Heinz-R. Köhler Cell Biology, Zoological Institute, University of Tübingen, Auf der Morgenstelle 28, D-72076 Tübingen, Germany. Tel.: +49 7071 2978864; fax: +49 7071 294634; e-mail: firstname.lastname@example.org
Thirty-eight populations of woodlice (Oniscus asellus, Porcellio scaber) and millipedes (Julus scandinavius) from 28 differently metal-polluted field sites were analysed for their 70-kDa stress protein (hsp70) level. Although ANOVA revealed significant dependence of the hsp70 level on the concentrations of water-soluble lead, cadmium and zinc and the soil pH, each of these parameters accounted for at most 18% of the intersite variability of the stress protein level only. A multivariate model based on multiple regression analysis explained more than 96% of hsp70 variance and revealed both the pollution history of a site (strong metal contamination for more than 70 years) and invertebrate species identity to act as the most important parameters. The model accounted for the observation that most of the populations from long-term polluted sites exhibited comparatively low stress protein levels in response to their own (contaminated) habitats. In contrast, isopods (O. asellus) from a control site were not able to maintain a low hsp70 level when they were exposed to either an artificial metal cocktail or soil taken from one of the contaminated field sites. They did not acclimatize to the exposure conditions within 3 months. We propose that selection of insensitive phenotypes in long-term polluted soils has taken place so as to minimize the stress protein level which, in turn, is indicative of high intracellular protein integrity. Long-term selection for a high hsp70 level to compensate for adverse metal impact was not observed, which suggests that such a strategy may trade off against other fitness consequences. In this context, insensitivity to metal stress involved increased selectivity in food choice and reduced variability in stress response. Multiple regression models showed species-specificity in those abiotic factors which determined (1) high hsp70 levels in sensitive populations as well as (2) low hsp70 levels in insensitive ones. Therefore, abiotic factors can be assigned to act as the main components of selection: lead and cadmium for J. scandinavius and O. asellus, zinc for P. scaber.
The so-called molecular stress response is a universal reaction of all biota to adverse conditions, particularly heavy metals (e.g. Nover, 1984). It comprises the increased production of a set of stress proteins (heat shock proteins, hsps) of different molecular weights (e.g. Sanders, 1993). So far, the best investigated family of stress proteins, hsp70, is induced by a variety of adverse effects on intracellular protein integrity (for molecular functioning, see Gething & Sambrook, 1992; Morimoto, 1993) and has therefore been assigned to integrate effectively overall proteotoxicity. The biochemical role of hsp70 is to stabilize nascent or partly affected polypeptide chains, to render possible a correct transmembrane protein passage (e.g. Pelham, 1986; Pfanner, 1990) and therefore to counteract proteotoxic action. Consequently, every organism which is exposed to proteotoxic environmental stressors like heavy metals either needs to elevate its hsp70 level to limit intracellular proteotoxicity or should avoid adverse impact on protein integrity by other protective mechanisms, as, for example, metallothionein induction (e.g. Dallinger et al., 1997 ). Thus, hsp70 has been used as a biomarker in organisms to indicate whether or not they originated from polluted sites ( Veldhuizen-Tsoerkan et al., 1991 ; Köhler et al., 1992 ; Sanders, 1993).
Numerous examples of the development of metal adaptation mainly in plants (e.g. Antonovics & Bradshaw, 1970; Linhart & Grant, 1996) but also soil invertebrates have been reported (summarized by Posthuma & van Straalen, 1993) that may reduce adverse effects of pollutants on terrestrial ecosystem functioning. With the exception of a few examples where tolerance was inherited ( Frati et al., 1992 ; Donker et al., 1993b ; Posthuma et al., 1993 ), it remains unclear whether adaptation of soil invertebrates to metals involves an acquired or a genetic component; in other words, can adaptation be based on physiological acclimatization only or also on selection of metal tolerance? Although a few studies exist which suggest that hsps are involved in the selection of individuals which are tolerant to stressors like elevated temperature or chemical contamination ( Koban et al., 1991 ; Sanders et al., 1991 ; Veldhuizen-Tsoerkan et al., 1991 ), only two studies consider evolutionary aspects of hsp inducibility in natural populations of soil invertebrates ( Eckwert & Köhler, 1997; Köhler et al., 1999a ). In the present study, investigation of 38 populations of isopods and diplopods from a variety of differently metal polluted sites was used to address the following questions:
1 Is there evidence for the selection of insensitive phenotypes in polluted areas, and which are the important parameters related to the compensation of metal exposure?
