Schizophrenia is characterized by positive symptoms (such as hallucinations and delusions), negative symptoms (such as blunt affect), cognitive impairment, and affective symptoms (1). Cognitive functions involving working memory—the ability to hold and manipulate pieces of information online for subsequent use in planning and acting—are particularly impaired in schizophrenia (2). Working memory is critically dependent on prefrontal cortical functioning (3). Patients suffering from schizophrenia show aberrant activation of the prefrontal cortex during working memory tasks, which suggests a role for the prefrontal cortex in the pathophysiology of cognitive deficits in schizophrenia (4). Animal and human studies have highlighted the significance of dopamine in regulating prefrontal cortical activity related to cognitive processing (5). Accordingly, the dopamine hypothesis of schizophrenia suggests that reduced dopamine function in the prefrontal cortex is responsible for the cognitive deficits encountered in patients with the disorder (6, 7).

At the receptor level, D₁ receptor function is essential for working memory function (8). It has been suggested that D₁ receptors in the prefrontal cortex, located mostly in apical dendrites and spines of pyramidal neurons (9), modulate the postsynaptic neuron’s response to incoming depolarizing, mostly glutamatergic, signaling (8). This modulation seems to be dependent on the state of the target neurons, and it seems to protect task-associated working memory representations from interfering and distracting stimuli by tuning the functional state of the target neurons (10). Too much or too little stimulation at the D₁ receptor can impair working memory, which suggests that the dependence of working memory on D₁ receptor stimulation may be described by an inverted U-shaped curve (11).

Schizophrenia is highly heritable (12), but the molecular basis of its heritability has remained largely unknown. The search for quantifiable mediators of this genetic vulnerability (intermediate phenotypes, or endophenotypes) could facilitate the exploration of the exact pathophysiological mechanisms behind the genetic vulnerability for schizophrenia (13). Recent evidence suggests that neuropsychological deficits (14), reduction of prefrontal cortical gray matter (15), and prefrontal cortical dopaminergic dysfunction (16) are associated with an increased genetic risk of schizophrenia. We recently demonstrated an increased density of dopamine D₂ receptors in the caudate nuclei of unaffected identical co-twins of patients with schizophrenia (17) that was similar in magnitude to the overall increase in density observed in a meta-analysis of studies of patients with schizophrenia (7). However, as yet, the role of D₁ receptors in the genetic etiology of schizophrenia has remained unaddressed.

We used positron emission tomography (PET) and the well-documented D₁ receptor tracer [¹¹C]SCH 23390 to examine the genetic and nongenetic contributions to D₁ receptor binding in schizophrenia by investigating twin pairs discordant for schizophrenia as well as healthy comparison twins. To explore the effects of genetic proximity to the patient, group comparisons were performed across three levels of genetic loading: comparison twins, dizygotic co-twins, and monozygotic co-twins. Monozygotic co-twins share 100% and dizygotic co-twins on average 50% of their segregating genes with the proband. All subjects also underwent comprehensive neuropsychological testing. Probands were included in the analyses, although they had experienced chronic antipsychotic drug treatment, which has been shown to down-regulate cortical D₁ receptors (18, 19).

The aim of the study was twofold: to investigate the effects of genetic liability to schizophrenia on D₁ receptor binding by comparing unaffected co-twins of patients with schizophrenia with matched healthy comparison twin pairs, and to examine disease- and treatment-specific contributions by comparing medicated patients with their own unaffected co-twins.

The study protocol adhered to the principles of the Helsinki Declaration and was reviewed and approved by the institutional review boards or ethical committees of the University of California Los Angeles, the University of Turku, the University of Pennsylvania, and the National Public Health Institute of Finland. All participants provided written informed consent after receiving a complete description of the study.

Sample Ascertainment and Clinical Evaluation

Using a procedure described previously, we drew subjects from a twin cohort consisting of virtually all same-sex twin pairs born in Finland from 1940–1957 (14, 17, 20). Nine discordant pairs were studied (four monozygotic and five dizygotic) along with two additional unaffected monozygotic co-twins and 13 twins from seven comparison twin pairs (four monozygotic and three dizygotic). Each co-twin was given a structured diagnostic interview.

