Comparison of instantaneous rate of population increase and critical-effect estimates in Folsomia candida exposed to four toxicants

ARTICLE IN PRESS

Ecotoxicology and Environmental Safety 57 (2004) 175–183

Comparison of instantaneous rate of population increase and critical-effect estimates in Folsomia candida exposed to four toxicants Iain N. Herbert, Claus Svendsen, Peter K. Hankard, and David J. Spurgeon
Centre for Ecology and Hydrology, Monks Wood, Abbotts Ripton, Huntingdon, Cambridgeshire PE28 2LS, UK Received 21 June 2002; received in revised form 4 March 2003; accepted 10 March 2003

Abstract The instantaneous rate of population increase (ri ) integrates several life cycle variables into one accessible statistic and has been proposed as a more practical alternative than assembling full life tables in the study of population-level responses to toxicant exposure. In this study the sensitivity of instantaneous rate of population increase is compared to critical-effect estimates for populations exposed to four toxicants with different modes of action. Populations of the Collembolan Folsomia candida were exposed to cadmium, copper, pyrene, and chlorpyrifos in artificial soil following the standardized ISO (International Organization for Standardization, Geneva, 1999) protocol. We calculated ri values and LC50, EC50juvenile, and NOEC values for each chemical. Comparison of the relative toxocity of the four chemicals indicated that chlorpyrifos had the lowest values and was thus the most toxic, followed by pyrene, cadmium, and copper. Significant changes in ri were seen to follow closely changes in the sublethal parameter measured (juvenile production) and showed populations in decline at concentrations as low as 40% of the LC50. The study showed ri to be a good measure of population response, and we conclude that the statistic gives a better understanding of effects on a population than through the sole use of traditional critical-effect estimates. r 2003 Elsevier Science (USA). All rights reserved. Keywords: Collembola; Demography; Extinction; Population effect assessment; Mode of action; Cadmium; Copper; Pyrene; Chlorpyrifos

1. Introduction This paper describes the results of four experiments in which Folsomia candida organisms were exposed to different toxicants (cadmium, copper, pyrene, chlorpyrifos) in artificial soil. The toxicants selected represented a range of modes of action and included metals, which act through specific mechanisms such as polar substitution of metalloproteins; polycyclic aromatic hydrocarbons (PAHs), which act by nonpolar narcosis; and pesticides, which are usually designed to act through targeting a specific molecular function (in this case acetylcholinesterase inhibition). On the basis of exposures to ranges of increasing concentrations of toxicants, LC50 and EC50 values were derived for each toxicant, and population growth at the different concentrations was measured by calculating ri : Analysis of the data allows comparison between the traditional critical-effect estimates (LC50, EC50) and variances in the population


Corresponding author. Fax: +44-1487-773-467. E-mail address: [email protected] (D.J. Spurgeon).

stability (measured through ri ), and how the relationship between these may vary depending on the mode of action of the tested chemical. In particular, it allowed us to test the null hypothesis that the relationship between lethality and effects on population growth is independent of the mode of action of the toxicant. In ecotoxicology, the ability to predict the effects of toxicants on the dynamics of natural populations can be seen as a major objective (Moe et al., 2001). To predict the effect of toxic chemicals on organisms, a number of standardized test procedures have been formulated, ring-tested, and adopted by international organizations. The primary aim of these toxicity tests is to estimate critical-effect levels (LCx, ECx) from concentration– response relationships for single life cycle variables such as mortality, growth, or reproduction (Kammenga et al., 1997). In this respect standard tests focus on protecting particular species, not necessarily communities or ecosystems. In contrast to these standardized singleendpoint tests, demographic studies allow the integration of several critical life cycle traits into a single variable (Van Leeuwen et al., 1985). This approach

0147-6513/03/$ - see front matter r 2003 Elsevier Science (USA). All rights reserved. doi:10.1016/S0147-6513(03)00033-2

ARTICLE IN PRESS 176

I.N. Herbert et al. / Ecotoxicology and Environmental Safety 57 (2004) 175–183

therefore has the potential to ensure a more considered evaluation of the potential ecological hazards of toxicants through monitoring population endpoints (Kammenga et al., 1997) and may be seen as more relevant (Forbes and Calow, 1999; Walthall and Stark, 1997a) and more dynamic (Ahmadi, 1983) than traditional methods of measuring the responses of single specific traits. Two approaches that can be used for predicting the effects of chemicals on population growth can be taken from population ecology. These are the intrinsic (rm ) and instantaneous (ri ) rates of population increase. The intrinsic rate of increase is used to measure the ability of a population to increase exponentially in an unlimited environment and is calculated using schedules of survivorship and fecundity generated from life tables. However, although rm has been shown to be an effective and ecologically relevant measure of population response (Forbes and Calow, 1999; Stark and Wennergren, 1995; Walthall and Stark, 1997a), it has remained underused in toxicological testing. This is mainly due to its labour- and time-intensive nature, which results from the need to measure chemical effects on all demographically relevant parameters throughout the full life cycle of the test organism. A potential alternative to rm that may be useful in ecotoxicological studies is the instantaneous rate of population increase (ri ). This parameter is a direct measure of population growth rate that integrates both survivorship and fecundity, as well as density-dependent regulatory mechanisms, but is less time-consuming and labour intensive than the development of a complete life table (Stark and Banks, 2000). The ri value for a population is derived from the equation ri ¼ lnðnf =no Þ=DT;

