Clinical correlation but no elevation of striatal dopamine synthesis capacity in two independent cohorts of medication-free individuals with schizophrenia

Abstract

Dysregulation of dopamine systems has been considered a foundational driver of pathophysiological processes in schizophrenia, an illness characterized by diverse domains of symptomatology. Prior work observing elevated presynaptic dopamine synthesis capacity in some patient groups has not always identified consistent symptom correlates, and studies of affected individuals in medication-free states have been challenging to obtain. Here we report on two separate cohorts of individuals with schizophrenia spectrum illness who underwent blinded medication withdrawal and medication-free neuroimaging with [¹⁸F]-FDOPA PET to assess striatal dopamine synthesis capacity. Consistently in both cohorts, we found no significant differences between patient and matched, healthy comparison groups; however, we did identify and replicate robust inverse relationships between negative symptom severity and tracer-specific uptake widely throughout the striatum: [¹⁸F]-FDOPA specific uptake was lower in patients with a greater preponderance of negative symptoms. Complementary voxel-wise and region of interest analyses, both with and without partial volume correction, yielded consistent results. These data suggest that for some individuals, striatal hyperdopaminergia may not be a defining or enduring feature of primary psychotic illness. However, clinical differences across individuals may be significantly linked to variability in striatal dopaminergic tone. These findings call for further experimentation aimed at parsing the heterogeneity of dopaminergic systems function in schizophrenia.

Introduction

Dysfunction of dopaminergic systems has been central to hypotheses of schizophrenia spectrum illness pathophysiology for decades [1]. Despite possessing a variety of receptor binding profiles, all mainstay medication treatments in schizophrenia currently share D₂ dopamine receptors as common targets [2], and psychotomimetic properties of pharmacological agents that boost synaptic dopamine concentrations have long been appreciated [3]. Instrumental to the modern understanding of dopaminergic abnormalities in schizophrenia, an accumulated body of in vivo, human molecular neuroimaging experiments has identified illness-associated presynaptic dopaminergic dysregulation in the basal ganglia. This work includes replicated reports of aberrantly amplified dopamine release from presynaptic striatal terminals, as inferred from amphetamine-driven D₂ dopamine receptor radioligand displacement studies [4–10], but also of excessive striatal presynaptic dopamine synthesis capacity, as measured by studies using radiolabeled L-DOPA analogues such as [¹⁸F]-FDOPA [11–19].

[¹⁸F]-FDOPA positron emission tomography (PET) permits in vivo measurement of presynaptic dopamine synthesis capacity by quantifying radiotracer-specific uptake by cells containing DOPA decarboxylase (DDC), which catalyzes its conversion to [¹⁸F]-fluorodopamine. This uptake is particularly robust in the dopamine-synapse-rich striatum, where neural relays responsible for cognitive, motivational, and motor functions are tightly regulated. Delineation of neurochemical pathways critical for the devastating and often difficult to treat psychopathological correlates of primary psychoses promises new routes to much needed biomarkers, therapeutic targets, and potential treatment personalization.

However, just as the clinical psychopathology of schizophrenia is characterized by marked heterogeneity, with affected individuals exhibiting variable patterns of negative, positive, affective, cognitive, and disorganized symptoms, so too do in vivo experimental neuroimaging assays of dopaminergic function reveal substantial variability in schizophrenia. The body of PET literature examining striatal presynaptic dopamine synthesis capacity in psychosis has identified group mean elevations in some [11–19], but not all [20–26] cohorts, and the wide distribution of values reported in patient groups includes many affected individuals whose results fall well within the normal range, raising questions about the nature and role of dopamine dysregulation in such individuals’ clinical presentations.

Accordingly, some of these studies have also looked for relationships between symptom ratings, generally measured with the Positive and Negative Syndrome Scale [27], and striatal-specific uptake; yet, many have not reported any significant relationships [12, 17, 19, 20, 22, 24], possibly due to modest sample sizes, medication treatment, or other sample characteristics. Investigations that have reported such relationships have yielded several notable findings. For instance, one study found positive relationships between overall Positive and Negative Syndrome Scale (PANSS) ratings and striatal presynaptic dopamine synthesis capacity in a cohort with prodromal symptoms considered at-risk for schizophrenia, but not in a schizophrenia comparison group [18]. Another study of ten individuals with schizophrenia found an inverse association between depressive symptoms and striatal [¹⁸F]-FDOPA specific uptake [14]. Two other studies identified a positive relationship between uptake and positive symptoms: the first was in a pooled group of both bipolar disorder and schizophrenia cohorts, and the second reported a weakened finding when demographic variables were included in the statistical model [25, 28]. Another recent study in first episode psychosis identified positive relationships across all tested symptom domains using a novel parcellation scheme [26]. Collectively, this diverse literature raises the possibility that in schizophrenia, the degree to which dopamine synthesis capacity may be affected might underlie interindividual differences in certain symptom domains.

