Phenotypic plasticity of thermal tolerance contributes to the invasion potential of Mediterranean fruit flies (Ceratitis capitata)

Abstract

1. The invasion success of Ceratitis capitata probably stems from physiological, morphological, and behavioural adaptations that enable them to survive in different habitats. However, it is generally poorly understood if variation in acute thermal tolerance and its phenotypic plasticity might be important in facilitating survival of C. capitata upon introduction to novel environments.

2. Here, by comparison of widely distributed C. capitata with a narrowly distributed congener, C. rosa, we show that both species have similar levels of survival to acute high and low temperature exposures under common rearing conditions. However, these species differ dramatically in the time-course of plastic responses to acute low temperature treatments.

3. The range of temperatures that induce rapid cold hardening (RCH) are similar for both species. However, C. capitata has two distinct advantages over C. rosa. First, at 5°C C. capitata develops RCH significantly faster than C. rosa. Second, C. capitata maintains a RCH response longer than C. rosa (8 vs. 0.5 h).

4. A simple population survival model, based on the estimated time-course of RCH responses determined for both species, was undertaken to simulate time to extinction for both species introduced into a similar thermally variable environment. The model showed that time to extinction is greater for C. capitata than for C. rosa, especially in habitats where temperatures frequently drop below 10°C.

5. Thus, variation in RCH responses may translate into significant variation in survival upon introduction to novel thermal habitats for C. capitata, particularly in cooler and more thermally variable geographic regions, and may contribute to their ongoing invasion success relative to other, more geographically constrained Ceratitis species.

Introduction

Overcoming environmental challenges is the first of several potential barriers determining whether a species becomes established, naturalised and, ultimately, invasive (Richardson & Pysek, 2006). Upon introduction to a novel environment, a species may be able to persist over short timescales either by having greater resistance to climate conditions, or by mounting a rapid response to these extremes and thereby avoiding potential detrimental effects. As such, short-term phenotypic plasticity could be an important mechanism enhancing survival of individuals upon a change in their environment (reviewed in Chown & Terblanche, 2007; Whitman & Ananthakrishnan, 2009). Indeed, acclimation responses to thermal extremes, which are a form of phenotypic plasticity, have been demonstrated to be a significant component of field fitness in insects (Kristensen et al., 2008) and may, in turn, be related to the habitat and the thermal variability a species experiences (Hazell et al., 2010). Phenotypic responses may also play a significant role in the immediate survival of alien species to novel environments (Lee et al., 2002; Chown et al., 2007; Slabber et al., 2007), without which it is unlikely that the species will become established (and see Hazell et al., 2010). Over longer timescales, rapid evolutionary adaptation (i.e. natural selection for novel phenotypes) may also aid in the naturalisation process (Huey et al., 2005; Lee et al., 2007), unless genetic constraints play a significant role (Gilchrist & Lee, 2007; Kellermann et al., 2009).

The Mediterranean fruit fly Ceratitis capitata has a detailed history of invasion success both in the New World and other parts of the world (e.g. Carey, 1991; Malacrida et al., 2007). The successful establishment of C. capitata outside its natural range probably stems from several physiological, morphological, and behavioural adaptations that enable them to survive in different habitats. Intrinsic factors that contribute to the species invasiveness probably include rapid generation times, polyphagy, and host-plant switching, while extrinsic factors likely include increased propagule pressure caused by repeated species introductions (Carey, 1991; Malacrida et al., 2007). Ceratitis species arose from East Africa (Baliraine et al., 2004) and C. capitata has become established in many countries worldwide while C. rosa has remained largely restricted to Africa (DeMeyer et al., 2008). In South Africa, C. capitata is widely distributed throughout agro-ecosystems, while C. rosa only occurs in cooler, wetter parts of the country (DeMeyer et al., 2008). These patterns suggest that variation in physiological tolerance to climatic stress may contribute to differences in present-day distributions. Moreover, climatic stress resistance is likely to be a significant factor during both active or passive dispersal (Parsons, 1991; Hazell et al., 2010). However, it is generally poorly understood if variation in acute thermal tolerance or its phenotypic plasticity might be important mechanisms facilitating survival of C. capitata upon introduction to novel environments. Indeed, to our knowledge, no studies of Ceratitis spp. to date have considered acute, rapid physiological responses to ecologically relevant temperature variation. Although sub-lethal effects may form an important component of the kinds of stress encountered by insects exposed to novel environments, and several processes are critical to sustained population growth (e.g. reproduction, resource availability and acquisition) during an invasion (discussed in Richardson & Pysek, 2006; and see, for example, Preisser et al., 2008), survival nevertheless serves as a good proxy for determining ecologically relevant variation among species (e.g. Addo-Bediako et al., 2000; Kimura, 2004). Indeed, survival responses probably form a major component of climatic stress resistance, and are used by insects to cope with temperature variation at daily timescales (Meats, 1973; Kelty & Lee, 2001; Kelty, 2007; Overgaard & Sørensen, 2008), but significantly, also upon introduction into new environments (Chown et al., 2007; Slabber et al., 2007; Kristensen et al., 2008; Preisser et al., 2008). Furthermore, survival is a critical first step in the invasion process, without which any further establishment through reproduction is impossible.