ð1Þ

where nf is the final number of animals, no the original number of animals, and DT the difference in time (number of days the experiment was run). Solving ri creates values between 1 and +1, where positive values represent a growing population, zero a stable population, and negative values a population in decline and heading toward extinction. Comparative studies have shown ri to be not statistically significantly different from rm under density-independent conditions (Walthall and Stark, 1997b), but to date, despite its potential practical advantages and ecological relevance, ri has not been widely applied as an endpoint in ecotoxicological studies. The springtail F. candida Willem (Insecta: Collembola) is an endogenic species that reproduces parthenogenically. The species reaches maturity at 15– 18 days, and females begin laying p100 eggs in alternate cycles of laying and moulting. Lifespan of the species in the laboratory at 20 C is on the order of 100–200 days. A standardized test protocol has been developed by the

ISO (1999), and the effects of a range of inorganic and organic chemicals (metals, herbicides, fungicides, insecticides) upon this species have been noted using a variety of experimental procedures (Chernova et al., 1995; Crommentuijn et al., 1995, 1997; Crouau et al., 1999; Frampton and Wratten, 2000; Pedersen et al., 1997; Sandifer and Hopkin, 1996). The widespread use of F. candida in toxicity tests and their relatively short generation time make this species ideally suited for use in studies targeting the effects of chemicals on population parameters such as ri :

2. Materials and methods 2.1. Test animals F. candida individuals obtained from Reading University were cultured in laboratory incubators at 2071 C at a 12 h light/12 h dark cycle, in plastic culture pots (200  100  70 mm3) with a moist substrate made up from plaster of Paris mixed with charcoal (8:1) and fed liberally with brewer’s yeast. Synchronized cohorts were obtained by transferring several hundred adults to fresh containers and allowing them to lay eggs over a period of 3 days. The adults were then removed from the containers and the eggs were monitored until hatching occurred. The experiments were carried out using juveniles of between 10 and 12 days old. 2.2. Preparation and contamination of the artificial soil The artificial soil used in the test is as described in the ISOs ‘‘Inhibition of Reproduction (F. candida)’’ guideline (ISO, 1999). The medium consisted of 70% quartz, 20% kaolinite clay, and 10% peat, with the pH adjusted to 6.070.5 by the addition of calcium carbonate and moisture content maintained at 35% (dry weight). After preparation of the test medium, 145 g (dry weight) per concentration was placed into kilner jars for dosing. The soil was contaminated with CdCl2 (2 12 H2O), chlorpyrifos, pyrene, and CuCl2 (all products supplied by SigmaAldrich). The concentrations to be tested were based on previously published data, spanning the reported LC50 and EC50 values for soil species that were available (Crommentuijn et al., 1993, 1995; Sandifer and Hopkin, 1996; Sverdrup et al., 2001). Six treatments and a control (with four replicates at each level) were utilized for each substance tested. Doses were spaced in geometric series such that the penultimate dose was approximate to estimated LC50 values based on available literature data. In the tests with the two metals the nominal concentrations used were 5, 20, 80, 320, 1280, and 5120 mg g1 dry soil for cadmium and 12.5, 50, 200, 800, 3200, and 12,800 mg g1 dry soil for copper. To spike the

ARTICLE IN PRESS I.N. Herbert et al. / Ecotoxicology and Environmental Safety 57 (2004) 175–183

soils, both metals (as chloride salts) were dissolved in deionized water to formulate a stock solution corresponding to the highest dose. The stock solution was diluted as required with deionized water and then added to the soil to achieve the desired nominal concentrations and water content in the test medium. The soils were dosed 1 week in advance of the experimental setup and allowed to rest in sealed kilner jars to allow equilibration of the added metal. The nominal concentrations for the pyrene experiment were 0.3125, 1.25, 5, 20, 80, and 320 mg g1 dry soil, and for chlorpyrifos, 0.001025, 0.0040625, 0.01625, 0.065, 0.26, and 1.04 mg g1 dry soil. Because these two chemicals are not water soluble, stock solutions were formulated with acetone. From the stock solutions, dilutions were made with acetone such that the nominal concentrations required could be added to the soil using a total of 15 mL acetone for all treatments. The controls for these two chemicals were also subject to the addition of 15 mL acetone. The dosed soils were left in a fume cupboard in open-topped vessels for 4 days to allow the acetone to evaporate off. Soils were then adjusted to 35% moisture content and allowed to rest for an additional 24 h before experimental setup. For all treatments the bulk prepared soil was gently mixed and its moisture content checked (and adjusted where necessary) before separation into replicates.