Clinical patient data have been described previously (14). All patients except one had received chronic antipsychotic treatment with conventional neuroleptics. Use of clozapine or flupenthixol was an exclusion criterion since these drugs have low D₂/D₁ selectivity (21). The drugs used by these probands were haloperidol (N=4, dose range=1–8 mg/day), levomepromazine (N=1, dose=25 mg/day), sulpiride (N=1, dose=100 mg/day), thioridazine (N=4, dose range=50–400 mg/day), and chlorprothixene (N=1, dose=50 mg/day). In chlorpromazine equivalents, daily doses ranged from 0 mg to 800 mg. The probands underwent the PET studies despite evidence that antipsychotic drug treatment may down-regulate D₁ receptors in the nonhuman primate cortex (18, 19). The duration of illness ranged from 10 years to 36 years.

Exclusion criteria for unaffected co-twins and comparison twins have been described previously (14, 17). The study groups were matched for demographic variables. The zygosity of the studied twins was confirmed by genetic markers as described previously (14).

All subjects were tested with a comprehensive neuropsychological battery, from which a canonical liability variable was constructed to reflect performance on tests of spatial working memory, divided attention, intrusions during recall of a word list, and choice reaction times to visual targets (14). Deficits on these tests have been shown to be heritable in twins discordant for schizophrenia, increasing in severity with increasing genetic proximity to the proband. Spatial working memory, as measured by the visual span subtest of the Wechsler Memory Scale—Revised, was selected from among these tests to be individually correlated with D₁ receptor binding. Findings on the neuropsychological data in the original twin sample have been previously published (14).

PET Studies

For the PET studies, a whole-body three-dimensional PET scanner (GE Advance; GE, Milwaukee) was used to acquire 35 slices of 4.25 mm thickness covering the whole brain. The scanning procedure and preparation of [¹¹C]SCH 23390 have been described in detail elsewhere (22). Briefly, an intravenous rapid bolus of 174–268 MBq of [¹¹C]SCH 23390 (mean=233.01 MBq, SD=24.4) was given to each subject (specific radioactivity, mean=63.5 MBq/nmol [SD=18.3]; amount of radiotracer, mean=1.14 μg [SD=0.31]). No statistically significant differences were observed between study groups in any of these parameters.

Quantification of [¹¹C]SCH 23390 Binding

For the calculation of regional time-activity curves, regions of interest were manually delineated using Imadeus software (version 1.15, Forima, Inc., Turku, Finland) on the putamen, caudate, angular gyrus, anterior cingulate cortex, cerebellum, dorsolateral prefrontal cortex, insular cortex, medial prefrontal cortex, orbitofrontal cortex, posterior cingulate cortex, supramarginal gyrus, subgenual anterior cingulate cortex, and superior, middle, and inferior temporal gyri, as previously described (22). There were no statistically significant differences between groups in region-of-interest size, except in the insular cortex, which was smaller in probands compared with other subjects (–5.0 to –7.1%, p<0.05), and the putamen, which was smaller in probands compared with monozygotic co-twins (–10.3%, p<0.05). The simplified reference tissue model (23) with cerebellum as a reference region was applied to the time-activity data of each region of interest to calculate binding potential values (f₂B_max/K_D), which represent the product of receptor density (B_max), apparent affinity (1/K_D), and the free fraction of the nondisplaceable tissue compartment (f₂). This method has been previously validated for [ ¹¹C]SCH 23390 (22, 23).

