Participants

All participants were assessed by a structured clinical interview for DSM–IV diagnosis (SCID). ^{Reference First, Spitzer, Gibbon and Williams11} The DSM–IV ¹² diagnosis of major depressive disorder was arrived at by the consensus of a research team based on information from an unstructured clinical interview as well as SCID. Individuals who met criteria for a major depressive disorder, refractory to previous antidepressant therapy and referred to a tertiary care mood disorders centre for ECT treatment were approached for participation in the study. The study was approved by the Clinical Research Ethics Board of the University of British Columbia. All participants provided written informed consent prior to completing any study procedures. Participants had no other Axis I diagnoses, substance or alcohol misuse within the past 6 months. All study participants had adequate trials with at least two antidepressant medications from two different classes and most of them had trials with several antidepressants and augmenting agents for the current episode. They were physically healthy and had a psychotropic medication washout period of at least 1 week prior to the baseline PET scan. The duration of washout period was determined for each individual based on the half-life of psychotropic medication they had been taking and the minimum washout period required was equivalent to at least five half-lives of the psychotropic medication. Individuals, however, were allowed to receive benzodiazepines for night sedation. Each participant had a magnetic resonance imaging (MRI) scan to exclude cerebral pathology and for co-registration of PET images. The severity of depressive symptoms in individuals before and after treatment with ECT was quantified using Hamilton Rating Scale for Depression (HRSD). ^{Reference Hamilton13,Reference Williams, Link, Rosenthal and Terman14}

PET scans

Each participant had two [¹⁸F]setoperone PET scans, one before the first ECT and the second within 1 week of the last ECT treatment. The [¹⁸F]setoperone was synthesised and the PET scans were performed as previously described. ^{Reference Yatham, Liddle, Shiah, Scarrow, Lam and Adam15} The participants were positioned in the scanner to obtain most of the cerebral cortex (frontal, temporal, parietal) apart from the upper regions near the vertex and some cerebellum. After participants had a transmission scan for 10 min to correct PET images for attenuation, they were injected with 4 to 7 mCi of [¹⁸F]setoperone intravenously and following this, 15 frame emission scans were performed over 110 min as previously described. ^{Reference Yatham, Liddle, Dennie, Shiah, Adam and Lane5} Radioactivity in the brain was measured with a PET camera ECAT/953B (Siemens, Knoxville, Tennessee, USA) that collects 31 axial slices. The spatial resolution of PET images is about 9 × 9 × 6 mm³.

ECT procedure

The procedure for ECT administration was similar to that described previously. ^{Reference Shiah, Yatham, Srisurapanont, Lam, Tam and Zis16} A Thymatron ECT apparatus was used to deliver electrical stimulus to individuals who had general anaesthesia induced with sodium thiopental and muscle relaxation with succinylcholine. All of them were given right unilateral ECT. Seizure threshold was estimated during the first treatment. Subsequent treatments were given at three times the seizure threshold, and further adjustments in stimulus intensity (an increase of 10–20%) were made for each ECT treatment to elicit a seizure of at least 20 s duration (measured with single strip electroencephalogram recording) and good amplitude with each treatment. Participants received ECT treatments three times a week and the number of ECT treatments for each individual was determined by the treating clinician based on clinical judgement of treatment response or non-response. The standard practice at the UBC Hospital is to give at least two more ECT treatments after remission of depression, but in those that had not shown any improvement, ECT is typically discontinued after eight ECT treatments. Each participant had a second PET scan between 3 and 7 days after the last ECT treatment.

Image processing and data analysis

The PET images were processed as previously described using Statistical Parametric Mapping software 2 (SPM2; Wellcome Department of Cognitive Neurology, Institute of Neurology, London, www.fil.ion.ucl.ac.uk/spm/software/spm2) run on Matlab 6 (Mathworks Inc., Natick, Massachusetts, USA). ^{Reference Yatham, Liddle, Dennie, Shiah, Adam and Lane5} Briefly, the PET scan frames 14 and 15 corresponding to the last 40 min of emission data were realigned to each other to create a mean image for baseline scan and a mean image for post-ECT scan. These two mean images were realigned to create a composite image that was then co-registered with the individuals' MRI. The PET and MRIs were then transformed into the standard coordinate frame used for templates in SPM2 using the spatial normalisation method specified in SPM2.

