.png){kind=icon}
.png){kind=icon}
Return
Cerebrospinal fluid-derived small extracellular vesicles reveal a compartment-specific diagnostic signature for NSCLC leptomeningeal metastasis
.png){kind=icon}
OA
AI
P
Y
T
M
J
K
J
X
Q
H
Y
W
B
DOI：10.1186/s12951-026-04847-8.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Leptomeningeal metastasis (LM) is a devastating complication of non-small cell lung cancer (NSCLC), yet timely diagnosis remains challenging because conventional approaches, including magnetic resonance imaging (MRI) and cerebrospinal fluid (CSF) cytology, have limited sensitivity, particularly at initial presentation. Small extracellular vesicles (sEVs), as nanoscale carriers of protected regulatory cargo, have emerged as promising substrates for liquid biopsy. However, in anatomically compartmentalized central nervous system malignancies, the extent to which disease-associated sEV-miRNA signals are preserved across biofluids remains unclear. We profiled sEV-associated miRNAs by small RNA sequencing in CSF and plasma samples from NSCLC patients with clinically adjudicated LM status (NC, LM−, and LM+), including paired specimens where available. Biofluid-specific miRNA programs were characterized using weighted gene co-expression network analysis (WGCNA), and diagnostic models were developed using machine learning with Random Forest and LASSO-based feature selection. Model performance was evaluated in an independent validation cohort relative to final adjudicated LM status and compared with initial and cumulative MRI and CSF cytology assessments. CSF-derived sEV-associated miRNAs showed markedly stronger disease-stratification capacity than plasma-derived signals (ARI = 0.813 vs. 0.255; cluster purity = 0.868 vs. 0.694), revealing profound biofluid asymmetry. Network analysis identified a dominant LM-associated module in CSF (r = 0.74, P = 9 × 10⁻⁷) with limited preservation in plasma, indicating that LM-related vesicle-associated molecular programs are compartment-restricted. From this CSF-specific nanoscale vesicle signature, we derived a four-miRNA diagnostic panel comprising hsa-let-7e-5p, hsa-miR-30d-5p, hsa-miR-486-5p, and hsa-miR-375. In the independent validation cohort, the single-sample sEV-miRNA panel achieved an AUC of 0.935, with 86.4% sensitivity and 87.5% specificity. By comparison, initial MRI and initial CSF cytology showed sensitivities of 50.0% and 77.3%, respectively, whereas cumulative MRI and cumulative CSF cytology showed sensitivities of 70.0% and 86.4%, respectively. CSF-derived sEV-associated miRNAs encode an LM-proximal molecular program that is not faithfully captured in plasma, underscoring the importance of biofluid compartmentalization in nanobiotechnology-based biomarker discovery. Our findings establish a compartment-aware sEV-miRNA diagnostic strategy for accurate LM detection and provide a conceptual framework for extracellular vesicle-based molecular diagnostics in central nervous system malignancies.
Keywords:
Small extracellular vesicles
MiRNA
NSCLC
Leptomeningeal metastasis
Biomarkers
Journal
IF:
12.6
Papers:
5.0K
Citations:
2.8W
Organization
Cited Papers