization of fly larvae suggests that ice formation displaces the tracheae, but does not fully collapse them (Sinclair et al., 2009). Finally, concentration (and presumably pH) changes associated with freezing might reduce the CO₂-buffering capacity of the hemolymph, driving CO₂ out of solution. Identifying whether the origin of this CO₂ burst is metabolic, mechanical or chemical will help clarify whether freezing is accompanied by metabolic activity in insects.

Repair and recovery after freeze–thaw, rather than avoidance of damage, may underlie insect freeze tolerance (Des Marteaux et al., 2018) and there are several reasons why these processes might be costly. First, if freezing disrupts ion and water balance, ion-motive ATPases will likely re-establish ionic and osmotic gradients during thawing and recovery (Boardman et al., 2011; Kristiansen and Zachariassen, 2001). Second, if anaerobic metabolites (e.g. lactate) accumulate while insects are frozen, they may need to use ATP-demanding processes during thawing to clear these products (Storey and Storey, 1985b). Finally, if insects are damaged by freezing (Izumi et al., 2005; Marshall and Sinclair, 2011; Des Marteaux et al., 2018), recovery may require recognition and repair of damage via energy-demanding protein synthesis, cellular proliferation and apoptosis (Toxopeus and Sinclair, 2018). Nevertheless, there is equivocal evidence of a decrease in energy stores after freezing (Marshall and Sinclair, 2011; Teets et al., 2011) and of elevated post-thaw metabolic rates in freeze-tolerant insects (Table 1). Among-species differences in post-thaw metabolic demands could underlie the discrepancies in these results, but so might differences in the respirometry techniques used in studies spanning four decades (Table 1). Although flow-through respirometry provides temporal resolution that closed-system approaches cannot, flow-through studies measure CO₂ production, rather than O₂ consumption. Although the former is more sensitive, it is affected by fuel use and CO₂ buffering in the hemolymph (Lighton, 2008; Sinclair et al., 2011). Thus, when determining any metabolic costs associated with recovery from freezing, the vagaries of these measurement systems must be considered.

The spring field cricket, Gryllus veletis (Alexander and Bigelow, 1960) (Orthoptera: Gryllidae), is emerging as a model system for studying insect freeze tolerance. Gryllus veletis is native to North America and overwinters in soil as a late-instar nymph (Alexander and Bigelow, 1960). These nymphs develop freeze tolerance when acclimated to declining temperature and photoperiod (Toxopeus et al., 2019a). Freeze-tolerant crickets have suppressed metabolic rates (Toxopeus et al., 2019a) and accumulate the cryoprotectants proline, myo-inositol and trehalose (Toxopeus et al., 2019c). Freeze tolerance is accompanied by differential expression of genes associated with cellular and macromolecular protection and metabolic and cell cycle suppression (Toxopeus et al., 2019b). Acclimated G. veletis freeze at ca. −6°C, and will survive for 1.5 h at ca. −12°C, or 1 week at −8°C (Toxopeus et al., 2019a). Once ice has formed, crickets remain frozen until the ambient temperature exceeds the melting point (ca. −1°C, based on reported hemolymph osmolality; Toxopeus et al., 2019a), at which point they thaw.

Here, we used G. veletis to elucidate the origin of the CO₂ released when insects freeze, and to explore the post-thaw metabolic costs of freezing. First, we hypothesized that the burst of CO₂ released during freezing results from a change in the CO₂-buffering capacity of hemolymph, rather than a metabolic response to ice formation. We disrupted aerobic metabolism with cyanide and the nervous system with thiacloprid to test our prediction that the CO₂ burst would persist in pharmacologically killed crickets. Second, if freezing disrupts ion and water balance and damages tissues, we hypothesized that recovery from freezing has a significant metabolic cost. We predicted that metabolic rate would increase post-thaw in frozen crickets compared with the metabolic rate of chilled, but unfrozen crickets.