2 Do these parameters act independently from one another and is apparent selection on different species based on the same parameters?
3 What is the result of the compensation of long-term metal exposure for the hsp70 level? Is hsp70 subject of selection and either triggered to elevated levels as a result of adaptation or kept to a minimum to avoid possible trade-offs against fitness consequences as detected in Drosophila ( Krebs et al., 1998 )? In this context, we used a mechanistic approach to explain selection.
Materials and methods
Field sites and sampling
Adults of the diplopod Julus scandinavius L ATZEL 1884 (Julidae) and the isopods Oniscus asellus L. 1758 (Oniscidae) and Porcellio scaber L ATREILLE 1804 (Porcellionidae), comprising a total of 38 separate populations, were sampled from 28 mainly deciduous forest sites in south-west Germany ( Fig. 1, Table 1). Isopods with juveniles in their marsupium were omitted from the study. Sampling was conducted during spring and autumn, predominantly on cloudy days, and in the morning to allow collection of a sufficient number of specimens even in those areas where they rarely occurred. Specimens were frozen directly in the field in liquid nitrogen and subsequently analysed for their hsp70 level. Sampling sites ( Fig. 1) were chosen for their comparability with respect to climate and for their differences in soil metal (lead, zinc, cadmium) concentration and pH. They included sites close to industrial areas (site no. 01 in Fig. 1 and Appendix 1), military areas (sites 02, 26–28), sites near a highway (site 04), and a former opencast lead and zinc mine (sites 10–15, centre of the mine at site 14). Also comparatively unpolluted areas (sites 09, 16) were sampled as reference sites. Although pollutants other than those measured may have been present in the soil, high concentrations of them are highly unlikely except for site 02 (manœuvre area), where organic pollutants are suspected to occur.
Table 1. Parameters used in multiple regression models significantly contributing to the explanation of variance in the stress protein (hsp70) level of isopod and diplopod populations.
Table 2. Table Sites, mean metal pollution in soil, soil pH, investigated species, and mean hsp70 level (± SD) of field populations
Due to vertical movements of diplopods and isopods in soil, the exact temperature the animals were exposed to prior to sampling is not known. However, in southern German woodlands, the temperature in the zone of decaying humic material which is the main habitat of these animals does not exceed 10–12 °C even in glades on sunny days. This temperature does not cause an elevated hsp70 level in isopods and diplopods ( Köhler et al., 1992 ; Köhler et al., 1999b ). Furthermore, these animals in summer and in the middle of the day tend to bury themselves in deeper soil horizons or under trunks, according to their temperature optimum. Thus, the influence of temperature on the hsp70 level is likely to be identical across sites.
To compare the potential of contaminated soil to induce stress proteins, a microcosm study was conducted with O. asellus from the control site Mauer (site 16). Adult isopods were kept outdoors in microcosms (polyethylene cylinders, diameter 21 cm, with holes drilled in the bottom) containing soil and partially decomposed leaf litter material either from the other control site (site 09, to change the habitat, but not into a polluted one) or from a heavily polluted site (site 14). There were two replicates of each treatment. The cylinders were covered above with fine-meshed polyethylene gauze and buried in soil in a shady place about 15 km distant from both sampling sites (sites 09, 14). The height of the soil-filled cylinders (26 cm) allowed the animals to choose their preferred habitat temperature. The microcosms were protected from excessive rain by a transparent plastic sheet 30 cm above the upper edges of the cylinders which allowed access of light and air to the isopods. The moisture in the microcosms was adjusted by spraying the leaf litter with tap water from above. After 3 months of exposure (25 May −17 August), all surviving specimens were removed from the microcosms, immediately frozen in liquid nitrogen and a maximum of five per replicate cylinder (i.e. a maximum of 10 specimens per treatment) analysed for their hsp70 level.
