Chronic Effects of an Insect Growth Regulator (teflubenzuron) on the Life Cycle and Population Growth Rate of Folsomia candida

Current standard toxicity tests on nontarget soil invertebrates mainly focus on the endpoints survival and reproduction. Such results are likely insufficient to predict effects at higher organizational levels, for example, the population level. We assessed the effects of exposure to the pesticide teflubenzuron on the collembolan Folsomia candida, by performing a full life‐cycle experiment exposing single individuals via contaminated food (uncontaminated control and 0.2, 0.32, 0.48, 0.72, 1.08, and 1.6 mg/kg dry yeast). Several life‐history traits were considered by following the growth and development of newly hatched individuals over a period of 65 days. We assessed survival, body length, time to first oviposition, cumulative egg production, and hatchability of eggs. A two‐stage model was applied to calculate the population growth rate (λ) combined with elasticity analysis to reveal the relative sensitivity of λ to the effects of teflubenzuron on each life‐history parameter. Body length was the least sensitive life‐history parameter (median effective concentration = 1.10 mg teflubenzuron/kg dry yeast) followed by time to first oviposition (0.96 mg/kg), survival (median lethal concentration = 0.87 mg/kg), cumulative egg production (0.32 mg/kg), and egg hatchability (0.27 mg/kg). Population growth decreased with increasing concentrations of teflubenzuron (λ = 1.162/day in control to 1.005/day in 0.72 mg/kg dry yeast, with populations going extinct at 1.08 and 1.6 mg/kg dry yeast). Elasticity analysis showed that changes in juvenile survival had a greater impact on the population growth rate compared with the other life‐history traits. Our study provides a comprehensive overview of individual‐level effects of long‐term exposure to teflubenzuron and integrates these effects to assess the potential risk to collembolan populations. Environ Toxicol Chem 2024;43:1173–1183. © 2024 The Authors. Environmental Toxicology and Chemistry published by Wiley Periodicals LLC on behalf of SETAC.


