1. Temperature- and time-dependent mortalities were studied and modelled in insects exposed in regimes with constant and alternating temperatures. In these experiments, freezing was not a cause of death.
2. Survival rates at a range of constant low temperatures (– 5 to + 1 °C) and for different exposure periods (1–14 days) were measured in the summer acclimated springtail Orchesella cincta.
3. Daily interruptions of the cold exposure with short intervals at high temperature reduced mortality or slowed the increase of mortality. This effect was stronger at higher temperature (19 vs 5 and 12 °C) and increased with the duration of the interruption (0·25–2 h).
4. The injury was reversible when the cold exposure was limited to 2 days.
5. Survival in desiccated animals (14% water loss) was reduced.
6. It is suggested that the mortality of summer acclimated springtails is caused by a complex metabolic disorder and membrane changes at low temperatures.
Salt (1950, 1961, 1963) published a series of studies describing the decrease in survival (or increase in freezing) with time in the Sawfly Cephus cinctus when frozen by internal nucleation or by external inoculation. The decrease in survival was faster at the lower exposure temperatures. It was subsequently shown in several insect species that mortality increased with prolonged exposure at a constant subzero temperature (Sømme 1982), but it is not clear whether freezing was the cause of death. The freezing and non-freezing mortality were distinguished in the Ladybird Beetle Ceratomegilla undecimnotata (Nedve˘d 1995) using the method of simultaneous measurement of supercooling point (SCP) and survival (Nedve˘d et al. 1995).
The importance of the duration of cold exposure and the relationship between time and temperature, when evaluating cold hardiness (survival) in invertebrates, has been stressed by several cryobiologists (Salt 1961; Bale 1989). The temperature of crystallization (SCP), the most often measured characteristic of cold hardiness, is time dependent (Salt 1966). However, the freezing-related mortality increased only during short exposures and was thereafter unaffected for 20 days in the noctuid moth Mamestra configurata (by Turnock & Bodnaryk 1991). Moreover, although the SCP may be used as a comparative measure of cold hardiness in some cases, it is often unreliable as an indicator of low-temperature mortality under field conditions (Baust & Rojas 1985; Bale 1987, 1989, 1993, 1996) and in many species or stages in some laboratory (Pullin & Bale 1988).
The importance of measuring survival in insects at temperatures fluctuating below and above zero, as happens in mild temperate winters, has been stressed by Bale (1989, 1991b). Survival of long exposure (almost 200 days) at constant – 5 °C and cycling – 5/+ 10 °C did not differ substantially in two nymphalid butterflies (Pullin & Bale 1989). Turnock & Bodnaryk (1991) reported positive effects on survival of high (≥ 15 °C) temperature exposure following cold stress (3 days at – 15 °C) in comparison with a lower (≤ 10 °C) poststress temperature in moth Mamestra configurata. The repair of cold injury at high temperature was repeated twice, but the duration of the interval at high temperature had a negative effect on the subsequent survival (Turnock & Bodnaryk 1991).
The aims of this study were: (1) to examine the relationship between exposure temperature and time with the resulting survival in summer acclimated individuals of the collembolan Orchesella cincta; (2) to investigate if freezing was responsible for the mortality; (3) to estimate the effect of high-temperature interruptions of the cold exposure on survival of the animals; and (4) to consider the role of desiccation on cold hardiness.
Cold hardiness in the collembolan O. cincta has been studied by measuring the SCP and survival at subzero temperature. There is a distinct seasonal variation in cold hardiness and, moreover, both summer and winter animals compose two subgroups according to their cold hardiness. Summer animals supercooled to either – 5 °C (high group) or – 11 °C (low group) (van der Woude 1987, 1988). LT50 after 24 h exposure period was either – 4 (high group) or – 6 °C (low group) (van der Woude & Verhoef 1986), while all the values were 2 or 3 degrees lower in winter animals. The percentage survival at constant – 2 °C for 1 week appeared to be a very useful characteristic to distinguish between the summer (0%) and winter (almost 100%) animals (van der Woude & Verhoef 1988).
