Synaptic and circuit mechanisms of working memory and their dysfunction in anti-NMDA receptor encephalitis and schizophrenia

Abstract

[eng] In this thesis, I investigate different synaptic and circuit mechanisms of working memory, how their interaction produces specific working memory biases, and how their disruption in psychiatric or neurological disease can contribute to abnormal working memory function. I show that PFC represents working memory contents not only in spiking, persistent activity, but also shows signatures of imprinted, synaptic traces of working memory. These traces can hold contents for an extended period of time, such as an ITI, without the need for firing rate-based maintenance. From locally facilitated synapses, stable working memory representations could be reactivated through unspecific network inputs, a result found in monkey PFC and indirectly in human EEG, and explained by a circuit model of PFC that exhibits bistability (i.e., stable, persistent activity) and is supported by a STP mechanism. Finally, memory reactivations as observed in monkey PFC, human EEG, and elicited with prefrontal TMS in humans increased systematic biases towards previous memories. These findings demonstrate the behavioral relevance as well as the prefrontal locus of the discussed mechanisms. I then designed experiments parallel to the first study to test the impact of NMDAR dysfunction on working memory precision and systematic serial biases in patients with anti-NMDAR encephalitis and patients with schizophrenia, and compared their data to that of healthy controls. Working memory precision in both patient groups was unaffected, but serial biases were drastically reduced in encephalitis patients, and completely disrupted in patients with schizophrenia. Moreover, biases normalized in patients with encephalitis, a sign of their relation to clinically relevant processes. By disrupting different NMDAR-related parameters in a prefrontal circuit model, I show that perturbations in E/I balance through reduced NMDAR-mediated currents cannot explain findings from patients. In contrast, reduced short-term potentiation successfully disrupted between-trial memory traces and the emergence of serial biases in the model. Finally, I tested whether the neural mechanism that underlies serial dependence in my first study would be disrupted in patients’ EEG. The findings validate the hypothesis generated in the second part of the thesis: While in healthy controls, memory representations from previous trials are reactivated during the ITI, and subsequently influence behavior, no such reactivation occurs in patients with anti-NMDAR encephalitis or schizophrenia. I show that the lack of memory reactivations is related to a weaker and less stable memory code during the delay period, and is explained by disrupted STP as proposed in the circuit model.