Altered cingulate and substantia nigra/ventral tegmental activation to novelty and emotional salience in antipsychotic naïve first episode psychosis patients

Abnormal salience processing has been suggested to contribute to the formation of positive psychotic symptoms in schizophrenia and related conditions. Previous research utilising reward learning or anticipation paradigms has demonstrated cortical and subcortical abnormalities in people with psychosis, specifically in the prefrontal cortex, the dopaminergic midbrain and the striatum. In these paradigms, reward prediction errors attribute motivational salience to stimuli. However, little is known about possible abnormalities across different forms of salience processing in psychosis patients, and whether any such abnormalities involve the dopaminergic midbrain. The aim of our study was, therefore, to investigate possible alterations in psychosis in neural activity in response to various forms of salience: novelty, negative emotion, targetness (task-driven salience) and rareness. We studied 14 antipsychotic naïve participants with first episode psychosis, and 37 healthy volunteers. During fMRI scanning, participants performed a visual oddball task containing these four forms of salience 1 . Psychosis patients showed abnormally reduced signalling in the substantia nigra/ventral tegmental area (SN/VTA) and the cingulate gyrus for novelty, negative emotional salience and targetness; reduced striatal signalling to novelty and negative emotional salience, and reduced signalling in the right amygdala to negative emotional salience. There was reduced cerebellar activation to targetness in the patients, consistent with abnormal efference copy transmission in preparation of a motor response. Our results indicate that generalised salience processing alterations in patients with psychosis, mainly involving the dopaminergic SN/VTA, the cingulate gyrus and the striatum. neurons in the SN/VTA and the ‘aberrant salience’ hypothesis of psychosis, we hypothesised that psychosis patients demonstrate altered SN/VTA and striatal responses to novelty, negative emotional salience and targetness. Furthermore, we predicted to find group differences in the prefrontal cortex in response to novelty, in the amygdala in response to emotional salience, and in the hippocampus in responses to all forms of salience. construct delusional explanations in order to explain these altered perceptions. Abnormal salience attribution is present from early and even prodromal stages of the disease 20,24 . Usually, this theory is investigated in the context of motivational salience 20,23 using reward prediction paradims. Here, however, salience was investigated in the absence of the reward, and constitutes an intrinsic property of the stimulus, rather than learned associations 2 . In healthy subjects, novelty identification is processed by a number of brain regions, including SN/VTA, striatum, parietal, and prefrontal cortices 1,25,42 . Consistent with this, we observed group differences in the SN/VTA and the striatum in response to novelty. Our findings extend recent results of a study by Schott and colleagues 27 . Although, this study did not detect clear differences in the midbrain or striatum, they found an increase in functional connectivity of the hippocampus and the orbitofrontal cortex with the rostral anterior cingulate gyrus and the ventral striatum. Our study also demonstrates significantly reduced activation in response to negative emotional salience compared to controls in right amygdala, the SN/VTA and the striatum in psychosis patients compared to controls. This result is consistent with the literature indicating reduced arousal to emotional stimuli 43 . Our study also supports findings of a PET study indicating tonic over-activation of the amygdala and impaired striatal signalling during emotional salience processing 30 . Jabbi and colleagues 28 reported increased dopaminergic releases in the amygdala and midbrain in response to emotional salience, which might be altered in psychosis. Our results, furthermore, reveal reduced activation in the thalamus of psychosis patients compared to healthy controls for negative emotional salience. The thalamus is a relay station of multiple neural connections and has dopaminergic synapses. Consistent with this and our findings, a study by Hadley and colleagues 44 reported reduced connectivity between the VTA/midbrain and the thalamus in schizophrenia patients. The results extend previous research by giving supportive evidence for the aberrant salience hypothesis of psychosis involving motivational and non-motivational forms of salience and the involvement of dopaminergic dysregulation in the development of psychotic disorder.


