Kappa opioid receptor availability predicts severity of anhedonia in schizophrenia

Abstract

The kappa opioid receptor (KOR) and its endogenous agonist dynorphin have been implicated in multiple psychiatric conditions including psychotic disorders. We tested the hypotheses that kappa expression is elevated and associated with psychotic symptoms in schizophrenia. We measured kappa expression in unmedicated patients with schizophrenia (7 female, 6 male) and matched controls (7 female, 6 male) with positron emission tomography (PET). We also acquired a measurement of cumulative dopamine activity over the life span in the same subjects using neuromelanin sensitive MRI. We hypothesized that neuromelanin accumulation would be higher in patients than controls and that in patients there would be a positive association between KOR availability and neuromelanin accumulation. Fourteen patients and thirteen controls were enrolled. Whole brain dynamic PET imaging data using the KOR selective tracer [¹⁸F]LY245998 were acquired. Distribution volume (V_T) was measured with region of interest analysis in 14 brain regions. Neuromelanin accumulation in midbrain dopaminergic nuclei was assessed in the same subjects. Positive and negative symptoms were measured by a clinical psychologist. We did not observe group level differences in KOR expression, neuromelanin accumulation or relationships of these to positive symptoms. Unexpectedly, we did observe strong positive associations between KOR expression and symptoms of anhedonia in the patients (Pearson r > 0.7, uncorrected p < 0.01 in 8 cortical brain regions). We also observed moderate associations between KOR expression and neuromelanin levels in patients. In conclusion, we did not observe a relationship between kappa and symptoms of psychosis but the observed relationship to the negative symptom of anhedonia is in line with recent work testing kappa antagonism as a therapy for anhedonia in depression.

Introduction

The kappa opioid receptor (KOR), one of three subtypes of opioid receptors, and its endogenous agonist dynorphin have been implicated in multiple psychiatric conditions including psychotic disorders. Evidence that KOR may be involved in psychotic disorders comes from studies showing that KOR agonists such as cyclazocine and salvinorin-A are psychotomimetic in healthy controls and the effects are blocked by pan-opioid antagonists such as naloxone or naltrexone [reviewed in refs. 1, 2]. Studies have also shown that pan-opioid antagonists have antipsychotic effects in schizophrenia [3–5]. Furthermore, there is evidence from preclinical studies demonstrating various interactions between the KOR-dynorphin system and the dopamine system, with KOR stimulation affecting dopamine release in the striatum and other brain regions involved in reward processing [6, 7]. Dopaminergic dysfunction has been extensively documented in schizophrenia [8–10], and dopamine transmission has been demonstrated to play an integral role in psychosis [10] as well as normal reward processing [11, 12]. We previously observed that hyperdopaminergic states in schizophrenia are mostly restricted to the striatum, particularly the associative striatum [13], while extra-striatal regions show deficits in dopamine release [14], suggesting that DA dysfunction in schizophrenia may be secondary to abnormal modulation by other systems, including the KOR.

Given these lines of evidence, we sought to investigate KOR expression, KOR-dopamine interactions, and their relationship to clinical symptoms in schizophrenia. We used Positron Emission Tomography (PET) imaging with [¹⁸F]LY2459989 (LY245) to measure KOR expression. LY245 is a recently developed antagonist KOR radiotracer with excellent kinetic properties and good test-retest reliability [15, 16]. Compared to the KOR agonist tracer [¹¹C]GR103545 used in many previous imaging studies, LY245 has faster kinetics allowing for better identifiability and as an antagonist, less likelihood of pharmacological reactions. We hypothesized that KOR availability would be higher in patients with schizophrenia compared to demographically matched healthy controls, particularly in the associative striatum. We further investigated whether enhanced expression of KOR relates to psychosis and to a novel measure of dopaminergic activity obtained with Magnetic Resonance Imaging (MRI), Neuromelanin-Sensitive MRI (NM-MRI). Neuromelanin is a metabolite of dopamine that accumulates in the substantia nigra over the lifetime. We have previously shown that NM-MRI is correlated with PET measures of striatal dopamine release and psychosis severity in schizophrenia [17] and could be used as a proxy measure of dopamine neurotransmission in patients with schizophrenia and demographically matched healthy individuals. We hypothesized that neuromelanin-related contrast, measured via NM-MRI, would be higher in patients than controls in a ventral subregion of the substantia nigra previously identified as related to psychosis (psychosis ROI), and that in patients there would be a positive association between KOR availability and NM-MRI CNR. Finally, to assess clinical symptoms in patients, we administered the Positive and Negative Syndrome Scale [PANSS, 18] and the Scales for Assessment of Positive and Negative Symptoms [SAPS-SANS, 19]. We also examined the relationships between KOR or NM-MRI and clinical scores in patients.