We used G. veletis nymphs from a colony reared at the University of Western Ontario that originated from individuals collected in 2010 from a wild population in Lethbridge, AB, Canada (Coello Alvarado et al., 2015). We reared crickets from hatching at 22°C under a long day photoperiod (14 h:10 h light:dark) and provided rabbit food and water ad libitum. To induce freeze tolerance, we acclimated fifth-instar nymphs to a declining photoperiod and temperature regime in a Sanyo MIR 154 incubator (Sanyo Scientific, Bensenville, IL, USA) over 6 weeks (Toxopeus et al., 2019a). The photoperiod began at 11.5 h:12.5 h light:dark and day length decreased by 36 min each week to 7.9 h:16.1 h light:dark. Daily maximum and minimum temperatures began at +16°C and +12°C and decreased systematically over 6 weeks to +1°C and 0°C (Toxopeus et al., 2019a).

We measured the rate of CO₂ production (⁠⁠) of G. veletis with a Sable Systems International (SSI, Las Vegas, NV, USA) flow-through respirometry system (Fig. S1; Sinclair et al., 2013). We weighed crickets (MX5 microbalance, Mettler-Toledo, Columbus, OH, USA) before each recording. We passed air that was normoxic, CO₂-free and dry through a 10 cm³ glass chamber containing an individual insect at 80 ml min⁻¹ [controlled by mass flow valves (Sierra Instruments, Monterey, CA, USA) and a mass flow controller (MFC2,; SSI)]. A Li7000 CO₂ analyzer (LiCor, Lincoln, NE, USA) measured CO₂ released by crickets into excurrent air. We included baseline measurements of dry, CO₂-free air before and after each measurement to correct for instrument drift. We monitored the activity of crickets using an AD-1 infrared activity detector (SSI). A UI2 interface and Expedata software (v.1.8.5, SSI) acquired and recorded data every second.

We used flow-through respirometry to measure  as a proxy for metabolic rate during freezing and thawing of G. veletis. Briefly, we cooled crickets by placing individual insects into a respirometry chamber in a plastic bag and submerging it into a Proline RP855 circulating bath (Lauda-Brinkmann LP, Delran, NJ, USA) that contained a 50% methanol solution. We cooled the crickets from +15°C to −9.5°C at 0.25°C min⁻¹, and held them at −9.5°C for 1.5–2 h, before re-warming to +15°C at 0.25°C min⁻¹. We recorded body temperature at 1 Hz with type-T (copper–constantan) thermocouples placed along each cricket's abdomen, and connected to a TC2000 interface (SSI) and recorded via a UI2 analog–digital interface (SSI). We identified freezing events from the heat released by the latent heat of crystallization (Sinclair et al., 2015). We identified thawing as a departure of the body temperature from the temperature of a reference thermocouple on top of the chamber during rewarming (Sinclair et al., 2013).

To determine whether the CO₂ emitted during freezing and thawing was the result of active metabolic responses, we measured  from pharmacologically killed crickets using flow-through respirometry. We used thiacloprid (a neonicotinoid that binds to acetylcholine receptors in the central nervous system; Matsuda et al., 2020) to determine whether the burst of CO₂ during freezing and thawing requires coordinated nervous system responses. We used cyanide (which inhibits cytochrome c oxidase in the mitochondrial electron transport chain) to determine whether the burst of CO₂ requires aerobic respiration. We injected crickets beneath the arthrodial membrane near the anterior edge of the pronotum with either 4 µl of 0.6 mmol l⁻¹ thiacloprid (Sigma-Aldrich, St Louis, MO, USA) solution or 4 µl of 1 mmol l⁻¹ cyanide (carbonyl cyanide 3-chlorophenylhydrazone, CCCP, Sigma-Aldrich) solution using a 5 µl Hamilton syringe (32-gauge needle; Hamilton Company, Reno, NV, USA). No crickets responded to touch 30 min after injection. We placed these crickets in respirometry chambers 30 min after injection and continuously measured  during freezing and thawing, as described above.