Laboratory exposure experiments
To investigate the inducibility of hsp70 to metal exposure, a laboratory experiment using Oniscus asellus from one control site (Mauer, site 16) and from one polluted mine site (Nußloch A, site 12) was conducted. From each population, 120 adult O. asellus (males and nongravid females) were exposed to partly decomposed leaf litter material soaked in an aqueous solution of 1 mg Cd2+ L−1, 10 mg Pb2+ L−1, 50 mg Cu2+ L−1 and 50 mg Zn2+ L−1 (as chlorides, respectively) resulting in a mean concentration of 4.8 μg Cd, 61.3 μg Pb, 261.2 μg Cu and 234.8 μg Zn per mg dry weight of the litter. The isopods were kept in plastic boxes (20 individuals per box) on a moist gypsum base, covered by leaf litter material which was supplemented with a few lime particles. Exposure was maintained for 10 days at 12 °C and a light/dark photoperiod of 12 h/12 h. After 10 days, the contaminated leaf litter was removed from the boxes and replaced by noncontaminated litter from control site 09 which was moistened with distilled water. The experiment lasted another 10 days. Samples were taken at days 0, 3, 6, 10 (exposed to the metal cocktail), 15 and 20, the latter two after 5 and 10 days of recovery, respectively. For both populations, the following parameters were recorded: (1) the relative hsp70 level (n=10 animals using as reference [1.0=100%]: population from site 16, 0 days of exposure); (2) the average dry mass of leaf litter which had been ingested per individual, and (3) O. asellus mortality. Because the calculation of ingested food required compensation for microbial activity during the exposure time, the loss of litter mass in six boxes comprising leaf litter material alone (four boxes with contaminated litter, two with noncontaminated litter) was recorded after 0, 3, 6 and 10 days (contaminated litter) or 5 and 10 days (noncontaminated litter), respectively.
After homogenization in extraction buffer (80 m M potassium acetate, 4 m M magnesium acetate, 20 m M HEPES pH 7.5), the volume of which was adjusted to the animal’s body weight, samples were centrifuged (15 min, 19 100g) and total protein concentration in the supernatant was determined according to the method of Bradford (1976). Constant protein weights (30 μg (isopods) or 100 μg (diplopods) of total protein per lane) were analysed by minigel SDS-PAGE and subsequent Western blotting ( Köhler et al., 1998 ) using a monoclonal mouse antihuman hsp70 antibody (Dianova, Germany). Grey value quantification of the Western blot protein bands took place with densitometric image analysis systems (Herolab E.A.S.Y. and Cybertech CS 2) after background subtraction. The mean grey value of the bands obtained for control specimens (J. scandinavius from site 09, O. asellus and P. scaber from site 16) was set arbitrarily to 1.00 as a standard reference. Relative grey values (rgv) ± standard deviations were calculated for all other treatments.
Replicates of control and exposed specimens were run on different gels to minimize the influence of methodological variation on mean grey value calculation. Although analyses comprise the entirety of the arthropods’ body including, for example, symbionts in the intestinal tract, the respective stress protein level can be assigned to the animals themselves due to the comparatively small mass of any associated organism.
Principal component analysis and model fitting
All mathematical analyses were carried out with the software package JMP 3.1 (SAS Institute Inc., Cary, NC, USA). As a prerequisite, both for regression analysis of the hsp70 level vs. the diverse abiotic parameters recorded for the different field sites and for multiple regression modelling, the independence of the variability of the abiotic field parameters were checked by principal component analysis (PCA). Both the entire data set and the subsets of each investigated species were analysed in this respect.