Statistical Analyses

SAS version 8.2 (SAS Institute, Cary, N.C.) was used for data analysis. The general linear mixed model with repeated measures was used, correcting for dependency (i.e., correlation) among the healthy co-twins by treating twin pair as a random variable and adjusting the model error terms accordingly (Satterthwaite option). Differences in mean binding potential between groups were tested by modeling risk group (probands, unaffected monozygotic co-twins, unaffected dizygotic co-twins, healthy twin pairs) as a fixed-effect predictor, while covarying for sex and age. To test for laterality, hemisphere was entered into the model as a within-subject repeated-measures factor, and a group-by-hemisphere interaction was entered into the model to test for differences in laterality between the groups. We did not test for sex-by-group interaction because of the small number of one sex or the other in some groups. Whenever one of these terms contributed significantly to the prediction of [ ¹¹C]SCH 23390 binding potential, contrast analyses were performed comparing conditions within the term collapsing over nonsignificant terms in the model. This approach maintains the hypothesis-wise type I error rate at 0.05 because a predictor’s contribution to particular dependent measures is evaluated only if its effect is found to vary at the multivariate level. The significance of each predictor was tested while accounting for all other model terms simultaneously. Since the SAS Proc Mixed procedure did not allow for the use of multiple repeated-measures factors and a random factor within the same model, a separate model was used for each region of interest. Because of the large number of brain regions studied, we applied a Bonferroni-corrected p value of 0.0033 (0.05/15) to control for multiple comparisons for the main effect of group. The relationships between D₁ binding potential and neuropsychological test scores were examined using mixed-model regression analysis predicting the neuropsychological measurement with D₁ binding potential, while covarying for age and sex and correcting for dependency (i.e., correlation) among the healthy co-twins, as we did in the linear mixed model. The relationship between D₁ binding potential and antipsychotic drug dose in the probands was tested using partial correlation, partialing out the effects of age and sex. Intraclass correlations (ICCs) for healthy comparison monozygotic twin pairs were calculated for the binding potential in each of the total regions. We did not calculate ICCs for the two healthy comparison dizygotic twin pairs. Differences between groups in demographic characteristics were tested using analysis of variance for continuous data and the chi-square test for nominal data with a significance level of p<0.05, correcting for dependency among twin pairs.

Confirmatory Voxel-Based Statistical Analyses

An independent confirmatory voxel-based receptor mapping analysis was performed as described previously (17). Briefly, parametric binding potential images were calculated using receptor parametric mapping based on the simplified reference tissue model (24). Preprocessing and statistical analysis of parametric images were performed using SPM99 (25) and Matlab 6.5 for Windows (MathWorks, Natick, Mass.). Parametric binding potential images were spatially normalized using summed PET images and a ligand-specific template for [ ¹¹C]SCH 23390 and were smoothed using a 12-mm Gaussian filter. To explore the effects of genetic load on [ ¹¹C]SCH 23390 binding potential, a multiple regression analysis with group indicator variables was performed. The effects of age and sex were eliminated by treating them as nuisance covariates. The search volume of the analysis was restricted to the same regions as the region-of-interest analysis using small-volume correction. A p value of 0.05, corrected for multiple comparisons, was considered the criterion for significance in the voxel-based analyses.

Table 1 summarizes the demographic and clinical characteristics of the study groups, and Table 2 lists the [¹¹C]SCH 23390 binding potential values for each group. Significant group effects were found in various brain regions: the anterior cingulate cortex, angular gyrus, caudate, dorsolateral prefrontal cortex, medial temporal gyrus, superior temporal gyrus, insular cortex, medial frontal cortex, orbitofrontal cortex, putamen, and supramarginal gyrus (all p<0.0033; statistics available on request). We then looked at individual brain regions to see if unaffected co-twins differed from healthy comparison twins. Monozygotic co-twins had higher D₁ binding potential values compared with healthy comparison twins in the medial prefrontal cortex (+35%, F=8.74, df=1, 20.9, p=0.008), superior temporal gyrus (+38%, F=7.51, df=1, 18.2, p=0.013), and angular gyrus (+22%, F=5.90, df=1, 19.1, p=0.025). The independent voxel-based receptor parametric mapping analysis with correction for multiple testing confirmed these findings and showed significant clusters in the inferior parietal and lateral temporal lobes as well as the medial prefrontal cortex (Figure 1). In unaffected co-twins, canonical liability scores were negatively correlated with D₁ binding potential in the putamen (N=9, R=–0.68, p=0.042), whereas spatial working memory performance was negatively associated with D₁ binding potential in the superior temporal gyrus (N=9, R=–0.48, p=0.045), suggesting that high D₁ binding potential was associated with poorer cognitive performance in these regions.

Patients with schizophrenia showed a significant widespread reduction in D₁ binding potential compared with their unaffected co-twins in almost all studied brain regions (see Table 2; statistics available on request). No associations were observed between antipsychotic drug dose (in chlorpromazine equivalents) and D₁ binding potential in individual regions of interest, but a significant negative correlation was observed between drug dose and average D₁ binding potential of the frontal and temporal regions (partial R=–0.86, p=0.012, controlling for the effects of age and sex), indicating that high drug doses were associated with low D₁ binding potential in these regions.

