Cortical stress regulation is disrupted in schizophrenia but not in clinical high risk for psychosis

See Gomes and Grace (doi:10.1093/brain/awy156) for a scientific commentary on this article.

Schifani et al. investigate the relationship between prefrontocortical dopamine release and stress response in schizophrenia. Associations between dopamine release and cortisol response to stress observed in controls and those at clinical high risk for psychosis are absent in schizophrenia, providing direct evidence of disrupted prefrontocortical dopamine-stress regulation in schizophrenia.

Abstract

While alterations in striatal dopamine in psychosis and stress have been well studied, the role of dopamine in prefrontal cortex is poorly understood. To date, no study has investigated the prefrontocortical dopamine response to stress in the psychosis spectrum, even though the dorsolateral and medial prefrontal cortices are key regions in cognitive and emotional regulation, respectively. The present study uses the high-affinity dopamine D_2/3 receptor radiotracer ¹¹C-FLB457 and PET together with a validated psychosocial stress challenge to investigate the dorsolateral and medial prefrontocortical dopamine response to stress in schizophrenia and clinical high risk for psychosis. Forty participants completed two ¹¹C-FLB457 PET scans (14 antipsychotic-free schizophrenia, 14 clinical high risk for psychosis and 12 matched healthy volunteers), one while performing a Sensory Motor Control Task (control) and another while performing the Montreal Imaging Stress Task (stress). Binding potential (BP_ND) was estimated using Simplified Reference Tissue Model with cerebellar cortex as reference region. Dopamine release was defined as per cent change in BP_ND between control and stress scans (ΔBP_ND) using a novel correction for injected mass. Salivary cortisol response (ΔAUC_I) was assessed throughout the tasks and its relationship with dopamine release examined. ¹¹C-FLB457 binding at control conditions was significantly different between groups in medial [F(2,37) = 7.98, P = 0.0013] and dorsolateral [F(2,37) = 6.97, P = 0.0027] prefrontal cortex with schizophrenia patients having lower BP_ND than participants at clinical high risk for psychosis and healthy volunteers, but there was no difference in ΔBP_ND among groups [dorsolateral prefrontal cortex: F(2,37) = 1.07, P = 0.35; medial prefrontal cortex: F(2,37) = 0.54, P = 0.59]. We report a positive relationship between ΔAUC_I and ¹¹C-FLB457 ΔBP_ND in dorsolateral and medial prefrontal cortex in healthy volunteers (r = 0.72, P = 0.026; r = 0.76, P = 0.014, respectively) and in participants at clinical high risk for psychosis (r = 0.76, P = 0.0075; r = 0.72, P = 0.018, respectively), which was absent in schizophrenia (r = 0.46, P = 1.00; r = 0.19, P = 1.00, respectively). Furthermore, exploratory associations between ΔBP_ND or ΔAUC_I and stress or anxiety measures observed in clinical high risk for psychosis were absent in schizophrenia. These findings provide first direct evidence of a disrupted prefrontocortical dopamine-stress regulation in schizophrenia.

Introduction

Schizophrenia is a debilitating mental disease with a complex aetiology. It is believed to be caused by both genetic predisposition and environmental factors. Environmental risk factors for schizophrenia include psychosocial stress such as developmental trauma, growing up in an urban environment, and social defeat, among others (van Os et al., 2010). The impact of psychosocial stress on schizophrenia can be explained by the vulnerability-stress hypothesis. This model proposes that an endogenous diathesis/vulnerability interacts with internal or external stressors in the development of psychotic disorders (Walker and Diforio, 1997). While the importance of striatal dopamine in psychosis and stress has been well studied (Laruelle, 2000; Mizrahi et al., 2012), the role of dopamine in the prefrontal cortex (PFC) is still poorly understood.

The PFC (including its subregions the dorsolateral and medial PFC, dlPFC and mPFC, respectively) is well known for its crucial role in planning, controlling and directing behaviour in response to changing environmental demands (Miller and Cohen, 2001). The dlPFC is extensively connected with sensory and motor cortices and key region to regulate cognitive demand (Goldman-Rakic, 1995). The mPFC is extensively connected to subcortical regions that generate emotional responses such as amygdala and hypothalamus (Öngür and Price, 2000). There is compelling evidence that cognitive and emotional regulation is impaired in schizophrenia (Green et al., 2004; Holt et al., 2009), hence the importance of these two brain regions.

