iNVERTOX: Rapid intelligent in silico prediction of sub-lethal ecotoxicological effects in invertebrates following pharmaceutical exposure

The iNVERTOX project will be the first of its kind to discover and predict both phenotypic and molecular level effects of trace pharmaceutical residues on small, but ecologically critical invertebrate organisms living in UK and international freshwaters which are now impacted by human activity. Pharmaceuticals are widely recognised as bioactive contaminants in our environment having been measured globally at very low concentrations. They enter the environment predominantly following excretion of consumed human and animal medicines and have been shown to be resistant to wastewater treatment. This leads to their consistent and prolonged infusion into receiving water catchments. Recently, three pharmaceuticals were placed on a "watch-list" of emerging priority pollutants following extensive studies of their toxicity to biota. However, the occurrence and diversity of pharmaceuticals contamination in the environment extends much further, with significantly more compounds detected in river water, sediments, soils and recently even in environmental species at any one time. Therefore given the scale of this problem, measurement of their effects on our environment is far too slow and laborious. More innovative and rapid approaches are required to understand and mitigate any effects these may have on our environment. Realistically, this must now involve some form of advanced computational modelling to use the limited information we have to predict the effects of additional pharmaceuticals. Moreover, traditional ecotoxicity testing for micro-pollutants use lethal doses and in the case of pharmaceuticals, these are often much higher than measured environmental concentrations. This suggests that more subtle effects need to be researched instead as a more accurate assessment of risk. In some cases, such small changes have resulted in a significant ecosystem imbalance which has indirect effects on wildlife, our environment and potentially also on human health. These so called, "sub-lethal phenotypic effects" are often more difficult to determine and establishing defined links to a pharmaceutical exposure is extremely challenging. The aim of this project is to study and model four sub-lethal phenotypic effects on a model freshwater benthic invertebrate species (Gammarus pulex) including growth rate, feeding rate, ventilation and locomotion following controlled exposure to low doses of over 60 pharmaceuticals typically found in the aquatic environment. In addition to this, changes in the organism at a molecular level will form a novel, central focus and enable knowledge discovery of how biota respond to such exposures at a fundamental level. This will be achieved via metabolomics, which is the measurement of thousands of small molecules present in a biological system following exposure to environmental contaminants. Lastly, and most importantly, this information will be used to build an set of advanced computational models using new machine learning tools to rapidly allow a user to screen potential phenotypic and molecular level effects of a pharmaceutical on biota in silico and minimise or remove the need for extended use of animals in ecotoxicity testing for this purpose. This project will therefore be pioneering in its approach and draw together the best academic and industry expertise from King's College London, The Francis Crick Institute, London and a global leader in pharmaceuticals, AstraZeneca, to rapidly and responsibly understand the effects of pharmaceuticals on environmental organisms.

This project will generate groundbreaking knowledge on the subtle effects of pharmaceuticals in the environment on a model freshwater benthic invertebrate, Gammarus pulex. As excellent indicators of surface water quality, these species are consistently impacted by pharmaceuticals and their metabolites at the ng-ug/L level mainly via sewage treatment plant effluents. Non-lethal phenotype-level effects, metabolomics studies and analytical measurements of >60 pharmaceuticals in G. pulex will be combined to generate biologically-inspired artificial neural networks and/or support vector machine models for rapid prediction of ecotoxicity from molecular level changes. In particular, models will be used to (1) predict growth rate, feeding rate, ventilation and locomotion effects; (2) identify metabolic pathways affected by pharmaceuticals; and (3) reduce the number of animals required for ecotoxicity testing in the future. The project will house five work packages (WPs): (1) Bioanalytical methods for G. pulex; (2) Pharmaceutical exposures and non-lethal effect measurement; (3) Metabolomics of exposed G. pulex and pharmaceutical residue measurement in biota; (4) Machine learning methods to model metabolomics/chemical measurement datasets to predict sub-lethal effects and/or affected pathways; and (5) Bioevaluation of novel biomarkers of exposure to pharmaceuticals. Metabolite/chemical analysis will be performed using gas and liquid chromatography coupled to (high resolution) mass spectrometry. Correlations with phenotypic effects will be identified using, for example, principal component analysis, Volcano plots and Z-transformation to rapidly identify dependent biomarkers. Linkage to pharmaceutical exposure will be built-in to models via internal pharmaceutical concentrations. Lastly, and in reverse, the prediction of molecular level changes will be investigated from quantitative structure-activity relationships and phenotype data for biomarker discovery and read-across.

This project will make a groundbreaking and timely contribution to environmental protection efforts by rapidly advancing the understanding of the effects of emerging contaminants such as pharmaceuticals and their metabolites on biota. Therefore, economic, academic and societal impact will lie in a novel and rapid approach to animal health assessment in UK and international rivers. The main beneficiaries are the pharmaceutical, chemical and environmental protection industries, policy-makers & regulators, healthcare (including third sector research institutions), the general public, early career researchers/students and academia in environmental and analytical science related fields. Academically, this is a timely issue to address now as the analytical technology and existing knowledge of this issue has advanced to a level which enables this groundbreaking work to be done. Combining two world-leading research institutions (King's and the Crick) with AstraZeneca as an IPA partner who prioritises and leads environmental health activity in this industry, our track record and expertise in this field will be focussed on delivering impact mainly across the pharmaceutical sector to inform practice and eventually policy. Firstly, the use of a robust in silico effects predictor will result in benefits to industry given that experimental determination is likely to be impractically long, complicated and at significant cost. Eventually, and if transferable across species, this may also enable better prioritisation of risk assessment strategies for both new and existing compounds as well as horizon scanning in drug discovery to ensure environmental effects are considered from the start. Long term, in silico predictors may enable environmental regulators to proactively rather than retroactively manage protection strategies in parallel with the pharmaceutical industry. As pharmaceuticals are organic molecules, in silico predictors may then also be transferable to other organic contaminant classes which will benefit the chemical industry as a whole. Dissemination of the findings during the project lifetime will be achieved by publicly accessible peer-reviewed publications, participation in world-leading environmental, metabolomics and computational modelling conferences, seminars at local and national level and via media engagement activity. The latter will be used to engage the public to raise the awareness of the proper use and disposal of unused pharmaceuticals and how they may affect organisms living in our rivers. The optimised in silico models themselves will be made freely and publicly accessible via the KCL website accompanied by the raw data generated. Furthermore, datasets will be uploaded to recognised scientific databases (e.g. MassBank, MetaboLights and MetaCyc) to ensure they can be effectively used by other researchers. Training and career progression of the researchers will form a central point of impact. Staff will be trained in state-of-the-art facilities at world-leading institutions in advanced technical skills for metabolomics, ecology, analytical chemistry and computational modelling as well as highlighting and supporting career development opportunities. In particular, working in such a collaboration ensures development of highly desirable professional knowledge, skills and attributes under both academic and industrial guidance and support will be offered to research staff to plan and advance in their careers afterwards. Additionally, both existing and prospective students will have involvement across the project via local outreach, internship and Masters research project activity. Therefore the next generation of scientists will avail of the industrial partnership activity with support for career mapping. Ultimately, this project will benefit the environment and society as well as the UK economy.