INTRODUCTION

The central dopaminergic system has been implicated in the modulation of emotional behavior, the pathophysiology of depression, and the mechanisms of antidepressant drugs, by clinical and preclinical evidence (Drevets et al, 1999; Nestler and Carlezon, 2006; Nutt et al, 2006). In experimental animals the dopaminergic projections from the ventral tegmental area (VTA) to the nucleus accumbens shell and medial prefrontal cortex (PFC) were important in learning associations between operant behaviors or sensory stimuli and reward, and in mediating the reinforcing properties of drugs of abuse and natural rewards (Wise and Rompre, 1989; Schultz, 1997). These observations lead to the hypothesis that reduced mesocorticolimbic DA function underlies the anhedonia, amotivation, and psychomotor slowing associated with major depression (Swerdlow and Koob, 1987; Fibiger, 1991; Nestler and Carlezon, 2006).

A variety of experimental data support this hypothesis. Reductions in dopaminergic function associated with α-methyl-para-tyrosine administration can induce depressive symptoms in susceptible individuals (Bremner et al, 2003; Hasler et al, 2008). Conversely, dopamine receptor agonists (eg pramipexole) exert antidepressant effects in placebo-controlled studies (Willner, 2000; Zarate et al, 2004). In MDD subjects DA turnover appears abnormally decreased, as concentrations of the DA metabolite, homovanillic acid (HVA), consistently are reduced in the cerebrospinal fluid (CSF) and jugular vein plasma in vivo (Lambert et al, 2000; Willner, 2000)—particularly in depressives who manifest psychomotor retardation or melancholic features (Asberg et al, 1984) and in the caudate and accumbens post mortem in suicide victims (Bowden et al, 1997). Neuroimaging studies of MDD showed reduced [11C]-L-DOPA uptake across the blood–brain barrier (Agren and Reibring, 1994) and increased striatal binding to D2/D3 receptor radioligands that were sensitive to endogenous DA concentrations, although these latter findings were limited to cases who showed psychomotor slowing (Ebert et al, 1996; Drevets et al, 2005). In such cases the elevated D2/D3 receptor binding may have reflected either reduced intrasynaptic DA concentrations, or compensatory upregulation of D2/D3 receptor density or affinity (Todd et al, 1996; Laruelle and Huang, 2001). Nevertheless, studies whose samples were not predominantly composed of psychomotor slowed cases found no difference in D2/D3 receptor levels during depression (Klimke et al, 1999; Parsey et al, 2001; Montgomery et al, 2007; Hirvonen et al, 2008). Similarly, some (Meyer et al, 2001) but not other (Brunswick et al, 2003; Argyelan et al, 2005; Yang et al, 2008) studies of striatal DA transporter (DAT) binding reported reduced availability in MDD subjects vs controls.

The specific DA receptor subtypes that mediate dopaminergic function in reward processing, emotional behavior, and depression remain incompletely understood, partly due to the paucity of highly selective agonists and antagonists. In mice phenotypic analysis of DA receptor knockouts identified roles for the D1, D2, and D3 receptor subtypes in mediating dopamine's effects on reward processing and/or emotional behavior. A complex role for D1 receptors in particular was supported by both preclinical and clinical evidence. In genetically engineered mice deletion of the D1 receptor attenuated the reinforcing properties of rewarding stimuli (reviewed in Holmes et al, 2004). Nevertheless, the euphoric effects of cocaine appeared blunted by D1 receptor-like antagonist administration in cocaine addicts (Romach et al, 1999; Waddington et al, 2001; Holmes et al, 2004). Moreover, in rats the reduction in sucrose consumption resulting from chronic mild stress, a purported model of anhedonia, was associated with increased D1 receptor density in the caudate–putamen (but not the accumbens or amygdala) (Papp et al, 1994), and in humans with schizophrenia, D1 receptor antagonists alleviated ‘negative’ symptoms such as anhedonia and amotivation (Den Boer et al, 1995; Karle et al, 1995).

