Introduction

The D2 dopamine receptor (D2R) is a G protein-coupled receptor coded by the DRD2 gene and is involved in essential brain functions such as learning, memory, locomotion, attention, motivation, sleep, emotion processing, reproductive behaviour1,2,3. The D2R is also the primary site of the therapeutic action of antipsychotics4,5,6,7. Furthermore, one of the schizophrenia-associated loci from Genome-Wide Association Studies (GWAS)8,9 includes the D2R coding gene (DRD2), implicating this gene in the pathophysiology of schizophrenia10.

Genetic variations within DRD2 have been associated with brain-related phenotypes, including working memory, sustained attention, variable attention control, emotion processing, dopamine binding in the striatum, suggesting that genetic mechanisms influence the effects of the D2R on brain function11,12,13,14,15,16,17. However, it is unlikely that genetic variations within a single gene explain the entire physiology related to specific brain phenotypes. In this regard, previous investigations have elucidated that genes involved in complex traits do not work in isolation but operate in networks of interacting genes18,19,20,21,22 acting via molecular pathways23,24,25. Genetic networks can be investigated in detail using methods for the analysis of gene co-expression patterns26,27. This approach is based on the evidence that the expression of different genes is influenced by common regulatory molecules, and that such gene expressions correlate24,28,29,30,31,32. Co-expressed genes are often related in terms of function33,34. A widely used technique to study gene co-expression is the weighted gene co-expression network analysis (WGCNA). WGCNA represents correlated gene expression into a graph that is designed to be scale invariant, hence reflecting the basic property of biological networks that include highly connected central hubs and more peripheral genes. Hierarchical clustering is used in WGCNA to define gene sets, called modules, that are tightly co-expressed. This approach has been used to identify, in post-mortem dorsolateral prefrontal cortex of healthy controls, a network of genes co-expressed with DRD235, including genes associated with schizophrenia identified in the PGC29 and PGC3 GWAS36. A follow-up study has supported with in vitro evidence the link between some of these genes and identified potential co-regulators32. Interestingly, a Polygenic Co-expression Index (PCI) predicting the brain-specific expression of this network of co-expressed genes was associated with response to antipsychotics and prefrontal inefficiency during working memory35, which has been consistently associated with schizophrenia37. Moreover, healthy subjects with higher PCI showed increased activation in the prefrontal cortex and longer reaction times when performing a working memory task35. Interestingly, in a recent network control theory study38 the same PCI has been shown to be related to dynamical brain state transitions during working memory in healthy volunteers. Furthermore, this PCI has been associated with within-subject variation of prefrontal cortex activity following pharmacological stimulation of D2R in a double-blind crossover design39.

While these studies focused on the frontal cortex, DRD2 has its highest expression in the striatum40. Interestingly, molecular imaging studies show evidence that presynaptic striatal dopamine dysfunction plays an important role in abnormal reward processing and anomalies of other aspects of cognitive function41,42. Moreover, elevated striatal dopamine synthesis and release capacity are associated with schizophrenia43,44,45,46,47,48,49,50,51,52,53, psychotic symptoms54 and risk of psychosis55,56.

Whilst the findings discussed above show that the DRD2 gene network is associated with cortical brain function relevant to cognitive phenotypes of schizophrenia57, it remains unknown if and how the genetic underpinnings of cortical dopaminergic function are related to striatal dopaminergic phenotypes associated with psychosis. The exploration of this relationship can be considered as particularly relevant in view of the connections between cortex and striatum58. The aim of the present study is to investigate the relationship between striatal dopamine synthesis capacity and co-expression of the DRD2-related genetic network35. To this aim, we analysed data from a cohort of healthy subjects that underwent 18F-DOPA PET and were genome-wide genotyped; we used the genetic variants indexing the co-expression of the DRD2-related genetic network to compute an individual PCI35. We hypothesised that higher PCI, which has been previously associated with greater prefrontal BOLD response (see also59) and longer reaction times during working memory processing35, would be associated with higher striatal dopamine synthesis capacity—thus outlining a consistent pattern of results resembling the physiological observations in patients with schizophrenia.