In contrast to the well-described striatal hyperdopaminergic state in schizophrenia, the PFC has been proposed to exhibit a hypodopaminergic response (Weinstein et al., 2017). This theory is supported primarily by preclinical schizophrenia models displaying attenuated mPFC dopamine signalling (Watt et al., 2009; Burke et al., 2010) and higher stress sensitivity (Gomes et al., 2016). However, direct evidence about a PFC hypodopaminergic state in living humans is still limited due, in part, to the lack of suitable PET tracers sensitive enough for the low density of dopamine D_2/3 receptors in cortex. The high-affinity radiotracer ¹¹C-FLB457 is a validated tool for cortical dopamine release quantification, reported to have ≤15% test-retest variability (Narendran et al., 2011b). Furthermore, ¹¹C-FLB457 is superior to other cortical D₂ receptor radioligands such as ¹¹C-fallypride as it displays a higher signal-to-noise ratio in cortical areas (Narendran et al., 2009). Using ¹¹C-FLB457, Slifstein et al. (2015) showed generally blunted dopamine release following an amphetamine challenge in various extrastriatal regions including a significant reduction in the dlPFC in first-episode psychosis. Furthermore, reduced dopamine release was associated with reduced working memory-related activation of the dlPFC in this population. While Hernaus et al. (2015) did investigate dopamine response to stress in mPFC previously using ¹⁸F-fallypride in a single-scan paradigm and reported no difference between individuals with non-affective psychotic disorder and healthy volunteers, so far, no data are available on PFC stress regulation in schizophrenia or its putative prodrome.

The current study aimed to examine stress-induced PFC dopamine response in patients with schizophrenia and those at clinical high risk (CHR) for psychosis using a validated two-scan paradigm with ¹¹C-FLB457 PET. We hypothesized a reduction in mPFC and dlPFC dopamine release in response to the stress challenge given the previous study by Slifstein et al. (2015) and supporting preclinical literature on cortical dopamine response to stress (Watt et al., 2009; Burke et al., 2010). Furthermore, we examined the relationship between stress-induced PFC dopamine release and salivary cortisol response, for the first time in CHR and schizophrenia, and explored associations with stress-related behaviours.

Materials and methods

Subjects

Forty-two participants (84 scans) were included in this study, comprising 14 patients with schizophrenia, 14 individuals at CHR and 14 matched healthy volunteers. However, two healthy volunteers had to be excluded from the analysis because of excessive head motion that could not be corrected. All patients were antipsychotic-free, corroborated by clean urine drug screens, with eight of them also being antipsychotic-naïve.

To be eligible, CHR individuals had to meet the following criteria: fulfilment of diagnostic criteria for prodromal syndrome as per the Criteria of Prodromal Syndromes (COPS) (Miller et al., 2003) with no current Axis I disorder, as determined with the Structured Clinical Interview for DSM-IV (SCID) (First et al., 2002) and no history of or current treatment with antipsychotic medication. Patients with schizophrenia had to have a diagnosis of schizophrenia, schizoaffective, delusional, or schizophreniform disorder as assessed with the SCID, with no current treatment with antipsychotic medication, and no concurrent Axis I disorder. Healthy volunteers did not meet criteria for any prodromal syndrome, had any history of psychiatric illness or psychoactive drug use, and had no first-degree relative with a major mental disorder. Participants were excluded for any of the following: current diagnosis of substance abuse or positive urine drug screen; pregnancy or currently breastfeeding; clinically significant medical illness; and the presence of metal implants precluding a MRI scan.

The clinical status and severity of symptoms were assessed with the Structured Interview for Psychosis-risk Syndromes (SIPS) and the Scale of Psychosis-risk Symptoms (SOPS) (Miller et al., 2003) (CHR group), the Positive and Negative Syndrome Scale (PANSS) (Kay et al., 1987) (schizophrenia group), the Global Assessment of Functioning Scale (GAF), the Zung Self-rating Anxiety Scale (SAS), the Social Interaction Anxiety Scale (SIAS), the Recent Life Events questionnaire (RLE), and the Trier Inventory of the Assessment of Chronic Stress (TICS). Assessments are referenced in the online Supplementary material. This study was approved by the Research Ethics Board at the Centre for Addiction and Mental Health in accordance with the Declaration of Helsinki. All subjects provided written informed consent after being informed of all study procedures.

Montreal Imaging Stress Task

Psychological stress was induced using the Montreal Imaging Stress Task (referred to as ‘stress’) (described in detail in Dedovic et al., 2005), which has been used and validated in various functional MRI and PET studies (Lederbogen et al., 2011; Mizrahi et al., 2012). In brief, subjects perform mental arithmetic presented on a computer screen, which also displays information about the total number of errors, expected average number of errors, time spent on the current problem, and performance feedback for each problem (correct, incorrect, time out). All subjects completed six blocks of arithmetic, each approximately 6 min in length, while lying in the scanner. The time constraint was adjusted individually to be slightly beyond each subject’s abilities by adjusting each block dependent on the performance in the previous block. Because of this manipulation of the difficulty level, the average performance was set at 20–30% correct answers. Additionally, participants were given negative verbal feedback between each block, telling them that they need to improve their performance in order to reach minimum performance requirements. On a separate day before the stress session, participants were scanned while performing a Sensory Motor Control Task (referred as ‘control’), using similar arithmetic but without any time constraints or negative verbal feedback. The control scan was always performed first, to avoid any residual effects of the stress task.