We measured total hemolymph CO₂ (TCO₂) according to Lee et al. (2018). Briefly, we collected 2.5–5 μl of hemolymph from acclimated crickets with a 5 µl 32-gauge Hamilton syringe from the hemocoel beneath the pronotal membrane, and injected this hemolymph directly into a sparging column (identical to that described by Lee et al., 2018) containing 1 ml of 0.01 mol l⁻¹ HCl and 0.2 ml of antifoam 204 (Sigma-Aldrich). We passed dry, CO₂-free air through the sparging column at 10 ml min⁻¹ from the column into a LiCor Li7000 infrared CO₂ gas analyzer. Bicarbonate and dissolved CO₂ in the hemolymph reacted with the acid to release gaseous CO₂ into excurrent air, which was measured in the CO₂ analyzer (Fig. S1). We calculated rates of CO₂ production using Expedata and integrated beneath each peak of CO₂ release to obtain the total volume of CO₂ produced and divided this by the molar volume constant (22.4141 l mol⁻¹) to yield the moles of CO₂ in the sample. We corrected measured hemolymph CO₂ to a calibration curve of sodium bicarbonate standards (2.5, 5, 10, 15, 20, 25 and 30 mmol l⁻¹; Sigma-Aldrich) injected into the sparging column before and after each set of hemolymph samples.

An elevated post-thaw metabolic rate could reflect the cost of recovery from chilling (MacMillan et al., 2012b) and not freezing per se. To account for this possibility, we cooled freeze-tolerant crickets to a temperature where approximately equal numbers of individuals froze or remained unfrozen and measured the metabolic rate of these individuals during recovery. We cooled crickets in a bath from +15°C to −8°C at 0.25°C min⁻¹ and held them at −8°C for 1.5 h. We monitored temperature with a 36 AWG type-T thermocouple placed along the abdomen of crickets within each tube. We connected the thermocouple to a Picotech TC-08 thermocouple interface and recorded temperature using PicoLog software (Pico Technology, Cambridge, UK). After 1.5 h at −8°C, we removed crickets from the block and placed them directly into individual 180 ml plastic cups (Polar Plastics, Summit Food Distributors Inc., London, ON, Canada) in a Sanyo MIR 154 incubator at +15°C, and provided them with water. We placed these crickets in respirometry chambers at 24, 48 and 72 h post-freeze or post-chill to measure CO₂ production and oxygen consumption during long-term recovery at +15°C. We compared the metabolic rate of a subset of acclimated male and female nymphs 72 h post-chill or after acclimation (controls); we did not observe any sex differences in metabolic rate (t₁₉=0.349, P=0.731), so we pooled male and female crickets for subsequent analyses.

We used stop-flow respirometry to measure O₂ consumption and CO₂ production in freeze- tolerant G. veletis during recovery from chilling and freezing (Fig. S1). We used a multiplexer (RM8, SSI) to sequentially flush three chambers containing crickets with air that was normoxic, CO₂-free and dry for 5 min at 200 ml min⁻¹, then sealed each chamber for 45 min before being passing this air through a LiCor Li7000 CO₂ analyzer and Oxzilla O₂ analyzer (SSI). We repeated this procedure 3 times over 4 h for each cricket. We measured activity using an AD-1 infrared activity detector (SSI). No cricket spent more than 5 min moving when the chamber was sealed (<10% of the 45 min measurement; Fig. S2). We included baseline measurements before and after each measurement to control for instrument drift, and recorded data as described above.

For continuous respirometry, we calculated the rate of CO₂ production (⁠⁠) by multiplying the fractional CO₂ concentration by the instantaneous flow rate using Expedata (Lighton, 2008). We estimated the critical thermal minimum (CT_min) as the temperature at which precipitously declined and spiracular control and activity ceased during chilling (Fig. S3A; cf. MacMillan et al., 2012a; Sinclair et al., 2004). We determined the volume of bursts of CO₂ at the onset of freezing (i.e. the supercooling point, SCP) by integrating relative to a 30 s baseline of  after each burst. We averaged  during periods of constant temperature: 30 min at +15°C prior to the onset of chilling; the last 30 min at −9.5°C; the first 30 min after rewarming to +15°C; and the last 30 min of each recording (18 h after rewarming to +15°C). We assessed the thermal sensitivity of  during periods of chilling and rewarming by regressing log₁₀ against temperature, and recorded 10^10×slope as the temperature coefficient (Q₁₀; Lake et al., 2013). We estimated the re-establishment of spiracular control (RESC) as the resumption of regular, sharp inflections in the  trace after rewarming to +15°C (Fig. S3B). We determined the volume of the CO₂ ‘overshoot’ after rewarming to +15°C by integrating relative to the lowest 30 min of  after rewarming to +15°C (cf. MacMillan et al., 2012b).

