The impacts of repeated cold exposure on insects

As small ectotherms, most insects experience changes in body temperature that reflect their thermal environment. Temperatures in the polar or temperate zones fluctuate on multiple, interacting time scales (Fig. 1). Long-term fluctuations (which impart long-term selection pressure) range from climate change (Bale and Hayward, 2010) to multi-year cycles such as the El Niño Southern Oscillation (Mysterud et al., 2001). At a scale relevant to the lifetime of individual insects, seasonal changes in temperature are generally predictable (at least in the Northern hemisphere; Fig. 1A) (Sinclair et al., 2003a). These longer-term cycles are periodically interrupted by weather patterns on the scale of days to weeks (Fig. 1B) (Kingsolver, 2000). Finally, the most predictable of these time scales is the daily fluctuation between day and night, which can lead to daily thermal ranges of 20°C or more (Fig. 1C,D) (Kingsolver, 2000; Irwin and Lee, 2003; Marshall and Sinclair, 2010).

Temperature regimes with large thermal ranges inevitably cross physiological thresholds (Fig. 1C,D) (Sinclair, 2001b), meaning that many insects are exposed to cycles of cold stress that repeat on a daily basis, although in nature temperature and climate variation at longer time scales can also affect the frequency of cold exposure. Together, these interact with the particular habitat of an insect, leading to the development of a microclimate. For example, Sinclair predicted a New Zealand alpine cockroach would experience almost daily freeze–thaw cycles in an El Niño year, but relatively few freeze–thaw cycles in a year with greater snow cover (Sinclair, 2001a). Thus, the timing and frequency of cold exposure can be affected directly by microclimate, and indirectly by variation caused by climate shifts at larger temporal and spatial scales. Snow buffers thermal variability in microhabitats on the ground (Fig. 1A,B) (Irwin and Lee, 2003), but the extent of snow cover in many regions is declining with climate change (Déry and Brown, 2007), which could lead to increased freeze–thaw cycles in soil microhabitats. Similarly, aspect and exposure to solar radiation can lead to cycles in microhabitat temperatures of much greater magnitude than seen in air temperatures (e.g. Sinclair et al., 2003b). If these cycles cross physiological thresholds, such as those that induce stress responses (Le Bourg, 2011), chill-coma (MacMillan and Sinclair, 2011a), mortality (Sinclair et al., 2003a) or freezing (Teets et al., 2011), then they are expected to impact upon the fitness of insects. The majority of laboratory-based studies of insect stress responses focus on a single exposure, so a pertinent question is whether these repeated stresses have impacts that are distinct from those of a single stress exposure. Here, we will address this question, focusing on the impacts of repeated cold exposure (RCE) on insects.

Insects risk freezing when exposed to temperatures below the melting point of their body fluids. Globally, individuals of most insect species die of chilling injury before they freeze (Lee, 2010). However, freeze-avoiding insects depress the temperature at which ice forms (the supercooling point, SCP) and survive as long as they do not freeze, while freeze-tolerant insects can withstand the formation of internal ice (Lee, 2010). In both cold tolerance strategies, the production of low molecular weight cryoprotectants (e.g. glycerol), the control of ice crystal growth with antifreeze agents, and the manipulation of the SCP are important physiological components of cold hardiness. Insect cold tolerance is plastic on evolutionary (e.g. Strachan et al., 2011), inter-annual (e.g. Horwath and Duman, 1984) and seasonal scales (e.g. Pio and Baust, 1988). Seasonal cold hardening in temperate insects is well characterised: insects increase their cold hardiness from a summer chill-susceptible state by accumulating carbohydrate cryoprotectants, synthesising antifreezes, reordering lipid membranes, and either retaining (in freeze-tolerant insects) or removing (in freeze-avoiding insects) ice nucleators (Lee, 2010).

