Using the PET tracer [11C]K-2 to image AMPAR density in vivo, the authors found that lower AMPAR availability correlates with greater illness severity and differs between patients with treatment‑resistant depression and healthy controls. Ketamine produced region‑specific changes in AMPAR density that correlated with clinical improvement and partially restored the abnormal AMPAR phenotype, supporting AMPAR dynamics as a mechanistic substrate of ketamine’s antidepressant effect in TRD.
Approximately 30% of patients with depression suffer from treatment-resistant depression (TRD). Ketamine has shown antidepressant efficacy for TRD. While glutamate α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPAR) has been demonstrated to play crucial roles in the process of pharmacological action of ketamine in experimental animals, it remains elusive how ketamine exhibits its efficacy through changes in AMPAR dynamics in patients with TRD. In this study, using a positron emission tomography (PET) tracer, [ 11 C]K-2, which depicts AMPAR density in the living human brain, we detected a negative correlation between AMPAR density and illness severity and differences in AMPAR distribution between patients with TRD and healthy participants. Furthermore, we detected brain areas where ketamine administration altered AMPAR density in significant correlations with ketamine-induced antidepressant effect in patients with TRD. AMPAR density alteration in these regions partially rescued AMPAR phenotype in the affected areas. Thus, AMPAR dynamics underlies the antidepressant effect of ketamine in patients with TRD.
Major depressive disorder is common and a substantial proportion of patients do not respond adequately to standard antidepressant treatment, creating treatment-resistant depression (TRD). Ketamine has shown robust antidepressant effects in TRD, and previous animal work has implicated AMPA receptors (AMPARs) in this action. However, the paper notes that it remained unclear how ketamine changes AMPAR dynamics in humans with TRD, and whether such changes relate to symptom improvement. The authors also frame this as an important translational gap because ketamine’s antidepressant effect is short-lived and its long-term safety remains a concern. Nakajima and colleagues therefore set out to map whole-brain AMPAR distribution in Japanese patients with TRD using the PET tracer [11C]K-2, compare these patients with healthy participants, and examine how ketamine treatment changes AMPAR density in relation to clinical response. They aimed to identify brain regions where baseline AMPAR density tracked illness severity, where TRD differed from healthy status, and where ketamine-induced AMPAR changes corresponded to antidepressant improvement. The study was designed to help clarify whether AMPAR dynamics could serve as both a mechanistic marker and a potential biomarker of ketamine response in humans.
Papers cited by this study that are also in Blossom
Abdallah, C. G., Charney, D. S., Duman, R. S. et al. · Neuron (2019)
Yang, Y., Cui, Sang, K. et al. · Nature (2018)
Murrough, J. W., Perez, A. M., Pillemer, S. et al. · Biological Psychiatry (2012)
Smith-Apeldoorn, S. Y., Veraart, J. K. E., Spijker, J. et al. · Lancet Psychiatry (2022)
The researchers combined data from three clinical studies conducted at Yokohama City University Hospital, Keio University Hospital, and Kyushu University Hospital between August 2016 and October 2023. The main study was a randomised, double-blind, placebo-controlled trial in patients with TRD, and two additional studies contributed healthy participant data for comparison. Ethical approvals and written informed consent were obtained for all studies, and the data were analysed retrospectively. Patients with TRD were adults aged 20-59 years who met DSM-5 criteria for major depressive disorder, had inadequate response to at least two antidepressants in the current episode, had a baseline MADRS score of at least 22, and had adequate decision-making capacity. After exclusions and discontinuations, baseline PET data were analysed for 34 participants, with 17 assigned to ketamine and 17 to placebo. Healthy participant data came from 54 individuals, although several were excluded after further assessment; the excerpt does not clearly state the final analysed number in that section, but the results indicate age-matched healthy controls were used for comparison. In the double-blind phase, participants were allocated 1:1 to intravenous ketamine 0.5 mg/kg over 40 minutes or placebo saline, twice weekly for two weeks, for a total of four infusions. Both participants and clinical raters were blinded, while drug preparation staff and the supervising anaesthesiologist were unblinded. Post-treatment clinical and PET imaging assessments were completed about 3.2 days after the fourth infusion. Participants in the placebo group were then offered open-label ketamine using the same dosing schedule, and outcomes were reassessed before and after that course. PET imaging used [11C]K-2 to estimate cell-surface AMPAR density, with standardised uptake value ratio (SUVR) 30-50 min after tracer injection calculated using whole brain as reference. MRI-based co-registration and spatial normalisation to MNI space were performed with SPM and DARTEL, and PET images were smoothed before analysis. The authors compared SUVR between TRD and healthy participants using two-sample t-tests, compared pre- and post-ketamine scans with paired t-tests, and used multiple regression in SPM12 to test associations between SUVR, change in SUVR, and MADRS scores. Age and sex were included as covariates in most models, with additional sensitivity analyses adjusting for illness duration, benzodiazepine use, and number of failed antidepressant trials. Multiple-comparison control used FDR correction at the cluster level, with analyses restricted to grey matter regions. The study also defined overlap maps to identify regions showing both baseline abnormalities and ketamine-related changes aligned with symptom improvement. This approach was used to test whether ketamine shifted AMPAR distribution back towards the pattern seen in healthy participants or towards the opposite direction, depending on the region.
