Critical thermal maxima in knockdown-selected Drosophila: are thermal endpoints correlated?

Virtually all organisms have evolved physiological, behavioral and developmental programs, as well as life history strategies, for confronting fluctuations in environmental temperature. As a consequence of global warming,both the minima and maxima of these fluctuations are increasing within many habitats (Houghton et al.,2001). A relevant concern of physiological and evolutionary ecologists is the capacity of organisms to deal with such challenges. The extent to which organisms can respond to thermal stress has a `profound effect on evolutionary fitness' (Huey and Bennett, 1990).

Drosophilids are model organisms in which thermotolerance has been studied extensively (Hoffmann et al.,2003) by a variety of methods, including assessment of mortality or survivorship following heat shock (Huey et al., 1991; Hoffmann et al.,1997; Berrigan and Hoffmann,1998; Stratman and Markow,1998; Berrigan,2000; Sørensen et al.,2001; Folk et al.,2006; Rashkovetsky et al.,2006), knockdown time (McColl et al., 1996; Hoffmann et al.,1997; Berrigan and Hoffmann,1998; Berrigan,2000; Sørensen et al.,2001; Kellett et al.,2005), knockdown temperature(Gilchrist and Huey, 1999; Berrigan, 2000; Folk et al., 2006), and locomotor functioning during or following heat shock(Krebs et al., 2001; Zatsepina et al., 2001; Roberts et al., 2003; Newman et al., 2005).

Although an array of traits linked to thermotolerance in flies has been studied, some of the traits may have different physiological bases and thus may not be correlated with each other. This idea was supported by Hoffmann et al. who examined the correlation of different measures of thermotolerance in replicate lines of Drosophila melanogaster selected for increased knockdown time at a high temperature(Hoffmann et al., 1997). They tested the flies for recovery time and survivorship following thermal stress. The flies selected for the highest knockdown times did not recover more quickly or show improved survivorship. In other words, the populations selected for highest knockdown had not evolved enhanced thermotolerance in a broad sense. If traits typically used to gauge thermotolerance lack correlation, the ecological and evolutionary relevance of some of the traits may require re-evaluation.

We are currently using populations of Drosophila melanogasterartificially selected for knockdown temperature as a model for investigating the complexities of thermotolerance(Gilchrist and Huey, 1999; Folk et al., 2006). Knockdown temperature (T_KD) is the temperature at which flies drop out of a heated knockdown column (Huey et al., 1992). This study system comprises four replicate populations that have undergone directional selection for high T_KD,four replicate populations that have undergone stabilizing selection on lower T_KD values, and four replicate control populations. The High T_KD populations have evolved high knockdown thermotolerance (average T_KD=∼41°C), whereas the Low T_KD populations experience knockdown at ∼35°C. The mechanistic underpinnings of the knockdown phenotypes are unclear.

Critical thermal maximum (CT_max) estimations have been used to define the thermal tolerance of vertebrate and invertebrate taxa for over six decades (Cowles and Bogert,1944; Lutterschmidt and Hutchison, 1997). Historically, CT_max is defined as the thermal endpoint at which “...locomotory activity becomes disorganized and the animal loses the ability to escape from conditions that will promptly lead to its death”(Cowles and Bogert, 1944). Lutterschmidt and Hutchison assert that “...CT_max is an excellent index and standard for evaluating the thermal requirements and physiology of an organism.”

The physiological states used traditionally to define CT_max are variable and include the `loss of righting response', `onset of muscular spasms', `heat paralysis' or `heat coma', and even `knockdown' (Lutterschmidt and Hutchison, 1997; Berrigan and Hoffmann, 1998). Here we define CT_max in the flies as the upper temperature at which normal locomotory functions and spiracular control are compromised. The central question addressed in the present study is: are thermal endpoints correlated? Specifically, we question whether selection for high (or low) knockdown temperature has resulted in high(or low) critical thermal maximum for heat paralysis. Exposure to either CT_max or T_KD would presumably result in disruption of normal neuromuscular functions.