With respect to other emotional states, in rats intra-amygdaloid injection of D1 receptor antagonists exerted anxiolytic effects (de la Mora et al, 2005) and impaired retention of inhibitory avoidance learning (fear-based memory) (Lalumiere et al, 2004). Moreover, D1 receptor knockout mice showed deficits in fear extinction and reversal learning (putative correlates of resilience to stress or adaptation to behavioral reinforcement, respectively), and abnormal long-term potentiation of synapses on PFC neurons (El-Ghundi et al, 1999, 2001; Huang et al, 2004). These data appeared consistent with evidence that an optimal range of D1 receptor stimulation is required to facilitate working memory, as intracortical administration of either agonists or antagonists for D1 receptors impairs working memory performance in rodents and primates (Arnsten et al, 1994; Williams and Goldman-Rakic, 1995; Granon et al, 2000; Robbins, 2000).

In MDD the D1 receptor has been assessed by genetic, post mortem and neuroimaging studies. Although polymorphisms in the human D1 receptor gene have not been associated specifically with the vulnerability for developing MDD (Koks et al, 2006), one study (Severino et al, 2005) (but not another (Kato, 2007)) found an association with bipolar disorder (BD). Post mortem studies of D1 receptor binding found no difference in the striatum (Bowden et al, 1997) or right amygdala (Klimek et al, 2002) in MDD subjects who were unmedicated at the time of death (n=28 and 11, respectively) vs controls, although one of these found increased D1 receptor density and reduced affinity in the accumbens in medicated suicide victims (Bowden et al., 1997). Finally, a PET-[11C]SCH-23390 study reported reduced D1 receptor binding in the striatum bilaterally in 10 depressed MDD subjects with anger attacks (Dougherty et al, 2006). It remained unclear whether this finding would generalize to other MDD subtypes. In subjects with BD (n=10), a PET-[11C]SCH-23390 study reported no difference in striatal D1 receptor binding relative to controls, although only three of the subjects were depressed (Suhara et al, 1992). The latter study also reported reduced binding in the frontal cortex of BD subjects, but because one-half of [11C]SCH-23390 binding in the frontal cortex is attributable to 5-HT2A receptor binding, this difference remained difficult to interpret (Ekelund et al, 2006).

The recent development of a PET radioligand using (+)-5-(7-benzofuranyl)-8-chloro-7-hydroxy-3-methyl-2,3,4,5-tetrahydro-1H-3-benzazepine (NNC-112), a potent and selective D1 receptor antagonist, provided sufficiently high specific-to-nonspecific binding ratios to permit meaningful assessment of D1 receptor binding in both extrastriatal and striatal tissues (Andersen et al, 1992; Halldin et al, 1998). The current study is the first application of [11C]NNC-112 in investigations of D1 receptor binding in mood disorders, and the first in vivo study of D1 receptor binding in depressed subjects selected more generally according to the criteria for MDD. We hypothesized that the mean D1 receptor BPND would be reduced in the striatum of MDD subjects vs controls, based on the data reported in MDD subjects with anger attacks (Dougherty et al, 2006). In addition, we sought to more specifically localize D1 receptor binding abnormalities in MDD using an MRI-based technique previously developed for characterizing the dopaminergic system in striatal subregions (Drevets et al, 1999, 2001).

PATIENTS AND METHODS

Participants

Right-handed volunteers 18–55 years of age, who either met DSM-IV criteria for recurrent or chronic MDD in a current major depressive episode (n=18), or had no personal history of a major psychiatric disorder (n=19) were studied (Table 1). Diagnosis was established by an unstructured interview with a psychiatrist and the Structured Clinical Interview for DSM-IV (APA, 1994). Exclusion criteria for all subjects included a history of psychosis, exposure to psychotropic drugs, cigarette smoking, or any medication likely to affect cerebral function within the 3 weeks before scanning (eight for fluoxetine), major medical or neurological illnesses, lifetime history of substance dependence, substance abuse within 1 year, recent suicidal behavior or serious suicidal ideation, current pregnancy, peri- or postmenopause status. Depressed subjects with secondary anxiety disorders were not excluded. Additional exclusion criteria for the controls included having a first-degree relative with a mood, anxiety, or psychotic disorder. Subjects provided written informed consent as approved by the NIMH IRB.