Materials and methods

Subjects

The protocol was approved by the Institutional Review Board of Stony Brook University. Fourteen patients with schizophrenia (SCZ) and thirteen healthy controls (HC) were enrolled in the study. SCZ participants met DSM-5 criteria for schizophrenia and were antipsychotic drug-naïve or drug-free for at least 3 weeks to avoid effects of antipsychotic exposure and other psychotropics on our outcome measures (See Table S4 for antipsychotic drug history). Inclusion criteria were age 18–45 yrs, fluency in English, and capacity to provide informed consent. Exclusion criteria were use of medications known to interact with KOR within 6 months prior to participation in the study, positive urine toxicology for illicitly used drugs or cannabis, psychiatric diagnoses as determined by SCID-5 (other than schizophrenia in the SCZ group), any substance use disorder other than nicotine use, history of severe medical or neurological illness including stroke or seizure, metal in the body, pregnancy or breast-feeding in women and recent or anticipated exposure to radiation that would cause the participant to exceed FDA guidelines for radiation exposure to research subjects.

PET methods

Radiochemistry

LY245 was synthesized according to a published method [20] with minor modifications to adapt the synthesis to a fully automated procedure using a GE FXNpro or FX2N synthesis module.

PET imaging

An arterial catheter was placed in participants’ radial artery for input function sampling and a venous catheter placed in the contralateral arm for radiotracer administration. Scans were performed on a Molecular Imaging Electronics Scintron, a Siemens HR+ which was retrofitted with modern digital electronics. Head motion was minimized with a polyurethane head immobilizer system molded around the participant’s head [21]. Following a 10 min transmission scan for attenuation correction, a bolus injection of LY245 (up to 4 mCi in men or 3.5 mCi in women due to sex differences in radiation dosimetry estimation; there were not significant sex differences in actual administered dose, see footnotes to Table 1, cold mass ≤0.79 µg) was administered. Emission data were collected for 120 min in list mode and binned into a sequence of frames of increasing duration (20 s to 10 min). Decay-corrected data were reconstructed by filtered backprojection with corrections for attenuation, scatter, randoms, and deadtime, using manufacturer-supplied software.

Table 1.

Demographics and scan parameters.

PET
	SCZ	HC	p
Age (yrs)	28.14 ± 4.79	29.41 ± 9.88	0.680
Sex	7F/6M	7F/6M
Smokers	1	2c
Drug-free/drug naïve	9 DF/4 DN
Injected activity (mCi)a	3.1 + 0.47	2.81 + 0.65	0.198
Injected cold mass (µg)b	0.52 + 0.19	0.56 + 0.2	0.607
fp	0.021 + 0.002	0.021 + 0.003	0.522
NM
	SCZ	HC	p
Age (yrs)	29.78 ≠ 7.67	28.13 ± 9.11	0.996
Sex	8F/6M	6F/6M
Smokers	1	1d
Drug-free/drug naive	10 DF/4 DN

^aNo significant differences by sex, (Female: 2.89 ± 0.39 mCi, Male: 3.03 ± 0.75 mCi, p = 0.59).

^bNo significant differences by sex, (Female: 0.51 ± 0.20 µg, Male: 0.57 ± 0.18 µg, p = 0.41).

^cOne tobacco smoker and one nicotine product vaper.

^dNicotine product vaper.