Insect cold tolerance can also vary on relatively short time scales. For example, there are daily cycles in the SCPs of freeze-avoiding Antarctic Collembola (Sinclair et al., 2003b), and chill-susceptible Drosophila melanogaster adults are more cold tolerant at the low points of daily temperature cycles (Kelty and Lee, 1999; Kelty, 2007). In the laboratory, many chill-susceptible insects show a rapid cold-hardening (RCH) response, whereby a short exposure to a mild temperature improves tolerance to a subsequent exposure to a more severe temperature (Lee et al., 1987; Lee, 2010). RCH can occur in minutes (Lee et al., 1987), does not appear to depend on de novo protein synthesis or transcription (Misener et al., 2001; Sinclair et al., 2007), and is variously associated with changes in apoptotic pathways (Yi et al., 2007), membrane composition (Overgaard et al., 2005) (but see MacMillan et al., 2009) and cryoprotectant concentrations (Chen et al., 1987; Overgaard et al., 2005) (but see MacMillan et al., 2009). Thus, we can draw two conclusions. First, insects can respond to thermal fluctuations in their environment almost as quickly as the environment itself changes. Second, following cold exposures, insects appear to be in an altered physiological state, which influences their response to subsequent cold exposures.

Recently, there has been a rise in interest in the role of ecophysiological models for predicting organismal responses to a changing environment (Buckley et al., 2010). This has precipitated a general recognition of the contrast between most laboratory studies (which examine a single stress exposure) and the repeated exposures experienced by insects in the field. Together, these have led to a surge of interest in the effects of repeated stress – particularly cold – exposure in insects. Here, we first examine the design and interpretation of RCE experiments in insects, before reviewing the general impacts of RCE on insect fitness and physiology. Finally, we identify patterns in insect responses to RCE and gaps in our current knowledge, and recommend some future directions and experimental designs.

In defining RCE experiments, we excluded studies focused on the effects of thermal variability that do not use potentially damaging temperatures (e.g. Paaijmans et al., 2010), because temperature regimes that vary without crossing a physiological threshold investigate different sorts of questions from RCE experiments, which focus on the effects of repeated physiological threshold crossing. As most insect species enter chill-coma (which is a clear physiological threshold) at or below 10°C (Goller and Esch, 1990), we chose crossing below this temperature to define a RCE. We also exclude fluctuating thermal regimes (FTRs) (e.g. Colinet et al., 2006), which are defined by a long period of cold exposure interrupted by short periods of warm exposure (e.g. 22 h cool:2 h warm) (Colinet et al., 2006). While we will touch on some major conclusions from FTR experiments, their focus on the repair mechanisms during the warm phase of the cycle is distinct from studies that focus on repeated exposure to a stressful temperature. We have identified more than 20 studies across five insect orders that use cold shocks at a cold temperature (≤10°C) expected to cause low temperature stress, and where the cold shocks are relatively short compared with the warmer temperature (≤50% of the experiment time; supplementary material Table S1). In the case of freeze-tolerant insects, it is important to distinguish between studies of repeated freeze–thaw (RFT) on its own (e.g. Brown et al., 2004; Marshall and Sinclair, 2011), and those that contrast RFT and RCE (e.g. Sinclair and Chown, 2005; Teets et al., 2011) (supplementary material Table S3).

Experiments investigating RCE are generally designed with reference to the ecology, tolerance and lifespan of the study organism. Thus, experiments vary in the intensity of cold shocks, their number and duration, and the length and temperature of the warm period between shocks (supplementary material Table S1). The most commonly used design involves a RCE group that is compared with controls kept at a relatively warm maintenance temperature (which may be an average temperature of the RCE group) which does not cause low temperature stress (‘RCE vs warm’; Fig. 2A) (e.g. Brown et al., 2004; Sinclair and Chown, 2005; Le Bourg et al., 2009). A second design contrasts a RCE group and both a warm control and a stressful prolonged low temperature (‘RCE vs cold’; Fig. 2B) (e.g. Yocum, 2001; Wang et al., 2006). This design is similar to FTR designs where the total time spent at the stressful low temperature is greater than in the RCE group. Finally, ‘RCE vs matched cold’ compares a RCE group with both a warm control and a stressful prolonged low temperature that is matched for total time spent by the RCE group at the stressful low temperature (Fig. 2C) (e.g. Marshall and Sinclair, 2010; Teets et al., 2011).