In all experiments, the control or stress task was started about 6–8 min before tracer injection (see Fig. 1 for task overview). The control task was also administered as a practice trial on a separate day before the PET experiments, to reduce novelty effects.

Figure 1.

Overview of study procedures including timeline of scans (A) and the explanation of control (B) and stress task (C)..

After each PET scan session, participants’ subjective perception of stress was assessed by a short version of the State Anxiety Questionnaire (SAQ) (Spielberger et al., 1977). Further, patients’ psychotic (schizophrenia group) or attenuated psychotic (CHR group) symptoms were evaluated before and after each scan session using the PANSS or SOPS positive subscale, respectively.

Physiological measures

Saliva samples were collected every 15 min throughout the PET scanning session (six samples total) to evaluate the physiological response to the stress paradigm, starting 15 min before tracer injection and 9 min before the arithmetic task started. Saliva-derived cortisol was analysed using a time-resolved fluorescence immunoassay and the normalized area under the curve (AUC_I) (g/dl/min) was calculated for each subject and each PET scan session as described elsewhere (Pruessner et al., 2003). Normalization to time point 1 was chosen due to differences in the scan time (between 9.00 am and 4.30 pm) as cortisol levels fluctuate over the course of the day (Castro et al., 2000). Cortisol data from six participants (one CHR and five schizophrenia) were not available for analysis. Change in AUC_I (ΔAUC_I) between control and stress task was defined as: ΔAUC_I = AUC_{I Stress} − AUC_{I Control}.

Image acquisition and reconstruction

Every subject underwent an MRI scan to acquire a proton density-weighted image, used for delineation of individual regions of interest after co-registering with the PET image. All PET scans were performed for 90 min following intravenous bolus injection of ∼9–11 mCi ¹¹C-FLB457 using a high resolution PET-CT scanner, Siemens-Biograph HiRez XVI (Siemens Molecular Imaging). Images were reconstructed using a 2D filtered back projection algorithm with a ramp filter at Nyquist cut-off frequency. Details of the image acquisition are summarized in the Supplementary material.

PET data analyses

Time–activity curves were extracted for the dlPFC and mPFC including both hemispheres (for detailed region of interest location, see Supplementary Fig. 5), and cerebellum using our validated in-house imaging software ROMI (Rusjan et al., 2006). All regions of interest were delineated using proton density-weighted image for each participant (Mizrahi et al., 2012). A quantitative estimate of binding was obtained from each time–activity curve with the Simplified Reference Tissue Model (SRTM) (Lammertsma and Hume, 1996) using the in-house software fMOD. The SRTM uses a within-brain reference region (cerebellar cortex in this case) instead of the arterial input function and provides an estimate of the binding potential (BP_ND) of the radiotracer, which is proportional to the more fundamental parameters of receptor number (B_max) and affinity (K_d) [BP_ND$\propto$ B_max/K_d]. It is a validated method and commonly used with ¹¹C-FLB457 (Olsson et al., 1999; Ito et al., 2001; Narendran et al., 2009). Although few studies suggest small specific binding of ¹¹C-FLB457 in cerebellum (Vandehey et al., 2010; Narendran et al., 2011a), no change in cerebellar distribution volume was observed following challenges with amphetamine and methylphenidate (Montgomery et al., 2007; Narendran et al., 2009). Previous studies with ¹¹C-FLB457 have successfully used SRTM with cerebellum as reference region (Mizrahi et al., 2007; Ko et al., 2009; MacDonald et al., 2009) and a recent study showed that SRTM is a valid modelling approach to measure the percentage change in BP_ND ($\Delta B{P}_{ND}\hspace{1em}=\hspace{1em}1-B{P}_{ND}^{Stress}/B{P}_{ND}^{Control}$) with ¹¹C-FLB457 (Sandiego et al., 2015). Right and left regions of interest were pooled together to create a single time–activity curve used to derive BP_ND. As quantifying ¹¹C-FLB457 is challenging, in part due to potential mass effects, a novel correction was applied in the current study (Gallezot et al., 2017). The correction takes competition between radioligand and dopamine in the stress condition into account while assuming negligible levels of occupied receptors in the control condition. These assumptions are supported experimentally as changes in cortical ¹¹C-FLB457 binding has not been observed in a dopamine depletion study (Frankle et al., 2010), while ∼1000% increase of dopamine has been measured with microdialysis following 0.3 mg/kg amphetamine in non-human primates (Narendran et al., 2014). The corrected change in BP_ND was calculated as (Gallezot et al., 2017):

