Effects of temperature on the development of an arctic Collembola (Hypogastrura tullbergi)


H.P. Leinaas. E-mail: h.p.leinass@bio.uio.no


  • 1Embryonic development, juvenile moulting and growth rates, and maximum size of the arctic collembolan Hypogastrura tullbergi were investigated at 5, 10 and 15 °C in laboratory experiments. The embryonic development was also investigated at 21 °C.
  • 2The lower temperature threshold of the embryonic development (t0) was −1·3 °C, possibly reflecting a slight cold adaptation. The temperature of maximum development rate exceeded 21 °C.
  • 3Instar duration rate was linearly related to temperature. Growth per instar, however, was thermally dependent, giving an overall non-linear correlation between growth and temperature. This emphasizes the importance of studying growth and moulting as separate processes in Collembola.
  • 4Development of genital area (number of hairs) over succeeding instars was affected by the temperature, suggesting that a higher proportion of individuals reached the adult stage in an earlier instar at 15 than at 10 °C.
  • 5Contrary to the general pattern in ectotherms of increasing size at lower temperatures, adult H. tullbergi reached a larger maximum size at 15 than at 10 °C.
  • 6No aspects of postembryonic development in H. tullbergi showed any signs of cold adaptation, probably because it is more important to be able to efficiently utilize high temperatures that frequently occur in the microhabitat during sunny periods in summer.


An important question for understanding population dynamics and life-history strategies of terrestrial arthropods in the high Arctic is how they are adapted to meet problems associated with the low-temperature sum received during the cool, short growth season. In many insects with a distinct voltinism (a fixed number of generations per year, or years per generation), a major problem is to what extent the temperature sum received may be sufficient to complete the life cycle within appropriate time ( MacLean 1983; Danks 1992; Strathdee et al. 1993 ). Moreover, since development rate (time to first reproduction) has a great impact on the intrinsic rate of natural increase, r (e.g. Stearns 1992) and the number of overwintering periods may affect juvenile survival ( Atkinson 1996; Coulson et al. 2000 ), adaptations to regional climate are also likely to occur in species with flexible life cycles. A number of authors have studied effects of temperature on development rates in relation to local or regional adaptations in insects (e.g. Groeters 1992; Ayres & Scriber 1994; Benson, Zungoli & Smith 1994; Honek 1996; Blanckenhorn 1997; Bryant, Thomas & Bale 1997), but apparently not in high-arctic terrestrial arthropods. Here we present a study on the high-arctic Collembola Hypogastrura tullbergi (Schäffer), focusing on reaction norms that might elucidate strategies to optimize performance in its thermal environment.

Collembola, having free-running life cycles, are useful model organisms for studying general aspects of climatic adaptations. Most species display all age classes throughout the year and are not dependent on seasonally occurring resources ( Addison 1981; Hertzberg, Leinaas & Ims 1994; Bale et al. 1997 ; Birkemoe & Sømme 1998). They are also highly successful in the high Arctic, exceeding all other groups of terrestrial arthropods both in abundance and species richness in most habitats ( Bengtson, Fjellberg & Solhøy 1974; MacLean et al. 1977 ; T. Birkemoe & H. P. Leinaas, personal observation).

Normally, development rates will show a sigmoid curve when measured over the whole temperature range where development may occur (e.g. Wagner et al. 1984 ). At lower temperatures the rates tend to level off, making it difficult to define a point when development stops. At high temperatures, increasing heat stress will gradually reduce the slope of the curve, reaching a maximum development rate before it decreases and finally stops due to heat stupor and death. (The point of maximum development rate is often referred to as the optimum temperature, but to avoid confusion with the Darwinian optimization concept, we here call it tdmax.) In intermediate temperature ranges, response curves are generally assumed to be linear ( Wagner et al. 1984 ; Atkinson 1996). Although only strictly valid for the linear range, the animal temperature response is commonly described by the inverse of the slope of the line and the interception of the extrapolated line with the temperature axis ( Wagner et al. 1984 ; Cossins & Bowler 1987; Lamb 1992; Honek 1996; Blanckenhorn 1997; Bryant et al. 1997 ). These two parameters are denoted thermal constant (K) and lower threshold temperature (t0), respectively. The t0 is a theoretical value, which may diverge from the real threshold because of the levelling of the curve at low temperature.