177

et al., 1977) programme version 1.5. EC50juvenile values were derived from measuring the reproduction of the original adults added to the system and were calculated through linear interpolation using the ICp (NorbergKing, 1988) programme version 2.0. The instantaneous rate of population increase was calculated using a modified form of Eq. (1). Because of the inability to process zero (extinction) through this equation, a modification similar to that used by Krebs and Boonstra (1978) and Mallorie and Flowerdew (1994) was chosen such that ri ¼ ðlnðnf þ 1Þ=ðno þ 1ÞÞ=DT;

ð2Þ

ri values were calculated for individual replicates and trends plotted by least-squared fitting of a logistic regression of ri against soil chemical concentration. From this a soil chemical concentration at which ri ¼ 0 was calculated. For the later comparisons all soil chemical concentrations greater than the calculated ri ¼ 0 value were deemed to represent negative ri values and thus a population in decline and heading toward extinction. Differences between the controls were analysed using one-way ANOVA and between the control and individual treatments using a General Linear Model (GLM) ANOVA followed by a post hoc Dunnetts test (5% level).

2.3. Experimental design 3. Results For each control and treatment concentration, for each tested chemical, four 100-mL glass jars (screw top) were filled with 35 g (dry weight) of the respectively undosed or spiked soil. Ten to twelve-day-old F. candida individuals were introduced to each jar, and 2 mg baker’s yeast (for food) was added. The jars were placed within incubators giving a controlled environment at 2071 C and a 12 h light/12 h dark regime. The jars were aerated and their moisture levels checked (by calculating weight loss) and adjusted as necessary twice a week. An additional 2 mg of baker’s yeast was added to the jars on day 14 of the experiment. The experiment was terminated after 28 days and the jars were flooded with deionized water, gently stirred, and the contents poured into 0.25-L polystyrene cups. A small quantity of blue ink was added to the cups to facilitate better definition of the springtails. Digital photographs were taken (Kodak DC4800 zoom, +7 diopter lens), and the number of surviving adults and juveniles was established by on-screen viewing and marking of adults and juveniles. Counts were carried out manually.

Adult survival and juvenile production in the controls for the four tests was lower than specified in the ISO guidelines but there were no statistically significant differences between adult survival (ANOVA, P40:05) or juvenile production (ANOVA, P40:05) between tests. No significant differences were found between the acetone controls in the pyrene and chlorpyrifos tests and nonacetone controls used with the metal tests (adult survival (ANOVA, P40:05) juvenile production (ANOVA, P40:05)). Thus the use of the solvent vehicle for dosing had no apparent effect on the performance of the springtails during the test. Adult survival juvenile production and total population numbers all decreased with increasing concentrations for each chemicals although the consistency of this trend varied (Figs. 1A–D). Calculated mean ri values (Figs. 2A–D) also declined in response to increases in test chemical concentrations (most notably at the highest concentrations). 3.1. Cadmium

2.4. Statistical analyses LC50 values for each treatment were initially calculated using the trimmed Spearman–Karber (Hamilton

In the cadmium experiment all the individuals of a population died at the top concentration of 5120 mg g1. Adult survival though seen to be reduced in the most

ARTICLE IN PRESS

6

60

4

40

2

20

0

0 5

80

320

1280

80

6

60

4

40

2

20

0 3.125 Control

100

8

80

6

60

4

40

2

20

0 1.25

5

20

80

200

800

12,800

8

80

6

60

4

40

2

20

0

0 0.00025 Control .001025 .0040625 .01625

.065

0.26

1.04

Soil chlorpyrifos concentration (µg g-1)

(D)

Soil pyrene concentration (µg g )

3200

100

320

-1

(C)

50

10

0 0.3125

12.5

Soil copper concentration (µg g-1)

(B)

10

0.075 Control

0

5120

Soil cadmium concentration (µg g-1)

(A)

Adult survival

20

8

Adult survival

1.25 Control

100

Juvenile production/total population

80

10

Juvenile production/total population

8

Adult survival

100

Juvenile production/total population

Adult survival

10

Juvenile production/total population

I.N. Herbert et al. / Ecotoxicology and Environmental Safety 57 (2004) 175–183

178

0.15

0.15

0.10

0.10

0.05

0.05

0.00

0.00

ri

ri

Fig. 1. Mean (7SEM) adult survival (solid bar), juvenile production (hatched bar), and total population (open bar) for F. candida exposed in four replicates per treatment to concentrations of (A) cadmium, (B) copper, (C) pyrene, and (D) chlorpyrifos during a 28-day period.

