Introduction
Reward processing is a complex but important feature in humans, and alterations in the brain reward system have been documented in medicated and in antipsychotic-naïve schizophrenia patients (Nielsen et al., Reference Nielsen, Rostrup, Wulff, Bak, Lublin, Kapur and Glenthøj2012; Radua et al., Reference Radua, Schmidt, Borgwardt, Heinz, Schlagenhauf, McGuire and Fusar-Poli2015). Several steps are involved in reward processing, e.g. the coding of motivational salient events and prediction error (PE) during outcome evaluation. Coding of motivational salience indicates the importance of stimuli that attract attention and behavioral resources (Zink, Pagnoni, Martin-skurski, Chappelow, & Berns, Reference Zink, Pagnoni, Martin-skurski, Chappelow and Berns2004). PE, which is the coding of a mismatch between expected and obtained outcome, forms the basis for learning (Schultz & Dickinson, Reference Schultz and Dickinson2000). If PE coding is absent, learning does not occur, and the stimuli may not be assigned salience which subsequently may diminish the anticipation of a reward value (Diederen & Fletcher, Reference Diederen and Fletcher2020). Abnormalities in the coding of motivational salience and outcome evaluation may add to an impaired ability to distinguish between relevant and irrelevant sensory information. This is hypothesized to misallocate attention and salience to otherwise neutral stimuli, resulting in false associations and development of psychotic symptoms (Fletcher & Frith, Reference Fletcher and Frith2009; Heinz, Reference Heinz2002; Heinz et al., Reference Heinz, Murray, Schlagenhauf, Sterzer, Grace and Waltz2019; Kapur, Reference Kapur2003).
In healthy individuals, the anticipation of both reward and punishment activates the caudate (Knutson, Adams, Fong, & Hommer, Reference Knutson, Adams, Fong and Hommer2001), whereas an attenuated blood oxygen level-dependent (BOLD) response during anticipation of motivational salient events has been reported in striatal regions in patients with psychosis (Nielsen et al., Reference Nielsen, Rostrup, Wulff, Bak, Lublin, Kapur and Glenthøj2012; Radua et al., Reference Radua, Schmidt, Borgwardt, Heinz, Schlagenhauf, McGuire and Fusar-Poli2015). Likewise, an aberrant signal during outcome evaluation has been found in the midbrain, striatum, thalamus, prefrontal cortex, including the anterior cingulate cortex (ACC) and dorsolateral prefrontal cortex (DLPFC) (Ermakova et al., Reference Ermakova, Knolle, Justicia, Bullmore, Jones, Robbins and Murray2018; Radua et al., Reference Radua, Schmidt, Borgwardt, Heinz, Schlagenhauf, McGuire and Fusar-Poli2015; White, Kraguljac, Reid, & Lahti, Reference White, Kraguljac, Reid and Lahti2015). An exaggerated BOLD response in the prefrontal cortex during negative PE, i.e. when the outcome is worse than expected, has been reported in unmedicated patients with schizophrenia (Schlagenhauf et al., Reference Schlagenhauf, Sterzer, Schmack, Ballmaier, Rapp, Wrase and Heinz2009). However, studies on medicated patients with schizophrenia report mixed findings of intact or exaggerated negative PE signal (Walter, Kammerer, Frasch, Spitzer, & Abler, Reference Walter, Kammerer, Frasch, Spitzer and Abler2009; Waltz et al., Reference Waltz, Xu, Brown, Ruiz, Frank and Gold2018) which stresses the importance of examining antipsychotic-naïve patients to exclude confounding effect of medication.
The use of N-methyl-D-aspartate antagonists (NMDA-A), which act upon the glutamatergic system (Weckmann et al., Reference Weckmann, Deery, Howard, Feret, Asara, Dethloff and Turck2019), produces delusional beliefs and modulates PE-dependent associative learning signals in the prefrontal cortex in healthy controls (HC) (Corlett et al., Reference Corlett, Honey, Aitken, Dickinson, Shanks, Absalom and Fletcher2006). Likewise, NMDA-A have been suggested to disrupt prior expectations and the signaling of violated expectations (Corlett, Honey, Krystal, & Fletcher, Reference Corlett, Honey, Krystal and Fletcher2010). Moreover, in preclinical studies, infusion of NMDA into the thalamus enhances dopamine neuron activity in the ventral tegmental area (Zimmerman & Grace, Reference Zimmerman and Grace2016), an area involved in reward processing (Robison, Thakkar, & Diwadkar, Reference Robison, Thakkar and Diwadkar2020). The association between glutamate and reward activity has only been investigated in vivo in a few studies, which report mixed findings depending on brain regions (Bossong, Wilson, Appiah-Kusi, McGuire, & Bhattacharyya, Reference Bossong, Wilson, Appiah-Kusi, McGuire and Bhattacharyya2018; Gleich et al., Reference Gleich, Lorenz, Pöhland, Raufelder, Deserno, Beck and Gallinat2015; Jocham, Hunt, Near, & Behrens, Reference Jocham, Hunt, Near and Behrens2014; White et al., Reference White, Kraguljac, Reid and Lahti2015). Reports of a positive association between ACC glutamate and BOLD response during cognitive task exist in schizophrenia patients (Cadena et al., Reference Cadena, White, Kraguljac, Reid, Maximo, Nelson and Lahti2018; Falkenberg et al., Reference Falkenberg, Westerhausen, Craven, Johnsen, Kroken, LØberg and Hugdahl2014). In another study, a correlation between glutamate levels in substantia nigra and PE was found in HC but not in schizophrenia patients (White et al., Reference White, Kraguljac, Reid and Lahti2015).