The SPM2 was used to determine the significance of the change in regional [¹⁸F]setoperone binding with ECT. Because non-specific binding for setoperone in cerebellum differs from that in cortex, ^{Reference Petit-Taboue, Landeau, Barre, Onfroy, Noel and Baron17} we used a reference region comprising all cortical grey matter voxels, excluding voxels in which there was a change (either decrease or increase) in binding significant at P<0.1 ^{Reference Desjardins, Kiehl and Liddle18} and the rationale for this is as follows: the ratio of concentration of ligand in a local region to that in a reference region is given by

\batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \[\ C_{\mathrm{local}}{/}C_{\mathrm{reference}}=[BP_{\mathrm{local}}+(k_{5}{/}k_{6})_{\mathrm{local}}]{/}[BP_{\mathrm{reference}}+(k_{5}{/}k_{6})_{\mathrm{reference}}]\ \] \end{document}

where BP denotes binding potential, and k ₆ and k ₆ are the rate constants for attachment to and dissociation from non-specific binding sites. If the local and reference regions are in cerebral cortex, it is reasonable to assume that (k ₅/k ₆) is similar in both regions. If the reference region is defined as that cortical region in which there is no change in binding between pre- and post-treatment scans and furthermore, if it is assumed that (k ₅/k ₆) is not affected by treatment, the change in

\batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \[\ C_{\mathrm{local}}{/}C_{\mathrm{reference}}{\ }\mathrm{is\ given\ by}{\ }{\Delta}[C_{\mathrm{local}}{/}C_{\mathrm{reference}}]=f.{\Delta}BP_{\mathrm{local}}\ \] \end{document}

where ΔBP _local is the change in local binding potential and

\batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \[\ f=1{/}[BP_{\mathrm{reference}}+(k_{5}{/}k_{6})]\ \] \end{document}

Therefore, a change in (C _local /C _reference), while systematically underestimating ΔBP _local, remains proportional to it. The reference region in which there was no change was identified by iterative exclusion of all voxels in which there was trend towards either an increase or decrease in setoperone binding during treatment, at level P<0.1 as stated earlier. ^{Reference Desjardins, Kiehl and Liddle18}

For inclusion of voxels in the SPM analysis, grey matter threshold was set at 120% of the mean image intensity as this threshold eliminated most white matter voxels without excluding any grey matter voxels. The theory of Gaussian fields as implemented in SPM2 was applied to assess the statistical significance of cluster of contiguous voxels in which change during treatment exceeded a threshold of P<0.025. This threshold was chosen because we predicted a widespread reduction in brain 5-HT₂ receptors with ECT. The Worsley's method in SPM takes into account extent within the cluster and applies correction for multiple comparisons in calculating cluster significance. The corrected cluster significance was set at P<0.01. We also assessed the significance of change in [¹⁸F]setoperone for each voxel using the Worsley's method based on the theory of Gaussian fields ^{Reference Worsley19} as implemented in SPM. ^{Reference Friston, Frith, Liddle and Frackowiak20} The SPM analysis provides corrected P-values based on family-wise error control (FWE-corr) as well as false-discovery rate. The former controls chances of any false positives in the brain, whereas the latter controls for the expected proportion of false positives among suprathreshold voxels. The false-discovery rate is more sensitive and is adaptive to the amount of signal observed in the data. ^{Reference Genovese, Lazar and Nichols21}

Computation of binding potential using Logan graphical method

Since the ratio method is susceptible to changes in ligand delivery/blood flow, we also computed the pre- and post-treatment values of the 5-HT₂ binding potential (BP _ND) for the regions identified as yielding a significant difference in the ratio values, using the Logan graphical method. ^{Reference Logan, Fowler, Volkow, Wang, Ding and Alexoff22} The BP _ND refers to the ratio at equilibrium of specifically bound radioligand to that of non-displaceable radioliogand in tissue and represents the ratio between the maximum free receptor density (B _max) and the dissociation constant (K _d). Since studies have shown ^{Reference Klimek, Zak-Knapik and Mackowiak23} no change in K _d with treatment, it can be assumed that a change in BP _ND is directly related to a change in receptor density.

Is 5-HT₂ receptor down-regulation intrinsic to mechanism of action of antidepressant treatments?

The finding that ECT and three different antidepressants with differing modes of action (desipramine – a noradrenaline reuptake inhibitor; ^{Reference Yatham, Liddle, Dennie, Shiah, Adam and Lane5} paroxetine – a serotonin reuptake inhibitor (SSRI); ^{Reference Meyer, Kapur, Eisfeld, Brown, Houle and DaSilva6} and nefazodone – an SSRI and a 5-HT₂ receptor antagonist) ^{Reference Mischoulon, Dougherty, Bottonari, Gresham, Sonawalla and Fischman7} all down-regulate brain 5-HT₂ receptors, suggests that down-regulation of 5-HT₂ receptors is either an intrinsic feature of the mechanism of antidepressant action or is a direct consequence of antidepressant action. Furthermore, this study of ECT and all three previous studies with antidepressants ^{Reference Yatham, Liddle, Dennie, Shiah, Adam and Lane5–Reference Mischoulon, Dougherty, Bottonari, Gresham, Sonawalla and Fischman7} have reported down-regulation of 5-HT₂ receptors in limbic and prefrontal areas, although with paroxetine this was seen mainly in younger participants. ^{Reference Meyer, Kapur, Eisfeld, Brown, Houle and DaSilva6} Limbic and frontal areas are implicated in the neurobiology of depression. ^{Reference Mayberg, Goldapple and MacIntosh27,Reference Mayberg28} The possibility that down-regulation of 5-HT₂ receptors in limbic and frontal areas is intrinsic to the mechanism of antidepressant action is further supported by the finding of a trend towards correlation between ECT-related changes in [¹⁸F]setoperone binding in medial prefrontal cortex as well as right parahippocampal gyrus and clinical improvement in depression as measured by changes in HRSD scores.