$$\Delta B{P}_{ND}^c\hspace{1em}=\hspace{1em}\frac{\Delta B{P}_{ND}\left(1+\left({\mu }^{Stress}/E{D}_{50}\right)\right)+\left(\left({\mu }^{Control}-{\mu }^{Stress}\right)/E{D}_{50}\right)}{1+\Delta B{P}_{ND}\left({\mu }^{Stress}/E{D}_{50}\right)+\left(\left({\mu }^{Control}-{\mu }^{Stress}\right)/E{D}_{50}\right)}$$ (1)

where $\mu$ is the ratio mass of radioligand injected to body weight and $E{D}_{50}$ the mass injected that would reduce $B{P}_{ND}^{Control}$ by 50%. The $E{D}_{50}$ was estimated as follows: The BP_ND of the dlPFC of the scans from 12 healthy volunteers and two CHR gathered under control condition of this study (age 18 to 38 years, $\mu$ ranging from 0.004 to 0.025 μg/kg) were pooled together with the BP_ND of the dlPFC of eight scans from healthy subjects with very high mass injected gathered under similar control conditions (age 20 to 29 years, $\mu$ ranging from 0.035 to 0.28 μg/kg) from a yet unpublished study, adjusted by age (Narendran et al., 2009) and plotted in Fig. 3. $E{D}_{50}\hspace{1em}$ = 0.124 μg/kg was estimated adjusting $B{P}_{ND}^{}\hspace{1em}=\hspace{1em}B{P}_{ND}^{max}\left(1-\frac{{\mu }_{}^{}}{E{D}_{50}+\mu }\right)$ to the data (Logan et al., 2012).

Figure 3.

Effect of injected mass of ¹¹C-FLB457 on the binding potential (BP_ND) in dlPFC. Graph shows original ¹¹C-FLB457 BP_ND values of the dlPFC (circles) or corrected for age (triangles) plotted against every subject’s injected mass gathered under control condition in this study (open symbols) or control conditions of an unpublished study (filled symbols), and the non-linear fitted curve (black line). The estimated ED₅₀ is 0.124 μg/kg.

Statistical analysis

Group differences in dopamine release were assessed using general linear models with ΔBP_ND value per region of interest (dlPFC or mPFC) as dependent variable and group (CHR, schizophrenia and healthy volunteers) as independent variable. BP_ND of the control scan was added as covariate to explore its effect on dopamine release. All analyses were two-tailed with the conventional α = 0.05. If significantly different, post hoc ANOVAs followed, using Bonferroni correction for multiple comparisons.

Group differences in the relationship between stress-induced dopamine release (ΔBP_ND) and salivary cortisol response (ΔAUC_I) were determined using two separate general linear models to examine the group by ΔAUC_I interaction, with ΔBP_ND value per region of interest (dlPFC or mPFC) as the dependent variable. The main analysis was followed by Pearson’s linear correlations, using Bonferroni correction for multiple comparisons.

Further, associations between ΔBP_ND or ΔAUC_I and scores in behavioural scales (stress and anxiety) were explored using Pearson’s linear correlation analysis. As these analyses were considered exploratory, P-values were not corrected for multiple comparisons. Descriptions of further statistics can be found in the Supplementary material.

We considered results to be significant at P ≤ 0.05 and at trend levels at P ≤ 0.1.

Results

Demographics, clinical characteristics and PET scan parameters

Our analysis comprised 14 CHR, 14 patients with schizophrenia, and 12 matched healthy volunteers (80 PET scans in total). Details of demographics, clinical characteristics and scan parameters are summarized in Table 1. There were no differences between groups in sex, but a difference present in age, with schizophrenia patients being older than CHR subjects (Bonferroni-corrected P = 0.016). As age affects the BP_ND of ¹¹C-FLB457 (Narendran et al., 2009) the ΔBP_ND was corrected for age (see ‘Materials and methods’ section and Fig. 3).

Table 1.

Participants’ demographics, clinical characteristics and radioligand injection parameters in a PET study of dopamine release in CHR and schizophrenia