In patients with TRD, baseline AMPAR density measured as SUVR 30-50 was negatively correlated with MADRS severity in several cortical and cerebellar regions, including the left frontal, temporal, parietal and occipital lobes, as well as the cerebellum. The authors describe these as illness-related or “state” regions. Compared with healthy participants, TRD patients showed lower SUVR in the anterior insula, anterior to middle cingulate cortex, frontal, parietal and occipital lobes, and higher SUVR in the cerebellum, temporal lobe, thalamus, posterior insula and basal ganglia. Age and sex were associated with some baseline AMPAR differences, whereas illness duration, number of failed antidepressants and benzodiazepine use were not significantly associated with baseline AMPAR density. After ketamine treatment, the average pre- versus post-treatment SUVR images did not differ significantly at the whole-brain level. However, regional changes in SUVR were related to antidepressant response. In the ketamine group, increases in SUVR change were positively correlated with symptom improvement in areas including the parietal lobe, middle cingulate cortex and part of the left frontal lobe, while negative correlations were seen in the cerebellum, thalamus, left parahippocampal gyrus and right basal ganglia. In the habenula, ketamine-related reduction in AMPAR density was negatively correlated with improvement, which the authors interpret as regionally specific regulation. No significant correlations between SUVR change and symptom improvement were found in the placebo group. Several of the ketamine-response regions overlapped with the regions that were abnormal at baseline in TRD. Regions with reduced AMPAR density in TRD, such as the parietal and occipital lobes, middle cingulate cortex and part of the left frontal lobe, showed positive associations between ketamine-induced SUVR increase and clinical improvement, suggesting partial normalisation. Regions with increased baseline AMPAR density, including the cerebellum and right basal ganglia, showed negative associations with response. In the sensitivity analyses, the main pattern remained broadly similar, although negative correlations also emerged in the bilateral basal ganglia. Baseline AMPAR density before ketamine also predicted response in a region-dependent way. Higher pre-treatment SUVR in the frontal, temporal, parietal, occipital lobes, insula, cingulate cortex and basal ganglia was positively correlated with percentage improvement in MADRS, while negative correlations were also reported in parts of the temporal, occipital and parietal lobes. After adjustment for covariates, the positive associations remained largely unchanged, whereas the negative associations no longer reached significance. Overall, the data suggest that regional AMPAR distribution is linked both to the severity of TRD and to the magnitude of ketamine response.
The authors interpret their findings as evidence that AMPAR density is altered in patients with TRD and that ketamine changes AMPAR dynamics in directions that are associated with antidepressant benefit. They argue that some regions reflect symptom-related “state” abnormalities, while others may represent more stable “trait” regions linked to the underlying biology of treatment resistance. They suggest that TRD may have regional AMPAR patterns distinct from non-TRD depression, consistent with a biologically different treatment trajectory. A major focus of the discussion is the habenula. The authors note that ketamine-induced reduction in AMPAR density in the habenula was associated with symptom improvement, aligning with earlier animal studies showing that ketamine suppresses lateral habenula activity through AMPAR-related mechanisms. They also emphasise overlap between ketamine-response regions and regions previously implicated in depression circuitry, especially the precuneus and superior parietal cortex, which are part of networks associated with the default mode network and dorsal prefrontal connectivity. They further discuss the occipital cortex, proposing that ketamine may restore visual-network dysfunction in depression through AMPAR-mediated plasticity. The authors also position their findings relative to earlier human imaging work. They state that their previous study in mainly non-TRD depression did not find a group difference in [11C]K-2 uptake versus healthy participants, whereas the current TRD cohort showed marked differences, supporting the idea that treatment resistance is associated with a more pronounced AMPAR phenotype. They suggest that some AMPAR-abnormal regions may be shared across psychiatric disorders, while others are disease-specific. The paper acknowledges several limitations. The imaging sample size was determined by the parent clinical trial and may be underpowered for regional PET analyses. The study was conducted only in a Japanese population, which may limit generalisability. The range of symptom improvement was constrained by the parent trial, and the authors note that more severe cases should be studied in future work. They also point out that no active placebo was used, so unblinding due to ketamine’s noticeable effects could have influenced symptom ratings, although the biological association between AMPAR change and clinical improvement was observed only in the ketamine group. The authors conclude that these findings support AMPAR dynamics as a plausible mechanistic target and potential biomarker of ketamine response in TRD, while noting that larger and more diverse studies are needed to confirm the results.