Materials and methods
A large sample of Orchesella cincta (L. 1758) (Collembola: Entomobryidae) was collected from a pine forest near Dronten in the Netherlands in October 1995. They were kept under controlled ‘summer’ and ‘winter’ conditions (19 °C, 16:8 light:dark; 5°C, 8:16 light:dark, respectively, both 75% RH), and fed with algae (Desmococcus sp.) growing on pine branches. Experiments started after 3 weeks of acclimation, when the mortality after chill exposure corresponded to that of either summer (100%) or winter (0%) animals from the field (van der Woude & Verhoef 1988). All experiments were repeated 1 month later.
Groups of 20 adult animals were put into small plexi-glass pots on wetted plaster of Paris, with a piece of foam polyurethane. Animals in which the cold exposure was interrupted with higher temperature were also given food (piece of bark with algae). To evaluate survival at low temperature, pots were placed in climate-controlled chambers at + 1, – 1, – 2, – 3 and – 5 °C ± 0·2 K. After different periods of time (see Table 1), the pots were removed, and the animals were allowed to recover at 19 °C for 2 h. Mortality was then established as the number of animals unable to walk and jump. Since control (animals at + 5 °C) had zero mortality, there was no need to adjust the data by the Abbots formula. Groups of winter animals were treated in the same way as the summer animals.
Table 1. . Basic data matrix of the proportion of surviving summer acclimated Orchesella cincta at a range of temperatures for different periods of time used for model fitting and constructing Figs 1 and 2. Most cells contain the average of two replications each of 20 animals
The experiments evaluating the role of high temperature were performed by interrupting the cold exposure at – 3 °C. The interruptions lasted 0·25, 0·5, 1, 2 and 4 h (timed at the first half of the photophase) at 19 °C, and 2 h at 5, 12 and 30 °C. Additional to daily interruptions, those of 2 h at 19 °C were also made every second and third day.
Parallel to the control group (summer animals in pots with wet plaster of Paris), survival was measured at – 3 °C in groups of 20 animals placed on dry plaster (otherwise untreated) and in groups of animals desiccated for 3 h at 19 °C in a dry atmosphere (above silica gel, on dry plaster). The decrease in water content was measured by weighing parallel groups of animals before and after desiccation. Haemolymph osmolality was determined after Ohlsson & Verhoef (1988) in both the control and desiccated individuals.
The proportion of survivors in the experimental groups decreased with decreasing temperature and increasing exposure time, with some initial time lag, giving a sigmoid shape of the survival curves (Figs 1–5). There is a series of usable functions that describe the sigmoid curve (e.g. Gompertz, Weibull), but available statistical software enables the users to fit these functions only by using the least squares loss function. Since the dependent variable (survival) is not normally distributed, the conditions for such a regression are violated and hence not correct, although they fit the data satisfactorily. Thus, a method using maximum likelihood loss function must be applied. The pattern of survival with time was modelled using logistic regression (Statistica software, see Statsoft 1997):
with a and b being positive parameters. Parameter a influences only the position along the x-axis, i.e. the lag time and the time of 50% survival (Lt50); parameter b influences both the position and the slope (steepness) of the middle part of the sigmoid curve. The time to 50% survival (Lt50) is equal to a/b.
MODELLING TIME–TEMPERATURE RELATIONSHIP
When the three-dimensional plot of survival rate, time and temperature was smoothed with a surface algorithm (DWLS method, using Systat™ software; SYSTAT 1992), two features became clear. Firstly, the (vertical) cross-section through the range of temperatures at a particular time showed sigmoid dependence of survival on temperature similar to the dependence on time. Secondly, the (horizontal) cross-section at a given level of survival revealed a hyperbolic relationship between negative temperature (–T) and time of exposure (t), i.e. the higher the temperature, the longer the time, in a multiplicative (not additive) manner, that was required for the same mortality. One asymptote of this hyperbola equals zero time, meaning that mortality starts to occur after a very short period of time at very low temperatures.