We analyzed data in R (v.3.6.1; https://www.R-project.org/), and performed data exploration and model validation according to Zuur et al. (2010). We checked for evidence of movement by crickets at 24 and 48 h after freezing and chilling. Crickets that were unresponsive to touch after 48 h were excluded from the analysis. The time required for RESC and the volume of the CO₂ ‘overshoot’ after rewarming were square-root transformed to meet the assumptions of linear models. We compared respirometry parameters from our flow-through experiments using general linear models with freezing status (frozen or unfrozen) as a factor. We included body mass as a covariate when analyzing  (Table 2). We compared the volume of CO₂ released at the onset of freezing between crickets that survived freezing, those that died from freezing, and those that were pharmacologically killed, with an analysis of variance. To test for differences in metabolic rate (and its proxies  and ⁠), RQ and activity as measured with stop-flow respirometry between frozen and unfrozen crickets during recovery, we used linear mixed-effect models with body mass as a covariate, time after rewarming and freezing status as fixed factors, and cricket as a random factor (Table 3).

During cooling from +15°C,  declined until the activity of crickets ceased at the CT_min (Fig. 1; Fig. S3), when  dropped precipitously to a low, stable baseline. There was no significant difference in the thermal sensitivity of  (i.e. Q₁₀) during chilling between crickets that froze and those that remained unfrozen (Table 2). In all living crickets that froze, we identified two superimposed bursts of CO₂ concurrent with the exotherm at the onset of freezing (Figs 1B and 2B): a small, transient burst of CO₂ and a large, prolonged burst of CO₂. In unfrozen crickets,  remained low and stable below the CT_min with no bursts of CO₂ (Fig. 1A). Because of this burst of CO₂ at the onset of freezing,  at −9.5°C was higher in frozen than in unfrozen crickets (Table 2, Fig. 3). Upon rewarming,  increased with temperature (Fig. 1), and the thermal sensitivity (Q₁₀) was not significantly different between frozen and unfrozen crickets (Fig. 1, Table 2). Although  did not differ significantly between frozen and unfrozen crickets after 0.5 h at +15°C (Fig. 3), the volume of CO₂ ‘overshoot’ after rewarming was significantly larger in crickets that froze (Table 2). Spiracular control was apparent as regular deflections of  (Fig. 1) that ceased below the CT_min, and resumed following rewarming to 15°C (i.e. RESC; Fig. S3). The time required for RESC after rewarming was greater in crickets that froze than in those that were chilled, but unfrozen (Table 2). After 18 h at +15°C,  did not differ significantly between frozen and unfrozen crickets (Table 2, Fig. 3).

At the onset of freezing, a large burst of CO₂ was released in crickets killed by cyanide or thiacloprid injection (Fig. 2B; Fig. S4). This burst of CO₂ was absent in chilled but unfrozen crickets, regardless of whether or not they were pharmacologically killed (Figs 1A and 2A). The volume of CO₂ released upon freezing was greater in crickets killed by thiacloprid than by living crickets or those killed by cyanide (F_2,10=6.29; P=0.017; Fig. S5). The hemolymph TCO₂ of unfrozen freeze-tolerant crickets was 9.5±2.2 mmol l⁻¹ (n=5).

After 1 day of recovery at +15°C, the mean metabolic rate of crickets that froze was approximately 40% higher than that of unfrozen crickets, and did not change significantly over time from 24 to 72 h after rewarming (Table 3, Fig. 4). Using either  or  as a proxy for metabolic rate did not change our interpretation of these effects (Table 3; Fig. S6A,B). RQ did not differ between frozen and unfrozen crickets (Table 3; Fig. S6C,D). Crickets recovering from freezing spent less time moving than their unfrozen and control (acclimated but unchilled) counterparts during stop-flow respirometry recordings (Fig. S2; Table 3).

We examined whether freezing and thawing induce active metabolic responses and whether recovery from freezing is metabolically costly in a freeze-tolerant insect. There was a large burst of CO₂ at the onset of freezing, and this burst was also released by individuals killed with cyanide or thiacloprid, implying that it is not coordinated by the nervous system or the result of mitochondrial respiration. We hypothesized that recovery from freezing is metabolically costly, and we observed a metabolic ‘overshoot’ immediately after thawing in crickets that were frozen compared with those that were chilled. The metabolic cost of freezing was apparent after 24 h, and persisted for several days.

