Partitioning variance in a physiological trait: desiccation resistance in keratin beetles (Coleoptera, Trogidae)

Introduction

Variation is a sine qua non of evolutionary physiology. Differences in the state of a given physiological trait, especially among populations and among species occupying different environments, have long been regarded as evidence for the importance of physiological evolution (Garland & Adolph 1991; Schmidt-Nielsen 1997). Indeed, it is widely accepted that one of the major achievements of physiological ecology is the demonstration of such a matching between organisms and their environment (Bartholomew 1987; Bennett 1987; Mangum & Hochachka 1998). This relationship between physiological performance and the occurrence of both populations and species has also been used as evidence in favour of the idea that physiological tolerances are important in determining the geographic distribution of animals (for review see Chown & Gaston 1999).

As the demand for information on the likely impacts of climate change on the distribution and abundance of organisms has increased (e.g. Karieva, Kingsolver & Huey 1993; Hoffmann & Parsons 1997), so the evidence in favour of physiological adaptation and its importance in determining species range limits has come under more careful scrutiny (Davis et al. 1998a,b; Chown & Gaston 1999; Spicer & Gaston, in press). The case for physiological adaptation has seen sharp criticism, and this is true particularly of comparative studies, which form the bulk of the evidence for such adaptation (e.g. Garland & Adolph 1994). Perhaps one of the most significant potential problems identified with regard to earlier comparative studies has been that of non-independence of data in multispecies comparisons.

A variety of authors working in a spectrum of fields has now clearly shown that species cannot automatically be considered independent data points because of phylogenetic non-independence (Felsenstein 1985; Harvey & Pagel 1991; Garland & Adolph 1994). Although the importance of taking phylogeny into account in comparative studies is being increasingly recognized in evolutionary physiology (e.g. Huey 1987; Garland, Huey & Bennett 1991; Garland & Carter 1994; Ward & Seely 1996; Degan et al. 1998; Mangum & Hochachka 1998), the necessity for doing so depends to a large extent on the hierarchical level at which most physiological variation is explained. Obviously, if most variation is found at the species level, then comparative methods incorporating phylogenetic independence at this level may not be required (although population level studies in this instance might benefit from the availability of a phylogeographic analysis). However, if variation were partitioned at a higher taxonomic level (e.g. genus or family), then a comparative method incorporating phylogenetic information would considerably improve confidence in the outcome of the study and other interpretations of the findings (see Harvey 1996; and also Ricklefs & Starck 1996; Price 1998 for additional critical discussion).

To date, few studies have attempted to determine at which hierarchical level a given physiological trait shows the greatest variance, despite the obvious importance of doing so (see also Spicer & Gaston, in press). Although a variety of studies has examined variation in physiological traits at the population, species, and individual levels, few studies have sought to partition variance at more than one or two levels within this hierarchy. In this respect, and unlike life-history theory (e.g. Read & Harvey 1989), and to an increasing extent macroecology (e.g. Gaston 1998), evolutionary physiology lacks explicit understanding of variation in the traits that form the substance of its discipline.

Here, we make a start at addressing this important lacuna. To do so, we examine variation in desiccation resistance within an insect family at the generic, species, population and individual levels. Desiccation resistance was chosen as the trait of interest because it is thought to be important in determining the ability of insect species to survive and reproduce in arid environments (Edney 1977; Hadley 1994; Sømme 1995), because it has been implicated more broadly in the evolution of latitudinal gradients in insect body size (Schoener & Janzen 1968; Cushman, Lawton & Manly 1993), and because it is also thought to be important in determining limits to species’ geographic ranges (Hoffmann & Parsons 1991; Nevo 1995). Keratin beetles (Coleoptera, Trogidae) were chosen in particular because the above arguments have regularly been applied to this group of insects (Scholtz 1980, 1981; Scholtz & Caveney 1988, 1992), and because we have recently investigated the mechanistic basis of their desiccation resistance (Le Lagadec, Chown & Scholtz 1998).