	Healthy volunteers n = 12	CHR n = 14	Schizophrenia n = 14	Comparisons
Demographics
Gender, male/female	7/5	6/8	8/6	χ2 = 0.81, P = 0.67
Age, years (SD)	26.00 (6.49)	22.07 (3.38)	28.29 (6.09)b	F(2,37) = 4.66, P = 0.016
Clinical characteristics
SOPS, mean (SD)
Positive	–	10.71 (3.45)	−	−
Negative	–	9.43 (6.22)	−	−
Disorganized	–	3.86 (1.75)	−	−
General	–	6.71 (3.67)	−	−
PANSS, mean (SD)
Positive	–	−	17.57 (3.30)	−
Negative	–	−	16.14 (7.16)	−
General	–	−	35.64 (8.37)	−
Medication
Antidepressants	–	3	1	−
Anxiolytics	–	0	1	−
Low dose antipsychotics	–	0	1c	−
PET measures (11C-FLB457)
Amount injected, mCi (SD)
Control task	9.83 (0.78)	9.92 (0.58)	10.18 (0.45)	F(2,37) = 1.21, P = 0.31
Stress task	10.28 (0.52)	10.06 (0.58)	10.11 (0.80)	F(2,37) = 0.40, P = 0.68
Specific activity, mCi/µmol (SD)
Control task	3536.85 (1172.20)	4011.15 (1977.21)	2738.32 (1107.81)	F(2,37) = 2.61, P = 0.087
Stress task	3934.05 (1659.07)	3264.46 (1642.78)	3503.29 (1599.33)	F(2,37) = 0.55, P = 0.58
Mass injected, µg (SD)
Control task	1.12 (0.29)	1.11 (0.47)	1.57 (0.55)a,b	F(2,37) = 4.59, P = 0.017
Stress task	1.15 (0.54)	1.42 (0.65)	1.30 (0.62)	F(2,37) = 0.65, P = 0.53

^aSignificantly different to healthy volunteers (P ≤ 0.05).

^bSignificantly different to CHR (P ≤ 0.05).

^cQuetiapine (100 mg) taken only after the PET scan session.

PANSS = Positive and Negative Syndrome Scale; SD = standard deviation; SOPS = Scale of Psychosis-risk Symptoms.

Control and stress scans were performed on average 9.36 ± 10.22 days apart. All subjects performed the tasks during the scans successfully. There was no significant group difference in any of the PET scan parameters except for higher injected mass in patients with schizophrenia only on the control PET scan.

Furthermore, there was no difference in (attenuated) psychotic symptoms measured before the control and stress scan (Supplementary Fig. 4) or medication status between scan sessions neither in CHR nor schizophrenia. Details on current medication status per participant can be found in Table 1.

Scan paradigm effects

As expected, the SAQ revealed that all subjects were less calm, satisfied, relaxed and pleasant, but more tense, strained, upset and confused following the stress task than following the control task (Fig. 2A; all P < 0.0006), suggesting that the stress paradigm was effective. Total SAQ scores (Fig. 2B; positive items reversed scored) were significantly elevated in all groups following the stress as compared with the control task [effect of task: F(1,37) = 155.30, P < 0.0001, Bonferroni-corrected P < 0.0001 for all groups]. Furthermore, a group difference between SAQ scores was observed [effect of group: F(2,37) = 7.25, P = 0.0022; for post hoc results see Fig. 2 legend] with no interaction between task and group [F(2,37) = 1.47, P = 0.24].

Figure 2.

Subjective stress response following the control and stress task in healthy volunteers, CHR and schizophrenia. Stress response was assessed with the state anxiety questionnaire in individual categories for all subjects (A) and as total scores per group (B). *P ≤ 0.05 (post hoc, after Bonferroni correction).

Patients with schizophrenia showed an increase in psychotic-like experiences following the stress task (PANSS positive subscore; t = 2.60, df = 13, P = 0.022; Supplementary Fig. 1). No significant increase in the SOPS positive subscore was found in CHR (t = 1.48, df = 12, P = 0.17).

All subjects performed significantly worse in the stress task [number of errors: 34.89 ± 11.61 (healthy volunteers), 36.30 ± 9.81 (CHR) and 35.39 ± 9.99 (schizophrenia)] than in the control task [number of errors: 4.54 ± 2.13 (healthy volunteers), 5.20 ± 3.35 (CHR) and 5.20 ± 3.56 (schizophrenia); effect of task: F(1,37) = 429.56, P < 0.0001], showing that the stress task was able to adapt to the level of performance of each person and produce a tailored programmed failure within each group.

Stress-induced dopamine response in prefrontal cortex across the schizophrenia spectrum

BP_ND at control conditions was significantly different between groups in dlPFC [F(2,37) = 6.97, P = 0.0027] and mPFC [F(2,37) = 7.98, P = 0.0013], with patients with schizophrenia exhibiting lower BP_ND compared to CHR (dlPFC: Bonferroni-corrected P = 0.0019; mPFC: Bonferroni-corrected P = 0.00089) and marginally lower compared to healthy volunteers (dlPFC: Bonferroni-corrected P = 0.17; mPFC: Bonferroni-corrected P = 0.16).