Glutamatergic abnormalities are believed to be involved in schizophrenia (Egerton & Stone, Reference Egerton and Stone2012; Moghaddam & Javitt, Reference Moghaddam and Javitt2012; Olney & Farber, Reference Olney and Farber1995) and have been found in ACC (Bustillo et al., Reference Bustillo, Rowland, Mullins, Jung, Chen, Qualls and Lauriello2010; Kegeles et al., Reference Kegeles, Mao, Stanford, Girgis, Ojeil, Xu and Shungu2012) and the thalamus (Bojesen et al., Reference Bojesen, Ebdrup, Jessen, Sigvard, Tangmose, Edden and Glenthøj2019; Théberge et al., Reference Théberge, Bartha, Drost, Menon, Malla, Takhar and Williamson2002, Reference Théberge, Williamson, Aoyama, Drost, Manchanda, Malla and Williamson2007), which are parts of major cortico-striato-thalamo-cortical networks believed to be disrupted in psychosis (Dandash, Pantelis, & Fornito, Reference Dandash, Pantelis and Fornito2017). Glutamatergic projections from the prefrontal cortex and thalamus may modulate striatal output (Carlsson, Waters, & Carlsson, Reference Carlsson, Waters and Carlsson1999; Dandash et al., Reference Dandash, Pantelis and Fornito2017) including responses to stimuli associated with a motivational value (Matsumoto, Minamimoto, Graybiel, & Kimura, Reference Matsumoto, Minamimoto, Graybiel and Kimura2001). Therefore, it seems likely that glutamatergic activity in ACC and thalamus may modulate the processing of motivational salience and outcome evaluation.
In the present study, we primarily compared signaling of motivational salience and negative outcome evaluation (NOE) between a large group of HC and antipsychotic-naïve patients with first-episode psychosis using a region of interest (ROI) approach. For ROIs with significant group differences, we further examined the relationship with glutamate levels in ACC and left thalamus. Explorative analyses of relationships were also performed for ROI without group differences in motivational salience or NOE signal, and for ROIs in the right hemisphere and on positive outcome (PO) signaling.
We hypothesized that patients would show an attenuated motivational salience signal and an altered NOE signal, as well as an abnormal association between these signaling and glutamate levels.
Methods
The study, approved by the Danish National Committee on Biomedical Research Ethics (H-3-2013-149), was carried out in accordance with Helsinki Declaration II. Participants received thorough information about the study before providing written informed consent.
Participants
Antipsychotic-naive patients with FEP were recruited from in and out-patient clinics in the Capital Region of Denmark Mental Health Services (2014–2019) as part of a larger study previously described (Bojesen et al., Reference Bojesen, Ebdrup, Jessen, Sigvard, Tangmose, Edden and Glenthøj2019). Patients were included if they were 18–45 years of age, lifetime antipsychotic-naïve, had no prior use of central nervous system stimulants (verified by medical records), and no substance abuse in the preceding 3 months. HC were recruited from the local community through advertisement (forsøgsperson.dk). HC were matched to FEP according to age, sex, and parental educational background. Inclusion and exclusion criteria are further specified in the online Supplementary material.
Spectroscopy data from the study partially overlap with data included in two other papers [N FEP = 34, N HC = 34 (Bojesen et al., Reference Bojesen, Ebdrup, Jessen, Sigvard, Tangmose, Edden and Glenthøj2019); N FEP = 51, N HC = 51 (Bojesen et al., Reference Bojesen, Broberg, Fagerlund, Jessen, Thomas and Sigvard2020)].
Clinical assessment
For patients, symptom severity was assessed by trained raters with the Positive and Negative Syndrome Scale (PANSS) (Kay, Fiszbein, & Opler, Reference Kay, Fiszbein and Opler1987) in the same week as magnetic resonance imaging (MRI) was performed. Prior to the MRI, participants were asked about the use of drugs and performed a drug urine test (Rapid Response, Jepsen Healthcare, Tune, Denmark).
