Proteomic systems evaluation of the molecular validity of preclinical psychosis models compared to schizophrenia brain pathology.

A systems biology approach compared protein changes in anterior prefrontal cortex tissue from four glutamatergic rodent models of schizophrenia and from human post-mortem schizophrenia patients. Liquid chromatography-tandem mass spectrometry proteomic profiling identified five functional domains of the disease, including "development and differentiation," represented across all rodent models. Kernel-based machine learning quantified that the chronic phencyclidine (PCP) model most closely resembled schizophrenia brain changes for four of these domains. The authors argue this methodology helps evaluate which rodent model best recapitulates neuropathological features of schizophrenia, supporting efforts to link functional behavioral dimensions with distinct biological processes.

Study at a glance

Abstract

Pharmacological and genetic rodent models of schizophrenia play an important role in the drug discovery pipeline, but quantifying the molecular similarity of such models with the underlying human pathophysiology has proved difficult. We developed a novel systems biology methodology for the direct comparison of anterior prefrontal cortex tissue from four established glutamatergic rodent models and schizophrenia patients, enabling the evaluation of which model displays the greatest similarity to schizophrenia across different pathophysiological characteristics of the disease. Liquid chromatography coupled tandem mass spectrometry (LC-MSE) proteomic profiling was applied comparing healthy and "disease state" in human post-mortem samples and rodent brain tissue samples derived from models based on acute and chronic phencyclidine (PCP) treatment, ketamine treatment or NMDA receptor knockdown. Protein-protein interaction networks were constructed from significant abundance changes and enrichment analyses enabled the identification of five functional domains of the disease such as "development and differentiation", which were represented across all four rodent models and were thus subsequently used for cross-species comparison. Kernel-based machine learning techniques quantified that the chronic PCP model represented schizophrenia brain changes most closely for four of these functional domains. This is the first study aiming to quantify which rodent model recapitulates the neuropathological features of schizophrenia most closely, providing an indication of face validity as well as potential guidance in the refinement of construct and predictive validity. The methodology and findings presented here support recent efforts to overcome translational hurdles of preclinical psychiatric research by associating functional dimensions of behaviour with distinct biological processes.