The biological meaning of the other asymptote can be described as the lowest temperature that does not cause chill mortality even after a very long (ecologically meaningful) period of time (upper limit of the cold injury zone, ULCIZ). Mortality does occur at any temperature, but a distinction should be made between that part of mortality attributed to the chill injury and the part caused by other factors. This limit temperature is not 0 °C, since there are some chill-resistant species that survive long periods of time at subzero temperatures and on the other hand, tropical species which suffer from the effects of cold at temperatures above zero. Therefore, another parameter (c) was needed to estimate the ULCIZ:
By evaluating this equation for given values of two of the variables (S, t, T) the third one can be derived. The general equations to obtain either time or temperature are awkward to handle:
but are simplified when the parameters are known and the proportion of animals surviving is given. For S = 0·5 we may write:
Hence, the ratio a/b is the key parameter describing the time–temperature relationship in cold injury.
The experimental data obtained for summer acclimated animals (Table 1) were successfully fitted by the extended logistic equation 2, with parameters a = 3·00, b = 0·22 and c = 1·41. The χ2 value of the likelihood estimation was 738, and thus very significant (P < 0·001). When run using least squares loss function, the model explained 94% of the variation (squared correlation coefficient R2). The graphical representation of the equation (Fig 1) provides insight on the shape of dependence of survival on time and temperature (with the sigmoid shape) and the hyperbolic interrelationship between the two factors themselves. The ratio a/b describing their relationship was in our case 13·6 (providing that the temperature is expressed in degrees centigrade and the time in days).
The value of parameter c represents an estimate of the upper limit of temperature that causes chill injury (ULCIZ) in summer acclimated O. cincta (+ 1·4 °C). Evaluating equations 5 and 6 we get (for S = 0·5): t = – 13·6/(T– 1·4) and T = 1·4 – 13·6/t. Thus, the predicted LT50 for 24 h exposure is – 12·2 °C, and the temperature that causes 50% mortality in a long exposure (in this case 1 month) is + 0·9 °C.
HIGH-TEMPERATURE INTERRUPTIONS OF THE COLD EXPOSURE
Compared with continuous subzero exposure, survival increased when the cold exposure was periodically interrupted by intervals of higher temperature. This enhanced survival was increased significantly with increasing interruption duration (Fig. 2) up to 2 h but there was no further increase with a 4 h interruption interval (see Table 2). When the interruptions were applied every second day, the effect on survival was slightly less than when they were applied daily, but were still strongly positive (Fig. 3, compare with Table 2). When the interruptions were applied every third day, survival was virtually the same as in the control (Fig. 3).
Table 2. . Results of logit analyses. Equation S(t) = exp(a–bt)/(1 + exp(a–bt)) fitted to experimental data as drawn in Figs 2–5. χ2 value for goodness of fit is very highly significant in all cases. a, b = parameters, LT50 = a/b = time to 50% survival in days. Letters in ‘diff.’ column indicate group treatments not significantly different according to Cox’s F-test (P = 0·05, adjusted according to Bonferroni for multiple comparison within a group of treatments (separated by solid lines))
The animals that survived 10 days of repetitive cold and warm exposures (daily interruptions) were subsequently exposed for 4 days in a constant exposure at – 3 °C. Some of the insects (0–60% in various groups) survived this potentially lethal exposure.
There was a positive effect of temperature during the interruption interval on survival, increasing from 5 to 19 °C (Fig. 4, Table 2). High exposure temperature (30 °C) resulted in less survival, between the values found at 5 and 12 °C.
Winter acclimated animals survived well (S=0·9–1·0) in all combinations of time and temperature, whether continuous or interrupted.
ROLE OF INOCULATION AND DESICCATION
Desiccation of summer O. cincta resulted in 10·8% loss of fresh weight, corresponding to a water loss of 13·8%. Osmolality increased from 343 ± 37 mOsm kg–1 to 423 ± 66 mOsm kg–1, corresponding to a water loss of 19%, a depression in the melting point of 0·15 K, and a depression of the SCP of 0·3 K. Desiccated animals survived less well than the animals on wet plaster (Fig. 5, Table 2); untreated animals put on dry plaster survived almost as well as those on wet plaster.
APPLICABILITY OF THE MODELS
The suggested logistic model for describing the dependency of survival on time (equation 1) fitted the experimental data well, as indicated by the high χ2 values (33–168) from the maximum likelihood method. Equation
2 describing survival as a function of both time and temperature also fitted well, as shown by the χ2 and the squared correlation coefficient of the least squares method. Thus, these equations are valuable in the analysis of survival data in cryoscience when parametric description is needed.
