Development and application of a multi-component 3-D in vitro model for predictive pharmacokinetics of environmental pharmaceuticals in fish

背景：许多药物最终进入环境，制药行业需要评估其对非靶生物的风险。鱼类是受环境药物危害风险最高的生物群体。因此，重要的是可以预测鱼类的吸收和影响。我们已经开发了一个现实的体外鱼鳃细胞系统（FIGCS），来自鳟鱼鳃细胞，研究污染物的吸收，排泄和毒性1，2，3。FIGCS在半渗透性培养插入物上生长，并在顶部耐受水中培养，而基底外侧隔室保持在代表血液的培养基中。该系统可以减少88%的鱼类数量，并节省时间，空间和成本。在之前的BBSRC项目（BB/J500483/1; BB/K501177/1; BB/M009513/1）中，我们发现药物是通过多种过程的组合吸收的，并使用ML来推导出决定吸收的分子描述符4，5。我们的数据表明，摄取不能单独从pKa预测和生理解释存在。使用在FIGCS数据上训练的ML，我们发现可电离药物的摄取也取决于溶解度、log D和分子量4。除了pH值之外，水化学变量也可能影响药代动力学（PK）1。ML和FIGCS的组合使我们能够快速地从机械上理解化合物在鱼中的PK并对其进行计算建模。然而，PK模型受益于多个隔室（包括靶器官）的数据。通过创新的BBSRC案例研究（BB/J500483/1），我们建立了FIGCS可以与肝球状体共培养。拟议的奖学金将验证这些共培养物，并使用它们来确定模型药物的鱼类PK及其对结构和水化学的依赖性。在这个高度跨学科的项目中，学生将部署FIGCS-肝脏类器官共培养和分析化学，使用LC-MS测量60多种药物的摄取，分布，代谢和排泄。将对肝脏球体提取物进行可疑筛选分析，使用我们的ML辅助全扫描高分辨率质谱（HRMS）方法识别代谢物6，7。ML模型将在R，Python和许可平台中构建，以通过遗传特征选择算法和每个模型的敏感性分析来确定FIGCS对药物特性和水化学的依赖性，以实现共识。使用上述和文献中的测量数据，将为所有化合物同时开发时间序列型ML模型，用于计算机模拟PK预测。

BACKGROUND: Many medicinal drugs end up in the environment and the pharmaceutical industry is required to assess their risk to non-target organisms. Fish is the group of organisms of highest risk of harm from environmental pharmaceuticals. Thus, it important that uptake and effects in fish can be predicted. We have developed a realistic in vitro Fish Gill Cell System (FIGCS), derived from trout gill cells, to study uptake, excretion and toxicity of pollutants1,2,3. FIGCS are grown on semi-permeable culture inserts and tolerate culture in water apically while the basolateral compartment is kept in culture medium, representing the blood. This system can bring an 88% Reduction of the number of fish needed for testing for accumulation of chemicals, and savings in terms of time, space and costs. In previous BBSRC projects (BB/J500483/1; BB/K501177/1; BB/M009513/1) we found that pharmaceuticals are taken up by a combination of processes and used ML to derive molecular descriptors determining uptake4,5. Our data show that uptake cannot be predicted from the pKa alone and physiological explanations exist. Using ML trained on FIGCS data, we found that uptake of ionisable pharmaceuticals is also dependent on solubility, log D, and molecular weight4. It is also likely that water chemistry variables other than pH influence pharmacokinetics (PK)1.The combination of ML and FIGCS allow us to mechanistically understand and computationally model PK of compounds in fish rapidly. However, PK models benefit from data on multiple compartments, including target organs. Through an innovative BBSRC Case studentship (BB/J500483/1) we established that FIGCS can be co-cultured with liver spheroids. The proposed studentship will validate these co-cultures and use them to determine fish PK of model drugs and their dependency on structure and water chemistry. In this highly interdisciplinary project, the student will deploy FIGCS-liver organoid co-culture and analytical chemistry to measure uptake, distribution, metabolism and excretion for 60+ pharmaceuticals, using LC-MS. Suspect screening analysis will be performed on liver spheroid extracts to identify metabolites using our ML assisted full-scan high resolution mass spectrometry (HRMS) methods6,7. ML models will be built in R, Python and licenced platforms to determine FIGCS dependency on pharmaceutical properties and water chemistry by genetic feature selection algorithms and sensitivity analysis of each model to achieve consensus. Using measured data from above and from the literature, time-series type ML models will be developed for all compounds simultaneously for in silico PK prediction.