The present work sought to provide further clarity to the variability in this literature through examination of striatal presynaptic dopamine synthesis capacity and clinical symptoms assessments in two separate cohorts of individuals with schizophrenia spectrum illness – a discovery cohort and a replication cohort – studied and closely monitored as inpatients while in an antipsychotic medication-free state. Additionally, a large group of healthy individuals with no psychiatric disorders provided a robust frame of reference with which to characterize dopamine function and its relationship to symptoms in the patient groups.

Materials and Methods

Participants

Two separate cohorts of individuals with schizophrenia or other schizophrenia spectrum illness – an initial discovery cohort (N=25) and a follow-up replication cohort (N=12), each assessed with a different PET camera – and two respectively matched groups of comparison healthy individuals (N=150 and N=12) were studied at the U.S. National Institutes of Health (NIH) Clinical Center (Table 1). All participants provided informed consent as approved by the NIH Combined Neuroscience Institutional Review Board and NIH Radiation Safety Committee. Complete medical and psychiatric evaluations were performed for patient cohorts and confirmed during extended inpatient stays on a dedicated schizophrenia ward, and control group participants were evaluated during outpatient visits. All screening and research procedures were performed at the NIH Clinical Center in Bethesda, MD, USA, with inpatient care, evaluation, and observation provided by the NIMH Psychotic Disorders Service on the Inpatient Clinical Research Unit of the NIMH Clinical and Translational Neuroscience Branch. All participants underwent eligibility screening that included history and physical examination, laboratory testing, and structural MRI of the brain to establish the absence of significant confounding medical, neurological, psychiatric or substance use disorder. Schizophrenia spectrum disorder diagnoses were based on DSM-IV criteria and confirmed via both clinician-administered standardized clinical interview (SCID) [29] and ongoing inpatient clinical evaluation.

Table 1:

Demographics.

	Discovery		Replication
	Patient Cohort 1	Control Cohort 1	Patient Cohort 2	Control Cohort 2
N	25	150	12	12
N (%) women	9 (36)	60 (40)	2 (16.7)	3 (25)
Age (years)	28±9	31±9	28±7	32±10
Self-reported race (White/Black/Other)	17/5/3	134/8/8	7/4/1	4/8/0
Handedness (R/A/L)	21/1/3	131/8/3*	12/0/0	10/0/2
Diagnosis
Schizophrenia	22		10
Schizoaffective disorder	0		1
Psychosis NOS	3		1
Duration of illness (years)	7±6		7±5
Total PANSS score	68±13		52±16
Positive PANSS subscore	10±4		8±3
Negative PANSS subscore	15±6		12±6
Disorganized PANSS subscore	8±3		5±2
Excited PANSS subscore	6±2		4±1
Depressed PANSS subscore	6±3		6±3
Duration medication-free at time of scan**	32±13		34±13

^*Data not available for 8/150 subjects

^**Ceiling value of 60 days applied to those medication-free for more than two months

Participants completed [¹⁸F]-FDOPA neuroimaging and symptom ratings in a confirmed medication-free state, achieved either through participation in a blinded medication withdrawal protocol (length = 4–6 weeks) [30] or for some individuals who presented to our service already in a medication-free state (two in the discovery cohort [including one medication-naïve individual] and two in the replication cohort), after at least two months free of all medications in the community. Symptoms were assessed by blinded, trained raters with the PANSS [27], and a 5-factor model approach [31] was employed to generate subscores for positive, negative, depressed, excited, and disorganized symptom clusters.

