The Neuroimmune and Neurotoxic Fingerprint of Major Neurocognitive Psychosis or Deficit Schizophrenia: a Supervised Machine Learning Study

Abstract

No studies have examined the immune fingerprint of major neurocognitive psychosis (MNP) or deficit schizophrenia using M1 macrophage cytokines in combination with chemokines such as CCL2 and CCL11. The present study delineated the neuroimmune fingerprint of MNP by analyzing plasma levels of IL-1β, sIL-1RA, TNFα, sTNFR1, sTNFR2, CCL2, and CCL11 in 120 MNP versus 54 healthy controls in association with neurocognitive scores (as assessed with the Brief Assessment of Cognition in Schizophrenia) and PHEMN (psychotic, hostility, excitation, mannerism and negative) symptoms. MNP was best predicted by a combination of CCL11, TNFα, IL-1β, and sIL-1RA which yielded a bootstrapped (n = 2000) area under the receiver operating curve of 0.985. Composite scores reflecting M1 macrophage activity and neurotoxic potential including effects of CCL11 and CCL2 were significantly increased in MNP. A large part of the variance in PHEM (38.4-52.6%) and negative (65.8-74.4%) symptoms were explained by combinations of immune markers whereby CCL11 was the most important. The same markers explained a large part of the variance in the Mini-Mental State examination, list learning, digit sequencing task, category instances, controlled word association, symbol coding, and Tower of London. Partial least squares analysis showed that 72.7% of the variance in overall severity of schizophrenia was explained by the regression on IL-1β, sIL-1RA, CCL11, TNFα, and education. It is concluded that the combination of the abovementioned markers defines MNP as a distinct neuroimmune disorder and that increased immune neurotoxicity determines memory and executive impairments and PHEMN symptoms as well.

Keywords: Cognition; Cytokine; Deficit schizophrenia; Immunological biomarkers; Machine learning.