−0.05

−0.05

−0.10

−0.10 −0.15

−0.15 Control 1.25

20

80

320

1280

Control 3.125

5120

Soil cadmium concentration (µg g-1)

(A)

(B)

0.15

0.15

0.10

0.10

0.05

0.05

0.00

0.00

ri

ri

5

−0.05

−0.05

−0.10

−0.10

50

200

800

3200

12,800

Soil copper concentration (µg g-1)

−0.15

−0.15 Control 0.075

(C)

12.5

0.3

1.25

5

20

80 -1

Soil pyrene concentration (µg g )

Control 0.00025 0.001025 0.0040625 0.01625

320

(D)

0.065

0.26

1.04

-1

Soil chlorpyrifos concentration (µg g )

Fig. 2. Instantaneous rate of population increase (ri ) per replicate (open square) exposed in four replicates per treatment at increasing soil concentrations of (A) cadmium, (B) copper, (C) pyrene, and (D) chlorpyrifos during a 28-day period. Mean (7SEM) values per concentration shown by open star. Solid line indicates the fitted logistic regression based on observed values.

ARTICLE IN PRESS I.N. Herbert et al. / Ecotoxicology and Environmental Safety 57 (2004) 175–183

contaminated soils (Fig. 1A) was not significantly different from the control (Dunnetts, P40:05) at concentrations as high as 1280 mg g1. The LC50 value for cadmium was calculated to be 900 mg g1 (Table 1). Soil cadmium concentrations significantly affected juvenile production (GLM, Po0:01) and significant reductions in comparison with the control were found at concentrations of 320 mg g1 and higher (Dunnetts, Po0:01). The EC50juvenile value (Table 1) was calculated to be 112 mg g1. ri values (Fig. 2A) were calculated to range from 0.086 (extinction) to 0.089 (in one replicate at 5 mg g1). The mean control ri was calculated to be 0.057. A significant effect of cadmium on ri was found (GLM, Po0:01) with significant differences between control and treatment ri values at concentrations of 320 mg g1 and higher (Dunnetts, Po0:01). Leastsquared fitting of ri values using a logistic model gave a good agreement (R2 40:8) between the model and the observed values (Fig. 2A). From this regression the concentration at which ri ¼ 0 was calculated to be 398 mg g1 (Table 1 and Fig. 2A). 3.2. Copper Three adults (mean 0.75 per replicate) survived at the top dose (12,800 mg g1) in the copper experiment (Fig. 1B). Adult survival was significantly affected by soil copper concentration (GLM, Po0:01), but was only significantly different in comparison with the control at a soil concentration of 12,800 mg g1 (Dunnetts, Po0:01). The LC50 value (Table 1) was calculated to be 6840 mg g1. A significant effect of copper on juvenile production was found (GLM, Po0:01), with significant reductions in comparison with the control at concentrations of 3200 mg g1 and higher (Dunnetts, Po0:01). The EC50juvenile value (Table 1) was calculated to be 813 mg g1. Calculated ri values (Fig. 2B) ranged from 0.086 (extinction) to 0.077 (in one replicate at 200 mg g1). The mean control ri was calculated as 0.041. Copper significantly affected ri (GLM, Po0:01), with significant differences found between the control and treatment ri values at concentrations of 3200 mg g1 and higher (Dunnetts, Po0:01). Least-squared fitting of

179

ri values using a logistic model gave a good agreement (R2 40:75) between the model and the observed values (Fig. 2B). From this regression the concentration at which ri ¼ 0 was calculated to be 2760 mg g1 (Table 1 and Fig. 2B). 3.3. Pyrene In the pyrene experiment all the individuals died at a concentration of 320 mg g1. Adult survival, though significantly affected by exposure (GLM, Po0:05) and reduced in a concentration-dependent manner (Fig. 1C), was not significantly different from the control (Dunnetts, P40:05) at concentrations as high as 80 mg g1 owing to high variability in the controls. The LC50 value (Table 1) was calculated to be 21.7 mg g1. Pyrene significantly affected juvenile production (GLM, Po0:01), with significant reductions in comparison with the control found at concentrations of 20 mg g1 and higher (Dunnetts, Po0:05). An EC50juvenile value (Table 1) of 7.79 mg g1 was calculated. ri values (Fig. 2C) were calculated to range from 0.086 (extinction) to 0.079 (in a replicate at 0.3125 mg g1). The mean control ri was calculated to be 0.043. Pyrene significantly affected ri (GLM, Po0:01), with differences between the control and treatment ri values close to significance at 20 mg g1 (Dunnets, Po0:06) and significant (Dunnetts, Po0:01) from 80 mg g1 onwards. Least-squared fitting of ri values using a logistic model gave a good agreement (R2 40:70) between the model and the observed values (Fig. 2C). From this regression the concentration at which ri ¼ 0 was calculated to be 17.5 mg g1 (Table 1 and Fig. 2C). 3.4. Chlorpyrifos In the chlorpyrifos experiment all exposed individuals died at a concentration of 1.04 mg g1. Adult survival (Fig. 1D) was not significantly different from the control (Dunnetts, P40:05) at concentrations of up to 0.26 mg g1, and there appeared to be only a limited concentration-related reduction in numbers surviving at tested concentrations o1.04 mg g1. From survival data an LC50 value (Table 1) was calculated of 0.18 mg g1

