Alterations in type 2 dopamine receptors across neuropsychiatric conditions: A large-scale PET cohort

Graphical abstract

Highlights

• •We compared human striatal and thalamic type 2 dopamine receptor (D₂R) availability between healthy controls, and subjects with Parkinson’s disease (PD), antipsychotic-naïve schizophrenia, severe violent behavior, pathological gambling, depression, and overweight.
• •We present the mean brain maps of group specific D₂R availabilities in NeuroVault (https://neurovault.org; https://identifiers.org/neurovault.collection:12799).
• •Dopamine type 2 receptor availability is lowered in PD in caudate nucleus, nucleus accumbens and thalamus.
• •Subjects with severe violent behavior had decreased correlation between the striatal and thalamic D₂R availability.
• •Altered regional D₂R availability in the striatum and thalamus is linked with motor disorders, while lowered interregional coupling in D₂R might relate to violence.

Abstract

Purpose

Aberrant dopaminergic function is linked with motor, psychotic, and affective symptoms, but studies have typically compared a single patient group with healthy controls.

Methods

Here, we investigated the variation in striatal (caudate nucleus, nucleus accumbens, and putamen) and thalamic type 2 dopamine receptor (D₂R) availability using [¹¹C]raclopride positron emission tomography (PET) data from a large sample of 437 humans including healthy controls, and subjects with Parkinson’s disease (PD), antipsychotic-naïve schizophrenia, severe violent behavior, pathological gambling, depression, and overweight. We analyzed regional group differences in D₂R availability. We also analyzed the interregional correlation in D₂R availability within each group.

Results

Subjects with PD showed the clearest decline in D₂R availability. Overall, the groups showed high interregional correlation in D₂R availability, while this pattern was weaker in violent offenders. Subjects with schizophrenia, pathological gambling, depression, or overweight did not show clear changes in either the regional receptor availability or the interregional correlation.

Conclusion

We conclude that the dopaminergic changes in neuropsychiatric conditions might not only affect the overall receptor availability but also how coupled regions are across people. The region-specific receptor availability more profoundly links to the motor symptoms, while the between-region coupling might be disrupted in violence.

1. Introduction

Neurotransmitter dopamine is centrally involved in motor, motivational and emotional processes (Beaulieu et al., 2015). The four major dopaminergic pathways are i) nigrostriatal pathway running from substantia nigra to the dorsal striatum (i.e. caudate nucleus and putamen), ii) mesolimbic pathway from ventral tegmental area (VTA) to the ventral striatum (i.e. nucleus accumbens), iii) mesocortical pathway from VTA to prefrontal and cingulate cortex, and iv) tuberoinfundibular pathway from hypothalamus to pituitary gland (Harsing, 2008, Purves et al., 2018). The nigrostriatal pathway is critically involved in motor function and Parkinson’s disease (PD) (Meltzer & Stahl, 1976), while the mesolimbic pathway (also known as the reward pathway) contributes to motivation (Merims and Giladi, 2008, Purves et al., 2018). Particularly, the striatal type 2 dopamine receptor (D₂R) availability has been linked with a variety of neuropsychiatric symptoms (De Keyser et al., 1988, Leggio et al., 2016, Usiello et al., 2000).

Lowered dopamine transmission in the nigrostriatal pathway is associated with motor symptoms (e.g. rigidity, and bradykinesia) in PD (Leggio et al., 2016), while hyperactivity of the dopamine neurons in the mesolimbic track might promote psychotic symptoms in schizophrenia (Meltzer & Stahl, 1976; but see also Stahl, 2018). Antipsychotic-naïve schizophrenia patients with first-episode psychosis show lowered availability of thalamic D₂Rs (Plavén-Sigray et al., 2022), possibly also linked to elevated presynaptic dopamine release, observed at least in the closely located striatum (Howes et al., 2012). However, while increased presynaptic dopamine synthesis and release is characteristic for schizophrenia, only some patients have altered postsynaptic D₂R, pointing towards subtypes of schizophrenia (Brugger et al., 2020).

Dopamine also contributes to impulsive behavior, such as aggressive outbursts and drug abuse (Schmidt et al., 2005, Schultz, 1998, Seo et al., 2008, Volkow et al., 2004). Mesolimbic dopamine pathway and upregulated striatal dopamine function is critical for aggressive behavior (Schmidt et al., 2005). Hyperactive striatal dopaminergic function is also consistently linked with acute effects of drug abuse in addicted subjects, who overall show diminished dopamine and D₂Rs in the striatum (Volkow et al., 2004). Baseline downregulation of dopamine in addictions might associate with their commonly comorbid affective symptoms (Marshall & Farrell, 2007), such as anhedonia (lack of pleasure) and dysphoria (dissatisfaction), that are coherently connected to diminished mesocorticolimbic dopamine (Bressan & Crippa, 2005). Overall, affective symptoms are involved in numerous neurological and psychiatric conditions, including PD, schizophrenia and depression (Bressan and Crippa, 2005, Knable et al., 1997).