Subsequently, linear regression models for the following data sets vs. the individual data on the hsp70 level for each specimen (294 data points as a maximum) were analysed by ANOVA: Pb, Zn and Cd concentrations and their respective log10, soil pH, and concentrations of the soluble portions of Pb, Zn and Cd. Significance probabilities (P) of 0.05 or less were considered evidence that there is at least one significant regression factor (the slope) in the model which differs from zero and, consequently, that the variability of the tested X variables (the abiotic parameters in the field) contributes to explain the variability of the response variable Y (the hsp70 level). Because the genetic variability and, consequently, the variability in physiological and biochemical responses to the environment are supposed to decrease in ‘tolerant’ (=insensitive) populations of soil invertebrates ( Posthuma & van Straalen, 1993), the standard deviation of the hsp70 level as a percentage of the mean hsp70 level was also tested vs. the soil metal concentrations by linear regression and ANOVA.
To reveal those parameters which contributed most to a mathematical explanation of the variability of the hsp70 level, multiple regression models were calculated in a stepwise way. For the categorial variables, the dataset was split to make the two groups of levels that most separated the means of the response parameter (the hsp70 level), e.g. a P. scaber group vs. a group comprising both O. asellus and J. scandinavius for the variable ‘species’. For the parameter ‘time >70 years polluted’, only two terms (‘yes’ and ‘no’) existed, but the split dataset for the variable ‘species’ had to be further subdivided into its most separated subgroups (e.g. the O. asellus plus J. scandinavius group into a O. asellus vs. a J. scandinavius subgroup). We started with simple models comprising only one of those variables for which analysis of variance has revealed a significant relationship with the hsp70 level, and went on towards more complex models including more than one variable. To account for possible nonlinear relationships and interactions between the abiotic variables separating the sites from one another on the one hand, and the mean hsp70 level on the other hand, (1) the log10 values of the Pb, Zn and Cd concentrations in soil were included in the list of variables (among the Pb, Zn, Cd concentrations, the soluble portions of Pb, Zn, Cd, the soil pH, the species and the pollution history) and (2) a factorial combination of variables to the degree of 2 was also included in the models. However, the multiplication of two variables which had opposing relationships with the hsp70 level (i.e. the soil pH, negative relationship, multiplied with the log10 of the Pb concentration, positive correlation) were excluded from the models.
The best model for the entire dataset indicated the importance of the parameters ‘species’ and ‘time >70 years polluted’. Furthermore, the principal component analyses showed independence of the parameters for species-related subsets of the dataset. Hence, separate models were calculated for each species and, within these subsets, for the short-time and minor polluted sites; for the >70 years polluted sites, the number of observations (investigated sites) was insufficient to calculate separate models for every species. Models were optimized in a pragmatic way in order to explain as much of the observed variability as possible with the highest number of variables contributing significantly to the model. The level of significance was set at P=0.01.
Since normal distribution of the data could not be assumed for all data sets, significance of two independent means was tested with the Mann–Whitney–Wilcoxon U-test. The significance levels were set at 0.05 ≥ P > 0.01 (slightly significant, *), 0.01 ≥ P > 0.001 (significant, *) and P ≤ 0.001 (highly significant, ***).
Stress protein (hsp70) levels in field populations and their relation to abiotic parameters
After division of the entire dataset into two subsets according to (1) long-term (>70 years) polluted and (2) short-term and minor polluted stands, the hsp70 level was related to log [Pb] in minor or short-term polluted stands, but not to log [Zn] and log [Cd] in these stands ( Fig. 4) nor to any absolute or logarithmic metal concentration in the long-term polluted stands ( Fig. 5). For species-specific analyses, the number of investigated sites was not sufficiently high. A significant trend towards lower variability (SD relative to the mean) of the hsp70 level in isopod or diplopod populations inhabiting areas with increasing (log) Pb concentrations could also be detected for the entire dataset ( Fig. 6) which could not be proven statistically for single species owing to the limited number of data points.
