The effect of respiratory pattern on water loss in desiccation-resistant Drosophila melanogaster

Because of their small size and large surface area to volume ratio, terrestrial insects face potentially high mass-specific rates of water loss. Physiological studies have demonstrated that specific characteristics of their cuticle, behavior, respiratory system and excretory system limit the rates of water loss (Bradley, 1985; Edney, 1977; Hadley, 1994).

In the case of the respiratory system, water loss is minimized by releasing as little water vapor as possible while still achieving the necessary oxygen uptake and carbon dioxide release. Many insects and other arthropods are known to open and close their spiracles in a clearly periodic pattern. These include blood-sucking bugs (Punt, 1944), moth pupae (Levy and Schneiderman, 1966a,b,c), cockroaches (Kestler, 1971), beetles (Lighton, 1988, 1991), grasshoppers (Hadley and Quinlan, 1993; Quinlan and Hadley, 1993), ants (Lighton and Berrigan, 1995), mites (Lighton and Duncan, 1995), ticks (Lighton and Fielden, 1996a) and solphugids (Lighton and Fielden, 1996b). It has been argued that this discontinuous gas exchange cycle (DGC) serves to reduce respiratory water loss (Schneiderman, 1960; Edney, 1977). When the spiracles are closed, carbon dioxide levels build up within the insect and water vapor is retained. When the spiracles subsequently open, carbon dioxide release, relative to water vapor loss, is maximized.

This hypothesis of water savings has been difficult to falsify, because arthropods that breathe using a DGC tend to use it consistently and without large changes in pattern (Lighton, 1996). This prevents us from determining whether a periodic DGC saves more water than some other pattern of spiracular control (e.g. a constant minimized opening or constant flutter).

Previous work with Drosophila melanogaster has found that selection for resistance to desiccation has resulted in populations of flies which show a decrease in water loss rate and an increase in survival time in dry air compared with their controls (Gibbs et al. 1997; Williams et al. 1997). Flies from these populations also show an increase in the height and periodicity of peaks of CO₂ release for the full survival period as measured by the standard error of a linear regression (SER) through the values of CO₂ release. In the present study, we wished to determine whether there is a direct relationship between respiratory pattern and reduced rates of water loss. To do this, we measured and concurrently from individual D. melanogaster that had been selected for resistance to desiccation (D flies). We examined four respiratory characteristics in detail: ⁠, , the SER of CO₂ release and the ratio . These characteristics are compared in flies that demonstrate a periodic release of CO₂ and flies that show non-periodic release of CO₂. We also examine how these variables are affected when an individual changes from a non-periodic to a periodic pattern of release of CO₂.

The 50 D flies used in this experiment were part of a larger comparison of three selection treatments of Drosophila melanogaster described by Williams et al. (1998). The raw data on and discussed here were also used in the treatment comparison study, but were analyzed differently. The remaining methods for the selection and handling of the flies and the techniques used to measure and analyze respiratory variables are described by Williams et al. (1998). In the present study, we used in one comparison an additional analysis, namely the standard error of a linear regression through the water release values (water SER).

When comparing groups of flies that demonstrated periodic release of CO₂, we performed analyses only on the 18 min recordings that showed the periodic release. Comparisons using these recordings were performed using one-way analyses of variance (ANOVAs) in Excel. We looked for correlations among respiratory characte. rs in .all 50 flies by calculating fly means for .each .variable: ⁠, , the SER of CO₂ release and the ratio. Each respiratory character was measured every 2 h and averaged for each individual fly to provide these fly means. Correlations and associated significance values were calculated using SPSS.

The classic pattern of the DGC consists of an open phase in which the spiracles are fully open, a flutter phase in which they open and close rapidly, resulting in a slight increase in , and a closed phase in which CO₂ release is indistinguishable from zero. The pattern of CO₂ release displayed by our populations of Drosophila melanogaster is generally different from that of a classic DGC. In our populations, CO₂ release does not drop to zero levels in most of our recordings. This non-zero release of CO₂ is punctuated by bursts of release of larger amounts of CO₂ that vary in peak height and frequency. We have used the standard error of the mean of the linear regression (the SER) drawn through the trace of CO₂ release as a means of objectively quantifying the degree of variability in the pattern of CO₂ release.