Table 1 LC50, EC50juvenile, ri ¼ 0 values, and NOECjuvenile values (95% confidence intervals in parentheses where appropriate) for 10–12-day-old F. candida exposed to a concentration series of four chemicals during a 28-day period Chemical

LC50 (mg g1)

EC50juveniles (mg g1)

ri ¼ 0 (mg g1)

NOEC (mg g1)a

Cadmium Copper Pyrene Chlorpyrifos

900 (691–1170) 6840 (5950–7860) 21.7 (15.4–30.7) 0.18 (0.13–0.25)b

112 (17.3–199) 813 (49.2–1710) 7.79 (2.85–15.9) 0.094 (0.004–0.17)

398 2760 17.5 0.26

80 800 5 0.065

a

The NOEC is the ‘‘no observed effect concentration’’, or the concentration below which no adverse effects are observed. This value appears to be an artefact of the method used by the trimmed Spearman–Karber method to adjust data that do not show monotonic increase before LC50 estimation. Use of an alternate probit model for estimation of the LC50 gave a value of 0.28 mg g1 (95% CI 1–48 mg g1). b

ARTICLE IN PRESS 180

I.N. Herbert et al. / Ecotoxicology and Environmental Safety 57 (2004) 175–183

using the trimmed Spearman–Karber method. Comparison of this value with the data suggests that this value is likely to be an underestimate of the actual LC50 value, because survival is comparable to controls at 0.18 mg g1. Instead, this LC50 is probably an artefact of the method used by the trimmed Spearman–Karber method to adjust data that do not show monotonic increase before LC50 estimation. To clarify this, an alternate probit model was used to estimate another LC50. This gave a value of 0.28 mg g1 (95% CI 148 mg g1). In this second calculation, the variability in the data for survival again appeared to give problems for the LC50 estimation, as indicated by the wide confidence intervals. However, the death of all individuals at 1.04 mg g1 does confirm that the LC50 is well below this concentration. Chlorpyrifos exposure significantly affected juvenile production (GLM, Po0:01), with significant reductions in comparison with the control found at concentrations of 0.26 mg g1 and higher (Dunnetts, Po0:05). The EC50juvenile value (Table 1) was calculated to be 0.094 mg g1. Calculated ri values (Fig. 2D) ranged from 0.086 (extinction) to 0.082 (in a replicate at 0.001025 mg g1). The mean control ri was calculated to be 0.040. Chlorpyrifos exposure significantly affected ri values (GLM, Po0:01), with significant differences between control and treatment ri values found at concentrations of 0.26 mg g1 and higher (Dunnetts, Po0:05). Least-squared fitting of ri values using a logistic model gave a good agreement (R2 40:75) between the model and the observed values (Fig. 2D). From this regression the concentration at which ri ¼ 0 was calculated to be 0.26 mg g1 (Table 1 and Fig. 2D). 3.5. Comparison among the four chemicals Comparison of the relative toxicity of the four chemicals indicated that chlorpyrifos had the lowest LC50, EC50, ri ¼ 0; and NOECjuvenile values (Table 1), and was thus the most toxic, followed by pyrene, cadmium, and copper. For cadmium, copper, and to a lesser extent pyrene, LC50 estimates were higher than the estimated value for ri ¼ 0: For chlorpyrifos the calculated value for ri ¼ 0 was higher than the LC50 estimate from the Spearman–Karber model and comparable with the value calculated using the probit model. In all cases EC50 values for juvenile production were the most sensitive parameter and gave the lowest values.