MRI data acquisition
Participants underwent structural MRI and proton magnetic resonance spectroscopy (1H-MRS) followed by functional MRI (fMRI) in one session in a 3.0 Tesla scanner (Achieva, Phillips Healthcare, Eindhoven, The Netherlands), using a 32-channel head coil (Invivo, Orlando, Florida, USA). Initially, a whole-brain 3D T1-weighted structural scan (TR 10 ms, TE 4.6 ms, flip angle = 8°, voxel size 0.79 × 0.79 × 0.80 mm3) was acquired for anatomical reference, spectroscopic voxel placement, and tissue classification of gray and white matter. Glutamate was measured using single voxel 1H-MRS [point-resolved spectroscopy sequence (PRESS): TR 3000 ms, TE 30 ms, 128 averages with multiply optimized insensitive suppression train (MOIST) water suppression] in a 2.0 × 1.5 × 2.0 cm3 voxel in the left thalamus and in a 2.0 × 2.0 × 2.0 cm3 voxel prescribed in ACC prior to the functional sequence. Mean voxel placement and spectra are shown in online Supplementary Fig. S1. The MRS voxels were prescribed in the ACC and left thalamus based on the previous findings of abnormalities in glutamatergic measures (Bustillo et al., Reference Bustillo, Rowland, Mullins, Jung, Chen, Qualls and Lauriello2010; Théberge et al., Reference Théberge, Bartha, Drost, Menon, Malla, Takhar and Williamson2002) and because these regions are part of cortico-striato-thalamo-cortical networks believed to be dysregulated in psychotic disorders and implicated in reward processing.
For the fMRI, 336 echo-planar images were acquired (TR 2000 ms, TE 25 ms, flip angle = 75°, 38 slices and voxel size of 2.8 × 2.97 × 2.4 mm3). To minimize motion artifacts, patients were instructed not to move their heads during scans.
fMRI task
Brain reward activity was examined with fMRI while participants played a variant of the MID task (Knutson, Westdorp, Kaiser, & Hommer, Reference Knutson, Westdorp, Kaiser and Hommer2000; Uldall et al., Reference Uldall, Nielsen, Carlsson, Glenthøj, Siebner, Madsen and Rostrup2020), a task widely used to probe the neural activity of anticipation and outcome. The paradigm used in the present study included trials with the possibility of winning or losing money and neutral stimuli only (Fig. 1). The task lasted 12 min and comprised 72 interactive trials that were evenly distributed between winning or losing money and neutral trials. The task adapted to the individual reaction time to provide a hit rate of 66%. Participants were instructed about the task, the meaning of the cues, the possibility of monetary gain, and practiced the task for 5 min before data acquisition. Participants were not informed about the adaptive hit rate. All participants correctly believed that they would receive money upon completion of the task. Hence, participants had expectations of monetary gain. For detailed description, see online Supplementary material.
The modified monetary incentive delay task used in the present study. First, a cue was presented on the screen representing a trial with possible win, possible loss, or a neutral condition. This was followed by a waiting phase (cross) and a target cue (white box), where participants pressed a button as fast as possible, expecting to win or avoid losses. Afterwards, the trial outcome appeared on the screen. In possible win trials, participants gained Euro 7 on a hit and Euro 0 on a miss. In possible loss trials, participants earned Euro 0 on a hit and lost Euro 7 on a miss. Neutral trials resulted in Euro 0 every time.
fMRI analysis
Analyses of fMRI data were performed using tools from FMRIB (Oxford Centre for Functional Magnetic Resonance Imaging of the Brain) Software Library fmrib.ox.ac.uk/fsl. First-level analyses were carried out using the FSL fMRI Expert Analysis Tool. Functional images were corrected for slice timing and motion effects, realigned, spatially smoothed with a 5 mm full-width at half-maximum Gaussian kernel. A high pass filter was applied with a 200 s cutoff. The images were co-registered to the corresponding T1-weighted image and normalized to Montreal Neurological Institute (MNI) space (MNI, Quebec, Canada). We used a general linear model consisting of nine predictors and their temporal derivative to analyze data. Three predictors modeled each of the cues, one predictor indicated button press, and five predictors defined the different outcomes: win, lose, hit, miss, and neutral. All predictors were convolved with the hemodynamic response function. Our contrasts of interest were a contrast of joint effect of the anticipation of win and loss v. neutral, and a contrast of outcome miss v. outcome neutral, see online Supplementary Fig. S3 for task design. The former contrast is hereinafter referred to as motivational salience, and the latter contrast as NOE.
Explorative analyses were performed for PO (outcome hit v. neutral outcome). The mean percent signal change for the contrasts was extracted from predefined ROIs to be used for group comparison and correlations with glutamate levels. For illustrative purpose, the contrasts of interest were taken to second-level analysis for a whole-brain group comparison. The resulting z-statistic images were thresholded using clusters determined by Z > 2.3 and corrected significance threshold of p = 0.05 (Worsley, Reference Worsley, Jezzard, Matthews and Smith2001).