Neuroimaging

Acquisition Procedures

For the discovery cohorts (Patient Cohort 1 and Control Cohort 1), PET scanning was performed using a General Electric Advance 3D PET camera. To limit head motion during data acquisition, an individually molded thermoplastic mask was created and applied to each participant. For the replication cohorts (Patient Cohort 2 and Control Cohort 2), PET scanning was performed using a Siemens ECAT HRRT PET camera equipped with a Polaris optical head tracking system. Data collected on the latter system were acquired in list-mode with motion correction applied. All PET scan sessions regardless of scanner included a transmission scan for attenuation correction. For all subjects, structural magnetic resonance imaging (MRI) at 3T was acquired separately in order to generate a T1-weighted anatomical image to guide PET data preparation as described below.

Caffeine and nicotine were not permitted for four hours before all scanning sessions. Additionally, a fasting state (minimum 6 hours pre-scan) was required in order to avoid competition for transport of tracer to the central nervous system (CNS) via the L-type large neutral amino acid carrier system. Oral administration of 200 mg of carbidopa, a peripheral amino acid decarboxylase inhibitor, was given one hour prior to [¹⁸F]-FDOPA injection (discovery: 15.3±2.0 mCi; replication: 10.5±0.4 mCi) to limit peripheral metabolism of circulating radiotracer and, thus, maximize its availability to the CNS.

[¹⁸F]-FDOPA emission scans were 90 minutes in duration immediately following intravenous radiotracer administration. Recorded activity was binned dynamically in 27 frames.

Data Processing Procedures

For discovery cohorts, image reconstruction was performed using filtered back-projection, and for replication cohorts, using ordered subset expectation maximization. Post injection radioactivity decay, dead time, and scatter were corrected. A registered attenuation correction algorithm adjusted for frame-wise head motion was also used.

Anterior-posterior commissure aligned, intensity normalized T1-weighted structural MRI volumes were segmented in each subject’s native space using Freesurfer (https://surfer.nmr.mgh.harvard.edu/) and AFNI (https://afni.nimh.nih.gov/) software, with manual quality checks and editing to derive a centroinferior gray matter cerebellar reference region that excluded vermis and lateral/superior parasinus regions.

Frame-wise realignment to the 21^st frame to account for motion was conducted using FLIRT software (http://fsl.fmrib.ox.ac.uk/fsl/), with the initial three activity-poor frames adopting the fourth frame’s transformation parameters. Structural MRI volumes and accompanying reference region maps for each subject were coregistered to the mean PET image using SPM software (https://www.fil.ion.ucl.ac.uk/spm/software/), and time-activity curves for the reference region were extracted for later entrance into kinetic modeling as described below. Subsequent MRI-driven spatial warping of PET data to common (MNI) brain space was accomplished with ANTS software (http://stnava.github.io/ANTs/). Striatal voxel-wise modeling was accomplished according to the non-invasive graphical linearization approach [32] using the cerebellar reference region time-activity curve as the input function and calculated with PMOD software (https://www.pmod.com/web/). This process yielded specific uptake rate constant, K_i, values at each voxel within the striatum, representing a measure of DOPA decarboxylase-driven dopamine synthesis activity and subsequent [¹⁸F]-fluorodopamine retention. To obtain a regional summary view of results, post-hoc region of interest data, with and without partial volume correction, were also generated by obtaining mean regional time-activity curves derived from native-space striatal parcellations. For partial volume corrected analyses, correction was applied to each PET frame for each scan using the PETPVC toolbox software (https://github.com/UCL/PETPVC) prior to time-activity curve extraction [33]. Parcellations were achieved with the assistance of Freesurfer software, individually inspected, and conservatively hand-edited (by staff blinded to the results) to ensure accurate labeling. Estimation of the K_i parameter was performed using the same graphical linear modeling approach described above.