Table 1 Demographic and Clinical Characterization of the Study Samples

Data Acquisition and Processing

The severity of depressive symptoms was rated using the Montgomery–Asberg Depression Rating Scale (MADRS) (Montgomery and Asberg, 1979) and the Inventory of Depressive Symptoms-Clinician Version (IDS-C). Anxiety symptoms were rated using the Hamilton Anxiety Rating Scale (HAM-A (Hamilton, 1959)), and anhedonia was assessed using the anhedonia subscale of the IDS-C. Psychomotor speed was assessed using the reaction time measured during performance of the Rapid Visual Information Processing (RVIP) task from the Cambridge Neuropsychological Test Automated Battery (CANTAB; Cambridge Cognition Ltd, Cambridge, UK). These computerized tasks were presented on an Advantech computer (Model PP-120T-RT) with a 10.5 in. touch screen monitor.

PET scans were acquired using a GE-Advance scanner in 3D mode (reconstructed 3D spatial-resolution=6 mm full-width at half-maximum (FWHM)). The PET data were reconstructed using a Hanning filter and Gaussian fit scatter correction method. An 8-min transmission scan was acquired using rotating rods of 68Ge/68Ga, and used to perform measured attenuation correction of the emission image. [11C]NNC-112 was synthesized as described previously (Halldin et al, 1998). Following intravenous bolus administration of a mean injected dose=19.5±1.24 mCi (range 15.5–20.8) of high specific activity [11C]NNC-112, a 90-min dynamic emission scan was acquired as 27 frames of increasing length (number × frame duration (min): 6 × 0.50; 3 × 1.0; 2 × 2.0; 16 × 5.0).

Head motion was minimized during scanning by stabilizing each subject's head position using a thermoplastic mask fixed to the scanner table. In addition, the PET data were corrected for head motion by aligning all frames to a mean image of frames 3 through 8, using AIR (Woods et al, 1992) as implemented in Medx (Sensor Systems Inc., Sterling, VA). To provide an anatomical framework for the PET analysis, MRI scans were acquired using a GE 1.5 or 3.0 T scanner and T1-weighted pulse sequence optimized to enhance tissue-contrast resolution (voxel size=0.86 × 0.86 × 1.2 mm). The realigned PET frames were coregistered to the anatomical MRI for each subject using fMRIB's Linear Image Registration Tool and a mutual information cost function (Jenkinson and Smith, 2001).

Data Analysis

The D1 receptor binding parameter estimations were performed voxel-wise using a multilinear reference-tissue model (MRTM) using a receptor-free reference region (cerebellum) without arterial data. The BPND values thus obtained were proportional to the receptor density (Bmax) and independent of blood flow (K1) (Ichise et al, 2003). The value of k2′ was estimated by the three-parameter MRTM using regions-of-interest (ROI) time–activity curves for the striatum. The R1 images, which reflect relative blood flow, were used for image co-registration with MRI. The generation of parametric BPND images from the dynamic [11C]NNC-112 PET data was performed in PMOD 2.5 (Mikolajczyk et al, 1998).

Hypothesis testing was performed in five striatal ROI defined over the anteroventral striatum (AVS; accumbens area, ventromedial caudate, anteroventral putamen), dorsal caudate (DC), middle caudate (MC), ventral putamen (VP), dorsal putamen (DP), as described previously (Bowden et al, 1997; Drevets et al, 2001). On the basis of the evidence that DA release during reward processing is lateralized in healthy humans (Martin-Soelch et al, 2007; Tomer et al, 2008), hemisphere was included as a factor in the analyses. To allow comparison with previously reported negative findings from post mortem studies of depression (Klimek et al, 2002), we examined binding in the amygdala post hoc. Additional exploratory analyses were conducted post hoc in cortical regions known to contain moderately high D1 receptor concentrations and to be implicated in the mesocorticolimbic dopaminergic system and the pathophysiology of MDD (Willner, 1995): right and left insula/lateral orbitofrontal cortex (ie sulcal BA47) (Ongur et al, 2003) and anterior cingulate cortex (ACC; including subgenual and pregenual areas). Amygdala, insular and ACC ROI were defined as described in (Cannon et al, 2007).

