Johanna Seitz-Holland*, Cornelius Berberich, Nora Penzel, Kevin Cho Kang-Ik, Ofer Pasternak, Suheyla Cetin-Karayumak, Yogesh Rathi, Sylvain Bouix, Mike J Coleman, Twishi Puri, Simone Veale, Elana Kotler Bayla, Nicholas Kim, Oemer Faruk Oeztuerk, Grace Jakobs, Amanda Lyall, Christian Sorg, Breno S. Diniz, Kathryn E. Lewandowski, Daphne Holt, Matcheri S. Keshavan, Dost Öngür, Alan Breier, Martha E. Shenton, Marek KubickiBrigham and Women’s Hospital, Harvard Medical School, Boston, Massachusetts, United States

Background: Research in recent years suggested the importance of transdiagnostic phenotyping when characterizing and treating mental disorders. Specifically, psychotic illnesses show high symptomatic overlap across diagnoses and common underlying alterations in brain networks. The Human Connectome Project for Early Psychosis (HCP-EP) was designed in the Research Domain Criteria (RDoc) spirit to study domains of psychopathology across the early phase of several psychotic disorders and integrate this with the HCP approach to acquire high-quality imaging data. The present study investigates cognitive and psychosocial functioning, white matter structure, brain volumes, and protein markers across the psychosis spectrum.

Methods: The HCP-EP study was carried out across four recruiting and three scanning sites. It recruited individuals within five years of the onset of a DSM-5 diagnosis of psychosis (schizophrenia, schizophreniform disorder, schizoaffective disorder, psychosis not otherwise specified, delusional disorder, brief psychotic disorder, major depression with psychosis, or bipolar disorder with psychosis). Individuals underwent extensive testing, including NIH Toolbox, a structural and diffusion-weighted magnetic resonance imaging (MRI) scan, and a blood draw.

Here, we analyzed 203 individuals with preprocessed and harmonized diffusion-weighted MRI data, of which 202 had a structural MRI, 182 had complete cognitive data, and 82 had protein markers analyzed by a commercial platform (O-link). First, we grouped individuals based on diagnosis (non-affective psychosis = 136, affective psychosis = 67). Then we run stepwise cluster analyses with internal validation. Clusters were based on 2) seven cognitive tests (Picture Vocabulary, Flanker Inhibitory, List Sorting Working Memory, Dimensional Change Card Sort, Pattern Comparison Processing Speed, Picture Sequence Memory, Oral Reading Recognition), and 3) clusters derived from protein markers previously associated with cell aging, inflammation, and cardiometabolic health (IGFBP2, TIMP1, GDF 15, MMP 1, MMP2, CTSD, CST3, CTSZ, ALCAM, IGFBP7).

For groups based on 1) diagnosis, 2) cognition, and 3) protein markers, we compared general functioning (GAF), NIH Toolbox measures of Perceived Stress, Life Satisfaction, Meaning and Purpose, and Self-Efficacy, voxel-wise cellular and extracellular white matter structure (fractional anisotropy and free water), and gray matter volumes (hippocampus, superior frontal, rostral middle frontal, medial orbital frontal, pars opercularis, middle temporal, and fusiform gyrus). Imaging analyses were corrected for age, age2, sex, and motion for diffusion-weighted analyses and age, age2, sex, site, and total intracranial volume for volume analyses.

Results: We derived two clusters with good stability and prediction strength based on cognitive measures: cluster 1 with lower (n = 82) and cluster 2 with higher performance (n = 100) and two protein-based clusters. The protein clusters were driven by a higher TIMP1, MMP1, and ALCAM expression in cluster 2 (n = 24) compared to cluster 1 (n = 38). There were more individuals with affective psychosis in cognitive cluster 2 (Χ2(1) = 5.70, V = 0.19, p = 0.017), but no association was found between diagnosis or cognitive clusters and protein-based clusters.

The two diagnostic groups demonstrated group differences in the GAF Social Functioning Scale (T(198) = −3.40, d = 0.51, p < 0.01). Individuals with non-affective psychosis also presented with higher fractional anisotropy and lower free water than individuals with affective psychosis (p < .005).

Groups based on cognitive clusters demonstrated lower Meaning and Purpose (T(176) = 3.38, d = 0.51, p < 0.01), a trend toward more Social Functioning (T(179) = −2.06, d = 0.31, p = 0.041) and Self Efficacy (T(180) = −2.23, d = 0.33, p = 0.027), and higher fusiform gyrus volumes (F(1, 176) = 16.69, p < .001) in cluster 2.

Protein-based clusters showed no differences in psychosocial functioning but demonstrated more extracellular free water in cluster 2 (p < .005).

Conclusions: Our findings demonstrated the complex interplay between diagnosis, cognition, and biological-based phenotypes for the pathophysiology of psychosis. Notably, we found that individuals with early psychosis can robustly be grouped based on their cognitive functioning and protein markers. While diagnoses are related to social functioning, cognition seems more indicative of self-reported measures of Efficacy, Meaning, and Purpose, all highly relevant for the quality of life but traditionally neglected in psychosis research. While we found few overall differences in brain structure, lower volume of the fusiform gyrus, a structure previously implicated in bipolar disorder, was related to poorer cognition. On the other hand, extracellular white matter abnormalities were related to a higher expression of proteins associated with cell aging, inflammation, and cardiometabolic health. Interestingly the three main proteins TIMP1, MMP1, and ALCAM are involved in extracellular matrix remodeling and blood-brain barrier regulation, which, together with the imaging findings, support the notion of a diagnostic independent inflammatory subtype of psychosis. Future studies should integrate the measures introduced in the present analyses to validate findings and examine their potential for treatment stratification and monitoring.

Keywords: Early Psychosis, Research Domain Criteria (RDoC), Multimodal Neuroimaging, Proteomics, Psychosocial Outcomes

Disclosure: Nothing to disclose.