¹¹C-FLB457 ΔBP_ND was not different among groups in any of the PFC regions of interest investigated [dlPFC: F(2,37) = 1.07, P = 0.35; mPFC: F(2,37) = 0.54, P = 0.59]. Even if BP_ND in control scan had a significant effect in the model [dlPFC: F(1,36) = 5.48; P = 0.025; mPFC: F(1,36) = 5.57; P = 0.024], including it as a covariate in the analysis did not change the results [dlPFC: F(2,36) = 0.82; P = 0.45; mPFC: F(2,36) = 1.06; P = 0.36]. It is worth mentioning that ΔBP_ND did not differ between groups either when using the conventional calculation (Sandiego et al., 2015) (without applying any correction for injected mass ¹¹C-FLB457).

Stress-induced dopamine response in the prefrontal cortex and salivary cortisol across the schizophrenia spectrum

No differences in salivary levels of cortisol between groups were found, neither in AUC_I during the stress scan [F(2,33) = 0.51, P = 0.60] nor in ΔAUC_I [F(2,33) = 1.91, P = 0.17]. However, the relationship between ΔAUC_I and ¹¹C-FLB457 ΔBP_ND differed significantly between groups [omnibus test in dlPFC: F(5,28) = 5.17, P = 0.0017; mPFC F(5,28) = 3.83, P = 0.0091; interaction between group and ΔAUC_I in dlPFC: F(2,28) = 3.22, P = 0.055; mPFC: F(2,28) = 3.91, P = 0.032]. The effect on ΔAUC_I on ¹¹C-FLB457 ΔBP_ND was significant in healthy volunteers (dlPFC: slope = 0.78, t = 3.60, Bonferroni corrected P = 0.0036; mPFC: slope = 0.83, t = 3.05, Bonferroni corrected P = 0.015) and partially in CHR (dlPFC: slope = 0.74, t = 2.91, Bonferroni corrected P = 0.021; mPFC: slope = 0.68, t = 2.13, Bonferroni corrected P = 0.13), but not in schizophrenia (dlPFC: slope = 0.030, t = 0.13, Bonferroni corrected P = 1.00; mPFC: slope = −0.24, t = −0.81, Bonferroni corrected P = 1.00) (Fig. 4), suggesting a direct relationship between dopamine release and salivary cortisol response due to the stress challenge in healthy volunteers and CHR, but not in schizophrenia.

Figure 4.

Associations between ΔBP_ND and ΔAUC_I in healthy volunteers, CHR and schizophrenia. Lines represent the best linear model fit of the data per group (healthy volunteers: dashed line; CHR = grey line; schizophrenia = black line). The correlations were significant in healthy volunteers (A: r = 0.72, P = 0.026; B: r = 0.76, P = 0.014) and CHR (A: r = 0.76, P = 0.0075; B: r = 0.72, P = 0.018), but not in schizophrenia (A: r = 0.46, P = 1.00; B: r = 0.19, P = 1.00). P-values were Bonferroni corrected for multiple comparisons. AUC_I = area under the curve; BP_ND = binding potential.

Associations of stress-induced dopamine and salivary cortisol response with stress/anxiety across the schizophrenia spectrum

CHR and schizophrenia subjects reported a stronger impact of stressful life events than healthy volunteers [total RLE score: F(2,37) = 6.08, P = 0.0052; CHR: Bonferroni-corrected P = 0.068, schizophrenia: Bonferroni-corrected P = 0.0045] with a higher number of stressful life events [F(2,37) = 6.24, P = 0.0046; CHR: Bonferroni-corrected P = 0.058, schizophrenia: Bonferroni-corrected P = 0.0041]. Furthermore, CHR and schizophrenia groups reported more chronic stress [TICS: F(2,37) = 13.43, P < 0.0001; CHR: Bonferroni-corrected P < 0.0001, schizophrenia: Bonferroni-corrected P = 0.00073], and displayed higher anxiety [SAS: F(2,37) = 15.22, P < 0.0001; CHR: Bonferroni-corrected P < 0.0001, schizophrenia: Bonferroni-corrected P = 0.00040; SIAS: F(2,37) = 9.65, P = 0.00042, CHR: Bonferroni-corrected P = 0.00060, schizophrenia: Bonferroni-corrected P = 0.0038] than healthy volunteers.

In CHR, but not in healthy volunteers or schizophrenia, chronic stress (total TICS score) was significantly negatively associated with ΔBP_ND specifically in mPFC (Fig. 5A; r = −0.62, P = 0.018). Furthermore, in CHR the number of stressful life events (RLE) was negatively associated with the ΔBP_ND in mPFC (Fig. 5B; r = −0.63, P = 0.015) as well as with cortisol levels during the stress scan (Fig. 5C; AUC_{I stress}; r = −0.72, P = 0.0060) and marginally with ΔAUC_I (r = −0.48, P = 0.093). No associations at all were observed in healthy volunteers and schizophrenia or in CHR ΔBP_ND in dlPFC (P > 0.05). In CHR, but not in healthy volunteers or schizophrenia, anxiety was associated with ΔBP_ND and ΔAUC_I. In detail, SAS score was negatively associated with ΔBP_ND in dlPFC (Fig. 5D; r = −0.54, P = 0.045) and mPFC (Fig. 5E; r = −0.78, P = 0.0011) as well as with ΔAUC_I (Fig. 5F; r = −0.56, P = 0.047), and SIAS score with ΔBP_ND in mPFC (r = −0.57, P = 0.034).

