The shallow cognitive map hypothesis: A hippocampal framework for thought disorder in schizophrenia

Memories are not formed in isolation. They are associated and organized into relational knowledge structures that allow coherent thought. Failure to express such coherent thought is a key hallmark of Schizophrenia. Here we explore the hypothesis that thought disorder arises from disorganized Hippocampal cognitive maps. In doing so, we combine insights from two key lines of investigation, one concerning the neural signatures of cognitive mapping, and another that seeks to understand lower-level cellular mechanisms of cognition within a dynamical systems framework. Specifically, we propose that multiple distinct pathological pathways converge on the shallowing of Hippocampal attractors, giving rise to disorganized Hippocampal cognitive maps and driving conceptual disorganization. We discuss the available evidence at the computational, behavioural, network, and cellular levels. We also outline testable predictions from this framework, including how it could unify major chemical and psychological theories of schizophrenia and how it can provide a rationale for understanding the aetiology and treatment of the disease.


The Dopamine hypothesis
It is widely accepted that dopamine dysfunction plays a role in Schizophrenia 112  Prediction 2, we discussed how Hippocampal dysfunction can drive the hyperdopaminergic state. Below we discuss how the hyperdopaminergic state could contribute to Hippocampal dysfunction.
Within this Hippocampus, dopamine is a key regulator of synaptic and network plasticity, particularly in relation to novelty or reward prediction errors 113 . Dopamine acting on D1-receptors enhances LTP and can alter Hippocampal spike timing dependent plasticity (STDP) rules, for example by retroactively converting spike-timing dependent LTD to LTP 114 118 . Moreover, another recent study identifies dopaminergic locus coeruleus inputs to the dorsal CA1 as being critical for linking memories 119 .
Assessing the effect of dopamine receptor blockade, or optogenetic manipulation of dopaminergic afferents, on Hippocampal attractor dynamics in Schizophrenia models would shed further light on the relationship between dopaminergic and hippocampal dysfunction in thought disorder.

The GABA hypothesis
In addition to dysregulated dopamine, there is robust evidence to support an association between loss of  126 for confounding factors that suggest NMDAR hypofunction involves a multitude of neuron types). Moreover, Gamma oscillations, known to be deficient in Schizophrenia patients and models, are generated by a mechanism that involves feedback inhibition by fast-spiking PV interneurons (PVN) 127 . The importance of early disruption of these circuits and the role of PVN has also been highlighted in a recent study where schizophrenic behaviours (novel object recognition, sensory-motor gating in Paired pulse inhibition) in PV-Cre; ErbB4 fl/fl mice could be reversed by chemogenetic activation of PVN in the frontal association cortex 128  The relationship between plasticity and Schizophrenia may be more complex, with enhancements in LTP also reported in Schizophrenia models. For example, the Schizophrenia associated DTNBP1 gene encodes Dysbindin protein that plays a role in the neuronal trafficking of proteins including receptors. In dysbindindeficient mice, enhanced AMPAR responses and LTP were observed in CA3-CA1 Hippocampal synapses 51 .
Nevertheless, as discussed above, an enhanced potentiation, if it is not specific to neuronal inputs representing rewarded or salient states, may give rise to aberrant or dysregulated associations, which would result in the same net effect of shallowing attractors and disorganizing Hippocampal maps ( Figure   3).
Overall, these findings suggest that dysregulated dopamine, compromised interneuron function and aberrant plasticity at glutamatergic synapses could all contribute to the shallowing of Hippocampal attractors underlying cognitive map function in the Hippocampus. This framing is useful in two ways.
Firstly, it provides a unified framework for understanding the role of these distinct transmitter systems in Schizophrenia. Secondly it provides a rationale for understanding the heterogeneity that famously characterizes the aetiology of Schizophrenia. Multiple distinct pathways could achieve a similar end result: a shallow, disorganized cognitive map.

