Pharmacological fMRI: Effects of subanesthetic ketamine on resting-state functional connectivity in the default mode network, salience network, dorsal attention network and executive control network

Mueller and colleagues contextualise their work within two parallel lines of investigation: first, subanesthetic doses of the NMDAR antagonist S-ketamine produce transient behavioural effects in healthy people that resemble positive and negative symptoms of acute psychosis, making ketamine a widely used pharmacological model of glutamatergic dysfunction in schizophrenia; second, resting-state functional MRI (RS-fMRI) permits non-hypothesis-driven characterisation of intrinsic brain networks and their connectivity, which may reveal neural mechanisms underlying those behavioural effects. Previous RS-fMRI studies of ketamine have reported heterogeneous findings, including increases or decreases of global or region-specific connectivity, variable effects across networks such as the default mode network (DMN), salience network (SN) and executive control network (ECN), and inconsistencies that may relate to dose, infusion protocol, seed choice and timing of acquisition. The authors note an unmet need to identify imaging markers that reliably index ketamine-induced psychotomimetic effects and that could serve as surrogate outcome measures in dose–response or translational studies. This study sets out to test whether subanesthetic S-ketamine alters functional connectivity within four large resting-state networks implicated in schizophrenia—the DMN, dorsal attention network (DAN), ECN and SN—and whether connectivity changes relate to concurrent clinical symptom changes. Using a double-blind, randomised, placebo-controlled cross-over design in healthy male volunteers, the investigators performed seed-to-voxel RS-fMRI analyses and assessed whether ketamine-induced connectivity changes could accurately discriminate drug versus placebo (via receiver operating characteristic analysis) and correlate with changes on the Positive and Negative Syndrome Scale (PANSS) and the 5D-ASC altered-states questionnaire. The aim was to evaluate whether such connectivity changes might qualify as candidate functional biomarkers for ketamine-induced side effects, particularly negative symptoms.

The study was conducted as a Phase I, randomised, double-blind, placebo-controlled cross-over trial in healthy male volunteers, approved by local ethics authorities. Twenty-four men were initially enrolled after medical, neurological and psychiatric screening; after dropouts and exclusions for missing or prematurely terminated RS-fMRI sessions, 17 participants (mean age 27.4 years, range 21–35, SD 4.42) contributed data to the analyses. Exclusion criteria included age outside 18–35 years, past medical/neurological/psychiatric disease, recent medication, drug abuse history, family history of schizophrenia, claustrophobia, metal implants and excessive head motion (>3 mm translation) during EPI acquisition. Each subject completed two scan sessions at least one week apart, receiving intravenous S-ketamine on one day and saline placebo on the other in randomised counterbalanced order. The infusion protocol used a bolus of 0.1 mg/kg followed by continuous infusion at 0.015625 mg/kg/min (maximum one hour) with a 10% dose reduction every 10 minutes to approximate stable plasma levels; ketamine plasma concentrations were not measured in the experiment, but the authors report simulation-based estimates (using STANPUMP and the Domino model) suggesting plasma levels between ~0.25 and <0.48 µg/ml. Psychopathology assessments (PANSS and the 5D-ASC subscales) were performed immediately before and after each scan. Imaging was performed at 3 T. The RS-fMRI session comprised 20 minutes of echo-planar imaging (33 slices, voxel size 3×3×3 mm, TR 3400 ms); because some ketamine sessions were prematurely terminated, all sessions were truncated to the first 150 volumes (8 min 30 s) for analysis, with the first five volumes discarded. Preprocessing in SPM12 included slice timing correction, realignment, coregistration, tissue segmentation, normalisation to MNI space and spatial smoothing (8 mm FWHM). Motion parameters and rotation were inspected and compared between conditions. Connectivity was analysed using the CONN toolbox (seed-to-voxel bivariate Pearson correlations). Seed ROIs were 6 mm radius spheres placed at predefined MNI coordinates: posterior cingulate cortex (PCC) for the DMN; left/right intraparietal sulcus for the DAN; left/right dorsolateral prefrontal cortex (DLPFC) for the ECN; and left/right fronto-insular cortex for the SN. Standard denoising steps in CONN were applied (white matter/CSF regression, CompCor, despiking, linear detrending, band-pass filtering 0.008–0.09 Hz), and session-specific realignment parameters were included as covariates. Second-level voxel-wise paired t-tests contrasted ketamine versus placebo; statistical thresholds were cluster-extent FDR corrected at p < 0.05 with an uncorrected voxelwise height threshold p < 0.001. To address multiple seed tests, the authors applied a Bonferroni-adjusted cluster threshold of p < 0.0125 (0.05/4). For clusters surviving that threshold, subject-level Fisher z-scores were extracted and ketamine-minus-placebo (delta) values computed. Discriminative accuracy of delta connectivity values was assessed with receiver operating characteristic (ROC) curves and area under the curve (AUC) metrics; AUC reflects the trade-off between sensitivity and specificity. Clinical delta scores (post–pre) on PANSS and 5D-ASC subscales were tested for ketamine–placebo differences using paired t-tests, and correlations between significant connectivity deltas and clinical deltas were evaluated with two-tailed Pearson correlations using Bonferroni-adjusted significance (p < 0.00625 for eight tests). The authors also tested whether differences in session length correlated with clinical changes to rule out confounding by prematurely terminated scans.

