Seed-based thalamic connectivity analyses and the pattern of thalamic

Functional connectivity analyses looking at the thalamus as a single seed, and differences between psychosis patients and matched healthy control group, showed general effects that have previously been reported by our and other research groups, with over- connectivity between the thalamus and large sensorimotor regions, but also the associative cortex. After stringent Type I error protection and thresholding at p < 0.05, increased thalamic connectivity effect survived for: large areas of bilateral primary motor and sensory cortex, associative sensory cortex, bilateral Supplementary motor area, left primary and bilateral secondary and associative visual cortex, auditory cortex, gustatory cortex, insular cortex, right fusiform gyrus, right posterior cingulate, and left supramarginal gyrus (Figure 5 and Table 7). Areas showing significant thalamic under-connectivity when comparing psychosis probands and healthy controls, unsurprisingly, included large cerebellar areas, but also right primary visual cortex, left secondary visual cortex, right superior parietal lobule, and bilateral MD nucleus, along with right VL nucleus, and left anterior thalamic nucleus (Figure 6 and Table 8). Restricting analysis to just schizophrenia patients and comparing them to a specifically matched subset of healthy controls, a similar overall pattern of thalamic over-connectivity with sensorimotor regions, and under-connectivity primarily with cerebellum survived. As shown in Figure 10 and listed in Table 9, in schizophrenia patients there was significant thalamic over- connectivity with bilateral primary and secondary sensorimotor cortical areas, including bilateral Supplementary motor area, as well as right insular cortex, left auditory cortex, right middle temporal gyrus (BA 39), and left secondary and right associative visual cortex. Interestingly, while insular cortex was primarily connected to disorders with significant emotion regulation impairment, over-connectivity with insular cortex robustly appear in analyses comparing SCZ population with healthy controls. Areas showing functional under-connectivity with the thalamus (Figure 11 and Table 10), in addition to the cerebellum, included bilateral secondary visual cortex, right VL and left MD thalamic nucleus.

It should be noted that in both wider psychosis sample and the schizophrenia patient population thalamic connectivity shows pronounced reduction in variation when compared to that in the healthy controls, and then a shift towards over- or under-connectivity.

106 The overall pattern seems to be strikingly similar when we look at schizophrenia separately and when we examine it as part of the psychosis spectrum, with SCZ conceptualized as just the far end that is characterized not by any pathognomonic signs or symptoms but by increased severity of symptoms and more pronounced functional impairments. Some differences were noted, however, including over-connectivity with primary visual cortex, gustatory cortex, fusiform gyrus, and posterior cingulate. We could explain these differences by the fact that inclusion of SCAD and BDp population is bound to result in changed connectivity with brain regions modulating mood or providing emotional salience and context like posterior cingulate and fusiform gyrus. Both regions, however, also play an important part in wider network implicating cognitive functions and intrinsic control networks. Another way of explaining differences in connectivity maps is the increased power in detecting regions with the larger sample, suggesting that the same regions might be identified even with an increase in size of the SCZ-only sample. Elements possibly supporting the latter explanation are non- significant difference in Emotional Distress scores between BD and SCZ populations (difference in scores on Emotional Distress was driven by SCAD population), and identifications of networks in SCZ sample that have previously primarily been linked to BD (i.e. insular cortex) (406).

Results of the current analyses are in line with our previously published findings (403), but also with the results from other research groups (401, 402, 404, 407), of likely correlated patterns of thalamic over-connectivity with sensorimotor areas and under-connectivity with the cerebellum. Cortical areas affected by connectivity changes, along with primary sensory and motor areas, include almost without exception areas taking part in the higher-order processing of sensory information. It is, therefore, not surprising that among areas showing thalamic over- connectivity we found the fusiform gyrus that plays a role in higher processing of visual information, faces, word recognition and reading (441-443), and the insular cortex that receives sensory input via the thalamus and is believed to play a role in sensorimotor processing, but also in providing emotional context to sensory information, and in higher cognitive functions (e.g. salience, attention, social cognition) (444). Another region centering on temporoparietal junction exhibited significant thalamic over-connectivity, and includes structures thought to be linked with Wernicke’s area, and, like angular gyrus, thought to play an important role in attention, semantic and number processing, default mode network, awareness, and social cognition (445).

