Supporting Information A: Elimination resistant: Characterizing multi-compartment toxicokinetics of the neonicotinoid thiacloprid in the amphipod Gammarus pulex using bioconcentration and receptor binding assays

Johannes Raths, Linda Schinz, Annika Mangold-Döring, Juliane Hollender* Department of Environmental Chemistry, Swiss Federal Institute of Aquatic Science and Technology Eawag, Überlandstrasse 133, 8600 Dübendorf, Switzerland Institute of Biogeochemistry and Pollutant Dynamics, ETH Zürich, Universitätstrasse 16, 8092 Zürich, Switzerland Department of Aquatic Ecology and Water Quality Management, Wageningen University, P.O. Box 47, 6700 Wageningen, The Netherlands


List of Tables
: Schedule of the online-SPE on the QExactive and the QExactive Plus mass spectrometers. S9 Table S3 Table S12: Comparison of the toxicokinetic parameters modelled for alive and dead gammarids ... S19

SI A1: Neonicotinoid insecticides
An overview of the seven commercially available neonicotinoid insecticides and nicotine is presented in Figure S1.

SI A2: Test animal acclimation
After transport to Eawag facilities, the gammarids were transferred to glass aquaria filled with river water from the sampling site. The aquaria were aerated and placed into an incubator programmed to 12:12 h light-dark cycles. During the acclimation period prior to the start of the experiments (≥ 3 d), the water in the aquaria was progressively replaced to 100 % (v/v) artificial pond water (APW) (pH = 7.9 ± 0.3, prepared according to Naylor et al. 2 ). In addition, the temperature in the incubator was increased gradually to 15.5 ± 1°C by the end of acclimation. The amphipods were fed ad libitum with the leaves collected at the sampling site. Juvenile, pregnant or animals visibly infected with acanthocephalan parasites were sorted out and not introduced into any experiment.

SI A3: Lipid content
The lipid content of gammarids was determined gravimetrically according to Raths et al. 3 . Samples were taken from the gammarid batch of experiments I to III (3 replicates, 4 gammarids each), and the batch for experiments V and VI (3 replicates, 4 gammarids each) shortly before (< 1 h) the corresponding experiments started. Prior to the lipid extraction, 1.5 mL glass vials were labelled incubated (90 °C, 2 h) and then weighed. Next, snap frozen gammarids were sampled into 2 mL Eppendorf vials and weighed. 300 mg of silica beads and 1 mL cyclohexane/isopropanol 5:4 (v/v) were added and samples homogenized in two runs (6 m s -1 , 15 s, FastPrep-24TM, MPITM). Next, 0.6 mL of nanopure water (NPW) were added, samples were vortexed (30 s) and centrifuged (1000 x g, 10 min, 4 °C). Then, the organic supernatant was transferred into the 1.5 mL glass vials using glass pasteur pipettes. Subsequently, 0.55 mL of cyclohexane/isopropanol 87:13 (v/v) were added to the residues and samples were vortexed (30 s) and centrifuged (1000 x g, 10 min, 4 °C) again. Then, the supernatant was once again removed and combined with the previous extract. The extraction solvent was then evaporated at 60 °C overnight and vials were weighed the next morning. Finally, the gammarid lipid content was calculated as the ratio of the lipid weight to the gammarid wet weight.

SI A4: Investigations on exoskeleton sorption
Previous research on gammarids and daphnids suggested sorption to the crustacean exoskeleton as a possible artefact in bioconcentration experiments. 4,5 Thus, the objective of this experiment was to determine the amount of thiacloprid bound to the exoskeletons of G. pulex. As muscular tissue of gammarids is very closely grown together with the exoskeleton, no sufficient dissectioning methods could be established. Instead, exuviae from moulted gammarids were collected as described earlier 5 .
Additionally, experiments where thiacloprid-exposed gammarids were washed with different detergents were performed.

