Inflammatory molecules called cytokines may be a peripheral biomarker of Parkinson’s progression, especially of its characteristic changes in motor abilities, according to new research using machine learning.
The study, “Parkinson’s progression prediction using machine learning and serum cytokines,” appeared in the journal npj Parkinson’s Disease.
Besides altered immune responses and distinct populations of immune cells, increased levels of cytokines — small proteins secreted by cells of the immune system — may link inflammation with Parkinson’s. This has been seen in asymptomatic carriers of the G2019S mutation in the LRRK2 gene, which accounts for 1–5% of all Parkinson’s cases. Still, the extent to which peripheral cytokines may trigger the disease remains unclear.
Increasing evidence suggests that Parkinson’s may start in the periphery (for example, in the gut), possibly enabling the identification of markers of disease progression that could improve outcomes for patients and lead to better trial design.
Researchers at The University of Sydney, Australia, used machine learning to further assess the correlation between peripheral inflammatory cytokines and Parkinson’s symptoms. They analyzed serum samples from 160 patients (mean age 68–69, ages 57–58 at diagnosis), 80 of whom (40 men) had the G2019S mutation and 80 who did not (54 men), all followed within the Michael J Fox Foundation Parkinson’s Progression Markers Initiative. They then used machine learning models to predict clinical outcomes at two years.
Comparing the two groups, patients who carried the G2019S mutation had milder motor disease and less severe hyposmia — a reduced sense of smell — as assessed with the Unified Parkinson’s Disease Rating Scale part 3 (UPDRS-III) and the University of Pennsylvania smell identification test. At baseline (study start), mutation carriers also had higher levels of the cytokines PDGF and MCP1 than those in the group of idiopathic (of unknown cause) Parkinson’s disease.
One year later, the scientists assessed blood serum samples from 126 of these patients. Results revealed that two cytokines, GCSF and interleukin (IL)-5, had the greatest variation. However, only the levels of one cytokine, IL-1RA, differed between the two groups.
Clinically, the patients showed more severe motor symptoms and depression, which were associated with a significantly decreased (worse) score in the Schwab and England activities of daily living (ADL) scale.
A subsequent analysis showed that, among a subset of 76 patients, higher baseline levels of 14 cytokines correlated with greater (worse) findings on the geriatric depression scale over two years. A similar link was found between seven cytokines and motor function. IL-5 and GCSF were among the cytokines whose levels correlated with both scales.
Using machine learning, researchers observed that two cytokines, MIP1 alpha and MCP1, made the biggest peripheral contribution to predicting motor symptom severity using the Hoehn and Yahr and UPDRS III scales, respectively.
As such, higher levels of these molecules were associated with faster motor deterioration.
In turn, the cytokines IL-6 and IL-4 were the primary contributors to predicting geriatric depression. All top cytokine contributors were also good predictors of the Schwab and England ADL scale scores.
Using cytokines improved predictions by 20% over clinical data alone. As for other analyzed variables, age and gender were among the top 10 contributors to predicting ADL and UPDRS-III scores, respectively.
“These results provide information on the longitudinal assessment of peripheral inflammatory cytokines in [Parkinson’s] and give evidence that peripheral cytokines may have utility for aiding prediction of [Parkinson’s] progression,” the scientists wrote.
Future studies should use a larger and more diverse group of patients, and assess the potential impact of medications on both clinical outcomes and cytokines levels, the researchers added.