Here we test the hypothesis that the highly invasive Mediterranean fruit fly, C. capitata, is aided by phenotypic plasticity of thermal tolerance as opposed to improved basal thermal tolerance, relative to a narrowly distributed congener C. rosa. Furthermore, we ask if the time-course of the rapid responses to thermal extremes might provide further advantages to C. capitata's invasion potential. First, we investigate survival of high and low temperature extremes, and the plasticity of survival to these extremes, in C. capitata and C. rosa. Essentially, we explore whether rapid cold hardening (RCH) or rapid heat hardening (RHH) occurs in each species. Second, we examine the range of temperatures that elicit RCH responses and the time-course of these responses in both species. Specifically, we investigate how long it takes to develop a maximal RCH response, and also the persistence of the enhanced survival effects. Finally, using a population extinction model in various thermal habitat scenarios, coupled with microclimate data from habitat in an area where C. capitata and C. rosa species' distributions overlap, we suggest that variation in RCH responses could translate into differential success in population establishment. Therefore, using a model based on the empirically determined time-courses of these responses, we estimate the potential effects on mortality that variation in RCH might afford C. capitata under natural conditions.

Materials and methods

Study animals and rearing conditions

Fruit flies (C. rosa and C. capitata) were reared in square Perspex™ cages (800 mm³) in the laboratory on a L:D 12:12 h photoperiod, at room temperature (25 ± 1°C) and 65 ± 10% relative humidity. Flies were provided with water, sugar, and hydrolysed yeast (MP Biomedicals, Aurora, Ohio) for food, and with bananas for oviposition. Cultures were first collected from Stellenbosch and Pniel in Western Cape Province of South Africa between December 2006 and March 2007 (summer months). Fruit fly colonies have been in culture for ∼2 years and once every month during summer, wild-caught flies were added to the colony to maintain genetic similarity to wild populations. The cultures of Ceratitis (10 000–12 000 individuals) were held in multiple cages and flies were randomised between cages to avoid inbreeding depression. All cages were held at similar low densities to avoid stressful crowding effects that might affect thermal tolerance estimates. All flies used in thermal tolerance experiments were of a similar age (24–48 h old) and had access to food and water ad libitum, but were of mixed genders since sex does not appear to affect thermal tolerance in either species (Nyamukondiwa & Terblanche, 2009).