Thermal adaptation in ectotherms may involve differential effects on processes, like embryonic development rate, instar duration, growth, ‘stages’ in sexual development and maximum size. To relate temperature responses to animal performance, we studied these developmental processes throughout the life span of H. tullbergi, with a thermal range (5 and 15 °C) based on daily mean summer temperatures recorded just below the soil surface in the New Ålesund area on Svalbard ( Coulson et al. 1993 ; Birkemoe & Sømme 1998; Hertzberg et al. 2000 ).

Materials and methods

Hypogastrura tullbergi is a high-arctic, circumpolar species, which usually is among the most common species in moderately dry to dry habitats ( Fjellberg 1994). The animals used in this study originated from a population sampled at Marmorpynten (80°11′ N 18°30′ E) during an expedition to the north-east part of Svalbard in August 1995. Soil samples collected in the field were stored for 6 months at 2 °C before the animals were extracted and transferred to 15 °C. They soon started to lay eggs, and all experiments were done with these eggs and the emerging F1 generation. The experimental eggs and animals were placed in culture boxes (d = 3·4 cm, h = 3 cm) with a bottom made of plaster of Paris mixed with charcoal. A few drops of water were added once a week to maintain high air humidity. All experiments were run in a 24-h photoperiod to simulate the arctic summer.

Collembola have indeterminate growth and continue to moult throughout their entire life span. After becoming adults, the animals alternate between reproductive and non-reproductive instars ( Niijima 1973; Rapoport & Aguirre 1973), which are easily recognized by distinct internal reproductive structures ( Fjellberg 1994). In large adults growth gradually ceases, and old animals may shrink in size ( Burn 1981; Leinaas & Bleken 1983). Owing to a diapause, H. tullbergi will not become reproductively active until exposed to cold, even though they continue to moult and grow ( Birkemoe & Leinaas 1999). Consequently it is difficult to determine exactly when they become adults if reared under constant temperatures. Several authors have described a distinct development in chaetotaxy (number and arrangement of hairs) of the genital area over successive juvenile instars in Collembola ( Agrell 1948; Petersen 1965; Walters 1968; Grégoire-Wibo 1974; Takeda 1976). The genital area of reproducing animals collected in the field was therefore analysed and, based on the numbers of hairs found in 32 males and 24 females, 22 and 16 were chose as approximate distinctions between juveniles and adults in the two sexes, respectively.

Eggs for studying embryonic development was obtained from cultures that were kept at 15 °C and inspected daily. Hypogastrura tullbergi deposits egg masses in batches. New batches (≥ five eggs) were transferred to clean culture boxes (day 1 of development). Altogether the following numbers of eggs were transferred (no. of batches in parentheses): 247 (14), 285 (18) and 389 (18 or 19), at 5, 10 and 15 °C, respectively. In order to compare the temperature response with similar studies on related species from the temperate zone, 21 °C was also included in the egg development study. The additional experiment was done later than the others, when fewer eggs were available, and therefore included 73 eggs (three batches). The boxes were inspected daily and the number of eggs hatched was recorded for each batch.

Postembryonic development includes traits such as growth, moulting and sexual development. Since Collembola continue to moult and grow as adults, we studied their postembryonic development throughout their entire life span. Algae (mainly Pleurococcus) growing on bark of deciduous trees were provided as food and replaced twice a week to ensure ad libitum supply. The animals were kept in groups in the boxes as single specimens had low survival.

Growth and moulting were studied in animals that were transferred to 5, 10 and 15 °C less than one day after hatching. They were reared in groups of 10 in five culture boxes placed at each temperature. The experiment lasted for 259 days. At that time 31 specimens were still alive at 10 °C and 15 animals at 15 °C. Owing to a defective climate chamber, the 5 °C treatment was terminated at day 122 when 32 specimens were left in the culture boxes. The number of exuviae and dead individuals were counted and removed twice a week. Body lengths of animals were measured (to an accuracy of 0·063 mm at 16× magnification) in the culture boxes after immobilizing the animals by a constant flow of CO2. Measurements were taken every 7 days during the first 10 weeks, and then at 2- or 4-week intervals depending on the development of the growth curve. As the experiment lasted beyond termination of growth, maximum sizes were seen directly from the growth curves. The sexes were distinguished from day 84. The sex ratio was close to unity in all treatments.