Finally, drugs that enhance striatal dopaminergic neurotransmission effectively treat motor symptoms (Poewe & Mahlknecht, 2020) but can also induce psychotic symptoms, impulsivity, and addictive behavior (e.g. gambling and overeating) (Meltzer and Stahl, 1976, Merims and Giladi, 2008, Witjas et al., 2005). Conversely, neuroleptics, pharmacological treatments that inhibit the striatal dopaminergic function by blocking the D₂Rs, reduce psychotic symptoms (Bressan and Crippa, 2005, Kapur and Mamo, 2003) and aggressive behavior (Schmidt et al., 2005), but induce Parkinsonian-like side effects, such as rigidity and bradykinesia (Harsing, 2008, Jackson and Westlind-Danielsson, 1994, Shiraiwa et al., 2018, Stahl, 2018), and anhedonia (Bressan & Crippa, 2005). Particularly the antipsychotics that affect the nigrostriatal track in addition to the mesolimbic track might be the ones with motor side-effects (Jackson & Westlind-Danielsson, 1994), further supporting the centrality of the nigrostriatal track in motor functions.

Taken together, existing data suggest that elevated striatal dopamine activity contributes to psychotic symptoms, impulsivity, aggressive outbursts, and acute effects of reward (namely intense hedonia / pleasure), while lowered dopamine activity relates to disturbed motor functions, and blunted affect (e.g., anhedonia, diminished reward experiences). Because several neuropsychiatric conditions share symptoms (Knable et al., 1997, Seo et al., 2008), perturbations in the dopamine system might link to specific neuropsychiatric symptoms regardless of the diagnosis. To resolve the links between the pathologies with shared dopamine-regulated symptoms of voluntary movement, motivation, and affect, it is crucial to study them together. Only this allows unravelling whether differential patterns of dopaminergic dysfunctions underlie different neuropsychiatric conditions. Here, by neuropsychiatric we refer to a combined set of pathologies with an organic (neurological) and/or psychological basis.

We investigated the differences in the striatal (caudate nucleus, nucleus accumbens, and putamen), and thalamic D₂R availability in a large (n = 437) sample of subjects using positron emission tomography (PET) data with radioligand [¹¹C]raclopride acquired in resting baseline state. The dopaminergic function includes tonic and phasic dopamine firing (Grace, 2016). Because our data pertain to baseline, resting state scans, these data reflect predominantly tonic rather than phasic dopaminergic function. Our sample included healthy controls, and subjects with PD, antipsychotic-naïve schizophrenia, severe violent behavior (violence), pathological gambling (gambling), depression, and overweight. We also analyzed between-region correlation of the D₂R availability in each group and assessed the effects of age and sex on D₂R availability in i) the healthy subjects and ii) the other six groups (i.e. groups of interest) together.

2. Material and methods

2.1. Data

Basic subject information is shown in Table 1. This retrospective register-based study includes 437 subjects (294 males, 143 females), including healthy controls (n = 239), and subjects with PD (n = 60), schizophrenia (n = 7), severe violent behavior, i.e. prison sentence for violence (n = 10), pathological gambling (n = 12), depression (n = 12), and obesity/overweight (n = 97) who had undergone [¹¹C]raclopride PET scan. The age range of the subjects was 19–82 years. The data were retrieved from the Turku PET Centre Aivo database (https://aivo.utu.fi), including data of imaging studies conducted at the site. See supplementary material for Original publications whose data are used in the current study. The images were acquired between the years 1989 and 2019 with six different PET scanners (GE Advance, HR+, HRRT, GE Discovery VCT PET/CT, GE Discovery 690 PET/CT, Ecat 931). The scanners are described in detail in our previous work (Malén et al., 2022), and the distribution of the different PET scanners across the subject groups are given in supplementary Table S1. The framing is given in the supplementary Table S2.

Table 1.

Age of the sample groups separately for males and females in the final sample (n = 437).