Input function

Arterial blood was sampled throughout the scan. Input functions were prepared as previously described [22]. Briefly, arterial plasma was extracted and counted in a gamma counter. A subset of samples was processed by HPLC for radio-labeled metabolite correction. The unmetabolized fraction was fitted to a Hill function and the total plasma activity was multiplied by the Hill function to obtain the measured arterial plasma concentration of LY245. This was fitted to a sum of 3 exponentials, starting from the peak concentration, to form a modeled arterial input function.

Image processing

PET data were processed as previously described [14]. Briefly, T1-weighted anatomical MRI images were acquired for each participant. PET frame data were realigned in SPM12 [23] and coregistered to participants’ T1 weighted MRI. Regions of interest (ROIs) were drawn on the MRI in a coronal view aligned to the AC-PC plane and transferred to the coregistered PET in MEDx software. Time activity curves were derived as the average activity in each frame and each ROI. ROIs included the functional subdivisions of the striatum (associative striatum, sensory-motor striatum, limbic striatum), prefrontal cortex (medial, dorsolateral and orbital) amygdala, insula, anterior cingulate, hippocampus, lateral temporal cortex, parietal cortex, occipital cortex and a midbrain region encompassing the substantia nigra and ventral tegmental area (SN/VTA).

PET data analysis

ROI time activity curves were fitted to 1 and 2 tissue compartment models (1TC, 2TC) using the modeled arterial input function, following correction for fixed 5% blood volume. Goodness of fit criteria were applied to determine which TC model was most parsimonious.

Neuromelanin MRI methods

NM-MRI imaging

Data acquisition methods were based on published methods [17, 24], and additional refinements implemented as in ref. [25]. MRI scans were performed on a 3T Siemens MAGNETOM Prisma scanner using a 64-channel head coil. NM-MR images (11 sequential volumes) were acquired as 20 transverse slices, each 1.5 mm thick, aligned in the AC-PC plane, and placed over the midbrain as in ref. [25], with 0.39 × 0.39 mm² in-plane resolution and FOV = 165 × 220 mm². The NM-MRI pulse sequence consisted of 2D gradient recalled echo (GRE) with magnetization transfer (MT) contrast of the midbrain (2D-GRE-MT, FA = 40°, slice gap = 0 mm, bandwidth = 390 Hz/pixel, TR = 555 ms, TE = 4.11 ms; MT frequency offset = 1200 Hz, MT pulse duration = 10 ms, MT flip angle = 300°).

NM-MRI data analysis

Following visual inspection for motion or other artifacts, data were coregistered to participant’s T1 weighted image. Volumes were realigned with Advanced Normalization Tools (ANTS), normalized to a Montreal Neurological Institute image (MNI Space) and averaged. A contrast to noise ratio (CNR) was derived as the difference in image intensity between each voxel in the substantia nigra and the modal intensity in a reference tissue (crus cerebri), normalized to the reference tissue intensity, The psychosis ROI, as defined in ref. [17], is the conjunction in MNI space of voxels strongly associated with positive symptoms in a prior sample of patients with schizophrenia and those strongly associated with attenuated symptoms of psychosis in a sample of clinically high-risk individuals. The ROI was determined in a separate set of participants from those in the current study and contains 48 voxels (See Fig. S1). On an exploratory bases, we also measured CNR in anatomical subregions in the midbrain dopaminergic nuclei. These included substantia nigra pars compacta (SNc), substantia nigra pars reticulata (SNr), and VTA as defined in ref. [17].

Clinical measures‘

The PANSS, SAPS and SANS were administered to all participants by a clinical psychologist (GP).

Statistics

PET V_T data were analyzed in a linear mixed model framework with group as factor and region of interest as repeated measure. NM-MRI data were tested for group mean differences with a 2 sided, 2 group t-test. Associations between PET or NM-MRI and clinical measures and between PET and NM-MRI were performed with Pearson product moment correlation coefficients. Significance level was set at p = 0.05.

Results

Subjects

Demographic characteristics of the sample are shown in Table 1. Groups were well matched for age and gender. All participants completed PET and NM-MRI. All data was included in analyses except for the following participants. PET data from one SCZ participant were disqualified due to technical issues during data acquisition, however, the NM-MRI CNR of this SCZ participant was included in group comparisons of NM-MRI and measures of NM-MRI association with clinical measures. NM-MRI data from one HC were excluded due to failure of the processing pipeline, most likely due to excessive motion. PET data for this HC participant were included in group comparisons of KOR availability. SAPS and SANS were not administered to one SCZ participant due to scheduling issues. Comparisons between PET or NM-MRI and SANS/SAPS were limited to the other SCZ participants (12 for PET, 13 for NM-MRI).