Thermal tolerance and rapid thermal responses

To determine survival at an acute temperature, and whether C. rosa and C. capitata rapidly cold- or heat-harden (RCH or RHH respectively), standard protocols were followed (Fig. 1) following Terblanche et al. (2008). In preliminary trials, we first assayed for upper and lower lethal temperatures that cause 70–100% mortality in both species using 2 h exposures (termed the ‘discriminating temperature’). For all hardening assays, five replicate 60 ml vials of 10 insects each were placed in a growth chamber at 25°C for 30 min, after which flies were exposed to a range of temperatures in programmable water baths (Grant GP200-R4, Grant Instruments Inc., Cambridge, U.K.) for 2 h before plunging vials containing flies directly into water baths set at lethal (discriminating) temperatures. After 2 h at the discriminating temperature (41°C for high temperature responses, −5°C for low temperature responses), flies were returned to 25°C for 24 h before scoring survival. In all cases, five replicate vials of 10 flies per vial were used as handling controls. These control flies were sorted into vials, placed at normal rearing temperatures (25°C) in a climate chamber for 30 min, then taken out of the climate chamber, handled for similar duration and with similar vigour to treatment flies, and placed back into climate chambers for 2 h. Next, control flies were exposed to discriminating temperatures for 2 h (at the same time as treatment flies) and then returned to climate chambers at 25°C for 24 h before scoring survival. Treatment and control flies had access to food and water during the recovery period. Survival was defined as a coordinated response to mild stimulation (e.g. prodding) or normal activities (e.g. mating, walking and flying).

Figure 1

Schematic diagram of experimental protocols used to induce rapid hardening responses in Ceratitis spp., following Terblanche et al. (2008) (see Materials and Methods for full description of treatments). (a) Rapid heat hardening, (b) rapid cold hardening.

Since there were no pronounced RHH effects following a 35°C pre-treatment for 2 h, we also investigated if a 36°C pre-treatment for 1 h would induce a hardening response. RCH responses were more pronounced in both species than RHH. Consequently, we only further explored time-courses of RCH responses. A range of pre-treatment temperatures (0–35°C at 5°C increments) were explored for the potential to improve low temperature survival (2 h at −5°C) (following Lee & Denlinger, 1991).

Time-course of rapid thermal responses

To determine how long it takes for C. rosa and C. capitata to develop a full hardening response (i.e. maximum RCH potential or maximum survival), flies were pre-conditioned at 5°C for different durations (15, 30, 60, and 120 min) before subjecting them to a discriminating temperature of −5°C for 2 h. In all cases, five replicate 60 ml vials of 10 insects each were used. A control batch of five replicate 60 ml vials of 10 flies each were taken directly to −5°C for 2 h without pre-treatment and survival was scored after 24 h.

We then investigated how long the full RCH response lasted after pre-conditioning. Flies were cold-hardened by exposing them to 5°C for 2 h and then returned to an environmental chamber at 25°C. Subsequently, at various time intervals (0.5, 1, 2, 4, 8, and 16 h) after hardening treatment, survival was tested following the above protocol.

Microclimate data and population extinction model

Shaded microclimate temperatures were recorded from an orchard (Lakenvlei farm, Ceres, Western Cape, South Africa; 33.34°S, 19.57°E; 1046 m a.s.l.), where both C. rosa and C. capitata are typically found, using Thermocron iButtons (0.5°C accuracy; 15 min sampling frequency). The iButtons were located in the centre of an apple tree at 1.25 m above the ground and a ca 3 month recording was obtained during late March to early June 2009.

Following this, we derived a theoretical model of the effects of RCH on the time to extinction for a finite population of non-reproductive, adult individuals of both species given the same variable, thermal opportunities. Simplified temperature variation for the population extinction model was created by generating sine waves of different amplitude and constant mean temperature (detailed below), and fitting the empirically derived survival curves of RCH for both species to these simulated temperature data. Thus, population size was simulated for at least 60 consecutive days to determine the influence that typical C. capitata or C. rosa RCH responses might have on the population survival over time. The best-fit non-linear (polynomial) equation was obtained by curve-fitting procedures in TableCurve2D software and comparison of Akaike weights to select the most likely model with the least number of model terms. These equations describing the proportion of survival after different durations of exposure to a RCH-inducing temperature (determined from the time-course experiments) were calculated and used for population extinction model simulations at three levels of temperature variation (see eqns 1–4). For C. capitata, the time-course describing the initiation of RCH is

(1)

and the termination of RCH is given by

(2)