	Males (n = 294)				Females (n = 143)
	N	Mean	SD	Range	N	Mean	SD	Range
Healthy controls	174	28	11	19–78	65	41	16	19–82
Parkinson’s disease	31	62	12	34–82	29	60	11	41–81
Schizophrenia	3	22	3	19–25	4	36	13	24–53
Violence	10	31	7	24–46	0	–	–	–
Pathological gambling	12	32	8	23–49	0	–	–	–
Depression	3	38	10	25–48	9	41	5	34–50
Overweight	61	29	9	19–56	36	43	13	23–74

We retrieved the following subject information at the time of each PET scan: subject age, sex, body mass index (BMI, available for 67 % of the subjects) and group, as well as the injected dose and the PET scanner used to acquire the imaging data. Although we do not have detailed medication information for all the subjects, the exclusion criteria for the original research projects where that the subjects were drawn from contained medication affecting the central nervous system. The schizophrenia group was antipsychotic-naïve, and the depression group drug-free (mainly drug-naïve subjects with moderate depressive episode). The PD sample contained both unmedicated and medicated patients and given the mean disease duration of 10 years (median = 9 years, range from 0 to 29 years, see supplementary Figure S1 under D₂R availability through PD duration), most of the patients were probably medicated at the time of imaging. The scans were, however, performed after a short washout period (subjects were withdrawn from medication at least from the night before the scan).

Patient categorization was based on their enrollment criteria in the original studies, except some healthy control subjects were classified to the overweight group if their BMI was higher than 25 (Engin, 2017). The BMI mean (standard deviation) of the remaining subjects were: PD 26.8 (5.1), schizophrenia 26.9 (3.0), severe violent behavior 27.2 (2.3), pathological gambling 27.8 (3.4), and depression 26.5 (5.2) (supplementary Table S3). Thus, some subjects in these groups are overweight, although their primary diagnosis was not obesity. However, none of the healthy controls were obese. These information regard 67 % of the total sample for who we had BMI information available (supplementary Table S3). None in the obese group are expected to have psychiatric or neurological disorders, as these disorders were a general exclusion criterion in the brain studies at the site (except in projects specifically studying those conditions). More detailed symptoms, alcohol use, or smoking status were not used in the analysis, as this information was not systematically available for all the subjects.

The potential sample of 455 subjects was cut down to 437 after excluding 18 subjects due to the following reasons. Three subjects with low injected dose < 90 megabecquerel (MBq) and seven without dose information were excluded from the sample to avoid low signal to noise ratio. The excluded subjects were from healthy control, PD, schizophrenia, and overweight groups. Due to lost or noisy data from interrupted PET scanning, we discarded eight subjects (including subjects from PD, schizophrenia, and severe violent behavior groups) whose kinetic model fits were inadequate.

2.2. PET data preprocessing and kinetic modelling

PET image preprocessing and kinetic modelling were carried out in MATLAB (The Mathworks Inc., 2021) using an in-house brain image processing pipeline Magia (Karjalainen et al., 2020) (https://github.com/tkkarjal/magia). We spatially normalized the dynamic PET images into MNI152 space with an in-house [¹¹C]raclopride PET-template created using a subset (n = 187) of healthy control subjects, because only a subset of the sample had magnetic resonance (MR) image available (n = 249) (Karjalainen et al., 2020) (see Validation of PET atlas based spatial normalization of the [¹¹C]raclopride binding estimates and Figures S2-S3 in supplementary material). Four bilateral regions of interest (ROI) from Harvard-Oxford atlas, caudate nucleus (caudate), nucleus accumbens (accumbens), putamen, and thalamus, were applied to the normalized PET-images. [¹¹C]raclopride is a commonly used tracer to investigate striatal and thalamic D₂R (Alakurtti et al., 2015, Elsinga et al., 2006), as its binding to D₂Rs in these regions is reliable, while negligible elsewhere (Hirvonen et al., 2003, Svensson et al., 2019).

Regional specific binding of [¹¹C]raclopride was quantified as nondisplaceable binding potential (BP_ND) using simplified reference tissue model (SRTM) and Harvard-Oxford atlas cerebellum as a reference region (Lammertsma & Hume, 1996). Tails beyond the cerebellum distribution's full width at half maximum were removed to ensure that the reference region excludes voxels with unusually high or low radioactivity (Karjalainen et al., 2020). BP_ND is a measure of receptor density and radioligand affinity and refers to the ratio at equilibrium of specifically bound radioligand to that of nondisplaceable radioligand in tissue (Innis et al., 2007). In the analyses, we used the mean of left and right hemisphere BP_ND as the regional estimate for each subject. Data on hemispheric symmetry in BP_ND is presented in supplementary Tables S4-S5 and Figures S4-S5.

Parametric BP_ND images were also calculated for illustrative purposes with basis function implementation of SRTM (bfSRTM) with 300 basis functions. Lower and upper bounds for theta parameter were set to 0.082 1/min and 0.6 1/min. Before the parametric image calculation, the dynamic PET images were smoothed using Gaussian kernel with 4 mm full width at half maximum to reduce the effect of noise in the voxel-level bfSRTM fit.