Participant flow and data inclusion: of 24 enrolled men, two subjects dropped out during their first session because of adverse events (one during ketamine, one during placebo) and five additional subjects were excluded because at least one RS-fMRI session was not acquired or was unusable, leaving 17 subjects for analysis. Group-level comparisons showed no significant difference in mean head motion or mean rotation between ketamine and placebo (motion p = 0.967, rotation p = 0.526). Seed-to-voxel contrasts (ketamine > placebo and ketamine < placebo) revealed network-specific effects. For the ECN (DLPFC seed), ketamine increased connectivity with left anterior cingulum and left superior frontal gyrus, and with right medial temporal lobe, right angular gyrus and right superior temporal lobe; ketamine decreased ECN connectivity with the calcarine fissure bilaterally. After Bonferroni correction, increased ECN connectivity with left anterior cingulum and left superior frontal gyrus and decreased ECN connectivity with left calcarine fissure remained significant. For the SN (fronto-insular seed), no significant increases were observed, but decreased connectivity with the right calcarine fissure survived Bonferroni correction, and decreased connectivity with left frontal gyrus was also reported. The DAN (intraparietal sulcus seed) showed increased connectivity with the left precentral gyrus under ketamine, with no significant decreases. The DMN (PCC seed) exhibited increased connectivity with the superior parietal lobule and decreased connectivity with the left Rolandic operculum under ketamine; these DMN changes did not survive the stricter Bonferroni cluster threshold. The authors calculated ROC curves for the four contrasts that survived Bonferroni correction to test the ability of delta connectivity to distinguish ketamine from placebo. Reported AUCs were: ECN–anterior cingulum 0.934, ECN–superior frontal gyrus 0.869, ECN–calcarine fissure 0.818, and SN–calcarine fissure 0.862, indicating high sensitivity and specificity in this sample. Clinical measures: all tested PANSS and 5D-ASC subitems increased significantly after ketamine relative to placebo. Correlation testing identified three significant associations between connectivity changes and symptom changes: negative symptoms on the PANSS correlated with ECN–left calcarine fissure connectivity and with SN–right calcarine fissure connectivity (direction described as a positive correlation between functional connectivity and clinical score). No significant correlations were reported between session length differences and clinical score changes, suggesting prematurely terminated sessions were not the principal driver of symptom–connectivity correlations.

Mueller and colleagues interpret their principal findings as demonstrating that subanesthetic S-ketamine induces network-specific RS-fMRI changes in healthy men, with the most robust effects in the ECN and SN after correction for multiple testing. Increased ECN connectivity, notably between DLPFC seeds and left anterior cingulum and superior frontal gyrus, is taken to reflect enhanced prefrontal connectivity. The investigators propose a physiological mechanism in which NMDAR blockade on GABAergic interneurons leads to pyramidal neuron disinhibition in prefrontal cortex, consistent with prior animal data showing increased prefrontal metabolic activity after ketamine. Conversely, ketamine decreased ECN and SN coupling with the calcarine fissure (primary visual cortex). The authors suggest this may indicate reduced visual information processing or impaired NMDAR-dependent transmission between frontal and occipital regions; they caution that calcarine findings are susceptible to motion and signal artefact because of proximity to the skull, although no group-level motion differences were detected. Importantly, decreased SN–calcarine connectivity correlated with greater PANSS negative symptoms, which the authors interpret as consistent with a model in which inhibition of visual input to attention/salience systems may relate to reduced reactivity or withdrawal-type negative symptoms. They emphasise that positive symptoms did not correlate with the cortical connectivity changes observed here, proposing that positive symptoms may arise from excitation or disinhibition in other brain regions (for example subcortical structures) not captured by their cortical seed set. The discussion places the results alongside prior ketamine RS-fMRI work. Differences from other studies are ascribed to methodological heterogeneity—dose, infusion method, sedation level, seed selection and whether whole-brain or ROI-based analyses were used. For example, some studies have reported global shifts in connectivity towards subcortical structures or ketamine-associated DMN alterations; by contrast, this study's seed-based cortical focus detected ECN increases and SN decreases. The authors acknowledge the limitations that may affect interpretation: the all-male sample limits generalisability, several ketamine sessions were prematurely terminated and sessions were truncated to a common length which may bias against the most strongly affected subjects, ketamine plasma levels were not measured, infusion dose was relatively low compared with some prior studies, vital signs were monitored but not synchronised with fMRI, and the seed-to-voxel approach confined analysis to predefined cortical ROIs and may not represent full network dynamics. They conclude that while the observed decreased SN–calcarine connectivity correlates with negative symptoms and shows good discriminative accuracy in this dataset, these findings require replication in larger and more varied samples and with more comprehensive measurement (including plasma levels and broader brain coverage) before being used as a validated surrogate biomarker.

The study reports that subanesthetic S-ketamine produced significant RS-fMRI effects in the executive control network and salience network in healthy male participants, with high discriminative accuracy between ketamine and placebo for several seed–cluster contrasts. Although ketamine increased several PANSS and 5D-ASC symptom measures, the principal neuroimaging finding highlighted by the authors is decreased connectivity between the salience network and the calcarine fissure, which correlated with negative symptom changes. The investigators suggest this specific connectivity reduction could be a candidate surrogate imaging biomarker for ketamine-induced negative symptoms in future drug trials, but they note that further research is needed to establish robustness and generalisability.