INTRODUCTION
The emphasis on "eco" in ecotoxicology dates back to 1988 when Cairns first proposed it.(Cairns, 1988).The environmental risk assessment (ERA) of chemicals has traditionally focused on measuring the effects of chemicals at the individual level (Accolla et al., 2021), even though the goal of ERA of chemicals is to protect higher organization levels (e.g., populations, communities, ecosystems).Thus there is a need to extrapolate the effects of chemicals measured in individuals to impacts at higher organization levels.One way to extrapolate is through ecological modeling, such as matrix population models (Charles et al., 2009), individual-based models (Martin et al., 2013), or ecosystem/food web models (Leuven & Poudevigne, 2002).Ecological modeling is promising but may require extensive data input for model parameterization, calibration, and validation (Raimondo et al., 2018).The current ERA of chemicals has generated much data on the effects of different chemicals at the individual level, using survival, growth, and reproduction as endpoints (Organisation for Economic Cooperation and Development [OECD], 2016a[OECD], , 2016b)).However, there are additional individual-level endpoints that are needed to extrapolate to the population level, for example, time to reach maturity, hatchability, and juvenile survival (Martin et al., 2014;Walthall & Stark, 1997), and it is therefore relevant to include more endpoints, especially demographic ones, in toxicity tests (Jager et al., 2004).Such endpoints can be, but are not limited to, adult survival, juvenile survival, time to reach sexual maturity, fecundity, and hatchability (Rubach et al., 2011;Sibly & Hone, 2002).Demographic endpoints can be collected from full life-cycle experiments that use an integrated approach by tracking life-history traits of an animal over a full life cycle.Full life-cycle experiments have been used in many toxicity studies (Jager et al., 2004;Scott-Fordsmand et al., 2022;Stark & Banks, 2003).With knowledge of the effect of the chemical on demographic endpoints, we can reveal the toxicological sensitivity of different endpoints and quantify the effect of the chemical at the population level by calculating the population growth rate.Reporting the population growth rate at different concentrations of chemical exposure provides a more ecologically relevant response than conventional individual-level toxicity test endpoints (Forbes & Calow, 2002;Stark & Banks, 2003).
Soil invertebrates are widely distributed and play essential roles in soil ecosystems (Lavelle et al., 2006;Rohr et al., 2016).Soil invertebrates actively affect biological, physical, and chemical processes in soil; their metabolic activity directly contributes up to 10% of the decomposition in soil (Griffiths et al., 2021;Seastedt, 1984).Soil microarthropods like collembolans also contribute to decomposition processes indirectly through stimulation of microbial activity (Hanlon & Anderson, 1979;Hopkin, 1997); they also regulate the flow of energy through food webs both above and below ground (Potapov et al., 2023).In ERA of chemicals, collembolans are commonly used to represent nontarget soil organisms.Despite their importance, only a few full life-cycle experiments have used collembolans.Instead, most full life-cycle experiments have employed either aquatic organisms (e.g., fish, mollusks, rotifers) or larger terrestrial organisms (e.g., earthworms; Ankley et al., 2005;Hazelton et al., 2012;Santos et al., 2022;Scott-Fordsmand et al., 2022;Tassou & Schulz, 2011;Zhang et al., 2016).
We studied the pesticide teflubenzuron, a chitin synthesis inhibitor (CSI).Because humans and other mammals do not have chitin, CSIs have minimal adverse effects on them, and insect pests rarely develop resistance to them.They are intensively utilized in various applications such as controlling flies, beetles, and moths in fruit trees and greenhouse crops (Chandi & Kaur, 2021;Lewis et al., 2016;Merzendorfer, 2013;Pener & Dhadialla, 2012;Rezende et al., 2016), protecting grains from pests during product storage (Abo-Elghar et al., 2004), and controlling gill lice on salmon farms (Harðardóttir et al., 2019;Macken et al., 2015).Even though teflubenzuron and several other CSIs (including chlorfluazuron, diflubenzuron, flucycloxuron, flufenoxuron, hexaflumuron, lufenuron, triflumuron, and novaluron) are currently not approved as active substances in plant protection products within the European Union (2009), many other countries still allow the use of CSIs, and several scientific studies have proved the efficiency of new CSIs, which may lead to their increasing use (Cruces et al., 2021;Hamadah et al., 2015;Kostyukovsky & Trostanetsky, 2006;Yasir et al., 2019).
CSIs act by blocking the postcatalytic step of chitin synthesis and ultimately altering the structure of the insect cuticle (Merzendorfer, 2013;Van Leeuwen et al., 2012).Because they are sprayed on the plants, soil organisms might be exposed to the chemicals.Effects of CSIs on nontarget organisms such as collembolans have been shown.Reductions in reproduction and molting were found in Folsomia candida and Yuukianura szeptyckii, and a field experiment found slow population recovery in earthworms, enchytraeids, mites, and collembolans (Beck et al., 2004;Campiche et al., 2006;Lee et al., 2019).Negative impacts on nontarget aquatic species have also been reported, including in fish (Gambusia affinis), shrimps (Penaeus Kerathurus), copepods (Tisbe battagliai), and cladocerans (Daphnia magna).These studies have shown different degrees of low survival, abnormal molting, and decreased reproduction in response to CSI exposure (Djeghader et al., 2014;Kashian & Dodson, 2002;Macken et al., 2015;Morsli et al., 2015;Zaidi & Soltani, 2011).However, no studies have applied full life-cycle experiments to evaluate the toxicity of teflubenzuron at the population level.Assessments of the consequences of teflubenzuron for population growth rate, and an analysis of the sensitivity of population growth rate to impacts on the lifehistory traits contributing to it, can increase our mechanistic understanding of individual-to population-level extrapolation and thereby improve ERAs.
Our study aimed to assess the toxicological sensitivity of F. candida's life-history traits to teflubenzuron and the relative sensitivity (i.e., elasticity) of population growth rate to changes in these traits.We measured survival, body growth, time to first oviposition, egg production, and hatchability of individual F. candida over a 2-month dietary exposure experiment encompassing their entire life cycle.Additionally, we integrated the traits using a two-stage population model to quantify the impact of teflubenzuron on population growth rate and conducted elasticity analysis to determine the relative sensitivity of population growth rate to toxic effects on the life-history traits.

Animals
The parental F. candida were taken from cultures maintained in our laboratory for several years, but originally came from the "Berlin strain" (Simonsen & Christensen, 2001).Because this species is parthenogenetic and has been cultured for hundreds of generations in constant laboratory conditions, the genetic variation among individuals is likely negligible (Simonsen & Christensen, 2001).The cultures were maintained at a temperature of 20 °C and a 16:8-h light:dark cycle and were fed with dried baker's yeast (Malteserkors) ad libitum.
For the production of age-synchronized juveniles, 30 adults were transferred into a Petri dish (90-mm diameter) with an approximately 0.5 cm-layer of a mixture of plaster of Paris and activated charcoal (8:1 w/w) for laying eggs.After 48 h, the adults were removed, and the eggs were left until hatching.Plaster moisture was maintained by adding deionized water to the saturation point twice weekly.