4. Discussion 4.1. Test performance in relation to published data The general performance of F. candida within the four tests was lower than expected. Adult mortality was

higher (mean control mortality 450%) and juvenile production lower (mean control production=41 juvenile/replicate) than the performance targets set by the ISO. The reduced survival and reproductive rates found here cannot be fully explained; however, for the comparisons of performance between concentrations and chemicals, it is fortunate that results were not significantly different between the controls in the different exposures (ANOVA, P40:05). This means that comparisons between the controls and the treatments were not adversely affected. Considering springtail performance in these tests, it should be noted that the ri values calculated herein reflect these mortality and reproduction rates and are thus pertinent to the populations studied at this time. This ability to represent the actual status of populations under the prevailing conditions is a strength of the overall ri approach, because the statistic integrates the biotic responses of juvenile production, juvenile survival, adult survival, density, and so on, under the prevailing ecological conditions. Comparisons of calculated-effect concentrations determined in this study with literature values generally indicated good agreement. For cadmium the recorded LC50 of 900 mg g1 was within the range reported by Crommentuijn et al. (1993) of 778–997 mg g1 (during 19–35 days exposure). Van Gestel and Van Diepen (1997) also found LC50 values for this metal ranging from 617 to 1275 mg g1 depending on moisture content. The EC50juvenile value for cadmium was also similar to those of Crommentuijn et al. (1995) of 123–154 mg g1. A value within this range and similar to that found in this study, of 129 mg g1, was reported by Crouau et al. (1999). The results show copper to be the least toxic of the chemicals tested, having the highest LC50 (6840 mg g1), EC50juvenile (813 mg g1), and ri ¼ 0 (2760 mg g1) values. Sandifer and Hopkin (1996) gave a EC50juvenile value for copper of 700 mg g1, which is similar to the value within this paper. No comparable LC50 data for copper were available for F. candida. Relatively few data are available concerning the effects of pyrene on invertebrates in general, and Collembola in particular. Sverdrup et al. (2001) have reported LC50 and EC50juvenile values for the toxicity of pyrene to F. fimetaria L. (53 and 16 mg g1, respectively), but the data presented here are believed to be the first values pertaining to F. candida. All three measured parameters (LC50=21.7 mg g1, EC50juvenile= 1 7.79 mg g , and ri ¼ 0 at 17.5 mg g1) show pyrene to be notably toxic to F. candida. The results clearly show chlorpyrifos to be the most toxic of the chemicals tested with an estimated LC50 of 0.18 mg g1. This is not surprising, in that this compound is a purposely designed insecticide. The LC50 value for chlorpyrifos (whether estimated by the trimmed Spearman–Karber or probit model) is close to that reported

ARTICLE IN PRESS I.N. Herbert et al. / Ecotoxicology and Environmental Safety 57 (2004) 175–183

by Crommentuijn et al. (1995), of 0.20–0.28 mg g1. No comparable EC50juvenile data were available. 4.2. Comparison between lethality and sublethal responses Cadmium and copper both exhibited sublethal effects in the form of significant reductions in juvenile production. This is in line with the findings of (Crommentuijn et al., 1993, 1995; Sandifer and Hopkin, 1996; Sverdrup et al., 2001) (this last in studies with F. fimetaria). Fountain and Hopkin (2001) noted delayed onset of egg laying in F. candida exposed to copper and cadmium in their food from 1 day old. This was attributed to slower growth resulting in a longer time to reach reproductive capability, although susceptibility to such growth retardations was lower when more mature organisms were exposed. For the two metals the significant effects on juvenile production were noticeable at nominal soil concentrations lower than the LC50. Thus the EC50 values for the effects of these metals on juvenile production were 112 mg g1 for cadmium (LC50=900 mg g1) and 813 mg g1 for copper (LC50=6840 mg g1). For chlorpyrifos and to a lesser extent for pyrene, the increased sensitivity of sublethal effects over lethal values was less clear. For these two compounds, significant sublethal reproductive effects were found at, or only slightly above the values estimated to cause a 50% lethal effect, although for chlorpyrifos this may be due in part to problem with accurate estimate of the LC50. Comparison of the EC50 with the LC50 for chlorpyrifos and pyrene (Table 1) indicated that the values were related by a factor of 2–3 depending on the model used. This compares to the factors of 8 for cadmium and copper. The higher ratios found for the metals suggests that the sublethal effects may be more sensitive for these toxicants that have potentially less specific and more diverse modes of action than for the insecticide and the PAH. 4.3. Usefulness of the demographic parameter ri In this study the demographic parameter ri has been has been used to integrate several life cycle traits (adult survival, reproductive rates, and juvenile survival) and environmental influences on the populations studied. In a review of life table data, Forbes and Calow (1999) demonstrated rm to be equally sensitive to the most sensitive individual parameter defined in life table studies, but concluded that rm was more relevant because of its integration of several factors. Walthall and Stark (1997b) have shown ri to be comparable to rm : Thus ri may be seen as a quicker and easier method of assessing population endpoints in toxicological testing than life tables, although the latter do provide a more