There are also other equations/models that describe the sigmoid shape of the relationship which can be used in the three-dimensional description of survival dependence on both time and temperature. The Gompertz formula is based on the exponential increase of the hazard function over time, therefore there is a double exponential, e.g. in the form (modified after Batschelet 1973):
The Gompertz function has been used by actuaries and has been applied in entomology, e.g. to mortality rate in Blow Fly (Readshaw & van Gerven 1983); however, the analysis uses an incorrect least squares loss function. The problem is that the computing of parameters using maximum likelihood estimation of the equation is not available in most statistical packages. In practice, the least squares method is widely used and does not cause severe deviations. If visual observation shows that the resulting curve runs through the data points then the use of the least squares method is justified.
The Weibull model has been used a few times to compare survival after cold exposure (Worland, Block & Rotheery 1992; Han & Bauce 1995) and it seems a reasonable choice in two-dimensional modelling (survival as a function of either time or temperature, modified after Dent 1997):
The model becomes complicated (with four parameters) in three-dimensional modelling (survival as a function of both time and temperature), but it represents an advantage, since separated parameters (and hence ‘weights’) are given to the two independent variables.
The listed methods (logistic, Gompertz and Weibull) have advantages over others in plotting survival data. Salt (1961, 1963) used a semilogarithmic plot of survival with time on a linear scale, and the number of unfrozen individuals on a logarithmic scale. This type of transformation produces an almost linear decrease of survival with time over a wide range of time. Values may, however, deviate in margins (very short and very long exposures). Zero survival is necessarily excluded from the plot, although it has its meaning. The logit transformation used by Turnock (1993) excludes even initial total (unit or 100%) survival.
There can be low precision in parameter estimation of any model if the range of S is low, mainly if it does not contain the point of inflection, or the value 0·5. Turnock’s method is useful in integrating and accumulating the chill injury through a series of different temperatures and exposure periods. In the cases where the temperature fluctuates and rises to the zone of repair of cold injury, however, the partial exposures may not be additive and a different approach may be necessary.
The calculated ULCIZ in summer acclimated O. cincta (+1·4°C) might be a realistic estimate. However, temperatures slightly above zero induce acclimation (van der Woude & Verhoef 1986); thus, after about 2 weeks of exposure, the ‘summer’ animals will change to ‘winter’ animals.
The parameters of equation 2 were estimated for the published data on survival of Lacanobia atlantica (Table 1 in Turnock 1993) as follows: a = 1·23, b = 0·0106, c = – 15·6, χ2 = 1124, variance explained (using least squares loss function) R2 = 0·69. The lower correlation coefficient is probably caused by the few survival values in the data matrix that are near to 50%. Estimation of the ULCIZ (– 15·6 °C) corresponds with the raw data and also with Turnock’s hyperbola (Fig. 2 in Turnock 1993). His estimation of ULCIZ, based on nonzero slope of fitting polynomial, was between – 12·6 and – 15 °C.
Where ULCIZ lies much below zero and does not involve acclimation, there might be other objections in considering this estimation to be valid in long-term experiments (or for long storage purposes). Chill injury at moderate temperatures and during moderate exposure periods may be attributable to metabolic disorder or phase changes in membranes, while other types of injury, such as desiccation of the supercooled organism, become more important after long exposure (see Ring & Danks 1994). Thus the actual ULCIZ will tend to be much higher, mainly depending on what is considered an ecologically valid duration of exposure. In tropical and parasitic species, the instantaneous cause of injury and death at low (but still positive) temperatures might be due to starvation of the inactivated insect.
Although the estimation of ULCIZ may seem valid, the estimation of the lower lethal temperature in brief exposures, based on equation 2 and parameters obtained for longer periods of time, is erroneous. The value LT50(24 h) = – 12·2 °C for summer acclimated O. cincta is incorrect, as the SCP of the freeze-susceptible animals fluctuates between – 6 and – 9 °C in summer (van der Woude 1987), and previous studies estimated the absolute lower lethal temperature LT100(24 h) at – 6·5 °C (van der Woude & Verhoef 1986). In Lacanobia atlantica, there was 18% survival at – 22·4 °C (Turnock 1993) with a mean SCP of – 26·4 °C; while the value we have calculated from equation 4 LT18%(4 days) = – 274 °C, a temperature of no physical sense.