For C. rosa, the time-course of initiation of the RCH response is

(3)

and termination of the response is described by

(4)

where S is survival (percentage of population) and a function of x the time (in minutes) kept at temperatures that induce RCH, or time (in minutes) at temperatures that are likely to result in a loss in RCH. For simplicity, we model survival as a function of time below or above a given threshold (10°C) for initiation and termination responses, respectively. Specifically, we assumed that RCH responses were triggered at temperatures <10°C, as suggested by the experimental results, and that these responses proceed in a time-dependent manner thereafter. Similarly, we assume that RCH is terminated above 10°C and thereafter proceeds in a time-dependent manner. A more realistic, although more complex, model could incorporate the non-linear relationship between time and temperature (e.g. Regniere & Bentz, 2007) across a wider range of ecologically relevant conditions and acclimation states, but data collection and model simulations would be logistically more challenging. Temperature simulations were undertaken for a total of 1440 points per day at a 1-min resolution. For the 5% simulation, 72 data points (= minutes) fall below 10°C, 288 min for 20%, and 576 min for 40%. Each species started with a similar population size of 1000 individuals and the population was considered extinct when it reached <0.5% of its original size.

Statistical analyses

To examine the effects of RCH and RHH on fruit fly survival, treatment groups were compared using a generalised linear model (GLZ) assuming a binomial distribution and a logit link function in SAS statistical software (Proc Genmod), with corrections for overdispersion (following, for example, Marais et al., 2009). In all cases, treatments were compared with controls from the same experiment only. Similar GLZ analyses, but including the categorical, ordered effect of time, were carried out to determine the time-course of rapid thermal responses in these two species. Tukey–Kramer's post hoc tests were used to identify statistically homogeneous groups. Microclimate data were analysed for the number of potential RCH events following methods outlined in Sinclair (2001) with 10°C as a threshold and 2 h duration as a minimum time for an event.

Results

Thermal tolerance and rapid thermal responses

A 2 h hardening at 35°C did not significantly improve survival during a 2-h exposure at 41°C in C. rosa (χ² = 8.78, d.f. = 2, P = 0.124; Fig. 2a). However, this treatment (2 h at 35°C) marginally improved survival in C. capitata at 41°C (χ² = 7.90, d.f. = 2, P = 0.0192; Fig. 2a). Pre-treatment of flies at 36°C (1 h) significantly altered survival during a 2-h exposure to 41°C in both C. rosa (χ² = 7.67, d.f. = 2, P = 0.022) and C. capitata (χ² = 15.37, d.f. = 2, P < 0.001). However, there was a marginal increase in C. rosa and a decrease in C. capitata survival (Fig. 2b).

Figure 2

Effects of (a) a 2-h pre-treatment at 35°C (rapid heat hardening) on high temperature survival (2 h at 41°C) of adult Ceratitis rosa and C. capitata; (b) a 1-h pre-treatment (rapid heat hardening) at 36°C on high temperature survival (2 h at 41°C) of C. rosa and C. capitata; (c) a 2-h pre-treatment at 10°C (rapid cold hardening) on low temperature survival (2 h at −5°C) of adult C. rosa and C. capitata; and (d) a 2-h pre-treatment (rapid cold hardening) at 5°C on low temperature survival (2 h at −5°C) of C. rosa and C. capitata. Values are means ± 95% CLs. n = 50 per group. NS, not significant. *Significant difference at P < 0.05.

Pre-treatment of flies for 2 h at 10°C improved survival at −5°C by ∼50% in C. rosa (χ² = 4.08, d.f. = 2, P = 0.013; Fig. 2c) and by ∼80% in C. capitata (χ² = 6.89, d.f. = 2, P = 0.032; Fig. 2c). Furthermore, following pre-treatment for 2 h at 5°C, there was a significant increase (80–90%) in survival at −5°C in both C. rosa (χ² = 31.34, d.f. = 2, P < 0.0001; Fig. 2d) and C. capitata (χ² = 22.86, d.f. = 2, P < 0.0001; Fig. 2d). However, the magnitude of the hardening effects were similar in both species and in both experiments (Fig. 2c,d).

Pre-treatment temperature significantly affected the development of RCH in both C. rosa (χ² = 50.67, d.f. = 9, P < 0.0001) and C. capitata (χ² = 71.94, d.f. = 9, P < 0.0001) (Temperature effect, Table 1). Two hours exposure at 0, 15, 20, 25, 30, and 35°C did not improve survival at −5°C in C. rosa (Fig. 3a). Only pre-treatment at 5 and 10°C elicited a RCH response in this species (Fig. 3a). In C. capitata, maximum survival was achieved only following pre-treatment at 5 or 10°C for 2 h (Fig. 3b). However, pre-treatment at 0, 15, and 35°C also resulted in significant increases in survival relative to the control flies (Fig. 3b), although survival following these pre-treatments did not exceed 50%.