Introduction
Salience is a property that enables a stimulus to attract attention, and to drive cognition and behaviour. It can be described as a product of matched/mismatched stimulus features and internal, driving factors of an individual, such as goals, beliefs and experiences at a particular point in time. Salience is a multifaceted concept 2 , including different dimensions, such as reward and threat prediction, prediction error, novelty, emotional salience or rareness. The literature well describes the role of dopamine (DA) for reward prediction error 3,4 , with neural signals originating in the substantia nigra/ventral tegmental area (SN/VTA) 5 . However, DA neuron firing is not exclusive to reward prediction error, but has also been reported in response to non-rewarding unexpected events, such as aversive or alerting 6 , as well as novel events 7 , suggesting that DA release, at least in some contexts, reflects general salience 8,9 . In psychosis, abnormal salience processing secondary to dysregulation of the dopaminergic system -described as the 'aberrant salience' hypothesis of psychosis 2,10,11 -has been linked to the formation and maintenance of psychotic symptoms [12][13][14] . It has been suggested that aberrant salience attribution in psychosis is caused by faulty DA signalling in the striatum, possibly driven by dysregulation from the prefrontal cortex (PFC) and hippocampus 15 . In psychosis, there is increased synthesis and release of DA in the striatum, which is present even at the prodromal stages of the disease 16,17 . Several studies reported reduced midbrain, striatal, and/or cortical processing of reward prediction errors [18][19][20][21] and non-reward related prediction errors in psychosis 22 . In our recent work, we documented meso-cortico-striatal prediction error deficits, involving midbrain, striatum and right lateral frontal cortex in medicated psychosis patients at different stages 20,22 and in unmedicated first episode psychosis patient and patients at clinical risk for developing psychosis 23 . Another study in people at clinical risk of psychosis showed a relation between striatal reward prediction signal and psychotic symptoms 24 .
Novel events activate DA neurons even in the absence of reward, which is associated with increased attention, memory and goal-directed behaviour 6 . Together with the fact that novelty exploration engages the areas of the brain involved in appetitive reinforcement learning (i.e. dopaminergic midbrain areas, striatum, medial prefrontal cortex) 1,25 , novelty may be intrinsically rewarding, irrespective of the choice outcome, and can provide a 'bonus' for exploration 26 . A recent study by Schott and colleagues 27 reported alterations in a fronto-limbic novelty processing 4 network in patients with acute psychosis. However, it is unclear whether novelty processing is disrupted in key dopaminergic regions for salience processing, such as the SN/VTA. Emotional events are also salient, capture attention, enhance memory and modify behavioural responses 25 . Presynaptic DA levels in the amygdala and SN/VTA predict brain activity in response to emotional salience 28 . Schizophrenia and first episode psychosis patients have problems processing emotions, especially in the context of facial recognition 29 . In a PET study, Taylor and colleagues 30 showed impaired neural processing in the ventral striatum in response to emotional salient events in chronic and acute psychosis patients. However, results regarding processing alterations in the amygdala were unclear. Furthermore, it is unknown whether processing of the dopaminergic SN/VTA is altered in psychosis in response to emotional salience.
Various studies suggest that SN/VTA neurons respond to a general form of salience (see reviews 8,9 ), often referred to as 'physical salience' or 'alerting' salience 3,31 , which is triggered by unexpected sensory events including surprise, attention, arousal, or novelty. If dopaminergic signalling is generally compromised in psychosis, it then follows that there should be overlapping patterns of abnormal activation to various forms of salient stimuli in the dopaminergic midbrain and associated target regions in psychosis patients. Under an alternative account, salience processing may still be generally impaired in psychosis, but this may be secondary to dysfunction of diverse neural systems. In the current study, we, therefore, investigated brain responses in the SN/VTA and other target areas to four types of salience 1 -stimulus novelty, negative emotional salience, rareness (or 'contextual deviance'), and targetness (task-driven attentional salience) -in patients with early psychosis and healthy volunteers. By focussing on early psychosis, we can avoid confounds of exposure to dopaminergic medications and other effects of chronic illness. We used a fMRI paradigm that previously was shown to significantly activate parts of the midbrain, amygdala and striatum to various forms of salience.
Based on the potentially general role in salience signalling of DA neurons in the SN/VTA and the 'aberrant salience' hypothesis of psychosis, we hypothesised that psychosis patients demonstrate altered SN/VTA and striatal responses to novelty, negative emotional salience and targetness.
Furthermore, we predicted to find group differences in the prefrontal cortex in response to novelty, in the amygdala in response to emotional salience, and in the hippocampus in responses to all forms of salience. author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/263020 doi: bioRxiv preprint

