Decrease in Brain Serotonin 2 Receptor Binding in Patients With Major Depression Following Desipramine Treatment:  A Positron Emission Tomography Study With Fluorine-18–Labeled Setoperone FREE

THE MECHANISMS by which antidepressants exert their therapeutic effects in major depression have remained elusive. Since there is a time lag of about 2 to 4 weeks between initiating antidepressant treatment and the clinical response, it has been suggested that alterations in neurotransmitter receptors occurring within this time frame might be relevant to therapeutic effects of these drugs.¹ Because 5-hydroxytryptamine (serotonin [5-HT]) has been implicated in depression,^2- 3 a number of animal studies^4- 15 have assessed the changes in 5-HT receptors after 2 to 4 weeks of the administration of antidepressants. Serotonin 2 receptors are the most widely studied,^{4- 9,11- 13,15} and these studies have shown that long-term administration of tricyclic antidepressants,^{4- 6,8- 9,11- 13} monoamine oxidase inhibitors,^4,6,8,12 atypical antidepressants such as iprindole^4,6,12 and mianserin hydrochloride,^7- 8,13 and most but not all serotonin reuptake inhibitors^{4,9,12,16- 19} decreases the density of cortical 5-HT₂ receptors in rats. It is, however, unknown if antidepressant treatment would lead to similar changes in 5-HT₂ receptor density in depressed patients.

The purpose of our study was to ascertain the effects of 3- to 4-week treatment with an antidepressant on brain 5-HT₂ receptors using positron emission tomography (PET) in patients with major depression. Ligands such as carbon 11 (¹¹C (R)-(+)-4-(1-hydroxy-1-[2,3-dimethoxy-phenyl]methyl)-N-2-(4-fluorophenylethyl)-piperidine (MDL 100,907),²⁰¹¹C-N-meth-ylspiperone,²¹¹⁸F-setoperone,²² and ¹⁸F-altanserin²³ are available for imaging 5-HT₂ receptors in human brain.²⁴ We chose ¹⁸F-setoperone because (1) it has a total-to-nonspecific binding ratio of greater than 2; (2) it was recommended by the European Task Force as a suitable ligand for imaging 5-HT₂ receptors²⁴; and (3) it has been used successfully by other groups to explore 5-HT function in psychiatric disorders.^25- 28 We chose to test the effects of the antidepressant desipramine hydrochloride on 5-HT₂ receptors because this medication consistently has been shown to decrease the density of 5-HT₂ receptors in rat studies.

Eleven patients who met DSM-IV criteria for major depressive disorder²⁹ and who were suitable for treatment with desipramine were recruited from the University of British Columbia Hospital, Vancouver. The study was approved by the University Ethics Committee, and subjects gave written informed consent. The diagnosis of major depression was made using the Structured Clinical Interview for DSM-III-R.³⁰ None had other Axis I diagnoses or current or past alcohol or substance abuse. All subjects were physically healthy as determined by history and results of physical examination. The severity of depression was quantified using the 21-item Hamilton Depression Rating Scale (HAM-D).³¹ All study subjects were free of psychotropic drugs for at least 2 weeks (5 weeks in the case of fluoxetine hydrochloride) before baseline PET scan, and none had ever received electroconvulsive therapy.

Radiosynthesis of ¹⁸F-setoperone was accomplished by a modified method of Crouzel and colleagues,²² as described by Adam et al.³² Radioactivity in brain tissue was measured with a PET system (ECAT 953B/31; CTI/Siemens, Knoxville, Tenn). The spatial resolution of images is 6 mm (full width at half maximum). Subjects were supine with the head slightly tilted to obtain brain slices parallel to the canthomeatal line. Each subject had a 10-minute transmission scan for correcting PET images for attenuation. Following this, subjects were given 148 to 259 MBq of ¹⁸F-setoperone intravenously, and a 15-frame scan was performed. The numbers and durations of frames were as follows: 5 frames for 2 minutes each, 4 frames for 5 minutes each, 4 frames for 10 minutes each, and 2 frames for 20 minutes each (total duration, 110 minutes).

Each patient started desipramine hydrochloride therapy, 50 mg, and the dose was increased to 100 mg on day 4. Patients underwent assessment weekly with a clinical interview and a 21-item HAM-D. If there was no improvement in symptoms by the second week, the dose of desipramine hydrochloride was increased to 150 mg; a further increase in dose was made if no improvement was seen by the third week. No other psychotropic medication was allowed except lorazepam or oxazepam as required for night sedation. Subjects had a second PET scan 3 to 4 weeks after the commencement of desipramine therapy, 12 to 20 hours after the last dose.