MRS analysis
PRESS acquisitions were analyzed using LCModel version 6.3-1L (http://s-provencher.com/lcmodel.shtml) (Provencher, Reference Provencher1993) and fitted in the spectral range 0.2–4.0 ppm, as previously described (Bojesen et al., Reference Bojesen, Ebdrup, Jessen, Sigvard, Tangmose, Edden and Glenthøj2019). Unsuppressed water reference spectra were acquired separately as inbuild sequences in the PRESS sequences. The in vivo water-scaled values of metabolites reported by LCModel were corrected for partial volume cerebral spinal fluid to estimate concentration in institutional units (IU) (Stone et al., Reference Stone, Dietrich, Edden, Mehta, De Simoni, Reed and Barker2012). Details of 1H-MRS acquisition, quality assessment and analyses are reported in online Supplementary material, including the illustration of mean voxel placements, representative spectra (online Supplementary Fig. S1), and analyses with the correction of gray matter and with glutamate + glutamine (glx).
ROIs
Since glutamatergic measures were assessed in ACC and in the left thalamus only, we also extracted fMRI activity based on ROIs in the left hemisphere. For explorative analyses, fMRI activity was extracted from ROIs in the right hemisphere. The striatal ROIs were defined as a 6 mm radius spherical region centered in the MNI coordinates: −10, 12, 8 (caudate) and −10, 14, −6 (accumbens), in accordance with published studies (Nielsen, Rostrup, Broberg, Wulff, & Glenthøj, Reference Nielsen, Rostrup, Broberg, Wulff and Glenthøj2018; Nielsen, Rostrup, Wulff, Glenthøj, & Ebdrup, Reference Nielsen, Rostrup, Wulff, Glenthøj and Ebdrup2016; Zink, Pagnoni, Martin, Dhamala, & Berns, Reference Zink, Pagnoni, Martin, Dhamala and Berns2003). The ROIs in ACC, DLPFC, and thalamus were defined as a 5 mm radius spherical region centered in the MNI coordinates −5, 39, 20 (ACC), −46, 38, 8 (DLPFC), and −7, −17, 5 (thalamus), in accordance with the previous findings of monetary outcome processing in DLPFC area within BA 46 (Waltz et al., Reference Waltz, Schweitzer, Ross, Kurup, Salmeron, Rose and Stein2010), ACC within BA 32 (Knutson et al., Reference Knutson, Westdorp, Kaiser and Hommer2000), and thalamus (Oldham et al., Reference Oldham, Murawski, Fornito, Youssef, Yücel and Lorenzetti2018) converted to MNI space using Talairach Daemon http://sprout022.sprout.yale.edu/mni2tal/mni2tal.html.
Overlap of the fMRI ROIs and spectroscopic voxels is provided in online Supplementary Fig. S2.
Statistics
Analyses were performed using IBM SPSS Statistics 25.
Group differences in demographic, clinical variables, motivational salience, and outcome evaluation signal were analyzed using independent t tests and χ2 tests. Glutamate levels were analyzed using ANOVA, with and without correction for covariates sex, age, and smoking due to previously shown impact on glutamate (Marsman et al., Reference Marsman, Van Den Heuvel, Klomp, Kahn, Luijten and Hulshoff Pol2013; O'Gorman, Michels, Edden, Murdoch, & Martin, Reference O'Gorman, Michels, Edden, Murdoch and Martin2011). To correct for multiple comparisons, significant level of group differences in the five left fMRI ROIs was set to p = 0.05/5 = 0.01.
Regression analysis was used to test the association between motivational salience or NOE signal and glutamate levels in ACC and thalamus separately for HC and FEP with and without correcting for covariates. For these regression analyses, the main outcome was analyses involving the left fMRI ROIs with significant group differences in BOLD responses, and to correct for multiple comparisons, the significance level for the association between glutamate voxels and two fMRI ROI with motivational salience signal and NOE signal was set to p < 0.05/2 × 2 × 2 = 0.006. Explorative regression analyses between three left fMRI ROIs with non-significant group differences in fMRI measures and glutamate voxels were set to p < 0.05/2 × 3 × 2 = 0.004.
Behavioral measures of hit rate and response time were analyzed using ANOVA with trial type (three levels: possible win, possible lose, neutral trial) as within-subject factor and group as between-subject factor.
Explorative analyses of correlations between PANSS scores and imaging measures were tested using spearman correlation and to correct for multiple correlations the significance level was set to p < 0.05/(4 PANSS subscores×12 imaging measures) = 0.001.
Results
A total of 103 participants were included, herein 52 HC and 51 FEP. The majority of FEP were diagnosed with schizophrenia (n = 39). Table 1 presents the demographic and clinical characteristics. There were no group differences in age, handedness, parental educational background, sex, and smoking status or cannabis use (all p > 0.15).
Demographic and clinical characteristics of participants
FEP, first-episode psychosis patients; HC, healthy controls; N, number of subjects; s.d., standard deviation; PANSS, Positive and Negative Syndrome Scale.
* Current use of cannabis was less than once a month, aindependent t test, bχ2 test, cFisher's exact test.
Patients were less educated (p = 0.001) and used benzodiazepine more often (p = 0.01).
Behavioral data
The mean monetary gain was Euro 76 with no group difference [T(101) = 0.10, mean difference 0.048, confidence interval (CI) −9.6 to 9.7; p = 0.99].