Methods

Participants

A total of 92 healthy subjects (demographics in Table 1) underwent 18F-DOPA PET scans60. The study was conducted in accordance with the Declaration of Helsinki and Good Clinical Practice. All participants gave informed written consent. The study was approved by the Administration of Radioactive Substances Advisory Committee (ARSAC), the South London and Maudsley/Institute of Psychiatry NHS Trust, the London Bentham Research Ethics Committee, and the Hammersmith Research Ethics Committee.

Table 1 Demographic characteristics of the sample.
Full size table

Inclusion criteria were: age range 18–65 years, no history of major medical condition, good physical health. Exclusion criteria were: significant medical disorder or treatment, history of psychiatric illness (assessed using the Structured Clinical Interview for DSM-IV Axis I Disorders) including alcohol or substance abuse or dependence. The dataset has been gathered from our publicly available imaging data archive (https://maudsleybrc.nihr.ac.uk/research/precision-psychiatry/neuroimaging/neuroimaging-database-node/). The PET data have been previously published61,62,63,64,65, but the integration with the PCI has not been published before.

Polygenic co-expression index

DNA was extracted from whole blood samples or cheek swabs using standard procedures66. Genome-wide genotyping was performed at Cardiff University, using HumanCore Exome 1.1 arrays ("Psych-chip", Illumina, San Diego, California, USA).

A PCI was calculated as previously described35. Briefly, a DRD2 co-expression gene set, including 85 genes, was identified with a Weighted Genes Co-expression Network Analysis67 using the post mortem frontal cortex mRNA expression Braincloud database68. A set of 8 SNPs (CHIT1 rs2486064, GPLD1 rs6902039, OSR1 rs851436, POP1 rs9297283, SDK2 rs1294071, DHX33 rs1805453, BTG4 rs1121391, AGR2 rs1037791) associated with the first principal component of gene set co-expression was used to compute the PCI; a weight based on the co-expression profile of the gene set was assigned to each genotype of each SNP (Table S1). Genotyping was conducted for these SNPs. Genotype quality control for these SNPs was performed according to standard parameters69. Briefly, these included an individual missingness rate < 0.98, a SNP call rate > 0.98 and a Hardy–Weinberg equilibrium (HWE) p value > 10−4, as computed by the PLINK v1.9 software70.

Population stratification

The Principal Components Analysis in Related samples (PC-AiR) method71 was used in R (GENESIS R/Bioconductor package72) on the full set of genotypes to generate the top 10 principal components of the sample, which were included as covariates of no interest in all the analyses, in order to correct for population stratification.

PET scanning

18F-DOPA PET scans were performed to measure dopamine synthesis capacity (indexed as the influx rate constant Kicer)73.

Image acquisition

Images were acquired in three-dimensional mode using three different PET scanners: an ECAT HR + 962 PET scanner (CTI/Siemens, Knoxville, Tennessee) and two Siemens Biograph HiRez XVI PET-CT scanners (Siemens Healthcare, Erlangen, Germany). After the administration of approximately 150 MBq of 18F-DOPA, dynamic PET data were acquired over a period of 95 min as previously described61,62,63,64,74.

Image processing

The frames were aligned using a mutual information algorithm75. A movement-corrected dynamic image was then used in the analysis. A tracer-specific (18F-DOPA) template76 was normalised together with a striatal probabilistic atlas77 to the individual PET summation images. The influx constant (Kicer) for striatum was calculated using the cerebellum as a reference region78. For the exploratory analyses, the striatum was sub-divided into limbic, associative and sensorimotor parts on the basis of function and the topography of brain projections from limbic, associative and sensorimotor cortical areas to the striatum48,77,78.

Statistical analysis

The effect of the PCI on whole striatal Kicer was tested using a linear model (lm) regression in R79 with age, gender, PET scanner and the first 10 genetic principal components as covariates of no interest in view of their potential effect on dopamine synthesis capacity80,81. To facilitate the interpretation of the results, PCI values were standardised using the scale() function in R before being entered in the model82. Injected dose of radiotracer was not considered, as it is not associated with 18F-DOPA Kicer estimates74. A significance threshold of α < 0.05 was used. Separate exploratory analyses were conducted to test the effect of the PCI on associative striatum, limbic striatum and sensorimotor striatum Kicer. R79 was used for all the statistical analyses. The R package ggplot283 was used to plot the main results. To exclude the presence of outliers, the Rosner's test function ("rosnerTest") of the R package EnvStats84 was used to remove extreme observations.