Instar duration, body growth and the development of the genital area over successive instars at different temperatures were studied in a separate experiment. Groups of 30–120 specimens of equal age (hatched within 24 h) were transferred to 10 culture boxes, of which 4 were kept at 10 and at 15 °C and 2 at 5 °C. The study lasted as long as moulting within each box occurred sufficiently synchronously to allow comparison of equal aged individuals of the same instar (moulted within the last 2 days). Exuviae were counted and removed five or six times a week. Immediately after each synchronous moult, 10 specimens from each of two or four cultures at each temperature were transferred to 70% alcohol and heated to 70 °C to stretch the animals. Some individuals, however, were lost during the subsequent preparation procedure. The body length of the first instar was common to all treatments and represented by 10 hatchlings. At 5 °C only the two first moults occurred synchronously, limiting the study to the first three instars at this temperature. At 10 and 15 °C the animals moulted synchronously six times (most material of instar 4 was accidentally lost and therefore not included in the analyses). In instars 5–7, the body length of each sex was treated separately. After recording the body length, each animal was cleared in Gisin’s mixture ( Gisin 1960) and the chaetotaxy of the genital area was investigated at 400× magnification.

Since it was impossible to rear individuals separately, or mark them individually when reared in groups, all animals in each culture box were used as the statistical unit when estimating maximum size and duration of instars. The egg batch was used as the unit to estimate embryonic development time. Variations in the number of animals per box and eggs per batch affect the precision of the estimates. Therefore, as a standard procedure the estimates at the box and batch level were weighted by the inverse of their variance. Thermal effects on development rates were analysed by linear regression; rate = a + bt (t = °C), giving t0 = −a/b and K = 1/b. Testing for effects of temperature on body length and genital development was done by ANOVA analyses. Both body length data and number of genital hairs were log-transformed prior to the statistical analyses to homogenize the variance.


Embryonic development time

Embryonic development time declined from 5 to 21 °C, corresponding to a linear increase in development rate ( Fig. 1), with a theoretical threshold value t0 = −1·32 °C, and a thermal constant K = 295 day–degrees ( Table 1). The linear curve suggests that tdmax is higher than 21 °C. But the increased variation in development rate at 21 °C ( Fig. 1) may reflect that some animals were affected by heat stress. Also a decline in hatching success, from 94, 98 and 99% at the three lowest temperatures, to 86% at 21 °C (contingency table, χ2 = 30·0, P < 0·001), indicated increasing heat stress. With t0 < 0 °C, the heat sum (day–degrees) needed to complete development increased with temperature: H = 226·7 + 2·81t;r2 = 0·72, P < 0·001 (t = degrees above 0 °C).

Figure 1.

Egg development time (○) and the rate (days-1) of embryonic development (●) of H. tullbergi at four temperatures. 95% confidence intervals are given. The estimates are based on 232, 213, 332 and 64 eggs hatched at 5, 10, 15 and 21 °C, respectively.

Table 1.  Embryonic development and instar duration against temperature (t) for H. tullbergi. Theoretical threshold values (t0) and thermal constants (K) of the rates (y) of embryonic development and instar duration (in days−1) are calculated from linear regression y = a + bt; where t0 = −a/b and K = 1/b. Regression statistics are given
 K (± SE)t0 (± SE)r2P
Embryonic development rate295 (±7)−1·32 (±0·27)0·98<0·001
Instar duration rate 88 (±3) 0·11 (±0·35)0·99<0·001

Postembryonic development

At 15 °C the growth pattern described a sigmoid curve, with some irregularities in larger animals, before a maximum size was reached after about 16 weeks ( Fig. 2a). Later, the animals showed increasing signs of senile degrowth. A similar pattern, but with slower growth, was seen at 10 °C, with maximum size obtained after 23 weeks. Females grew larger than males, and both sexes became largest at 15 °C. The maximum mean body lengths (in mm) reached by females were 1·89 vs 1·93, and by males 1·65 vs 1·73 at 10 and 15 °C, respectively. A two-way ANOVA showed a significant effect on maximum mean size of both temperature (10 vs 15 °C) (F1,17 = 8·07, P = 0·011) and of sex (F1,17 = 94·16, P < 0·001). At 5 °C growth was much slower and during a 1-month period (week 5–9) it was hardly detectable. The much larger difference in growth between 5 and 10 °C than between 10 and 15 °C, is illustrated by the heat sum needed to reach 1·0 mm being twice as large at 5 °C than at the two other temperatures (about 500 vs 250 and 255 day-degrees).