Traditionally, CT_max has been estimated by determining body (or ambient) temperature at the onset of one of the physiological states mentioned above (Lutterschmidt and Hutchison, 1997). To estimate CT_max in individual flies we used a more objective method, namely thermolimit respirometry, in conjunction with constant monitoring of activity(Lighton and Turner, 2004). Thermolimit respirometry allows the estimation of CT_maxfor involuntary muscle function, as indicated by loss of spiracular control,whereas activity monitoring allows estimation of CT_max for voluntary muscle function, as indicated by cessation of movement. The loss of spiracular control is evidenced by a distinct alteration in the pattern of CO₂ output, characterized by a dramatic reduction in the variance. In addition, we explore the feasibility of using thermolimit respirometry and activity monitoring to detect T_KD, which would be possible only if the patterns of spiracular and locomotor activity were altered significantly at knockdown.

The history of the original and new knockdown lines is detailed fully elsewhere (Folk et al., 2006). Briefly, in 1992 a subset of flies from a founding population of Drosophila melanogaster (collected by L. Harshmann and M. Turelli in 1991 in Escalon, CA, USA) was used to establish discrete lines. From these lines, six High Knockdown lines (`High T<sub>KD</sub>'), three Low knockdown lines (`Low T_KD'), and six Control lines were established (Fig. 1). These lines were subjected to artificial selection (described below) for 46 generations until September 1997. Selection was resumed for 7 generations in June 2004. In January 2005, we combined flies from all replicate lines within each selection (and control) group and randomly chose individuals to generate new replicate High T_KD, Low T_KD, and Control lines (Fig. 1),referred to as the HN_1-4, LN_1-4 and CN_1-4lines, respectively. We generated the new knockdown selection lines to reduce inbreeding by crossing the original lines within each treatment group, and to create equal numbers of lines within the treatment groups.

The knockdown protocol is detailed fully elsewhere(Folk et al., 2006). To begin a knockdown experiment, ∼1000 flies from each line are released into the inner tube of a Weber column (Weber,1988). The inner tube is surrounded by a water jacket, through which water circulates via a heat pump (Haake-Buchler Inc., Paramus,NJ, USA) immersed in a water bath. The flies are transferred immediately from a 25°C incubator into the column, which is initially heated to 30°C. The temperature setting of the heat pump is then increased to 50°C. Consequently, the inner tube heats up (∼0.4°C min^-1); the flies fall out and are collected at 0.5°C intervals between 32-46°C. After falling out of the heated column, the flies resume an upright posture nearly instantaneously (Gilchrist and Huey,1999). Following knockdown, the flies are anesthetized with CO₂, sexed and counted. The distribution of knockdown temperatures is computed separately for males and females from each line.

The selection regime and fly maintenance are detailed fully elsewhere(Folk et al., 2006). Briefly,after each High T_KD line (HN_1-4) is run through the knockdown column, ∼30% of the flies with the highest T_KD are retained for breeding. For each Low T_KD line (LN_1-4), ∼30% of the flies with T_KD values of ∼35.5-37°C are retained for breeding. After each Control line (CN_1-4) is run through the column, ∼30% of all the flies are randomly chosen for retention and breeding. The selected flies are maintained for 6-7 days at 25°C to ensure remating (Gilchrist and Huey,1999), following which eggs are collected for rearing of the subsequent generation. Selection on adults occurs 4-5 days post-eclosion.

For all respirometry experiments, eggs (50 eggs/food vial; 7 vials/line)from each HN_1-4, LN_1-4, and CN_1-4 line were collected from a subset of adults removed from selection. The flies were reared at 25°C. At 2-4 days post-eclosion, 12 flies (6 males, 6 females)from each line were haphazardly chosen, anesthetized with CO₂,weighed to the nearest 0.001 mg, and placed individually into food vials for 20-28 h. We then haphazardly chose four individuals (2 males, 2 females) from the larger group of 12, and measured CO₂ output and activity of individuals across a temperature range from 30°C to >45°C. All respirometry and activity measurements of flies from a line were made on the same day.

To measure CO₂ output, we used a flow-through respirometry system (Sable Systems International, Las Vegas, NV, USA). The respirometer chambers were constructed from 2-dram, glass shell vials (19 mm diameter×51 mm length, Kimble Glass, Inc., Vineland, NJ, USA), each fitted with a two-holed, solid-rubber stopper (no. 2). A piece of aluminum tubing (∼50 mm length, 4.76 mm i.d.; K&S Engineering, Chicago, IL,USA) was pushed into each hole of the stopper until the end of the tube was flush with the bottom surface. A piece of fine mesh was glued to the bottom surface of the stopper to keep flies from escaping. Space between the stopper holes and the aluminum tubes was filled with silicone sealant (Devcon, Riviera Beach, FL, USA). The two sections of aluminum tubing (∼23 mm length)protruding from the top of the stopper were used to attach tubing (1.5875 mm i.d., PharMed, Cole-Parmer Instrument Co., Vernon Hills, IL, USA), which directed air flow to an RM8 Intelligent Multiplexer (Sable Systems International). A thermocouple (Type T) wire was inserted into the PharMed tubing and threaded into the respirometer chamber, allowing us to record the chamber temperature at 1 s intervals. [The thermocouple wire was connected to a digital thermometer (Omega HH509R, Stamford, CT, USA).] The wire insertion site was sealed with silicone sealant.

