Jakob Kaminski

Validating the construct of aberrant salience in schizophrenia — Behavioral evidence for an automatic process

Suspecting significance behind ordinary events is a common feature in psychosis and it is assumed to occur due to aberrant salience attribution. The Salience Attribution Test (SAT; Roiser et al., 2009) measures aberrant salience as a bias towards one out of two equally reinforced cue features as opposed to adaptive salience towards features indicating high reinforcement. This is the first study to validate the latent constructs involved in salience attribution in patients. Forty-nine schizophrenia patients and forty-four healthy individuals completed the SAT, a novel implicit salience paradigm (ISP), a reversal learning task and a neuropsychological test battery. First, groups were compared on raw measures. Second and within patients, these were correlated and then used for a principal component analysis (PCA). Third, sum scores matching the correlation and component pattern were correlated with psychopathology. Compared to healthy individuals, patients exhibited more implicit aberrant salience in the SAT and ISP and less implicit and explicit adaptive salience attribution in the SAT. Implicit aberrant salience from the SAT and ISP positively correlated with each other and negatively with reversal learning. Whereas explicit aberrant salience was associated with cognition, implicit and explicit adaptive salience were positively correlated. A similar pattern emerged in the PCA and implicit aberrant salience was associated with negative symptoms. Taken together, implicit aberrant salience from the SAT and ISP seems to reflect an automatic process that is independent from deficient salience ascription to relevant events. Its positive correlation with negative symptoms might reflect motivational deficits present in chronic schizophrenia patients.

Epigenetic variance in dopamine D2 receptor: a marker of IQ malleability?

Genetic and environmental factors both contribute to cognitive test performance. A substantial increase in average intelligence test results in the second half of the previous century within one generation is unlikely to be explained by genetic changes. One possible explanation for the strong malleability of cognitive performance measure is that environmental factors modify gene expression via epigenetic mechanisms. Epigenetic factors may help to understand the recent observations of an association between dopamine-dependent encoding of reward prediction errors and cognitive capacity, which was modulated by adverse life events. The possible manifestation of malleable biomarkers contributing to variance in cognitive test performance, and thus possibly contributing to the “missing heritability” between estimates from twin studies and variance explained by genetic markers, is still unclear. Here we show in 1475 healthy adolescents from the IMaging and GENetics (IMAGEN) sample that general IQ (gIQ) is associated with (1) polygenic scores for intelligence, (2) epigenetic modification of DRD2 gene, (3) gray matter density in striatum, and (4) functional striatal activation elicited by temporarily surprising reward-predicting cues. Comparing the relative importance for the prediction of gIQ in an overlapping subsample, our results demonstrate neurobiological correlates of the malleability of gIQ and point to equal importance of genetic variance, epigenetic modification of DRD2 receptor gene, as well as functional striatal activation, known to influence dopamine neurotransmission. Peripheral epigenetic markers are in need of confirmation in the central nervous system and should be tested in longitudinal settings specifically assessing individual and environmental factors that modify epigenetic structure.

Overestimating environmental volatility increases switching behavior and is linked to activation of dorsolateral prefrontal cortex in schizophrenia