Despite a lack of evidence for dependence of the hsp70 level on absolute or logarithmic metal concentrations other than log [Pb] in the unifactorial regression analysis, a trend towards increasing stress protein levels with increasing metal concentrations in soil plus a shift to comparatively low hsp70 levels in long-term and extremely metal-polluted stands becomes clear when the average hsp70 levels of all investigated populations are plotted upon logarithmic scales of soil lead and zinc concentrations in a three-dimensional diagram ( Fig. 7). Further corroboration of this shift in average hsp70 level between soil arthropod populations taken from low and moderately contaminated areas on the one hand and those residing at long-term polluted sites on the other is provided by multiple regression modelling using the entire dataset (Table 1, model ): the most significant variable (combination) contributing to the best model was the product of pollution history (‘time >70 years polluted’) and species identity. There was a split of the dataset within the model according to these two categorial variables. Thus, the model also seemed to recognize the diverging biochemical response of the abundant populations in the long term polluted areas to their own habitat. Eight of 15 ‘>70 years populations’ had average hsp70 levels which were even lower than those recorded for control populations in control stands ( Fig. 7).
Microcosms and laboratory experiments
The microcosm study supported the field observations on different physiological responses of the diverse populations. Specimens from an O. asellus control population kept on uncontaminated soil in the microcosms for 3 months largely resembled their free-living relatives in the natural habitat with respect to the level of the stress protein, hsp70 (field population 1.00 ± 0.46 (SD), microcosms 1.52 ± 0.94). In contrast, most animals from the same population which were kept on soil material from one of the long-term polluted stands (mine site 14) did not adapt to these conditions nor survive 3 months of exposure, other than a single surviving individual which had a remarkably high hsp70 level (4.38). These results differ strongly from what was observed for the mine site resident O. asellus population (1.38 ± 0.50).
Such differing responses by two divergently pre-exposed isopod populations to metal-contaminated substrate also became obvious in the exposure-and-recovery laboratory experiments. The O. asellus control population ingested artificially metal-enriched leaf litter material at the same rate as unpolluted food ( Fig. 8, top). In contrast, a population from a site in the vicinity of the mine centre was able to distinguish between metal-polluted and clean food, and reduced feeding drastically during exposure time ( Fig. 8, top). Consequently, the hsp70 level rose significantly during exposure time in the site 16 population only, while the mine site isopods maintained a constantly low hsp70 ( Fig. 8, bottom). During the period of recovery, the elevated stress protein level in the control site isopods disappeared 10 days after removal of the contaminated litter, and the hsp70 level in the mine site woodlice even declined to significantly lower values than those obtained for this population in their original (heavily contaminated) habitat ( Fig. 8, bottom). None of the exposed animals died during the experiment.
Using multiple regression models, it was possible to explain the variability of the mean hsp70 level in the field populations to a considerable extent. The variables contributing significantly to these models are listed in Table 1. It was possible to explain 96.48% of the total variability of mean stress protein levels in all investigated arthropod populations (excluding the one from the presumably multiple polluted manoeuvre area 02) using 25 variables or factorial combinations of them (Table 1, model ). Eleven variables (or combinations) contributed significantly to the model, from which the species and the pollution history were most important. With a single parameter only, it was not possible to explain more than 26.34% of mean hsp70 variability (Table 1, models –). In contrast, more complex models drawn up for subsets of the recorded data explained almost 100% (J. scandinavius, Table 1, model ) or slightly more than 84% (both isopod species, Table 1, models  and ) of the variability in hsp70 level means. Only 1–4 variables (or factorial combinations) contributed significantly to these models, and it was evident that soil Pb and Cd concentrations (or their log values) were most important to explain the variability in hsp70 observed for the diplopod, J. scandinavius, and the isopod, O. asellus. This could be shown both for all sites (except 02) (Table 1, models  and ) and for the respective subsets of the minor and short-term polluted sites alone (Table 1, models  and ). The soil (log) Zn concentration was most important to explain the stress protein level variability recorded for the other isopod species, P. scaber (Table 1, models  and ).