Chemicals
The F. candida were exposed to teflubenzuron through food.The method was adapted from a study by Crommentuijn et al. (1997).The active ingredient teflubenzuron (greater than 98% purity) was provided by Merck (CAS no.83121-18-0).Teflubenzuron was dissolved in acetone (Sigma-Aldrich) to get a 400-mg/L stock solution.Serial dilution was conducted subsequently to reach final concentrations of 4, 2.7, 1.8, 1.2, 0.8, and 0.5 mg/L.An amount of 0.25 g dry baker's yeast was spiked with 0.1 mL of the final solution and left in a fume hood overnight until the acetone had fully evaporated.The concentrations of teflubenzuron in yeast were then 1.6, 1.08, 0.72, 0.48, 0.32, and 0.2 mg/kg dry yeast, respectively.Water and solvent control yeasts were prepared similarly using deionized water and acetone, respectively.The recommended field application dose is 0.15 mg/kg of soil, and the highest concentration in our study is equivalent to approximately 10× the recommended field dose.The lowest concentration used in our study (0.2 mg/kg dry yeast) was less than the concentration causing 10% lethality in a 28-day reproduction test under International Organization for Standardization (2023) standard 11267 with F. candida (0.46 mg/kg dry soil; Campiche et al., 2006;Cycoń et al., 2012).
An amount of 0.25 g of spiked yeast was dissolved in 0.75 mL of deionized water with a needle and later mixed by a vortex mixer for a few seconds until the yeast suspension became homogenous.An amount of 20 μL of the yeast suspension was added to the center of a filter paper disc (0.6-cm diameter), which was later provided to each individual collembolan.The teflubenzuron-contaminated yeast suspension was renewed every week.

Chemical analysis
To confirm that teflubenzuron accumulated in F. candida via dietary exposure over time, we performed a separate experiment for chemical quantification alone.We exposed 10 to 12day-old F. candida over 7 and 14 days, respectively, to nominal concentrations of teflubenzuron: 0.043, 0.14, 0.48, and 1.6 mg/ kg dry yeast.We used 25 individuals/replicate and five replicates/concentration.Due to the highly lipophilic properties of teflubenzuron, the lipid fraction of springtail samples were extracted using the method of Xu et al. (2020).Quantification analysis was performed on an Acquity ultra-performance liquid chromatography device, coupled with Xevo TQ Absolute mass spectrometry (Waters).The limits of detection and quantification of teflubenzuron were 0.022 and 0.073 ng/mL, respectively.The detailed procedures can be found in the Supporting Information S1 (Chemical analysis).Quantified teflubenzuron data are listed in the Supporting Information S1, Table S1.

Experimental set-up
Newly hatched juveniles (2-3 days old; 12 individuals [replicates] in each treatment) were exposed for a period of 65 days.Each individual was placed in a well of a 24 multiwell plate (Merck) containing a floor of moist plaster of Paris and received one paper disc where it held the yeast suspension, as previously described in the Chemicals section.Wells were closed with rubber stoppers that had been soaked in water overnight prior to use, to remove static electricity.The advantages of using this method are that animals cannot hide but must stay on the surface of the plaster of Paris, allowing continuous observations of body size and egg production (Crommentuijn et al., 1997;Folker-Hansen et al., 1996).
To measure body length, pictures of each animal were taken on days 1 (i.e., the first exposure day), 4, 8, 11, 15, 21, 29, 40, 47, 57, and 65 of exposure.The definition of collembolan body length was adopted from Folker-Hansen et al. (1996), which was "from the posterior end of the abdomen to the anterior end of the head, between the antennae."Pictures were taken with a Dino-Eye digital microscope (Dino-Lite model AM7025X), and collembolan length was measured using the DinoCapture Ver.2.0 software (AnMo Electrics).Length calibration of the DinoCapture was done according to the user's manual at 12.5× magnification.Survival was scored on the same day as the pictures were taken.
Folsomia candida normally has its first oviposition when it is approximately 21 to 24 days old at 20 °C (Fountain & Hopkin, 2005).To ensure correct recording of time to first oviposition, we checked for reproduction daily from age 14 days.The eggs were counted weekly for 7 consecutive weeks, collected with a soft brush, and transferred to a Petri dish (35-mm diameter) containing moist plaster of Paris.Eggs were incubated in Petri dishes at 20 °C for 21 days to ensure all viable juveniles had hatched.Eggs of F. candida hatch in approximately 10 days at 20 °C (de Lima e Silva et al., 2021).Eggs overgrown with mold or any other abnormal conditions were excluded from the analyses.
We performed this experiment twice to have 24 replicates/ treatment.However, the animals used in the second trial were 1 day older on the first day of teflubenzuron exposure than those in the first trial.We confirmed there was no difference between the two batches before we analyzed the merged data (see the Statistical analysis section for more details).