Using the multipurpose imaging (MPI) tool³³ to define regions in frontal, temporal, and parietal cortex and cerebellum, we confirmed previous observations³⁴ that cortical- to -cerebellar ratio is constant, indicating pseudoequilibrium, between 70 to 110 minutes after injection of ¹⁸F-setoperone in depressed patients.

Statistical parametric mapping (SPM96) software^35- 36 was used to align PET images in the 70- to 110-minute period, to coregister them to magnetic resonance images, and to transform the magnetic resonance images (and PET images) into the standard coordinate frame used for templates in SPM96. Images were smoothed by applying a 12-mm (full width at half maximum) isotropic Gaussian filter to improve the signal-to-noise ratio.

For trace doses of setoperone at equilibrium, the amount of setoperone bound to receptors is proportional to the binding potential, B_max/K_d, where B_max is the total number of receptors and K_d is the equilibrium-binding constant. To determine the binding potential from the measured concentration of setoperone in cerebral tissue, it is necessary to allow for variation in the amount of setoperone in plasma. This can be performed by comparing the setoperone concentration in the region of interest with that in a reference region. Provided the capacity for nonspecific binding is identical in the region of interest and the reference region, the ratio of setoperone concentation in the region of interest (C_local) to that in the reference region (C_reference) is given using the following formula:

C_local/C_reference=f⋅BP_local+1,

where BP_local is the local binding potential, and f is a quantity that reflects the total binding capacity (specific+nonspecific) in the reference region. Although cerebellum has been used as a reference region, Petit-Taboue et al³⁷ have demonstrated that in humans, nonspecific binding of setoperone in cerebellum is different from that in cortex. Hence this equation is not accurate, and its use leads to errors of approximately 20% in the estimation of f⋅BP_local if cerebellum is used as a reference region. However, Petit-Taboue et al³⁷ demonstrated that nonspecific binding does not vary substantially between different cortical regions. Therefore, when the object is to measure regional changes in setoperone binding, an alternative approach is to use the whole cortex as the reference region.

The following equation is used when the whole cortex is treated as the reference region:

f=1/(BP_global+k5/k6),

where BP_global is mean global binding potential, and k5 and k6 are the transfer coefficients for attachment to and detachment from nonspecific binding sites. If BP_global+k5/k6 is not substantially affected by treatment, the change in C_local/C_global is proportional to the change in binding potential during treatment. If BP_global+k5/k6 changes during treatment, change in C_local/C_global is proportional to change in ratio of local binding potential to mean global binding capacity (specific+nonspecific), but there is reduced power to detect local changes in binding potential in regions where local change is in the same direction as the mean global change. Provided there is no substantial change in global binding, it is preferable to use the whole cortex as the reference region for the determination of local changes, as this avoids errors due to differences between cortical and cerebellar nonspecific binding.

We evaluated C_global using the middle 10 slices of the image and C_cerebellum in a cerebellar region drawn using the MPI tool. The pretreatment value of the ratio C_global/C_cerebellum was 1.173, whereas the posttreatment ratio was 1.148 (paired t₉, 0.514; P=.62). The minimal observed change in the ratio C_global/C_cerebellum justifies the use of whole cortex as the reference region for determining local changes in setoperone binding.

Using SPM96, we then determined the change in ratio of local setoperone concentration to mean global setoperone concentration during treatment with proportional scaling. The threshold for including voxels in the analysis was set at 1.3 times mean global cerebral image intensity to exclude non–gray matter voxels. For each voxel, the general linear model was used to estimate the mean change in C_local/C_global. The significance of the change for each voxel was determined applying the method developed by Worsley³⁸ based on the theory of Gaussian fields, as implemented in SPM96.^35- 36 Although there are approximately 58,000 voxels in the image, the image intensity in adjacent voxels is strongly correlated, so that there are only about 90 independent measurements. In effect, Worsley's method determines the number of independent measurements and applies the appropriate Bonferroni-type correction.³⁸ Voxels in which the significance of z satisfies this correction can be accepted as significant irrespective of whether they are a part of a contiguous set of voxels exhibiting significant change.

The theory of Gaussian fields was also applied to determine the statistical significance of clusters of contiguous voxels in which change during treatment exceeded a specified threshold. For this calculation, we set the threshold of z=1.96 corresponding to P<.025,one-tailed (or P<.05 two-tailed). Worsley's method³⁸ calculates the significance of clusters, taking account of the extent and peak z value within the cluster, and then applies a correction allowing for the multiple comparisons performed. Cluster significance level was set at P<.05.