Subjects
We recruited 14 antipsychotic naive individuals with first-episode psychosis and active psychotic symptoms from the Cambridge, early intervention service for psychosis, CAMEO. Other inclusion criteria were as follows: age 16-35 years, meeting ICD-10 criteria for a schizophrenia spectrum disorder (F20, F22, F23, F25, F28, F29) or affective psychosis (F30.2, F31.2, F32.3). Age, gender and handedness matched healthy volunteers (n=37) were recruited as control subjects. Demographic and clinical characteristics of those participants included in the final analysis are presented in Table 1 and 2. None of the healthy volunteers reported any personal or family history of severe neurological, psychiatric or medical disorders. All participants had normal or corrected-to-normal vision, and had no contraindications to MRI scanning. At the time of the study, none of the participants were taking antipsychotic medication or had drug or alcohol dependence.
Before scanning, each of the participants underwent a general interview and clinical assessment using the Positive and Negative Symptom Scale (PANSS) 32 , the Scale for the Assessment of Negative Symptoms (SANS) 33 and the Global Assessment of Functioning (GAF) 34 . The Beck Depression Inventory (BDI) 35 was used to assess depressive symptoms during the last two weeks.
IQ was estimated using the Culture Fair Intelligence Test 36 .
The study was approved by the Cambridgeshire 3 and National Health Service research ethics committee. All participants supplied written informed consent after they had read a complete description of the study.

Novelty task
We used a visual oddball paradigm 37 in order to investigate four types of salience, which were novelty, negative emotional salience, targetness and rareness. Participants were presented with a series of greyscale images of faces and outdoor scenes. 66% of those had a neutral emotional valence ('standard'). The four types of rare or contextually deviant events were randomly intermixed with these; each occurred with a probability of 8.3%. These deviant events were: neutral stimuli that required a motor response ('target oddball'); stimuli that evoked a negative author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/263020 doi: bioRxiv preprint emotional response ('emotional oddball', angry face or image of car crash); novel stimuli ('novel oddball', different neutral images that appear only once); and neutral stimuli ('neutral oddball', neutral image of face or scene) ( Figure 1). All participants completed four blocks with 60 trials each, resulting in a total of 240 trials (160 standard oddballs, and 20 each of target, neutral, emotional and novel oddballs). The task contained 50% faces and 50% outdoor scenes, this allowed prevention of category-specific habituation. These categories were chosen instead of abstract images to make stimulus exploration biologically relevant. Participants were introduced to the target stimulus prior to the experimental session for 4.5s, and they were required to make a simple button press with their right index finger in response to each of its subsequent appearances during the experiment within the fMRI-scanner. No motor responses were associated with any of the other stimulus types.
During the fMRI-experiment, the pictures were presented for 500ms followed by a white fixation cross on a grey background (grey value=127) using an inter-stimulus interval (ISI) of 2.7s. ISI was jittered with ±300ms (uniformly distributed). The order of stimuli was optimised for efficiency with regard to estimating stimulus-related haemodynamic responses.
All of the stimuli were taken from Bunzeck and Düzel 1 . The scalp hair and ears of faces were removed artificially, the outdoor scenes did not include faces. All pictures were grey scaled and normalised to a mean grey value of 127 (SD 75). The pictures were projected on to the centre of a screen, and the participants watched them through a mirror mounted on the head coil, subtending a visual angle of about 8°. The negative emotional scene depicted a negatively rated car accident (without any people). The contrast between stimuli allowed us to examine brain responses to the pure stimulus novelty ('novel' vs. 'neutral'), targetness ('target' vs. 'neutral'), negative emotional valence ('emotional' vs. 'neutral') and rareness/deviance per se ('neutral' vs. 'standard') ( Figure 1). Irrespective of whether participants were left or right handed, they used their right hand to press the buttons on the button box for the target picture.