Logistic growth model
Because our growth data covered animals at a young age, the logistic growth model described growth better than the von Bertalanffy model (Folker-Hansen et al., 1996).The growth data of each individual in each treatment were used to fit a logistic growth model as follows: where L t is the length of the animal at time t (mm), ∞ L is the maximum length of the animal (mm), G is the instantaneous rate of growth at the origin of the curve (mm day -1 ), and t 0 is the animal age at the inflection point of the curve where the absolute growth rate begins to decline (day).
Individuals that did not survive the entire experiment but had sufficient data points to fit the logistic growth model were included as valid replicates.However, individuals with insufficient data points for the model to fit were excluded.

Population growth rate (λ)
The population growth rate (λ) was calculated from lifehistory traits at different concentrations of teflubenzuron by using a simplified two-stage model (Calow & Sibly, 1990;Forbes et al., 2001).The model is described as follows: where n is the number of eggs/brood, S j is the average of ju- venile survival from hatch to t j calculated as p j t j where p j is the average juvenile survival between census days, λ is the pop- ulation growth rate/day, t j is the average time to first re- production in days, S a is the average of adult survival between census days after , and t a is the average time between broods (estimated from the period between census days).
We used an average hatchability/treatment as the egg survival instead of actual hatchability for each replicate in calculating S j .This is because some individuals did not have valid hatchability due to their eggs being covered with mold.
The total variance of λ is the sum of the variance component of each life-history parameter (i.e., n S S t t , , , , and j a j a ).The calculation of the total variance of λ is based on Sibly et al. (2000).With a known total variance of λ, the 95% confidence interval can be calculated as √ × ( λ) total variance of 1.96 . The effect of teflubenzuron on population growth rate was considered significant if there was no overlap between the 95% confidence interval around the mean value of λ for each treatment and the control.

Elasticity analysis (relative sensitivity)
Elasticity analysis of population growth rate quantifies the relative sensitivity of λ to changes in the life-history traits (i.e., n, S j , t j , S a , and t a ) contributing to it.
The elasticity was calculated as ( /λ) × (Δλ/Δ ) i i , where i de- notes the life-history traits.(Δλ/Δ ) i was calculated by implicit differentiation, as described in Forbes et al. (2001).Because S j is dependent on t j , i.e., = S p j j t j , we used p j in the elasticity analysis of juvenile survival, as follows: Estimation of LC50 and EC50 We estimated the median lethal and effective concentrations (LC50 and EC50) for the directly measured endpoints, namely, adult survival, time to first oviposition, cumulative egg production, and hatchability plus maximum body length ( ∞ L ).Details are described below.We also estimated L(E)C50 for the five parameters used in the two-stage model, namely, n S S t , , , j a j , and t a that were described in the Population growth rate (λ) section.
The LC50 was based on the number of individuals surviving until the last day of the experiment.For time to first oviposition, individuals that had survived (not necessarily through the entire experiment) and laid eggs were considered valid replicates.However, individuals that survived the entire experiment but never laid eggs, as well as those that did not survive and therefore had no data on time to first oviposition, were both excluded.For cumulative egg production/replicate, all individuals were considered valid.This means that for individuals that died during the experiment but laid eggs, we calculated the cumulative egg production as adding up all the eggs laid by each individual throughout the experiment.For individuals that died but did not lay eggs and also those that did not die but did not lay eggs, we calculated the cumulative egg production as 0. For hatchability, only individuals with valid egg clutches were considered as replicates.This means that if an individual laid multiple egg clutches but all eggs exhibited abnormal conditions during hatching (e.g., overgrown mold), that individual was not included as a replicate in the hatchability data.We estimated the number of juveniles as the cumulative egg production/replicate × average hatchability.
All LC50 and EC50 values were determined using a threeparameter log-logistic model with a lower limit of 0. Model fitting was done using the R package drc (Ritz et al., 2015).The EC50 values could not be estimated for growth rate (G) and inflection point ( ) t 0 because the highest concentration did not result in a maximal effect of 50%.To estimate the EC50 based on time to first oviposition (t egg ), we transformed the data as "−t egg + t maximum egg " such that the response decreased with in- creasing concentration (as assumed by the log-logistic model).