Because the pattern of CO₂ release in Drosophila melanogaster is so different from that normally considered to be a DGC, we do not generally refer to this pattern as a DGC. Instead, we discuss the effect of selection on the SER of CO₂ release. We did record patterns from two flies that showed CO₂ release dropping to zero between periodic bursts (see Results). This indicates to us that D. melanogaster is capable of a classic DGC, but that it is rarely used.

We have found that the SER of CO₂ release is variable, and we describe these differences as follows. A large SER indicates large peaks of CO₂ release, either in regular periodic bursts or in bursts that are not visibly periodic. We use the term ‘periodic release of CO₂’ to describe a pattern that is visibly periodic with a measurable frequency. Neither a large SER nor periodic release of CO₂ need be associated with values of CO₂ release dropping to zero between bursts.

Among the populations of flies we have examined, periodic release of CO₂ occurs only in the D flies and is not seen in their control or ancestor populations. Periodic CO₂ release is rather rare even within this group, occurring in only 13 of the 50 D flies examined and in only three of the five populations. We found that peaks in the water loss trace were visible only in conjunction with periodic peaks of CO₂ release. Fig. 1 provides examples of periodic release from four flies, with the concurrent recordings. Interestingly, periodic release of CO₂ did not guarantee periodic water loss; only six of the 13 flies demonstrating periodic CO₂ release showed visible peaks of water loss. In Fig. 1A,C, the peaks of water loss are visible; in Fig. 1B,D, they are not. Clearly, bursts of CO₂ release are not necessarily correlated with bursts of water loss in these four flies.

We wished to determine what additional differences might exist between flies in which periodic water release occurred simultaneously with periodic CO₂ release and flies in which this did not occur. In seeking these differences, we measured ninech aracteristics in the CO₂ and water vapor recordings: ⁠, , the SER of and , the frequency of peaks of CO₂ release, baseline levels of and , the volu. me of CO₂ found within the peaks and the percentage of total found within the peaks. Table 1 summarizes these variables in the 18 min recordings from the 13 D flies that showed periodic CO₂ release. While most of these variables were significantly different between the two groups, those describing peak and baseline values of CO₂ are of most interest. When the majority of CO₂ is released within peaks, the interburst levels of CO₂ and H₂O are relatively low. As can be seen in Fig. 1A,C, this low H₂O baseline results in enough contrast between interburst and burst water vapor release for peaks to be seen.

Fig. 2 indicates that the peak height and frequency of CO₂ release is associated with the peak height and frequency of water vapor loss (r²=0.64, P<0.01). When the CO₂ SER increases in flies that show periodic CO₂ and H₂O release, the H₂O SER also increases. At low CO₂ SER values (below 3 μl CO₂ h⁻¹), the H₂O SER in all flies was similar to that obtained from an empty chamber. The H₂O SER in flies with periodic H₂O release was 0.013 μl CO₂ h⁻¹, in flies without periodic H₂O release it was 0.011 μl CO₂ h⁻¹ and in flies without periodic CO₂ release it was 0.010 μl CO₂ h⁻¹. An empty chamber gives an H₂O SER of 0.010 μl CO₂ h⁻¹ (P=0.24).

From the recordings that show peaks of water loss corresponding with peaks of CO₂ release, it is possible to estimate the fraction of evaporative water loss represented by respiratory as opposed to cuticular loss. Such calculations are valid only if there is evidence that the spiracles are completely closed between bursts of CO₂ release. Using recordings from two D flies in which CO₂ release was indistinguishable from zero between bursts, we found that 27 % and 23.8 % of the total water vapor was lost within the peaks of water release associated with open spiracles. This limited evidence leads us to estimate that approximately 25 % of evaporative water loss in the desiccation-resistant D. melanogaster is respiratory.

As discussed above, of 13 D flies that showed periodic CO₂ release, six showed measurable periodic water vapor loss while seven did not. A comparison between these two groups of flies showed no significant difference in (P=0.25) or (P=0.14, see Table 1). We randomly selected 13 additional flies from the 37 D flies that showed no periodic release of CO₂, and used recordings that were matched for the degree of desiccation (determined by the time spent in dry air) to those of the 13 flies showing periodic release. This analysis enabled us to compare statistically a group of D flies that exhibit periodic release of CO₂ with a group without periodic release (Table 2). This comparison indicated that flies without periodic release of CO₂ M had a significantly reduced CO₂ S ER (P<0.001) and H₂O SER (P=0.017), but no difference in (P=0.25) relative to flies showing periodic release. Most importantly, there was no significant difference in between these two groups (P=0.19), indicating that periodic CO₂ release is not related to a reduction in respiratory water loss. In addition, for flies that did show periodic CO₂ release, there was no significant difference in between individuals showing periodic H₂O release and those in which this did not occur (Table 2).

