Repeatability of standard metabolic rate and gas exchange characteristics in a highly variable cockroach, Perisphaeria sp.

For a trait to be considered an adaptation, natural selection must have been responsible for its origin and must be responsible for its maintenance,or, if one is prepared to blur the distinction between exaptation and adaptation, must be responsible for its maintenance(Coddington, 1988; Baum and Larson, 1991; Ketterson and Nolan, 1999). In turn, for natural selection to take place there are three prerequisites(Endler, 1986; Bech et al., 1999; Dohm, 2002). First, the trait in question must show consistent variation among individuals. Second, there must be a consistent relationship between that variation and variation in fitness. Third, the trait must be heritable. Whilst these assumptions, and particularly that of consistent among-individuals variation, are being increasingly well explored for physiological traits in vertebrates (e.g. Huey and Dunham, 1987; Hayes and Chappell, 1990; Hayes et al., 1992; Chappell et al., 1995; Berteaux et al., 1996; Chappell et al., 1996; Bech et al., 1999; Fournier and Thomas, 1999; Koteja et al., 2000; McCarthy, 2000), the same is not true of invertebrates.

In insects, the genetic variability and heritability of several physiological traits have been investigated, mostly in Drosophila(usually melanogaster) and often in the context of selection experiments (e.g. Parsons,1980; Hoffmann and Parsons,1989a; Graves et al.,1992; Gibbs et al.,1997; Gibert et al.,1998; Hoffmann et al.,2003). Likewise, variation in traits among populations and as a consequence of acclimation has also been well explored, especially for thermal tolerance and desiccation resistance(Hoffmann, 1990; Hoffmann et al., 2001; Klok and Chown, 2003). Whilst these studies provide evidence that adaptation has probably been responsible for variation in thermal tolerance and desiccation resistance (see also Chown et al., 2002), explicit exploration of the assumptions underlying the hypothesis of adaptation remains scarce for most traits. This is especially true of metabolic rate and gas exchange characteristics. The few explicit studies that have been undertaken have generally demonstrated a metabolic response to laboratory selection for desiccation resistance in Drosophila melanogaster, which implies that the conditions for selection must have been met (Hoffmann and Parsons, 1989a,b, 1993; Gibbs et al., 1997; Djawdan et al., 1998; Williams et al., 1998). Nonetheless, it is widely assumed that among-species and among-population variation in whole-organism metabolic traits in most insect taxa is adaptive(for reviews and examples, see Lighton,1996; Chown and Gaston,1999; Addo-Bediako et al., 2001, 2002; Gibbs et al., 2003). Metabolic rate is of particular significance in this regard. Not only is it thought to be closely linked to variation in life history characteristics and body size(Hoffmann and Parsons, 1991; Graves et al., 1992; Kozłowski and Gawelczyk,2002), but variation therein apparently also has a profound influence on broad-scale variation in diversity(Allen et al., 2002).

Adaptive explanations for variation in metabolic rate and the patterns in exchange underlying oxygen delivery and CO₂ removal in insects generally take two major forms. First, variation in metabolic rate is thought to take place in response to either dry conditions, when it is reduced to conserve water, or to short seasons, when it is elevated to enable more rapid development (for reviews and discussion, see Chown and Gaston, 1999; Addo-Bediako et al., 2002; Chown, 2002). Second,alterations in gas exchange patterns are thought to have taken place to effect a respiratory water savings under dry conditions. In particular, it has long been thought that discontinuous gas exchange, which is present in many insect species at rest (Lighton, 1996, 1998), evolved as a means to limit respiratory water loss and that it continues to serve this major function (Levy and Schneiderman,1966; Kestler,1985; Sláma and Coquillaud, 1992; Lighton et al., 1993a; Duncan et al.,2002a). Discontinuous gas exchange is typically cyclic with each cycle consisting of a Closed (C) period, during which the spiracles are tightly closed, a Flutter (F) period, during which the spiracles partly open and close in rapid succession, and an Open (O) period, during which the spiracles are open (Lighton,1996). The principal explanations for the contribution of discontinuous gas exchange cycles (DGCs) to water economy are that spiracles are kept closed for a portion (the C-period) of the DGC thus reducing respiratory water loss to zero, and that a largely convective F-period restricts outward movement of water(Kestler, 1985). Moreover, it has also been argued that there is adaptive variation in the durations of the C-, F-, and O-periods to further reduce water loss. That is, a reduced O-period, and prolonged C- and F-periods are likely to further restrict respiratory water loss (Lighton,1990; Lighton et al.,1993b; Davis et al.,1999; Bosch et al.,2000; Duncan et al.,2002a; Duncan,2003). Whilst several other hypotheses for the evolution and maintenance of DGCs have been proposed(Lighton and Berrigan, 1995; Lighton, 1998; Bradley, 2000), these are also largely adaptive in nature (though for an exception, see Chown and Holter, 2000).