Statistical analysis
We merged data from the two experiments/batches into one data set after we had visually examined and confirmed the overlap of 95% confidence intervals for mean values of endpoints (including three parameters from the logistic growth model ( ∞ G L t , , 0 ), time to first oviposition, cumulative egg production, hatchability, five input data for the two-stage model, and population growth rate; Supporting Information S1, Figure S1).For survival data, we merged data from the two batches by using the average exposure day in a Kaplan-Meier estimation.Overlapping 95% confidence intervals between LC50 values estimated from the first and second "batch" (0.74; [95% CI: 0.35-1.1]and 1.03 [95% CI: 0.3-1.8]mg/kg dry yeast, respectively) indicated no difference between the two "batches," and therefore we used the merged survival data to estimate the LC50.
There was no significant difference between the water and acetone controls, as tested by simple t-tests.The data for the acetone control were used as the control in subsequent data analyses.Survival data (over 11 time points) were analyzed using Kaplan-Meier estimation and log-rank test for multiple comparisons.Survival data at exposure day 65 were fit using a log-logistic model to estimate the LC50 (see the Logistic growth model section).Growth data were analyzed using a logistic growth model (Equation 1), effects of teflubenzuron on the parameters ∞ L and G were tested using linear models, and the parameter t 0 was tested using a generalized linear model (GLM) with Gamma distribution and inverse as the link function.The data on time to first oviposition and cumulative egg production were analyzed using GLM with Poisson distributed errors and log as the link function.Hatchability data were analyzed using GLM with binomially distributed errors and logit as the link function.Differences in population growth rates under different treatments were considered significant if their 95% confidence intervals did not overlap (Sibly et al., 2000).
All statistical analyses were done in R studio (Ver.4.0.3;R Core Team, 2020).The Kaplan-Meier survival estimation was done using the R package survival (Therneau, 2022).The logistic growth models were constructed using the R function nls.The GLMs were constructed using the function glm.The population growth rate and elasticity analysis were calculated in Microsoft Excel with the Solver Add-in (Ver.2016).Data visualization was done using the R package survminer, corrplot, and ggplot2 (Kassambara et al., 2021;Wei & Simko, 2021;Wickham, 2016).

Survival over time
In general, F. candida had high survival at low concentrations (i.e., 0.2-0.48mg/kg dry yeast), and more than 70% until the end of the experiment (Figure 1).After 30 days of exposure, a clear drop in survival probability was seen at high concentrations, ranging from 0.72 to 1.6 mg/kg dry yeast.The F. candida exposed to high concentrations, that is, 0.72 to 1.6 mg/kg dry yeast, had significantly lower survivals than controls at the end of the experiment (p < 0.001; see complete results of the log-rank test in the Supporting Information, Table S2).

Logistic models for growth
A logistic growth model was fitted to the growth of F. candida over time (see values of parameters and their 95% confidence intervals for each individual in the Supporting Information, Table S3).The F. candida grew at concentrations up to 1.08 mg/kg dry yeast, but only obtained half the maximum body length compared with the control (Figure 2).The effect of teflubenzuron on the reduction in body length was significant ( ∞ L ; LM, p < 0.001).The growth rate (i.e., the steepness of the growth curve) significantly decreased with increasing concentrations of teflubenzuron (G; LM, p < 0.01).Teflubenzuron did not decrease the inflection point of the growth curve (t 0 ; GLM, p = 0.237).

Reproduction
Teflubenzuron significantly delayed the time to first oviposition with increasing concentration (Figure 3A; GLM, p < 0.001).In the control group, F. candida took an average of 20 days to become sexually mature, whereas it took approximately 30 days at 0.72 mg/kg dry yeast.The times to first oviposition for the 1.08-and 1.60-mg/kg dry yeast treatment were removed from the analysis due to the death or a longer time to maturity than the experiment period.The cumulative egg production/individual decreased with increasing concentrations of teflubenzuron (Figure 3B; GLM, p < 0.001).At 0.72 mg/kg dry yeast, the number of eggs was 7 times lower than in the control group.The hatchability declined with exposure to increasing teflubenzuron concentrations (Figure 3C; GLM, p < 0.001).The hatchability in the control was 72% on average, and it was reduced to 3% on average at 0.72 mg/kg dry yeast.