Figure 5.

Associations between the Trier inventory of the assessment of chronic stress (TICS), number of stressful life events or Zung SAS score and ΔBP_ND or ΔAUC_I in CHR (n = 13–14). The line represents the best linear model fit of the data. All correlations were significant (A: r = −0.62, P = 0.018; B: r = −0.63, P = 0.015; C: r = −0.72, P = 0.0060; D: r = −0.54, P = 0.045; E: r = −0.78, P = 0.0011; F: r = −0.56, P = 0.047). P-values were not corrected for multiple comparisons. AUC_I = area under the curve; BP_ND = binding potential.

Discussion

Our results suggest that PFC dopamine release in response to stress in healthy volunteers and CHR, but not in schizophrenia, is associated with salivary cortisol response, implying abnormal PFC stress regulation in schizophrenia. In addition, individuals at CHR with higher distress and anxiety had lower dopamine release in mPFC and salivary cortisol response following the stress challenge, associations that were absent in the schizophrenia group. A similar association between stress-induced dopamine release and cortisol response was reported in striatum for CHR (Mizrahi et al., 2014) and healthy volunteers (Pruessner et al., 2004; Mizrahi et al., 2013). This suggests an overall disrupted stress response in schizophrenia.

So far, only two studies investigated dopamine release in PFC in schizophrenia. Slifstein et al. (2015) reported lower dopamine release in dlPFC in schizophrenia following an amphetamine challenge, with no significant changes in mPFC. Similar to our study, Hernaus et al. (2015) found no change in dopamine release in mPFC in response to a similar psychosocial stress challenge in individuals with a psychotic disorder (brief psychotic episode, schizophrenia or psychosis not otherwise specified) using ¹⁸F-fallypride and a one-scan paradigm. The present study is also consistent with previous investigations reporting no association (or rather a lack of association) between dopamine release in PFC and clinical symptoms in schizophrenia (Hernaus et al., 2015; Slifstein et al., 2015). Differences between studies include the challenge conditions (amphetamine versus psychosocial stress), tracers (¹⁸F-fallypride versus ¹¹C-FLB457) and clinical populations.

Stress-induced extrastriatal dopamine response has not been studied in CHR, but Lataster et al. (2014) reported comparable dopamine release in first-degree relatives of patients with schizophrenia and control subjects in ventromedial PFC.

A model proposed by Grace and others can explain our results. Acute stress induces an increased population activity of the ventral tegmental area (VTA) dopamine neurons (Valenti et al., 2011) leading to an increased striatal dopamine release (Rougé-Pont et al., 1993). One major regulator of the mesolimbic dopaminergic system is the mPFC, which makes direct and indirect connections to the hippocampus and amygdala (Belujon and Grace, 2015), as well as directly to the VTA (Sesack and Carr, 2002; Gabbott et al., 2005). Inhibition of the infralimbic subdivision of the mPFC was sufficient to increase the VTA dopamine neuron activity and this effect was modulated by the hippocampus (ventral subiculum) (Patton et al., 2013). Chronic stress, however, has been shown to lead to loss in dendritic material in the mPFC (Holmes and Wellman, 2009) and hippocampus (McEwen et al., 2016). This suggests that chronic stress weakens the structures that provide negative feedback for the stress response (Arnsten, 2009). This is also in line with structural changes reported in schizophrenia (Glausier and Lewis, 2013; Haijma et al., 2013). Recently Gomes and Grace (2016) observed that a mPFC lesion with ibotenic acid combined with stress during adolescence led to a long-lasting increase of VTA dopamine neuron activity accompanied by higher striatal dopamine release (measured by increased amphetamine-induced locomotor activity) and anxiety. Interestingly, we could not statistically observe an overall decreased PFC dopamine release due to the stress challenge in schizophrenia, even though the patients reported increased chronic stress and high number of past stressful events. As there is strong evidence that hippocampal hyperactivity underlies the dopamine hyperfunction in striatal regions (Lodge and Grace, 2011), it is possible that a deficient regulation of the hippocampus, rather than PFC, leads to the stress-induced increased dopamine release seen in striatal areas in schizophrenia and CHR (Mizrahi et al., 2012) and in substantia nigra only in schizophrenia (Tseng et al., 2017). Another explanation for the lack of significant difference in stress-induced PFC dopamine release among groups in our study might be a potential compensatory mechanism by recruiting more cortical dopamine, since the PFC has direct and indirect connections to the dopamine cell bodies in substantia nigra and VTA, in order to regulate its dopamine output (Arnsten, 2009). Chronic stress further compromises the plasticity in the hippocampus–PFC pathway (Rocher et al., 2004; Cerqueira et al., 2007; Garcia et al., 2008) and previous data suggest that the hippocampus-PFC pathway is compromised in patients with schizophrenia (Godsil et al., 2013). This could explain why we observed dissociation between endocrine and PFC dopamine-stress response in schizophrenia but not in CHR (or healthy volunteers). Interestingly, preclinical data using the neonatal ventral hippocampal lesion model, a validated schizophrenia model (Tseng et al., 2009), support our findings. While acute stress increased the nucleus accumbens dopamine release stronger in neonatal ventral hippocampal lesioned rats than sham-lesioned rats, there was no difference in stress-induced dopamine increase in frontal cortex between both groups. Furthermore, although the corticosterone response to the acute stressor was associated with frontal cortex dopamine release in sham-lesioned rats, this association was absent in neonatal ventral hippocampal lesion rats (Chrapusta et al., 2003), similar to the present study. We acknowledge that the present study is only a starting point in understanding how the complex dopamine-stress regulation is compromised in schizophrenia and its putative prodrome.