An air pump (Topfin, XP-125, Pacific Coast Distributors, Inc., Phoenix, AZ,USA) pulled air into the respirometry system. The air stream was then pushed through three scrubbing columns: (1) silica gel (Type III, Sigma-Aldrich, St Louis, MO, USA), (2) silica gel and (3) Drierite/Ascarite II/Drierite(Drierite™: Fisher Scientific, Pittsburgh, PA, USA; Ascarite II™:Thomas Scientific, Swedesboro, NJ, USA), and then through a mass flow control valve (Side-Trak, Sierra Instruments, Inc., Monterey, CA, USA) at a mass flow rate of 50 ml min^-1, maintained by a Mass Flow Controller (Sable Systems International, Version 1.1). The air stream flowed next through tubing and a coil of aluminum tubing (350 mm length; 3 mm i.d.) held in a temperature-control incubator (DigiTherm Incubator, Tritech Research, Los Angeles, CA, USA), to promote thermal equilibration of the air (see Lighton and Turner, 2004). The temperature-equilibrated air then flowed through the respirometer chamber and into the Li-Cor 7000 CO₂ infrared gas analyzer (LI-COR, Inc.,Lincoln, NE, USA). (Before the air stream reached the CO₂ analyzer,H₂O released by the fly was scrubbed with magnesium perchlorate.)

In order to measure CO₂ output within a temperature-controlled environment, we kept the RM8 Intelligent Multiplexer (Sable Systems International, the RM8 controls the air flow through different respirometer chambers) inside the incubator. We used only two respirometer chambers in each experiment: no. 1 for obtaining the zero-CO₂ baseline, and no. 2 for measuring CO₂ output from a fly. The temperature of the incubator was set initially at 30°C. When the temperature inside chamber no. 2 reached 30±0.5°C, we inserted a fly and allowed it to equilibrate for 5 min, during which time we set zero-CO₂ baseline from the empty chamber (no. 1). After equilibration, we started measuring CO₂ output; and at the same time, we increased the incubator temperature setting to 50°C. The temperature within the fly respirometer chamber increased at an average rate of 0.24±0.02°C min^-1. We continued to measure CO₂ output of each fly until the temperature inside the chamber reached >45°C, well beyond the lethal temperature. ExpeData software (Sable Systems International, Version 1.01) was used to control the respirometry system and for data acquisition.

An AD-1 activity detector (Sable Systems International, Version 2) was used to monitor fly activity in conjunction with the CO₂ data acquisition. The respirometer chamber fit securely into the cradle of the AD-1, which generates an infrared light (880 nm) field. Movements of the fly cause fluctuations in the light field intensity, which are detected and recorded (by ExpeData software) as deviations from a zero-activity baseline. Activity was recorded at 1 s intervals.

We used the methods of Klok et al. to estimate both CT_max endpoints (Klok et al., 2004). CT_max for spiracular function was indicated on each respirometry tracing by a distinctive and abrupt alteration in the pattern of CO₂ output, characterized by a dramatic reduction in the variance (Fig. 2). Using the data analysis functions of ExpeData software, we were able to select the transitional point on each tracing at which the CO₂ output was altered. The CO₂ data were aligned with the temperature data (using the timing of both recordings), allowing us to estimate the temperature at which spiracular function was compromised(CT_max). The CO₂ data were transformed using R, version 2.2.1 (R Development Core Team, 2006) by: (1) correcting for baseline drift, (2)converting from p.p.m. to μl h^-1 (mass flow rate=50 ml min^-1) and (3) normalizing with body mass (μg). CT_max for locomotor function was estimated in a similar manner. Using the data analysis functions of ExpeData software, we marked the point on each activity tracing beyond which activity was no longer apparent(Fig. 2). The activity data were aligned with the temperature data (as above), allowing us to estimate this CT_max.