LC50 and EC50 of teflubenzuron
Results of LC50 and EC50 of teflubenzuron on individuallevel endpoints are summarized in Table 1 together with other studies reporting the effects of teflubenzuron on Collembola.In the present study, hatchability and cumulative egg production were the two most sensitive endpoints, and body length was the least.
A rank of toxicity sensitivity based on the estimation of LC50 and EC50 values of teflubenzuron for the five parameters used in the two-stage model are shown in Figure 4. Egg production  , 20, 23, 22, 16, 18, and 19 for teflubenzuron concentrations 0 to 1.6 mg/kg dry yeast, respectively.Data are presented as means.
(n) was the most sensitive trait, and adult survival (S a ) was the least sensitive.Note that values for t a , p j , and S a are ex- trapolated beyond the concentrations tested.

Population growth rate and elasticity analysis
The population growth rate decreased with increasing concentrations of teflubenzuron (Figure 5).At 0.32, 0.48, and 0.72 mg/kg dry yeast, the population growth rate was significantly lower than in the control group, whereas for the 1.08and 1.6-mg/kg dry yeast treatment groups, animals experienced complete reproductive failure, meaning that population went extinct for these cases.
The elasticities of juvenile survival (p j ), adult survival (S a ), and egg production (n) were positive, meaning that increasing each of these traits while holding all other traits constant would have a positive effect on population growth rate, whereas the elasticities of time to first oviposition (t j ) and time between broods (t a ) were negative, meaning that increasing these traits would have a negative effect on population growth rate (Figure 6).Population growth rate was most sensitive to relative changes in juvenile survival (p j ) and least sensitive to changes in time between broods (t a ; Figure 6).
The elasticities of trait t a and n were flat compared with p j and t j , which converged to 0 as concentrations of teflubenzuron increased.Elasticity of S a , on the other hand, increased slightly as concentration increased.L and G) and generalized linear models (GLM; t 0 ) are plotted as blue lines, with 95% confidence intervals shown as gray shades.The number of replicates is indicated in parentheses.The data for 1.6 mg/kg dry yeast were removed from the analysis due to only one replicate.

DISCUSSION
Individual-level effects of teflubenzuron on F. candida Our study showed that exposure to teflubenzuron under environmentally realistic concentrations affected all life-history traits of F. candida including lowered survival at high concentrations, reduced growth, delayed first oviposition, and reduced egg production and hatchability.Folsomia candida belongs to the arthropod group, and it relies on molting to grow (Fountain & Hopkin, 2005).Lack of chitin due to teflubenzuron exposure could lead to reduced molting frequency, as has been shown, for example, by reduced growth in Yuukianura szeptyckii (Collembola; Lee et al., 2019).A side effect of slower growth is delayed time to first oviposition.The F. candida in our control treatment took an average of 20 days after hatching to reach sexual maturity, which was within the normal range of this species (Fountain & Hopkin, 2005;Hopkin, 1997).The F. candida in the 0.72-mg/kg dry yeast teflubenzuron treatment were approximately 10 days late compared with the control.Delayed sexual maturity affects lifetime reproductive success, as well as potential survival because an immature individual is more vulnerable to predation for a longer period.
It is possible that reduced reproduction and hatchability are related to the lack of chitin.Chitin is the main component of the cuticle of F. candida, provides the building materials for egg shells, and is needed by embryos to complete their development and hatching (Nickerl et al., 2014;Rezende et al., 2016;Vargas et al., 2021).In addition, the reduced reproduction could be the consequence of shifts in energy usages, such that resources are used for survival rather than reproduction.For both reasons, F. candida may produce fewer eggs and more vulnerable eggs than normal.We observed a high fraction of egg clutches overgrown by mold in individuals exposed to high concentrations of teflubenzuron (Supporting Information S1, Figure S3).The lack of hydrophobicity may induce mold growth due to excessive moisture.Whether this was directly caused by exposure to the chemical requires further study to confirm.
Although our findings align with previous studies on the effects of teflubenzuron on Collembola, it should be noted that other nontarget arthropods, for example, lobsters, locusts, and shrimps, may respond differently to teflubenzuron, including deformity of various organs, in addition to reduced fecundity   ( Acheuk et al., 2012;Cresci et al., 2018;Olsvik et al., 2019;Samuelsen et al., 2014).It is assumed that the effects of teflubenzuron mentioned above can be related to premature molting, which is the direct consequence of reduced chitin content (Schmid et al., 2021).