Results

Demographic (± SD) and Kicer values included are reported in Table 1.

The Rosner's test did not reveal any outliers. PCI was significantly associated with whole striatal dopamine synthesis capacity (t value = 2.106, p = 0.038). Figure 1 illustrates a positive correlation between whole striatum Kicer (y axis) and PCI. PET scanners, included as covariates of no interest, did not show a statistically significant association with dopamine synthesis capacity (t value = 1.603, p = 0.112).

Figure 1
figure 1

Scatterplot illustrating the correlation between whole striatum Kicer (y axis) and PCI.

Full size image

The exploratory analyses in the striatal subdivisions revealed an effect of the PCI on dopamine synthesis capacity for the associative striatum only (t value = 2.063, p = 0.042) (Fig. 2), while there was not a significant correlation with limbic striatum (t value = 1.957, p = 0.054) or sensorimotor striatum (t value = 1.841, p = 0.069). The interaction among striatal subdivision, PCI, and Kicer was not significant (p = 0.738).

Figure 2
figure 2

Scatterplot illustrating the correlation between associative striatum Kicer (y axis) and PCI.

Full size image

Discussion

The present study shows for the first time an in vivo association between striatal dopamine synthesis capacity and a DRD2-related co-expression score in a cohort of healthy subjects. Specifically, high polygenic co-expression index, reflecting greater prefrontal co-expression of a DRD2-related genetic network, was associated with elevated striatal dopamine synthesis capacity. These results suggest that, besides DRD2, several genes and related products may be relevant to the modulation of striatal dopamine function.

Gene co-expression networks have been instrumental in identifying gene sets associated with antipsychotic treatment response30,35, phenotypes associated with schizophrenia25,28,35,85, clinical state and risk for schizophrenia31, and changes in prefrontal function after D2R stimulation39. As reviewed previously, increased dopamine synthesis capacity represents a phenotype associated with schizophrenia47,86,87. Thus, the results of the present study suggest a possible relationship between the DRD2-related co-expression network identified and schizophrenia. Consistently, the exploratory analyses in the different striatal subdivisions suggest that the PCI is associated with dopamine capacity in the associative striatum, which is the striatal region showing greater dopaminergic dysfunction in patients with schizophrenia compared with other striatal subdivisions according to meta-analytic evidence88. Nevertheless, it is important to note that the analyses in the different striatal subdivisions were only exploratory and there was no significant difference between effect sizes across striatal subdivisions. Additionally, the association with the associative striatum Kicer would not survive correction for multiple comparisons. Therefore, these results should be considered as exploratory and require further evaluation in a larger sample.

The hypothesis of a positive relationship between PCI and striatal dopamine synthesis capacity was based on a study35 demonstrating in healthy subjects an association of this index with prefrontal inefficiency during working memory, another phenotype related to schizophrenia. However, it is noteworthy that previous studies have shown both positive89,90,91 and negative92,93 correlations between striatal dopamine synthesis capacity and working memory. Thus, a study investigating striatal dopamine function, working memory efficiency and PCI in the same sample would be necessary to elucidate the relationship between these factors.

Interestingly, fifteen genes of the DRD2 Co-Expression Network (ACR, ALDH3A1, BTN3A1, CALHM3, CES3, DRD2, EFCAB6, GALNT10, GATAD2A, GLI1, HIST1H1E, HIST1H3G, IL31, RBM6, SLC28A1) are located within schizophrenia-associated loci in the latest Psychiatric Genetic Consortium investigation36. Notably, GATAD2A is among the genes resulting from the PGC3 prioritisation analysis due to its eQTL co-localisation profile36. Accordingly, it is considered a plausible causal gene for schizophrenia94. This gene codes for the protein GATA zinc finger domain containing 2A, a transcriptional repressor95, which is preferentially expressed during foetal brain development96. Its involvement in cell proliferation97 indicates a key role in development94. Furthermore, it has been implicated in schizophrenia through its involvement in the regulation of gene expression98,99. Consistently, it is upregulated in the hippocampus of patients with schizophrenia compared with healthy controls94.