Analysis of hit rate showed no effect of group [F (1, 101) = 0.02, p = 0.89] and no significant group×trial type interaction. There was a main effect of trial type [F (1, 101) = 64.9, p < 0.001], and post hoc tests showed a difference between hit rates of neutral trials and trials with possible win (p < 0.001) or possible loss (p < 0.001), with the lowest hit rates in neutral trials.
The analysis of response time showed a main effect of group [F (1, 101) = 6.2, p = 0.015], with FEP showing a higher response time but no effect of trial type [F (1, 101) = 0.84, p = 0.36] and no group×trial type interaction, see online Supplementary material and Table S7.
Thus, all participants understood the importance of the cues, and no group difference in hit rate omit the confounding effect of behavioral data on the analyses in motivational salience and NOE signal.
Motivational salience and NOE signal
For the motivational salience signal, there were no group differences in any ROIs [caudate (T(101) = −0.9, p = 0.35, CI −0.11 to 0.04), accumbens (T(101) = −1.3, p = 0.18, CI −0.10 to 0.02), DLPFC (T(101) = −0.08, p = 0.94, CI −0.09 to 0.08), ACC (T(101) = −0.02, p = 0.98, CI −0.06 to 0.06), and thalamus (T(101) = −0.2, p = 0.80, CI −0.9 to 0.07)], see Fig. 2.
Responses in predefined left regions of interest to motivational salience and negative outcome evaluation divided by groups. Extraction of parameter estimates was performed using FMRIB (Oxford Centre for Functional Magnetic Resonance Imaging of the Brain) Software Library tools. Group differences were found for negative outcome evaluation in the caudate (p = 0.001) and in the dorsolateral prefrontal cortex (p = 0.003).
For NOE signal, there was a group difference, with FEP showing a positive contrast signal which was not found in HC in the caudate [T(101) = 3.4, p = 0.001, CI 0.07–0.28] and in DLPFC [T(101) = 3.1, p = 0.003, CI 0.07–0.34] but not in accumbens [T(101) = 1.8, p = 0.07, CI −0.009 to 0.19], thalamus [T(101) = 1.9, p = 0.06, CI −0.006 to 0.21], or ACC [T(101) = 1.7, p = 0.09, CI −0.1 to 0.17], see Fig. 2.
Explorative analyses for NOE signal on ROIs in the right hemisphere were the same as for the left hemisphere, and analyses for PO signal showed no group difference, see online Supplementary material and Table S1.
Glutamate levels in ACC and left thalamus
Glutamate measures in ACC in three FEP were excluded, while glutamate measures in the thalamus in one FEP and three HC were excluded.
There was no group difference in glutamate levels [thalamus: F (1, 97) = 0.39, p = 0.53, CI −0.24 to 0.46; ACC: F (1, 98) = 0.98, p = 0.32, CI −0.42 to 0.14], nor when controlled for covariates (thalamus: p = 0.24, CI 0.54–0.14; ACC: p = 0.33, CI −0.14 to 0.41). The main effects of age, sex, and smoking status are reported in online Supplementary results as well as mean values, 95% CI, and glutamate measures corrected for the content of gray matter.
Association between motivational salience or NOE signal and glutamate levels
There was a different association between NOE signal in the left caudate and left DLPFC and thalamic glutamate levels in FEP and HC [significant interactions: caudate F (1, 95) = 7.2, p = 0.009; DLPFC: F (1,95) = 7.1, p = 0.009] due to a negative correlation in FEP (caudate: β = −0.40, p = 0.004, CI −0.62 to −0.12; DLPFC: β = −0.39, p = 0.005, CI −0.69 to −0.13), also after adjustment for covariates (caudate: β = −0.55, p = 0.001, CI −0.44 to −0.15; DLPFC: β = −0.48, p = 0.002, CI −0.82 to −0.19), but not in HC (caudate: p = 0.37, CI −0.15 to 0.40; DLPFC: p = 0.46, CI −0.12 to 0.39) (Fig. 3).
Correlation between thalamic glutamate levels and negative outcome evaluation signaling in the left caudate (a) and left DLPFC (b) in antipsychotic-naïve patients with first-episode psychosis and healthy controls. Using regression analysis, a negative correlation was found in the former (caudate: β = −0.44, p = 0.004, DLPFC: β = −0.39, p = 0.005) but not in the latter (p = 0.37).
For ROIs with no group differences in fMRI measures, explorative correlations between BOLD response and glutamate levels in FEP showed an association between thalamic glutamate and NOE signal in left thalamus (r = −0.39, CI −0.65 to −0.12, p = 0.005), which did not survive Bonferroni correction. No other correlations were found between motivational salience, NOE signal or PO signal, and glutamate levels in thalamus or ACC, nor when correcting for covariates, when correcting for gray matter, or when performing analyses with glx measures, see online Supplementary material and Tables S2–S4.
Correlations between PANSS scores in patients and imaging measures
Explorative correlations showed no correlations between PANSS scores and imaging measures (all p > 0.07), see online Supplementary material and Table S6.