Statistical Analyses

Group comparisons of demographic measurements employed standard and general linear model-based methods using R software (https://www.r-project.org/). For frequency data, Chi-squared tests (or, in the case of low cell counts, Fisher’s exact tests) were used. General linear model analyses of K_i were conducted on voxels within the striatum using SPM software (https://www.fil.ion.ucl.ac.uk/spm/software/) with a voxel-wise threshold of p<0.05, false discovery rate (FDR)-corrected for multiple comparisons. Detection of mean differences across groups utilized voxel-wise t-tests. For tests of association between K_i and symptoms, multiple regression models incorporated age and sex as nuisance covariates. Voxel-wise meta-analytic tests used the inverse variance weighted Hedges’ g with bias correction method and were calculated using AFNI software (https://afni.nimh.nih.gov/) with an FDR-corrected threshold of p<0.05. Additional non-parametric permutation-based group comparisons were performed with SnPM13 software (http://nisox.org/Software/SnPM13/; 10,000 permutations conducted per comparison). For post-hoc, region-wise analyses, Shapiro-Wilks and Levene’s tests as well as two-sided student’s t-tests and linear models were completed in R (www.r-project.org) for analyses of deviations of PET data from the normal distribution, group variance differences, group mean differences and tracer specific uptake-negative symptom associations (controlling for age and sex), respectively.

Results

Demographics and Clinical Characteristics

Demographics for discovery and replication cohorts are described in Table 1. Characteristics of the two cohorts were similar, and there were no significant sex or age differences, respectively, between patient and control groups in either discovery (sex: χ² (1, N=175)=0.25, p=0.87; age: t(173)=1.44, p=0.15) or replication datasets (sex: Fisher’s exact test p=1; age: t(22)=1.14, p=0.26). There were no significant differences in injected tracer dose between patient and control groups in either discovery (t(173)=1.91, p=0.056) or replication datasets (t(22)=0.92, p=0.37). Patients had an overall symptom burden in the mild to moderate range based on the PANSS total scores [34] and had a mean duration of illness of seven years at the time of study.

Neuroimaging

Between-Group Comparisons

In the discovery cohorts, voxel-wise t-tests comparing patient and control groups on striatal [¹⁸F]-FDOPA specific uptake (K_i), a measure of presynaptic dopamine synthesis capacity, yielded no significant differences. This result was unchanged when controlling for injected tracer dose. Similarly, replication patient and control cohorts showed no significant group differences in striatal K_i. Post-hoc examination at a less stringent statistical threshold (p<0.005, uncorrected) identified three small (k<20) clusters in the bilateral putamen only in the discovery cohort where control subjects had greater K_i than patient subjects and no voxels in which patient subjects had greater K_i than controls. Additionally, there were no between-group findings in either direction in the replication cohort, even at this more liberal threshold. Furthermore, meta-analytic combination of the two cohort results also failed to yield significant findings. Additionally, no group differences were seen when comparisons were performed using non-parametric tests. No group differences were seen in post-hoc analyses controlling for handedness, race, or season of scan. Finally, in post hoc, region-wise analyses, there were no significant deviations from normality, Levene’s tests for unequal variances were not significant, and no regions showed significant group mean differences for either the discovery or replication cohorts using either standard or partial volume corrected data.

Symptom Severity Effects

In the discovery cohort, voxel-wise regression analyses controlling for age and sex identified a significant inverse relationship between total PANSS score and K_i in clusters in the bilateral striatum, localized largely to putamen locales (left cluster: 331 voxels, t=3.50, p=2.31×10⁻⁴, location (x,y,z)=(−27,−3,4.5); right cluster: 57 voxels, t=3.48, p=1.12×10⁻³, location(x,y,z)=(31.5,−9,3)). When symptom cluster subscores were evaluated, only negative symptom subscores showed significant correspondence with K_i (Figure 1). Greater negative symptom burden was associated with lower K_i robustly and widely throughout the bilateral putamen (Figure 1; Table 2). Importantly, examination of this same relationship in the replication cohort yielded the same inverse correlation between negative symptom subscores and K_i, with localization that encompassed not only bilateral putamen, but also bilateral caudate. In post-hoc analyses removing individuals without a strict schizophrenia diagnosis (3 individuals in the discovery group, and 2 in the replication group), this relationship remained evident in both discovery and replication cohorts, although in the discovery cohort, results met an FDR-corrected voxel-wise threshold of p<0.06 rather than p<0.05. In post-hoc analyses controlling for duration of medication free period (ceiling value of 60 days applied to those medication-free for more than two months), handedness, race, or season, significant inverse relationships between negative symptoms and striatal K_i were still observed for both the discovery and replication cohorts. Post hoc, region-wise analyses were also consistent with these results. In the discovery cohort, negative symptoms showed a significant inverse relationship with tracer specific uptake (K_i) in the bilateral caudate (t(21)=2.45; p=0.023) and putamen (t(21)=3.24; p=0.004). In the replication cohort, negative symptoms were again inversely associated with K_i in the bilateral caudate (t(8)=3.18; p=0.013) and putamen (t(8)=3.63; p=0.0067). These results remained consistent in partial volume corrected data analyses (Supplementary Table 1).