The reference region was defined in the cerebellar cortex, with the ROI situated at least one FWHM ventral to the occipital and temporal cortices, and away from the cerebellar vermis and brain edge.

Statistical Analysis

The mean BPND for each ROI was compared between groups using a linear mixed model and the heterogeneous compound symmetry covariance matrix performed using SPSS. For ROI where the model indicated significant group effects, post hoc comparisons of BPND across groups were performed using unpaired t-tests. In regions where group effects were detected the relationship between D1 receptor BPND and clinical ratings of depression, anxiety, and anhedonia were assessed post hoc by computing Spearman's bivariate correlation coefficients (rho, ρ).

RESULTS

The groups did not differ significantly for mean age or gender composition (Table 1). Mean depression, anxiety, and anhedonia ratings were greater (p<0.001, t=−4 to −17) in MDD subjects than controls (Table 1). Reaction times on the RVIP did not differ significantly between the MDD and control groups (t=−1.50, p=0.15). Performance measured by the number of omission (HC 4.5±2.7, MDD 6.9±5.5, t=−1.64, p=0.12) or commission errors (HC 0.33±1.29, MDD 1.35±1.83, t=−1.794, p=0.083) on the RVIP also did not differ significantly between groups. Five MDD subjects had comorbid anxiety disorders (generalized social phobia (n=4), specific phobia (n=1)). Two MDD subjects had a remote history of alcohol abuse and one a remote history of marijuana abuse.

The rank order of the [11C]NNC BPND values was caudate, putamen>AVS>amygdala>insula>ACC>cerebellum (Table 2), consistent with the relative D1 receptor densities measured post mortem in humans (Cortes et al, 1989).

Table 2 The Mean and Standard Deviation of Regional [11C]NNC-112 Binding Potential Values

The mean D1 receptor BPND was 14% lower in the left MC of the MDD subjects vs controls (t=2.21, p=0.03; Figures 1 and 2; Table 2). The magnitude of this difference was similar after removing the 3 subjects with a remote history of substance abuse (12.6%), the 7 previously medicated MDD subjects (13.5%), or the 5 subjects with co-morbid phobias (11%). In the remaining ROIs mean D1 receptor BPND did not differ between the control and MDD groups (Table 2).

Figure 1
figure 1

Dopamine D1 receptor binding potential in healthy control and major depressive disorder (MDD) groups in the left and right hemispheres for the striatal regions-of-interest. *p<0.05: D1 receptor binding was significantly lower in the left middle caudate (MC) in the MDD group relative to control group. **p<0.05, ***p<0.001: the main effect of hemisphere on binding potential (BPND) was significant in the anteroventral striatum (AVS), ventral putamen (VP), dorsal putamen (DP), middle caudate (MC), and dorsal caudate (DC) regions-of-interest.

Figure 2
figure 2

Representative parametric dopamine D1 receptor binding potential (BPND) images through the striatum in a control subject (left panel) and an major depressive disorder (MDD) subject (right panel). Within each panel the upper row of images show axial (left) and coronal (right) sections through the anatomical MRI, on which the regions-of-interest (ROI) were defined (Drevets et al, 2001), and the lower row of images show parametric images of the nondisplaceable component of the D1 receptor BPND modeled from PET emission images. The MRI and PET images for each subject are coregistered. The axial images show the plane containing both the anterior and posterior commissures. The orthogonal lines locate approximately the center of the caudate head on this bicommissural plane. The coronal sections pass through the anterior striatum, which at the level shown is composed predominantly of the caudate head. The middle caudate region is evident in the coronal sections, where it is situated immediately above the bicommissural plane (marked by the horizontal line). The D1 receptor BPND consistently was higher in the right caudate than the left in both groups (Figures 1 and 3), as evinced in these representative images. The mean BPND value in the left middle caudate of the MDD group was lower than in the control group (Figure 1).