KOR

The 1TC model was found to be most parsimonious, as 2TC V_T had poor identifiability in 39 of the 364 fits (coefficient of variation >10%) whereas all 1TC fits met identifiability criteria (Summed images, Fig. 1). The two methods were highly correlated for the 2TC fits that were identifiable (r = 0.98 across all 325 fits, r = 0.95 ± 0.02 by region). There was a significant main effect of region (F(13,24) = 113, p < 0.001) but not of group (F(1,24) = 0.03, p = 0.857) or group by region interaction (F(13,24) = 1.63, p = 0.145). Table 2 shows V_T by region with p values from 2 group t-tests for illustrative purposes (no correction for multiple comparisons) and effect sizes (Cohen’s d). There were no detectable group-level differences in plasma free fraction of LY245 (Table 1) and correction for free fraction (V_T/f_p) did not alter the statistical results (data not shown).

Fig. 1. Summed LY245 images averaged across 13 HC subjects in MNI space, showing high binding regions (Insula, Amygdala).

MRI is the MNI single subject T1 weighted image in SPM12. Black lines in the MRI show slice levels. Arbitrary units normalized to 0 to 100%.

Table 2.

LY245 distribution volume.

Region	SCZ VT ± SD	HC VT ± SD	p value	Effect size (d)
AST	4.38 ± 0.59	4.35 ± 0.86	0.928	0.036
SMST	4.39 ± 0.8	4.37 ± 0.71	0.95	0.025
VST	6.63 ± 1.05	6.65 ± 1.34	0.972	–0.014
MFC	4.9 ± 0.66	5.02 ± 0.87	0.704	–0.151
DLPFC	4.23 ± 0.71	4.53 ± 0.73	0.299	–0.416
OFC	4.6 ± 0.55	4.64 ± 0.73	0.875	–0.062
Amygdala	8.41 ± 1.22	8.48 ± 1.34	0.893	–0.053
Insula	6.81 ± 0.88	6.87 ± 1.1	0.894	–0.053
ACC	6.33 ± 0.83	6.47 ± 1.18	0.717	–0.144
Hippocampus	4.51 ± 0.72	4.37 ± 0.81	0.64	0.186
TEM Ctx	4.75 ± 0.6	4.95 ± 0.82	0.485	–0.278
PAR Ctx	4.11 ± 0.48	4.18 ± 0.67	0.741	–0.131
Occ Ctx	3.89 ± 0.48	3.87 ± 0.64	0.922	0.039
Midbrain	3.15 ± 0.37	3.08 ± 0.43	0.658	0.176

NM-MR

We did not observe any significant group-level differences in any of the NM-MRI ROIs, though HC were trend level higher than SCZ in SNr (Table S1).

Relationship of clinical measures to imaging measures

Comparisons to clinical measures were limited to the SCZ sample due to floor effects in HC (most HC had the minimal score on all items). We observed significant associations (uncorrected for multiple comparisons) between cortical KOR V_T and the SANS-A composite score of anhedonia [26] (Table 3). For the PANSS, there were no significant relationships observed between KOR V_T and composite scores in the general, positive, or negative symptom domains on the PANSS. However, in an exploratory item-by-item analysis of the PANSS, we observed significant associations between the negative item N2 (emotional withdrawal), which is a behavioral manifestation of anhedonia, and V_T in most cortical regions (not corrected for multiple comparisons) but not in subcortical regions (Table 3). As to be expected, N2 and SANS-A were strongly related to each other (r = 0.72, p = 0.005) as were their relationships to KOR availability (the correlation between regional r values = 0.89, Table 3), providing converging support for the idea that KOR signaling is related to anhedonia in patients. We did not detect significant associations between NM-MRI CNR and any item in the PANSS or any cumulative subtotal of the PANSS scores. However, we did observe (uncorrected) associations between NM-MRI CNR and SANS-A in SNr (r = 0.624, p = 0.023) and in the The psychosis ROI (r = 0.557, p = 0.048, Table S2).