5-HT₂ receptors and right hemisphere

Although reduction in brain 5-HT₂ receptors was observed in both right and left cortical regions, changes were slightly prominent on the right side. Since all participants in this study received right unilateral ECT, it raises the possibility that slightly more prominent reduction in 5-HT₂ receptors in right cortical regions might be as a result of the fact that the electrical stimulus was delivered to the right brain. However, previous studies that assessed the effects of antidepressants on brain 5-HT₂ receptors have also observed such asymmetry with changes slightly more prominent on the right side. ^{Reference Yatham, Liddle, Dennie, Shiah, Adam and Lane5,Reference Mischoulon, Dougherty, Bottonari, Gresham, Sonawalla and Fischman7} Furthermore, some studies have also reported changes in brain 5-HT₂ receptors only on the right side in untreated individuals with depression. ^{Reference D'haenen, Bossuyt, Mertens, Bossuyt-Piron, Gijsemans and Kaufman29,Reference Biver, Wikler, Lotstra, Damhaut, Goldman and Mendlewicz30} Given that the density of 5-HT₂ receptors in human brain appears to be similar in both hemispheres, ^{Reference Schotte, Maloteaux and Laduron31} these findings suggest that changes in brain 5-HT₂ receptors in the right brain might be more relevant to neurobiology of depression and therapeutic effects of treatments.

Brain 5-HT₂ receptors and efficacy of ECT in refractory depression

Interestingly, the percentage reduction in brain 5-HT₂ receptors was only 3.8% for the entire cluster, 5% in medial frontal gyrus and 6.7% in right parahippocampal gyrus. These percentage reductions in 5-HT₂ receptor binding with ECT are somewhat smaller than those reported with desipramine (8–15.7%) ^{Reference Yatham, Liddle, Dennie, Shiah, Adam and Lane5} and paroxetine (10%). ^{Reference Meyer, Kapur, Eisfeld, Brown, Houle and DaSilva6} However, it should be noted that many of the participants in this study had already received extensive treatment with antidepressants before ECT, so it is possible that the reductions related to ECT were superimposed on prior reduction associated with antidepressant medication. Furthermore, some ^{Reference Yatham, Liddle, Shiah, Scarrow, Lam and Adam15,Reference Mintun, Sheline, Moerlein, Vlassenko, Huang and Snyder32} but not all ^{Reference Meyer, McMain, Kennedy, Korman, Brown and DaSilva33,Reference Bhagwagar, Hinz, Taylor, Fancy, Cowen and Grasby34} studies suggest that untreated individuals with major depression have up to 29% reduction in brain 5-HT₂ receptors compared with age- and gender-matched healthy individuals and we have previously hypothesised that such down-regulation might be a compensatory homeostatic response of the brain to the state of depression. ^{Reference Yatham, Liddle, Shiah, Scarrow, Lam and Adam15} Thus, it is plausible that, in some individuals, natural homeostatic processes might produce sufficient reduction in brain 5-HT₂ receptors to alleviate depression; in others, antidepressant medication is required, whereas in treatment-refractory cases, ECT is required to produce additional reduction in 5-HT₂ receptor density. This hypothesis is supported by the findings of this study that showed that the antidepressant-refractory participants had further reduction in 5-HT₂ receptors and that there was a trend for a correlation between reduction in [¹⁸F]setoperone binding in medial prefrontal cortex as well as right parahippocampal gyrus and clinical improvement in depression with ECT. The reductions in 5-HT₂ receptors observed in this study in participants who were refractory to antidepressant treatment may explain not only its superior efficacy to antidepressants but also its benefit in those non-responsive to pharmacotherapy.