181

detailed mechanistic understanding. The results presented here show ri to have a level of sensitivity similar to the sublethal parameter that was measured (juvenile production), with significant reductions in ri closely following the significant reductions in juvenile numbers. This suggests that when sublethal effects occur they can be displayed and therefore measured through the use of an integrating parameter such as ri : However, to integrate fully the effects of toxicants on all life cycle parameters, exposures should be long enough to encompass the full life cycle. In this respect, the standard protocol used herein (ISO, 1999) may not fully assess a populations response; effects on some important demographic parameter such as ‘‘time to sexual maturity’’ will not be fully integrated. The results shown here display that whereas adult survival and juvenile reproduction levels were lower than expected within the experiment, traditional criticaleffect estimates remained comparable to previously presented data. Estimation of these parameters thus allows the comparison of toxicity results between situations but actually provides little information concerning the actual fates of the exposed populations. Despite the limitations of the test method, a more complete estimate of population fates in the different treatments can be gained by careful examination of the trends in ri data. In this respect, the most significant change for a population will occur when ri changes from a positive to a negative. This transition represents a change from a growing or stable population to one in decline. For cadmium and copper, logistic regression of ri values (Figs. 2A-D) suggested that these populations were heading toward extinction (ri ¼ 0) at concentrations of between 40% and 80% of LC50 estimates. For pyrene, ri ¼ 0 was marginally below the calculated LC50, whereas for chlorpyrifos the ri ¼ 0 value was above the LC50 values estimated from the two models used. The relationship between ri ¼ 0 and LC50 thus also suggests that populations are more vulnerable to sublethal exposure to pollutants with complex modes of action than those for which there is a specific response. The relationship between LC50 and population growth rate has been investigated in a limited number of studies. Bechmann (1994), in experiments with the marine copepod Tisbe furcata, found that significant reductions in demographic parameters leading to a 61% reduction in rm could be directly related to toxicants at levels as low as 32% of the LC50. Snell and Serra (2000) utilized long-term modelling of rotifer populations to conclude that reductions in rm greater than B30% (compared with the controls) could destine populations to extinction, whereas reductions of as little as 5% could double the probability of extinction. These studies and the work conducted here demonstrate that populationlevel endpoints can lead to a better understanding of the

ARTICLE IN PRESS 182

I.N. Herbert et al. / Ecotoxicology and Environmental Safety 57 (2004) 175–183

real relevance of toxicant effect than critical-effect estimates and can provide valuable information concerning the fates of populations exposed to diverse pollutants. An approach for predicting the critical concentration of chemicals resulting in the long-term decline to extinction of exposed natural populations would be greatly valued by managers and policymakers. As a method to estimate chemical effects on population growth rate, instantaneous rate of population increase is attractive because it is more easily measured than full life cycle responses. This simplicity, however, comes at the expense of knowledge concerning responses of individual parameters. For predicting the actual effects of chemicals on populations in the field, both of the demographic approaches can be of value. Their correct application is, however, dependent on the development of complex ecological models that can integrate the effects of density dependence, environmental stochasticity, predation, and parasite and disease load on population growth and size. Such models require detailed knowledge concerning the basic biology and ecology of the species being considered.

Acknowledgments Thanks to S. Hopkin and M. Fountain for supplying the F. candida. This work was supported by the National Environment Research Council through the Centre for Ecology and Hydrology sandwich student placement scheme.

References Ahmadi, A., 1983. Demographic toxicology as a method for studying the dicofol-two spotted spider mite (acari: tetranychidae) system. J. Econ. Entomol. 76, 239–242. Bechmann, R.K., 1994. Use of life-tables and LC50 tests to evaluate chronic and acute toxicity effects of copper on the marine copepod Tisbe furcata (Baird). Environ. Toxicol. Chem. 13, 1509–1517. Chernova, N.M., Balabina, I.P., Ponomareva, O.N., 1995. Changes in population growth of springtails (Collembola) under the influence of herbicides. Pol. Pismo Entomol. 64, 91–98. Crommentuijn, T., Brils, J., Vanstraalen, N.M., 1993. Influence of cadmium on life-history characteristics of Folsomia candida (Willem) in an artificial soil substrate. Ecotoxicol. Environ. Saf. 26, 216–227. Crommentuijn, T., Stab, J.A., Doornekamp, A., Estoppey, O., Van Gestel, C.A.M., 1995. Comparative ecotoxicity of cadmium, chlorpyrifos and triphenyltin hydroxide for 4 clones of the parthenogenetic collembolan Folsomia candida in an artificial soil. Funct. Ecol. 9, 734–742. Crommentuijn, T., Doornekamp, A., Van Gestel, C.A.M., 1997. Bioavailability and ecological effects of cadmium on Folsomia candida (Willem) in an artificial soil substrate as influenced by pH and organic matter. Appl. Soil Ecol. 5, 261–271. Crouau, Y., Chenon, P., Gisclard, C., 1999. The use of Folsomia candida (Collembola, Isotomidae) for the bioassay of