All statistical analyses were performed using R, version 2.2.1,and are discussed in detail in the Results.

The evolved responses of our Drosophila melanogaster populations to selection for high and low knockdown temperature (T_KD)are described in full elsewhere (Folk et al., 2006). Essentially, the High T_KD lines have evolved significantly enhanced knockdown thermotolerance with an overall mean T_KD that is >5°C above that of the Low T_KD lines, and >4°C above that of the control lines. Populations of D. melanogaster typically have a bimodal distribution for T_KD(Gilchrist and Huey, 1999). We show in Table 1 the mean T_KD values of the upper and lower modes of the knockdown populations (Folk et al.,2006). In response to selection, the HN lines have a significant reduction in the numbers of individuals in the lower mode; conversely, the LN lines have relatively few individuals in the upper mode (see Table 1 legend).

The metabolic profiles of the flies during the temperature ramp resemble those observed in a tenebrionid beetle (Gonocephalum simplex) and in ants (Pogonomyrmex rugosus and P. californicus)(Klok et al., 2004; Lighton and Turner, 2004). During the initial phase of the profile, CO₂ output increases exponentially and reaches a maximum, followed by a short plateau(Fig. 2). The profile then shows a `mortal fall' (Lighton and Turner,2004), during which CO<sub>2</sub> output drops, and oscillates dramatically in some flies. Following the mortal fall, the pattern of CO<sub>2</sub> release is distinctively altered by an abrupt reduction in variance, indicating CT<sub>max</sub> for spiracular control(Fig. 2). At temperatures> CT<sub>max</sub>, the ability to modulate spiracle opening is apparently lost and CO<sub>2</sub> is released continually. The release of CO<sub>2</sub> during this phase, termed the `postmortal peak', may be due to the discharge of bound and dissolved CO₂ or to mitochondrial respiration that continues beyond CT_max(Lighton and Turner, 2004). Interestingly, preliminary observations indicate that the flies seemingly recover when switched to a milder temperature (∼23°C) following 2-4 min into the postmortal peak phase, suggesting that physiological processes are not irreversibly damaged at the onset of the postmortal peak phase. These few minutes may constitute a window of recovery, beyond which death is inevitable if the fly is not moved to milder thermal conditions. Following this final peak of CO₂ release, exponential decay in CO₂output was observed in all flies (Fig. 2).

The critical thermal maximum for locomotor function was estimated from the activity patterns of individual flies and was defined as the highest temperature beyond which movement was no longer detectable(Fig. 2).

The effect of knockdown selection on both CT_maxendpoints, as well as on T_MetMax (the temperature at which maximal metabolic rate occurred) among the HN, LN and CN groups was examined using Model I ANOVA. CT_max for locomotor function did not differ among the groups (F_[2,45]=0.1228, P=0.8848), nor did CT_max for spiracular control(F_[2,45]=0.2122, P=0.8096) nor T_MetMax (F_[2,45]=0.1262, P=0.8817). We used boxplots to summarize graphically the data for the CT_max thermal endpoints(Fig. 3A,B). Homogeneity of variances for CT_max of locomotor function among the groups was examined using a Bartlett's test, which showed that the LN lines had a significantly lower variance than the CN and HN lines for CT_max of locomotor function (P<0.001).

To determine whether the dual CT_max measures are functionally related within individual flies we used least squares-linear regression analysis (Fig. 4). Specifically, we tested whether CT_max for locomotory function of a fly is predictive of CT_max for its spiracular control. The relationship between the two CT_maxendpoints was essentially unity for the HN and CN lines: the slopes of the regression lines were 0.99 (P=1.28×10^-14) and 1.04(P=3.72×10^-7), respectively. The slope of the regression line for the LN group was not significantly different from zero(P=0.1233). The lack of significance is likely due to the exceedingly low variance in CT_max in the LN group.

To test for the effects of selection and sex on the pattern of mass-specific V̇_CO2 (μl h^-1 mg^-1) over a temperature range of 32°C to CT_max (∼41°C), we used a second-order orthogonal polynomial analysis for repeated measures. Both V̇_CO2 and the temperature data were log-transformed to attain a normal distribution. We tested for homogeneity of the second-order orthogonal polynomial coefficients among the sex X selection groups and were unable to reject the null hypothesis of homogeneity of slopes (F_[4,42]=0.0088, P=0.9998). We then fit common second-order orthogonal polynomial coefficients to the sex X selection groups and tested for homogeneity among the intercepts using ANCOVA. Knockdown selection did not have a significant effect on mass-specific V̇_CO2 over the experimental range of temperatures (F_[2,42]=0.0048, P=1.0000), whereas sex had a strong effect(F_[1,42]=32.5884, P=1.042×10^-6). All males had a significantly higher mass-specific V̇_CO2 when heated than the females (Fig. 5).