Whole-brain analysis of motivational salience and NOE signal
NOE
Analysis showed significant group differences in parts of several brain areas (left thalamus, left superior and inferior frontal gyrus, left middle and superior temporal gyrus, intra-calcarine cortex, and occipital fusiform gyrus), including regions partly overlapping predefined ROIs with higher signaling in FEP. HC showed no brain areas with increased NOE signal compared to FEP (online Supplementary Fig. S3).
Motivational salience
No group difference was found, though HC displayed increased signal in areas of striatum, midbrain, and occipital regions, and FEP did not (online Supplementary Fig. S4).
Discussion
Our primary outcome examining group differences in motivational salience and NOE signal in predefined ROIs showed a positive signal change in FEP during NOE in the caudate and DLPFC compared to HC. For the motivational salience signal, no significant group differences were found. Our secondary outcome examining associations between glutamate levels and motivational salience or NOE signal showed an inverse correlation between thalamic glutamate levels and NOE signal in the caudate and DLPFC of FEP only.
To our knowledge, this is the largest study to date investigating signaling of motivational salience and NOE using a MID task in a large cohort of antipsychotic-naïve patients with FEP. In addition, this is the first study exploring associations between glutamate levels and coding of motivational salience and NOE in antipsychotic-naïve patients with FEP.
In general, fMRI studies in patients with schizophrenia report hypoactivation in the ventral striatum during anticipation of monetary reward (Radua et al., Reference Radua, Schmidt, Borgwardt, Heinz, Schlagenhauf, McGuire and Fusar-Poli2015). Our group has previously reported attenuated signal using a comparable contrast of motivational salience in antipsychotic-naïve schizophrenia patients (Nielsen et al., Reference Nielsen, Rostrup, Wulff, Bak, Lublin, Kapur and Glenthøj2012). We did not replicate these findings in the present study. Importantly, the MID task used in the present study was modified to include cues with possible win or loss only, leading to higher frequency of salient trials which may have introduced habituation in HC (Avery et al., Reference Avery, McHugo, Armstrong, Blackford, Woodward and Heckers2019). Further, an effect of uncertainty on tonic dopamine firing has been suggested (Mikhael & Bogacz, Reference Mikhael and Bogacz2016), thus the presence of uncertainty in the cues involved may confound the response to motivational salience. In addition, the present study included patients with FEP and not exclusively patients with schizophrenia. Moreover, previous studies have reported associations between attenuated response to anticipation of salient events and increased level of positive (Esslinger et al., Reference Esslinger, Englisch, Inta, Rausch, Schirmbeck, Mier and Zink2012; Nielsen et al., Reference Nielsen, Rostrup, Wulff, Bak, Lublin, Kapur and Glenthøj2012) and negative symptoms in patients with schizophrenia (Radua et al., Reference Radua, Schmidt, Borgwardt, Heinz, Schlagenhauf, McGuire and Fusar-Poli2015). In the present study, the mean PANSS scores were lower compared to other studies of unmedicated/antipsychotic-naïve patients with schizophrenia (Esslinger et al., Reference Esslinger, Englisch, Inta, Rausch, Schirmbeck, Mier and Zink2012; Nielsen et al., Reference Nielsen, Rostrup, Wulff, Bak, Lublin, Kapur and Glenthøj2012).
We found a positive signal change during NOE in the caudate and DLPFC in FEP. Previous studies of HC report striatal activation during unexpected rewards and hypoactivation in striatal and prefrontal regions during unexpected unsuccessful outcomes, i.e. omission of rewards or loss of money (Delgado, Nystrom, Fissell, Noll, & Fiez, Reference Delgado, Nystrom, Fissell, Noll and Fiez2000; Kim, Shimojo, & O'Doherty, Reference Kim, Shimojo and O'Doherty2006; Knutson, Fong, Bennett, Adams, & Hommer, Reference Knutson, Fong, Bennett, Adams and Hommer2003; Morris et al., Reference Morris, Vercammen, Lenroot, Moore, Langton, Short and Weickert2012; Schlagenhauf et al., Reference Schlagenhauf, Sterzer, Schmack, Ballmaier, Rapp, Wrase and Heinz2009). Studies of patients with schizophrenia have shown, in contrast to HC, an exaggerated response in prefrontal and striatal regions when expected rewards were not delivered (Schlagenhauf et al., Reference Schlagenhauf, Sterzer, Schmack, Ballmaier, Rapp, Wrase and Heinz2009; Walter et al., Reference Walter, Kammerer, Frasch, Spitzer and Abler2009) which is in line with our results.
These findings may be explained by altered responses to neutral events in people with psychosis (Maia & Frank, Reference Maia and Frank2017) or that FEP coded NOE in an unsigned fashion, which indicates surprise without valence (Haarsma et al., Reference Haarsma, Fletcher, Griffin, Taverne, Ziauddeen, Spencer and Murray2020). Our NOE contrast was defined as miss v. neutral outcome, however, additional explorative analyses on neutral outcome, and on PO signaling, showed no alterations in FEP, see online Supplementary material. Hence, the NOE signal in FEP in this study may not be influenced by alterations in neutral responses or unsigned coding.