The above-mentioned contradictions are examples of different processes causing injury and death additional to the chill injury (metabolic disorder and membrane phase changes), i.e. freeze injury. Probability of freezing increases dramatically at lower temperatures and freezing prevails over chill injury, making equation 2 of little value. When describing mortality within the whole range of subzero temperatures, equation 2 should be combined with another equation for freezing probability. This may be a negative exponential, without a lag in the initial period. In the case of O. cincta, two such terms should be added to equation 2 as the distribution of SCPs in O. cincta is bimodal (van der Woude 1987).
IMPROVEMENT OF SURVIVAL
Survival of exposure at – 3 °C, a temperature that normally causes death in 50% of animals after 3 days and in most individuals after 4 days, was much higher when interruption procedures were applied. The warm interval resets the physiological state of the animals towards the initial values. The chill injury during subsequent cold exposure was almost the same as for the initial exposure. Therefore, chill injury did not accumulate from day to day; the effects of chilling were almost completely reversed or repaired by returning the insects to higher temperatures.
Chill injury was reversible for about 2 days, corresponding to the time lag after which survival started to decrease steeply. Following the reversible phase (after 3 days’ exposure), even living animals did not recover during the warm interval and died during the following cold exposure. This suggests a limit of interpreting results of survival experiments when survival is counted shortly after recovering from the cold exposure, as was the case in the present study. Observing survival for some substantial time after the end of the exposure or establishing viable progeny is highly recommended (Bale 1987; Baust & Rojas 1985).
The ability to recover without accumulating chill injury might be of ecological importance for many temperate invertebrates, since alternating cooling and warming periods during autumn and spring are more common (measured on the surface, not deeply in soil) than prolonged periods of constant cold. The temperatures that occur during the day time, alternating with freezing nights, will be lower than the 19 °C used in the present experiments. The improvement of survival acquired after 2 h at 5 °C was weak (similar to the effect of 19 °C lasting only 0·5 h). However, when considering the effect of the duration of the warm interval, a day at + 5 °C may ensure surviving a night with temperatures below zero.
The possibility of improving survival by the alternation of cold and warm intervals for a long period of time may have an application in long-term storage of insects. Recently, Gagne & Coderre (1996) studied survival in larvae of the Ladybeetle Coleomegilla maculata, at 4 and 8 °C, and their subsequent development and voracity. Survival was high after 2 weeks’ exposure but decreased dramatically after 3 weeks. Application of high-temperature intervals might extend the storage period.
Turnock & Bodnaryk (1991, 1993) reported on the importance of poststress temperature and its duration on the expression of chill injury in Mamestra configurata. In O. cincta, the non-lethal chill injury was repetitively reversed by a warm interval to provide a chance to survive subsequent cold exposure. Turnock et al. (1983) found that survival of M. configurata pupae given a continuous exposure to – 10 °C and an exposure interrupted by 28 days at 0 or – 5 °C did not differ, indicating that repair of cold injury does not occur at or below 0 °C. Survival was even slightly less in the butterfly Inachis io when the cold exposure (– 5 °C) was interrupted daily by high temperature (+ 10 °C) (Pullin & Bale 1989). Survival of the weevil Rhynchaenus fagi at temperatures cycling from + 2 to – 15 °C over 48 h was higher than survival at a constant – 15 °C, but survival at cycling + 10 to – 5 °C was less than in constant – 5 °C (Coulson & Bale 1996). Therefore a distinction should be made between temperatures which promote repair of injury and those that cause deacclimation.
Winter acclimated animals survived in high numbers at all combinations of time and temperature, either continuous or interrupted. Thus, it can be concluded that short warm intervals did not cause a higher mortality, i.e. did not deacclimate the winter animals to the summer level. The possibility that the summer animals exposed to repetitive cold and warm intervals do acclimate to low temperature cannot be excluded. This response occurs in diverse insect species at temperatures fluctuating around zero. It seems that the summer acclimated O. cincta were able to achieve a moderate cold hardiness in an interrupted exposure in these experiments. Since this epedaphic species does not show strong avoidance behaviour (as other collembolans do, see van der Woude & Verhoef 1986), a rapid cold hardening ability is quite important.
