Validation of ketamine as a pharmacological model of thalamic dysconnectivity across the illness course of schizophrenia

N-methyl-D-aspartate receptor (NMDAR) hypofunction is a leading pathophysiological model of schizophrenia. Resting-state functional magnetic resonance imaging (rsfMRI) studies demonstrate a thalamic dysconnectivity pattern in schizophrenia involving excessive connectivity with sensory regions and deficient connectivity with frontal, cerebellar, and thalamic regions. The NMDAR antagonist ketamine, when administered at sub-anesthetic doses to healthy volunteers, induces transient schizophrenia-like symptoms and alters rsfMRI thalamic connectivity. However, the extent to which ketamine-induced thalamic dysconnectivity resembles schizophrenia thalamic dysconnectivity has not been directly tested. The current double-blind, placebo-controlled study derived an NMDAR hypofunction model of thalamic dysconnectivity from healthy volunteers undergoing ketamine infusions during rsfMRI. To assess whether ketamine-induced thalamic dysconnectivity was mediated by excess glutamate release, we tested whether pre-treatment with lamotrigine, a glutamate release inhibitor, attenuated ketamine’s effects. Ketamine produced robust thalamo-cortical hyper-connectivity with sensory and motor regions that was not reduced by lamotrigine pre-treatment. To test whether the ketamine thalamic dysconnectivity pattern resembled the schizophrenia pattern, a whole-brain template representing ketamine’s thalamic dysconnectivity effect was correlated with individual participant rsfMRI thalamic dysconnectivity maps, generating “ketamine similarity coefficients” for people with chronic (SZ) and early illness (ESZ) schizophrenia, individuals at clinical high-risk for psychosis (CHR-P), and healthy controls (HC). Similarity coefficients were higher in SZ and ESZ than in HC, with CHR-P showing an intermediate trend. Higher ketamine similarity coefficients correlated with greater hallucination severity in SZ. Thus, NMDAR hypofunction, modeled with ketamine, reproduces the thalamic hyper-connectivity observed in schizophrenia across its illness course, including the CHR-P period preceding psychosis onset, and may contribute to hallucination severity.

averages of WM and CSF voxel timeseries. The subset of significant (p < .05) noise components, for each run, was selected using a Monte Carlo simulation 5 and retained for subsequent use as nuisance variables in the first-level connectivity analyses (described below). After implementing ART and aCompCor, FEAT was used to carry out high-pass temporal filtering (100 sec 8 ) and spatial smoothing (6 mm kernel).
For each participant, the following nuisance regressors were entered in the first-level connectivity model: (i) seven motion parameters from ART (i.e., temporal derivatives of six motion parameters and the composite of total motion across translation and rotation), (ii) vectors flagging single outlier data points identified with ART to censor them, (iii) significant noise components from aCompCor, and (iv) task-condition event vectors convolved with the hemodynamic response function to regress out task-related activity.
Correlations between ketamine-induced thalamic dysconnectivity and ketamine-induced positive and negative symptoms in healthy male volunteers As noted above, ketamine induced schizophrenia-like symptoms in a sample of healthy men. In separate group-level models using FSL's FEAT (https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FEAT/), we assessed whether ketamine-induced thalamic dysconnectivity effects correlated with ketamine-induced increases in positive or negative symptoms (assessed with the BPRS). For each participant, we created positive and negative symptom change scores that mimicked the between-day, within-day double contrasts from the main grouplevel analysis (e.g., positive symptoms during [active ketamine -saline] > [placebo ketamine -saline]); the resulting change scores were mean-centered. Subsequently, we regressed whole brain ketamine-induced thalamic dysconnectivity maps on ketamine-induced positive and negative symptom change scores in separate voxel-wise analyses. We used a voxel-wise height threshold of z > 3.29 (p < .001), and a more stringent corrected cluster significance threshold of p < .025, two-tailed, to account for the separate positive and negative symptom models (i.e., p < .05/2). These results were not significant, indicating that psychotomimetic effects in healthy volunteers were unrelated to thalamic connectivity changes.
analogous to the healthy control participants (HC) z-score adjustments we used in the clinical data sets. The resulting distributions are shown in Figure S7a.
These analyses show, as expected, that ketamine similarity coefficients were higher in the ketamine study participants (mean = 0.39, SD = 0.13) than in SZ (mean = 0.18, SD = 0.20) or ESZ (mean = 0.11, SD = 0.15), but that the variances were generally similar across the data sets.
We also note that the variance of similarity coefficients did not differ between HC and SZ in the Schizophrenia versus Healthy Control Study clinical data set (F177,182 = 1.11, p = .48; Figure S7b). For ESZ and CHR-P from the Early Illness Schizophrenia, Clinical High-risk, versus Healthy Control Study clinical data set ( Figure S7c), there was no significant difference in variance between HC and ESZ (F84,73 = 1.44, p = .11), or ESZ and CHR-