A significant age effect was noted in the whole sample, with D₁ binding potential tending to decline with age in the angular gyrus (F=5.99, df=1, 13.1, p=0.029), caudate (F=7.49, df=1, 13.1, p=0.017), insular cortex (F=4.66, df=1, 14.7, p=0.048), putamen (F=5.20, df=1, 14.6, p=0.038), and supramarginal gyrus (F=6.79, df=1, 14.1, p=0.021). We observed a significant sex difference in the whole sample, with women having higher binding potential values than men in the medial prefrontal cortex (F=7.43, df=1, 12.2, p=0.018) and anterior cingulate cortex (F=5.75, df=1, 10.4, p=0.037). A significant hemisphere effect was observed in the inferior temporal gyrus (F=14.31, df=1, 40.3, p<0.001) and the middle temporal gyrus (F=7.28, df=1, 40.7, p=0.010). Left hemisphere values were higher than right hemisphere values in these regions. No significant group-by-hemisphere interactions were seen.

Moderate to high ICCs for binding potential in the healthy monozygotic twins were noted for both striatal and cortical regions, namely, the inferior temporal gyrus (0.84), putamen (0.80), middle temporal gyrus (0.80), caudate (0.71), supramarginal gyrus (0.66), and superior temporal gyrus (0.63). For other regions studied, ICCs were below 0.50.

In this study, we used PET with [¹¹C]SCH 23390 to investigate brain D₁ receptors in monozygotic and dizygotic twin pairs discordant for schizophrenia. The study produced two major findings. First, unaffected monozygotic co-twins of patients with schizophrenia showed increased D₁ receptor binding in the medial prefrontal cortex, superior temporal gyrus, and angular gyrus compared with healthy comparison twins. Monozygotic co-twins displayed the highest values compared with the comparison twins, while dizygotic co-twins had values intermediate between monozygotic co-twins and comparison twins, which suggests an association with increasing genetic loading for schizophrenia. To our knowledge, this is the first report of D₁ receptor alterations related to genetic risk for schizophrenia. The other major finding is that patients with schizophrenia who had experienced chronic antipsychotic treatment (probands in the discordant pairs) had a widespread reduction of D₁ receptor binding in the brain, which was associated with antipsychotic medication dose, a finding consistent with previous preclinical work (18, 19).

Binding potential represents a combined estimate of receptor density and apparent affinity, and the latter is affected by both the propensity of the radioligand to bind to the receptor and competition with endogenous dopamine for binding to the receptor. Acute fluctuations of endogenous dopamine do not change the in vivo binding of [ ¹¹C]SCH 23390 in the striatum or cortex in humans (26) or monkeys (27, 28), and we interpret the results of our study as reflecting altered D₁ receptor density. It seems unlikely that the up-regulation of D₁ receptors in unaffected monozygotic co-twins observed in this study would be mediated by primary variations of the gene coding for the D₁ receptor. It is currently unknown whether some polymorphism in the D₁ receptor gene might affect receptor binding in vivo or even carry a risk for schizophrenia. SCH 23390 shows non-negligible in vitro affinity for serotonin 5-HT₂ receptors, and there is preliminary evidence from a recent nonhuman primate study (29) that a small proportion of the cortical specific in vivo binding of both [ ¹¹C]SCH 23390 (14%) and [ ¹¹C]NNC 112 (28%) may originate from 5-HT_2A receptors. Human data are currently lacking for the in vivo pharmacological selectivity of these radioligands, but we cannot exclude the possibility that our findings may be due, to some extent, to hypothesized group differences in 5-HT_2A receptor density.