In post hoc assessments, a main effect of hemisphere was evident in several regions (Figures 1 and 3), including AVS (F=87.8, p=8.01e−11), VP (F=8.84, p=3.10e−2), DP (F=25.5, p=1.56e−5), MC (F=308, p=1.09e−5), and DC (F=98.5, p=1.77e−4). A group-by-gender-by-hemisphere interaction (F=7.33, p=0.011) was detected in the DP, which was accounted for by a greater interhemispheric difference in BPND in female controls vs female MDD subjects (Figure 4). This interaction remained significant after including age as a covariate, or after removing from analysis the subjects who had a remote history of substance abuse (F=4.44, p=0.044), previous psychotropic medication exposure (F=6.17, p=0.020) or comorbid phobia (F=5.83, p=0.023).

Figure 3
figure 3

Dopamine D1 receptor binding potential in the left and right hemispheres for each striatal region-of-interest showing the consistency of laterality effects. The left- and right-sided values corresponding to a single participant are connected by a line. Main effects of hemisphere on regional binding potentials were significant in all five regions-of-interest (see Figure 1 and ‘Results’ section for significance levels).

Figure 4
figure 4

Mean (±SEM) binding potential values for the right and left dorsal putamen separated by sex and diagnosis. In women with major depressive disorder (MDD) the difference in D1 receptor binding potential (BPND) between the right and left hemispheres of the dorsal putamen was significantly less than that of the female controls. In the dorsal putamen the group-by-gender-by-hemisphere interaction was significant (F=7.33, p=0.011). Abbreviations: HC, healthy control; MDD, major depressive disorder.

Age correlated with BPND in the left VP (ρ=−0.34, p=0.04), left DP (ρ=−0.33, p=0.04), right MC (ρ=−0.38, p=0.02), bilateral DC (left: ρ=−0.39, p=0.02; right: ρ=−0.38 p=0.02), and left insula (ρ=−0.42, p=0.01). The relationship between BPND and age showed a trend toward significance in right VP (ρ=−0.32, p=0.06), left MC (ρ=−0.29, p=0.08), and left insula (ρ=−0.32, p=0.052) but did not approach significance (p>0.10) in the right DP, AVS, ACC, or amygdala. Analyses were covaried for age where relevant.

A main effect of gender on BPND was significant in the AVS (F=4.21, p=0.048), VP (F=7.28, p=0.043), MC (F=9.72, p=0.026), and DC (F=9.77, p=0.026) but not in the DP (F=1.38, p=0.249). There were no significant group-by-gender interactions in any region (p>0.1).

Exploratory, post hoc clinical correlations showed that BPND values in the left MC correlated inversely with illness duration (r=−0.53, p=0.02; Figure 5), but did not correlate significantly with depression (MADRS), anxiety (HAM-A), or anhedonia (IDS-C anhedonia subscale) severity (ie p>0.05). Reaction times on the RVIP correlated negatively with D1 receptor binding in the left MC in controls (r=−0.453, p=0.007) but not MDD subjects (r=−0.096, p=0.358). The ratio of left-to-right BPND values in the MC correlated inversely with IDS-C anhedonia subscale ratings (r=−0.65, p=0.0040) but not with depression or anxiety ratings (p>0.4). MDD subjects with comorbid phobias did not differ in D1 receptor BPND in the left MC than nonphobic depressives.

Figure 5
figure 5

Relationship between dopamine D1 receptor binding potential in the left middle caudate and illness duration in the major depressive disorder (MDD) sample. The D1 receptor binding potential in the left middle caudate correlated inversely (r=−0.531, p=0.02) with illness duration (calculated as time since onset of illness).

Depression severity correlated negatively with BPND in right DP (MADRS ρ=−0.61, p=0.007; Figure 6). Anxiety severity (HAM-A) correlated negatively with BPND in right DP (ρ=−0.48, p=0.045). Anhedonia ratings, age at onset, and illness duration did not correlate significantly with BPND in the DP (p>0.05). In right DP, MDD subjects with comorbid phobias showed lower D1 receptor BPND than nonphobic depressives (Z=−2.3, p=0.019).