Findings in the literature, however, are not consistent since a negative signal change and no group difference have also been reported (Koch et al., Reference Koch, Schachtzabel, Wagner, Schikora, Schultz, Reichenbach and Schlösser2010; Morris et al., Reference Morris, Vercammen, Lenroot, Moore, Langton, Short and Weickert2012; Waltz et al., Reference Waltz, Xu, Brown, Ruiz, Frank and Gold2018). Importantly, group differences in signal during outcome evaluation may depend on the design of the contrast used to analyze brain responses, where increased activity during the evaluation of neutral outcome or expected rewards (Jensen et al., Reference Jensen, Willeit, Zipursky, Savina, Smith, Menon and Kapur2008; Murray et al., Reference Murray, Corlett, Clark, Pessiglione, Blackwell, Honey and Fletcher2008) in patients with schizophrenia may affect the signal of the contrast defined in various studies. Moreover, some argue that brain responses may differ between receiving a punishment and not receiving a reward depending on individual sensitivity to punishment and reward (Boksem, Tops, Kostermans, & De Cremer, Reference Boksem, Tops, Kostermans and De Cremer2008) and, to some degree, may involve different neural processes (Boksem et al., Reference Boksem, Tops, Kostermans and De Cremer2008; Matsumoto, Reference Matsumoto2008; Matsumoto & Hikosaka, Reference Matsumoto and Hikosaka2007).
Processing of aversive outcome in HC seems to involve DLPFC, thalamus, and ACC, where the brain responses to aversive outcomes decrease, with decreased expectation of the outcome indicating learning (Dunsmoor, Bandettini, & Knight, Reference Dunsmoor, Bandettini and Knight2008). In contrast, patients with schizophrenia have shown impaired learning of PE, with an inverse association between learning rate and brain response activity in DLPFC and thalamus (Koch et al., Reference Koch, Schachtzabel, Wagner, Schikora, Schultz, Reichenbach and Schlösser2010). Thus, impaired learning may contribute to the altered NOE signal in FEP.
Contrasting previous findings on alterations in positive PE coding in early psychosis patients (Ermakova et al., Reference Ermakova, Knolle, Justicia, Bullmore, Jones, Robbins and Murray2018), we found no group difference in PO signaling, maybe because the task involved was less suited for the evaluation of positive PE as there was a high hit rate (Waltz et al., Reference Waltz, Schweitzer, Ross, Kurup, Salmeron, Rose and Stein2010).
Our secondary outcome showed an inverse association between thalamic glutamate level and NOE signal in the caudate and DLPFC in FEP but not in HC. Patients with higher levels of glutamate displayed a less positive contrast signal during NOE, and this may suggest a possible compensatory mechanism of glutamate on NOE signaling in these patients. Suggested to be a key component in regulating the reward circuit (Haber & Knutson, Reference Haber and Knutson2009), the thalamus may balance disturbances in striatal NOE signaling through glutamatergic projections to inhibitory striatal GABAergic neurons (Carlsson et al., Reference Carlsson, Waters and Carlsson1999; Dandash et al., Reference Dandash, Pantelis and Fornito2017; Nanda, Galvan, Smith, & Wichmann, Reference Nanda, Galvan, Smith and Wichmann2009) or through the ventral tegmental area (Zimmerman & Grace, Reference Zimmerman and Grace2016). In line with our findings, a study on individuals at ultra-high risk of psychosis has shown a negative correlation between thalamic glutamate levels and DLPFC functional responses, and suggests that variations in thalamic glutamate can affect cortical function (Fusar-Poli et al., Reference Fusar-Poli, Stone, Broome, Valli, Mechelli, McLean and McGuire2011). However, the association between NOE signal and thalamic glutamate levels was insignificant when corrected for gray matter and when investigating the association with glx levels (see online Supplementary Tables S3 and S4).
We did not find any correlations between ACC glutamate levels and NOE signal. Other studies have reported an association between ACC glutamate and BOLD response in patients with schizophrenia (Cadena et al., Reference Cadena, White, Kraguljac, Reid, Maximo, Nelson and Lahti2018; Falkenberg et al., Reference Falkenberg, Westerhausen, Craven, Johnsen, Kroken, LØberg and Hugdahl2014), which was not present or reversed in HC (Falkenberg et al., Reference Falkenberg, Westerhausen, Craven, Johnsen, Kroken, LØberg and Hugdahl2014; Gleich et al., Reference Gleich, Lorenz, Pöhland, Raufelder, Deserno, Beck and Gallinat2015). This, however, was primarily observed when examining cognitive functions with the Stroop color task and the auditory speech perception task. To some extent, the mixed results can be explained by experimental design and the ROIs examined, as well as the effects of medication and number of subjects included (Bossong et al., Reference Bossong, Wilson, Appiah-Kusi, McGuire and Bhattacharyya2018; Cadena et al., Reference Cadena, White, Kraguljac, Reid, Maximo, Nelson and Lahti2018; Falkenberg et al., Reference Falkenberg, Westerhausen, Craven, Johnsen, Kroken, LØberg and Hugdahl2014; Jocham et al., Reference Jocham, Hunt, Near and Behrens2014; White et al., Reference White, Kraguljac, Reid and Lahti2015).