CO₂ release by crickets declined during cooling, until the magnitude and variability of  precipitously decreased at ca. −2°C (Fig. 1, Table 2). This decrease in variability coincided with the cessation of activity (Fig. S3), and was most likely due to the loss of coordinated spiracle control below the CT_min (MacMillan et al., 2012a; Robertson et al., 2017). Estimates of CT_min vary, sometimes by several degrees, with acclimation (Everatt et al., 2013) and methodology (Chown et al., 2009; Terblanche et al., 2007), and our estimates (Table 1) were ca. 1°C lower than visual measures of the CT_min of cold-acclimated, but freeze-intolerant, G. veletis nymphs (Coello Alvarado et al., 2015).

Although patterns of CO₂ release by insects near the critical thermal maximum are qualitatively similar between species (Boardman and Terblanche, 2015; Heinrich et al., 2017; Klok et al., 2004; Lighton and Turner, 2004; Mölich et al., 2012; Stevens et al., 2010), this does not seem to be the case near the CT_min. Our observation that the continuous gas exchange of G. veletis becomes abruptly low and stable at the CT_min was consistent with patterns in some freeze-intolerant and freeze-tolerant insects (Boardman et al., 2016; MacMillan et al., 2012a; Sinclair et al., 2004). By contrast, other insects appear to release a burst of CO₂ at the CT_min followed by a low stable  (Oyen and Dillon, 2018; Robertson et al., 2017; Stevens et al., 2010). In locusts, this final burst of CO₂ is preceded by a separate pulse associated with neuronal hyperactivity (Robertson et al., 2017). The diversity of CO₂ release patterns near the CT_min could be due to different mechanisms of neuromuscular failure, which can be driven by localized muscle depolarization (Andersen et al., 2017; MacMillan et al., 2014; Overgaard and MacMillan, 2017) or systemic loss of central nervous system function (Robertson et al., 2017; Rodgers et al., 2010). Alternatively, relaxation of spiracle muscles can lead to either spiracle opening or closing, depending on the species (Chapman, 1998), and could produce discrepancies in the pattern of CO₂ release near the CT_min. We might expect that cold-hardy, acclimated, G. veletis will show less response than less cold-hardy, unacclimated, insects as used in other studies (Oyen and Dillon, 2018; Robertson et al., 2017; Stevens et al., 2010). Broader comparative investigations of the physiological basis of CO₂ release near the CT_min are needed to explain this diversity of patterns in insects.

We observed two superimposed bursts of CO₂ at the onset of freezing in living crickets (Fig. 1B): a small initial spike of brief duration, and a larger and longer burst. These bursts were coincident with the SCP in our crickets, unlike the release that occurs below the CT_min, but before freezing, in beetles (Stevens et al., 2010), and unlike the preparatory response to ice formation in wood frogs, where CO₂ production increases with cooling just before freezing (Sinclair et al., 2013). Although an initial small spike was not present in caterpillars during freezing (Sinclair et al., 2004), anatomical differences, including the presence of air sacs in gryllid crickets (Greenlee et al., 2013; Xu et al., 2016), may explain this discrepancy. We hypothesize that this small burst arises from the displacement of the tracheal system and air sacs by the formation of ice in the hemocoel (Sinclair et al., 2009).

