Distinct immune landscapes characterize highly versus minimally invasive brain metastases
Sarah M. Maritan, Elham Karimi, Matthew Dankner, Aldo Hernandez-Corchado, Miranda W. Yu, Matthew G. Annis, Yashar Aghazadeh Habashi, Morteza Rezanejad, Bridget Liu, Nebras Koudieh, Emilie Pichette, Parvaneh Fallah, Benoit Fiset, Yuhong Wei, Ali Nehme, Chun Geun Lee, Jack A. Elias, Morag Park, Yasser Riazalhosseini, Hamed Najafabadi, Kevin Petrecca, Marie-Christine Guiot, Daniela F. Quail, Logan A. Walsh, Peter M. Siegel
Sarah M. Maritan, Elham Karimi, Matthew Dankner, Aldo Hernandez-Corchado, Miranda W. Yu, Matthew G. Annis, Yashar Aghazadeh Habashi, Morteza Rezanejad, Bridget Liu, Nebras Koudieh, Emilie Pichette, Parvaneh Fallah, Benoit Fiset, Yuhong Wei, Ali Nehme, Chun Geun Lee, Jack A. Elias, Morag Park, Yasser Riazalhosseini, Hamed Najafabadi, Kevin Petrecca, Marie-Christine Guiot, Daniela F. Quail, Logan A. Walsh, Peter M. Siegel
View: Text | PDF
Research Article Immunology Oncology

Distinct immune landscapes characterize highly versus minimally invasive brain metastases

Abstract

Brain metastases (BrMs) occur in approximately 30% of cancer patients, causing nearly one-fifth of cancer deaths. While immune checkpoint inhibitors (ICIs) benefit some BrM patients, responses remain highly variable. This variability partly reflects distinct histopathological growth patterns that include minimally invasive (MI) and highly invasive (HI) brain BrMs. Here we show that MI BrMs exhibit robust immune infiltration, whereas HI lesions are immunosuppressed. However, histological differentiation between MI and HI can be challenging because of subjective margin assessment. Here, using highly multiplexed spatial proteomics on 119 tumor sections from 46 patients with BrMs, we identify CHI3L1 as a key mediator of the immunosuppressive microenvironment in HI BrMs. In preclinical models, genetic deletion of CHI3L1 converts immune-cold metastases into lymphocyte-rich, ICI-responsive lesions infiltrated by granzyme B+ CD8+ T cells. In BrM patients treated with ICI, immunohistochemical quantification of CHI3L1 expression was a stronger predictor of ICI response than traditional MI/HI classification. Thus, CHI3L1 represents a promising biomarker and therapeutic target for BrMs.

Authors

Sarah M. Maritan, Elham Karimi, Matthew Dankner, Aldo Hernandez-Corchado, Miranda W. Yu, Matthew G. Annis, Yashar Aghazadeh Habashi, Morteza Rezanejad, Bridget Liu, Nebras Koudieh, Emilie Pichette, Parvaneh Fallah, Benoit Fiset, Yuhong Wei, Ali Nehme, Chun Geun Lee, Jack A. Elias, Morag Park, Yasser Riazalhosseini, Hamed Najafabadi, Kevin Petrecca, Marie-Christine Guiot, Daniela F. Quail, Logan A. Walsh, Peter M. Siegel

×

Figure 2

NanoString Digital Spatial Profiling reveals augmented interferon-γ signaling in MI BrMs.

Options: View larger image (or click on image) Download as PowerPoint
NanoString Digital Spatial Profiling reveals augmented interferon-γ sign...
(A) Schematic depicting cancer cells selected for NanoString Digital Spatial Profiling in HI (left panels) and MI (right panels) patient BrMs. Invaded cancer cells (HI samples) or cancer cells at the brain-tumor interface (MI samples) underwent profiling. Scale bars: 750 μm. (B) Volcano plot illustrating differentially expressed genes (red, n = 21) in MI versus HI lung cancer BrM samples. Genes to the left of center are upregulated in HI samples, while genes to the right of center are upregulated in MI samples. (C) Heatmap depicting normalized gene expression (z score of log10-normalized gene counts) of differentially expressed genes (Padj < 0.1; n = 21) in HI (blue) and MI (brown) lung cancer BrMs. (D) Gene set enrichment analysis enrichment plot for the interferon-γ (IFN-γ) response pathway in lung cancer BrMs. (E) Quantification of p-STAT1 staining (H score) of BrM tissue microarrays. Samples were taken at the tumor core (left; n = 125 MI, n = 258 HI) and tumor-brain margins (right; n = 9 MI, n = 28 HI). P values were calculated using 2-tailed Student’s t test. (F) Representative images of p-STAT1 staining (brown) in MI (top) and HI (bottom) margin samples, corresponding to data in E. Scale bars: 1 mm (left), 100 μm (middle and right). (G) Quantification of p-STAT1 staining (H score) in tumor cells in patient-matched core and margin regions of BrMs from any primary site (n = 16 HI, 14 MI). H scores were calculated by multiplication of staining intensity scores (0 to 3) by the percentage of positively stained tumor cells (1%–100%) for a maximum H score of 300. For each patient tissue, 10 regions of interest (ROIs) (200,000 μm2 each) were captured at the brain-tumor interface, and 10 ROIs were captured within the metastasis core (more than 1.5 mm distance from any brain-tumor interface). Margin/core H score ratio was calculated per sample. (H) Representative p-STAT1 staining (brown) in 2 lung cancer BrM samples (MI, top; HI, bottom), corresponding to data in G. Images of the brain-tumor interface (margin) and within the metastatic lesion (core) are shown for each sample, with high-magnification panels provided (right). Scale bars: 100 μm (all images).