Figure 2.

Growth and moulting of H. tullbergi kept at 5 (○), 10 (▵) and 15 °C (+). All estimates are based on pooled data from five culture boxes, each initially with 10 animals. (a) Mean body lengths with 95% confidence intervals. (b) Mean cumulative number of moults per individual.

The moulting frequency remained fairly constant at each temperature throughout the entire life of the animals, except for a 1-month period at 5 °C with no growth and hardly any moulting ( Fig. 2b). Animals surviving 260 days at 15 °C moulted on average 30 times. Instar duration ( Fig. 3) remained stable through the six moulting cycles studied at 10 and 15 °C, with 8·00 (SE = 0·07) and 5·25 (SE = 0·03) days per instar, respectively (r2 > 0·99; P < 0·001 in both regressions). Also the two first instars at 5 °C were almost identical in duration with 17·83 (SE = 0·12) days per moult. Moulting rate increased linearly with temperature from 5 to 15 °C, with t0 close to 0 °C ( Table 1), and the heat sum needed to reach the third instar was constant across temperatures (H = 89·6 − 0·08t;r2 = 0·02, P < 0·72).

Figure 3.

Mean duration (days) of successive instars at 5 (○), 10 (▵) and 15 °C (+). Each point is mean ± 1 SE of the number of replicate cultures given by figures.

At each of the first three instars, the body length did not differ between 10 and 15 °C, but was significantly shorter at 5 °C ( Fig. 4). Later (instar 5–7), there were also some differences between 10 and 15 °C, with males being significantly larger at 15 °C independent of instar, whereas females became larger at 15 °C in the seventh instar only ( Fig. 4, Table 2).

Figure 4.

Mean body length with 95% confidence intervals of H. tullbergi at successive juvenile instars, at 5 (○), 10 (▵) and 15 °C (+). The number of specimens measured is given. (a) For the first three instars (not possible to sex) growth was recorded at all three temperatures. (b) In instars 5–7, the sexes were separated, but data were only available for 10 and 15 °C.

Table 2.  Two way ANOVAs on the effect of temperature (10 vs 15 °C) and instar number (5–7) on body length of H. tullbergi. Males and females are analysed separately
SexCategorical variablesdfSSFP
MaleTemperature  10·01 7·83 0·006
 Instar  20·2782·21<0·001
FemaleTemperature  10·00 0·74 0·391
 Instar  20·2065·64<0·001
 Temperature × instar  20·02 7·07 0·001

The sexes can be separated in the fourth instar by a vertical (males) or horizontal (females) split in the genital area. This stage has no genital hairs, but thereafter the number of hairs increased for each instar ( Fig. 5) in a pattern that was affected by temperature. In females there was also an interaction between instar and temperature ( Table 3). At 15 °C, both sexes tended to show a unimodal frequency distribution in the number of hairs at each instar, giving an impression of discrete ‘stages’ in development. In males, however, there was considerable overlap in number of hairs between instars 6 and 7. As judged from the number of genital hairs in reproducing animals from the field, some males may have already reached the adult stage by the sixth instar. But instar 7 appeared to be the first mainly adult instar at 15 °C in both sexes. The situation was more complicated at 10 °C, where a number of animals lagged behind in this development in both instars. This was most clearly seen in females, but a similar, less marked tendency was also seen in males.

Figure 5.

Frequency distributions in the number of genital hairs on males and females over three successive instars of H. tullbergi at 10 and 15 °C. Vertical dashed lines represent the lowest number of setae found on reproducing specimens collected in the field. The scales given apply to all three instars.