Individual D flies can exhibit marked changes in respiratory pattern from non-periodic to periodic (see, for example, Fig. 1B). Fig. 3 presents data for three flies in which the respiratory pattern changed from non-periodic to periodic. Note that the mean CO₂ SER values are highly elevated during bouts of periodic CO₂ release. No concomitant reduction in is observed in these measurement periods. A correlational analysis of the CO₂ SER versus data presented in Fig. 3 revealed no correlation between these variables for each of the three flies (P=0.07, P=0.35, P=0.10).

To determine whether the strength of the CO₂ SER is associated with water savings, we measured correlations between survival time, ⁠, , the CO₂ SER and the ratio in the 50 D flies from this experiment. Increased .survival time is significantly co.rrelated with increased (r=0.34, P=.0.009.), decreased (r=—0.46, P<0.001) and an increased ratio (r=0.61, P<0.001). Surprisingly, however, survival time is negatively correlated with the SER of CO₂ release (r=—0.31, P=0.013) (Fig. 4).

We also found that is positively correlated with the CO₂ SER (r=0.24, P=0.050), and not negatively correlated, as one might expect. Fig. 4 shows the relationships between CO₂ SER and both water loss rate and survival time.

Previous studies have demonstrated that, following selection for desiccation resistance, D flies show reduced rates of water loss and increased survival time in dry air (Graves et al. 1992; Gibbs et al. 1997). In addition, D flies show marked changes in their respiratory patterns compared with their controls and ancestral populations, as shown by significant increases in the SER of the CO₂ release pattern (Williams et al. 1997, 1998).

In the present study, we have taken advantage of the marked variations in respiratory pattern displayed among the D flies. Some flies show periodic release of CO₂ while others show only an increase in the SER of CO₂ release. Some flies show marked changes in respiratory pattern within a 20 min recording period or between recordings. No previous study of the effects of periodic release of CO₂ on water loss rates has been able to examine a single insect as it modified its respiratory behavior over a short period and without experimental manipulation (Lighton, 1996). Analysis of the variation within a single recording also reduces the uncertainty associated with comparisons between insects or treatments.

The D flies show a variety of patterns of CO₂ and H₂O release. We can utilise this physiological diversity to address whether periodic release of CO₂ is associated with decreased rates of water loss compared with the patterns of non-periodic CO₂ release recorded at other times. We used the relevant time periods from the 13 flies that exhibited periodic release of CO₂ and matched these recordings with those from 13 flies that never showed periodic release. from flies with periodic release of CO₂ was not significantly different. from those without. In addition, there was no difference in between the groups, which precludes the functioning of periodic release of CO₂ to allow a greater propagation of CO₂ to be released with the same amount of water loss.

The pattern of CO₂ release in many insects studied to date includes closed-spiracle phases, in which levels of CO₂ are low or near zero. We found that only half the flies that showed periodic release had low enough levels of CO₂ and therefore of water release during the interburst period to allow us to see concurrent peaks of water loss. The rarity of the periodic H₂O release pattern (only recorded in six flies out of 50) suggests that tight closure of the spiracles between open phases is not a critical trait in desiccation resistance. There was no significant difference in the rates of water loss between the six flies with low interburst levels of CO₂ and the seven flies with high levels. Also, was not significantly different between these two groups. It is thus unlikely that a greater need to release CO₂ resulted in the spiracles being more widely open in the group with a high interburst release.

We also examined recordings of individual D flies that exhibited isolated bouts of periodicity within a pattern of non-periodic CO₂ release. The water loss rates of these flies could be followed in concert with the changes in respiratory pattern. There was no evidence that periodic CO₂ release caused any reduction in the rate of water loss in these flies. In fact, an increase in often occurred within 2 h of the bout of periodicity, although the increase might occur before or after the pattern of CO₂ release changed (Fig. 3).

These studies have permitted us to examine rates of water loss associated with a periodic release of CO₂ among several flies and within a single fly. We conclude from these results that bouts of periodic CO₂ release do not reduce water loss in these insects.

