Characterization of a Myeloid Activation Signature that Correlates with Survival in Melanoma Patients

Abstract

Understanding the cellular interactions within the tumor microenvironment (TME) of melanoma paved the way for novel therapeutic modalities, such as T cell-targeted immune checkpoint inhibitors (ICI). However, only a limited fraction of patients benefits from such therapeutic modalities, highlighting the need for novel predictive and prognostic biomarkers. As myeloid cells orchestrate the tumor-specific immune response and influence the efficacy of ICI, assessing their activation state within the TME is of clinical relevance. Here, we characterized a myeloid activation (MA) signature, comprising the three genes Cxcl11, Gbp1, and Ido1, from gene expression data of human myeloid cells stimulated with poly(I:C) or cGAMP. This MA signature positively correlated to overall survival in melanoma. In addition, increased expression of the MA signature was observed in melanoma patients responding to ICI (anti-PD-1), as compared to non-responders. Furthermore, the MA signature was validated in the murine B16F10 melanoma model where it was induced and associated with decreased tumor growth upon intratumoral administration of poly(I:C) and cGAMP. Finally, we were able to visualize co-expression of the MA signature genes in myeloid cells of human melanoma tissues using RNAscope in situ hybridization. In conclusion, the MA signature indicates the activation state of myeloid cells and represents a prognostic biomarker for the overall survival in melanoma patients.

Keywords: innate immunity; melanoma; myeloid cells; prognostic gene signature; tumor immunity.

Figure 1

Identification of a three gene signature characteristic to activated myeloid cells. (A) Dendritic cells (DC) and CD14⁺ monocytes were isolated from healthy donors and stimulated with poly(I:C) (10 µg/mL) or cGAMP (10 μg/mL) for 6 h in vitro. Gene expression counts were quantified using the NanoString nCounter Immunology Panel. Unsupervised t-Distributed Stochastic Neighbor Embedding (t-SNE) analysis showed significant clustering between unstimulated and stimulated cells for both monocytes and DC. (B) IDO1, CXCL11, and GBP1 expression values (RNA-seq, microarray) of unstimulated and stimulated myeloid cells using public Gene Expression Omnibus (GEO) datasets: GSE57494 (LPS + IFNγ), GSE82227 (IFNγ), GSE2706 (LPS), and GSE1925 (IFNγ) plotted as 3D scatter plots using the R package plotly and show clustering between unstimulated and stimulated cells.

Figure 2

Expression of the myeloid activation (MA) signature genes positively correlated with M1 macrophage and ‘mature DC’ signatures in various cancer types. (A) RNA-seq data from peripheral blood mononuclear cells (PBMC) were obtained from GEO (GSE107011) and analyzed for the expression of the signature genes for every cell type available. Transcripts per million (TPM) values were scored with a non-parametric, rank-based method using the R package singscore based on the co-expression of Cxcl11, Gbp1, and Ido1. (B) Bulk tumor RNA-seq gene expression data was obtained from The Cancer Genome Atlas (TCGA ) cohorts using GDCRNATools in R. The MA signature score was compared to previously described M1 macrophage (upper panel) and mature DC (lower panel) signatures by Pearson correlation using R. X-axis represents TCGA study abbreviations.