This study comprised three clinical studies that were registered under the following IDs: jRCTs031210124, UMIN000025132, and jRCTs031200083. The main ketamine study (jRCTs031210124) was approved by the Certified Review Board of Keio in accordance with the Ethical guidelines for medical and health research involving human participants by the Japan Ministry of Health, Labour and Welfare and the Clinical Trials Act in Japan. For comparison of healthy participants, the data were used in the other two studies (UMIN000025132, jRCTs031200083) which approved by Yokohama City University Human Investigation Committee and Yokohama City University Certified Institutional Review Board following the same ethical guidelines and the Clinical Trials Act in Japan. These studies were conducted at Yokohama City University Hospital, Keio University Hospital, and Kyushu University Hospital between August 2016 and October 2023. The specific roles of each site were as follows: recruitment, informed consent, administration of study drugs, and clinical assessments for patients with TRD in the main study were all conducted at Keio University Hospital. In the main study, PET/MRI imaging was performed at Keio University Hospital and Yokohama City University Hospital. Imaging healthy participants in the other two studies were conducted at Yokohama City University Hospital and Kyushu University Hospital. All participants provided written informed consent after receiving detailed information about the protocol. The CONSORT diagram is provided in Supplemental Fig.. Data from these three studies were combined and analyzed retrospectively.
Patients with treatment resistant depression. The inclusion criteria were as follows: patients who (a) were in-and outpatients 20-59 years of age; (b) met the Diagnostic and Statistical Manual of Mental Disorders Fifth Edition (DSM-5) criteria for MDD using the structured clinical interview for DSM-5, the Research version (SCID-5-RV); (c) had an inadequate response defined as < 50% subjective improvement to approved doses of at least two antidepressants in the current episode based on a visual analogue scale ranging from 0 (no improvement) to 100 (complete improvement); (d) had a total score of ≥ 22 on the Montgomery Åsberg Depression Rating Scale (MADRS) at screening; and (e) had sufficient decision-making capacity confirmed by the MacArthur Competence Tool for Clinical Research. The exclusion criteria are provided in the Supplementary information. Baseline PET imaging data were analysed for 34 participants. Baseline demographic and clinical characteristics did not differ significantly between the ketamine and placebo groups (Table). The mean age for the entire sample was 41.4 ± 9.4 years, 11 (32.4%) participants were female, and the mean duration of illness was 11.6 ± 8.2 years. The mean baseline MADRS total score was 28.1 ± 7.6. Details of concomitant medications are provided in Supplemental Table.The changes in the MADRS total scores and % improvement of the MADRS scores were -9.1 ± 9.9, 30.7 ± 31.4% in the ketamine group, -2.7 ± 5.1, 8.4 ± 22.5% in the placebo group (double-blind period), and -11.1 ± 8.7, 40.1 ± 26.5% in the placebo group (open-label period), respectively (the detail in Supplemental Table, 3). These clinical assessments were performed by one of the trained investigators who was blind to the results of the PET and MRI imaging. Healthy participants. In the first study (UMIN000025132) the inclusion criteria were: healthy male participants who were 30-79 years of age and did not fulfil any diagnostic criteria for psychiatric conditions according to the DSM-IVusing the SCID-I/DSM-IV, DSM-5or ICD-10. Among them, age-matched (i.e., 30-59 years) healthy participants were included. In the second study (jRCTs031200083), the selection criteria were the same as those in the first study, other than the age range (i.e., 20-49 years) and sex (i.e., both men and women were included). The exclusion criteria are provided in the Supplementary information. Fifty-four healthy participants were included. After inclusion, thorough assessments identified that four participants met criteria for social communication disorder according to DSM-5 (n = 2) or had cognitive impairment (n = 2), and they were therefore excluded from further analysis. Additionally, one participant was excluded due to incompatibility of spatial normalization of [ 11
A detailed explanation of this trial is described elsewhere. Briefly, enrolled participants were randomly allocated to either the ketamine or placebo group in a 1:1 ratio. Both all participants and clinical raters were blind to the allocation during the double-blind period. The investigator responsible for study drug preparation, the pharmacist in charge of dispensing, and the supervising anesthesiologist were unblinded, but they had no involvement in any clinical or imaging assessments throughout the trial. During the double-blind treatment period, the baseline assessments including PET scan and clinical assessment were performed before the first administration of the study drug. In the ketamine group, intravenous ketamine (0.5 mg/kg) was administered over 40 min twice a week for two weeks (i.e., four times in total), whereas in the placebo group intravenous placebo (0.9% sodium chloride for injection) was administered over 40 min twice a week for two weeks. This repeated-infusion design was chosen to investigate the neurobiological mechanisms of ketamine's sustained antidepressant effects rather than the transient acute effects. Both posttreatment clinical and PET imaging assessments were performed at 3.2 ± 1.5 days after the fourth administration of the study drug on the same day. Participants assigned to the ketamine group were terminated from the study at that time, while those assigned to the placebo group were then offered the opportunity to receive open-label, intravenous ketamine treatment. During this open-label treatment period, the participants received the treatment with intravenous ketamine (0.5 mg/kg) over 40 min twice a week for two weeks and received clinical assessment before and after this course of treatment. Participants continued the same dose of psychotropic medications that they were receiving at screening in both arms throughout the study period (see Supplemental Table). The CONSORT diagram illustrates the flow of the participants (Supplemental Fig.). A total of 34 participants were assigned to receive either ketamine (n = 17) or placebo (n = 17) during the double-blind period. Three discontinued participations during the double-blind phase due to their intention to withdraw (n = 2 in the ketamine group) or fatigue (n = 1 in the placebo group). Subsequently, 16 participants in the placebo group received ketamine in the open-label extension period. Consequently, baseline PET imaging data were analysed for 34 participants (n = 17 in the ketamine group and n = 17 in the placebo group). For the analysis of the association between ΔSUVR 30-50 and % improvement of MADRS, the matched pre-and post-treatment PET imaging data were analysed for 31 participants (n = 15 in the ketamine group and n = 16 in the placebo group).