Table 3.

Correlations between PET and PANSS N2 (emotional withdrawal), SANS-A and The psychosis ROI NM-MRI.

	PET v PANSS N2		PET v SANS -A		PET v The psychosis ROI NM
Region	r	p	r	p	r	p
AST	0.377	0.204	0.319	0.313	0.461	0.113
SMST	0.085	0.783	0.138	0.669	0.399	0.177
VST	0.326	0.277	0.210	0.512	0.470	0.105
MFC	0.860	0.000	0.613	0.034	0.248	0.414
DLPFC	0.805	0.001	0.464	0.128	0.118	0.701
OFC	0.792	0.001	0.480	0.114	0.227	0.456
Amygdala	0.487	0.091	0.250	0.434	0.348	0.244
Insula	0.710	0.007	0.578	0.049	0.504	0.079
ACC	0.766	0.002	0.604	0.038	0.403	0.172
Hippocampus	0.253	0.404	0.083	0.798	0.310	0.303
TEM Ctx	0.821	0.001	0.445	0.147	0.258	0.395
PAR Ctx	0.757	0.003	0.532	0.075	0.502	0.081
OCC Ctx	0.778	0.002	0.673	0.017	0.540	0.057
Midbrain	0.287	0.342	0.322	0.308	0.351	0.239

AST associative striatum, SMST sensory motor striatum, VST ventral striatum, MFC medial frontal cortex, DLPFC dorsolateral prefrontal cortex, OFC orbito-frontal cortex, ACC anterior cingulate cortex, TEM Ctx temporal cortex, PAR Ctx parietal cortex, OCC Ctx occipital cortex

Relationship between KOR and NM-MRI

In SCZ, associations between KOR V_T and the psychosis ROI NM-MRI CNR were in the range 0.4 ≤ r ≤ 0.54 in 7 of the 14 regions; all were trend-level or non-significant (Table 3). Moderate associations were also observed between KOR V_T in striatal subregions and NM-MRI CNR in anatomical divisions of the midbrain (Table S3). To give a general sense of the associations in SCZ between KOR in striatum and NM-MRI in midbrain, correlation between the first principal components of standardized values of KOR V_T in striatal subregions and NM-MRI CNR in midbrain subdivisions was r = 0.43 (not significant).

Discussion

In this initial study, contrary to our hypothesis, we observed no significant differences in KOR availability between SCZ and HC in any of the measured brain regions. We also observed no group-level differences in NM-MRI measures of CNR in the psychosis ROI or anatomic subregions of the midbrain, contrary to our hypothesis, and to prior reports using this methodology [27]. Furthermore, there were no significant associations between KOR availability or NM-MRI CNR with positive symptom domain items on the PANSS in patients, but we did observe a relationship between KOR levels and ratings of anhedonia.

To the best of our knowledge, this is the first study to examine in vivo KOR availability in schizophrenia, and our results suggest there is not an overall dysregulation of receptor levels, though these results do not preclude abnormal receptor function. Additional studies targeting measures of KOR receptor affinity (for the radiotracer or for the endogenous agonist dynorphin), as well as measures of dynorphin levels in cerebral spinal fluid (CSF) [28] or plasma [29], might provide additional insights into KOR function and its possible dysregulation in schizophrenia. While our sample size was small, the small effect sizes (Table 2) suggest these results are not due to lack of power for KOR availability estimates. One possible explanation for the lack of association of KOR V_T with positive symptoms relates to the limited time resolution inherent to PET parameters. Evidence from preclinical data suggests that effects of KOR stimulation on D2 receptor activity are highly dependent on the relative timing and duration of the KOR activation, with opposite effects observed for acute versus chronic KOR stimulation and for sequential versus simultaneous coadministration of DA and KOR stimulation [1]. PET-derived KOR availability measurements do not provide information about these timing issues. Our failure to observe a difference in receptor expression or relationship to positive symptoms does not necessarily preclude a therapeutic effect of KOR antagonists on psychosis, in light of the previous work showing therapeutic potential for KOR antagonism [2]. A larger study is needed to provide more conclusive information in this regard.