We tested whether the larger burst was metabolic or chemical in origin. Crickets killed with thiacloprid release a large burst of CO₂ at freezing (Fig. S4), suggesting that this burst is independent of any coordinated nervous system response or active ventilation. Further, crickets killed with cyanide and frozen still produced the large burst of CO₂ (Fig. 2B), excluding the possibility that it originated from aerobic respiration by mitochondria. We measured ca. 6–16 µl of CO₂ release at freezing from living crickets (Fig. S5), a volume that could be accounted for by CO₂ buffered in the hemolymph of crickets. Assuming that fifth-instar nymphs have a total hemolymph volume between that of fourth-instar (8 µl) and adult (80 µl) Gryllus crickets (Coello Alvarado et al., 2015; MacMillan et al., 2012b), our TCO₂ measures predict ca. 2–17 µl of CO₂ in hemolymph. Thus, we propose that the freezing process is passive, and that CO₂ buffered in the hemolymph is released at the onset of freezing. We expect that the liquid fraction of hemolymph would decrease rapidly during freezing, reducing the volume of hemolymph available to buffer CO₂. Increasing solute concentrations, including [H⁺], would likely acidify hemolymph and further liberate CO₂. While insect hemolymph acid–base status is usually well regulated, acid–base homeostasis can be disturbed by temperature and activity, among other factors (Harrison, 2001). For example, under anoxia, the insect hemolymph is acidified (Ravn et al., 2019), mostly likely by the accumulation of lactate (Woods and Lane, 2016), which liberates stored CO₂ as a burst (Lighton and Schilman, 2007) akin to what we observed in crickets at freezing (Fig. 1B). Severe acidification of hemolymph is sufficient to damage tissues (Ravn et al., 2019) and may be an overlooked source of freeze injury.

We found that the CO₂ release of crickets was low but detectable for at least 2 h while frozen (Fig. 1B). Because O₂ and CO₂ diffusion through ice is negligible (Hemmingsen, 1959), we interpret this CO₂ release as evidence that the tracheal system of partially frozen insects exchanges gas between unfrozen tissues and the atmosphere (Sinclair et al., 2004). However, frozen insects appear to shift to anaerobic metabolism (Doelling et al., 2014; Storey and Storey, 1985b), implying that oxygen is limited (or not consumed) while frozen. We suggest that measures of internal oxygen over a freeze–thaw cycle could clarify the extent to which ice formation makes tissues anoxic in partially frozen insects, although such measurements would be technically challenging because of the physical disruption caused by ice formation. Freeze-tolerant insects can tolerate hypoxia well, and some survive several days in anoxic atmospheres (Storey and Storey, 1990). This anoxia tolerance can be explained, in part, by the fact that insects that acquire freeze tolerance may also suppress metabolic rate in response to seasonal acclimation (Storey, 2005), or buttress antioxidant defense (Joanisse and Storey, 1996), as is the case in G. veletis nymphs (Toxopeus et al., 2019a; Toxopeus et al., 2019b). Nevertheless, some capacity to deliver oxygen to unfrozen tissues while partially frozen, as implied by CO₂ release in frozen crickets (Figs 1B and 3), could mitigate both the accumulation of harmful anaerobic end products and oxidative damage upon re-oxygenation during thaw (Toxopeus and Sinclair, 2018). We did not measure the O₂ consumption of frozen crickets, and therefore cannot exclude the possibility that the gradual CO₂ release we observed was from ongoing liberation of CO₂ buffered in hemolymph and not aerobic metabolism per se.

To our surprise, crickets recently killed with cyanide or thiacloprid continued to release CO₂ in a temperature-sensitive manner (Fig. 2). This CO₂ release slowly declined over 24 h after death (A.S., unpublished data), and could represent a gradual liberation of stored CO₂ from hemolymph. Crickets killed by thiacloprid released larger bursts of CO₂ at freezing than our living or cyanide-injected crickets (Fig. S5), which might indicate that metabolic rate (and CO₂ buffered in hemolymph) was elevated in these insects prior to death, as is the case in insects exposed to sublethal doses of some insecticides (Karise and Mänd, 2015). In addition, gradual CO₂ release after death could arise from the (still-living) microbiome of dead crickets, which is dominated by anaerobic bacteria (Domingo et al., 1998b; Ferguson et al., 2018) within a possibly fermentative hindgut (Domingo et al., 1998a).

Gryllus veletis crickets exhibited stereotyped low variability in CO₂ release during the initial recovery period from chilling and freezing (Fig. 1). RESC was delayed in crickets that froze relative to unfrozen (but chilled) crickets (Fig. 1, Table 2). Although RESC coincided with the recovery of movement in crickets that did not freeze (Fig. S3), this was not the case for crickets that froze (which required at least 24 to recover movement; Fig. S2). This loss of coordinated spiracular control at low temperatures, and its re-establishment after rewarming, is consistent with prior observations of both chilled (Lalouette et al., 2011; MacMillan et al., 2012b) and frozen insects (Sinclair et al., 2004). Qualitatively, the pattern of  during recovery from freezing in crickets bears a striking resemblance to insects recovering from anoxia, where spiracles are likely forced open by the acidification of hemolymph (Lighton and Schilman, 2007; Ravn et al., 2019; Woods and Lane, 2016). A similar acidification of hemolymph following freezing could explain these similarities.