Figure 6
figure 6

Relationships between age and D1 receptor binding potential (BPND) in the middle caudate and ventral putamen for all subjects. Age correlated negatively with [11C]NNC-112 BPND in the middle caudate, ventral putamen, and most of other striatal regions examined (ρ=−0.33 to −0.42, p=0.01–0.04; see text).

Mean BPND values for depressed subjects who previously were exposed to psychotropic medications (n=7) showed no significant difference vs drug–naïve subjects (n=11) in any region (p>0.10).

DISCUSSION

The mean D1 receptor binding was 14% lower in the left MC of MDD subjects vs healthy controls (Figure 1). Moreover, although significant differences in D1 receptor binding between the right and left hemispheres existed in controls, this normal asymmetry was absent in the DP in depressed women (Figure 4). Post hoc assessments in the MDD sample showed that D1 receptor binding in the left MC correlated inversely with illness duration and anhedonia ratings. In addition, the left MC D1 receptor binding correlated inversely with psychomotor speed in the controls but not the depressives.

The abnormalities in striatal BPND in MDD likely reflect alterations in D1 receptor density and/or affinity for [11C]NNC-112. Although [11C]NNC-112 binding is reversible in vivo (Abi-Dargham et al, 2000), it is insensitive to displacement by endogenous dopamine (Chou et al, 1999). Among DA receptors NNC-112 is highly selective for D1 receptors (eg the D1 : D2 receptor binding ratio=4500) (Ekelund et al, 2007). The major limitation in specificity is that [11C]NNC-112 is only 2- to 3-fold more selective for D1 receptors than 5-HT2A receptors (Ekelund et al, 2007). Thus in the frontal cortex, where the 5-HT2A receptor and D1 receptor concentrations are comparable, up to 25% of [11C]NNC-112 binding is attributable to 5-HT2A receptor binding (Ekelund et al, 2006; Slifstein et al, 2007). In the striatum, however, where the ratio of D1 : 5-HT2A receptor density is relatively higher, the 5-HT2A contribution to the measured BPND is negligible (Slifstein et al, 2007). Thus, the abnormalities observed in the MC and DP in MDD most likely reflect differences in D1 receptor binding.

The mean 14% decrement in D1 receptor BPND in the left MC in our MDD subjects appeared consistent with the 13% reduction in binding in the left striatum reported by Dougherty et al (2006) in MDD subjects with anger attacks. Our data extended this previous finding by showing that the left striatal D1 receptor binding was also decreased in a depressed sample selected more generally according to MDD criteria. We did not observe any abnormality in the right striatum, however, raising the possibility that a deficit in right striatal D1 receptor binding may be specific to MDD subjects with anger attacks.

Our data more specifically localized the abnormality in left striatal D1 receptor binding in MDD to the left MC, which showed the greatest magnitude-of-difference and effect size across the striatal ROI (Figures 1 and 3). Because of the proximity of the MC to other striatal ROI, measured signals from the MC were weakly influenced by those from the AVS and DC which were partly continuous with the MC (Drevets et al, 1999, 2001). As parts of these ROI were separated by less than 6 mm resolution of our PET measures, demonstrating differential regional abnormalities in MDD depended upon showing that the mean difference in BPND was greater in the MC than in the DCA or AVS. This approach was facilitated by the low [11C]NNC-112-specific binding in extrastriatal areas, which reduced the number of comparisons required across adjacent regions. The ability to assess relative differences in radiotracer concentration across conditions in ROI separated by less than the FWHM resolution is central to PET's utility in localizing voxels of maximal difference in brain mapping studies (Fox et al, 1986; Friston et al, 1996). Because of these spatial resolution limitations, however, the nonsignificant trends seen in the left AVS and left DC (Figure 1) are difficult to interpret as they may reflect spilling in of radioactivity from the MC (Links et al, 1996).