Figure 1: [¹⁸F]-FDOPA Specific Uptake (K_i) and Negative Symptoms.

Statistical T-maps (left upper (Discovery) and middle (Replication) images; p<0.05, FDR-corrected) are displayed overlaid on a grayscale anatomical T1-weighted MRI template in MNI standard space. Color bars indicate T-values. Left lower panel represents an overlap map showing regions where [¹⁸F]-FDOPA specific uptake (K_i), a measure of presynaptic dopamine synthesis and retention, was significantly associated with negative symptoms in either cohort (D=Discovery [cyan], R=Replication [lime]), or in both (O=Overlap [red]). The panels on the right show plots of negative symptom severity versus mean bilateral dorsal putamen K_i (upper: Discovery, lower: Replication).

Table 2: [¹⁸F]-FDOPA Specific Uptake (K_i) and Negative Symptoms.

Multiple regression results are tabulated for the voxel-wise regression of [¹⁸F]-FDOPA specific uptake (K_i) on negative symptom score severity, controlling for sex and age, within Discovery and Replication cohorts. All voxels within tabulated clusters met voxel-wise FDR correction for multiple comparisons (p<0.05), with uncorrected peak p-values reported.

Discovery
	Region	Cluster Size	T-value	P-value	Location (x,y,z)
	Left Putamen	730	4.64	7.10E-05	−24,1.5,9
	Right Putamen	517	3.98	3.44E-04	24,7.5,-1.5
	Left Caudate	61	2.96	3.76E-03	−18,12,10.5
	Right Caudate	5	2.54	9.48E-03	18,10.5,10.5
Replication
	Region	Cluster Size	T-value	P-value	Location
	Right Putamen	1932	6.22	1.27E-04	28.5,1.5,4.5
	Left Putamen	2324	4.91	5.92E-04	−30,4.5,4.5
	Right Caudate	32	2.37	2.23E-02	12,3,13.5

Discussion

Dopamine dysregulation in psychotic illness has been considered a core tenet of schizophrenia pathophysiology and therapeutics, supported in part by [¹⁸F]-FDOPA PET imaging findings in affected individuals and even those at high risk of illness [11–19]. The present data do not identify any significant group differences in striatal [¹⁸F]-FDOPA specific uptake, a measure of presynaptic dopamine synthesis and retention, between individuals living with schizophrenia spectrum illness and healthy comparison participants in either a discovery study or a replication cohort. However, among those with illness, this work delineates an association between tracer-specific uptake and the severity of negative symptom burden, replicated in both groups. These results may provide information to help interpret the variability in findings across the literature and may ultimately guide biomarker development.

The lack of group-mean elevations here, while unlike several previously reported studies [11–19], aligns with prior literature finding that not all cohorts demonstrate [20–26] illness-associated elevations of [¹⁸F]-FDOPA uptake, suggesting the possibility that this type of presynaptic dopaminergic abnormality may not be an enduring signature of schizophrenia for some individuals. Although random measurement error and unaccounted variability due to unassayed factors may obscure a true mean difference, the substantial size of the healthy comparison group in this study, the fact that the same result emerged in two separate patient cohorts studied with two different PET scanners, and the lack of mean differences even at more lenient statistical thresholds strengthens confidence in this negative observation for the current datasets. In fact, the only voxels to appear at these more lenient thresholds were small clusters in the opposite direction, which were absent in the replication cohort and unlikely to represent important signal. Although this study was well-powered to detect effect sizes consistent with estimates from prior literature (the discovery cohort had a sufficient sample to detect an effect size of 0.6 at an alpha of 0.05 and power of 0.8, in a two-tailed regional comparison), we cannot rule out the possibility of two false negative results. It may be that [¹⁸F]-FDOPA-measured presynaptic hyperdopaminergia in schizophrenia spectrum illness is, in fact, a phenotype with much greater clinically-related or other variability than previously appreciated. If so, these negative findings underscore the importance of developing a better knowledge of such variability in this PET phenotype, which has been increasingly employed to understand schizophrenia pathophysiology and which has drawn interest as a potential biomarker [24].