Materials and methods

STUDY SITES AND ANIMALS

The Trogidae is a monophyletic family (Scholtz 1986) represented by two genera in Africa: Trox Fabricius comprises mostly smaller-bodied species (mean body length = 9·8 mm), while those of Omorgus Erichson tend to be larger (mean body length = 14·1 mm) (Scholtz 1980). Adult beetles from both genera and a variety of environments were collected between November 1991 and March 1995 from ungulate carcasses and skins in South Africa. The following species were collected: Omorgus asperulatus (Harold), O. elevatus (Harold), O. freyi (Haaf), O. melancholicus (Fahraeus), O. radula (Erichson), O. rusticus (Fahraeus), O. squalidus (Olivier), O. tuberosus (Klug), O. zumpti (Haaf), Trox consimilis Haaf and T. squamiger Roth. A minimum of two populations of each of O. melancholicus, O. radula, O. squalidus, T. consimilis and T. squamiger were also sampled, and in these cases population collection localities were separated by a minimum distance of ≈ 200 km (Table 1). All beetles were returned to the laboratory within one week of collection and kept at 27 ± 1·0 °C (12L : 12D) with free access to food and water for between 2 and 4 weeks.

Table 1. . Climatic data for the sites at which keratin beetle adults were collected for this study

Results

In those analyses where the effects of body mass were not removed, most of the variance in fresh mass, water content and maximum tolerable water loss was found among genera, with species contributing a lower, though significant, amount. Contributions from the population and individual levels were generally less than 10%, irrespective of whether the analysis included only those species where more than a single population was sampled (Table 2), or all of the species (Table 3). Variance in rate of water loss tended to be partitioned more or less equally among genera and species (Tables 2 and 3), while variance in survival time was partitioned mostly at the species level (Tables 2 and 3).

Table 2. . Distribution of variance in fresh mass, water content, maximum tolerable water loss, rate of water loss and survival time in keratin beetle species where more than one population was sampled. Tabulated values are percentages of the total variance accounted for at each successive level. The individual level includes the error term in the data. The second data series represents equivalent values for the traits once the effects of fresh mass were taken into account. Note that there is no difference for survival time between the two analyses because the regression of this trait on body mass was not significant

Table 3. . Distribution of variance in fresh mass, water content, maximum tolerable water loss, rate of water loss and survival time in all of the keratin beetle species sampled. Tabulated values are percentages of the total variance accounted for at each successive level. The individual level includes the error in the data. The second data series represents equivalent values for the traits once the effects of fresh mass had been taken into account.

Linear least squares regressions of the physiological traits on fresh mass were significant in all cases except for survival time in the analysis incorporating species where more than a single population was sampled (Table 4). Once the effects of body mass had been taken into account, variance partitioning among the hierarchical levels changed considerably. None of the variance in the traits examined was partitioned among genera. Rather, the greatest variance in water content, lipid content and maximum tolerable water loss was among individuals, and in rate of water loss and survival time among species and to some extent populations (Tables 2 and 3).

Table 4. . Results of the linear least squares regression analyses of water content, lipid content, maximum tolerable water loss (MAX), rate of water loss (RATE), and survival time (TIME) for the two sets of analyses undertaken for the keratin beetles (see text for additional details)

Discussion

When body size was not controlled for, variance in water and lipid content, as well as in rate of water loss and maximum tolerable water loss was partitioned in a way similar to that of variance in body size, although this was not the case for survival time. Thus, in four of the five physiological traits examined, most of the variance was accounted for among genera (≈ 50–70%), a somewhat lower quantity among species (≈ 20–50%), and least variance was found among populations and individuals. This situation appears to be different from that in mammals and birds (with the exception perhaps of taxa such as Procellariiformes) where the bulk of the variation in body mass, and many life-history traits, is partitioned among orders (e.g. Read & Harvey 1989). It should be kept in mind, though, that in a study such as this one, inclusion of additional species belonging to more distantly related higher taxa might substantially alter this picture. Be that as it may, variance partitioning of this nature is probably not unusual for beetles (and many other animal taxa), where size variation within families can be marked, but where genera tend to have similar body sizes (see Scholtz & Holm 1985; CSIRO 1991). Therefore it appears that when using the comparative method to investigate variation in both body size and desiccation resistance/tolerance in beetles where size effects are not removed, phylogenetic non-independence should be explicitly accounted for as part of such an analysis (see Harvey & Pagel 1991; Garland, Harvey & Ives 1992 for methods).