2.3. Statistical analysis

Statistical analysis was carried out in RStudio (Posit team, 2023) using R (R Core Team, 2023). We assessed the normality of the data with Shapiro-Wilk test with package stats (R Core Team, 2023). We analyzed the regional data using Bayesian modeling tools of brms (Bürkner, 2017, Bürkner, 2018, Bürkner, 2021) utilizing rstan (Stan Development Team, 2020). We analyzed the interregional BP_ND correlation in each subject group using correlation matrix with corrplot (Wei, 2021), and Mantel test (Mantel, 1967) for similarity of two matrices using ape (Paradis & Schliep, 2019). We used R (R Core Team, 2023) (particularly R packages ggplot2 (Wickham, 2016), ggsci (Xiao, 2023), bayesplot (Gabry, 2022), superheat (Barter, 2017), corrplot (Wei, 2021), and GGally (Schloerke, 2021)), MATLAB (The MathWorks Inc., 2021), and MRIcroGL (https://www.nitrc.org/projects/mricrogl) to visualize the data and findings.

2.3.1. Descriptives

We created group-specific maps of the BP_ND i) means and ii) coefficients of variation, to describe the homogeneity of D₂R distribution between and within the groups, respectively. We calculated the statistics using MATLAB (The Mathworks Inc., 2021) and illustrated them as maps using MRIcroGL (https://www.nitrc.org/projects/mricrogl).

2.3.2. Regional group differences in D₂R availability

We used linear mixed effects regression to compare regional D₂R availability across the seven subject groups. In the analysis, we adjusted for the scanner used in the image acquisition, and the subject age and sex that affected striatal D₂R availability in healthy controls (Malén et al., 2022).

The regional log-transformed (log) binding potentials (BP_ND) were modeled as a dependent variable. We analyzed the group differences in log-BP_ND in caudate, accumbens, putamen, and thalamus, while adjusting for age, sex, and scanner. Specifically, we applied group, standardized age, and sex as fixed effects and allowed these effects to vary between ROIs (regionally varying a.k.a. random slopes). We applied random intercepts for subjects, scanners, ROIs, and scanner-ROI combinations. This means that we allowed the log-BP_ND intercept to vary by subject, region, scanner, and scanner within each region. For the residual variances, we similarly applied random intercepts for scanners, ROIs, and the scanner-ROI combinations, but not for subjects as we did not expect the residual variances to depend on subject. We validated the model by comparing it with three other candidate models (see Model diagnostics and comparison in supplementary material).

The age effect was calculated for the whole sample together (and not separately for each subject group by interaction of age and group), because we wanted to maximize the age range and the sample size in the estimation of the age effect, while some of the studied groups had relatively limited age range (e.g. schizophrenia 19–25 years). Group-specific age effects were also not hypothesized as e.g. prior studies suggest that neural changes in PD do not reflect ‘accelerated aging’ (Kaasinen et al., 2015). Similarly, the sex effect was calculated for the whole sample because some of the studied groups consisted solely of males. Nevertheless, of all the studied groups, the healthy controls hold the greatest number of observations altogether and separately for males and females, as well as the widest age range. Thus, the age and sex effects observed in the healthy control group were expected to have a strong representation in the regression analysis due to large proportion of healthy controls in the total sample.

Shapiro Wilk testing (R Core Team, 2023) did not clearly support the normality of region-specific BP_ND in the original nor the logarithmic scale. We decided to use the log-transformed instead of the original-scale BP_ND as a dependent variable in the regression model, because log-transformed dependent variable allows us to directly estimate the relative (e.g. 10 %) rather than absolute (e.g. 1 unit) change in BP_ND between predictor levels, particularly the group specific intercepts. Relative change is more intuitive, while BP_ND is a ratio that describes receptor availability, and is not a direct measure of absolute receptor density (Innis et al., 2007).

We used linear regression, because based on our previous assessment (Malén et al., 2022), the relationship between age and logarithmic BP_ND is well approximated by a linear function in a healthy sample, that is the majority of the current sample, and as stated previously, we are not aware of group-specific age-effects. Further, modeling higher-degree associations increases the risk of overfitting, potentially resulting in poor generalization of the results (McElreath, 2020). For the remaining fixed predictors (group and sex), the estimated parameters are intercepts and not slopes.

In the absence of appropriate priors for Bayesian statistical inference from independent data, we gave a normal distribution prior (with expected value = 0 and variance = 1) for the fixed effects (age, sex, group). Such zero-centered normal prior assigns symmetrical probabilities for both positive and negative regression coefficients and does not restrict the parameter to certain values (i.e., no regression coefficient is given zero probability), while it still regularizes the model’s fit. For the standard deviations of the random effects, we set the same normal prior. As standard deviation is always positive, brms-package automatically restricts this prior to positive values, making it a half-normal distribution (Bürkner, 2017, Bürkner, 2018, Bürkner, 2021). For the rest of the modeling parameters, we used weakly informative default priors of brms (Bürkner, 2017). The model was carried out using Markov Chain Monte Carlo sampling with each four chains running 4000 iterations (including 1000 warm-up samples).

