PD-L1 expression in cerebrospinal fluid for leptomeningeal metastasis from solid tumors: preliminary assessment of clinical implications

TL;DR

This study shows that PD-L1 in cerebrospinal fluid can be reliably tested and may help guide immunotherapy for brain cancer spread.

Contribution

A reliable method for detecting PD-L1 in cerebrospinal fluid using ThinPrep LBC is established for leptomeningeal metastasis.

Findings

PD-L1 detection via ThinPrep LBC showed high concordance with immunohistochemistry for certain antibodies.

PD-L1 positivity rates in CSF samples ranged from 42% to 59% depending on cell count and antibody used.

PD-L1 expression in CSF showed poor agreement with extracranial lesions, but higher response rates were observed in PD-L1-positive patients.

Abstract

Intrathecal immune checkpoint inhibitors have shown promise for leptomeningeal metastasis (LM). Programmed cell death ligand 1 (PD-L1) is currently the primary predictive biomarker for immunotherapy response. This study assessed the feasibility of PD-L1 detection via immunocytochemistry based on ThinPrep liquid-based cytology (LBC), explored its application in cerebrospinal fluid (CSF) from LM patients and preliminarily evaluated clinical implications.

Technical validation used six human tumor cell lines (lung, breast and gastric) with validated high/low PD-L1 expression, which were processed into ThinPrep LBC slides and cell block sections. PD-L1 immunocytochemistry and immunohistochemistry were performed using four antibodies (Dako 22C3, Ventana SP263, Abcam 28–8 and SP142). Clinically, CSF samples from LM patients were ThinPrep-processed and PD-L1 immunocytochemistry-stained. PD-L1 expression was assessed via tumor proportion score. Concordance of different antibody assays, as well as that of PD-L1 expression between paired CSF and extracranial lesions, was analyzed using Cohen’s κ. The association between CSF PD-L1 expression and response to intrathecal immunotherapy was assessed.

Immunocytochemistry based on ThinPrep LBC reliably detected membrane PD-L1, showing high concordance with immunohistochemistry on cell blocks for clones 22C3/SP263 across all cell lines (NCI-H358, MDA-MB-231, SGC-7901, A549, MCF7, and AGS). Clones 28-8/SP142 showed false positives and were excluded. Using clones 22C3/SP263, 130 CSF samples from 65 LM patients (55 lung, 8 breast, and 2 gastric cancers) were analyzed. Overall PD-L1 positivity rates were 48% with 22C3 and 51% with SP263 (κ=0.815). In the subset of 68 slides containing ≥100 tumor cells, PD-L1 positivity rates were 53% with 22C3 and 59% with SP263 (κ=0.881). For 62 slides containing 20–100 tumor cells, the rates were 42% (13/31) with both antibodies (κ=0.735). PD-L1 expression showed poor agreement between 29 paired CSF and extracranial lesions (κ=0.175 with 22C3; κ=0.179 with SP263). Among 45 patients receiving intrathecal immunotherapy, A numerical increase in response rate was observed in PD-L1-positive patients (61.9% vs. 33.3%; p=0.055).

Our study establishes a robust methodology for detecting CSF PD-L1 expression using ThinPrep LBC with immunocytochemistry for LM patients. This approach suggests potential utility of CSF PD-L1 expression as a biomarker for guiding intrathecal immunotherapy for LM from solid tumors.

Figures