Brain 5-HT₂ receptors and improvement in depression

If ECT and antidepressant medications all down-regulate brain 5-HT₂ receptors, and if such down-regulation mediates antidepressant effects, other treatments that down-regulate 5-HT₂ receptors should also work as antidepressants. Indeed, atypical antipsychotics, which block and down-regulate brain 5-HT₂ receptors, ^{Reference Tarazi, Zhang and Baldessarini35} have recently been shown to have significant antidepressant properties in clinical trials of people with bipolar depression. ^{Reference Tohen, Vieta, Calabrese, Ketter, Sachs and Bowden36,Reference Calabrese, Keck, Macfadden, Minkwitz, Ketter and Weisler37} In contrast, conventional antipsychotics such as haloperidol do not alter brain 5-HT₂ receptors ^{Reference Burnet, Chen, McGowan, Franklin and Harrison38} and have no significant antidepressant properties or, indeed, can induce depressive symptoms. ^{Reference Tohen, Goldberg, Arrillaga, Azorin, Vieta and Hardy-Bayle39} Furthermore, in another PET study, we have shown that, in healthy volunteers, 5-HT₂ receptor down-regulation is associated with lack of depressive symptoms induced by tryptophan depletion. ^{Reference Yatham, Liddle, Shiah, Lam, Adam and Zis40}

Limitations

Some limitations of our study warrant comment. First, although setoperone binds with higher affinity to 5-HT₂ receptors (K _i = 0.37 nmol/l), it has some affinity to α₁ adrenergic as well as dopamine D₂ receptors. ^{Reference Blin, Sette, Fiorelli, Bletry, Elghozi and Crouzel9,Reference Crouzel, Guillaume, Barre, Lemaire and Pike10} However, pre-treatment with α₁ antagonist prazosin or D₂ antagonist sulpiride did not change specific/non-specific binding ratios in humans, indicating that setoperone signal in humans is reflective of 5-HT₂ binding. ^{Reference Blin, Sette, Fiorelli, Bletry, Elghozi and Crouzel9} Setoperone also has a good specific to non-specific binding ratio in cortex and hence is considered a suitable ligand for assessing brain 5-HT₂ receptors in humans. ^{Reference Blin, Sette, Fiorelli, Bletry, Elghozi and Crouzel9,Reference Crouzel, Guillaume, Barre, Lemaire and Pike10} It is likely that setoperone binding provides an estimate of brain 5-HT_2A receptors as it was fully displaced by EMD281014, a more selective 5-HT_2A antagonist. ^{Reference Mamo, Sedman, Tillner, Sellers, Romach and Kapur41} Second, since each ECT treatment was given under general anaesthesia, one could argue that the anaesthetic agent contributed to reduction in brain 5-HT₂ receptors observed in this study. As no study to date examined the effects of barbiturates anaesthesia on brain 5-HT₂ receptors, we cannot exclude this possibility. However, a previous study that examined the effects of other anaesthetic agents such as ketamine, zoletile mixture, isoflurane or halothane reported that none had any effect on either 5-HT₂ receptors or cerebellar non-specific binding. ^{Reference Elfving, Bjornholm and Knudsen42} Third, we cannot exclude the effects of the muscle relaxant succinylcholine on brain 5-HT₂ receptors as no previous study assessed its effects. Fourth, the method used to assess changes in 5-HT₂ receptors in this study does not provide an absolute quantification of 5-HT₂ receptor binding. In fact, the changes in 5-HT₂ receptor binding reported here are relative changes between pre- and post-treatment scans in total bound concentration of ligand. Provided that non-specific binding did not change during treatment (an assumption also required if we had used a cerebellar reference region), these changes reflect a change in specific binding. Further, these reported changes are relative to total pre-treatment binding at the reported locations. Assuming the changes in non-specific binding are negligible, our method, although providing an estimate of change in specific binding that is proportional to the change in 5-HT₂ binding potential, systematically underestimates the true change in 5-HT₂ receptors. This may have been one of the reasons for a smaller observed effect of ECT on 5-HT₂ receptors in this study. Fifth, in the absence of a control group, we cannot exclude the possibility of a systematic change in 5-HT₂ binding over time that is unrelated to treatment. However, the observation by Chow et al ^{Reference Chow, Mamo, Uchida, Graff-Guerrero, Houle and Smith43} of test–retest variability of 1.9% in frontal lobes and 3.3% in temporal lobes without evidence of significant systematic change makes it unlikely that the change we observed (3.8% averaged over the entire cluster, with a 6.7% peak reduction) was as a result of systematic change unrelated to treatment.

In conclusion, this study provides evidence that, in contrast to the effects of electroconvulsive shock on 5-HT₂ receptors in rodents, ECT in individuals with depression down-regulates brain 5-HT₂ receptors in the limbic and prefrontal cortical brain areas that have previously been implicated in the neurobiology of depression. The findings of this study taken in conjunction with the findings of previous studies of the effects of antidepressants on brain 5-HT₂ receptors, consolidates the role of brain 5-HT₂ receptors in mediating the antidepressant effects and may put to rest the controversy over the last four decades.