MRI images scanned at Yokohama City University Hospital and Kyusyu University Hospital were segmented into probability maps of grey matter, white matter, and cerebrospinal fluid using a unified framework for tissue segmentation. In MRI images scanned at Tokyo University, multiple segmentation was conducted using T1-and T2-weighed MRI. PET images scanned at Keio University Hospital were processed with a 5.0 mm fullwidth at half-maximum (FWHM) Gaussian filter to match the resolution of raw PET images scanned at other sites. PET image taken between 30-50 min after injection of [ 11 C]K-2 were summed and normalized to whole brain radioactivity, creating standardized uptake value ratio (SUVR) 30-50 images. SUV of whole brain didn't change in pre-and post-treatment with ketamine (pre-treatment: 2.44 ± 0.45, post-treatment: 2.34 ± 0.45, p = 0.471 (paired t-test)). All SUVR 30-50 images were co-registered to each individual's MRI. A template for anatomical normalization was created using the high-dimensional nonlinear warping algorithm DARTEL. SUVR 30-50 images were spatially normalized into MNI standard space using the template with statistical parametric mapping (SPM). Further details on the methodology were described in our previous study. An 8 mm FWHM Gaussian filter was applied to the spatially normalized PET images.
The binary images were created from reduction and increase regions compared with age-matched healthy participants, regions that showed a negative correlation between SUVR 30-50 and MADRS, and regions that showed a positive and negative correlation between ΔSUVR 30-50 and % improvement in MADRS. We identified the overlapped areas showing ketamine induced increase or decrease SUVR 30-50 at affected areas in patients with TRD.
A group comparison of SUVR 30-50 between TRD patients and age-matched healthy participants was performed using two-sample t-test. Comparison of SUVR 30-50 between before and after treatment with ketamine was using paired t-test. For the analysis of the association between ΔSUVR 30-50 and % improvement in MADRS, we generated ΔSUVR 30-50 images by subtracting pre-treatment from post-treatment SUVR 30-50 images. MADRS scores improvement were calculated as: -(MADRS post-treatment -MADRS pre-treatment) / MADRS pre-treatment × 100%. Associations between SUVR 30-50 and MADRS scores, ΔSUVR 30-50 and MADRS improvement, and pre-treatment SUVR 30-50 and MADRS improvement were assessed using multiple regression design implemented in SPM12. Covariates (age, sex) were included in the models for the analyses between SUVR 30-50 and MADRS scores, and between pre-treatment SUVR 30-50 and MADRS improvement, and group comparison, but not for the analysis between ΔSUVR 30-50 and MADRS improvement, as the subtraction process minimizes the influence of these covariates. To further assess potential confounding effects, we performed additional regression analyses between SUVR 30-50 and each of the demographic and clinical variables (age, sex, illness duration, benzodiazepine use, and number of failed antidepressant trials). These variables were then included as covariates in sensitivity analyses to confirm the robustness of our primary results. Statistical significance was set at p < 0.05 (peak-level uncorrected), false discovery rate (FDR) was corrected at p < 0.05 (cluster-level) for multiple comparisons across all in-mask voxels. To minimize false-positives, analyses were restricted to grey matter regions exceeding 10% probability based on standard tissue probability maps in SPM12. Additional correlation analysis and scatter plot were generated using GraphPad Prism 9 (Graph Pad Software, Massachusetts, USA). The assumption of homogeneity of variance for the group comparison was confirmed using Levene's test. All data fulfil the normality assumption.