In RCE vs warm designs, it is not possible to distinguish between the effects of cold exposure in general or RCE in particular. While this is useful in some ecological contexts (e.g. stable subnivean vs fluctuating exposed microclimates) (Irwin and Lee, 2003), it is not a useful design for identifying RCE-specific responses. However, RCE vs warm designs do allow evaluation of the potential for recovery from cold stress (and in that way are similar to FTR experiments). Similarly, RCE vs cold designs mismatch the ‘amount’ of stress they deliver between experimental groups. In RCE vs cold designs, RCE groups always spend less total time at the stressful temperature than the group kept only at low temperatures, so detection of RCE effects is confounded by the reduced stress on that group. We believe that RCE vs matched cold designs are most appropriate for disentangling the impacts of RCE from other effects of cold exposure because they match the total amount of stress applied in the prolonged cold and RCE experimental groups.

The difficulty of choosing the appropriate intensity and duration of exposure is a limitation of the RCE vs matched cold design: the sum of the repeated exposures must fall within the range that the insect can survive. As a result, RCE vs matched cold designs may be forced to use RCE treatments that do not reflect the most extreme exposures possible. This constraint is particularly pertinent when working with species that accumulate lethal chilling injury with chronic cold exposures, as there is an interaction between the effects of time and intensity of cold exposure on recovery time from chill coma that may extend to effects on survival (see MacMillan and Sinclair, 2011b). For example, the cold exposures used by Marshall and Sinclair were based on the survival of Drosophila melanogaster after 10 h at –0.5°C (Marshall and Sinclair, 2010), whereas these flies can survive acute exposures down to –5°C (Nyamukondiwa et al., 2011). While the temperatures used are important, other variables to consider in a RCE study include the frequency, number and duration of cold shocks (supplementary material Table S1).

Despite the various time scales of temperature fluctuations (Fig. 1), the majority of studies focus on daily temperature cycling (supplementary material Table S1), although there have been a few studies showing that frequency of exposure is an important predictor of survival (e.g. Bale et al., 2001; Brown et al., 2004). Depending on the life history of the insect (particularly its generation time), it may be appropriate to incorporate cold cycles on other time scales (e.g. Marshall and Sinclair, 2011). Similarly, despite the majority of studies utilising an immediate transfer between temperatures (supplementary material Table S1), in the field, insects generally do not experience abrupt shifts of 10–15°C (Sinclair, 2001b), so an appropriate cooling rate must be chosen (Chown et al., 2009). Indeed, some of the negative effects of RCE and RFT may be due to rapid cooling, although with the limited number of studies it is difficult to separate the effect of cooling rate from the overall effect of RCE and RFT.

Another under-appreciated aspect of the design of RCE studies is the effect of ageing on the physiology of cold hardiness. This is particularly problematic in D. melanogaster where age decreases reproductive output (Marshall and Sinclair, 2011) and up-regulates immune function after cold exposure on the scale of days (Le Bourg et al., 2009). Ageing is less of a concern in univoltine insects, although the response to RCE in Eurosta solidaginis depends on the month of collection and may be linked to diapause status, although this may be complicated by prior cold exposures in field conditions (Pio and Baust, 1988). The problem of age effects has a simple solution – age matching of experimental groups (e.g. Zhang et al., 2011).

Although the cold tolerance and life history of the insect will be the primary determinant of the thermal regime chosen, the nonlinear effects of temperature on many biological processes need to be considered when designing RCE experiments. In some experiments (e.g. Petavy et al., 2001; Yocum, 2001), the stable cold stress group received the same mean temperature as the fluctuating insects. However, if stress and temperature have a non-linear relationship (as is seen in many biochemical and metabolic processes), Jensen’s inequality dictates that the mean of the response to the fluctuating temperature will be different from the mean of a response to a stable temperature as a result of simple mathematics (see Ruel and Ayres, 1999).

The effects of RCE on survival are relatively straightforward – when compared with any prolonged cold exposure, RCE almost always results in higher survival (supplementary material Table S2) [with the exception of the lepidopterans Aglais urticae and Inachis io, in which RCE-exposed caterpillars show similar or decreased survival, but greatly increased mass loss, compared with constant low temperature groups (Pullin and Bale, 1989)]. Increasing recovery time between repeated exposures increased survival to the end of the experiment in repeatedly frozen Hydromedion sparsutum (Bale et al., 2001), but other studies exploring the impact of recovery duration are lacking. Experiments with freeze-tolerant species that involve repeated freezing generate more mixed results (supplementary material Table S3). Two studies found decreased survival of RFT relative to a period of prolonged freezing in a hoverfly (Brown et al., 2004) and a caterpillar (Marshall and Sinclair, 2011). By contrast, there were no differences in mortality between caterpillars that received RCE or RFT (Sinclair and Chown, 2005), and midge larvae exposed to RFT had increased survival relative to those that received a matching, sustained freeze exposure, but decreased survival compared with RCE (Teets et al., 2011). So, while it seems clear that RCE benefits survival, the effects of RFT are much more mixed.