The MC region is the major striatal target of predominantly ipsilateral, afferent projections from the ACC and orbitofrontal cortex (OFC) (Ferry et al, 2000; Haber et al, 2006). As the grey matter volume and/or neuronal counts in these cortical areas is reduced in MDD (Drevets et al, 2008), the regional specificity of the reduction in D1 receptor binding to the MC raises the possibility that this abnormality reflects a reduction in afferent neuronal terminals from the cortex, and thus in the density of synapses where D1 receptors are expressed postsynaptically. This hypothesis appears compatible with findings that in MDD both the reduction in D1 receptor binding in the MC (Figure 1) and the reduction in grey matter volume in the ACC and OFC are predominantly left-lateralized (Drevets et al, 2008). This hypothesis may also be compatible with the finding that the D1 receptor binding in the left MC correlated inversely with illness duration (Figure 5), as the reductions in ACC volume reportedly worsen across time in mood disorders (Koo et al, 2008). In contrast, perturbations in D1 receptor binding in MDD may be less likely to reflect local changes in receptor expression associated with changes in DA release, as the severe DA deficiency state accompanying Parkinson's Disease is not associated with changes in D1 receptor binding or density (Pimoule et al, 1985; Shinotoh et al, 1993). Moreover, although changes in D1 receptor regulation at the level of G-protein coupling may exist in the DA-depleted striatum (Gerfen et al, 2002), such a change would not likely be evinced by corresponding changes in D1 receptor antagonist binding to PET radioligands (Pimoule et al, 1985; Shinotoh et al, 1993) which typically bind to receptors in high vs low-affinity states with comparable potency.

Post hoc analyses revealed an altered interhemispheric ratio of D1 receptor BPND that differed significantly between depressed and control women in the DP. Specifically, the depressed women showed a loss of the asymmetry in D1 receptor binding that was evident in the controls (Figure 4). This difference was explained by D1 receptor BPND values that were nonsignificantly lower in the left DP and nonsignificantly higher in the right DP in depressed women vs control women (Figure 4).

In other striatal regions the D1 receptor binding also showed highly significant laterality effects in healthy controls, which did not differ significantly between groups. In the caudate and AVS (accumbens area), the BPND was lower in the left hemisphere than the right. In contrast, in the putamen the direction of the laterality effect was reversed (left>right). This difference between caudate and putamen may explain the absence of laterality effects in striatal [11C]SCH-23390 binding in Dougherty et al (2006), who combined caudate and putamen into a single ROI. Other studies employing neuroimaging or autoradiographic techniques to assess D1 receptor binding in depressed or healthy humans (De Keyser et al, 1988; Suhara et al, 1991, 1992; Bowden et al, 1997) or experimental animals (Cortes et al, 1989; Camps et al, 1990) did not examine laterality effects.

Asymmetrical D1 receptor binding across hemispheres is consistent with several reports of lateralized function within the dopaminergic system. The amount of DA release in human striatal subregions in response to unpredicted reward (Martin-Soelch et al, 2007), DA release in the rodent caudate during voluntary behavior (Yamamoto et al, 1982), and preference for self-stimulation induced by localized amphetamine administration in rats all show prominent laterality effects (Glick et al, 1980, 1981). In healthy humans greater D2 receptor availability in the left relative to the right striatum was associated with greater positive incentive motivation (Tomer et al, 2008). In right-handed humans DA concentrations were higher in the left putamen than the right (Glick et al, 1982; de la Fuente-Fernandez et al, 2000), and in marmosets levels of DA and HVA were higher in the right caudate and putamen than the left (Silva et al, 2007). Consistent with these data humans also showed higher dopamine transporter density in the left putamen and caudate vs the right, irrespective of handedness (van Dyck et al, 2002). (Although the lateralization of motor dominance (‘handedness’) has been associated with variation in elements of the dopaminergic system including COMT (Savitz et al, 2007), it remains unclear whether handedness also would influence D1 receptor BPND.) Finally, our finding that the loss of normal asymmetry in D1 receptor BPND in MDD appeared specific to women was notable given evidence that the direction of asymmetry preference for self-stimulation in rats showed prominent sex effects (Glick and Badalamenti, 1986).