2.3.3. Validation of the age and sex effects on regional D₂R availability

For validation purposes, we assessed the age and sex effects in the current sample. We tested the replicability of the age and sex effects (Malén et al., 2022) in the i) healthy control sample (almost identical sample to our earlier work (Malén et al., 2022)) and ii) groups of interest together (subjects with PD, schizophrenia, severe violent behavior, pathological gambling, depression, and overweight). These analyses are presented in Validation of the age and sex effects on regional D₂R availability in supplementary material).

2.3.4. Between-region correlation of the D₂R availability

We used Spearman correlation to assess how similar the BP_ND rank-order across subjects was between the regions of interest. This was done separately in each subject group to assess if the regional correlation patterns were consistent across groups. We computed a 4 × 4 between-region BP_ND correlation matrix for each subject group. These matrices were computed with the original binding estimates, because taking the logarithm of the data (group and region specific BP_ND) did not consistently transform it closer to normal distribution (Shapiro-Wilk), and the rank-order based Spearman correlation readily accounts for monotonic associations in the data and is unaffected by log transformations. Using Spearman and not Pearson correlation is also supported by the limited sample sizes in some of the studied groups. One-tailed testing was used as regional coupling estimates for D₂R are positive rather than negative (Malén et al., 2022).

We tested for the similarity of each correlation matrix pair (e.g. correlation matrix of healthy controls compared to correlation matrix of PD patients) using one-tailed Mantel tests with default number (9 9 9) of permutations (Paradis & Schliep, 2019). These analyses allowed us to assess, whether interregional D₂R coupling is altered in neuropsychopathology, and whether different neuropsychiatric conditions differ in their striatal and thalamic correlation patterns.

3. Results

3.1. Descriptives

Group-specific maps for mean and coefficient of variation for D₂R are shown in Fig. 1. Based on the Fig. 1, the groups show broadly similar receptor distribution in the subject groups, and the binding level shows greatest variation in the PD group. Across groups, the coefficient of variation appears greatest in thalamus. The mean D₂R maps are also available in NeuroVault (https://identifiers.org/neurovault.collection:12799). Mean regional non-transformed BP_ND estimates for each group are shown in Fig. 2. D₂R availability decreases through PD duration, please see D₂R availability through PD duration and Figure S1 in the supplementary material.

Fig. 1.

Maps of the original-scale [¹¹C]raclopride BP_ND estimates in the subject groups. Left: Mean BP_ND. Right: Coefficients of variation (standard deviation / mean) for BP_ND within the regions of interest (Harvard-Oxford atlas masks). The maps are presented on an MNI-template.

Fig. 2.

Barplot of the group-specific mean original-scale BP_ND in each ROI. The black intervals show the group-specific standard errors (sd(BP_ND)/sqrt(n), where sd = standard deviation, sqrt = square root, n = number of observations).

3.2. Regional group differences in D₂R availability

Bayesian linear mixed effects modeling (Fig. 3), adjusting for age, sex, and scanner, revealed that compared to the other groups, PD patients had approximately 10–20 % lower [¹¹C]raclopride BP_ND (values transformed back to original scale) in caudate, accumbens and thalamus, while the BP_ND in putamen was approximately 10 % higher (Fig. 4). The other groups of interest (schizophrenia, severe violent behavior, pathological gambling, depression, and overweight) did not show clear differences in regional D₂R availability when compared with healthy controls or each other. However, subjects with schizophrenia and severe violent behavior showed more support for lowered than elevated D₂R availability (more probability mass below zero), although the 80 % posterior intervals (i.e. probability mass) crossed zero. Pathological gambling, depression and overweight groups showed the least support for altered D₂R availability with practically non-existent difference compared to healthy controls.

Fig. 3.

Posterior probability distributions (posteriors), as well as their median (point), 80% (thick line) and 95% (thin line) posterior intervals, describing the between-group differences in log-transformed BP_ND. Posterior located on the negative side of the zero line suggests lower log-BP_ND in the reference group on the left compared with the group on the right.

Fig. 4.

Summary of the regional group comparisons. The values represent the (median) proportional BP_ND (values transformed back to original scale) of the reference group (on the left) compared with the group on the right. Asterisks represent comparisons were 80% (*) and 95% (**) posterior interval did not cross zero.

3.3. Validation of the age and sex effects on regional D₂R availability

Similarly as in our previous analysis with healthy subjects (Malén et al., 2022), D₂R availability decreased through age, both in the healthy control group that is almost identical to the sample in our previous work (Malén et al., 2022) and in the groups of interest. The lower striatal D₂R availability in males versus females was only found in healthy controls. However, in thalamus, females had higher thalamic availability in both the healthy controls and the groups of interest. The results are illustrated in Figure S6. For comparison, Figure S6 also shows the age and sex effects of the main model and confirms that these effects are roughly similar between the models.