Not unexpectedly, a strong relationship was found between fresh mass and water and lipid contents, and fresh mass and the two major mechanisms modulating survival under dry conditions, maximum tolerable water loss (i.e. desiccation tolerance), and rate of water loss (desiccation resistance) (see Edney 1977; Chown 1993; Lighton et al. 1994; Chown et al. 1995 for discussion). However, there was either no relationship or a weak relationship between fresh mass and survival time, the ultimate measure of desiccation resistance/tolerance. This is largely a consequence of the fact that maximum tolerable water loss (MAX) scales more or less as Mass (M)¹, and rate of water loss (RATE) as M¹ or, where more species are involved as M^0·8 (implying that time to maximum water loss (TIME) must scale as MAX/RATE ≈ M⁰ or M^0·2), and the fact that MAX and RATE have opposing influences on time to maximum water loss (Table 5).

Table 5. . Results of a multiple regression analysis with survival (TIME) as the dependent variable, and fresh mass (MASS), maximum tolerable water loss (MAX) and rate of water loss (RATE) as the independent variables. Partial correlations and significance of the independent variables are shown, overall R² for the regression was 0·788

An important consequence of these differences in scaling relationships among traits is that once body mass is controlled for, the level at which most variation is explained differs substantially among traits. Thus for survival time (prior to controlling for body mass), and for rate of water loss and survival time where the effects of body size had been taken into account, most of the variance was found among species ≈ 40–70%), whereas in the case of maximum tolerable water loss and water and lipid contents most of the variance was found among individuals (i.e. interindividual variation and measurement error). Nonetheless, in the analysis involving only those species where more than a single population was sampled, at least 25% of the variance in rate of water loss and survival time was accounted for at the population level. This situation is like that found for at least some life-history variables in mammals (see Read & Harvey 1989), where much of the variation initially found among higher hierarchical levels is found among species once the effects of body size have been taken into account. High species level variation also appears to be characteristic of geographic ranges, at least as far as instantaneous assessments are concerned (Gaston 1998; but see also Jablonski 1987).

In the context of insect water balance this finding is not entirely surprising. In general, insects tend not to show large variation in maximum tolerable water loss when this is expressed as a function of fresh mass or initial body water content (Edney 1977; Hadley 1994). Rather, it is rate of water loss that shows most variation among species, and in some instances populations, occupying different environments (see Zachariassen et al. 1987, 1988; Chown 1993; Hadley 1994; Chown et al. 1995; see also Arad, Goldenberg & Heller 1992, 1993a; Arad et al. 1993b for data on snails). Thus large variation among individuals in water content, lipid content and maximum tolerable water loss, but high variance among species in rate of water loss in these keratin beetles are in keeping with most water balance studies. Of course, this does not mean that methods taking phylogenetic non-independence into account should not be applied at this level where this can be done. Rather, these findings suggest that conclusions of the majority of past investigations of adaptive variation in water balance characteristics of insects, especially those that take body mass into account (i.e. provide mass-specific data), may be accepted with somewhat more confidence than if variance had been partitioned mostly at higher taxonomic levels.

In sum, we have demonstrated that in keratin beetles variance in desiccation resistance and tolerance is partitioned mostly among genera before the effects of body size are taken into account, but mostly among species once this has been done. To some extent the study is likely to have been influenced by differences in sample sizes between species and populations (for discussion see Gaston & McArdle 1994), and the fact that some species show considerably more population level variation than others. Nonetheless, it has shown that considerable insight into the evolution of a physiological trait can be obtained when such a partitioning of variance is undertaken, and that this insight may extend well beyond an understanding simply of the variable of interest. In consequence, we recommend that more analyses of this kind be undertaken.

Acknowledgements

Jaco Klok and James Harrison are thanked for collecting some of the beetles, and E. F. Donkin for providing ungulate skins. Melodie McGeoch reviewed an earlier draft of the manuscript. South African National Parks, Kwazulu-Natal Conservation Services, and Gauteng Provincial Administration are thanked for providing us with access to National Parks and Reserves where some of this research was undertaken. The data for this paper were drawn from the PhD thesis of MDLL. John Spicer and Kevin Gaston kindly gave us permission to cite their book, prepublication. Kevin Gaston and an anonymous referee are thanked for their constructive comments which improved the MS. This work was supported by grants to SLC and CHS from the Foundation for Research Development and the University of Pretoria. The Mazda Wildlife Fund provided logistic support in the field.