Although mortality prevents reproduction, and therefore has rather large fitness consequences, most impacts of the environment on insects are mediated through sub-lethal effects, reducing performance, reproductive output and, therefore, Darwinian fitness. In many insects, fitness-related parameters like growth, development and reproductive output can be measured with relative ease. The observed effects of and responses to RCE suggest that there should be sub-lethal fitness consequences of RCE for insects. For example, the energetic costs of repair of chilling or freezing injury, especially during winter when insects often do not feed (Bale and Hayward, 2010), could lead to decreased investment in reproduction. Similarly, the immediate physiological responses include pathways with demonstrated fitness costs; for example, the production of heat shock proteins with RCE (Silbermann and Tatar, 2000). RCE decreases development time in D. melanogaster (Petavy et al., 2001), which could indicate a fitness benefit. In D. melanogaster, although RCE increases survival relative to matched time in a single cold exposure, it decreases subsequent fitness through a male-biased sex ratio (Marshall and Sinclair, 2010). Thus, when only survival is measured, it is possible to conclude that RCE has a positive effect on fitness when sub-lethal effects could accumulate to a net depression of fitness. Disentangling the trade-offs between fitness components and how RCE impacts these trade-offs is a topic of clear importance for future research.

The studies reported here all focus on experiments conducted in the laboratory, where thermal regimes may easily be manipulated. In nature, variation in temperature depends greatly on snow cover, altitude, aspect, distance to ocean and other geographical parameters. There are also potential interactions between seasonal acclimation and the additional acclimation that appears to occur during RCE. To date, only Pio and Baust appear to have addressed this concern, showing that repeated cold exposure only increases cold hardiness in E. solidaginis during the months of December and January – early winter in temperate North America (Pio and Baust, 1988). To add to this complexity, snow cover in temperate environments may only be present during a portion of the winter, so soil freeze–thaw may occur more frequently in spring and autumn when snow is absent. These interactive effects remain poorly understood.

RCE appears to have fitness impacts on insects that are distinct from the impact of a single cold exposure. These differences could arise from either a cumulative effect of the physiological impact of repeated cold and warm cycles, or because repeated exposure initiates a set of responses that set the insect on a different physiological path to that followed after just a single exposure. Microarrays indicate that the transcriptomic responses of D. melanogaster to single and repeated exposures are divergent (Zhang et al., 2011) and these responses could well underlie the increasingly well-documented physiological responses that are unique to repeated cold.

Both the RCH response and acclimation responses lead to an expectation that RCE will improve cold tolerance in insects, and this appears to be broadly true (supplementary material Table S2). Low temperature tolerance in D. melanogaster is improved after RCE relative to controls kept at room temperature (Le Bourg, 2007). Similarly, repeated freezing lowered the supercooling points (Bale et al., 2001; Brown et al., 2004) and increased cryoprotectant concentrations (Marshall and Sinclair, 2011; Teets et al., 2011) of freeze-tolerant larvae. However, RCE in (chill-susceptible) Orchesella cincta did not decrease chill-coma recovery time relative to individuals held at a constant low temperature (van Dooremalen et al., 2011). It is possible that lowered SCPs and perceived strategy switches may result from artificial selection on individuals with lower SCPs, resulting in individuals with lower cold hardiness being removed from the experiment at the first exposure. This is likely to be an issue with any measurements of mortality, and measurements of cold hardiness should be made apart from the cold exposures that constitute the experimental treatments to avoid potential biasing by the death of less cold-hardy individuals.