It remains unclear what factors are responsible for this heterogeneity in striatal dopamine synthesis capacity in schizophrenia. In one recent [¹⁸F]-FDOPA report [24], individuals with treatment-resistant schizophrenia taking clozapine showed tracer-specific uptake that was diminished not only relative to a treatment-responsive group, consistent with a prior study [23], but also relative to a group without psychiatric illness – suggesting that treatment or clinical factors may play a potent role in lending systemic variability to presynaptic dopaminergic tone in schizophrenia [24]. While several of the prior [¹⁸F]-FDOPA schizophrenia investigations featuring patient cohorts that did not show significant differences from control groups included individuals studied during active antipsychotic treatment [21–24], the present data reflect measurements taken during an established, carefully observed, and well documented antipsychotic-free state in inpatients, indicating that the replicated lack of differences observed here were not attributable to acute neuroleptic effects [20, 25]. By the same token, the two patient cohorts studied in the current work largely represent individuals who were not medication naïve (with a single exception), often having tried many neuroleptics over the course of their illness, and one might speculate that this, or perhaps other clinical qualities, may have some bearing not only on the absence of group main effects but also on variability of the [¹⁸F]-FDOPA measurement more generally.

Thus, hypothesizing that symptom profile may be related to dopamine synthesis capacity in schizophrenia spectrum illness, the present investigation found that greater severity of negative symptoms was associated with lower tracer specific uptake. While a relationship with positive [25, 28] or depressive [14] symptoms was not observed, as reported in some prior work, the relationship with negative symptoms was consistently significant in both discovery and replication cohorts. Negative symptoms are a debilitating dimension of schizophrenia spectrum illness and are generally more often treatment refractory than are positive symptoms [35]. Furthermore, excessive dopamine blockade with antipsychotic agents has been linked clinically to “secondary” negative symptoms [36], and prodopaminergic medications, which have served as important experimental probes of pathobiology in schizophrenia [37], have garnered interest as possible therapeutic agents, although clinical trials conducted to date have not yet shown robust effects [38]. While the theory that the volitional and motivational deficits that are captured in measures of negative symptoms in schizophrenia arise from aberrant salience processing in dopamine-modulated frontostriatal circuitry has found some experimental support [39], the precise nature of dopaminergic contributions to negative symptoms has not been defined. Diminished ventral striatal activation during reward expectation has been frequently observed in schizophrenia and in several studies corresponds to greater negative symptom burden [40]. Here, a direct association between presynaptic dopamine tone and negative symptoms lends further credence to the contention that even when not excessive, presynaptic subcortical dopaminergic drive in schizophrenia may have relevance for psychopathology in schizophrenia. This may be a critical caveat to recently proposed dichotomous neurochemical models of schizophrenia (i.e., hyperdopaminergic treatment-responsive versus eu- or hypo-dopaminergic treatment-resistant) [24, 41, 42] and provides a distinct impetus to better delineate the foundations of presynaptic dopamine systems pathology in primary psychotic illness. Extended [¹⁸F]-FDOPA studies, such as those designed to determine additional kinetic parameters (e.g., washout rates and effective distribution volumes) may help further characterize the current observations [43].

In summary, these data identify further evidence for important heterogeneity in presynaptic dopamine synthesis tone in schizophrenia, not only demonstrating that a lack of hyperdopaminergia can occur even when assayed without the confound of concurrent therapeutic dopaminergic blockade, but also providing more direct, novel neurochemical support for hypotheses that presynaptic dopaminergic mechanisms in the striatum have relevance for negative symptomatology. While it is possible that technical or clinical factors, such as tracer metabolites, reliance on reference region modeling without arterial measurements [43], genetic biases [44], neuroendocrinological status [45], environmental exposures [46] or pyridoxal 5′-phosphate adequacy [47], may have played a role in the current results, the consistency observed in two independent cohorts, the relatively ample study size, and the highly controlled inpatient setting for antipsychotic-free patient participant observations are particular strengths of these data. Future work is needed to better understand the molecular foundations and neurofunctional implications of this heterogeneity, and longitudinal, interventional studies will be important to parse the degree to which identified negative symptom-related dopaminergic signal might represent a meaningful biomarker in schizophrenia.