Longitudinal single-cell (sc) immune profiling of cerebrospinal fluid (CSF) from melanoma patients (pts) with leptomeningeal disease (LMD) treated with intrathecal (IT) and intravenous (IV) nivolumab (N).

Abstract

2015 Background: LMD is associated with an extremely poor prognosis in melanoma pts, who currently have very limited treatment options. Our group previously reported initial safety and efficacy results from the dose escalation portion of a first in human phase I/IB trial using IT and IV N (NCT03025256) in melanoma LMD pts. Here we present longitudinal sc profiling of immune cells in CSF from a subset of pts in this trial to better understand the immune composition of this unique microenvironment and features associated with treatment and outcomes. Methods: 39 CSF samples were collected longitudinally from 5 long survivors (LS; IT N number of doses received 10-92) and 5 rapid progressors (RP; IT N doses 2-9) and profiled by sc RNA sequencing. Raw FASTQ reads were aligned to the GRCh38 human reference genome using Cell Ranger v6.0.0 to generate the count matrix. All samples were aggregated by Seurat v4.3.0 and data integration was performed using Harmony v0.1.0. Cells were filtered by following criteria: (1) > 20% mitochondrial content, or (2) express < 100 or > 2,500-5,000 genes, according to sample specific distribution. Empty droplets and doublets were identified by EmptyDrops function from the R package DropletUtils v1.18.0 and DoubletFinder v2.0.3, respectively. We applied normalization, FindVariableFeatures (n = 2,000), scaling, PCA, UMAP and FindClusters to summarize the integrated samples and determine cell clusters. Cell types were annotated based on (1) canonical marker genes using FindAllMarkers, (2) published gene expression signatures, and (3) reference-based tool using Symphony v.0.1.0 on published data: GSE164378 and GSE158803. Statistical analysis was conducted using the Wilcoxon rank-sum test. Results: A total of 10 patients were included in this pilot analysis. Five pts were LS (2 males; IT dose 5 mg, 1 pt; 10 mg, 2 pts; 20 mg, 2 pts; 2 BRAF mutant; 5 negative CSF cytology at baseline) and 5 pts were RP (3 males; IT dose 5 mg, 1 pt; 20 mg, 2 pts; 50 mg, 2 pts; 4 BRAF mutant; 4 positive CSF cytology at baseline). At baseline [prior to cycle (C)1 of IT N], LS pts had a higher percentage of NK cells (relative to total immune cells; P = 0.036) and mucosal-associated invariant T cells (relative to total T cells, P = 0.021) than RP pts. Prior to C2, LS had a significantly higher percentage of T cells (relative to all cells; P = 0.036). The abundance of immune cells and tumor cells were not different between the two groups at C3, at which point pts have received two doses of IT N and one dose of IV N. Conclusions: Exploratory sc analysis of longitudinal CSF samples of LMD pts suggest that baseline immune features may predict outcomes with IT N. Analysis of additional pts is ongoing to provide further insights.