The synthesis of additional cryoprotectants in response to RCE (Churchill and Storey, 1989; Teets et al., 2011) would be expected to deplete other energy stores, particularly in species that do not feed during winter; however, this does not consistently appear to be the case for cold-hardy insects. Repeated freezing did not change metabolic rate or energy reserves in Pyrrharctia isabella (Marshall and Sinclair, 2011). Similarly, there was no change in adenylate energy charge in the freeze-intolerant caterpillar Epiblema scudderiana after RCE (Churchill and Storey, 1989). However, although RFT decreased energy reserves in larvae of the Antarctic midge Belgica antarctica, RCE without freezing appeared to increase energy reserves (Teets et al., 2011). By contrast, in laboratory colonies of D. melanogaster, both glycogen stores (Marshall and Sinclair, 2010) and starvation resistance (Le Bourg, 2007) were decreased after RCE, even though food was not limiting during the warm parts of the cycles. It is possible that these conflicting results are due to the cycling of energetic reserves to other, unmeasured, components or that changes in energy reserves do not result directly from RCE.

The acquisition of energy reserves is a key component of the performance of the juvenile stages of many holometabolous insects, so observations of decreased feeding rates after RCE in caterpillars (Sinclair and Chown, 2005) could imply a mechanism that links physiological impacts to fitness costs. Cellular damage in Malpighian tubules of P. isabella larvae was greater following repeated freezing relative to sustained freezing or controls (Marshall and Sinclair, 2011), and repeatedly frozen B. antarctica had lower cell survival in their midgut relative to those frozen in a single period [although RCE without freezing improved cell survival relative to a single long chill (Teets et al., 2011)]. Thus, the full extent of the relationship between RCE, gut damage, feeding performance and subsequent fitness remains to be determined, and probably varies considerably among species.

Recently, several lines of evidence indicate that RCE may result in an enhanced immune response. RCE increased survival of infection by entomopathogenic fungi in both the chill-susceptible D. melanogaster (Le Bourg et al., 2009) and freeze-tolerant P. isabella (Marshall and Sinclair, 2011). Immune-related genes were upregulated in response to cold in D. melanogaster, but not differentially so in response to RCE (Zhang et al., 2011). These observations indicate that there may be complex interactions with pathogens over winter. It is not clear whether RCE in these cases is priming the immune response in preparation for potential attack, if RCE causes damage that directly elicits the immune response, for example to encapsulate damaged tissue or respond to gut flora that escape into the haemocoel (e.g. MacMillan and Sinclair, 2011a), or whether the cold exposure is increasing the pathogenicity of existing microbes. Given that many insects overwinter in environments in direct contact with soil, this topic appears to be of direct importance to studies not only of RCE effects but also of insect overwintering in general.

The driving force of RCE studies is the observation that natural temperature regimes vary. In response to this observation, researchers have hypothesised that insect physiological responses to cold depend on the frequency and timing of cold exposure. This hypothesis has been supported by three decades of study, and mechanisms of these differing responses are beginning to be investigated. From these beginnings, we have identified several areas that we believe are particularly important for understanding the effects of RCE.

Reproductive output and development time in insects are nearly as important to population growth as the survival of individuals. However, the effects of RCE on these parameters have barely been investigated, and translation of population effects from the laboratory to natural settings is complicated by the costs and benefits of phenotypic plasticity (Driessen et al., 2011). Given also that gene expression patterns change significantly between RCE and prolonged cold exposure (Zhang et al., 2011), it seems likely that extrapolation of laboratory studies of cold exposure to field settings will require a more thorough understanding of RCE in the context of the complexity, frequency and number of ecologically relevant cold exposures.

Of the cold exposure studies reported here, the majority deal with either a single species or comparisons between two species that are unlikely to directly interact [e.g. comparisons between E. solidaginis and E. scudderiana (Churchill and Storey, 1989)]. However, there are clearly differences between insect species in cold hardiness and strategy even within the same environment, and these differences seem likely to hold between larger taxonomic classes. Thus, a clear next direction in RCE research seems to be the integration of several trophic levels, including both insects and their pathogens and parasites. Similarly, studies on community assemblages could be a fruitful area of research as it might be important to understand individual species’ responses and how these play out in a community setting. These studies will be complex to plan and perform, but there are simple systems that could be employed. Microarthropod mesocosms have already been used (Sjursen et al., 2005), but other possibilities could include single pathogen–host interactions or further investigations of the insect assemblage associated with galls in Solidago (Judd, 1953).