The profiles of AMPAR in patients with TRD We have recently reported the phenotypes of the distribution of cell surface AMPAR in patients with depression that mainly consisted of patients with non-TRD using PET scanning with [ 11 C] K-2. In the current study, we focused on patients with TRD and performed PET scanning on them. We performed a regression analysis between standardized uptake value ratio (SUVR) using the whole brain as a reference during 30-to 50-minutes after [ 11 C]K-2 injection (SUVR 30-50 ) and illness severity using statistical parametric mapping (SPM). We found a negative correlation between SUVR 30-50 and the Montgomery Asberg Depression Rating Scale (MADRS) total score (higher values represent greater illness severity of depression) in the cortical areas including the left frontal, temporal, parietal, occipital lobes, and cerebellum (Fig., Supplemental Fig.and Supplemental Table), indicating that alteration in AMPAR density within these regions may contribute to depressive states ("state" regions). Most affected regions spanning frontal to parietal cortex were similar to those detected in our previous study. The occipital lobe and the cerebellum were newly identified regions specific to patients with TRD (Fig., Supplemental Fig.and Supplemental Table). Next, we compared SUVR 30-50 between patients with TRD and the healthy participants. We found that the decrease in SUVR 30-50 in TRD in the anterior insula, anterior to middle cingulate cortex, frontal, parietal, occipital lobes, and the increase in the cerebellum, temporal lobe, thalamus, posterior insula, and basal ganglia compared to the healthy participants (Fig., Supplemental Fig.and Supplemental Table). These regions were uniquely detected in patients with TRD, but not in patients with non-TRD. To assess the influence of demographic factors, we examined associations between SUVR 30-50 and age or sex within the TRD group. We found significant negative correlations between SUVR 30-50 and age (Supplemental Fig.and Supplemental Table), and significant increases and decreases in AMPAR density in male patients compared with female patients (Supplemental Fig.and Supplemental Table). Given these findings, age and sex were included as covariates in the above analyses. Additionally, within the TRD group, neither illness duration, number of failed antidepressant trials, nor benzodiazepine use showed significant correlations with baseline AMPAR density (all p > 0.05, uncorrected). Sensitivity analyses including these variables as additional covariates in the within-group models yielded essentially unchanged results, confirming the robustness of our primary findings (Supplemental Fig.).
Next, we investigated how ketamine administration alters the dynamics of AMPAR in patients with TRD. First, we compared the average SUVR 30-50 images of the first scan (before the treatment with ketamine) to those of the second scan (after the treatment with ketamine) and there was no significant difference between them using a paired t-test with SPM. Next, we performed a regression analysis between ΔSUVR 30-50 and improvement in depressive symptoms. We found the significant positive correlation between ΔSUVR 30-50 and % improvement in MADRS in some brain areas such as the parietal lobe, lobe, middle cingulate cortex, and a part of left frontal lobe in the ketamine group (Fig., Supplemental Fig.and Supplemental Table). Interestingly, we found brain areas such as the cerebellum, thalamus, left parahippocampal gyrus, right basal ganglia showed a significant negative correlation between ΔSUVR 30-50 and % improvement in MADRS in the ketamine group (Fig., Supplemental Fig.and Supplemental Table). Notably, in the habenula, there existed a negative correlation between ΔSUVR 30- 50 and % improvement in MADRS (Fig.). This suggested that ketamine-induced reduction in cell surface AMPAR in the habenula resulted in the antidepressant effect of ketamine. In contrast, there were no significant correlations in the placebo group (Fig.). Thus, the effect of ketamine on AMPAR dynamics, which relieves depressive symptoms of patients with TRD, could depend on the brain regions. As described above, we detected brain regions such as the parietal cortex and the occipital cortex that exhibited a negative correlation between SUVR 30-50 and the MADRS scores in patients with TRD (Fig.and Supplemental Fig.). These regions were partially overlapped with brain areas, such as the parietal, occipital lobes, and a part of the frontal lobe that exhibited positive correlations between ketamine administration-induced increase in ΔSUVR 30-50 and % improvement in MADRS (Fig., Supplemental Fig.and Supplemental Table). Next, we examined the effect of ketamine administration on the cell surface density of AMPAR in brain regions where we detected the difference between patients with TRD and healthy participants. We detected brain regions such as the parietal, occipital lobes, middle cingulate cortex, and a part of the left frontal lobe that exhibited reduced SUVR 30-50 in patients with TRD compared with healthy participants and presented with a positive significant correlation between ΔSUVR 30-50 and % improvement in MADRS by the ketamine administration (Fig., Supplemental Fig.and Supplemental Table). In addition, we also found brain regions such as the cerebellum, right basal ganglia which exhibited increased cell surface AMPAR density in patients with TRD compared with healthy participants and showed a negative significant correlation between ΔSUVR 30-50 and % improvement in MADRS by the ketamine administration (Fig., Supplemental Fig.and Supplemental Table). Additionally, sensitivity analysis was performed including age, sex, illness duration, benzodiazepine use and the number of failed medication trials as covariates. While the overall pattern of results remained essentially unchanged (Supplemental Fig., 7B, 8B, 8D), we observed some differences: negative correlations between ketamine-induced improvement in depressive symptoms and changes in AMPAR density emerged in the bilateral basal ganglia (Supplemental Fig.). These results are consistent with the concept that ketamine regulates the dynamics of AMPAR trafficking toward an antidepressant direction in patients with TRD. Cell surface distribution of AMPAR predicts a response to ketamine A significant positive correlation was found between SUVR 30-50 before ketamine administration and % improvement in MADRS by the ketamine administration in brain areas such as the frontal, temporal, parietal, occipital lobes, insula, cingulate cortex, and basal ganglia (Fig., Supplemental Fig.and Supplemental Table). We also found brain regions such as the left temporal, occipital lobes and a part of the parietal lobe with significant negative correlations between SUVR 30-50 before ketamine administration and % improvement in MADRS by the ketamine administration (Fig., Supplemental Fig.and Supplemental Table). Additionally, a sensitivity analysis was performed including age, sex, illness duration, benzodiazepine use and the number of failed medication trials as covariates. In this analysis, the positively correlated regions remained largely unchanged, whereas the negative correlations described in Fig.did not reach significance (Supplemental Fig.). Therefore, relative cell surface AMPAR density in some specific brain regions can determine the responsiveness to ketamine among patients with TRD.