The volume of CO₂ ‘overshoot’ during early recovery was greater in crickets that froze than in unfrozen crickets (Fig. 1, Table 2), suggesting that initial recovery from freezing is energetically costly. We hypothesize that the metabolic ‘overshoot’ could be due to the re-establishment of ion gradients through ATP-dependent ion transport after thawing (Boardman et al., 2011), akin to the metabolic overshoot that follows chill-coma recovery in the cricket Grylluspennsylvanicus (MacMillan et al., 2012b). However, we note that rates of absolute CO₂ release were similar between frozen and unfrozen crickets for the first half hour after return to +15°C (Fig. 3). Our estimates of metabolic rate and RQ derived from CO₂ release in recently thawed insects, and those from previous studies (Table 1), should be interpreted cautiously. If buffered CO₂ is liberated by freezing (Fig. 1; discussed above), then we would expect hemolymph to have a very high CO₂ buffering capacity immediately after thaw. Thus, CO₂ release would be underestimated until the hemolymph partial pressure of CO₂ (P_CO2) returns to normal. Although freezing and thawing are likely to compromise electrodes used to measure hemolymph P_CO2, we suggest that it should be possible to measure P_CO2 during the post-thaw recovery period to examine this process.

Metabolic rate was elevated in crickets for at least 3 days after thaw, indicating that the costs associated with freeze–thaw persist beyond the recovery of neuromuscular function (Fig. 4; Fig. S2), and presumably the restoration of ion balance. Because both  and  were elevated (Fig. S6), this magnitude of the metabolic increase (Fig. 4) cannot be solely attributed to changes in fuel use (e.g. switching between lipid and carbohydrate metabolism; Sinclair et al., 2011). We hypothesize that this prolonged cost is due to the recognition and repair of freeze injury. Although the mechanisms of freeze injury and their role in survival of freeze–thaw remain unclear (Toxopeus and Sinclair, 2018), at least some tissues are susceptible to cryoinjury (Des Marteaux et al., 2018; Izumi et al., 2005; Marshall and Sinclair, 2011; Neufeld and Leader, 1998; Rozsypal et al., 2019; Teets et al., 2011). In the fat body of larval Chymomyza costata, some forms of freeze-attributed cytoskeletal damage are repaired during recovery (Des Marteaux et al., 2018), a process that requires energy expenditure. Freezing also drives permanent lipid droplet coalescence in fat body cells (Des Marteaux et al., 2018; Lee et al., 1993; Salt, 1959), and this decrease in lipid surface area could cause long-term changes to energy use. Here, we only measured metabolic rate for the first 3 days after thaw, so it is unclear how long the cost of freeze–thaw persists in G. veletis. In addition to repair, other processes such as the processing of toxic anaerobic intermediates (Storey and Storey, 1985b), the neutralization of reactive oxygen species (Doelling et al., 2014) and the production of cryoprotectants (Marshall and Sinclair, 2011; Teets et al., 2011) could all contribute to increased metabolic rate during recovery. Our uncertainty about the nature of this metabolic cost speaks to the need for fine-scale examination of the damage, repair and recovery that follows freeze–thaw (e.g. Des Marteaux et al., 2018). To identify the origin of the cost of freeze–thaw, future studies should attempt to link variation in cryoinjury and the timing of repair to patterns of whole-organism metabolic rate.

Our observations support the hypothesis that freezing is a metabolically passive process in insects. Although crickets release a burst of CO₂ at the onset of freezing, we conclude that it is not metabolic in origin, but is most likely a chemical liberation of CO₂ buffered in the hemolymph. By contrast, recovery from freezing is metabolically costly and we hypothesize that this elevation in metabolic rate arises initially from the restoration of ion balance and, later, from the repair of freeze-attributed damage. Further investigations are needed to identify the mechanistic origin of these costs, and their contribution to variation in insect freeze tolerance.

We would like to thank James F. Staples, Christopher Guglielmo and two anonymous reviewers for insightful comments on an earlier version of this manuscript.