Exuviae collection
Exuviae from moulted gammarids were collected during the kinetic experiment (TK kinetic experiment I, 50 μg L -1 ). The exuviae were sampled, extracted, and analysed in the same manner as other gammarid samples.
During the uptake phase, the shedded exoskeleton accounted for 1.7 to 2.4 % (m/m) of thiacloprid compared to the total body burden at the same time point (Table S1). During the elimination phase, the contribution increased to 6.6-17 %. It should be noted that the number of exuviae sampled was small, especially during the elimination phase (n = 1) leading to high uncertainty for the upscaling of the results. However, the elimination resistant fraction accounted for approximately one third of the total tissue concentration at the end of the uptake phase and would be assumed to account for the major proportion of the total tissue concentration at the end of the elimination phase. Thus, the exoskeleton bound fraction would not explain the elimination resistant residues observed during the toxicokinetic experiments and our findings were different from observations reported for daphnids. 4 Higher proportions of exoskeleton associated fractions (i.e. up to 24%) were reported by Miller et al. 5 for ionised pharmaceuticals where this mechanism may be of more relevance. In this experiment, gammarids (collected in October, together with experiment II and III) were exposed to 50 μg L -1 thiacloprid for two days and then transferred to uncontaminated medium for another two days. The amphipods were sampled in triplicates after two and four days, respectively. This experiment was performed with live and dead (heat shock-euthanised) gammarids. Each treatment was tested in a separate aquarium. Only the tank containing live gammarids was aerated and received leaf disks during the elimination phase. In contrast to the standard method, at the end of the experiment the amphipods were removed from the aquaria, rinsed with one of three solutions, and subsequently rinsed with nanopure water. Thereby, it was tested whether the detergents were able to remove thiacloprid potentially sorbed to the exoskeleton of G. pulex. The solutions were the detergent triton X-100, dimethyl sulfoxide (DMSO) and a commercially available soap (MayaEasy Soft). DMSO is a solvent often used to accelerate the dissolution of chemicals in toxicology and pharmacology. 6 Triton X-100, on the other hand, is a widely used surfactant for lysing cells and to make cells permeable. 7 The triton X-100 and DMSO solutions were prepared in 0.1% dilution in NPW. The soap detergent was produced by adding 1 drop of soap to 0.5 L of NPW. As a control treatment, gammarids were washed S6 twice with NPW. Afterwards, the gammarids were homogenized and extracted with methanol as described in the manuscript.
The internal concentrations in live gammarids at the end of the uptake (48 h) and elimination phase (96 h) rinsed with different detergents are presented in Figure S2. The internal concentrations of alive and dead gammarids washed with DMSO, triton X-100 and soap did not differ significantly from the control gammarids washed with NPW or each other (one-way ANOVA, p ≥ 0.05). All measured tissue concentrations were in the range of previous toxicokinetic experiments. A larger variance of data for dead gammarids was observed and may be explained by disintegration caused by microbial growth and slower kinetics in lifeless tissue (i.e. because of no sterilisation).
The detergent rinsing experiment demonstrated that the elimination resistant fraction could not be washed of the gammarids using any of the present solutions. Because some of the detergents are specifically used to remove organic contaminants sorbed to the surface of a sample it was concluded that sorption to the exoskeleton of amphipods would not explain the non-eliminating thiacloprid fraction.

Figure S2: Internal thiacloprid concentrations (mean ± SD, n = 3) in alive (left) and dead (right) gammarids washed with NPW, and triton X-100, DMSO and soap after 48 h of exposure (grey) and after 48 h of elimination (white).
Individual data points are presented as dots. No significant influence of washing solution was detected (one-way ANOVA, p ≥ 0.05).

SI A5: Protein content
The protein contents were determined in gammarids from the batch of experiments I to III (18 gammarids), and the batch of experiments V and VI (49 gammarids). Samples were analysed in sets of approximately 200 mg gammarid per vial (= 1 mL DM per 200 mg ww).
The complete (membrane) protein extraction workflow is illustrated in Figure S3. The total protein concentration was determined from the combined extracts SN1 and SN2. The present, adapted approach using a cooled pre-cooled tissue lyser recovered an approximately 20% higher total protein content compared to homogenization using an automatic pestle motor as described by Maloney et al. 8 Detailed methods on the Pierce BCA Protein Assay Kit (ThermoFisher Scientific): In short, 200 μL of the Working Reagent were added to wells of a 96-well plate. Afterwards, 25 μL of protein suspension or calibration standard were added to the wells. The calibration standards were produced by diluting bovine serum albumin (BSA) with dissociation medium as described in the BCA Protein Assay Kit manual. All standards and samples were prepared in duplicates. After adding standards and samples to the 96-well plate, the covered well plate (in a plastic bag to avoid evaporation) was incubated (30 min, 37 °C). The plate was cooled down to room temperature and read within 10 min after incubation using a plate reader (Tecan InfiniteR 200 PRO). Protein concentrations were then determined with a linear standard calibration curve. The R 2 of the calibration curve of all assays was ≥ 0.98. Protein suspensions were stored at -20 °C until performance of the ligand binding assays.
The membrane protein content and the protein content in the supernatant (FProt, in % wet weight) were calculated as the protein weight normalised by the gammarid wet weight: where CProt (mg L -1 ) is the protein concentration determined in the BCA assay, VExtract (L) the extraction volume (4 mL for membrane proteins and 7 mL for proteins remaining in the supernatant) and WGammarid (mg) the gammarid wet weight (ww). The total protein is the sum of the membrane protein and the supernatant protein content.
In a direct comparison, the optimised protein extraction methods adapted from Maloney et al. 8 provided a 50% higher total protein yield (3.8 ± 0.2% ww versus 2.5 ± 0.2% ww, mean ± SD, n = 2) than by our previous methods 3 which were based on Janssen et al. 9 . Figure S3: Illustration of the total protein and membrane protein extraction method. SN1 and SN2 designate the supernatants after the first and the second centrifugation run at 1000 x g.

