Inflammatory subgroups of schizophrenia and their association with brain structure: A semi-supervised machine learning examination of heterogeneity

Written by Katherine Reed

Background:

In schizophrenia patients, there has been evidence showing a spike in circulating cytokines before the first psychotic episode occurs. The inflammation that is accompanied by these cytokines, as well as acute proteins, is also shown to be correlated with a decrease in grey matter volume. Some of these different cytokines with pro-inflammatory qualities have also been associated with cognitive impairment in schizophrenic patients.

It is hypothesized that immune system dysfunction contributes to changes of brain structure via aberrant synaptic pruning in schizophrenia. Although the current data is mixed, it is found to be likely that inflammation increases could be found in only a sub-group of schizophrenia patients.

Current study:

In this study, it was tested that multiple inflammatory subgroups would be identified, as well as a decrease in gray matter volume in the brain accompanied by lower cognitive function in the elevated inflammatory subgroups.  Participants in an earlier stage of schizophrenia were chosen, from the Australia Schizophrenia Research Bank (ASRB) dataset, as to uncover possible stage-specific insights within the diagnosis. 1,067 participants were involved in this study, comprising 467 patients with chronic schizophrenia and 600 health controls.

            HYDRA (HeterogeneitY through DiscRiminant Analysis), a multivariate clustering and classification technique, was employed. HYDRA distinguished Health Control patients, and sub-categorized patients based on disease-based heterogeneity using cytokines g CRP, IFN-γ, IL-10, IL-12, IL1-β, IL-2, IL-6, IL-8, and TNFα. After the HYDRA output was obtained the inflammatory and neurocognitive profiles within subgroups were tested to highlight differences, in order to further distinguish the clusters that resulted from the HYDRA testing. Cytokine levels, demographics, and neurocognitive performance of identified clusters were compared using one-way ANOVA models corrected for multiple comparisons (Tukey’s HSD).

Results:

            Five main schizophrenia clusters were found in separation from the Healthy Controls: Low Inflammation, Elevated CRP, Elevated IL-6/IL-8, Elevated IFN-γ, and Elevated IL-10 with an adjusted Rand index of 0.573. Temporal and Hippocampal grey matter volume loss was evident in all schizophrenia clusters in comparison to the healthy controls. The cluster that showed the most widespread grey matter volume reduction, with anterior cingulate included, was cluster IL-6/IL-8. The least amount of grey matter volume loss, as well as the lowest cognitive function impairment, was shown by the IFN-γ inflammation cluster. The data also showed dominance of the CRP and Low inflammation clusters within the younger external dataset that was collected.

The relationship between inflammation and schizophrenia is more complicated than just a matter of having low or high inflammation; data is showing that pluripotent, heterogeneous mechanism involvement. These mechanisms can then potentially be detected via accessible, peripheral measures, may then be used to aid in the production of targeted interventions.

Looking Forward:

            Moving forward with future research, it should be noted that in this study, there was a loss of information due to the combination of two clusters (Low Inflammation and Anti-Inflammatory) that had similar mechanisms, as to enhance the available statistical power within the neurocognitive comparisons.  The use of an external database set in this study was noted as a strength, however, utilizing an external database comprising of patients with like illness characteristics could have been more beneficial for this study. In addition, while not included in this study, continued work should include cytokines belonging to the pathway IL-17. The cytokines within the IL-17 pathway are noted as important in the development of schizophrenia within patients.

References:

 Paris Alexandros Lalousis, Lianne Schmaal, Stephen J. Wood, Renate L.E.P Reniers, Vanessa L. Cropley, Andrew Watson, Christos Pantelis, John Suckling, Nicholas M. Barnes, Carmine Pariante, Peter B. Jones, Eileen Joyce, Thomas R.E. Barnes, Stephen M. Lawrie, Nusrat Husain, Paola Dazzan, Bill Deakin, Cynthia Shannon Weickert, Rachel Upthegrove. (2023). Inflammatory subgroups of schizophrenia and their association with brain structure: A semi-supervised machine learning examination of heterogeneity. Brain, Behavior, and Immunity, Vol. 113 166-175. DOI: https://doi.org/10.1016/j.bbi.2023.06.023 .