Abteilung Umwelttoxikologie

We applied machine learning methods to predict chemical hazards focusing on fish acute toxicity across taxa. We analyzed the relevance of taxonomy and experimental setup, showing that taking them into account can lead to considerable improvements in the classification performance. We quantified the gain obtained throught the introduction of taxonomic and experimental information, compared to classification based on chemical information alone. We used our approach with standard machine learning models (K-nearest neighbors, random forests and deep neural networks), as well as the recently proposed Read-Across Structure Activity Relationship (RASAR) models, which were very successful in predicting chemical hazards to mammals based on chemical similarity. We were able to obtain accuracies of over 93% on datasets where, due to noise in the data, the maximum achievable accuracy was expected to be below 96%. The best performances were obtained by random forests and RASAR models. We analyzed metrics to compare our results with animal test reproducibility, and despite most of our models "outperform animal test reproducibility" as measured through recently proposed metrics, we showed that the comparison between machine learning performance and animal test reproducibility should be addressed with particular care. While we focused on fish mortality, our approach, provided that the right data is available, is valid for any combination of chemicals, effects and taxa.

Understanding the ability of fish intestinal cells to act as a barrier for nanoparticle (NP) uptake and their effects is of significance from an environmental perspective as well as for human health, for which fish serves as an important nutrient source. We used an in vitro intestinal barrier model, based on rainbow trout intestinal (RTgutGC) cells, to elaborate the toxicity and translocation of five types of metal-based NPs. The NPs were polyvinylpyrrolidone (PVP)-coated Ag NPs, uncoated Ag NPs, CuO NPs, ZnO NPs and TiO₂ NPs. In conventional monolayers on impermeable supports, cell viability declined according to classical sigmoidal dose-response curves with EC50 values between 0.28 mg L^-1 and >100 mg L^-1 in the following rank order, from the most toxic (lowest EC50) to the least toxic (highest EC50): PVP–Ag NPs < uncoated Ag NPs < CuO NPs < ZnO NPs < TiO₂ NPs. When cells were cultured on permeable membranes to mimic an intestinal lumen and a blood-facing side, however, a much higher resistance of the cells towards NP-induced stress was noted with little to no impact on cell viability or barrier integrity. Yet, increased levels of Ag, Cu and Zn but not Ti were measured in the blood-side mimicking (basolateral) compartment, indicating translocation of Ag, Cu, and Zn-based NPs or ions liberated from them through the epithelial cell layer. Since especially CuO NPs appeared to be translocated as intact particles, they were investigated in more detail. A time- and temperature-dependent analysis, involving different endocytosis inhibitors, suggested that CuO NPs were translocated through the epithelium by apical caveolae-mediated endocytosis followed by delayed export onto the basolateral side. These data give valuable insights into NP uptake by, and translocation through, the fish intestinal epithelium and will be of value for future research on the molecular mechanisms of NPs that enter the fish via this critical uptake route.

Land-use changes often have significant impact on the water cycle, including changing groundwater/surface-water interactions, modifying groundwater recharge zones, and increasing risk of contamination. Surface runoff in particular is significantly impacted by land cover. As surface runoff can act as a carrier for contaminants found at the surface, it is important to characterize runoff dynamics in anthropogenic environments. In this study, the relationship between surface runoff and groundwater recharge in urban areas is explored using a top-down water balance approach. Two empirical models were used to estimate runoff: (1) an updated, advanced method based on curve number, followed by (2) bivariate hydrograph separation. Modifications were added to each method in an attempt to better capture continuous soil-moisture processes and explicitly account for runoff from impervious surfaces. Differences between the resulting runoff estimates shed light on the complexity of the rainfall–runoff relationship, and highlight the importance of understanding soil-moisture dynamics and their control on hydro(geo)logical responses. These results were then used as input in a water balance to calculate groundwater recharge. Two approaches were used to assess the accuracy of these groundwater balance estimates: (1) comparison to calculations of groundwater recharge using the calibrated conceptual HBV Light model, and (2) comparison to groundwater recharge estimates from physically similar catchments in Switzerland that are found in the literature. In all cases, recharge is estimated at approximately 40–45% of annual precipitation. These conditions were found to closely echo those results from Swiss catchments of similar characteristics.