Table 1. Summary of results for species effects from three thermal tolerance experiments.

Experiment	Effect	d.f.	Wald χ^2	P
Temperature variation effects
	Temperature	8	82.48	<0.0001
	Species	1	0.00	0.9997
	Temperature × species	8	9.73	0.2848
Duration to achieve RCH
	Time	4	61.40	<0.0001
	Species	1	9.72	0.0018
	Time × species	4	6.87	0.1427
Duration RCH persists
	Time	7	273.04	<0.0001
	Species	1	179.31	<0.0001
	Time × species	6	23.43	0.0007

• In all cases, survival was compared among treatments using a generalised linear model (GLZ) assuming a binomial distribution and a logit link function in SAS (Proc Genmod) with corrections for overdispersion. d.f., degrees of freedom; χ², chi-square statistic. Temperature variation effects = the effect of different pre-treatment temperatures on survival at −5°C; Duration to achieve RCH = effects of pre-treatment duration at 5°C on survival at −5°C; Duration RCH persists = time taken for increased survival (determined at −5°C) after a 5°C pre-treatment to return to control levels at 25°C. For full details, see Materials and methods.

Figure 3

Effects of pre-treatment temperature (2 h exposures) on the ability to survive 2 h at −5°C in (a) Ceratitis rosa and (b) C. capitata. Control = pre-treatment for 2 h at 25°C. Values are means ± 95% CLs. n = 50 per group. Tukey–Kramer's post hoc tests were used to separate statistically homogeneous groups at P = 0.05. Group means with the same letter are not significantly different.

Time-course of rapid thermal responses

Duration of hardening at 5°C significantly affected the development of a RCH response in both C. rosa (χ² = 35.70, d.f. = 5, P < 0.0001) and C. capitata (χ² = 54.13, d.f. = 5, P < 0.0001) (time effect, Table 1). Overall, C. capitata had significantly higher survival in this experiment [species effect, Table 1; C. capitata: 58.5 ± 25.2%, C. rosa: 38.5 ± 38.2% (mean ± SD)]. However, there was no time × species interaction effect (Table 1). In C. rosa, up to 30 min hardening at 5°C did not significantly improve survival during 2 h exposure to −5°C (Fig. 4a). In C. capitata, there was a significant increase in survival (40–50%) at −5°C for 2 h after only 15 min hardening at 5°C (Fig. 4b), which was significantly greater survival than in C. rosa at the same time point (χ² = 4.42, d.f. = 1, P = 0.0355). However, in both species 80–100% survival was achieved following 2 h pre-treatment at 5°C indicating RCH (Fig. 4a,b).

Figure 4

Time course of rapid cold hardening (RCH) in Ceratitis capitata and C. rosa. Time to induce a full RCH response (100% survival) in (a) C. rosa and (b) C. capitata and duration that the RCH response lasts in (c) C. rosa and (d) C. capitata. Control = no pre-treatment. Values are means ± 95% CLs. n = 50 per group. Tukey–Kramer's post hoc tests were used to identify statistically homogeneous groups. Means with the same letter are not significantly different.

Time after a 2 h long 5°C pre-treatment significantly affected survival at −5°C in both C. rosa (χ² = 63.51, d.f. = 8, P < 0.0001) and C. capitata (χ² = 48.43, d.f. = 8, P < 0.0001). RCH was lost after only 30 min at 25°C in C. rosa (Fig. 4c), but in C. capitata, survival at −5°C remained significantly higher than controls for up to 8 h after 5°C pre-treatment (Fig. 4d). Overall, average survival across all time treatments was higher in C. capitata (74.3 ± 28.6%) compared with C. rosa (26.6 ± 34.9%).

Microclimate data and model simulation

The number of temperature events that fell below 10°C for at least 2 h and that therefore might result in RCH responses were determined from a field season in 2009 (Fig. 5a). In an 86-day recording period during the late austral summer, 63 potential RCH events occurred in the field setting.