Schizophrenia versus Healthy Control Study
Additional participant details SZ met diagnostic criteria for schizophrenia based on the Structure Clinical Interview for DSM-IV-TR. 12 SZ and HC were excluded given a history of major medical illness, MRI contraindications, drug dependence in the past five years or current substance abuse, or an IQ less than 75. SZ with movement disorder symptoms were excluded (e.g., significant extra-pyramidal symptoms). Additionally, HC participants were excluded given a current or past history a of major neurological or psychiatric disorder, or a first-degree relative with a psychotic disorder diagnosis. The study protocol was approved by the Institutional Review Boards at the University of California, Irvine, the University of California, Los Angeles, the University of California, San Francisco, Duke University, the University of North Carolina, the University of New Mexico, the University of Iowa, and the University of Minnesota. Written informed consent was obtained from all participants; this included permission to share-de-identified data across study sites and with the broader research community.

Neuroimaging data denoising
As described in the Ketamine -Lamotrigine Study, ART and aCompCor were used to identify outlier volumes and noise components. Motion parameters from ART, vectors flagging outlier datapoints identified by ART, and significant noise components from aCompCor were entered as nuisance regressors in the first-level connectivity models.
Overlap between ketamine-induced thalamic dysconnectivity with schizophrenia thalamic dysconnectivity In addition to the computation of ketamine similarity coefficients, we qualitatively evaluated the resemblance of the ketamine thalamic dysconnectivity pattern with the thalamic dysconnectivity pattern observed in schizophrenia. To this end, we performed a conjunction analysis of the group-level [active ketamine -saline] > [placebo ketamine -saline] contrast and the group-level SZ > HC contrast (previously reported in 6 ). Both contrast maps were converted into z-maps, warped into MNI space, and thresholded (voxel-z > 3.29, corrected cluster-p < .05) using the same procedures to enable a direct comparison. Next, we identified which voxels passed this significance threshold for both contrasts. Figure S2 shows the conjunction map, overlaid on the separate ketamine-induced thalamic dysconnectivity and schizophrenia thalamic dysconnectivity maps. We observed overlap in several sensory regions, including the superior temporal gyrus, pre and postcentral gyrus, lingual gyrus, and superior parietal lobule.
Ketamine similarity coefficients and sex Because our ketamine template map was generated from only male participants, we tested for sex differences in the ketamine similarity coefficients derived from that template. Specifically, we used an ANOVA to test for sex differences in similarity coefficients between HC and SZ, while controlling for study site. There was no significant interaction between group and sex when predicting similarity coefficients (F1,351 = 1.63, p = .20), nor a main effect of sex after removing the non-significant interaction term (F1,352 = 0.09, p = .77). Figure S6 shows that the similarity coefficient means were similar across sexes for both groups.   Abbreviations: CHR-P, clinical high-risk for psychosis participant.  (voxel-z > 3.29, corrected cluster-p < .05); anatomical location details for clusters 1 through 7 are found in Table   S4. Asterisks reflect significance levels from follow-up pairwise tests (based on the corresponding repeated measures ANOVA for that cluster).