3.4. Between-region correlation of the D₂R availability

The interregional correlations of original-scale BP_ND are shown in Fig. 5. Additional between-region scatter, and density plots are given in supplementary Figure S7. Despite the regional BP_ND changes, PD patients showed high between-region BP_ND correlation, comparably with healthy controls. Across all groups, the weakest correlations were observed between thalamus and other regions. Subjects with severe violent behavior showed the weakest correlation between the regions.

Fig. 5.

Group-wise interregional original-scale BP_ND correlations (one-tailed Spearman, p-values > 0.05 marked with a circle, while the higher the p-value, the less support for non-zero correlation).

When we compared the correlation matrixes of the groups (Fig. 5) with Mantel tests, we observed similarity in the correlation patterns of the groups. Healthy controls showed some support for similarity with PD (p = 0.04), schizophrenia (p = 0.055), depression (p = 0.037), and overweight (p = 0.04). Further, similarity was suggested between PD and schizophrenia (p = 0.07), PD and depression (p = 0.038), PD and overweight (p = 0.089), and depression and overweight (p = 0.034). The remaining group comparisons, mainly including subjects with severe violent behavior and pathological gambling, did not support similarity, while the p-value was greater than 0.1 in the similarity tests.

Altogether, the groups showed strong dopaminergic correlation patterns, particularly between the striatal regions. The weakest correlations were most notably manifested with thalamus, where the estimates might be less reliable than in the striatum (Hirvonen et al., 2003). The thalamic decoupling was emphasized in the subjects with severe violent behavior, who had lowest correlation pattern also within the striatum.

4. Discussion

Striatal dopamine neurotransmission has a central role in motor functions that are disturbed particularly in PD (Kaasinen and Vahlberg, 2017, Kaasinen et al., 2021), and motivational processes that are impaired in many neuropsychiatric disorders (Bressan and Crippa, 2005, Knable et al., 1997, Schmidt et al., 2005, Schultz, 1998, Volkow et al., 2004). Our findings show that in PD, striatal (caudate and accumbens) and thalamic D₂R availability is downregulated, while interregional correlation might be lowered in severe violent behavior. These results suggest that in the striatothalamic dopamine circuit, overall changes in receptor levels and specific regional decoupling patterns associate to different neuropsychiatric conditions.

4.1. Regional D₂R changes in Parkinson’s disease

Striatal (caudate and accumbens) and thalamic downregulation in D₂R availability was most salient in PD patients. There was limited evidence for striatal downregulation also in schizophrenia and severe violent behavior, but even the 80 % posterior intervals for these groups overlapped with zero. Our findings support a previous meta-analysis describing that the effect of the striatal D₂R in schizophrenia is negligible (and possibly linked to the antipsychotic treatment) compared to the presynaptic dopamine synthesis and release (Howes et al., 2012). The other groups did not show clear alterations in striatum nor thalamus.

In the regression modeling, we used normal prior with zero mean and variance of one for the regression coefficients (beta) of the fixed effects (including group). With variance of one, the prior distribution was wide. The prior placed 50 % probability mass for beta from −0.67 to 0.67 that roughly corresponds to a range from 50 % reduction to 100 % increase in BP_ND in the comparison group compared to the reference group. In this sense, the prior is weakly informative and allows a wide range of regression coefficients for groups. The prior, however, places the highest probability to zero difference between groups, which could override the data from small groups and overlook some of the group differences. We, thus, believe that in our analysis of the regional group differences including small groups, there was higher probability of finding false similarity than false difference between the groups.

Compared with the other groups, PD patients exhibited both lowered (accumbens, caudate, thalamus) and elevated (putamen) striatal D₂R availability when adjusting for age and sex. PD subjects had lowered D₂R availability in nuclei caudate and accumbens, and thalamus. In putamen, PD patients had high D₂R availability, although this effect is not discernible in the raw data where age and sex are not adjusted for.

Although age and group were both accounted for in the regression modeling, downregulation of D₂R in PD might have been partly misinterpreted as an age effect, as PD patients are on average older than the other subjects (Table 1). This possible overcorrection of the age effect might have contributed to the sign switch of the observed alteration of D₂R in the PD group (positive in the putamen, negative in all other ROIs). However, the age and sex effects were roughly similar in the main model (with all groups included) as in the model of only healthy controls (Figure S6). These findings suggest that the low D₂R availability of PD patients did not accelerate the negative age effect in the main regression model for the group comparison. Finally, there are a few putamen observations of PD patients that exceed the ones of healthy controls (Figures S4-S5), that might have also lifted the PD intercept in putamen.