Figure 3

MA signature positively correlates with increased overall survival and the presence of M1 macrophages and CD8⁺ T cells in melanoma. (A) RNA-seq data from various tumor types were obtained from TCGA and the MA signature score was assessed with a non-parametric, rank-based method in R using the singscore package. The horizontal line indicates a p value of 0.05. (B) Kaplan–Meier survival curves of patients with high and low MA signature expression plotted as -log10 p values of log-rank tests of survival data for skin cutaneous melanoma (SKCM) with high vs. low MA signature gene expression (split by median) performed in R using the survminer package (n = 118). (C) Pearson correlation of the MA signature score and the presence of immune cell subsets in SKCM patient data obtained from TCGA. The estimated abundance of various immune cells was determined by Cell-type Identification By Estimating Relative Subsets Of RNA Transcripts (CIBERSORT). Each dot represents an individual patient. (D) Signature score in melanoma patients receiving anti-PD-1 (Nivolumab, Pembrolizumab) treatment. Patients were stratified into responders (complete response and partial response, n = 13) and non-responders (progressive disease, n = 14). Each dot represents an individual patient and only patients sampled pre- and post-treatment were included in the analysis. Dataset was obtained from GSE91061. Box plot defines the maximum, third quartile, first quartile, and minimum values. p-values were determined by two-sided Welch’s t-test (* p < 0.0332; ** p < 0.0021; *** p < 0.0002; **** p < 0.0001).

Figure 4

The MA signature expression was significantly induced in bone marrow-derived macrophages (BMDM) but not in melanoma cells upon stimulation with poly(I:C) or cGAMP in vitro. BMDM from three biological replicates were stimulated with poly(I:C) (10 μg/mL) or cGAMP (10 μg/mL) for 6 h (A) and 24 h (B), following gene expression analysis for Cxcl11, Gbp1, and Ido1 by qPCR (n = 6). (C) B16F10 melanoma cells were cultured with poly(I:C) (10 µg/mL) or cGAMP (10 μg/mL) for 24 h following gene expression analysis by qPCR (n = 6). Data are represented as mean ± standard error of log₂ transformed values. Data were normalized to the housekeeping gene Rplp0 [21]. (D) Cell viability of B16F10 and human melanoma cell lines SK-Mel-37 and D10 upon stimulation with poly(I:C) and cGAMP was assessed by AlamarBlue assay. Cell lines were incubated with poly(I:C) (10 µg/mL), cGAMP (10 µg/mL) for 24 h. After 6 h of incubation with AlamarBlue (10% v/v), fluorescence intensity was measured. Statistical analysis was performed using an unpaired, two-tailed Students t-tests and the p values are indicated as follows: p > 0.05 (ns), p ≤ 0.05 (*), p ≤ 0.01 (**), p ≤ 0.001 (***), p ≤ 0.0001 (****).

Figure 5

Intratumoral poly(I:C) and cGAMP significantly reduced tumor growth and induced the expression of the MA signature genes in vivo. C57BL/6J wt mice were injected with 2 × 10⁵ B16F10 melanoma cells and treated with poly(I:C) (50 μg/mouse), cGAMP in lipofectamine (10 μg/mouse), or PBS on day 7 and 11 after tumor injection. (A) Tumor size was measured using a caliper and tumor volume was calculated using the following formula: V = (length × width²)/2 (n = 12 per group). Statistical significance was calculated by a two-way ANOVA followed by Šidák’s multiple comparisons test. Data are represented as mean ± SEM and the p values are indicated as follows: p > 0.05 (ns), p ≤ 0.05 (*), p ≤ 0.01 (**), p ≤ 0.001 (***), p ≤ 0.0001 (****). (B) On day 12, tumors were isolated and analyzed for the MA signature expression by qPCR. Data were normalized to the housekeeping gene Rplp0 and are represented as mean ± SE of log₂ transformed values (n = 12 per group) [21]. Statistical analysis was performed using an unpaired, two-tailed Students t-tests, and the p values are indicated as follows: p > 0.05 (ns), p ≤ 0.05 (*), p ≤ 0.01 (**), p ≤ 0.001 (***), p ≤ 0.0001 (****). (C) Fluorescent detection of RNA transcripts in human melanoma tissue. FFPE tissue section was hybridized with Opal-labeled probes for CD68 (Opal 520), CXCL11 (Opal 570), GBP1 (Opal 620), and IDO1 (Opal 690). Nuclei were counterstained with DAPI (blue). Adjustment of brightness and color merging was performed using ImageJ. Scale bar = 30 μm.