Immune cell profiles distinguish patients with juvenile-onset systemic lupus erythematosus (SLE) from healthy individuals and can be used for patient stratification of disease severity, according to study results published in Lancet Rheumatology.

Using machine-learning approaches, the investigators aimed to characterize the link between immune cell profiles of patients with juvenile-onset SLE and disease outcomes.

Peripheral blood mononuclear cells were isolated from 67 patients with juvenile-onset SLE (81% girls) and 39 healthy control participants (56% girls). Frequencies of 28 immune cell subsets were assessed using flow cytometry. Prognostic parameters of the immune cell profiles were identified using a balanced random forest machine learning algorithm, a sparse partial least squares-discriminant analysis, and logistic regression. The identified parameters were then used to perform k-means clustering and combined with clinical features for patient stratification into disease trajectories.

Compared with healthy control participants, patients with juvenile-onset SLE had disrupted immune cell profiles, characterized by increases in total and naive CD8 T cells, total monocytes, and plasmablasts, as well as decreases in total CD4 T cells and memory B and T cell populations. Many of the associations that existed between immune cell populations in the healthy control participants were inverted or exacerbated in juvenile-onset SLE. 


Continue Reading

Machine learning was used to generate a predictive model based on the immune cell profiles. After optimization, the model distinguished patients with juvenile-onset SLE from the healthy control participants with a classification accuracy of 86.8%. The diagnostic sensitivity was 89.6% and specificity was 82.1%. After 10-fold cross-validation analysis, the classification accuracy remained 87.8%, and the sensitivity and specificity were 89.6% and 84.7%, respectively.

The top contributing factors that segregated patients with juvenile-onset SLE from healthy control participants were CD19 unswitched memory B cells, Bm1 B cells, and CD14 monocytes. Using logistic regression analysis, 12 out of 28 immune cell subsets were significantly associated with juvenile-onset SLE, including reduced frequency of CD19 unswitched memory B cells (odds ratio, 0.71; 95% CI, 0.60-0.82).

In a sparse partial least squares-discriminant analysis, CD19 unswitched memory B cells were also identified as the highest weighted parameter for discriminating between healthy controls and juvenile-onset SLE patients, followed by Bm1 B cells. The top discriminating factors matched those identified by both logistic regression and machine learning approaches.

Using k-means clustering, 8 of the top discriminating immune cell subsets (total CD4, total CD8, CD8 effector memory, CD8 naive, and invariant natural killer T cells; Bm1 and unswitched memory B cells; and total CD14 monocytes) were used to stratify juvenile-onset SLE patients into 4 groups which distinguished between patients with longitudinally active and inactive disease. Between-group differences were primarily driven by differences in the frequency of CD8 and CD4 T cells.

Researchers noted that the majority of patients with juvenile-onset SLE had well-controlled disease, which limited the investigation of a severely active phenotype.

“[T]he application of machine-learning approaches to immune phenotyping data has identified immunological biomarkers that could potentially help to unravel underlying disease mechanisms in juvenile-onset SLE and explain the differences in long-term outcomes of patients with this disease,” the researchers concluded. “Such immunological signatures could facilitate better stratification of patients for optimal treatment choices and provide information to improve interventional clinical trial design.”

Reference

Robinson GA, Peng J, Dönnes P, et al. Disease-associated and patient-specific immune cell signatures in juvenile-onset systemic lupus erythematosus: patient stratification using a machine-learning approach. Lancet Rheumatol. 2020;2(8):e485-96.