On the other hand, the lack of significant association with positive symptoms in the psychosis ROI NM-CNR may be related to limited power due to the small sample size. Other studies have detected significant associtions between PANSS positive symptoms and NM-CNR in the psychosis ROI but with small effect sizes in larger samples. Cassidy et al. [17] and Wengler et al. [30] observed partial correlation coefficients of 0.38 and 0.35, respectively. Power analysis applied to these suggests sample sizes of 49 and 59, respectively, to have power 1–β = 0.8 with significance level α = 0.05.

In contrast to the absence of relationship to psychosis, we did observe a strong positive association between LY245 V_T in many cortical regions and clinical measures related to anhedonia, a negative symptom in patients with schizophrenia. These included the SANS-A composite score of anhedonia-related attributes and the PANSS negative item N2, which characterizes emotional withdrawal (Table 3, Fig. 2). The convergence in findings across the two different assessment tools provides cross validation to this observation. This result, though initially unexpected, is intriguing and in accord with our meta-analysis concluding that negative symptoms may respond to KOR antagonism [2]. It will be important to confirm these KOR-anhedonia relationships in a larger patient cohort, particularly given the lack of therapies for negative symptoms in schizophrenia as well as recent interest in KOR antagonism as a therapeutic strategy for treating anhedonia in individuals with major depressive disorder [31, 32]. Anhedonia, the absence of derived pleasure from experiences that are usually rewarding, is a common feature of many psychiatric conditions, including schizophrenia [33], major depressive disorder [34–36], and substance use disorders [37, 38]. In schizophrenia, anhedonia is associated with functional deficits and poorer outcome [39] and is thought to be a central feature of the disease. Furthermore, deficits in anticipation of reward, or the enhanced value of reward due to anticipation, are frequently reported in schizophrenia, while the hedonic experience of reward is less affected [40]. In the study of Krystal et al. [31] in depression, it was also anticipation of reward that related to KOR antagonism. These studies, considered together, suggest the possibility that KOR may be involved in anticipation of reward and in anticipatory anhedonia by extension [41]. Behavioral and neuroimaging tasks that can disentangle these subcomponents of reward as they relate to anhedonia in schizophrenia and other psychiatric conditions and test these for associations with in vivo KOR imaging data will be an important future direction. We do note that Krystal et al. [42] observed increased BOLD activity in ventral striatum following KOR antagonism, whereas our findings are associations of KOR availability with anhedonia primarily in cortical areas. The degree to which these observations represent different aspects of the same dysregulated circuitry remains to be determined.

Fig. 2. Relationship between KOR availability and clinical measures of anhedonia in SCZ participants.

LY245 V_T (mL*cm^–3, horizontal axes) in Anterior Cingulate (ACC, bottom row) or Orbito-frontal cortex (OFC, top row) and clinical rating scores (vertical axes), SANS-A (left) or PANSS N2 (right). Graphs were similar in other cortical regions, see Table 3. Pearson correlation coefficients are displayed in upper left corners.

We observed moderate associations in SCZ between KOR availability in many brain regions and the psychosis ROI NM-CNR, consistent with 15% to 25% of shared variance between the psychosis ROI NM-CNR and KOR availability in associative striatum, limbic striatum, and several areas of cortex (Table 3). These correlations did not reach statistical significance, possibly due to insufficient power for detecting associations of these magnitudes in this relatively small sample. Still, these observations are again intriguing and should be confirmed in larger samples, to provide definitive tests of the associations. Such associations are anticipated considering that dopamine plays a pivotal role in reward processing [12, 43–46]. Animal models show that KOR agonists reduce dopamine release in the nucleus accumbens, striatum, and medial prefrontal cortex [6, 47, 48] and that experimentally induced stress activates the transcription factor CREB in the nucleus accumbens, which induces increased production of dynorphin, the endogenous agonist of KOR [49, 50]. KOR stimulation by dynorphin reduces firing of the dopaminergic cells in the VTA, reducing dopamine release in mesocorticolimbic projections, including frontal cortex, disrupting this well-established node in reward circuitry [11]. While there was no apparent relationship (r = 0.123) between NM-MRI in the VTA and KOR availability in midbrain, our PET midbrain region for KOR encompasses SN, VTA, and some surrounding structures due to the limited spatial resolution of our PET system. Future studies on higher-resolution PET systems may be more informative on this issue.