Behaviour analysis
An analysis of variance (ANOVA) was used to investigate group differences in pressing the buttons in response to the target stimuli and assessing reaction times. All runs in which participants missed author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/263020 doi: bioRxiv preprint more than five button presses were excluded. Behavioural data were analysed using SPSS 21 (IBM Corp.).

fMRI data acquisition and analysis
A Siemens Magnetom Trio Tim syngo MR B17 operating at 3 T was used to collect imaging data.
Gradient-echo echo-planar T2*-weighted images depicting BOLD contrast were acquired from 35 non-contiguous oblique axial plane slices of 2mm thickness to minimise signal drop-out in the ventral regions. We did not retrieve images of the whole brain; the superior part of the cortex was not imaged. The relaxation time was 1620ms, echo time was 30ms, flip angle was 65°, in-plane resolution was 3×3×3mm, matrix size was 64×64, field of view was 192×19mm, and bandwidth was 2442Hz/px. A total of 437 volumes per participant were acquired (35 slices each of 2mm thickness, inter-slice gap 1mm). The first five volumes were discarded to allow for T1 equilibration effects.
The data were analysed using FSL software (FMRIB's Software Library, www.fmrib.ox.ac.uk/fsl) version five. Participants' data (first-level analysis) were processed using the FMRI Expert Analysis Tool (FEAT). Functional images were realigned, motion corrected (MCFLIRT) and spatially smoothed with a 4mm full-width half-maximum Gaussian kernel. A high-pass filter was applied (120s cut-off). All images were registered to the whole-brain echo-planar image (EPI) (i.e., functional image with the whole-brain field of view), and then to the structural image of the corresponding participant (MPRAGE) and normalised to an MNI template. The five explanatory variables (EVs) that we used were the onset times of the standard, target, emotional, novel and neutral pictures. They were modelled as 1s events and convolved with a canonical double-gamma response function. We added a temporal derivative to the model to take into account possible variables in the haemodynamic response function. To capture residual movement-related artefacts, six covariates were used as regressors of no interest (three rigid-body translations and three rotations resulting from realignment). We used four contrasts: target-neutral, emotionneutral, novel-neutral, and neutral-standard. In the second-level analysis, we averaged the four blocks of the task for each participant. For estimation of higher level, we used FEAT with FMRIB's Local Analysis of Mixed Effects (FLAME2). author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/263020 doi: bioRxiv preprint Our main analysis was based on a region of interest (ROI) approach as follows. For novelty, our primary ROI was the dopaminergic SN/VTA using the probabilistic atlas 38 , in which traditional anatomical segmentation was replicated using a seed-based functional connectivity approach and which provides a mask that consists of the SN and VTA, also used in our previous work 23 . In our two secondary ROIs, we investigated the striatum (using a single hand-drawn mask), encompassing both associative and limbic striatum, based on operational criteria 39,40 , and the right lateral frontal cortex (utilising a sphere, 10mm, centred at x=50, y30=, z=28, based on our previous work 22,23 ). For negative emotional salience, our two primary ROIs were the dopaminergic SN/VTA and the amygdala (anatomically derived mask using the Harvard-Oxford cortical and subcortical structural atlases). Our secondary ROI was the striatum. For targetness and rareness, we used the dopaminergic SN/VTA as our primary ROI and the striatum as our secondary ROI.
Within our ROIs, we used FSL randomise in order to test our hypothesis that on the contrast of interest activation pattern: controls>patients. Multiple regions were combined in one mask in order to control for multiple comparisons. For illustrational purposes only, we then extracted contrast values (contrast of parameter estimates, or COPEs in FSL) for each individual from voxels in which significant group differences were found (See bar chart in Figure 2B and 3B). In order to compare groups on a whole brain level, we used easythresh, a whole brain cluster corrected FSL analysis, with a threshold of p<0.05, family-wise error (FWE) corrected, with cluster size z>2.6 (results presented in Table 3).
Furthermore, parameter estimates are presented in the supplementary materials for all conditions and ROIs that show a significant effect. The parameter estimates show the driving condition for the group effects.

Movement differences during fMRI scan
As the task was relatively long (46min) and mostly passive (button presses were only required in 20 out of the total of 240 trials), we split the task into four blocks of 11.5min. Still many participants, independent of group, exhibited movements, possibly due to tiredness. We, therefore, excluded those blocks in which movement exceeded 3mm on average or 10mm maximum. In total, we identified 14 runs that fulfilled the movement exclusion criterion. Of those 14 runs, 10 were either from testing block 3 or 4, 2 were from testing block 2 and 2 were from author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/263020 doi: bioRxiv preprint testing block 1. Additionally, three runs had to be interrupted and were therefore not completed by the participants. If for a single participant only one or two runs remained for analysis, we excluded this participant entirely. Based on these criterions, we excluded three controls and one psychosis patient entirely, as well as one run in five psychosis patients and one run in three controls.
In the remaining sample, we compared the two groups in two separate repeated measure ANOVAs across the four testing blocks, one for movement means and one for maximum (Supplementary Figure 1 and 2). We did not find any significant group, run or interactions effect, neither for mean movement nor for maximum movement (all p>0.1).
We excluded those four individuals from all analyses within this study.