Figure 5

(a) Microclimate temperature recording from an orchard (Lakenvlei farm, Ceres, Western Cape, South Africa) where both Ceratitis rosa and C. capitata are typically found. Shade temperature was recorded using Thermocron iButtons (0.5°C accuracy; 15 min sampling frequency) in the centre of an apple tree at 1.25 m above the ground. From experimental results, temperature events that fall below 10°C (stippled horizontal line) for at least 2 h can induce a rapid cold-hardening response. (b) Three simulated temperature scenarios used to model population extinction at different levels of temperature variability. The 40% model contained temperatures that spent 40% of their time <10°C. The 20% and 5% models contained temperatures with 20% and 5% of the time <10°C. (c) Results from the population extinction model simulation exploring differences in RCH responses between C. rosa and C. capitata at the three simulated temperature variation conditions.

The population extinction model showed that time to extinction for C. capitata is 7.5 days at a 40% temperature variability scenario, 4.5 days in the 20% scenario, and 3 days in the 5% scenario. By contrast, C. rosa population times to extinction were 2.5 days at 40% and 20% variability, and 2 days in the 5% simulation (Fig. 5b,c).

Discussion

The major physiological factors contributing to invasion success once an insect species is introduced into a novel environment, typically include either increased environmental tolerance or increased plasticity of environmental tolerance of the invasive species (Lee et al., 2002; Richardson & Pysek, 2006; Chown et al., 2007; Slabber et al., 2007). However, it is less well documented that the time-course of plastic responses of thermal tolerance can also have marked effects on differential survival among populations or species. When estimated as proportion surviving at acute high or low temperatures, both the widespread C. capitata and more restricted congener C. rosa have similar levels of basal thermal tolerance (although see Nyamukondiwa & Terblanche, 2009). In addition, these lethal temperatures are similar to those reported for other tropical–temperate insects (e.g. Addo-Bediako et al., 2000; Kimura, 2004; Hazell et al., 2010) suggesting that variation in basal tolerance may not be a major factor contributing to C. capitata's invasion success.

At high temperatures there was significant variation in heat hardening effects between species, but the magnitude of plasticity was relatively low. However, this lack of hardening effect may be partly a consequence of the exact experimental protocol used, as it has been previously shown that including a recovery period of 30–60 min at an intermediate temperature (20–25 °C) can elicit stronger responses (Hoffmann et al., 2003; Chown & Nicolson, 2004). RHH has been observed in a number of Drosophila species (Chen et al., 1991; Loeschcke et al., 1997; reviewed in Hoffmann et al., 2003) and recently also in C. capitata (Kalosaka et al., 2009), but to our knowledge, such high temperature responses have not been documented in C. rosa. The physiological mechanisms underlying RHH involves the production of heat-shock proteins (HSPs) which can act as molecular chaperones by preventing accumulation of structurally damaged proteins (reviewed in Sørensen et al., 2003; Hoffmann et al., 2003). Consequently, HSPs are thought to be a significant component of field thermal tolerance in Drosophila (Feder et al., 2000; Overgaard et al., 2010) and also other insect species (e.g. Rinehart et al., 2006).