In addition to the limited timing resolution of PET described above, limitations of this study include the small sample size and the fact that our PET measures of receptor availability may reflect levels of receptor expression, competition from endogenous dynorphin, or a combination of these. Additionally, the outcome measure V_T is the sum of specific binding to KOR (binding potential, BP_P) and V_ND, the equilibrium nondisplaceable uptake, which is independent of KOR [51]. This means that our measures combines specific and non-specific, or non-target related parameters. To isolate the binding potential measurement we would need either a reference region in which V_T ≈ V_ND or a receptor blocking study to estimate V_ND, for example, by Lassen plot [52]. Martinez et al. blocked binding of the KOR selective radiotracer [¹¹C]GR103545 with naltrexone in human research participants and observed substantial reduction of V_T in all brain regions, consistent with KOR binding throughout the brain [53]. The relative contributions of BP_P and V_ND to V_T could differ between the tracers, such that LY245 distribution volume in the low V_T regions could be more representative of V_ND, but this can only be investigated with blocking studies.

Conclusions

In this study, we did not observe differences in KOR availability between SCZ and HC, nor an association between KOR availability and psychosis in SCZ, contrary to what we predicted based on the links between kappa agonism and propensity for psychosis. However, we did observe an unexpected but strong association between KOR availability and clinical measures of anhedonia in SCZ. Future work should replicate and expand this observation in a larger study. If this finding proves to be reliable, it will lay the groundwork, building also on recent work implicating a role for the KOR-dynorphin system in anhedonia, for future clinical trials testing the hypothesis that KOR antagonism has potential for treating the clinically significant symptom of anhedonia, and/or additional negative symptoms related to reward dysfunction, in schizophrenia. This is an exciting and needed new direction for reducing disease burden in patients.

Author contributions

Mark Slifstein had oversight of all data acquisition, study design, analysis, statistical analysis and primary manuscript preparation. Wenchao Qu developed the radiochemistry methods and had oversight of radiopharmaceutical production. Roberto Gil performed all patient recruitment and screening, medical assessment, guidance of research participants through the imaging protocol and editing of the manuscript. Jodi Weinstein oversaw acquisition and performed analysis of neuromelanin MRI and participated in manuscript preparation. Greg Perlman performed all clinical assessments (PANSS, SAPS/SANS, SCID-5) and participated in manuscript preparation. Thomas Jaworski-Calara was the primary research coordinator on the study. He guided all participants through all aspects of the study and had primary responsibility for healthy control recruitment. Jiayan Ming performed all preprocessing of the PET data and manual delineation of regions of interest for PET analysis. Bao Hu was the chief production chemist who contributed to radiochemistry methods and had oversight of all radiopharmaceutical production. Scott Moeller provided guidance on interpretation of clinical measures and provided editorial oversight and participation in manuscript preparation. Guillermo Horga provided guidance on neuromelanin MRI methods and interpretations and contributed editorial oversight during manuscript preparation. Anissa Abi-Dargham was the principal investigator of NIMH 5R21MH125454-02. She developed the study design and had oversight of all aspects of the study performance and manuscript preparation.

Funding

This study was supported by an NIMH R21 grant (NIMH 5R21MH125454-02).

Competing interests

MS is a consultant to Neurocrine Biosciences Inc. GH has an investigator-initiated research agreement with Terran Biosciences and has filed patents for neuromelanin sensitive MRI in central nervous system disorders. AA is on the Scientific Advisory Boards of Neurocrine Biosciences Inc, Boehringer Ingelheim International, and Abbvie, has received honoraria from Sunovion, holds stock in Herophilus, has stock options in Terran Life Sciences and is on the DSMB for Merck. WQ, RG, JJW, GP, TJ-C, JM, BH, and SJM have nothing to report.

Supplementary information

The online version contains supplementary material available at 10.1038/s41386-024-01975-3.