As the mesocorticolimbic dopaminergic system is important in the neural processing of reward and motivation, it is conceivable that either a deficit in dopaminergic function in the left MC, or alterations in the interhemispheric ratio of D1 receptors that may result from such a deficit, may disrupt the normally lateralized function of this system and thereby contribute to the depressed mood, anhedonia, and amotivation associated with MDD (Nestler and Carlezon, 2006; Liu et al, 2008). Compatible with this hypothesis, the ratio of left-to-right D1 receptor binding in the MC correlated inversely with the IDS-C anhedonia subscale (r=−0.65, p=0.0040). In the left MC functional abnormalities have been identified during reward processing in MDD. In independent samples of MDD subjects the magnitude of DA release during unpredicted monetary reward and the hemodynamic response to anticipated monetary reward were abnormally blunted specifically in the left MC (Drevets, 2008).

The relationship between the response latencies on a sustained attention task (RVIP) and D1 receptor binding in the left MC in healthy controls suggests that the abnormal D1 binding group in MDD also may play a role in the psychomotor or cognitive impairment associated with MDD. The mean response latency was 12% slower in depressives than controls, as expected (Tavares et al, 2003), although we were underpowered to establish the statistical significance of this difference. Although reaction times measured on the RVIP reflect both psychomotor speed and sustained attention, these variables generally are confounded on neuropsychological assessments of psychomotor speed. In controls the response latencies correlated inversely with BPND in the left MC (explaining 20.5% of the variance) implying the reduction in D1 receptor binding in this region in MDD may contribute to the psychomotor slowing associated with depression. The absence of a normal correlation between reaction time and MC-BPND in the MDD group also may be compatible with this hypothesis. Notably, psychomotor slowing previously was associated with reduced dopaminergic function in the left caudate in MDD, as (Martinot et al, 2001) reported that [18F]DOPA was reduced specifically in the left caudate in MDD patients with psychomotor retardation compared to both healthy controls and MDD patients without psychomotor slowing.

The inverse correlation between D1 receptor binding and age is consistent with the results of previous PET (Suhara et al, 1991; Iyo and Yamasaki, 1993; Wang et al, 1998) or autoradiography studies (Klimek et al, 2002) conducted in humans or experimental animals (Suzuki et al, 2001). In humans Wang et al, (1998) reported a 6–7% decline in [11C]SCH-23390 binding to D1 receptors per decade in the caudate and putamen, comparable to the 3–6% decline in the BPND for [11C]NNC-112 per decade detected herein (assessed relative to the y-intercept for predicted value at birth; Figure 6). In contrast, the relationship between BPND and age was not significant in the amygdala, consistent with the data of Klimek et al. (2002) who observed no significant age effect on D1 receptor density measured post mortem in the human amygdala.

A limitation of our study was that we modeled the BPND for [11C]NNC-112 using a simplified reference tissue model to avoid arterial cannulation. This approach used the time-tissue radioactivity concentration in a reference region, the cerebellum, to estimate the concentration of free plus nonspecifically bound radiotracer in all regions. However, without arterial plasma we were unable to measure the distribution volume of [11C]NNC-112 in the cerebellum. Consequently, a difference in BPND between groups could have been accounted for by abnormal [11C]NNC-112 binding either in the region-of-interest or the reference tissue. However, an abnormality of tracer uptake in the reference tissue could not account for the regionally specific group difference or the group-by-gender-by-hemisphere interaction found in the left MC and the DP, respectively.

In conclusion, reduced D1 receptor binding in the left MC, and in women an altered symmetry in D1 receptor binding in the DP, may contribute to dysfunction within the central dopaminergic system during depression. The prominent laterality effects we observed for striatal D1 receptor binding converges with other evidence indicating that DA concentrations are lateralized within the striatum (Glick et al, 1982; de la Fuente-Fernandez et al, 2000; Silva et al, 2007), and that the function of the central dopaminergic system in the striatum is lateralized during reward processing (Martin-Soelch et al, 2007), voluntary movement (Yamamoto et al, 1982), and self-stimulation behavior (Glick et al, 1980, 1981; Glick and Badalamenti, 1986). As dopaminergic projections into the striatum modulate the neural processing of reward learning, motivated behavior, and psychomotor activity, alterations in the interhemispheric ratio of D1 receptor binding conceivably may disrupt the normally lateralized function of this system, and thereby contribute to the depressed mood, psychomotor slowing, anhedonia, and amotivation that characterize depression.