By contrast, substantial RCH responses were found in C. capitata and C. rosa, which were comparable in magnitude to those reported from other tropical fly species to date (e.g. Lee et al., 1987; Coulson & Bale, 1990; Kelty & Lee, 2001; Shreve et al., 2004 reviewed in Lee & Denlinger, 2010). However, RCH has not been demonstrated previously in any Ceratitis species [although see work undertaken on other Tephritids e.g. Dacus tryoni (Meats, 1973), Bactrocera oleae (Koveos, 2001), and Eurosta solidaginis (Bale et al., 1989)]. The results showed a significant improvement in low temperature survival following 2 h pre-treatment at 5 and 10°C in both C. rosa and C. capitata. Indeed, survival at −5°C for 2 h increased from ∼20% to 80–100% in both species. Experiments investigating which temperatures elicit RCH responses showed that only 5 and 10°C pre-treatments (for 2 h) induce a full RCH response in both C. rosa and C. capitata (see Fig. 3). However, a pre-treatment at 0 and 35°C also significantly improved survival in C. capitata, although a full hardening response was not achieved. High temperature exposures may confer low temperature tolerance (see Chen et al., 1991; Sinclair & Chown, 2003; Rajamohan & Sinclair, 2008). However, high and low temperature could be different stressors, and it has been argued that the mechanisms for high and low temperature tolerance may be fundamentally different (Hoffmann et al., 2003; but see discussions in Chown & Nicolson, 2004; Sinclair & Roberts, 2005). There may be some overlap in these mechanisms, for example through HSP production and membrane phospholipid composition changes (Murray et al., 2007; MacMillan et al., 2009a), and clearly inhibition of apoptosis occurs (Yi et al., 2007), although HSP production and membrane lipid remodelling may not explain the short-term survival increase conferred by RCH (MacMillan et al., 2009b). Regardless, the cross-tolerance observed in this study could translate to a survival advantage for C. capitata relative to C. rosa, as the former species can likely respond not only to low temperatures but also to increasing temperature variation.

The marked RCH responses documented here probably allow C. rosa and C. capitata to track diurnal changes in environmental temperature and may optimise feeding and mating during otherwise unfavourable temperature conditions (Kelty & Lee, 2001; Shreve et al., 2004; reviewed in Chown & Nicolson, 2004). The physiological mechanisms underlying RCH are unclear, but may include accumulation of glycerol, trehalose, polyhydric sugars, and alcohols (Denlinger & Lee, 1998), changes in whole body supercooling points (Czajka & Lee, 1990; Lee et al., 2006) and production of HSPs (reviewed in Hoffmann et al., 2003; Sørensen et al., 2003; Chown & Nicolson, 2004; although see discussion in Sinclair & Roberts, 2005). However, improved supercooling capacity may be of little significance in RCH if the cold shock temperature is much higher than the insect's freezing temperature unless they are correlated in some way (reviewed in Lee & Denlinger, 2010). Nevertheless, a range of species do not adjust thermal tolerance over short time periods. For example, little evidence exists for RCH in adults of the tsetse fly Glossina pallidipes (Terblanche et al., 2008), adults of the Antarctic midge Belgica antarctica (Lee et al., 2006), the sub-Antarctic kelp fly Paractora dreuxi (Marais et al., 2009), larvae of Pringleophaga marioni (Sinclair & Chown, 2003), or adult false codling moth Thaumatotibia leucotreta (Stotter & Terblanche, 2009).

The results of the present study are therefore novel since they show that the time-course of developing a RCH response may also provide a distinct advantage in the invasive C. capitata. Ceratitis rosa showed a significant increase in low temperature survival following a 1 h hardening at 5°C. By contrast, survival of low temperature significantly increased after only 15 min of pre-treatment at 5°C in C. capitata, indicating a quicker RCH response in this species. However, a full RCH response was only realised following 2 h hardening at 5°C in both C. rosa and C. capitata (Fig. 4a,b). Similar RCH responses have been observed in, for example, Sarcophaga crassipalpis, where only 30 min pre-treatment at 0°C increased survival at −10°C, although maximum survival was only achieved following 2 h pre-treatment (Lee et al., 1987).

After the RCH response has been acquired we found further differences in the duration of protection afforded between the two species (Fig. 4c,d). Survival at −5°C for 2 h was significantly higher than control levels up to 8 h after RCH in C. capitata had been returned to 25 °C, although by 16 h post-RCH, all effects were reversed. By contrast, in C. rosa the RCH effects were rapidly lost. After only 30 min, improvements in survival were significantly lower and were indistinguishable from control values (Fig. 4c). Overall, this shows that in C. capitata, RCH occurs more quickly and lasts longer as compared with C. rosa indicating, on average, a survival advantage for C. capitata relative to C. rosa in these experiments. Previous studies have also found that RCH lasts for only a few hours, although this is dependent on the temperature post-hardening (e.g. Meats, 1973). Nevertheless, if S. crassipalpis was returned to 25°C after low temperature pre-treatment, the effects of RCH disappeared rapidly (Chen et al., 1991; see also Meats, 1973; Coulson & Bale, 1990; Czajka & Lee, 1990; reviewed in Chown & Nicolson, 2004). In a broader context, however, asymmetric plastic responses to changes in stressful conditions have been found previously (e.g. Palumbi, 1984; Dekinga et al., 2002) and suggest that rates of acquisition of stress resistance might typically be faster than loss of stress resistance (but see also Murray et al., 2007), although this probably depends in part on the relative costs and benefits of acclimation responses (see discussions in Palumbi, 1984; Deere & Chown, 2006; Chown & Terblanche, 2007). However, such clear-cut differences in the time-course of plasticity in insect thermal tolerance are seldom demonstrated, even among closely related species.