In this study, we profiled AMPAR density on the cell surface in patients with TRD and how ketamine affected the dynamics of AMPAR in the living human brain. We detected cerebellar areas where we found a negative correlation between [ 11 C]K-2 uptake and MADRS score in patients with TRD (Fig.and Supplemental Figs.). In our previous paper, we reported that there existed brain areas in the cerebellum which exhibited a negative correlation between [ 11 C]K-2 uptake and Young Mania Rating Scale (YMRS) in patients with bipolar disorder. In this report, these two areas were not overlapped so that there exist cerebellar area-specific roles on emotional processing. In our previous study, which primarily included patients with non-TRD who had mild to moderate illness severity and showed a good response to conventional antidepressants, we detected no difference in [ 11 C]K-2 uptake between patients with depression and healthy participants. In contrast, we found remarkable differences between patients with TRD and healthy participants in this study (Fig.and Supplemental Fig.). Previously, we found regional differences in AMPA receptor density between psychiatric patients (e.g., bipolar disorder, schizophrenia and ASD) and healthy participants, and majority of these regions did not show significant correlations with symptom severity scores. We hypothesized that these areas may be representative of alterations more closely related to the disease process itself, or a "trait region". Similarly, we posit that patients with TRD share common trait-related regions across psychiatric disorders, which may lead them to present with distinct pathophysiology compared to non-TRD patients who experience a more normative treatment course. This supports the well-established notion that there are distinct biological processes subserving the treatment resistance trajectory. Part of these trait regions commonly exist across diseases (bipolar disorder, schizophrenia, ASD and TRD), while there exist disease specific trait regions. As was observed in the previous findings, we observed potential trait regions specific to TRD. Thus, while there exist common trait regions across psychiatric diseases, disease-specific trait regions also exist. It remains to be elucidated how these two types of traits emerge, what is the sequence of these events, and how these traits produce state regions. We found that the habenula was included in the brain area that showed a significant correlation between the ketamine-induced antidepressant effect (% improvement of MADRS) and the reduction of cell surface AMPAR density (Fig.). This is consistent with previous studies in experimental animals, which showed that ketamine administration in a depressive animal model reduced burst firing in the lateral habenula and that the changes in burst firing depended on AMPAR. Interestingly, [ 11 C]K-2 uptake in the habenula did not differ between TRD patients and healthy participants (Fig.and Supplemental Fig.), nor was it associated with baseline depressive severity (Fig.and Supplemental Figs.). A large body of evidence has been accumulated regarding the roles of the lateral habenula in depression in experimental animals, while it is still poorly elucidated how the lateral habenula contributes to the pathophysiology of depression in human. The lateral habenula has attracted wide attention in experiments with macaque monkey in which recordings of lateral habenula exhibited increased firing when animals fail to receive an expected reward or receive a cue predicting aversive stimuli. While this is an acute response to negative rewards, repetitive negative rewards can cause depression. Animal models often employ such repetitive negative reinforcement to induce depression. However, these models differ substantially from TRD patients in chronicity, etiology, and genetic variability. Notwithstanding such a complicated situation, the decrease in cell surface AMPAR density in response to ketamine in Fig.Overlapping regions where changes in AMPAR density correlate with clinical response to ketamine and regions where AMPAR density is altered in association with symptoms. A, Brain regions showing a significant negative correlation between SUVR 30-50 and the MADRS scores in patients with TRD (blue) (p < 0.05, t < -1.7, one-tailed, FDRc, adjusted covariate (sex and age)) and a significant positive correlation between ΔSUVR 30-50 and % improvement of MADRS in patients with TRD in the ketamine group (red) (p < 0.05, t > 1.77, one-tailed, FDRc, no adjusted covariate). Green region shows where the two regions overlap. Significant clusters and overlapping regions displayed on an axial, coronal, sagittal slices. B, Overlapping regions included the precuneus and the superior parietal cortex (x = 12, y = -66 and z = 60, Top) and zoomed view (arrow heads show the precuneus and the superior parietal cortex, Bottom). The (x, y, z) value represents MNI coordinates of represented sections. FDRc; false discovery rate correction, R; Right, L; Left, A; Anterior, P; Posterior. the habenula in patients with TRD in the present study is consistent with the acute effect of ketamine on the lateral habenula of experimental animals. This consistent finding is highly promising and critically important, considering that many biological outcomes differ between experimental animals and real-world human patients with depression in