Supplementary Discussion 2 -Avenues for treatment
Our framework provides rationale for the following: 1-Prognosis: In addition to Schizophrenia patients, relatives of schizophrenics also score higher on thought disorder severity than non-psychiatric controls 84 . Identifying mutations and epigenetic changes in Hippocampal plasticity-related genes that predispose individuals to disease not only allows identifying atrisk individuals and families, but also may allow more precise prediction of their probable disease progression and predisposition to thought disorder. Example: Using model-based approaches to classify individuals as at risk of developing thought disorder based on the proximity of their genetic profiles to those of Schizophrenia datasets associated with this symptom, guided by functional data linking such genes to shallowing of Hippocampal cognitive maps. Example: exercise 134 and environmental enrichment 128,135 can reverse Schizophrenia symptoms in animal models by enhancing interneuron neurogenesis or function in the dentate gyrus respectively. This provides a mechanistic understanding of exercise-induced improvements in Hippocampal function in Schizophrenia patients 136 . Our framework explains such an effect as a result of rescuing imbalance of PS/PC by enhancing the former, thereby promoting the formation of organized, deep cognitive maps. Combining such behavioural therapies with pharmacological treatments that target PV interneurons could provide a robust therapeutic strategy, especially for non-responders to dopaminergic treatments.

Supplementary Discussion 3 -Additional Detailed Experimental Predictions
1-Impaired replay in Schizophrenia is the result of "jumps" between Hippocampal sequences representing distinct memories: e.g., in Schizophrenia mouse models experiencing two or more linear tracks, awake replay (in CA3 and CA1) in one track should exhibit a greater intermixing of non-random fragments of sequences from multiple tracks (Figure 1b).
2-Jumps between sequences for distinct memories may correlate with enhanced performance on tasks that require associating distinct memories, but should also impair tasks that necessitate suppressing memory interference. This pattern should be seen in human patients at early or subclinical stages of the disease (e.g., at risk relatives of patients).
3-Is thought disorder associated with dysregulated encoding or retrieval of relational maps?
An impairment in encoding of relational structures (e.g. due to excessive associations driven by aberrant salience) would involve anterograde deficits restricted to relational structures learned after disease onset.
Conversely, retrograde deficits would indicate some level of retrieval impairment too. However, this is complicated by the difficulty in determining disease onset and the mounting evidence for premorbid cognitive deficits (e.g. 137 ). It is also the case that memories are subject to modification both offline and when explicitly retrieved (memory reconsolidation 138 ), which would mean that even an anterograde deficit could affect relational structures learned prior to disease onset. This could be resolved by more controlled experiments in which acute models of thought disorder are investigated (perhaps those involving ketamine induced psychosis, which is known to include thought disorder 139 ). This would allow precise determination of impairment onset and better control of memory processing. 4-Molecular changes that impair pattern separation, enhance pattern completion or induce aberrant salience should have a similar end result of shallowing Hippocampal attractors in the CA3, causing jumps during awake replay (see point 1) and predisposing subjects to thought disorder (Figure 3). Functional grouping of genes based on their roles in deepening hippocampal attractors should allow a more accurate prediction of patient symptoms and prognosis than functional grouping based on lower level processes (e.g. LTP, interneuron function...etc). This may emphasize genes with promotors/regulatory elements that are preferentially transcriptionally activated in the Hippocampus.
5-Molecular changes that give rise to either impaired LTP or enhanced yet dysregulated LTP at synapses onto CA1 (including CA3-CA1 and nucleus reuniens-CA1 synapses) should effectively result in a net shallowing of attractors towards goals relative to non-goal locations -giving rise to impaired goal overrepresentation and goal-directed navigation. Where plasticity is enhanced, our framework predicts that this enhancement will exhibit little or no specificity for salient compared to non-salient stimuli. This may involve enhanced heterosynaptic plasticity, where synapses other than those exhibiting plasticity inducing patterns of activity are potentiated/depressed. Alternatively, this could involve a reduced threshold for inducing homosynaptic plasticity, such that weak patterns of activity that typically don't induce plasticity at a given synaptic input in controls are effective in driving plasticity in schizophrenic models. An enhanced LTP that maintains input specificity (e.g. if induction threshold is maintained but LTP expression is enhanced) should have the opposite effect, causing deeper attractors to form and hence being protective against thought disorder.
6-Cognitive-mapping-related changes outlined in points 1-5 should be enriched in Schizophrenia patients with a high conceptual disorganization score on the PANSS compared to those with lower scores.