Table 3.  Two way ANOVAs on the effect of temperature (10 vs 15 °C) and instar number (5–7) on the number of genital hairs in H. tullbergi. Males and females are analysed separately
SexCategorical variablesdfSSFP
MaleTemperature  10·08  4·01 0·047
 Instar  24·78115·21<0·001
FemaleTemperature  10·11  5·32 0·023
 Instar  28·10204·86<0·001
 Temperature × instar  20·22  5·58 0·005


Embryonic development

The relatively low theoretical threshold value (t0 < 0 °C) of embryonic development indicates a degree of cold adaptation in H. tullbergi. Many surface-living, temperate Collembola have considerably higher values ( Hale 1965; Niijima 1973; Mertens & Blancquaert 1980; von Allmen & Zettel 1983; Takeda 1984; van Straalen & Joosse 1985). Also subzero values have been found in several surface-dwelling species of the family Hypogastruridae from the temperate zone (calculated from Thibaud 1970), but as they were sampled in caves their low t0 could reflect an adaptation to the stable, cool climate there. On the other hand, a tdmax in H. tullbergi exceeding 21 °C is fairly high. For instance H. manubrialis from Poland, with a very high t0 = 6·5 °C, had a tdmax of only 20 °C ( Mikulski & Wosiak 1960). None of the cave species studied by Thibaud (1970) showed a reduction in development rates towards the highest temperatures when hatching occurred (18·5–27·5 °C). But compared with H. tullbergi at 21 °C (86% hatching), their hatching success was low. Four species did not hatch at 22·5 °C and the other seven had a hatching success of (20–70%) at that temperature.

Postembryonic development

For a given temperature, the moulting rate remained fairly constant throughout most of the animal’s life, even after growth had ceased. The fact that the thermal sum needed to complete an instar remained constant over the temperature range studied (t0 very close to 0 °C) indicated an adaptation to variable rather than to low temperature ( Pritchard, Harder & Mutch 1996).

Growth per moult has been treated as a constant factor in several studies on Collembola ( Agrell 1948; Walters 1968; Joosse & Veltkamp 1970), and moulting has been used to describe variation in growth across temperatures ( Burn 1984). In the present study, however, we found that growth per instar varied not only with age, but also with temperature. This emphasizes the importance of studying moulting and growth as separate processes. Until the third instar, when growth per moult could be compared for all three temperatures, it was distinctly lower at 5 °C. Combined with the linear increase in moulting rate with temperature, this gives a non-linear relationship between growth per unit time and temperature. The growth curves ( Fig. 2) confirmed this pattern by a much larger difference between 5 and 10 °C than between 10 and 15 °C. This non-linearity shows that growth rate under fluctuating and constant temperature with the same mean may differ considerably, but the effect of temperature fluctuation depends on both amplitude and mean temperature, and may be positive as well as negative. This emphasizes the differential approach of focusing on thermal adaptations, which requires standardized constant temperature regimes, and the study of development and population dynamics under field conditions, when natural temperature fluctuations have to be considered.

The disproportionately strong reduction in growth at 5 °C may partly reflect a behavioural response leading to reduced feeding activity at low temperature (see Burn 1981), which automatically would reduce the growth rate. Such an effect is probably seen in weeks 5–9 of the growth study at 5 °C, when neither moulting nor growth occurred. However, during the last juvenile instars growth increased even from 10 to 15 °C, indicating also a physiological effect underlying the thermal dependency of growth per instar. Nevertheless, the premature termination of the 5 °C experiment after 4 months makes our description of the differential growth pattern between the three temperatures incomplete. For instance, we do not know if low temperature affects size at maturation and thus the period of juvenile growth rate.

Our results point to instar 7 as the first adult instar at 15 °C, but possibly with some fast-developing individuals becoming adults in the sixth instar. This agrees with an experiment at 15 °C, where diapause had been terminated by cold exposure, and a single female reproduced at an age corresponding to the sixth instar ( Birkemoe & Leinaas 1999). However, it is unlikely that transition to the adult stage can be linked to a definite number of genital hairs. Our results show that juvenile instars cannot be identified unambiguously by genital chaetotaxy, and most importantly, it is evident that temperature affects the development rate (per instar) of the genital area. We believe that the bimodal frequency distribution in number of hairs at 10 °C indicates that some individuals became adults at a later instar than at 15 °C. Thus high temperatures may shorten the juvenile stage more than just its effect on instar duration, and thus further speed up maturation in warm summers.