xenobiotic substances and soil pollutants. Appl. Soil Ecol. 12, 103–111. Forbes, V.E., Calow, P., 1999. Is the per capita rate of increase a good measure of population-level effects in ecotoxicology? Environ. Toxicol. Chem. 18, 1544–1556. Fountain, M.T., Hopkin, S.P., 2001. Continuous monitoring of Folsomia candida (Insecta: Collembola) in a metal exposure test. Ecotoxicol. Environ. Saf. 48, 275–286. Frampton, G.K., Wratten, S.D., 2000. Effects of benzimidazole and triazole fungicide use on epigeic species of collembola in wheat. Ecotoxicol. Environ. Saf. 46, 64–72. Hamilton, M.A., Russo, P.C., Thurston, R.V., 1977. Trimmed Spearman–Karber method for estimating median lethal concentrations in toxicity bioassays. Environ. Sci. Technol. 12, 417. ISO, 1999. Soil quality—inhibition of reproduction of collembola (Folsomia candida) by soil pollutants. Report No. ISO11267, International Organization for Standardization, Geneva. Kammenga, J.E., Korthals, G.W., Bongers, T., Bakker, J., 1997. Reaction norms for life-history traits as the basis for the evaluation of critical effect levels of toxicants. In: Van Straalen, N.M., L^kke, H. (Eds.), Ecological Risk Assessment of Contaminants in Soils. Chapman & Hall, London, pp. 293–304. Krebs, C.J., Boonstra, R., 1978. Demography of the spring decline in populations of the vole, Microtus townsendii. J. Anim. Ecol. 47, 1007–1015. Mallorie, H.C., Flowerdew, J.R., 1994. Woodland small mammal population ecology in Britain—a preliminary review of the mammal-society survey of wood mice Apodemus sylvaticus and bank voles Clethrionomys glareolus. Mammal Rev. 24, 1–15. Moe, S.J., Stenseth, N.C., Smith, R.H., 2001. Effects of a toxicant on population growth rates: sublethal and delayed responses in blowfly populations. Funct. Ecol. 15, 712–721. Norberg-King, T.J., 1988. An interpolation estimate for chronic toxicity: the ICp approach. Document support for Release 1.0 and 1.1, Report No. 05-88, US Environmental Protection Agency, Duluth, MN. Pedersen, M.B., Temminghoff, E.J.M., Marinussen, M., Elmegaard, N., Van Gestel, C.A.M., 1997. Copper accumulation and fitness of Folsomia candida Willem in a copper-contaminated sandy soil as affected by pH and soil moisture. Appl. Soil Ecol. 6, 135–146. Sandifer, R.D., Hopkin, S.P., 1996. Effects of pH on the toxicity of cadmium, copper, lead and zinc to Folsomia candida Willem, 1902 (Collembola) in a standard laboratory test system. Chemosphere 33, 2475–2486. Snell, T.W., Serra, M., 2000. Using probability of extinction to evaluate the ecological significance of toxicant effects. Environ. Toxicol. Chem. 19, 2357–2363. Stark, J., Banks, J., 2000. The toxicologists’ and ecologists’ point of view—unification through a demographic approach. In: Kammenga, J., Laskowski, R. (Eds.), Demography in Ecotoxicology. Wiley, Chichester, pp. 9–23. Stark, J.D., Wennergren, U., 1995. Can population effects of pesticides be predicted from demographic toxicological studies. J. Econ. Entomol. 88, 1089–1096. Sverdrup, L.E., Kelley, A.E., Krogh, P.H., Nielsen, T., Jensen, J., Scott-Fordsmand, J.J., Stenersen, J., 2001. Effects of eight polycyclic aromatic compounds on the survival and reproduction of the springtail Folsomia fimetaria L. (Collembola: Isotomidae). Environ. Toxicol. Chem. 20, 1332–1338. Van Gestel, C.A.M., Van Diepen, A.M.F., 1997. The influence of soil moisture content on the bioavailability and toxicity of cadmium for Folsomia candida Willem (Collembola: Isotomidae). Ecotoxicol. Environ. Saf. 36, 123–132.

ARTICLE IN PRESS I.N. Herbert et al. / Ecotoxicology and Environmental Safety 57 (2004) 175–183 Van Leeuwen, C.J., Luttmer, W.J., Griffioen, P.S., 1985. The use of Cohorts and Populations in Chronic Toxicity studies with Daphnia magna: a Cadmium example. Ecotoxicol. Environ. Saf. 9, 26–39. Walthall, W.K., Stark, J.D., 1997a. A comparison of acute mortality and population growth rate as endpoints of toxicological effect. Ecotoxicol. Environ. Saf. 37, 45–52.

183

Walthall, W.K., Stark, J.D., 1997b. Comparison of two population-level ecotoxicological endpoints: the intrinsic (rm) and instantaneous (ri) rates of increase. Environ. Toxicol. Chem. 16, 1068–1073.