This finding requires additional validation because in PD, the D₂R changes in putamen are dynamic and dependent on the disease stage. Compensatory D₂R upregulation in putamen changes to downregulation around four years from PD onset (Kaasinen et al., 2021, Rinne et al., 1990). The PD duration greatly varied between the studies subjects (range: 0–29 years) and our findings cannot directly discern the effects in early and late stages, as the disease duration information was available for only half of the patients. Thus, we are unable to specify the weight of early and late stage patients in our data and analysis. For the subjects with available information, the correlation between disease duration and BP_ND was subtle but nevertheless negative in all tested regions (-0.17 in putamen) suggesting no prolonged upregulation with progressed PD. Consequently, our findings do not provide clear evidence against late stage D₂R downregulation in putamen. Overall, the findings nonetheless highlight the region and motor symptom specificity of within-region D₂R changes.

4.2. Lowered between-region correlation in severe violent behavior

Dopaminergic changes in severe violent behavior were more prominent in the regional coupling than in the overall regional receptor availability. Although in all groups we observed that thalamus showed the lowest regional correlation, this decoupling was most salient in the subjects with severe violent behavior, who also had lowest regional coupling within the striatum. Aberrant coupling in this striatal dopamine network might reflect altered communication and weakened links between the regional nodes. Alternatively, the weakened correlation of D₂R availability between different regions might also be explained by selective alteration of one specific dopaminergic region, while other regions are unaffected (without interaction between regions). However, this hypothesis was not notably supported by the regional analysis where the binding in subjects with severe violent behavior did not clearly differ from healthy controls in any of the regions.

We found some support for the similarity in the regional coupling patterns (correlation matrices) between the groups, except for group comparisons including severe violent behavior and pathological gambling. However, as the matrices we compared only had six individual data points (correlation estimates of 1) caudate-accumbens, 2) caudate-putamen, 3) caudate-thalamus, 4) accumbens-putamen, 5) accumbens-thalamus, 6) putamen-thalamus), the tests might be underpowered to show significant similarity between the matrices even if similarity exists, particularly if the reliability of the six input correlations (in the matrices) are calculated from a small sample size.

Altogether, the sample size for violent offenders was small which might limit the generalizability of our findings, although some of the other groups showing stronger correlation pattern were also small. Our findings nevertheless suggest different dopaminergic correlation profiles between motor symptoms and violent aggression. The D₂R (de)coupling of the regions in violence and affective disorders such as schizophrenia, pathological gambling, and depression (for which we had limited sample sizes), should be further examined in larger studies. Studies investigating the D₂R correlates of specific behavioral symptoms would also complement our findings regarding D₂R variation in relation to more specific clinical characterizations. Finally, we encourage future studies to address D₂R alterations also in the cortex using other radioligands such as [¹¹C]FLB457.

4.3. Unaltered baseline D₂R in overweight

We observed neither regional D₂R alterations nor aberrant regional coupling in schizophrenia, pathological gambling, depression, or overweight groups. While some of the effects in the first three groups might have result from limited sample sizes, overweight group included almost a hundred subjects making these findings more robust. Our results, thus, underline that dopaminergic mechanisms do not notably contribute to general weight gain (without any subgrouping).

Several recent studies have addressed the conflicting findings regarding the link between overweight and D₂R (Dagher, 2012, Darcey et al., 2023, Horstmann et al., 2015). They suggest that BMI and striatal D₂R availability may not be strictly linear. Mild obesity might link with low tonic and high phasic dopamine, while severe obesity would show the opposite pattern (high tonic, low phasic dopamine) (Horstmann et al., 2015). In our previous study we assessed both linear and non-linear associations between BMI (range 18–38) and D₂R availability finding only subtle support for positive linear association between the measures (Malén et al., 2022). Here we instead compared the baseline intercepts of the studied groups. Our findings accords with previous empirical studies (Karlsson et al., 2015, Malén et al., 2022) as well as meta-analyses (Pak and Nummenmaa, 2023) finding no clear connections between altered D₂R availability and obesity. Although overweight subjects did not differ from other studied groups in their baseline D₂R availability, this does not exclude the existence of a more complex or nuanced links between weight and dopaminergic function.

Finally, our results highlight that motor symptoms, rather than affective disorders or weight gain, are most consistently linked with regional alterations in D₂R availability. According to our data, PD is associated with changes in regional D₂R availability, while severe violent behavior might link to disturbed correlation between the regional availabilities. Our results regarding unaltered D₂R in overweight suggests that the postsynaptic type 2 dopamine receptor in the striatum and thalamus is not a central contributor of general weight gain, and possibly neither to schizophrenia, depression nor pathological gambling, although more data are needed to clarify the issue. Overall, this large multi-group dataset suggests that the dopaminergic changes observed in neuropsychiatric pathology are multifaceted, involving region-specific and between-region mechanisms of the D₂R. We conclude that the changes in baseline D₂R might be more profoundly linked with motor symptoms and violent aggression than with affective disorders or weight gain.