Demographic and questionnaire results
The demographic and rating results are summarised in Tables 1 and 2. Bonferroni-corrected posthoc tests demonstrated significant differences in IQ between controls and the psychosis group (p=0.02). More importantly, however, the groups were matched in maternal educational, which was similar across both groups (p=0.36). Alcohol consumption was significantly lower in psychosis patients compared to controls (p=0.002).

Behavioural responses to pictures and reaction times
In order to maintain engagement with task, participants were required to press a button in response to the target picture. Due to technical problems, button presses were not recorded for eight controls, and one psychosis patient. Analysing the number of missed button presses and reaction times of the remaining participants across the four testing blocks (Supplementary Table   1), we did not find any significant effects for group, testing block or any interactions (all p>0.3). On average, participants missed to press the button once (mean: 1.0 SE±0.2) and generally required approximately 550ms (SE±0.02) to make a response, which is consistent with previous findings 1 .
Due to the high performance across all groups, we included the data of the nine participants author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/263020 doi: bioRxiv preprint without recorded button presses in all further analyses, in order to increase statistical power. We also repeated the analysis after excluding those participants. The results were very similar.
On whole brain analysis correcting for multiple comparisons (thresholded at p<0.05, z>2.6), psychosis patients showed a significant reduction of activation in the occipital lobe, including the lingual gyrus and fusiform gyrus, and the posterior cingulate gyrus (Table 3 and Supplementary material).

Negative emotional salience (emotion-neutral oddballs)
In our primary ROIs, the amygdala and SN/VTA, we found five clusters in which psychosis patients show significantly reduced activation, one in the right amygdala (  Figure 3A and B.
On whole brain analysis correcting for multiple comparisons (thresholded at p<0.05, z>2.6), psychosis patients showed a significant reduction of activation in thalamus, lingual gyrus, caudate and the anterior cingulate gyrus (Table 3).

Targetness (target-neutral oddballs)
In our primary ROI, the SN/VTA, psychosis patients showed a marginal reduction of activation compared to the controls (t=3.84, p=0.066 FWE corrected, 5 voxels; maximal difference at x=0, y=-

22, z=-8).
On whole brain analysis correcting for multiple comparisons (thresholded at p<0.05, z>2.6), psychosis patients showed a significant reduction of activation in occipital lobe, including the fusiform gyrus and lingual gyrus, the cerebellum (vermis VI and crus II) and the anterior cingulate gyrus (Table 3).

Rareness (neutral-standard oddballs)
Our ROI analysis in the SN/VTA was not significant. Similarly, on the whole brain analysis, there were no group differences that passed our statistical threshold, corrected for multiple comparisons.

Correlations of symptom score and brain responses in patients
We found positive correlations between SN/VTA signalling and the total score of negative symptoms (SANS; rho=0.66, p=0.047; Figure 4A), and the hallucinatory behaviour (P1; rho=0.77, p=0.002; Figure 4B) in response to novelty.
Furthermore, in response to negative emotional salience, we found a positive correlations between striatal signalling and the total score of positive symptoms (PANS positive; rho=-0.61, author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/263020 doi: bioRxiv preprint p=0.028; Figure 4C), a positive correlation between amygdala signalling and the Beck Depression Inventory (BDI; rho=0.64, p=0.048; Figure 4D), and a positive correlation between amygdala signalling and delusional behaviour (P3; rho=0.59, p=0.035; Figure 4E). Correlations were computed using a nonparametric Spearmen's correlation.
We emphasize that we are presenting exploratory correlation analyses (given the small sample size); when controlling for multiple comparisons using a strict Bonferroni-approach only the correlation between SN/VTA signalling and hallucinatory behaviour in response to novelty would be retain as statistically significant.