Background Reward-based decision-making is impaired in schizophrenia, as reflected by increased switching between choices. The underlying cognitive mechanisms and associated neural signatures remain unknown. Reinforcement learning (RL) and hierarchical Bayesian learning account for this behavior in different ways. We hypothesized that enhanced switching during flexible reward-based decision-making in schizophrenia relates to higher-order beliefs about environmental volatility and examined the associated neural signatures. Methods 46 medicated schizophrenia patients and 43 controls underwent a reward-based decision-making task requiring flexible behavior to changing action-outcome contingencies during functional Magnetic Resonance Imaging (fMRI). Computational modeling of behavior was performed, including RL and the Hierarchical Gaussian Filter (HGF). The estimated learning trajectories informed the analysis of fMRI data. Results A three-level HGF accounted best for the observed choice data and revealed a heightened prior belief about environmental volatility and a stronger influence of volatility on lower-level learning of action-outcome contingencies in schizophrenia. This finding was replicated in an independent sample of unmedicated patients. Beliefs about environmental volatility were reflected by higher activity in dorsolateral prefrontal cortex (dlPFC) of patients compared to controls. Conclusions This study suggests a mechanistic explanation for instable behavior in schizophrenia: patients inferred the environment as being too volatile and thus overestimated environmental changes, leading to maladaptive choice switching. Our data suggest enhanced dlPFC activity related to beliefs about environmental volatility as a neural learning signature of instable behavior. Such detailed ‘computational phenotyping’ may provide useful information to dissect clinical heterogeneity and could improve prediction of outcome.Abstract Background Reward-based decision-making is impaired in patients with schizophrenia (PSZ) as reflected by increased choice switching. The underlying cognitive and motivational processes as well as associated neural signatures remain unknown. Reinforcement Learning (RL) and hierarchical Bayesian learning account for choice switching in different ways. We hypothesized that enhanced choice switching, as seen in PSZ during reward-based decision-making, relates to higher-order beliefs about environmental volatility and examined the associated neural activity. Methods 46 medicated PSZ and 43 healthy controls (HC) performed a reward-based decision-making task requiring flexible responses to changing action-outcome contingencies during functional Magnetic Resonance Imaging (fMRI). Detailed computational modeling of choice data was performed, including RL and the hierarchical Gaussian filter (HGF). Trajectories of learning from computational modeling informed the analysis of fMRI data. Results A three-level HGF accounted best for the observed choice data. This model revealed a heightened initial belief about environmental volatility and a stronger influence of volatility on lower-level learning of action-outcome contingencies in PSZ as compared to HC. This was replicated in an independent sample of non-medicated PSZ. Beliefs about environmental volatility were reflected by higher activity in dorsolateral prefrontal cortex of PSZ as compared to HC. Conclusions Our study suggests that PSZ inferred the environment as overly volatile, which may explain increased choice switching. In PSZ, activity in dorsolateral prefrontal cortex was more strongly related to beliefs about environmental volatility. Our computational phenotyping approach may provide useful information to dissect clinical heterogeneity and could improve prediction of outcome.

S222. Decreased Striatal Reward Prediction Error Coding in Unmedicated Schizophrenia Patients

Abstract Background Reinforcement learning involves flexible adaptation towards a changing environment and is driven by dopaminergic reward prediction error (RPE; outcome (R) – expectation (Q)) signaling in the midbrain and projecting regions, such as the ventral striatum (Schultz, 1998). Schizophrenia patients show heightened dopamine levels in the striatum (Howes et al., 2012) as well as deficits in reinforcement learning (Waltz, 2016) which may be mediated by disrupted prediction error signaling (Heinz and Schlagenhauf, 2010; Schlagenhauf et al., 2014). Using model-based fMRI, the present study aims to assess these neural signals during a reversal learning paradigm in unmedicated schizophrenia patients and healthy individuals. Methods In the current study, 19 schizophrenia patients and 23 age- and gender-matched healthy controls completed a reversal learning paradigm (Boehme et al., 2015) during fMRI scanning where subjects had to choose between two neutral stimuli to maximize their reward. A Rescorla Wagner learning model (Single Update, one learning rate) was fitted against the individual choice data using a softmax function. Individual RPE trajectories from the fitted Rescorla Wager learning model were correlated with the BOLD response during feedback onset. Parameter estimates of ventral striaral RPE trajectories were correlated with psychopathology scores from the PANSS (Kay et al., 1987). Results In the reversal learning task, schizophrenia patients chose the correct stimulus less often compared to healthy individuals (percent correct choices: 65.7 ± 10.7 vs. 76.7 ± 7.7; t=3.7, p=0.001). Across all participants, the RPE trajectories correlated with BOLD response in the bilateral ventral striatum (left ventral striatum [-10 12 10], t=7.40, pFWE <0.001, right ventral striatum [10 12 -10], t=6.56, pFWE=0.006). Schizophrenia patients displayed decreased RPE coding in the right ventral striatum compared to healthy individuals ([14 14 -10], t=3.69, pSVC for nucleus accumbens = .015). In patients, extracted parameter estimates from the right ventral striatum correlated negatively with the PANSS total symptoms score (Spearman’s rho =-0.55, p=0.018). Discussion We found that unmedicated schizophrenia patients performed worse in the reversal learning task and displayed decreased striatal prediction error signaling. This neural deficit was increased in patients with overall higher symptom severity. While RPE coding seems to be intact in patients receiving antipsychotic medication (Culbreth et al., 2016), our findings are in line with previous studies in unmedicated schizophrenia patients (Reinen et al., 2016; Schlagenhauf et al., 2014). Therefore, deficient neural coding of this core reinforcement learning mechanism may reflect a characteristic of the disorder of schizophrenia and does not result from antipsychotic medication.