Selection of insensitive phenotypes
It is indisputable that isopods and diplopods can inherently tolerate exposure to metals to some extent. This ability is mainly based on the intracellular precipitation of metal ions as phosphates or organic complexes in membrane-surrounded granules (spherites), which withdraw them for the most part from the animal’s metabolism (summarized in Hopkin, 1989). Nevertheless, high concentrations of cadmium, lead or zinc have been shown to surpass this detoxification system and to result in significantly reduced growth or moult rates in P. scaber ( Drobne & Štrus, 1996; Donker et al., 1998 ; Witzel, 1998) and may therefore exert a strong selection pressure toward the evolution of mechanisms to overcome the effect of these ions. It was an aim of this study to reveal the existence of adapted phenotypes in metal-contaminated habitats, and several of our results indicate selection of stress-insensitive phenotypes in some of the investigated populations. Although the stress response level covaried with certain abiotic parameters (soil pH, soluble metal fractions), models using these variables alone were only able to explain a few per cent of the intersite variability in the hsp70 level.
In contrast, the importance of the site-specific pollution history suggests that time is required for the formation of populations indicative of the evolution of stress insensitivity. Statistical models incorporating time explained much variability in hsp70 and, furthermore, suggested additional parameters, metal ions, which comprised the selection of stress insensitivity. Further support for the selection of insensitive phenotypes in the long-term polluted mining area and its surrounding was given by the laboratory and, particularly, the microcosm experiments. In the laboratory, isopods of different origin displayed a divergent behaviour and, concomitantly, different biochemical responses to stress situations posed by metal contamination. Although the differences in food uptake and stress response may result from pre-exposure and may represent after-effects of different acclimation, the long-term microcosm experiment clearly showed that any acclimation of control isopods to the conditions at the mine site was impossible within 3 months of exposure. This finding suggests the involvement of a genetic component to tolerance at least for the tested Oniscus populations. The observed trend towards a decrease in hsp70 variability (relative standard deviations) also supports the notion of the evolution of tolerance ( Posthuma & van Straalen, 1993).
Metal insensitivity was associated with a low stress protein level. However, a low hsp70 level in metal-exposed animals does not necessarily indicate stress insensitivity. Although heavy metals were shown usually to induce stress proteins in isopods ( Eckwert et al., 1997 ) and diplopods ( Zanger et al., 1996 ), the hsp70 level follows a curve with a maximum at rather high but not extreme concentrations. In response to very high metal concentrations, the stress protein level decreases ( Guven et al., 1994 ; Eckwert et al., 1997 ; Pyza et al., 1997 ) due to severe pathological impact upon target organs, such as the hepatopancreas of isopods ( Köhler et al., 1996 ). It is doubtful that a population of pathologically injured animals can maintain itself under field conditions. Indeed, Gräff (1997) and Schill (1999) did not find pathologically damaged organs in free-living animals, even if they derived from heavily metal-polluted stands. That low hsp70 levels of the populations at the long-term contaminated sites arise through histopathological disintegration of tissues can hence be refuted.
Parameters of importance
Interestingly, the main factors besides pollution history explaining both high hsp70 levels in nontolerant populations and low but increasing hsp70 levels with increasing metal pollution in stress-insensitive populations differed between the investigated species. For O. asellus and J. scandinavius, cadmium and lead were of greater importance, while zinc was associated with P. scaber response to metal pollution. Zinc appears responsible for the absence of P. scaber in a number of sites in the surrounding of a metal smelting works close to Avonmouth, south-west England ( Hopkin, 1990). Hopkin & Hames (1994) argued that zinc amongst a metal ‘cocktail’ will be the limiting factor for P. scaber, if the zinc to cadmium ratio is >10:1 and the zinc to lead ratio is >20:1. The latter conjecture was not confirmed by our investigations since in none of our sites was the zinc concentration 20 or more times higher than the lead concentration. Nevertheless, zinc was found to be significantly associated with stress insensitivity. Therefore, we conclude that under natural conditions, regardless of the actual total concentration of heavy metals in soil, zinc will either limit the survival of nontolerant populations of P. scaber (as shown by Donker, 1992; Jones & Hopkin, 1998) and/or will act as a main component of selection for metal stress tolerance.