Pearson correlations between changes in relative setoperone binding in the significant clusters and changes in 21-item HAM-D score and HAM-D suicide item score between pretreatment and posttreatment conditions were computed. Unless otherwise indicated, data are given as mean±SD.

One patient could not tolerate the side effects of desipramine and dropped out on day 4. The other 10 patients completed the study, and the data analysis is presented for these 10.

The subjects had a mean age of 40.6±9.2 years. Three patients had no previous depressive episodes, and 2 had 1 previous episode. The other 5 patients had from 3 to 12 previous episodes of depression (Table 1). The duration of current depressive episode ranged from 20 to 120 weeks. Four patients had been free of psychotropic medication for more than 6 months, and 1 patient was drug naive. The other 5 patients were drug free from 2 to 5 weeks. Three subjects received nighttime sedation with lorazepam and/or oxazepam during the study. The mean dose of desipramine hydrochloride was 160.0±21.0 mg (range, 150-200 mg). The patients had a mean 21-item HAM-D score of 27.0±5.6, and the mean HAM-D score decreased to 7.0±7.8 following treatment. Eight of 10 patients improved (defined as ≥50% reduction in 21-item HAM-D score) with treatment by second PET scan.

Depressed patients showed a significant reduction in setoperone binding in several cortical areas, including frontal, temporal, parietal, and occipital regions, following desipramine treatment. Analysis using SPM96 demonstrated an extensive cluster of voxels in which there was a significant decrease in setoperone binding following treatment with desipramine (Table 2, Figure 1, and Figure 2). This cluster was highly significant after correcting for multiple comparisons (P<.001). It included 26,695 voxels, and this corresponds to approximately 46% of the volume of gray matter that was in the field of view. The entire cluster embraced medial frontal cortex, insula, and lateral frontal cortex and extended back to temporoparietal and occipital cortexes bilaterally. Within this cluster, 383 voxels satisfied the criteria for significance irrespective of their membership in the cluster. Table 2 gives the location and z value for change in setoperone binding at those individual voxels where z exceeded 4.00 (P<.05 after stringent correction for multiple comparisons). The mean decrease in setoperone binding for the entire cluster was 8.1%, and it ranged from 8.0% to 15.7% for individual voxels that were significant (Table 2). The sites where most significant decreases in setoperone binding were observed included left medial and orbitomedial frontal gyri, left inferior frontal gyrus, left middle and inferior temporal gyri, right inferior frontal gyrus, right lingual gyrus, and right middle occipital gyrus (Table 2 and Figure 1 and Figure 2). At all these sites, decreases in setoperone binding occurred in all 10 subjects, including the 2 nonresponders. The magnitude of decrease in binding at these sites was similar in responders and nonresponders (Figure 3; data shown only for left medial frontal gyrus) and drug-naive and drug-free patients.

There were, however, no significant correlations between change in setoperone binding and change in 21-item HAM-D or suicide item scores (data not shown). As predicted, depressed patients did not show a significant increase in setoperone binding in any of the brain areas.

This is the first study to assess the effects of desipramine treatment on brain 5-HT₂ receptors in living, depressed patients. The results indicated that depressed patients had a significant decrease in brain setoperone binding in various cortical areas, including frontal, temporal, parietal, and occipital regions, following desipramine treatment. The decrease in setoperone binding was observed bilaterally, with decreases highly significant in left medial and orbitomedial frontal gyri, left inferior frontal gyrus, left middle and inferior temporal gyri, right inferior frontal gyrus, right lingual gyrus, and right middle occipital gyrus.

Several interpretations must be considered for a decrease in setoperone binding in our study subjects. First, the methods used in our study would not permit an independent determination of B_max and K_d. Hence, a decrease in setoperone binding observed following desipramine treatment in our study could be due to a decrease in B_max or an alteration in K_d. However, postmortem studies of brain 5-HT₂ receptors in depressed patients have shown a decrease in B_max without any alteration in K_d for 5-HT₂ receptors.^39- 43 Similarly, studies of the effects of antidepressant treatments on brain 5-HT₂ receptors in rodents also have shown a decrease in B_max without a change in K_d for 5-HT₂ receptors.^{4,7- 8,13,17- 18} Therefore, the decrease in setoperone binding observed in depressed patients following desipramine treatment in this study is likely to be due to a decrease in B_max, indicating a decrease in 5-HT₂ receptor density.