literature. Thus, biological events in the habenula can share the same mechanisms between experimental animals and patients with TRD at least in acute effects of ketamine. As presented in our results, we identified brain regions showing either a positive or negative significant correlation between ketamine-induced improvement (% improvement in MADRS) and AMPAR density changes in patients with TRD (Fig.and Supplemental Fig.). Brain areas with a positive correlation between ketamine-induced improvement and AMPAR density change were partially overlapped with brain regions where we detected a negative correlation between illness severity measured with MADRS and AMPAR density (Fig.and Supplemental Fig.). These overlapping regions included the precuneus and the superior parietal cortex. Recent LNM studies have suggested that multiple brain regions, with the left DLPFC serving as a central node, are involved in the pathophysiology of depression. Specifically, brain areas such as the precuneus and the superior parietal cortex, which exhibit strong positive functional connectivity with the DLPFC, have been reported to show reduced activity and connectivity in depression, correlating with depressive symptoms. Our current findings in the precuneus and superior parietal cortex align with these LNM observations, reinforcing the idea that these regions play a pivotal role in depression, especially in the context of ketamine's effect on AMPAR dynamics. The Fig.Overlapping regions where changes in AMPAR density correlate with clinical response to ketamine and regions where AMPAR density is different compared to healthy participants. A, Brain regions showing a significant relative reduction of SUVR 30-50 in patients with TRD compared to healthy participants (blue) (p < 0.05, t < -1.66, one-tailed, FDRc, adjusted covariate (age, sex)) and a significant positive correlation between ΔSUVR 30-50 and % improvement of MADRS in patients with TRD in the ketamine group (red) (p < 0.05, t > 1.77, one-tailed, FDRc, no adjusted covariate). Green region shows where the two regions overlap. Significant clusters and overlapping regions displayed on an axial, coronal, sagittal slices. B, Overlapping regions included the precuneus, the superior parietal cortex and the middle cingulate cortex (x = 12, y = -72 and z = 48, Top) and zoomed view (arrow heads show the precuneus and the superior parietal cortex, arrows show the middle cingulate cortex, Bottom). C, Brain regions showing a significant relative increase of SUVR 30-50 in patients with TRD compared to healthy participants (red) (p < 0.05, t > 1.66, one-tailed, FDRc, adjusted covariate (age and sex)) and a significant negative correlation between ΔSUVR 30-50 and % improvement of MADRS in patients with TRD in the ketamine group (blue) (p < 0.05, t < -1.77, one-tailed, FDRc, no adjusted covariate). Green region shows where the two regions overlap. Significant clusters and overlapping regions displayed on an axial, coronal, sagittal slices. D, Overlapping regions included the putamen and the pallidum (x = -18, y = 8 and z = 4, Top) and zoomed view (arrow heads show putamen and the pallidum, Bottom). The (x, y, z) value represents MNI coordinates of represented sections. FDRc; false discovery rate correction, R; Right, L; Left, A; Anterior, P; Posterior. precuneus is a core region of the DMN, but its role, whether contributing to increased or decreased functional connectivity after ketamine treatment, remains unclear. Given the positive correlation between AMPAR density and both short-range and long-range functional connectivity densities, AMPAR density may serve as a molecular determinant of functional connectivity. Therefore, our findings suggest that ketamine may increase AMPAR density in association with depressive symptom improvement within specific DMN regions, potentially shifting them toward a normalized functional connectivity. Additionally, brain areas, such as the precuneus, the superior parietal cortex and the left middle cingulate cortex, where a positive correlation between ketamine-induced improvement and change in AMPAR density was also partially overlapped with brain regions with reduced AMPAR density compared with healthy participants (Fig.and Supplemental Fig.). Notably, the precuneus and middle cingulate cortex are components of para-cingulate network, which is associated with reward anticipation and working memory. We also detected large brain regions in patients with TRD, including the putamen, the pallidum and the cerebellum, with increased AMPAR density than healthy participants. Some of these regions exhibited a negative correlation between ketamineinduced improvement and AMPAR density (Fig.and Supplemental Fig.). Among these, the pallidum plays a critical role in reward-based decision-making and provides input to the habenula. In particular, pallidal neurons projecting to the lateral habenula exhibit anti-reward characteristics. Given that ketamine inhibits pallidal activity, it may suppress lateral habenula hyperactivity, thereby activating reward-related neural circuits. These results indicate that ketamine regulates AMPAR dynamics in a direction that rescues the phenotype of cell surface AMPAR in patients with TRD. Our findings in the occipital cortex also warrant specific discussion. As shown in fMRI studies, patients with depression commonly report perceptual alterations, such as the "world