We believe the greatest unknown is why the effects of RCE and RFT differ so substantially from single cold exposures. There must be a mechanistic reason for this difference, yet choice of study design and experimental animal can create difficulties when comparing impacts between studies, and understanding of the integration of fitness and physiological impacts of RCE remains relatively unexplored. Given that these differences can be traced from gene expression through to fitness effects, it should be possible to determine the connecting physiological links. Signalling pathways, enzymes and intermediates involved in carbohydrate metabolism, wounding responses and vitellogenesis are likely candidates for some of the observed effects.

We have several recommendations for the design of future studies of repeated cold exposure. First, experimental designs should reflect the research question. We recommend RCE vs matched cold design for isolating the effects of RCE, but other questions may require different sets of controls. Second, designs must acknowledge both the thermal biology and life history of the study species. Species that have multiple generations per year may need age controls in the design, but may also facilitate detection of cross-generational effects. Third, prior to embarking on a matched cold design we strongly suggest performing a pilot study to determine the maximum survivable single prolonged exposure for the study organism. This pilot study would allow the selection of appropriate temperature and time of exposures throughout the experiment. Finally, we encourage measurement of sub-lethal impacts. Few studies have linked how immediate physiological changes impact long-term fitness, yet this question underpins the purpose of laboratory studies of RCE.

 
• Chilling injury Injury

  accrued during cold exposure that is not caused by or associated with internal ice formation.

 
• Chill coma

  The reversible state of loss of neuromuscular function and paralysis at low temperatures.

 
• Chill susceptible

  Describes insect species that experience mortality unrelated to freezing at low temperatures (i.e. mortality is due to chilling injury).

 
• Cryoprotectant

  Compounds synthesised within the body that increase survival at low temperatures. These may be classified as either colligative or non-colligative. Colligative cryoprotectants are usually simple sugars (e.g. glucose) or polyols (e.g. glycerol). These compounds frequently accumulate to high concentrations during the overwintering phase of many insect species, are usually highly water soluble, and are non-toxic even at high concentrations. The action of colligative cryoprotectants is generally proportional to their concentration. Non-colligative cryoprotectants interact with ice and function either as antifreezes, by preventing the growth and propagation of ice crystals, or as recrystallisation inhibitors, preventing the expansion of ice crystals in the frozen haemolymph. Non-colligative cryoprotectants include antifreeze proteins in insects, antifreeze glycoproteins in fishes and recently described antifreeze glycolipids in some insects.

 
• FTR

  Fluctuating thermal regimes. An experimental temperature regime that typically involves long periods of time spent at a low temperature interrupted by short warming periods (often 22 h cold and 2 h warm).

 
• Freeze avoiding

  Insect species that cannot survive freezing, but are able to reduce their probability of freezing by depressing their supercooling point to low subzero temperatures through the accumulation of cryoprotectants and avoiding ice nucleation.

 
• Freeze tolerant

  Insect species that can survive internal ice formation.

 
• Ice nucleating agents

  Particles, molecules or proteins that initiate freezing of a supercooled solution. Ice crystals also act as nucleators of supercooled solutions.

 
• Microclimate

  The abiotic conditions that an individual insect experiences as a result of its microhabitat.

 
• RCH

  Rapid cold hardening. The physiological phenomenon in which an individual insect is able to increase its cold hardiness after a very short cold exposure (typically 2 h or less).

 
• RCE

  Repeated cold exposure. An experimental temperature regime in which a low temperature stress (typically below 10°C) is applied two or more times, and the total amount of time spent at the low temperature is equal to or less than the time spent at the warm temperature.

 
• RFT

  Repeated freeze–thaw. An experimental temperature regime in which a freeze-tolerant insect is repeatedly frozen and thawed.

 
• SCP

  Supercooling point. The lowest temperature measured in an insect before the initiation of freezing; the temperature at which supercooling stops.

We particularly thank Jeremy McNeil, Steven Chown, Janice Lord, Kath Dickinson and Caroline Williams for contributing to our thinking on this topic over the years. Heath MacMillan, Hugh Henry, Art Woods and an anonymous referee made constructive comments on an earlier draft of the manuscript.