Toxic effects of teflubenzuron on different life-history traits
By comparing the toxicity values across different endpoints, cumulative egg production and hatchability were both found to be the most sensitive, with estimated maximum body length being the least sensitive of the endpoints that we measured.Other studies (Table 1) that investigated the effect of teflubenzuron on Collembola primarily focused on survival and reproductive endpoints, but not developmental endpoints such as body length or time to first oviposition, as was done in the present study.Previous investigations identified juvenile number as the most sensitive measure (Campiche et al., 2006;Fernandes et al., 2023;Lee et al., 2019).Additionally, one study assessed survival, cumulative egg production, and juvenile size as endpoints, highlighting cumulative egg production as the most sensitive endpoint (Guimarães et al., 2019).The reason that survival was not the least sensitive endpoint in the present study is that all individuals were used to estimate the LC50, whereas only individuals where we could obtain at least seven consecutive body length data points to fit the logistic growth model were used to estimate the EC50 of maximum body length.Hence, only a fraction of the individuals that eventually died during the test were included in estimation of growth and this would tend to bias the EC50 for growth toward less sensitive individuals.Across studies, reproductive endpoints are consistently found to be more sensitive than survival despite differences in exposure route, exposure duration, animal age, and species (Campiche et al., 2006;Fernandes et al., 2023;Guimarães et al., 2019;Lee et al., 2019).

Exposure route
Earlier research reports the LC50 of teflubenzuron in F. candida ranging from a minimum of 0.1 (Guimarães et al., 2019) to a maximum of 0.35 mg/kg dry soil (Fernandes et al., 2023), all evaluated using OECD (2016b) test guideline 232.These values are approximately 2.5 to 8.7 times lower than ours, which implies that F. candida exposed through food may accumulate less teflubenzuron than through soil exposure.Compared with constant exposure via soil, whereby F. candida would be exposed to teflubenzuron continuously in multiple ways, including dermal contact, ingestion of contaminated soil particles, and absorption of contaminated porewater (Ogungbemi & van Gestel, 2018), in our study, they would be exposed to teflubenzuron when they ate and when they walked on the contaminated filter paper.It is worth considering that if the test is carried out with natural soil, greater effects may be a result of the natural fungi in the soil, because they contain chitin in their cell walls.Collembola feed on fungi in nature and may suffer more if food resources are also affected.If artificial soil is used, this compounded effect may not be as significant, because their primary food source would be the same yeast as in our study.
Teflubenzuron accumulation through dietary exposure can be influenced by several factors.First, different life stages of F. candida require varying amounts of energy, which affects their feeding frequency.It has been seen that newly hatched F. candida explore the environment rather than start feeding right away (Hamda, 2014).However, regardless of life stage, the amount of food eaten is proportional to body size, and hence the amount of teflubenzuron accumulated/unit body weight is expected to remain the same during the life of individual F. candida.Second, we cannot rule out the possibility that teflubenzuron directly impacts feeding rate by making the yeast unpalatable.If teflubenzuron affects feeding rate, F. candida will eat less, resulting in a slow growth rate, but also reduce their accumulation of teflubenzuron.These two effects may counter each other, but they are difficult to tease apart because it is difficult to quantify the animals' consumption of contaminated yeast.We did confirm that teflubenzuron accumulated in F. candida increased with exposure concentrations (Supporting Information S1, Table S1 and Figure S2).We also saw that concentration of teflubenzuron in animals did mirror the measured concentrations in the contaminated yeast, which is consistent with the concentration-dependent patterns we observed for toxic effects on multiple endpoints.

Population-level effect of teflubenzuron on F. candida
The population growth rate of F. candida was significantly lower in the 0.32-mg/kg dry yeast and above treatments than in the control, but it remained greater than one until the populations went extinct at the two highest concentrations due to no eggs hatching successfully.To our knowledge, comparable data on the effect of teflubenzuron at the population level are not available.One study focused on the abundance of Folsomia quadrioculata after application of a 10× recommended dose of diflubenzuron in a field experiment (Beck et al., 2004).Diflubenzuron and teflubenzuron are in the same chemical group of CSIs, that is, benzoylurea (Sun et al., 2015).The field experiment of Beck et al. (2004) showed that the abundance of F. quadrioculata had not recovered 1.5 years after the one-time application (Beck et al., 2004).This finding implies that chemicals alike to teflubenzuron could threaten the population of Collembola.
The ERA of chemicals tends to focus on the endpoints that are the most toxicologically sensitive (Forbes et al., 2010).This ignores the fact that the same effects (e.g., of chemicals) on different life-history traits can result in very different impacts at the population level because traits vary in their elasticity.A small change in a highly elastic trait could have a larger impact on population growth than a large change in an inelastic trait.In our study, egg production (n) was the most toxicologically sensitive trait, but its elasticity was the smallest; juvenile survival (p j ) was relatively insensitive to teflubenzuron, but highly elastic.A negative relationship between toxicological sensitivity and elasticity of life-history traits seems to be fairly common, and there are good evolutionary reasons for this (Forbes et al., 2010;Pfister, 1998).Variability in population growth rate (i.e., fitness) increases the risk of extinction.Traits that are more important for fitness (i.e., highly elastic traits) would be selected to be less variable in response to environmental changes (Pfister, 1998).Due both to differences in elasticity among life-history traits and to the nonlinear relationship between traits and population growth rate, using single traits (such as survival, growth, or reproduction) as simple proxies for population-level impacts can be misleading depending on the species used for ERA.In the example of F. candida, the number of eggs produced is generally high (even in exposed animals), which is why reductions in this trait do not have a large impact on population growth rate, whereas reductions in juvenile survival and time to first reproduction do.We propose that integration of these individuallevel responses into population models can provide more robust estimates of risk to populations and improve mechanistic understanding of the consequences of chemical exposure at ecologically relevant scales (Forbes et al., 2006(Forbes et al., , 2010)).