Nonetheless, with the exception of the laboratory selection experiments on D. melanogaster, there have been few explicit attempts to investigate the assumptions underlying these claims for adaptation, which have largely been made on the grounds of comparative studies, of which the majority have not been undertaken in an explicitly phylogenetic context (for discussion, see Chown, 2002; Chown and Gaston, 1999; for recent studies, see Davis et al.,1999; Duncan and Byrne,2000; Addo-Bediako et al.,2001; Gibbs et al.,2003). Whilst such comparative studies are useful, they are not without their problems (e.g. Leroi et al.,1994), and it is widely accepted that comparative work should be supported by more explicit investigations of the extent to which selection is responsible for variation in physiological traits (e.g. Huey and Kingsolver, 1993; Kingsolver and Huey, 1998). We therefore undertook this study to investigate the repeatability of metabolic rate and the characteristics of discontinuous gas exchange cycles in an insect species that not only exchanges gases intermittently, but also shows considerable variation in its gas exchange pattern. We reasoned that if these traits show significant repeatability in this species, then it is likely that repeatability will be even more pronounced in most other insect species, which are generally not as variable (see Chown,2001).

Repeatability of metabolic rate, and cyclic gas exchange and its components were examined in the cockroach, Perisphaeria sp. (Blattodea,Blaberidae). In short-term pilot studies this species showed not only a clear DGC typical of many other insect species, but also several other gas exchange patterns that were not associated with activity, which generally disrupts typical DGC patterns (Lighton,1994). Seventeen females and three males (males of species within this genus are rarely found; Picker et al., 2002) were collected at altitudes higher than 950 m above sea level on Stellenbosch Mountain (33°57′S, 18°53′E) and returned to the laboratory in Stellenbosch. Here they were held for 2 months before experiments commenced to exclude possible variation associated with microhabitat differences (Huey and Berrigan, 1996). Throughout this period and over the course of the experiments, the cockroaches were kept individually in marked Petri dishes in a humidified incubator at 22±1.0°C (on a 12 h:12 h L:D cycle). They were fed dry dog- and horse-feed pellets ad libitum and sliced apple was provided on a regular basis. Subsequent culturing has shown that this species is able to reproduce successfully under these conditions (E. Marais,unpublished data).

Prior to each gas exchange assessment, the individual in question was starved for 24 h to reduce variability associated with specific dynamic action(McEvoy, 1984; Lighton, 1989; Duncan et al., 2002b). Assessments were made during the day only, in a well-lit room, because we were concerned only with discontinuous gas exchange and standard metabolic rate. The species is nocturnal, and at night activity and metabolic rate are high(Fig. 1A). Assessments were also made in dry air because under these conditions a discontinuous gas exchange cycle would seem most likely as a means to conserve water(Quinlan and Hadley, 1993; Duncan et al., 2002b). Each individual was weighed (to a resolution of 0.0001 g), using an analytical balance (Toledo AX504, Mettler, Columbus, OH, USA), and placed into a 5 ml cuvette kept at 20±0.2°C using a water bath (Grant LTD20,Cambridge, UK). Air, scrubbed of water (using Drierite, Krugersdorp, South Africa) and CO₂ (using soda lime) was pushed through the cuvette at a flow rate of 200 ml min^-1 (regulated using a Sidetrack Mass Flow Controller, Monterey, USA) and into a calibrated infrared gas analyzer (Li-Cor Li7000; Henderson, USA) set in differential mode to measure CO₂production. A Sable Systems (Henderson, USA) AD-1 activity detector was used to detect any movement of the cockroach in the cuvette during the experiment,and the output of the detector was fed into the auxillary channel of the Li7000. The AD-1 presents activity as a value between -5 V and 5 V, where 0 V is an accurate indication that the specimen is inactive (for more detail, see www.sablesys.com/ad1.html). Inspection of several individuals confirmed lack of activity detected by the AD-1. To avoid the potential influence of pheromones on the behaviour of individuals, the cuvette was cleaned thoroughly with ethanol after each experimental trial. Each experimental assessment also lasted for at least 3 h(for rationale, see Chown,2001). The data file generated by the Li7000 software was exported, via Microsoft Excel, to DATACAN V (Sable Systems), which was used for initial analysis of the respirometry data (corrected to standard temperature and pressure).