CAUSE OF DEATH
Mortality in summer acclimated O. cincta is not caused by freezing but by non-freezing cold injury (some metabolic disorder and failure, or membrane lipid changes). The positive effect of the interruption of cold exposure on survival might be explained as interrupting the period during which the chance (stochastic probability) of freezing (nucleation or inoculation) increased, subsequently starting again from zero time. In that case, no clear positive effects of the length of the warm interval and the temperature during the interruption would be expected. Instead, it would be likely that of the animals that survived 3 days at – 3 °C (50% of initial number), 50% of the remainder would survive the second 3-day period. This was, however, not the case. Therefore, it does not seem plausible to explain the positive effect of the interruptions on survival simply by diminishing the chance of freezing.
The injury could also be attributed to the desiccation that takes place in supercooled animals surrounded by ice (Zachariassen 1991). This may be important after a longer period of time, but this cannot explain the effect on survival of the duration and of the temperature of the interruption interval. Partial dehydration may be favourable in several ways in overwintering insects (Block 1996), either in freeze-susceptible species by depressing the SCP or in freeze-tolerant ones, but no survival improvement was described in chill (non-freeze) injury in insects.
The deleterious effect of artificial desiccation before cold treatment suggests that the natural desiccation during cold exposure can be dangerous when water loss is extreme (> 20%). However, among diverse collembolans, O. cincta is relatively strongly resistant to desiccation (Verhoef & Witteveen 1980). It tolerates fast dehydration and long dry periods, resulting in up to 47% water loss (Verhoef & Prast 1989). Haemolymph osmolality of about 450 mOsm kg–1 can be commonly found in the field during frost periods (van der Woude 1987). The expected depression of SCP related to desiccation was too small to improve survival even if mortality were caused by nucleated freezing. There was no death caused by inoculation by external ice as well, because survival was the same on the dry surface. Moreover, it was proven elsewhere (Lavy 1996) that gut evacuation during starvation in O. cincta, which should void efficient nucleators from the body, did not improve survival, although the SCP depression has been found to correlate with gut clearance in previous studies (van der Woude 1987, 1988).
Turnock & Bodnaryk (1991, 1993) ascribed the chill injury in insects to phase changes in membrane lipids. It is suggested here that the nature of the chill injury in summer acclimated O. cincta may be also a complex metabolic disorder. It may include a non-proportional decrease (imbalance, decoupling) in enzymatic reactions and transports, resulting in the accumulation of some harmful compounds (e.g. lactic acid, the nitrogenous waste substances) and lack of other, essential compounds. During the warm interval, a substantial part of the waste is metabolized or excreted, essentials are produced and distributed, some injuries are repaired. Rebuilding the order in an organism takes time, which explains the positive effect of the duration of the interruption. As temperature is important in many metabolic processes, a relatively high temperature during the interruption resulted in higher survival.
Prolonged chilling might cause disorder to such an extent that it is impossible to be repaired, or it may cause injury to the reparative mechanisms themselves (waste compounds reaching toxic concentrations). Death would then follow even in virtually healthy individuals (Bale 1991a), or developmental malformations and delays, and a decreased reproductive potential, will occur (see Sehnal 1991; Turnock et al. 1983). The potential of insects to recover from short exposures to temperatures that may become lethal in longer exposures complicates the issue of calculating a cumulative injury at fluctuating temperatures. Predictions of field mortality will have to include the effects of such warmer intervals during the entire exposure. Turnock & Bodnaryk (1993) also consider the importance of postexposure conditions in predicting field mortality.
The chill resistance of winter acclimated animals might be ascribed to the lower metabolic activity (hence a relationship between cold hardiness and diapause), to the rebuilt metabolic and transport systems (membrane lipid composition), and possibly also to changed neural and hormonal control of the organism.
O. Nedve˘d was awarded a fellowship by NUFFIC, D. Lavy was supported by the Life Sciences Foundation (SLW, BION) which is subsidized by the Netherlands Organization for Scientific Research (NWO). The first author also thanks Faculty of Biology, Vrije Universiteit, Amsterdam, for the opportunity to work at its laboratories. We are grateful to members of the Department of Ecology and Ecotoxicology of Animals for diverse kind assistance.