SI A6: Online SPE LC-HRMS/MS settings
Thiacloprid concentrations were determined on three different mass spectrometers (Thermo Fisher Scientific Inc.), which are the QExactive (kinetic experiment I, concentration dependence II and III), QExactive Plus (in vivo and in vitro nAChR binding assays V and VI) and Exploris 240 (exoskeleton rinsing, pulsed exposure IV).
On the QExactive, thiacloprid was measured in polarity switching mode. On the QExactive Plus and the Exploris 240, samples were measured in positive ionisation mode. Additional information on the online SPE and LC-ESI-HRMS/MS settings are provided in Table S2 to Table S8. Peaks were integrated and concentrations were determined with the software TraceFinder 5.1 (Thermo Fisher Scientific Inc.), using a linear standard calibration and the THI/ISTD area ratio. Calibrants were weighted by the inverse of their quantity in the least squares optimisation. The R 2 of the calibration curves of all measurements was ≥ 0.99.
For the measurement on the QExactive and QExactive Plus, 20 mL of NPW and 100 μL of sample extracts were added to 20 mL glass vials. For the measurement on the Exploris, 100 μL of the samples were added to 10 mL of NPW.

SI A7: Online SPE LC-HRMS/MS quality control
Quality control parameters of the online SPE LC-HRMS/MS performance are presented in Table S9. The high matrix factors in the receptor binding assays are likely to be caused by phosphate salts, which originate from the dissociation medium, and were not completely removed in the online-SPE process. Limits of quantification were not limiting any of the gammarid extract and medium analysis. A suspect screening on biotransformation products of thiacloprid was performed based on reported biotransformation products in bacteria 10 . For this purpose, the by full scan acquired HRMS/MS data were screened for the exact masses [M+H] + of the reported biotransformation products. Additionally, a reference standard of the main biotransformation product thiacloprid amide was analysed in the calibration and relative recovery samples. No precursor ion of any biotransformation product was detected in gammarid samples, but thiacloprid amide was found in the calibration and recovery samples as expected.

SI A8: In vitro receptor binding assay -specific and unspecific binding model
The saturation curve of the specific binding model at concentrations of 1 to 25 nM (n = 4) is presented in Figure S4.

S13
The corresponding unspecific binding model and the calculated parameters are presented in Figure S5 and Table S10, respectively. The estimates of BMax and Kd were very similar (< 4% deviation) than the values calculated from the specific binding model.

SI A9: Details on the toxicokinetic-receptor model Assumptions
The in vivo and in vitro assays of this study showed evidence of an elimination resistance predominantly present in the membrane protein fraction. Moreover, the nACh receptors, as membrane proteins, are the molecular target of thiacloprid. 11 Due to this evidence, the underlying assumptions of this model are the following: 1) The organism is represented as a two-compartment system with a structure compartment representing the debris and supernatant fraction of the in vivo assay, and a membrane protein compartment representing the fraction of the membrane protein, including receptors. Both compartments are homogeneous (well-mixed) compartments. The external water (i.e., environment) compartment is also well-mixed, infinitely large, and not influenced by the organism.
2) The uptake flux of the chemical from the water into the structure compartment is proportional to the external water concentration, and the uptake flux of the chemical from the structure to the membrane protein compartment is proportional to the concentration in the structure compartment as well as the concentration within the membrane protein compartment (saturation effect).
3) The elimination flux of the chemical from the structure compartment to the environment is proportional to the concentration of the structure compartment. There is no elimination of the chemical from the membrane protein compartment.
4) The measured chemical concentrations in the membrane protein compartment approximate the concentration of ligand-bound receptors in the organism.