The effect of selection on mass-specific V̇_CO2 at the critical thermal endpoints, namely CT_max for spiracular control, CT_max for locomotor function and T_MetMax, was tested using Model I ANOVA. V̇_CO2 of the treatment groups did not differ significantly at any of the critical thermal endpoints (CT_max for spiracular control: F_[2,45]=0.4528, P=0.6387; CT_max for locomotor function: F_[2,45]=0.2634, P=0.7696; and T_MetMax: F_[2,45]=0.2501, P=0.7798). Mean V̇_CO2 at each critical temperature for all selection groups is listed in Table 1.

We used Model I ANOVA to test for the effects of sex and selection on Q₁₀ for CO₂ production during the temperature ramp,focusing on sensitivity of metabolic rate to increasing temperature during the exponential phase (32-38°C). When calculating Q₁₀, we corrected for the <10°C change in temperature. Although sex did not affect Q₁₀ (F_[1,42]=0.3399, P=0.5630),selection treatment had a significant effect(F_[2,42]=4.2193, P=0.0214). The LN flies had a lower Q₁₀ than the HN flies (Tukey HSD, P=0.043) and the CN flies (Tukey HSD, P=0.039). Q₁₀ did not differ significantly between the CN and HN flies (Tukey HSD, P=0.999). These data suggest that the thermal sensitivity of metabolic rate in the LN flies was relatively reduced. Q₁₀ values for all groups are listed in Table 2.

The response to knockdown selection of the HN and LN lines has been well-characterized (Gilchrist and Huey,1999; Folk et al.,2006). The HN lines have evolved the ability to maintain normal posture within a heated knockdown column at temperatures ⩾40°C. In contrast the LN lines fall down and out of the column at much cooler temperatures (∼35°C). A primary aim of this study was to estimate the critical thermal maxima for locomotor and spiracular function in the divergent knockdown lines in order to explore the correlation, or lack thereof, between various thermal endpoints. Our data indicate that, although T_KD has diverged between the HN and LN lines, the CT_max endpoints have not shifted in accordance. These findings underscore the importance of the notion that the “choice of indices for study (of thermotolerance) could be critical”(Berrigan, 2000).

The CT_max endpoints for locomotor function and spiracular control of each fly are essentially equivalent(Fig. 4). The thermal ceiling of both the locomotory and spiracular control systems are strikingly similar. We propose that a shared physiological constraint may dictate the upper thermal boundary of the dual CT_max endpoints. Hochachka and Somero (Hochachka and Somero,2002) proposed that all physiological systems participate in setting thermotolerance limits of an organism, but that `weak links' assert the strongest influence. They further propose that the `weakest links' in thermotolerance are cellular membranes. Alternatively, Pörtner proposed that at high temperatures metabolic rate may be limited by inadequate delivery of oxygen to the mitochondria(Pörtner, 2002). Klok et al. provide evidence that this may not be the case in terrestrial insects(Klok et al., 2004). Thermal stress can disrupt membrane homeostasis by altering composition, permeability,ion channel activity, etc. A membrane property that is highly sensitive to thermal stress is synaptic transmission, the thermal impairment of which would compromise both effective locomotion and control of ventilation(Robertson, 2004). The narrow range of the CT_max endpoints in our flies supports the hypothesis of a common `weakest link', such as synaptic transmission(Fig. 3A,B). The LN lines, in particular, had very little variability for CT_max of locomotor function. (Stabilizing selection on a limited range of lower T_KD values appears to have eliminated most of the variability for this trait.) Overall, our findings are suggestive of a mechanistic correlation between both measures of CT_max.