4.4. Effects of age and sex on D₂R availability

We found that baseline striatal D₂R availability is lowered through age both in healthy controls and in the groups of interest. Previously, we observed sex differences (female > male) in healthy controls consistently in the striatothalamic regions (Malén et al., 2022), whereas here in the groups of interest, sex differences were only found in the thalamus. However, this effect pertains to the joint analysis of the groups of interest together. Because all the groups did not include both males and females, we pooled the groups together for the analysis of age and sex, which might have overlooked group specific sex differences. Nevertheless, our data indicate that the effect of age on D₂R is more consistent than the effect of sex, and generally found also in clinical populations. Accordingly, adjusting for age and sex is necessary when assessing the effects of other variables on striatothalamic D₂R availability.

5. Limitations

Our data were not fully balanced. Although the dataset was large, the group size, age, sex, and scanner distribution were compromised in some of the studied groups. Combined with our statistical analysis of the regional group differences that adjusted for the age, sex, and scanner related variation in the data, some group differences may have been undetected (McElreath, 2020). In the between-region analysis, our findings mainly indicate that in the healthy control, PD, and overweight groups (where we had largest number of subjects) the between-region binding correlation is strong. Larger samples are required to validate whether the striatothalamic correlation pattern is similarly strong or decoupled in the other groups with limited sample sizes. Subjects with severe violent behavior, pathological gambling and depression were imaged with a single scanner (severe violent behavior group with GE Discovery VCT PET/ CT, pathological gambling, and depression groups with Ge Advance) (Table S1). However, the data from GE Discovery VCT PET/ CT and Ge Advance included 9 and 69 healthy control images, respectively. Additional non-healthy subjects were also imaged with these two scanners. This means that none of the scanners represented exclusively one subject group, theoretically making it possible for the model to differentiate the effects of scanner and group. In a single PET scan, it is not possible to demonstrate the exact molecule-level mechanism for altered receptor availability. Not all subjects were drug naïve. Due to the retrospective nature of the data derived from multiple projects, the collection of clinical characteristics of patient populations were not standardized. Patient categorization was based on their enrollment criteria in the original studies. Finally, it must be noted that unaltered baseline D₂R level does not mean the dopaminergic responses in different environmental contexts would be similar across all studied groups. It is possible that differential patterns of dopamine firing in motivationally triggering environments exist for example, towards gambling in pathological gamblers and towards feeding in obese individuals (Grace, 2016). Investigating such effects however requires standardized challenge or activation paradigms, which could not be incorporated in the current analysis.

6. Conclusions

Dopaminergic function is altered across multiple psychiatric and neurological conditions. The regional receptor availability patterns, as well as interregional coupling of D₂R levels, might distinguish between the specific conditions. PD was associated with region-specific changes in D₂R availability. The role of D₂R in violent behavior might rather relate to the interregional connections, as in this subject group, the between-region receptor availabilities were less correlated. In pathological gambling and overweight groups, no clear changes were observed in the striatothalamic D₂R level nor the correlation structure. Our results suggest that in striatum and thalamus, motor symptoms are more profoundly linked to the region-specific D₂R availability, while violence might be more associated with lowered correlation between the regions.

Ethics approval

Finnish legislation does not require ethical approval for register-based studies.

CRediT authorship contribution statement

Tuulia Malén: Writing – review & editing, Writing – original draft, Visualization, Validation, Project administration, Methodology, Formal analysis, Data curation, Conceptualization. Severi Santavirta: Writing – review & editing, Visualization, Validation, Methodology, Conceptualization. Sven De Maeyer: Writing – review & editing, Visualization, Software, Methodology. Jouni Tuisku: Writing – review & editing, Visualization, Validation, Software, Methodology, Formal analysis, Data curation. Valtteri Kaasinen: Writing – review & editing, Resources, Investigation, Data curation, Conceptualization. Tuomas Kankare: Writing – review & editing, Resources, Investigation, Data curation. Janne Isojärvi: Writing – review & editing, Software, Data curation. Juha Rinne: Writing – review & editing, Resources, Investigation. Jarmo Hietala: Writing – review & editing, Resources, Investigation. Pirjo Nuutila: Writing – review & editing, Resources, Investigation. Lauri Nummenmaa: Writing – review & editing, Visualization, Supervision, Resources, Investigation, Funding acquisition, Conceptualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article:

Data availability

We present the mean brain maps of group specific D₂R availabilities in NeuroVault (https://neurovault.org; https://identifiers.org/neurovault.collection:12799).