Characterizing memory T helper cells in patients with psoriasis, subclinical, or early psoriatic arthritis using a machine learning algorithm

Background Psoriasis patients developing psoriatic arthritis (PsA) are thought to go through different phases. Understanding the underlying events in these phases is crucial to diagnose PsA early. Here, we have characterized the circulating memory T helper (Th) cells in psoriasis patients with or without arthralgia, psoriasis patients who developed PsA during follow-up (subclinical PsA), early PsA patients and healthy controls to elucidate their role in PsA development. Methods We used peripheral blood mononuclear cells of sex and age-matched psoriasis patients included in Rotterdam Joint Skin study ( n =22), early PsA patients included in Dutch South West Early Psoriatic Arthritis Cohort (DEPAR) ( n =23) and healthy controls (HC; n =17). We profiled memory Th cell subsets with flow cytometry and used the machine learning algorithm FlowSOM to interpret the data. Results Three of the 22 psoriasis patients developed PsA during 2-year follow-up. FlowSOM identified 12 clusters of memory Th cells, including Th1, Th2, Th17/22, and Th17.1 cells. All psoriasis and PsA patients had higher numbers of Th17/22 than healthy controls. Psoriasis patients without arthralgia had lower numbers of CCR6 - CCR4 + CXCR3 + memory Th cells and higher numbers of CCR6 + CCR4 - CXCR3 - memory Th cells compared to HC. PsA patients had higher numbers of Th2 cells and CCR6 + CCR4 + CXCR3 - cells, but lower numbers of CCR6 + CCR4 + CXCR3 + memory Th cells compared to HC. The number of CCR6 + Th17.1 cells negatively correlated with tender joint counts and the number of CCR6 + Th17 cells positively correlated with skin disease severity. Conclusions Unsupervised clustering analysis revealed differences in circulating memory Th cells between psoriasis and PsA patients compared to HC; however, no specific subset was identified characterizing subclinical PsA patients.