Symbols, parameters and equations
A summary of all model parameters and state variables used in the model development, including the best fit parameters of the calibration, is provided in Table S11. The final model equations are summarized in Eq. S3 to S5.

Model calibration
The plots of the model calibration are presented in Figure S6 (tissue concentrations) and Figure S7 (parameter space).   Table S11.

Model validation
The plots of the model validation are presented in Figure S8 and Figure S9.

SI A10: Impact of physiological activity on toxicokinetics
To differentiate between active (i.e. filtration, enzyme activity) and passive (i.e. diffusion) toxicokinetic processes, an additional setup, analogous to toxicokinetic experiment I, was conducted with dead gammarids. Gammarids were heat shock inactivated (few seconds at 53 °C) and exposed in a separate aquarium. Aeration and feeding were omitted to reduce bacterial growth that promotes disintegration of the dead gammarids. Due to disintegration, the elimination phase lasted only 6 days. For toxicokinetic receptor modelling the parameter kon was fixed to 200, as the dataset would not allow a proper estimation of the parameter.
The measured and modelled concentrations in dead gammarids are presented in Figure S10. At the end of the uptake phase, the total tissue concentration in dead gammarids was with 0.72 ± 0.03 µmol kg -1 slightly lower compared to 0.84 ± 0.11 µmol kg -1 in alive gammarids. At the end of the elimination phase, the tissue concentrations in dead gammarids were 0.30 ± 0.02 µmol kg -1 compared to 0.23 ± 0.02 in alive gammarids. The measured concentrations were similar to experiments for investigations on exoskeleton sorption (SI A4, exoskeleton rinsing) for both alive and dead gammarids. Despite slower kinetic rates, both the steady state concentration as well as the non-eliminating fraction of accumulate thiacloprid were on very similar levels between both alive and dead gammarids. The higher residual thiacloprid concentration in dead gammarids at the end of the elimination phase could potentially be caused by disintegration of the dead gammarids, leading to a lower sample weight (average reduction of 37% after two days of elimination).
The toxicokinetic parameters are compared with the parameters for alive gammarids in Table S12. The present experiment showed that a loss of physiological activity (as is the case in dead gammarids) had a considerable impact on the kinetic rates (strongest on ku, -50%) but much smaller effect (-20%) on the BCFkin,Structure or the elimination resistant fraction CBMax. A similar effect of physiological activity (i.e. filtration) on toxicokinetics was indicated earlier 3,12,13 . The heat shock inactivation of gammarids has shown to inhibit biotransformation enzymes 3 and would imply that also membrane proteins, such as nAChRs, denature. However, Saitoh et al. 14

S19
(few seconds). Thus, it appears likely that the nAChR in dead gammarids were still mostly intact and irreversible binding explains the elimination resistant thiacloprid fraction also in dead gammarids.

SI A11: Pulsed exposure
In the 5 µg L -1 treatment ( Figure S11), internal concentrations were only significantly higher after the third pulse compared to elimination phase samples (and 24 h exposure of the first pulse). The internal concentrations were slightly higher during the uptake phase (0.28 ± 0.06 µmol kg -1 , n = 18) than during the elimination phase (0.20 ± 0.03 µmol kg -1 ).
The observed pattern provides a more blurry vison on the pattern of the 50 µg L -1 pulsed exposure scenario. The low concentration after 24 h of the first pulse may be caused by the initial uptake (ku) into gammarid tissue limiting the transfer of thiacloprid from the exposure medium towards the membrane protein, as explained with the toxicokinetic-receptor model. The estimated time to steady state at 5 µg L -1 exposure is about 3 to 4 days. This may explain the slight increase in uptake phase concentration across the three pulses. S20 SI A12: Gammarid cross section A gammarid sagittal cross section of G. pulex indicating the high proportion of muscle tissue is presented in Figure S12.

SI A13: Possible receptor binding assay optimizations
Even though the sensitivity acquired by online-SPE LC-HRMS/MS measurements was sufficient for determination of the thiacloprid concentrations in the assay, time and resource requirements exceed the ones of the radio ligand assay. However, the present method would allow to distinguish between multiple ligands (i.e. competition assays) and potential transformation products. Furthermore, centrifugation was much more time consuming compared to filtration resulting in much lower sample throughput, higher workload, and potentially higher membrane protein loss. Due to the higher volumes required for ultracentrifugation, higher initial amounts of membrane protein were needed in order to obtain reasonable ratios of membrane protein and assay medium. The reduction of the assay to a microplate layout would minimise most of the drawbacks of the present method and should be considered for future applications for non-radioactively labelled compounds.