CONCLUSIONS
Using realistic exposure concentrations, we found that teflubenzuron negatively affected all F. candida's life-history traits and that the reproductive endpoints were more sensitive than survival and growth.The negative effects on the reproductive traits accumulated into negative population effects depending on the concentrations.At the last two highest concentrations, the F. candida population went extinct due to complete reproductive failure.In such cases, modeling of the full life cycle will not be needed.Our study has provided comprehensive insights into the toxic effects of teflubenzuron on an individual level, highlighting the potential risks to the population.This knowledge can aid in the development of more precise and accurate regulations for the use of CSIs, thereby ensuring the safety of our environment and its inhabitants.
Supporting Information-The Supporting Information is available on the Wiley Online Library at https://doi.org/10.1002/etc.5850.

FIGURE 1 :
FIGURE 1: Kaplan-Meier survival estimates of Folsomia candida exposed to teflubenzuron via diet.The contaminated food was renewed weekly.The sample sizes of each concentration were 21, 20, 23, 22, 16,  18, and 19  for teflubenzuron concentrations 0 to 1.6 mg/kg dry yeast, respectively.Data are presented as means.

FIGURE 3 :
FIGURE 3: (A) Time to first oviposition of Folsomia candida exposed to teflubenzuron via food.We had only one replicate for 1.08 and 1.6 mg/kg dry yeast, and therefore they are excluded from the analysis.(B) Cumulative egg production for each replicate under different treatments.(C) Hatchability of F. candida eggs exposed to teflubenzuron via food.Data are plotted as gray dots, and the mean value is shown as a solid black dot.The fitted generalized linear models were plotted as blue lines, with 95% confidence intervals.The number of replicates is shown in parentheses.No eggs hatched at 1.08 and 1.6 mg/kg dry yeast.

FIGURE 2 :
FIGURE 2: Mean of the parameters extracted from the logistic growth models of each individual Folsomia candida as shown by solid black dots.The raw data points are shown as gray dots.The fitted linear models ( ∞L and G) and generalized linear models (GLM; t 0 ) are plotted as blue lines, with 95% confidence intervals shown as gray shades.The number of replicates is indicated in parentheses.The data for 1.6 mg/kg dry yeast were removed from the analysis due to only one replicate.

FIGURE 4 :
FIGURE 4: The median lethal and effective concentration values of teflubenzuron estimated for five parameters used in the two-stage model, that is, egg production (n), time to first oviposition (t j ), time between broods (t a ), juvenile survival (p j ), and adult survival (S a ).The error bars indicate the upper range of the 95% confidence interval.

FIGURE 5 :
FIGURE 5: Estimated population growth rate of Folsomia candida (λ) under dietary exposure to teflubenzuron.The population growth rate cannot be estimated at concentrations 1.08 and 1.6 mg/kg dry yeast because no eggs hatched successfully.The error bars represent the 95% confidence interval.

FIGURE 6 :
FIGURE 6: Elasticity of egg production (n), adult survival (S a ), juvenile survival (p j ), time between broods (t a ), and time to first oviposition (t j ) of Folsomia candida at different concentrations of teflubenzuron via dietary exposure.To aid in visualization, parameters shown as filled symbols indicate that reductions in them reduced population growth rate, whereas in parameters shown as nonfilled symbols, reductions in them increased the population growth rate.

TABLE 1 :
Median lethal and effective concentration (LC50 and EC50) values of teflubenzuron in the present and other studies using Collembola Values are estimated toxicity values and their 95% confidence intervals in parentheses.The most sensitive trait in each study is shown in bold.