Each individual was assessed five times: twice between 07:00 h and 11:00 h,once between 11:00 h and 14:00 h and twice between 14:00 h and 18:00 h. This was done because circadian patterns in metabolic rate have been found in other insect species (Takahashi-Del-Bianco et al., 1992). Typically, at least 5 days elapsed between each assessment of an individual, and the order of assessment with regard to time of day was randomised. The interval between repeated measures is important because the shorter it is the greater the likelihood that a high repeatability will be found (Chappell et al.,1996; Bech et al.,1999). Although we have no data on lifespan for this species, we have cultured adults for more than a year, and other blaberids are known to have an adult lifespan of several years(Scholtz and Holm, 1985). Therefore, an interval of at least 5 days is appropriate for this species,though perhaps biased somewhat in the direction of higher repeatability. The total time taken for the study was approximately 5 months.

Because of the small number of males available, we generally restricted our analyses of repeatability to females. Somewhat surprisingly, we found four major patterns of gas exchange, of which three were intermittent and cyclic(see Results), and the fourth was continuous. Data from the continuous pattern were excluded because metabolic rate was significantly higher (approximately twofold) during this pattern of gas exchange than during the others[repeated-measures analysis of variance (ANOVA), F(3,41)=6.79, P=0.0008, Tukey's HSD for unequal sample sizes, Table 1], even though the individuals were inactive. In consequence, investigations of the repeatability of gas exchange components were undertaken for each for the three major cyclic patterns, and across the dataset as a whole. The variables investigated were duration (s), CO₂ volume (μl) and CO₂ emission rate V̇_CO2 (μl h^-1) for each period, and mean V̇_CO2 and mean frequency of the cycles. Where the analyses were done across the three cyclic patterns, the Flutter period typical of discontinuous gas exchange was compared with the `Interburst' period associated with the other forms of cyclic gas exchange. In these cases we also included data for males. Repeatability (r) was calculated using the intraclass correlation approach(Berteaux et al., 1996; Falconer and Mackay, 1996),based on analyses of variance and the equations provided by Lessells and Boag(1987). Because variation in body mass affects variation in metabolic rate and DGC characteristics in arthropods (Peters, 1983; Lighton, 1991; Lighton and Fielden, 1995; Davis et al., 1999), and because there was a reasonable range in the body mass of the specimens we examined (females: mass 0.3397±0.0184 g, mean ± S.E.M., range 0.1795–0.4643 g; males: mass 0.2357±0.0425 g, range 0.1793–0.3189 g) the effects of body size were taken into account in a second round of repeatability analyses. Usually, to do this the residuals from the regression of body mass and the characteristic of interest are used(Berteaux et al., 1996; Fournier and Thomas, 1999). Here, this was not done. Rather, in all cases, body mass was included as a covariate in the initial ANOVAs (for rationale, see Freckleton, 2002). Where mass did not explain a significant portion of the variance in the independent variable, r was not determined including mass as a covariate. Confidence intervals for r were calculated using the formulae provided by Krebs(1999). A significant repeatability value of 1 indicates that individuals are perfectly consistent in their performance over time, whereas a non-significant repeatability value,or one of 0, indicates no consistent variation among individuals. In all cases a sequential Bonferroni test (α=0.05) was used to correct table-wide significance values for multiple tests(Rice, 1989).

To further investigate the likely sources of variation in these traits,nested (hierarchical) analyses of variance(Sokal and Rohlf, 1995) were used. This method allows ready identification of the level at which most variation can be explained, and has been used for this purpose in several other studies (Berteaux et al.,1996; Chown et al.,1999; Addo-Bediako et al.,2002). For each of the major gas exchange patterns, variance was partitioned between error nested within (<) trial<time of day<individual<gender. Gender was not used as a level of partitioning in the pulsation pattern because males never showed the pattern. A similar analysis was also undertaken across all three cyclic gas exchange patterns. In the case of frequency and mean metabolic rate, the trial level was excluded because metabolic rate and frequency are calculated across all the cycles,rather than just for each individual cycle as can be done for the characteristics of each of the periods. A sequential Bonferroni correction(α=0.05) was also applied here.

This Perisphaeria sp. showed four major gas exchange patterns at rest (confirmed by inspection and by the AD-1), here termed Continuous(Fig. 1B), DGC(Fig. 1C), Interburst-Burst(Fig. 1D), and Pulsation(Fig. 1E). Metabolic rate did not differ among the latter three patterns, although this was not the case with Continuous gas exchange (see Analyses above). Whilst one individual showed all four patterns over the course of the study, most individuals showed two or three of them. However, neither females nor males preferred a particular pattern (females: log-linear analysis, MLχ ²=65.3, P=0.05, d.f.=48; males: log-linear analysis,ML χ²=7.18, P=0.31, d.f.=6), although males never showed the Pulsation pattern. Time of day made no difference to the patterns that were shown (log-linear analysis, ML χ²=3.94, P=0.68, d.f.=6), and no individuals switched from one pattern to another over the course of a recording.