A novel finding of our study is the association of increased D₁ receptor density with increasing genetic risk for schizophrenia in the medial prefrontal cortex, superior temporal gyrus, and angular gyrus. This finding was confirmed with an independent voxel-based receptor mapping analysis. The interpretation of our finding of D₁ receptor binding elevation as an intermediate phenotype for schizophrenia is hindered by the fact that previous PET studies of D₁ receptor binding have produced contradictory findings, with studies showing decreased (30), unaltered (31), and increased (32) binding in unmedicated patients compared with comparison subjects. These studies used different radioligands for D₁ receptors, namely [ ¹¹C]SCH 23390 (30, 31) and [ ¹¹C]NNC 112 (32). The radioligand used may be important, since it has been shown that the cortical binding of [ ³H]SCH 23390 and [ ¹¹C]NNC 112 behave differently following chronic dopamine depletion in rodents: [ ¹¹C]NNC 112 binding increases, whereas [ ³H]SCH 23390 binding remains unchanged, a difference hypothesized to relate to differential affinity for internalized versus externalized receptors (33). Clearly, more studies are needed to validate these assumptions and to shed light on the discrepant findings in subjects with unmedicated schizophrenia.

Our second important finding was that patients with schizophrenia who had chronic antipsychotic treatment showed a profound and widespread reduction in D₁ receptor binding throughout the brain, including regions such as the striatum as well as frontal, temporal, and parietal cortices. Studies of nonhuman primates have shown that chronic administration of D₂ receptor antagonist medications down-regulates D₁ receptor binding in the frontal and temporal association areas but not in motor, visual, or somatosensory cortex or striatum (18), as well as levels of D₁ and D₅ receptor mRNAs in the prefrontal cortex but not in the striatum (19). Consistent with this preclinical data, we found that high doses of antipsychotic medication were associated with low frontal and temporal D₁ receptor binding potential in the probands. Some of the antipsychotic drugs used by the patients in our study may also directly occupy D₁ receptors and thereby cause a reduction in [ ¹¹C]SCH 23390 binding potential. Although the use of clozapine and flupenthixol was excluded, other drugs (such as thioridazine, which has only around a 10-fold D₂/D₁ selectivity and was used by four patients) might contribute to a reduction in binding potential as well. However, we regard direct occupancy unlikely because a previous clinical PET study with medicated patients indicated that a daily dose of 300 mg of thioridazine did not induce detectable D₁ occupancy as assessed with [ ¹¹C]SCH 23390 (21), and only one patient in our study had a daily dose over 300 mg. Because of this potential confounder, our results on our patient group are not directly comparable with previous D₁ receptor PET studies of unmedicated schizophrenia (30–32) and are most likely reflective of D₁ receptor down-regulation, at least in the frontal and temporal cortices, induced by antipsychotic medication.

Dopaminergic hypofunction in the prefrontal cortex has long been regarded as one of the key neural substrates for deficit aspects of schizophrenia (6), and a postmortem study has provided direct evidence for abnormal dopaminergic innervation of the medial prefrontal cortex in schizophrenia (34). The increased D₁ receptor density observed in our study might be viewed as a compensatory reaction to decreased cortical dopamine levels, which has been suggested to confer genetic liability for schizophrenia via polymorphisms in the gene coding for catechol O-methyltransferase (16). Another hypothesis is that increased D₁ receptor binding is a compensatory phenomenon secondary to reduced neuropil (35)—dendritic processes where D₁ receptors are preferentially located (9)—which may be responsible for genetic vulnerability-related gray matter loss (15). The functional significance of D₁ receptors in the superior temporal gyrus and inferior parietal areas is currently unknown. Neurons in these regions express D₁ receptors (36), and they participate in working memory functions in concert with neurons of prefrontal cortex (37, 38), suggesting that they may be relevant for the pathophysiological process of schizophrenia. Consistent with the notion of marked inherited contributions, we observed high ICCs for striatal and parietotemporal D₁ receptor densities in the healthy monozygotic twins.

Sex differences in D₁ receptor density in the anterior cingulate cortex and medial prefrontal cortex are novel findings: women had higher receptor densities than men. Women have also been shown to have higher D₂ receptor binding in the anterior cingulate cortex (39). The significance of the sex difference in D₁ receptor density remains to be established in future studies. The observed significant lateralization in D₁ binding potential in the lateral temporal cortex is in line with the well-documented structural and functional asymmetry in these regions that has been attributed to the specialization of the left hemisphere to language functions (40).

One limitation of this study is the relatively small sample size. Nevertheless, the effect sizes for the group contrasts of interest were of sufficient magnitude to detect significant differences in cortical regions of theoretical importance in the pathophysiology of schizophrenia. It is important to emphasize, however, that with larger samples, similar effects may be detected in other brain regions.