Discussion
We investigated brain responses to four different types of salience, including novelty, negative emotional salience, targetness and rareness in healthy volunteers and first episode psychosis patients. In psychosis patients, our results show reduced SN/VTA (primary ROI), striatal (secondary ROI) and cingulate (whole brain) signalling to novelty, reduced SN/VTA and amygdala (both primary ROIs), striatal (secondary ROI) and cingulate (whole brain) signalling to negative emotional salience, and reduced SN/VTA (primary ROI) and cingulate (whole brain) signalling to targetness. The 'aberrant salience' hypothesis of psychosis postulates that dysregulated dopaminergic signalling in the mesolimbic system in people with psychosis results in the attribution of salience to irrelevant or non-significant stimuli 11,41 . These unusually salient representations may lead to the formation of hallucinations or generally altered perceptions. As a result, patients may author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/263020 doi: bioRxiv preprint construct delusional explanations in order to explain these altered perceptions. Abnormal salience attribution is present from early and even prodromal stages of the disease 20,24 . Usually, this theory is investigated in the context of motivational salience 20,23 using reward prediction paradims. Here, however, salience was investigated in the absence of the reward, and constitutes an intrinsic property of the stimulus, rather than learned associations 2 . In healthy subjects, novelty identification is processed by a number of brain regions, including SN/VTA, striatum, parietal, and prefrontal cortices 1,25,42 . Consistent with this, we observed group differences in the SN/VTA and the striatum in response to novelty. Our findings extend recent results of a study by Schott and colleagues 27 . Although, this study did not detect clear differences in the midbrain or striatum, they found an increase in functional connectivity of the hippocampus and the orbitofrontal cortex with the rostral anterior cingulate gyrus and the ventral striatum.
Our study also demonstrates significantly reduced activation in response to negative emotional salience compared to controls in right amygdala, the SN/VTA and the striatum in psychosis patients compared to controls. This result is consistent with the literature indicating reduced arousal to emotional stimuli 43 . Our study also supports findings of a PET study indicating tonic over-activation of the amygdala and impaired striatal signalling during emotional salience processing 30 . Jabbi and colleagues 28 reported increased dopaminergic releases in the amygdala and midbrain in response to emotional salience, which might be altered in psychosis. Our results, furthermore, reveal reduced activation in the thalamus of psychosis patients compared to healthy controls for negative emotional salience. The thalamus is a relay station of multiple neural connections and has dopaminergic synapses. Consistent with this and our findings, a study by Hadley and colleagues 44 reported reduced connectivity between the VTA/midbrain and the thalamus in schizophrenia patients.
In addition to reduced SN/VTA processing in response to novelty and negative emotional salience, we also found reduced signalling in response to targetness in patients. Therefore, our study is first to provide clear evidence for reduced SN/VTA processing in response to these different forms of non-motivational salience in psychosis. Together with the striatal findings, the findings in the patients support the aberrant salience hypothesis for general salience dysfunction. As both the midbrain and the striatum are dopaminergic key regions, it also provides supporting evidence for a dysregulated dopaminergic system during salience processing in psychosis 2 . In healthy controls, Bunzeck and Düzel 1 reported significantly enhanced SN/VTA activation in response to novelty, and author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/263020 doi: bioRxiv preprint also positive, but not statistically significant, activation in response to negative emotional salience, providing supportive evidence for a differential activation of the SN/VTA in response to novelty.
Using a larger sample size than previous studies and a slightly different regional specification used for the SN/VTA, we, however, find significant SN/VTA activation to novelty, negative emotional salience and targetness in controls, although activation in response to novelty is greatest. Our results, therefore, support the view of general processing of salience in the SN/VTA 8,9 , including novelty, negative emotional salience and targetness. An account reconciling these results with those of Bunzeck and Düzel 1 , may be that SN/VTA is highly sensitive to novelty, but is also sensitive to other forms of salience. On the other hand, our activation is slightly more rostral as compared to Bunzeck and Düzel 1 but still lies within the SN/VTA ROI as defined by Murty et al. 38 .
Given the spatial resolution of the current study, we cannot determine within this region whether the exact same neurons activate to diverse or specific stimuli 45 . However, future studies should employ higher resolution to answer these questions.