Among ectotherms high temperatures commonly lead to early maturation and reduced adult size (e.g. Charnov 1993; Atkinson 1994; Partridge & French 1996; Atkinson & Sibly 1997; Hutchings & Jones 1998). In H. tullbergi, however, maximum size was found to increase from 10 to 15 °C. Thus in this species increased growth per instar at higher temperatures may have more than compensated for a tendency to mature at an earlier instar.


Any species that is adapted to the generally cool environment of the high Arctic will show some kind of cold adaptation, at least to survive winter cold. The question about cold adaptation (temperature compensation) in biological processes during the growth season is more complex. Even if not evident in growth or development, other forms of cold adaptation might have taken place, e.g. at the molecular level ( Clarke 1991). Here, however, we have focused on phenotypic characteristics that are closely related to the species’ fitness (generation time and adult size) ( Stearns 1992), in order to elucidate ultimate factors underlying the climatic adaptation of the species. From that perspective, the lack of cold adaptation in postembryonic development is striking.

The most likely reason for this lack of cold adaptation in H. tullbergi is the high temperatures it may experience during sunny periods. Such conditions vary greatly both within and between summers. During the exceptionally warm summer of 1998, soil surface temperatures exceeding 35 °C were recorded (in a study field 20 km SE of Longyear byen), while at the other extreme some summers are persistently cool and wet, with the soil temperature rarely exceeding 10 °C (H. P. Leinaas, unpublished data). However, in most years there are sunny periods when the soil temperature in the middle of the day often reaches 15–20 °C (e.g. Birkemoe & Sømme 1998). With such a highly variable and unpredictable climatic regime, there may be more to gain by efficiently utilizing higher rather than lower temperatures, depending on the long-term (over generations) contribution of low and high temperatures to the total heat sum of the microhabitat. This is in contrast to the stable temperature of marine polar environments, where cold adaptation may be defined to include all aspects of an organism’s physiology that allow it to live there ( Clarke 1991).

Since thermal adaptation seems strongly constrained by diurnal variation in microclimate, it may be more dependent on the expected number of sunny days during summer than of latitude. For instance, in the Maritime- and Subantarctic the summers are typically cool and cloudy ( Danks 1999) and cold adaptation may well be more common there than in many areas of the high Arctic. This may explain why both the collembolan Cryptopygus antarcticus and the oribatid mite Alaskozetes antarcticus at Signy Island showed clear signs of cold adaptations, by a very slow increase in moulting frequency from 5 to 15 °C, and a low tdmax for growth (about 7 °C), respectively ( Burn 1984; Convey 1994a), and C. antarcticus even continue to grow during mild winter periods ( Convey 1994b). On the other hand, the coexisting collembolan Parisotoma octoculata did not show a similar cold adaptation to that of C. antarcticus ( Burn 1984), emphasizing our limited understanding of why species adapt differentially to the same environment. Also the lack of studies that takes the same set of developmental parameters into account, makes regional comparisons of thermal adaptations in polar terrestrial ectotherms premature.

The more cold adapted embryonic development of H. tullbergi agrees with the fact that eggs deposited in the soil or under small stones, etc., experience less temperature fluctuation (and somewhat lower mean summer temperature) than the animals on the soil surface. A similar effect of differential temperature fluctuation is described for egg development rates of soil surface dwelling vs. deeper living collembolans ( van Straalen 1994).


We have here reported a detailed study of a series of reaction norms in animals from one population of a high-arctic Collembola. The lack of cold adaptation in animal growth and development may be explained by the great diurnal variations in ground temperatures. However, to improve our understanding of the complicated processes underlying thermal adaptations in these animals, more populations from areas with different climate should be compared. This study also emphasizes the importance of including different developmental processes when studying a species’ adaptation to climatic conditions. Stages may be adapted to different microclimate (eggs vs hatched animals), and different processes within the same developmental stage (moulting, growth, sexual development) may show different norms of reaction.


The study was financed by the Norwegian Research Council under the Arctic Terrestrial Ecology Program. Søren Bondrup-Nielsen, Steve Coulson, Ger Ernsting, Katarina Hedlund, Sten Rundgren, Lauritz Sømme and Asbjørn Vøllestad gave valuable comments on an earlier version of the manuscript (TB’s PhD thesis). Bente Einan Eriksen kindly helped with technical assistance in the laboratory.

Received 12 November 1999; revised 19 April 2000;accepted 26 April 2000