Fronto-parietal connectivity and its relation to frontal glutamate in patients suffering from schizophrenia

Cognitive deficits like working memory impairment in schizophrenia are of great importance for clinical outcome, but the underlying neurobiology is not fully understood. During working memory (WM) altered connectivity patterns in the fronto-parietal network are present in patients suffering from schizophrenia [1–3]. One candidate biochemical marker for the integrity of connectivity is glutamate [4]. Here we tested for group differences in fronto-parietal connectivity and it’s relation to possible glutamatergic influences. During a functional magnetic resonance imaging (fMRI) scan, a sample of 42 medicated patients (SZ) and 41 age and gender matched healthy controls (HC) were asked to perform a numeric n-back working memory task, which consisted of two conditions “2-back” and “0-back”. FMRI was conducted on a 3T Siemens Trio scanner with a 12 channel head coil using gradient-echo echo-planar imaging. Data were further preprocessed using the statistical parametric mapping (SPM 8; Welcome Department of Imaging Neuroscience, London, UK; htt://www.fil.ion.ucl.ac.uk/spm) in MATLAB 2009b. We performed dynamic causal modeling [5] (DCM) on a model space comprising regions of interest for a visual input, parietal (PC) and dorsolateral prefrontal cortex (DLPFC). We calculated Bayesian model averages to obtain weighted connectivity parameters. One-sample t-tests were calculated on parameters for PC->DLPFC and DLPFC->PC connectivity within each group in order to evaluate modulatory effects of working memory on the executive network. Glutamate levels were measured in left DLPFC using magnetic resonance spectroscopy. We used LCmodel (Linear Combination of Model spectra, a commercial spectral fitting package) to estimate local glutamate concentration. Absolute glutamate concentrations were adjusted for grey and white matter volume. We performed group comparisons (two-sample t-test) on connectivity parameters as well as on glutamate levels. Due to non-normality two-sided Spearman correlation analysis were calculated between connectivity parameters and glutamate levels. Working memory dependent connectivity effects (between PC and DLPFC) could be observed in the left hemisphere on backward connections (PC->DLPFC, HC: t=2.77, p=0.008; SZ: t=2.62, p=0.012), whereas no significant working memory effects were present on DLPFC->PC connectivity. We found no group difference in fronto-parietal connectivity parameters and no difference in glutamate levels. Nonetheless numerically controls show higher glutamate levels as compared to patients. To explore possible effects of local Glutamate levels on working memory dependent connectivity, we correlated parameters for PC->DLPFC connectivity (where we found significant working memory dependent effects) with glutamate levels. We found a significant negative association between parieto-frontal connectivity and glutamate in DLPFC in patients (rho=-0.47, p=0.0035). Controls did not show any significant association (rho=-0.12, p=0.53). Comparing correlation coefficients, we found a trend-wise significant difference between groups (z=1.55, p=0.06). Although our data neither showed a difference in connectivity nor in glutamate levels, our findings suggest that glutamate is differentially related to working memory dependent connectivity from parietal to frontal areas in patients as compared to controls.