Second, although desipramine does not inhibit 5-HT reuptake, long-term treatment with desipramine might lead to an increase in brain 5-HT levels through some other mechanism. An increase in brain 5-HT levels could occupy 5-HT₂ receptors, leaving a decreased number of receptors available for setoperone binding, thus accounting for a decreased binding we observed. However, a recent PET study⁴⁴ using ¹⁸F-setoperone reported no significant change in 5-HT₂ receptor binding following an acute challenge dose of paroxetine, the most potent 5-HT reuptake inhibitor in healthy subjects. This would suggest that elevation in 5-HT levels has no measurable impact on 5-HT₂ receptor binding as assessed with PET, thus suggesting that decreased setoperone binding observed in the study subjects is not due to a confounding effect of elevation in endogenous brain 5-HT levels.

Third, since patients underwent scanning while receiving treatment with desipramine, we cannot exclude the possibility that the decrease in setoperone binding is due to desipramine binding to 5-HT₂ receptors and not due to a decrease in 5-HT₂ receptor density. However, this is unlikely, because desipramine binding would be expected to occur wherever there are 5HT₂ receptors, but we observed no substantial global change in setoperone binding.

Our study has some limitations. First, although setoperone binds with higher affinity to 5-HT_2A receptors (K_i=0.37 nmol/L),³⁴ it also has some affinity (K_i=50 nmol/L) to 5-HT_2C receptors,⁴⁵ and hence we cannot tell whether a reduction in binding observed is due to a down-regulation of 5-HT_2A or 5-HT_2C or both receptor populations. Second, most patients in our study improved by the time the second PET scan was performed. Hence, we cannot tell if the decrease in 5-HT₂ receptor binding is due to clinical improvement that may or may not have been induced by desipramine or due to an effect of desipramine that was unrelated to clinical improvement. The fact that all 10 patients showed a reduction in 5-HT₂ receptor binding but only 8 improved substantially supports the latter possibility. Conversely, in a preliminary PET study,⁴⁶ we reported that electroconvulsive therapy, another effective treatment for depression, induced a significant decrease in 5-HT₂ receptor binding in all 6 depressed patients who improved, supporting the former possibility.

Based on our finding of a decrease in 5-HT₂ receptor binding following desipramine treatment, one would expect an increase in 5-HT₂ receptor binding to be associated with depression. In this regard, several^39- 42,47 although not all^48- 51 postmortem studies that examined 5-HT₂ receptor binding in brain samples from suicides reported increased B_max with no changes in K_d in various brain regions, particularly in frontal cortex, compared with control subjects. Similarly, 5-HT₂ receptor binding was also reported to be increased in postmortem brain samples of depressed patients who died of natural causes, compared with controls,^43,52 although not in all studies.⁴⁸ Of the 2 studies that assessed brain 5-HT₂ receptor density in living, depressed patients, the single-photon emission tomography scan study with 2–iodine 123–ketanserin reported an increase in B_max within parietal cortex and right inferior frontal region,⁵³ whereas the PET study by Biver et al⁵⁴ using [¹⁸F]altanserin reported a decrease in 5-HT₂ receptor binding in the right posterolateral orbitofrontal cortex and anterior insular cortex in depressed patients compared with controls. The discrepancy in findings between the latter study and previous single-photon emission tomography and postmortem studies that showed an increased 5-HT₂ receptor binding in depression might be related to differences in duration of drug-free periods before ascertaining 5-HT₂ receptor binding.

In conclusion, to our knowledge, we demonstrated for the first time in humans that depressed patients show a decrease in brain 5-HT₂ ligand binding following desipramine treatment. Our results, however, cannot tell us whether the decrease in 5-HT₂ receptor binding was related to clinical improvement or to an effect of desipramine that was unrelated to clinical status. Further studies assessing the effects of antidepressant treatment on brain 5-HT₂ receptors in depressed patients are needed to verify the hypothesis that 5-HT₂ receptors are an important target for antidepressant therapies.

Reprints: Lakshmi N. Yatham, MBBS, FRCPC, Mood Disorders Clinical Research Unit, The University of British Columbia, 2255, Wesbrook Mall, Vancouver, British Columbia, Canada V6T 2A1 (e-mail: yatham@unixg.ubc.ca).

Accepted for publication April 13, 1999.

This study was supported by a National Alliance for Research on Schizophrenia and Depression Young Investigator Award (Dr Yatham).

Preliminary results of this study were presented at the 21st Collegium Internationale Neuro-Psychopharmacologicum Congress, Glasgow, Scotland, July 13, 1998.