For most of the characteristics examined here, repeatability was significant and large (Fig. 2,Bonferroni correction did not alter significance values appreciably). Within patterns, repeatability tended to be highest, as might be expected, with values for Burst or O-period characteristics generally above 0.3 (with the exception of Burst duration in the Pulsation pattern when mass was included as a covariate). By contrast, Interburst or C-period characteristics tended to have lower repeatabilities (with the significant exception of emission rate, Fig. 2). Across patterns,repeatabilities were also high for the Burst period and somewhat lower for the Interburst period, with emission rate now having the lowest repeatability. This is not surprising because the three patterns differ in the extent to which individuals close their spiracles. In the DGC pattern the spiracles are held closed, whilst this is generally not the case in the other patterns(Fig. 1). There was consistent among-individual variation in metabolic rate (excluding mass: r=0.51 for males and females, 0.48 for females only; including mass r=0.22 for males and females, 0.29 for females only) and frequency (excluding mass: r=0.25 for females only, 0.31 for males and females; including mass: r=0.29 for females only, 0.35 for males and females) (see Supplemental data, Appendix 1A). In general, repeatabilities tended to decline when mass was included as a covariate, but this was not always the case (Supplemental data, Appendix 1B).

The nested analyses of variance generally bore out our repeatability results (Table 2). Moreover,they provided additional insight into the level at which variation that was not a function of individual identity was partitioned. Thus, it is clear that DGC patterns tended to vary much more between trials (the Trial term in Table 2) than within a given trial (the Error term in Table 2), whilst the converse was true of the Pulsation pattern and, to a lesser extent, of the Interburst–Burst pattern. In this context it is important to realize that the error term includes both error and variation between individual cycles in a particular trial. The nested ANOVAs also revealed that there is generally little variation amongst genders in most of the traits examined here. Although this does not appear to be the case when the analyses are undertaken across all three intermittent patterns, this is solely the consequence of the absence of a pulsation pattern in the males.

The presence of four, or at the least three, very different patterns of gas exchange at rest is unusual for most insects, but perhaps not entirely so for cockroaches. Although most studies acknowledge that there is some variation in gas exchange characteristics (excluding that associated with body size and/or treatment temperature; see Lighton,1991; Davis et al.,1999; Rourke,2000), the majority have not found the range of variation within a single species documented here (for a discussion, see Lighton, 1998; Chown, 2001). The only investigations that have suggested that there might be pronounced variability within a species have been those on Blaberus cockroaches by Miller(1973, 1981), who investigated patterns in CNS firing associated with spiracle control, although he thought that some of the variability was probably associated with activity. Nonetheless, among different taxa a range of periodic gas exchange patterns has been found, varying from purely cyclic, with no spiracle closure, to discontinuous gas exchange cycles of the kind originally documented so carefully for lepidopteran pupae (Punt et al., 1957; Levy and Schneiderman, 1966; Lighton, 1988, 1991; Duncan and Byrne, 2000;Shelton and Appel, 2000, 2001). Our findings for Perisphaeria indicate either that the variety of gas exchange patterns documented in insects does not represent as much of a major difference between taxa as comparative studies might suggest(Lighton, 1998), or that basal taxa such as cockroaches and termites differ substantially from more derived groups.

Despite this variety in gas exchange patterns, repeatability values for metabolic rate, frequency and the other gas exchange characteristics were generally high and always significant when examined across the three patterns that were typical of animals at rest with low metabolic rates. These high repeatabilities were not a consequence of pronounced differences between the genders, with the notable exception of the absence of a Pulsation pattern in males. However, the exclusion of body size variation did tend to result in lower repeatabilities. Although most studies first remove the effects of size variation before examining repeatability, it might also be argued that this should not be done. This is most readily demonstrated in the context of metabolic rate variation. Several models have demonstrated the importance of metabolic rate for body size evolution (e.g. Kozłowski and Weiner,1997), and Kozłowski and Gawelczyk(2002) have clearly shown that the major factors influencing optimal size are the size dependence of production rate (which is influenced by metabolic rate; see Sibly and Calow, 1986) and the size dependence of mortality rate (which could be influenced by metabolic rate; see Chown and Gaston,1999). Thus, it seems much more likely that selection will act on the metabolic rate of an animal of a given size than on the residual variation of that trait once size has been taken into account. McNab(1999) arrived at a similar conclusion, pointing out “...that total units of metabolism are the ecologically and evolutionary relevant units”. This argument can readily be applied to all of the other traits we examined, and indeed, in our view, to most other physiological and life history traits.