A null‐space Monte‐Carlo (NSMC) approach was applied to account for uncertainty in the calibration of the hydraulic conductivity (K) field for a three‐dimensional groundwater flow model of a major water supply system in Switzerland. The approach generates different parameter realizations of the K field using the pilot point methodology. Subsequently, particle tracking (PT) was applied to each calibrated model, and the resulting particles are interpreted as the spatial pathline density distribution of multiple sources. The adopted approach offers advantages over classical PT which does not provide a means for treating uncertainty originating from the incomplete description of subsurface heterogeneity. Uncertainty in the K field is shown to strongly influence the spatial pathline distribution. Pathline spreading is particularly evident in locations where the information content of the head observations does not sufficiently constrain the estimated parameters. Despite the predictive uncertainty, the pumped drinking water at the study site is most likely dominated by artificially‐infiltrated groundwater originating from the local infiltration canals. The model suggests that within the well field, the central pumping wells could be extracting regional groundwater, although the probability is relatively low. Nevertheless, a rigorous uncertainty assessment is still required since only a few realizations resulted in flow paths that support the field observations. Model results should therefore not be based on only one model realization; rather, an uncertainty analysis should be carried out to provide a sufficiently large suite of equally probable simulations that include all potential sources and pathways.

An in vitro model of the fish intestine is of interest for research and application in diverse fields such as fish physiology, aquaculture and chemical risk assessment. The recently developed epithelial barrier model of the fish intestine relies on the RTgutGC cell line from rainbow trout (Oncorhynchus mykiss), cultured in inserts on permeable membranes. Our aim was to extend the current system by introducing intestinal fibroblasts as supportive layer in order to reconstruct the epithelial–mesenchymal interface as found in vivo. We therefore initiated and characterized the first fibroblast cell line from the intestine of rainbow trout, which has been termed RTgutF. Co-culture studies of RTgutGC and RTgutF were performed on commercially available electric cell substrate for impedance sensing (ECIS) and on newly developed ultrathin, highly porous alumina membranes to imitate the cellular interaction with the basement membrane. Cellular events were examined with non-invasive impedance spectroscopy to distinguish between barrier tightness and cell density in the ECIS system and to determine transepithelial electrical resistance for cells cultured on the alumina membranes. We highlight the relevance of the piscine intestinal fibroblasts for an advanced intestinal barrier model, particularly on ultrathin alumina membranes. These membranes enable rapid crosstalk of cells cultured on opposite sides, which led to increased barrier tightening in the fish cell line-based epithelial–mesenchymal model.
 

We studied the role of the fish intestine as a barrier for organic chemicals using the epithelial barrier model built on the rainbow trout (Oncorhynchus mykiss) intestinal cell line, RTgutGC and the newly developed exposure chamber, TransFEr, specifically designed to work with hydrophobic and volatile chemicals. Testing 11 chemicals with a range of physicochemical properties (logK_OW: 2.2 to 6.3, logHLC: 6.1 to 2.3) and combining the data with a mechanistic kinetic model enabled the determination of dominant processes underlying the transfer experiments and the derivation of robust transfer rates. Against the current assumption in chemical uptake modeling, chemical transfer did not strictly depend on the logK_OW but resulted from chemical-specific intracellular accumulation and biotransformation combined with paracellular and active transport. Modeling also identified that conducting elaborate measurements of the plastic parts, including the polystyrene insert and the PET filter, is unnecessary and that stirring in the TransFEr chamber reduced the stagnant water layers compared to theoretical predictions. Aside from providing insights into chemical uptake via the intestinal epithelium, this system can easily be transferred to other cell-based barrier systems, such as the fish gill or mammalian intestinal models and may improve in vitro–in vivo extrapolation and prediction of chemical bioaccumulation into organisms.

Nanomaterial risk governance requires models to estimate the material flow, fate and transport as well as uptake/bioavailability, hazard and risk in the environment. This study assesses the fit of such available models to different stages during the innovation of nano-enabled products. Through stakeholder consultations, criteria were identified for each innovation stage from idea conception to market launch and monitoring. In total, 38 models were scored against 41 criteria concerning model features, applicability, resource demands and outcome parameters. A scoring scheme was developed to determine how the models fit the criteria of each innovation stage. For each model, the individual criteria scores were added, yielding an overall fit score to each innovation stage. Three criteria were critical to stakeholders and incorporated as multipliers in the scoring scheme; the required time/costs and level of expertise needed to use the model, and for risk assessment models only, the option to compare PEC and PNEC. Regulatory compliance was also identified as critical, but could not be incorporated, as a nanomaterial risk assessment framework has yet to be developed and adopted by legislators. In conclusion, the scoring approach underlined similar scoring profiles across stages within model categories. As most models are research tools designed for use by experts, their score generally increased for later stages where most resources and expertise are committed. In contrast, stakeholders need relatively simple models to identify potential hazards and risk management measures at early product development stages to ensure safe use of nanomaterials without costs and resource needs hindering innovation.