The reason for the species-specificity of selection pressure components may lie in the kinetics of uptake, accumulation and loss of metals, and in the specific cellular, subcellular and physiological responses to these metals which most likely form the material upon which selection acts. The amount of accumulated zinc, lead and cadmium during artificial exposure is higher in O. asellus than in P. scaber ( Hopkin, 1989; Hopkin, 1990; Drobne & Hopkin, 1995). Though apparently contradictory to the view that zinc has a profound effect on P. scaber, Hopkin’s (1989) experiments were conducted with nontolerant specimens. Secondly, zinc was retained by contaminated P. scaber fed on uncontaminated leaf litter while it was rapidly lost from the hepatopancreas of O. asellus ( Hopkin, 1990). It therefore seems crucial for the survival of O. asellus in polluted stands that it selects food carefully from less contaminated patches and, consequently, allows zinc to be excreted from the midgut gland. This is all the more true since Hopkin (1990) estimated the ‘critical’ concentrations of zinc in the midgut gland (the maximum storage capacity) beyond which survival is impossible, to about 15 000 μg per g dry weight in O. asellus but to about 25 000 μg per g in P. scaber. The observations on feeding selectivity from the present study concur with this suggestion and indicate, furthermore, that this selectivity is not a general behavioural criterion of O. asellus, but seems to be selected for only in few populations. Our observations of selective feeding by O. asellus may mean that the external zinc concentrations are less important for its stress status than other metals.
In contrast to P. scaber, cadmium and, to a lesser degree, lead seems to be the limiting factor for the diplopod, J. scandinavius. In an ultrastructural study on cytopathological effects of metals on the main target organs of a variety of soil animals, the midgut epithelium of J. scandinavius was particularly sensitive to cadmium and displayed cytopathological alterations at cadmium concentrations which were about one order of magnitude lower than those which induced comparable effects in the midgut gland of P. scaber. For other metals, differences between species were not detected ( Köhler & Triebskorn, 1998). This particular sensitivity to cadmium is probably mirrored by the present finding that cadmium, in combination with lead and long-term pollution, seems to have acted as the main component of selection for tolerance in J. scandinavius.
Survival strategies and trade-offs
Since either excessive accumulation of metals or their pathological impact limits survival of isopods and diplopods under chronic exposure, selection should favour those response mechanisms and life-cycle traits which led to the avoidance of the stressor or its impact with little cost in terms of survival and reproduction. Avoidance of metal uptake can never be complete and thus both isopods and diplopods exhibit increasing metal concentrations with increasing age due to storage-excretion ( Hopkin, 1989). Also, altered behaviour may limit the ingestion of excessively polluted food and, consequently, the damage exerted by it. It is most likely that adaptation of the O. asellus populations from the polluted sites around the Wiesloch mine (site 14) is based on selective feeding. This, rather than physiological acclimation, enables O. asellus to survive under high metal ion concentrations, since a control population from an unpolluted site could not acclimatize to the same conditions within 3 months. Reduction of food uptake, however, is also a general response of nontolerant isopods to extreme metal concentrations (e.g. Joosse et al., 1981 ; Donker, 1992; Drobne & Hopkin, 1995) and does not necessarily have a genetic basis. Yet tolerance of long-term exposed soil invertebrate populations may at least partly be based on the evolution of behavioural patterns that reduce contact with or consumption of contaminated material.