Moreover, the whole brain analysis revealed reductions in anterior and posterior cingulate gyrus activity in psychosis patients compared to healthy controls in response to novelty, negative emotional salience and targetness. The cingulate cortex, as part of the salience network, has been found to show aberrant connectivity and structure in psychosis 46,47 . We previously showed that the severity of psychotic symptoms in healthy volunteers induced by methamphetamine, significantly correlated with the degree of drug induced disruption of the incentive value signal disruption in the posterior cingulate cortex, suggesting a dopamine mediated mechanism in this region 48 . A study by Gradin and colleagues 49 reported dysfunctional connectivity between the salience network and the midbrain during a reward learning task leading to abnormal reward processing in schizophrenia patients. Furthermore, structural alterations have consistently been documented in patients with psychosis [50][51][52] . Therefore, our results may provide a first indication that possible dysfunctional interactions between the salience network and the SN/VTA may also lead to aberrant processing of different types of salience. Furthermore, in the whole brain analysis we found reduced cerebellar activity in response to targetness in patients compared to controls.
This finding can be explained by a potential inability to appropriately generate and use efferences copies to predict sensory consequences of motor events. The generation of the efference copy has been localised in the cerebellum 53,54 and has been linked to action prediction failures in schizophrenia (see review 55 ). author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/263020 doi: bioRxiv preprint paradigm across all stimuli 42,56 . Here, we observed group differences in response to novelty, negative emotional salience and targetness, this is in line with impaired visual perceptions often reported in schizophrenia (see review 57 ). In contrast with the previous literature, which reported hippocampal activity in response to salience 1,27,58 , we did not find any activity in the hippocampus, neither in a group difference nor in a healthy volunteers separately. It is possible that signal in this region may not have been reliably captured during fMRI scanning.
In an exploratory analysis, we found positive correlations between SN/VTA activity to novelty and symptom scores for hallucinations and negative symptoms, between amygdala signalling to negative emotional salience and the Beck Depression Inventory and delusions, and between striatal signalling and total score for positive symptoms. However, when controlling for multiple comparisons, only the correlation between SN/VTA activation to novelty and hallucinations remains significant. Here, we would have rather predicted a negative correlation, showing a decrease of SN/VTA activation with increased symptom scores, especially given the group difference that showed lower activation in the patient group as a whole. It is thought-provoking that in this small study, several forms of salience showed reduced activation across regions in the average patient, but greater activation associated with greater symptoms. One speculation is that reduced activation (between group results) could reflect a trait abnormality, and superimposed on this are state dysfunctions closely linked to symptom expression. However, symptom correlations with functional imaging have often yielded inconsistent results in schizophrenia research 59 . One of the most important difficulties to reliably detect symptom correlations is gathering a large enough sample, and our small sample size of 14 patients is a clear limitation to assess symptom correlations. However, we report it to generate future hypotheses and to be available for future meta-analysis.
In conclusion, this study provides concise evidence for aberrant SN/VTA, striatal and cingulate signalling during non-motivational salience processing in a sample of antipsychotic naïve early psychosis patients. The results extend previous research by giving supportive evidence for the aberrant salience hypothesis of psychosis involving motivational and non-motivational forms of salience and the involvement of dopaminergic dysregulation in the development of psychotic disorder.
author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/263020 doi: bioRxiv preprint White TP, Joseph V, Francis ST, Liddle PF. Aberrant salience network (bilateral insula and author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/263020 doi: bioRxiv preprint    author/funder. All rights reserved. No reuse allowed without permission.

Tables and Figures
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/263020 doi: bioRxiv preprint Figure 2. Group effects in primary and secondary region of interest (ROI) analysis of activation associated with novelty processing. A) Primary ROI (colour coding yellow-red): SN/VTA, maximal difference at x=0, y=-20, z=-6. Secondary ROI (colour coding light-blue-dark-blue), striatum, two clusters maximal difference at x=8, y=-2, z=14 and x=-8, y=-2, z=12 (p<0.05 FWE corrected). B) Bar chart shows the mean contrast (COPEs, FSL) values to group, extracted from significant clusters determined by FSL randomise ANOVA results of primary and secondary ROI analysis. Multiple significant clusters are combined. Error bars show ±1 SE. author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/263020 doi: bioRxiv preprint Figure 4. Correlation between signal strength and symptom score in patients. Panels A and B show significant symptom correlations with activation in the SN/VTA in response to novelty. Panel C shows significant correlation between total PANS score and striatal activation in response to negative emotional salience. Panel D and E show significant symptom correlations with activation in the amygdala in response to negative emotional salience. Lines indicate fitted regression lines. author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/263020 doi: bioRxiv preprint