Moreover, it has been recently demonstrated32 that the expression of genes of the DRD2 co-expression module can be regulated by NURR1, a transcription factor regulating genes involved in the dopaminergic system100. As D2R is a potent NURR1 activator101,102, it has been hypothesised that antipsychotics, through the blockade of the D2R, can impact the expression of NURR1, which in turn can regulate the transcription of the genes included in the DRD2 co-expression module32. The results of the present study—indicating a relationship between the DRD2 co-expression network and an established phenotype linked to schizophrenia such as PET-estimated dopamine synthesis capacity—are consistent with the hypothesis of the involvement of the genes of this module in the pathophysiology of schizophrenia and mechanisms underlying the response to antipsychotics.

Notably, the approach used in the present study is data-driven and the genes within the network are not pre-defined; thus, the mechanisms through which the proteins coded by the genes of this network interact with the dopaminergic pathway still need to be clarified. In fact, it needs to be understood how the DRD2 co-expression network influences striatal presynaptic dopamine synthesis capacity. In this context, it should be considered that post-synaptic D2 receptors play a role in the regulation of dopamine synthesis and release through inhibitory feedback loops103,104. It is also possible that the PCI reflects a different expression of the D2 autoreceptors, which regulate dopamine synthesis, although it should be noted that this score was developed analysing the expression of transcripts including exon 6, which is characteristic of the long isoform of D2R more often found post-synaptically105,106. Moreover, in view of the fact that the DRD2 co-expression gene-set indexed by the PCI is enriched for “negative regulation of dopamine secretion (GO:0033602)”35,39, preclinical studies are needed to test the hypothesis that the transcriptomic context of DRD2 influences dopamine presynaptic signalling.

Furthermore, in order to examine the potential involvement of this co-expression network in the regulation of expression and availability of the post-synaptic D2 receptors, it would be helpful to investigate the in vivo relationship between PCI and D2 receptor availability through studies using other PET tracers (e.g. 11C-raclopride).

The present study was conducted on healthy subjects; thus, the results were not influenced by medication or disease status. A key next step is thus to explore the effect of the DRD2-Polygenic Co-expression Index on dopamine function in disorders where involvement of the dopamine system has been demonstrated, such as psychosis, addiction, bipolar disorder54,107,108.

A potential limitation of the study is the use of data from three different PET scanners. However, the scanner was used as a covariate of no interest. Furthermore, we did not find a statistically significant association of PET scanner with Kicer; consistently, our recent investigation on the effect of the scanner in a similar dataset acquired from three different PET tomographs (Siemens Biograph 6 Hi-Rez, Siemens Biograph 6TruePoint, ECAT/EXACT3D) with an injected radioactivity below 200 MBq and acquisition time of 95 min did not reveal significant effects109.

Moreover, it should be considered that the DRD2 co-expression pathway, and therefore the PCI used in this study, was calculated by using post mortem mRNA from the frontal cortex35,68, thus it would be interesting to test if the DRD2 co-expression network remains the same in the striatum and the midbrain, where the dopamine neuron cell bodies are located.

Furthermore, in the present study, we did not examine dopamine function in the frontal cortex, due to lower 18F-DOPA signal reliability in frontal cortical regions when quantified without arterial blood input function76,110. Therefore, a study using PET tracers more suitable for the measurement of the cortical dopamine system59,111 would be helpful in understanding the relationships between PCI, striatal and cortical dopamine systems.

Conclusions

The results from the present study indicate that a polygenic score indexing a DRD2-related co-expression network is associated with striatal dopamine function measured in vivo with 18F-DOPA imaging. Our findings suggest that the same genetic variants associated with prefrontal inefficiency during working memory are also associated with greater estimated dopamine synthesis in the striatum. In view of the hypothesised link between striatal hyperdopaminergia and prefrontal hypodopaminergia59,112, it is tempting to observe that these variants originally found analysing the prefrontal cortex may have more widespread system-level correlates38.