If stress insensitivity in our study populations involved selection of phenotypes which are able to choose their food more carefully than others and restrict their metal absorption and proteotoxicity, this would explain the low hsp70 levels recorded in populations exposed to long-term pollution. Is then a low hsp70 level only a by-product of altered behaviour or might it be subject itself to evolution? At least for O. asellus the latter possibility cannot be excluded. After ingestion of contaminated food, it is essential for O. asellus to feed selectively on less contaminated litter to excrete zinc from its midgut gland ( Hopkin, 1990). Accumulation is quicker and the critical zinc level lower in O. asellus than in P. scaber. One therefore should expect zinc to be the most important ion involved in the selection of behavioural adapted phenotypes of O. asellus. If the stress protein level was just a side-effect of altered behaviour, zinc must also be assumed as being most relevant for the hsp70 level. However, in our study, zinc has been shown to be of only marginal importance to explain selection of insensitive populations in O. asellus and, thus, the hsp70 level itself (or any other factor determining it) must be regarded as being the subject to evolution in the investigated populations.
Even though a transiently elevated hsp70 level undoubtedly helps organisms to restrict the effects of a stress factor, our insensitive populations have been seemingly selected for a low stress protein level. If the stress protein level is subject to evolution, and a low hsp70 level is favoured instead of a constitutively elevated level, the latter likely trades off against other fitness consequences across the entire life cycle. The constitutive formation of stress proteins at a high rate would be an energy-intense strategy to cure the symptoms of proteotoxicity and, surely, would alter processes of energy allocation. In this context, it has been shown that the higher stress tolerance of individuals with an elevated hsp70 level or extra copies of hsp70 genes in the genome ( Welte et al., 1993 ; Feder et al., 1996 ; Krebs et al., 1998 ) was associated with reduced juvenile-to-adult survival and lengthened development in Drosophila melanogaster ( Krebs & Feder, 1997a, b; Krebset al., 1998 ).
The evolution of greater protein integrity in long-term stressed populations to mitigate the cost of hsp70 chaperoning, however, may result in other costs, such as, for example, increased energy expenditure due to selective feeding or increased detoxification mechanisms, reduced heterozygosity, reduced food uptake or nutrient absorption. Interestingly, Donker et al. (1993b ) found that P. scaber from around a smelter at Budel, The Netherlands, showed increased mortality and slower growth rates than control populations, but persist through earlier reproduction and increased reproductive allocation, suggestive of trade-offs ( Southwood, 1988; Sibly & Calow, 1989). Further studies indicate that P. scaber from the Budel population had a greater capacity to detoxify zinc ( Donker et al., 1996 ) and cadmium ( Donker & Bogert, 1991) relative to those from other sites. The energy costs of detoxifying high concentrations of metals in the diet most likely were responsible for the smaller average size of P. scaber in the contaminated areas ( Donker, 1992; Donker et al., 1993a ). Both P. scaber and O. asellus from a metal gradient at Avonmouth, UK, were also smaller than typical individuals from noncontaminated sites ( Jones & Hopkin, 1998), but there was no change in their reproductive allocation ( Jones & Hopkin, 1996).
Although the molecular mechanisms of metal detoxification which may have been induced in insensitive populations of isopods and diplopods remain unclear, they may represent variables that selection has also acted upon in the polluted southern German sites. One may argue that an induction of metal-scavenging proteins could be responsible for tolerance but, as yet, such proteins are unknown for both isopods and diplopods.
We are grateful to Sabine Ludwig for her help during field sampling, to Klaus Dietz for mathematical advice, and to the Baden-Württemberg Research Agency for Forestry, Freiburg, Gemany, as well as to the District Magistracy of Tübingen, Germany, for the provision of numerous analytical data. We also greatly acknowledge the critical prereview of the manuscript by Rita Triebskorn and Robert Paxton. The study was financed by the EU Environment and Climate Programme (BIOPRINT-II, ENV4-CT96–0222).