Taking into account that the primary input to the thalamus comes from sensory areas, and the confirmed pattern of thalamic functional over-connectivity with primary and associative sensory cortical areas, it is tempting to interpret this finding through theories of sensory gating disruptions in schizophrenia and psychotic disorders. However, one of the regions, previously

107 used to explain thalamic over-connectivity with sensory cortical regions via reduction in top- down regulatory tone, was conspicuous for its absence – the prefrontal cortex. In our previously published analysis of thalamo-cortical dysconnectivity (403), we postulated that there is a case to be made for changes in functional connectivity including reduced prefrontal-thalamic connectivity, increased connectivity between prefrontal and sensory regions, and finally increased thalamic connectivity with sensory regions, with prefrontal regions playing an inhibitory/regulatory role and in schizophrenia exhibiting inadequate top-down control indirectly through basal ganglia (446). It is important to note that reduced connectivity with prefrontal areas is indeed noticeable in unthresholded Figures 3 and 8, but these effects do not survive statistical thresholding. Given the wider focus on prefrontal functions disruption in schizophrenia (447-449), explanation including dysfunction of top-down prefrontal inhibitory tone seems to be mechanistically sound. Although the absence of significant prefrontal- thalamic connectivity differences between psychosis/SCZ and healthy control groups in the current analysis does not rule out the role of prefrontal regions, it naturally warrants discussion on possible explanations of thalamic over-connectivity in SCZ patients in the absence of prefrontal effects.

We already mentioned models that move away from conceptualizations of the thalamus as a simple relay, or relatively passive ‘switchboard’ on the path of sensory information flow, and these models are exactly the ones that could explain primary dysfunctions located in the thalamus as driving downstream effects (e.g. observed increased connectivity with sensory cortical areas). Dysfunction of glutamatergic neurotransmission, primarily through NMDA receptor dysfunction on GABA interneurons, has repeatedly been identified as a possible point of origin for defects seen in schizophrenia, and NMDA receptor dysfunction has also been hypothesized to be present in the thalamus and to underlie cortico-thalamo-cortical network deficits. One of the hypothesized ways in which this could take place is by indirectly compromising sensory driver inputs via NMDA receptor-mediated GABA interneuron dysfunction, or through thalamic higher-order nuclei dysfunction caused by direct NMDA receptor-related attenuation of driver feedforward excitatory inputs projecting from cortical areas (450).

Reticular thalamic nucleus, a thin sheet of inhibitory GABAergic interneurons that receives collateral afferents and connects widely with the rest of the thalamus providing a strong inhibitory signal, represents an appealing target for the hypothesized primary thalamic dysfunction leading to deficits in thalamo-cortical communication and large scale cortical communication as well as higher cognitive functions (311). RTN is suggested to be a vital communication hub between cortical areas and the thalamus, and disinhibition in thalamo- cortical circuits has been proposed to significantly affect behavior and cognitive functions

108 (311). We could, therefore, imagine NMDA receptor dysfunction affecting GABAergic inhibitory tone in both prefrontal cortex and the thalamus separately, resulting in the disruption in widely distributed networks communication, affecting complex cognitive functions and the underlying symptoms of psychosis. Important interactions between the thalamus and prefrontal cortex would most likely make these changes correlated, but the correlation would not necessarily mean causation. Shifting the point of origin towards primary thalamic dysfunction would be in line with conceptualizations of ‘cognitive thalamus’ (259), and, although it would remove the need for prefrontal effect as the one driving these disruptions, it in no way invalidates the critical role the prefrontal cortex plays in the entire model. Interestingly, in our already published analysis of thalamo-cortical disturbances (403), we found specifically thalamic over- connectivity with sensorimotor cortical areas to correlate with general psychopathology score, while the same was not seen for under-connectivity with prefrontal areas. This in itself, of course, does not allow any clear inference about a possible hierarchy or directionality, the least of all any inference about relative importance of prefrontal effects (as over-connectivity effects might just be closer to behavioral expression). However, at the very least it confirms that over- connectivity with large sensorimotor areas presents a stable finding, valuable in overall conceptualizations of thalamo-cortical dysconnectivity, its nature, and phenomenological/clinical consequences.

One of the findings regarding thalamic connectivity changes in schizophrenia and psychotic disorders, consistent across almost all investigations, like the sensorimotor cortex effect, is rarely mentioned or additionally analyzed – the thalamic over-connectivity with almost all cerebellar areas in schizophrenia patients. The same effect was clearly replicated in the current analysis, in wider psychosis-spectrum population, as well as in schizophrenia patients when they were analyzed separately. The cerebellum has been somewhat neglected in biological models explaining psychotic disorders or their symptoms, but that has been slowly changing with the expanding knowledge of the cerebellar role in non-motor functions. It has been shown that schizophrenia patients exhibit changes in the size of the cerebellum, density and size of Purkinje cells, as well as decreased blood flow in a range of tasks (e.g. tasks evaluating memory, attention, social cognition) (145). Andreasen and Pierson (145) have outlined possible models of cerebellar function extending beyond motor control, and motor and associative learning, that include modulation of cognitive processes ranging from timing to detecting patterns and providing feedback to the cortex following detection of errors (in thought as well as in movement). Realization of the role of the cerebellum in schizophrenia, and development of wider function models, lead to the conceptualization of dysfunction of integration in cortical-subcortical-cerebellar circuits as underlying schizophrenia, with

109 ‘cognitive dysmetria’ coined to describe phenomenological outcome of the circuit disruption (144, 145, 323).