The Morphological Quality Index (MQI) and the Morphological Quality Index for monitoring (MQIm) have been applied to eight case studies across Europe with the objective of analyzing the hydromorphological response to various restoration measures and of comparing the results of the MQI and MQIm as a morphological assessment applied at the reach scale, with a conventional site scale physical-habitat assessment method. For each restored reach, the two indices were applied to the pre-restoration and post-restoration conditions. The restored reach was also compared to an adjacent, degraded reach. Results show that in all cases the restoration measures improved the morphological quality of the reach, but that the degree of improvement depends on many factors, including the initial morphological conditions, the length of the restored portion in relation to the reach length, and on the type of intervention. The comparison with a conventional site scale physical-habitat assessment method shows that the MQI and MQIm are best suited for the evaluation of restoration effects on river hydromorphology at the geomorphologically-relevant scale of the river reach.

Long-term arsenic exposure poses serious health hazards. A 4-year project investigated the extent and magnitude of groundwater arsenic contamination in Burkina Faso. Low-cost arsenic removal filters were also tested and arsenic uptake pathways via food explored.

Transfer of compounds across cellular barriers is a critical step of compound uptake into organisms. Using in vitro barrier systems to evaluate such transfer is attractive because of the higher throughput and reduced resource needs compared to animal studies. Thus far, however, studying the transfer of hydrophobic and volatile compounds was hampered by the unavailability of in vitro exposure systems that allow for stable and predictable chemical exposure concentrations. To overcome this limitation, we constructed a novel exposure chamber, TransFEr, and tested it with an in vitro epithelial barrier model using the rainbow trout (Oncorhynchus mykiss) intestinal cell line, RTgutGC. Key features of the chamber are its closed design and rotatable silicon segments, which can serve for chemical dosing and sampling. Using the fragrance damascone beta (log K_OW: 3.7, log HLC: −3.9) as a pilot chemical, we were able to demonstrate that our exposure chamber provides for stable chemical exposure concentrations and full mass balance. The RTgutGC epithelium served as barrier for damascone beta transfer, which we attribute to chemical retention and biotransformation in the intestinal cells. Nevertheless, substantial transfer of the chemical across the epithelium occurred. When a chemical sink, i.e. a silicon segment, was included in the basolateral chamber to mimic blood constituents binding in vitro, transfer was about three-fold enhanced. We suggest that the presented methodology can help to obtain insights into chemical uptake mechanisms via the intestinal or other epithelia of fish and other animals for hydrophobic and volatile chemicals.

Permanent fish cell lines constitute a promising complement or substitute for fish in the environmental risk assessment of chemicals. We demonstrate the potential of a set of cell lines originating from rainbow trout (Oncorhynchus mykiss) to aid in the prediction of chemical bioaccumulation in fish, using benzo[a]pyrene (BaP) as a model chemical. We selected three cell lines from different tissues to more fully account for whole-body biotransformation in vivo: the RTL-W1 cell line, representing the liver as major site of biotransformation, and the RTgill-W1 (gill) and RTgutGC (intestine) cell lines, as important environment-organism interfaces, which likely influence chemical uptake. All three cell lines were found to effectively biotransform BaP. However, rates of in vitro clearance differed, with the RTL-W1 cell line being most efficient, followed by RTgutGC. Co-exposures with α-naphthoflavone as potent inhibitor of biotransformation, assessment of CYP1A catalytic activity, and the progression of cellular toxicity upon prolonged BaP exposure revealed that BaP is handled differently in the RTgill-W1 compared to the other two cell lines. Application of the cell-line-derived in vitro clearance rates into a physiology-based toxicokinetic model predicted a BaP bioconcentration factor (BCF) of 909-1057 compared to 920 reported for rainbow trout in vivo.