Thalamic glutamate levels in IU were not increased in FEP, but after adjustment for gray matter, there was a trend of higher glutamate levels in FEP compared to HC. We have previously found increased measures of thalamic glutamate levels (Bojesen et al., Reference Bojesen, Ebdrup, Jessen, Sigvard, Tangmose, Edden and Glenthøj2019). These variations may be explained by differences in symptom severity or diagnosis, which may affect glutamate levels (Merritt, Egerton, Kempton, Taylor, & McGuire, Reference Merritt, Egerton, Kempton, Taylor and McGuire2016). No group difference in ACC glutamate levels was found, which is in line with a recent meta-analysis of antipsychotic-naïve/free patients (Iwata et al., Reference Iwata, Nakajima, Plitman, Mihashi, Caravaggio, Chung and Graff-Guerrero2018). Variations in studies may be explained by prior exposure to antipsychotic medication (Bustillo et al., Reference Bustillo, Rowland, Mullins, Jung, Chen, Qualls and Lauriello2010; Kegeles et al., Reference Kegeles, Mao, Stanford, Girgis, Ojeil, Xu and Shungu2012) or examinations in more dorsal regions of ACC (Reid et al., Reference Reid, Salibi, White, Gawne, Denney and Lahti2019; Wang et al., Reference Wang, Pradhan, Coughlin, Trivedi, Dubois, Crawford and Barker2019).
Our data did not indicate that glutamate levels in the thalamus or ACC are associated with reward processing on a more general level, since no association was found with motivational salience. Rather, glutamate levels in the thalamus were specifically related to caudate and DLPFC NOE signaling in FEP. Complex cognitive processes appear to be involved in the encoding of outcome evaluation (Heinz et al., Reference Heinz, Murray, Schlagenhauf, Sterzer, Grace and Waltz2019), which theoretically may be sensitive to glutamatergic activity in patients with schizophrenia (Corlett et al., Reference Corlett, Honey, Aitken, Dickinson, Shanks, Absalom and Fletcher2006, Reference Corlett, Honey, Krystal and Fletcher2010; Honey et al., Reference Honey, Honey, Sharar, Turner, Pomarol-Clotet, Kumaran and Fletcher2005; Vinckier et al., Reference Vinckier, Gaillard, Palminteri, Rigoux, Salvador, Fornito and Fletcher2016). It has been suggested that higher levels of glutamate can act as a buffer, preventing patients with schizophrenia from showing marked cognitive impairments (Falkenberg et al., Reference Falkenberg, Westerhausen, Craven, Johnsen, Kroken, LØberg and Hugdahl2014) which is in line with our results of a more normalized NOE signal in FEP with higher levels of thalamic glutamate.
A strength of the present study was the multimodal approach and the inclusion of antipsychotic-naïve FEP, thus excluding the possible impact of antipsychotic medication. One of the limitations is that the study cohort represented only moderately ill patients, which may affect results. In addition, the task and contrast used did not measure a learning estimate of outcome evaluation, and may not formally test a PE model, however our findings are compatible with a disturbed processing of outcome evaluation in FEP. Applying larger cortical ROIs would be preferable, as previous studies show more widespread cortical activity for reward tasks (Bartra, McGuire, & Kable, Reference Bartra, McGuire and Kable2013; Garrison, Erdeniz, & Done, Reference Garrison, Erdeniz and Done2013).
Finally, the assessment of glutamate was limited to voxels in the thalamus and ACC, and the inclusion of MRS voxels in the striatum and DLPFC would be preferable.
In conclusion, we found an altered NOE signal in caudate and DLPFC in FEP as part of the pathophysiology of schizophrenia. In addition, our findings indicated a possible link between the levels of thalamic glutamate and signaling of NOE in patients with first-episode psychosis.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0033291721003305
Acknowledgement
We thank Gitte Saltoft, Helle Schæbel, and Mikkel Sørensen for assistance with participants and producing the figures.
Financial support
This work was supported by a Ph.D. grant from the Mental Health Services, Capital Region of Denmark (Tangmose); a Ph.D. grant from the Faculty of Health and Medical Sciences, University of Copenhagen (Bojesen); Ph.D. grants and a postdoc grant from the Mental Health Services, Capital Region of Denmark (Sigvard, Jessen, Nielsen); an independent grant from the Lundbeck Foundation (R155-2013-16337) to the Lundbeck Foundation Center of Excellence Center for Clinical Intervention and Neuropsychiatric Schizophrenia Research (CINS) (Glenthøj) and support from the Mental Health Services, Capital Region of Denmark (Glenthøj). The funding sources played no role in the study design, collection, analysis, and interpretation of data, writing the report, or the decision to submit the paper for publication.
Conflict of interest
None.
Ethical standards
The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008.