Accurate quantification of ¹¹C-FLB457 is challenging as the BP_ND and ΔBP_ND values are rather small [i.e. smaller than those obtained using ¹¹C-(+)-PHNO in striatal regions]. Although some studies argued against the use of the SRTM for quantification of ¹¹C-FLB457 binding in view of the presence of specific binding in cerebellum and change in cerebellum distribution volume (V_T) by the D₂ partial agonist aripiprazole (Narendran et al., 2011a), others supported its suitability with cerebellum as reference tissue (Olsson et al., 1999; Ito et al., 2001; Olsson and Farde, 2001). Cerebellum V_T and V_T/fp were not observed to change pre- and post-amphetamine challenge, supporting the use of cerebellum as reference tissue in challenge-based experiments only (Sandiego et al., 2015). Moreover, SRTM was more reliable than the two-tissue compartment model in detecting ΔBP_ND following amphetamine challenge and had lower relative standard error. Additionally, while SRTM may lead to underestimation of BP_ND compared to arterial input-based models (Innis et al., 2007), this underestimation applies to both scan sessions, control and stress, such that the potential bias cancels out when calculating the ΔBP_ND.

The current study has limitations, many of which are inherent to neurochemical PET studies, particularly when investigating cortical dopamine. First, the resolution of the PET scanner does not allow differentiation of histological subdivisions of mPFC and dlPFC (i.e. ventromedial and dorsomedial PFC) and nearby structures. Second, the mass of ¹¹C-FLB457 may not be at tracer dose (Narendran et al., 2011b), and hence we present our results using a novel approach to account for this issue. Third, the potential effect of specific binding of ¹¹C-FLB457 in cerebellum may not be negligible, so we compared the cerebellar tracer uptake between scans, and showed nearly complete overlap in tracer uptake between both scans (Supplementary Fig. 2). Fourth, since our control condition may be expected to recruit dopamine activity (Egerton et al., 2009), it does not permit estimation of a true baseline D_2/3 receptor availability but serves as an excellent control for the cognitive part of the stress protocol. Fifth, although our sample size provides sufficient power to detect a group effect in stress-induced dopamine release (n = 40) and its interaction with salivary cortisol response, the number of participants within each diagnostic group is small and we cannot correct for the number of exploratory correlations with questionnaires. This, however, does not change our general conclusion. Sixth, although all patients were antipsychotic-free, only eight were antipsychotic-naïve. Exploratory comparisons of dopamine release in antipsychotic-free and -naïve patients revealed no significant difference (Supplementary Fig. 3). However, since the sample size is small, an effect of past exposure to antipsychotics cannot be completely ruled out. Overall, while we acknowledge the limitations of both task and radioligand, these would not have been possible to overcome as (i) the stress task we used is the only validated one in PET imaging studies; and (ii) arterial sampling was impossible as all participants were doing the task with their hands while lying in the scanner. Thus, to date, there is no better methodology available to examine PFC dopamine response to a stress challenge in humans.

Conclusion

In summary, our results suggest that PFC dopamine release in response to stress in healthy volunteers and CHR, but not in schizophrenia, was associated with salivary cortisol response. These findings provide first direct evidence of a disrupted cortical dopamine-stress regulation in schizophrenia.

Glossary

Abbreviations

CHR

clinical high risk for psychosis

dl/mPFC

dorsolateral/medial prefrontal cortex

SRTM

Simplified Reference Tissue Model