Nevertheless, species comparisons make several simplifying assumptions regarding the evolution of thermal tolerance. For example, one of the species may have had significantly greater basal thermal tolerance in the past which was subsequently lost over evolutionary timescales. Alternatively, C. capitata may have evolved greater plasticity of low temperature tolerance as its range expanded with repeated introductions into novel habitats, rather than having a greater invasion success as a consequence of its present physiological responses. Thus, the direction of causality, and any co-relationship with biogeography, is not clear. Nevertheless, a major question posed by this study is whether or not extant variation in physiological responses may contribute to present-day differences in invasion success. To answer this question, we ran a population extinction model that predicts the time to extinction under various hypothetical temperature scenarios. Except for variation in the persistence of RCH responses, this model assumes that all else was equal among these two species (e.g. thermal requirements for development and growth rate), a simplifying assumption, but one which nevertheless gives an indication of the potential fitness benefits of variation in plastic time-courses of cold-hardening responses. Indeed, the model showed clear-cut differences in population persistence over time with the given variation in time-courses of RCH responses, and suggests that, upon introduction of both species to similar thermally variable environments, this may well constitute a fitness advantage for C. capitata. However, in the simulation in which both species are introduced to a similar thermally variable environment with a relatively high mean temperature (e.g. a tropical, equatorial environment), the variation in RCH time-courses is not likely to make a significant difference for population establishment. However, such a model does not consider the fitness costs of plasticity (see discussions in Kristensen et al., 2008), a major aspect of thermal tolerance plasticity that is poorly understood in insect hardening responses (Chown & Terblanche, 2007). Clearly further work is required to understand the costs and benefits of variation in plastic responses of thermal tolerance in Tephritid flies. Nevertheless, additional circumstantial evidence for this model is provided by microclimate temperature recordings from an area in South Africa where C. capitata and C. rosa distributions overlap. These climate data showed that temperatures eliciting these responses are frequently encountered under agro-ecosystem conditions in a cool, temperate environment. This suggests that RCH may be occurring frequently in the wild in this habitat, and possibly protecting flies against sudden temperature changes, as has been shown in Drosophila under semi-field (Kelty & Lee, 2001; Kelty, 2007) and field conditions (Shreve et al., 2004; Overgaard & Sørensen, 2008), assuming that opportunities for behavioural thermoregulation are similar for both species (e.g. Huey & Pascual, 2009). Therefore, RCH probably affords significant fitness advantages to C. capitata relative to C. rosa under natural conditions.

In conclusion, the results of the present study document significant variation in the time-course of phenotypic plasticity of acute low temperature tolerance in adult C. rosa and C. capitata. Ceratitis capitata rapidly cold-harden faster, have a wider range of temperatures that elicit RCH responses, and have RCH survival benefits that last longer than in C. rosa. These results suggest that plasticity in acute thermal tolerance might be an important mechanism facilitating survival of C. capitata upon introduction to novel environments, particularly in cool, temperate habitats. Broadly, these results suggest variation in physiological responses may aid C. capitata's survival upon introduction to novel thermal habitats.

Acknowledgements

We thank Juanita Liebenberg for fly rearing and Corné van Daalen for mathematical advice. We are also grateful to Brent Sinclair, Ray Huey and several anonymous referees for their constructive comments on an earlier version of this manuscript. This research was financially supported by the Deciduous Fruit Producers Trust (DFPT), Citrus Research International (CRI) and THRIP. Water baths and iButtons were purchased with Sub-Committee B and Faculty of AgriSciences funding support (Stellenbosch University).