Systemic cytokines and GlycA discriminate disease status and predict corticosteroid response in HTLV-1-associated neuroinflammation

Abstract

Abstract Background HTLV-1-Associated Myelopathy/Tropical Spastic Paraparesis (HAM/TSP) is an incapacitating neuroinflammatory disorder for which no disease-modifying therapy is available, but corticosteroids provide some clinical benefit. Although HAM/TSP pathogenesis is not fully elucidated, older age, female sex and higher proviral load are established risk factors. We investigated systemic cytokines and a novel chronic inflammatory marker, GlycA, as possible biomarkers of immunopathogenesis and therapeutic response in HAM/TSP, and examined their interaction with established risk factors. Patients and methods We recruited 110 People living with HTLV-1 (PLHTLV-1, 67 asymptomatic individuals and 43 HAM/TSP patients) with a total of 946 person-years of clinical follow-up. Plasma cytokine levels (IL-2, IL-4, IL-6, IL-10, IL-17A, IFN-γ, TNF) and GlycA were quantified by Cytometric Bead Array and 1NMR, respectively. Cytokine signaling and prednisolone response were validated in an independent cohort by nCounter digital transcriptomics. We used multivariable regression, machine learning algorithms and Bayesian network learning for biomarker identification. Results We found that systemic IL-6 was positively correlated with both age (r = 0.50, p < 0.001) and GlycA (r = 0.45, p = 0.00049) in asymptomatics, revealing an ‘inflammaging” signature which was absent in HAM/TSP. GlycA levels were higher in women (p = 0.0069), but cytokine levels did not differ between the sexes. IFN-γ (p = 0.007) and IL-17A (p = 0.0001) levels were increased in untreated HAM/TSP Multivariable logistic regression identified IL-17A and proviral load as independent determinants of clinical status, resulting in modest accuracy of predicting HAM/TSP status (64.1%), while a machine learning-derived decision tree classified HAM/TSP patients with 90.7% accuracy. Pre-treatment GlycA and TNF levels significantly predicted clinical worsening (measured by Osame Motor Disability Scale), independent of proviral load. In addition, a poor prednisolone response was significantly correlated with higher post-treatment IFN-γ levels. Likewise, a transcriptomic IFN signaling score, significantly correlated with previously proposed HAM/TSP biomarkers (CASP5/CXCL10/FCGR1A/STAT1), was efficiently blunted by in vitro prednisolone treatment of PBMC from PLHTLV-1 and incident HAM/TSP. Conclusions An age-related increase in systemic IL-6/GlycA levels reveals inflammaging in PLHTLV-1, in the absence of neurological disease. IFN-γ and IL-17A are biomarkers of untreated HAM/TSP, while pre-treatment GlycA and TNF predict therapeutic response to prednisolone pulse therapy, paving the way for a precision medicine approach in HAM/TSP.