sCSF1R levels in plasma and CSF1R expression in LCH lesions
We determined plasma sCSF1R levels by ELISA to confirm the proteomic findings. The results showed that sCSF1R levels were greater in 104 LCH patients than in 10 healthy controls (median level: 38,823.8 vs.18,172.6 pg/ml, P < 0.001; Figure 3A) and significantly differed among the three disease extents, with the most increased level occurring in the MS RO+ LCH group (P < 0.001; Figure 3B).
We further investigated the association between sCSF1R levels and the clinicopathological characteristics of LCH patients (Supplementary Figure S2). Patients younger than two years at diagnosis with skin, lung, or risk of organ involvement who were harboring theBRAF- V600E mutation in lesion tissues or cell-free DNA had increased sCSF1R. In addition, the plasma levels of several other DEPs were evaluated via ELISA in LCH patients (Supplementary Figure S3).
We analyzed the differential expression of CSF1R in LCH lesion cells by IF staining (Figure 3C). Consistently, patients with MS RO+ LCH had significantly greater CSF1R expression than did those with MS RO- LCH (P < 0.01) or SS-LCH (P < 0.001).
Prognostic significance of sCSF1R levels at diagnosis
ROC curve analysis revealed that plasma sCSF1R levels at diagnosis could efficiently predict the prognosis of pediatric LCH patients (n=104) treated with standard first-line treatment (AUC = 0.782, P< 0.001; Figure 4A), and an optimal cutoff for sCSF1R was chosen at 42,401.2 pg/ml by maximizing the Youden index. Using the cutoff values, we divided the whole cohort into high- and low-sCSF1R groups (72 and 32 patients, respectively). The 3-year PFS was significantly poorer in the high-sCSF1R subgroup than in the low-sCSF1R subgroup (21.9% ± 7.3% vs. 78.8% ± 4.9%, P < 0.001; Figure 4B). The prognostic impact of high-sCSF1R expression was significant in MS LCH patients (P < 0.001; Figure 4D) but not in SS-LCH patients (P = 0.85; Figure 4C).
Univariate analysis revealed that age at diagnosis; sex; bone, skin, liver, spleen, hematologic system, lung, pituitary, eye, ear, or lymph node involvement; and sCSF1R level were associated with the 3-year PFS of LCH patients (Figure 4E). Moreover, multivariate Cox proportional hazards regression analysis revealed that the sCSF1R concentration was an independent prognostic factor for 3-year PFS in children with LCH after we adjusted for these univariate risk factors (Figure 4F).