Therefore, we can conclude that for most of the characteristics we examined variation among individuals was typically significant, and often considerable. These results provide strong evidence that one of the conditions for considering natural selection an important process in the evolution both of gas exchange traits and standard metabolic rate has been met(Endler, 1986; Bech et al., 1999). They also provide a line of evidence, independent of that of comparative analyses,suggesting that variation in these traits among species and populations might well be adaptive. The only exceptions appeared to be the characteristics of the Closed period (in DGC) and Interburst period (in the other cyclic patterns), where repeatability was generally low. Thus, of the gas exchange characteristics examined, those associated with the Closed and Interburst periods are least likely to be the subject of selection. This finding is in keeping with evidence demonstrating that among species with discontinuous gas exchange cycles it is most often the F- and O-periods that vary in a way consistent with adaptive change (Lighton,1988; Lighton et al.,1993a; Bosch et al.,2000; Duncan and Byrne,2000; Duncan,2003; Chown and Davis,2003).

To date, no other studies have convincingly demonstrated consistent among-individual variation in standard metabolic rate and gas exchange characteristics in insects. Prior to this investigation, repeatability in one or more of these traits had only been examined on two occasions. Buck and Keister (1955) reported, but did not provide the statistics for, analyses of variance, which apparently revealed that among-individual variation in O-period volume in diapausing moth pupae was larger than that within individuals, but that several other characteristics of the DGC showed “about as much variation between different cycles of a single pupa as between pupae”. Much later,Chappell and Rogowitz (2000)reported repeatability of standard metabolic rate and DGC characteristics for two species of longicorn beetles (see also Rogowitz and Chappell, 2000),but included both species in their analysis without distinguishing them,factored out body size before the analyses, and considered their non-significant results a consequence of small sample size. Our work takes these initial, useful analyses a step further and demonstrates that, in general, both standard metabolic rate and gas exchange characteristics are significantly repeatable, so meeting one of the major requirements for selection.

Although repeatability estimates for physiological traits in insects and other arthropods are comparatively rare, our data are in keeping with the work that has been undertaken to date. For example, Chappell and Rogowitz(2000) found r values in the range 0.26–0.57 for DGC characteristics in the longicorn beetles they examined. Our values for DGC characteristics not only span a broader range,but unlike theirs were also all significant. This difference is particularly important in the context of metabolic rate. Their analysis indicated a low and non-significant repeatability (0.38), whilst ours suggested that repeatability of metabolic rate was both higher (0.48–0.51) and significant. Nonetheless, it should be kept in mind that the inclusion of body mass as a covariate makes a considerable difference to the value of r in our analysis,but not to its significance (repeatability declined from 0.51 to 0.21 when the effects of mass were controlled for). Considering other physiological traits,in Melanoplus grasshoppers, repeatability of tethered flight duration varies between 0.6 and 0.7 (Kent and Rankin, 2001), whilst in Rhizoglyphus mites,repeatability of sperm competitive ability is much lower (0.22)(Radwan, 1998).

Our repeatability estimates for metabolic rate in Perisphaeria sp. were also well within the range of values typically found in vertebrates. For example, repeatability estimates ranged from 0.35 to 0.52 in breeding female kittiwakes measured over an interval of one year(Bech et al., 1999) and, in a variety of small mammals and birds, varied between 0.261 in meadow voles measured over an interval of 42 days(Berteaux et al., 1996) and 0.64 in kittiwakes measured over the course of a single day(Fyhn et al., 2001).

In conclusion, we have provided evidence that at least one of the prerequisites for natural selection for metabolic rate and gas exchange characteristics in insects is satisfied, and therefore that variation in these traits might be considered adaptive. Whilst our work does not provide conclusive evidence for adaptation in these traits, when considered in conjunction with selection experiments (reviewed in Gibbs, 1999), and comparative analyses (reviewed in Chown and Gaston,1999), it does make the argument for adaptive variation more compelling than it has been. In the past, investigations of variation in gas exchange characteristics in particular have suffered from an unduly Panglossian approach.

We thank Brent Sinclair and Jaco Klok for assistance in the laboratory and Sue Jackson, Melodie McGeoch and two anonymous referees for helpful comments on an earlier version of this manuscript. Collections were made under Western Cape Nature Conservation Board Permit No. 234/2001. This work was funded by South African National Research Foundation Grant GUN 2053804 to S.L.C.