T164. NETWORK CONNECTIVITY SUPPORTING REWARD LEARNING DIFFERENTIALLY DISRUPTED IN TREATMENT RESISTANT SCHIZOPHRENIA
Abstract Background It is estimated that one third of patients with schizophrenia fail to adequately respond to antipsychotic medication, termed ‘treatment-resistance’. This occurs despite adequate blockade of D2 receptors in the brain. The parsimonious options are that treatment resistance could arise through a failure of cognitive control over the dopaminergic dysfunction in the striatum; or has a different primary non-dopaminergic mechanism that isn’t targeted by current antipsychotics. Contemporary models suggest that schizophrenia is associated with reduced reward prediction errors (RPE) and consequent aberrant salience driven by increased dopamine levels that ‘drown out’ phasic signals. This causes positive symptoms and impaired reward learning. However, RPE signalling in treatment-resistant patients appears intact despite sub-optimal behavioural performance. It is therefore unclear how reward learning is impaired in these patients. Methods We investigated how reward learning is disrupted at the network level in 21 medicated treatment-responsive and 20 medicated treatment-resistant patients with schizophrenia compared with 24 healthy controls (HC). Participants learnt to associate one of two emotional faces with a reward during a reinforcement learning task in an MRI scanner. Functional MRI BOLD signal was extracted from four brain regions (fusiform cortex, amygdala, caudate and anterior cingulate cortex (ACC)) activated in response to face cues and RPEs. These formed a network of interacting brain regions supporting reward learning. Dynamic Causal Modelling assessed how effective connectivity between regions in this cortico-striatal-limbic network is disrupted in each patient group compared to HC. Connectivity was also examined with respect to symptoms and salience. Finally, cognitive control and the role of glutamate were assessed by relating top-down connectivity from the ACC with glutamate levels measured from the same region of the ACC. Results In responsive patients, there was enhanced top-down connectivity from the ACC to sensory regions (fusiform and amygdala) and reduced input to the caudate compared to HC. Increased top-down connectivity was inversely correlated with symptom severity and sensory-salience. This suggests the presence of an effective compensatory mechanism for unreliable sensory information in responsive patients. Resistant patients however showed normal network connectivity compared to HC except abnormal connectivity within the ACC. This supports an alternative, non-dopaminergic mechanism disrupting reward learning in this refractory group. Increasing connectivity from ACC to caudate was related to positive symptom severity and salience in this group. Moreover, ACC glutamate levels were related to key top-down connections in HC and responsive patients but were not related to any connections in resistant patients. This suggests that glutamate may not be modulating connectivity effectively in this network to exert cognitive control and update reward predictions. Discussion In summary, differential mechanisms underlie disrupted reward learning between responsive and resistant groups. Resistant patients show similar RPE signalling and network connectivity to HC suggesting their dopaminergic functioning is intact. Impaired glutamate function may present a key mechanism that disrupts reward learning – and why dopaminergic drugs are ineffective. This finding is important for developing new drugs (e.g. glutamatergic targets) and guiding treatment strategies (e.g. giving clozapine earlier) in resistant patients. Future research probing cognitive control mechanisms and glutamate function will be useful to elucidate this putative pathology in treatment resistance.