Abstract
Background
A hyperdopaminergic state in the striatum is a corner stone in psychosis research (1, 2). However, how this neurocognitively translates into symptoms is still unknown. In that sense, the reward prediction error (RPE; eq.1) reflecting the difference between a received reward and its anticipation has been proposed as such a mediating mechanism. Previous studies in unmedicated schizophrenia patients revealed deficient RPE coding (3, 4, 5), while these signals seemed to be intact when patients received antipsychotic medication (6). Thus, the literature points towards a relation between dopamine and neural RPE coding. To our knowledge, both measures have never been tested within one schizophrenia sample. The aim of the current study was to investigate striatal dopamine synthesis capacity and RPE coding as well as their potential correlation in a sample of unmedicated schizophrenia patients.
Methods
In this multimodal imaging study, 21 unmedicated schizophrenia patients (Sz) and 23 matched healthy controls (HC) underwent 1) an 18F-fluorodopa PET scan measuring presynaptic dopamine synthesis capacity as well as 2) fMRI scanning where they performed a reversal learning task (7). Via computational modeling of choice data in this task (3.), individual RPE trajectories were fitted. We compared Sz and HC in their Ki-values (influx rate constants of 18F-fluorodopa) for the striatal subregions (sensorimotor, associative, limbic) as well as regarding their striatal RPE coding. In a voxel-based analysis, we correlated individual RPE coding in the striatum with striatal Ki-values while also including a group factor.
Results
Sz did not differ from HC in any of the striatal subregions in terms of dopamine synthesis capacity (F=0.64, p=0.70). A post-hoc analysis revealed that when taking into account comorbid alcohol abuse, that Sz without alcohol abuse displayed increased Ki-values, whereas Sz with alcohol abuse displayed Ki-values comparable to HC (F=4.65, p=0.016 for sensorimotor striatum; F=5.11, p=0.011 for associative striatum). The model-based fMRI analysis revealed a main effect for Reward Prediction Error in the bilateral ventral striatum; [-10 12 10], t = 7.40, pFWE < 0.0001 and [10 12 -10], t = 6.56, pFWE = 0.006. Sz showed decreased striatal RPE coding compared to HC (t = 3.69, pSVC for nucleus accumbens = .015) and this decrease correlated with negative symptoms ([12 8 -10], t = 3.23, pSVC for striatal PE = 0.035). In a voxel-based analysis there was a significant interaction with group, so that the RPE signal positively correlated with the limbic dopamine synthesis capacity only in HC, but not in Sz ([12 8 -12], F=11.2, pSVC for striatal PE=0.031).
Discussion
While we could not replicate the finding of heightened striatal dopamine synthesis capacity in the associative striatum of Sz at first, our control analyses revealed increased dopamine levels only in Sz who were not diagnosed with a comorbid alcohol abuse, while those with a respective comorbidity showed comparably normal dopamine levels. In our comparably small sample we cannot rule out other confounding factors of this group and thus future studies with larger samples are warranted for testing whether this was due to a direct effect of alcohol on dopamine synthesis capacity. In line with previous findings, the striatal RPE coding was significantly decreased in Sz and this deficit correlated with PANSS negative symptoms. In healthy controls, increased limbic striatal dopamine correlated with higher RPE coding. In Sz, this correlation was absent. Further, our findings highlight the functional dissociations between the striatal subdivisions, since the RPE finding and correlation both refer to the limbic and thus not to the associative part where dopamine levels were increased.