appears colorless" or "foggy", which are thought to reflect dysfunction in the visual network within the occipital lobe. We observed that ketamine's antidepressant effects were positively correlated with increases in AMPAR density in the occipital lobe (Fig.and Supplemental Fig.). A MEG study similarly found that clinical response to ketamine in TRD participants was associated with gamma-band activation in the early visual cortex. Moreover, prior imaging studies have shown that ketamine modulates occipital lobe function, as measured by increased glucose metabolismand functional connectivity. Taken together, these findings suggest that ketamine may restore visual network dysfunction in depression through AMPAR-mediated plasticity in the occipital lobe, offering a possible neurobiological mechanism for its antidepressant efficacy. We also detected brain regions where there exists a positive or negative correlation between AMPAR density before the administration of ketamine and its subsequent antidepressant effect (Fig.and Supplemental Fig.). [ 11 C]K-2 may be a predictive tool for response to ketamine and may help guide treatment selection in patients with TRD. Although significant associations were observed between AMPAR density and demographic factors such as age and sex, the absence of significant correlations with clinical variables including illness duration, number of failed antidepressant trials, and benzodiazepine use suggests that the observed alterations are not secondary to illness chronicity or treatment history. While most results remained essentially unchanged, we observed some differences: negative correlations between ketamine-induced improvement in depressive symptoms and changes in AMPAR density emerged in the bilateral basal ganglia (Supplemental Fig.). In the analysis between pre-treatment SUVR 30-50 and % improvement in MADRS by the ketamine administration, positive correlations remained, but no significant negative correlations were observed (Supplemental Fig.). These findings suggest that the main conclusions of our study are robust to these potential confounders, although subtle regional differences may exist. This study has several limitations. First, the sample size of the parent RCTwas determined based on the clinical score changes in TRD. Therefore, it may not be optimal for imaging analysis, limiting the interpretability of our findings. Further studies with a larger and more sex-balanced sample are warranted to confirm our results. Second, this study was conducted exclusively in a Japanese population, which may restrict the generalizability of the results to other ethnic or demographic groups. Third, our findings were analyzed within the range of MADRS improvement observed in the parent RCT, and variations in this range may influence the results. Thus, future studies should include patients with more severe baseline depression to investigate AMPAR dynamics across a broader range of clinical severity changes. Fourth, this study did not use an active placebo to avoid its potential confounding neurobiological effects. However, the adverse events associated with ketamine may have led to unblinding for the participants and raters. Although subjective expectations or allocation guesses were not assessed, such factors could have contributed to variability in symptoms through placebo or nocebo effects, thereby influencing symptom outcomes. Nevertheless, the robust correlation between changes in AMPAR density and clinical improvement observed only in the ketamine group suggests that [ 11 C]K-2 PET signals can capture a biological association between AMPAR dynamics and the therapeutic effect of ketamine. Overall, our findings demonstrate that AMPAR distribution is altered in patients with TRD and that ketamine partially normalizes these abnormalities in association with its antidepressant effects. These findings indicate that AMPAR dynamics underlies the antidepressant effect of ketamine in patients with TRD, highlighting AMPARs as a potential effector and biomarker of treatment response.
As for one trial (jRCTs031210124), the trial protocol or the data that support the findings of this study are available from the corresponding author upon reasonable request. The Certified Review Board of Keio did not permit the deposit of raw data in a publicly accessible data archive or repository at the time of approval, as the procedure was not included in the protocol or informed consent document. Regarding the other two trials (UMIN000025132, jRCTs031200083), all requests for raw and analyzed data are promptly reviewed by the Yokohama City University Research Promotion Department to determine whether the request is subject to any intellectual property or confidentiality obligations and, further, inspected by the Institutional Review Board of Yokohama City University Hospital. Upon these approvals, derived data will be released via a material transfer agreement from the corresponding author.
Create a free account to open full-text PDFs.
Raja, S. M., Guptill, J. T., Mack, M. et al. · Clinical Pharmacology and Therapeutics (2024)
Zanos, P., Moaddel, P. J., Morris, P. J. et al. · Nature (2016)
Leal, G. C. · Journal of Affective Disorders (2023)
Mcintyre, R. S., Rosenblat, J. D., Nemeroff, C. B. et al. · American Journal of Psychiatry (2021)
Gilbert, J. R., Galiano, C. S., Nugent, A. C. et al. · Frontiers in Psychiatry (2021)