The majority of insects and other arthropods examined for respiratory pattern exhibit a periodic DGC, but this pattern may or may not contain all the elements of the ‘classic’ cycle of open, closed and flutter phases. Our five populations of D flies have been selected for resistance to desiccation and show a change in respiratory pattern compared with their controls. This change is demonstrated by a change in the CO₂ SER, but not always by the introduction of a periodic pattern of CO₂ release. All. five populations show an increased CO₂ SER and decreased , while periodicity was recorded in only three of the five populations. Might it be the magnitude of the CO₂ SER that establishes water savings and not its periodicity? We measured correl.ations. between survival time, ⁠, , CO₂ SER and the ratio in the 50 D flies from this experiment.

The correlations between CO₂ SER and both and survival time were not in the direction that we would expect if an elevated CO₂ SER were to increase desiccation resistance. The peak height and frequency of CO₂ release are negatively correlated with survival time and positively correlated with water loss rate. These findings indicate that not only does an increased SER of CO₂ release tend not to reduce water loss or lengthen survival in the D flies but that it may in fact be detrimental to the fly.

We conclude then that an increased SER is of no help in reducing water loss. If the spiracles must be open to some extent, it is not apparent that opening and closing the spiracles in an exaggerated or periodic DGC saves more water than do other options such as constant flutter or keeping the spiracles open a consistently small amount.

We must add that our experimental system prevents us from drawing conclusions that would be universal among all insect species. Because Drosophila melanogaster do not generally show a clear closed-spiracle phase, we cannot reliably separate respiratory and cuticular water loss. We are forced to infer changes in respiratory water loss by examining respiratory plus cuticular water loss. Also, the models of discontinuous respiration ascribe some water savings to the flutter phase (Levy and Schneiderman, 1966c), when CO₂ release drops to zero and there is an inward bulk flow of air. Because we have no evidence of inward convection of air in our flies, water savings during this portion of a DGC are not likely.

What our experimental model does test is whether the respiratory pattern that is seen in these flies saves water. Our data indicate that, while the SER of CO₂ release has increased in response to selection for resistance to desiccation, it does not seem to reduce water loss.

In the present study, we have shown that no correlation can be demonstrated between a highly periodic pattern of CO₂ release or an increase in CO₂ SER and a reduction in respiratory water loss. What then is the physiological basis for reduced water loss and increased desiccation resistance in the D flies?

Hoffman and Parsons (1991) suggested that a reduction in metabolic rate might be a mechanism for resistance to environmental stresses in general and desiccation stress in particular. However, our results and those of other researchers argue against this as a physiological mechanism for resisting desiccation in our populations of flies (Djawdan et al. 1997; Williams et al. 1998).

Other studies have found differences in the D flies compared with their controls (C flies). Gibbs et al. (1997) have shown that the D flies have an increased somatic water content and decreased rates of water loss compared with the C flies. Williams et al. (1998) indicate that the D flies lose water at approximately two-thirds the rate of the C flies and that their total loss of body water is twice that of the C flies. In other work, D females also showed statistically significant reductions in excretory water loss compared with C females, but these represented a small proportion (approximately 10 %) of total water loss rates (Gibbs et al. 1997).

Differences in metabolic storage may also help to explain increased desiccation tolerance in the D flies. These flies have, in the course of selection for desiccation resistance, increased their glycogen levels and reduced their somatic lipid levels (Graves et al. 1992; Gibbs et al. 1997; Djawdan et al. 1996). Gibbs et al. (1997) estimate that approximately 50 % of the water in D flies is associated with stored glycogen. Because our work does not indicate that a larger CO₂ SER or more periodic release of CO₂ saves water, we are left with the conclusion that factors other than respiration strongly affect water loss.

If increased desiccation resistance in D flies is due principally to increased water storage and reduced cuticular water loss, why are changes in CO₂ SER correlated with selection for desiccation resistance? One possibility is that the changes in respiratory pattern are correlated with changes in locomotory behavior. D flies appear to be more quiescent than C and O flies. Studies using other insect species have found that insects are more likely to show a DGC when inactive. It may be that the increases in CO₂ SER observed in the D flies are associated with behavioral changes in the flies. We are currently conducting studies to examine this hypothesis.

The authors would like to thank M. R. Rose for the use of selected lines of D. melanogaster, and A. Gibbs for the use of his equipment. A. Bennett, A. Gibbs, J. Harrison and R. K. Josephson contributed valuable comments on earlier manuscripts. This work was supported by NSF grant IBN 9723404.