Knockdown temperature has been defined generally as the “upper temperature at which a fly falls from a Weber column”(Huey et al., 2003). In natural populations, the distribution for knockdown temperature is typically bimodal (Gilchrist and Huey,1999). Selection for High T_KD causes flies to lose the lower mode of the distribution, whereas selection for Low T_KD tends to eradicate the upper mode. The simplest explanation for the shift to a unimodal distribution is that we are selecting for enhanced (in HN lines) or diminished (in LN lines) performance of a complex polymorphic trait. We propose otherwise. Specifically, we propose that the upper and lower modes of knockdown may represent altogether different traits. Selection for High T_KD may be targeting CT_max, the thermal ceiling for heat paralysis. Interestingly, both T_KD of the HN flies and CT_max of the HN and LN flies are set at∼41°C (Table 1). If High T_KD is essentially CT_max, then the same mechanisms (discussed above) involved in establishing CT_max would be involved in setting the upper limits of knockdown temperature.

Selection on Low T_KD may be targeting something altogether different, possibly behavior. When we first place the LN lines into the knockdown column during an experiment, the column temperature is moderately high (30°C). As the column temperature rises, the LN flies tend to walk along the surface of the baffles, move to the lower edge, and then drop to a lower baffle. They continue to drop to successively lower baffles,until eventually they drop out of the column. The HN flies do not display these behaviors. They localize within the top half of the column when first transferred to the column and do not tend to travel downwards during knockdown. These behavioral peculiarities have led us to question whether we are selecting for a behavioral phenotype in the LN flies. Specifically, are we selecting for negative phototaxis, positive geotaxis, or escape behavior? If this were the case, it might explain the lack of correlation between T_KD and CT_max in the LN flies.

Historical selection experiments, using D. melanogaster and D. pseudoobscura, generated flies with strong positive or negative geotactic(also phototactic) behaviors (Hirsch and Erlenmeyer-Kimling, 1961; Dobzhansky and Spassky, 1962; Del Solar, 1966). The flies had sufficient genetic variation for these behaviors and responded quickly to selection. Interestingly, two desert species (D. mettleri and D. nigrospiracula) share photonegative behavior, but display contrasting geotactic behaviors (Markow and Fogleman,1981). Toma et al. have identified a number of genes involved in phototaxic and geotaxic responses, providing evidence that they are highly complex polygenic behaviors (Toma et al.,2002). We speculate that our selection lines, in response to selection for high or low knockdown performance, may have diverged in the expression of such behaviors.

The composite metabolic profile during the mounting heat stress was similar among the HN, LN and CN lines (Fig. 5). Yet, as indicated by the Q₁₀ values, the thermal sensitivity of metabolism was significantly reduced in the LN lines(Table 2). We are now presented with a paradox: why do flies selected for a lower knockdown temperature have reduced metabolic thermal sensitivity? Selection for Low T_KD performance may have resulted in photo- and geotactic behaviors (discussed above) genetically correlated to metabolic performance. Selection on phototaxis and geotaxis in Drosophila results in correlated changes in multiple traits, including eye size, testis color, wing venation and mating behavior (Del Solar,1966; Dobzhansky and Spassky,1969). Toma et al. compared gene expression of 250 genes in populations of D. melanogaster selected for negative or positive geotaxis and found that ∼5% of the genes were differentially expressed(Toma et al., 2002). Further testing strongly implicated three candidate genes, whose mechanistic involvement in geotaxis is unclear. We speculate that selection on photo- or geotaxis (via selection on low knockdown) may lead to genetic changes influencing the thermal sensitivity of metabolism. An informative study would be to select on photo- or geotaxis and monitor metabolism, particularly Q₁₀ values during heat stress.

In summary, ambient temperature of small ectotherms, such as Drosophila melanogaster, influences directly all physiological networks. We have explored here the thermal limits of respiratory and locomotor performance of flies. We hypothesized that the critical thermal maxima of these performances would co-evolve with knockdown temperature. The dual CT_maxvalues were strongly correlated with each other and with knockdown temperature in the HN lines, suggesting a shared physiological `weakest link' in thermal sensitivity. The CT<sub>max</sub> values were not correlated with knockdown temperature in the LN flies, leading us to conclude that some other`weakest link', perhaps behavioral, sets knockdown in these populations. Finally, respiratory and gross activity patterns were not altered significantly in the LN flies at T_KD, resulting in an inability to detect T_KD using thermolimit respirometry or activity monitoring.

This work was supported by NSF DEB-0344273 to G.W.G. We would like to acknowledge Jon Lighton, Robin Turner, and Barbara Joos at Sable Systems Intl. for their technical and moral support. We are grateful for the valuable comments provided by Ray Huey and two anonymous reviewers.