Single-cell characterization of human GBM reveals regional differences in tumor-infiltrating leukocyte activation

  1. Philip Schmassmann  Is a corresponding author
  2. Julien Roux
  3. Steffen Dettling
  4. Sabrina Hogan
  5. Tala Shekarian
  6. Tomás A Martins
  7. Marie-Françoise Ritz
  8. Sylvia Herter
  9. Marina Bacac
  10. Gregor Hutter  Is a corresponding author
  1. Brain Tumor Immunotherapy Lab, Department of Biomedicine, University of Basel, Switzerland
  2. Bioinformatics Core Facility, Department of Biomedicine, University of Basel, Switzerland
  3. Swiss Institute of Bioinformatics, Switzerland
  4. Roche Pharmaceutical Research and Early Development, Roche Innovation Center Munich, Germany
  5. Roche Pharmaceutical Research and Early Development, Roche Innovation Center Zürich, Switzerland
  6. Department of Neurosurgery, University Hospital Basel, Switzerland
7 figures, 1 table and 8 additional files

Figures

Figure 1 with 2 supplements
Single-cell RNA-seq of cells from tumor center, periphery and blood.

(A) Experimental workflow for single-cell analysis of cells isolated from tumor center, periphery and peripheral blood mononuclear cells (PBMC), including fluorescent-activated cell sorting and 3’-scRNA-seq. (B) Axial T1 with contrast (left) and T2 (right) MRI brain in a patient with a left temporal GBM. Fresh tumor biopsies were taken according to neuronavigation (green cross). The tumor center was defined as contrast enhancing, whereas the tumor periphery was defined as T2 hyperintense. (C) Nuclear DAPI staining of resected tissue specimens. ×40 magnification (scale bar = 20 μm). n=3 patients, 4 field of view (FOV) per patient. Statistics: ***p<0.001, two-tailed Mann Whitney U test (Figure 1—source data 1).

Figure 1—source data 1

Related to Figure 1C.

DAPI + cell count per field of view (FOV) between tumor Center and Periphery.

https://cdn.elifesciences.org/articles/92678/elife-92678-fig1-data1-v1.xlsx
Figure 1—figure supplement 1
CD45+CD11b+immune cells gating strategy and quality control of scRNA-seq data.

(A, B) Gating strategy for paired tumor-derived (A) and PBMCs (B); after debris, doublet and dead cell removal, immune cells were assessed as CD45+and/or CD11b+. (C–E) Percentage of mitochondrial (MT) reads (C) number of detected genes (D) and cell-cycle phase (E) overlaid on tSNE representation. Please see Supplementary file 2 for quantification of C–E.

Figure 1—figure supplement 2
Patient representation among clusters.

(A, B) tSNE map stratified according to site (A) and patient (B). (C–E) tSNE maps showing all cells, colored by site (C), patient (D) and sample (E). (F) Cell-mixing score (Lütge et al., 2021) overlayed on tSNE representation.

Figure 2 with 2 supplements
Single-cell RNA-seq analysis identifies main immune cell populations.

(A) Dimensionally reduced tSNE projection of the scRNAseq data showing the annotated cell types. (B) Heatmap displaying centered and scaled normalized average expression values of characteristic cell-type specific genes used to annotate clusters. Columns are ordered by site and cell type, and rows show centered and scaled expression values, hierarchically clustered. Heatmap displaying genes whose expression is most specific to each cell type is shown in Figure 2—figure supplement 2. (C) Principal component analysis of pseudo-bulk scRNAseq samples aggregated by patient and cell type. Symbols represent individual patients and cell lineage is displayed by different colors. (D) Relative frequencies of immune populations among leukocytes between tumor center and periphery, shown as boxplots. Symbols represent individual patients (n=5) and paired samples are indicated by connecting lines. p-Values were calculated using diffcyt-DA-voom method (Figure 2—source data 1). (E–G) Differential abundance testing of the tumor innate immune compartment (E) using the miloR package which tests the abundance of each neighborhood of cells separately between tumor center und periphery (F, G).

Figure 2—source data 1

Related to Figure 2D.

Differential abundance analysis between tumor center and periphery.

https://cdn.elifesciences.org/articles/92678/elife-92678-fig2-data1-v1.xlsx
Figure 2—figure supplement 1
Cross-referencing scRNA-seq data with published datasets.

(A–D) Using hierarchical clustering, identified cell clusters at different levels of granularity (A, B) which were then annotated into nine distinct cell types for the immune subset and four cell types for the CD45neg subset (C, D). (E, F) Immune cell types were annotated by referencing to a dataset of bulk RNA-seq samples of sorted immune cell types from human PBMC (E) (Monaco et al., 2019) and MdMs and microglia were annotated by comparing to a dataset of bulk RNA-seq samples of sorted immune cell types from the tumor microenvironment of human gliomas (F) (Klemm et al., 2020). Clusters are highlighted which were annotated using each respective reference dataset. (G–K) Subset analysis on tumor innate immune cells (G), showing re-clustering on the subset (H), annotation by referencing to a 10 X genomics scRNA-seq dataset of TAMs from the GBM TME of seven newly diagnosed human patients (I) (Pombo Antunes et al., 2021) and signature scores defined from scRNA-seq of glioma TAMs (J, K) (Müller et al., 2017). (L–O) Subset analysis on CD45neg cells (L), showing re-clustering on the subset (M), annotation by whole-transcriptome comparison to a scRNA-seq dataset of IDH1wt GBM (N) (Neftel et al., 2019) and an endothelial score which was defined by averaging the center and scaled expression levels of the genes CDH5, VWF, CD34, and PECAM1 (O).

Figure 2—figure supplement 2
Cell type specific gene expression.

(A) Heatmap displaying genes whose expression is most specific to each cell type. Columns are ordered by site and cell type, and rows show centered and scaled expression values, hierarchically clustered.

Figure 3 with 1 supplement
MG display regionally resolved transcriptional profiles that differ from those of DCs and MdMs.

(A) Microglia cluster highlighted on tSNE map and scatterplots showing differentially expressed genes (FDR <5%, indicated by blue and yellow) in Microglia (MG) cells from tumor periphery versus center. Volcano plot showing p value versus fold-change (left) and MA plot showing fold-change versus mean expression (right). For a complete list of differentially expressed genes per cell cluster between tumor periphery and center, please refer to Supplementary file 4. (B) Heatmap representation of Gene set enrichment analysis (GSEA) results between peripheral and center microglia using Gene Ontology (GO) collection (Biological Processes). The fraction of overlap between gene sets is calculated as Jaccard coefficient of overlap between the gene sets. (C, D) Heatmap representation of GSEA of DCs (C) and monocyte-derived macrophages (MdMs) (D) from tumor periphery versus tumor center using Hallmark collection of major biological categories. (E) Unsupervised hierarchical sub-clustering of the MG population revealed two transcriptionally distinct subsets of MG, termed MG_1 and MG_2, displayed on the tSNE map. (F) Heatmap displaying the cluster-specific genes identifying MG_1 subcluster. Columns are ordered by site and cell type, and rows show centered and scaled normalized average expression values, hierarchically clustered. A complete list of cluster specific genes for MG_1 and MG_2 subcluster is provided in Supplementary file 5. (G) Heatmap representation of GSEA between MG_1 and MG_2 subclusters using Hallmark collection of major biological categories. (H) Heatmap displaying previously described reactivity markers of MG. Columns are ordered by site and cell type, and rows show centered and scaled normalized average expression values, hierarchically clustered.

Figure 3—figure supplement 1
Regionally dependent transcriptional profiles of innate immune subsets and MG subclusters MG_1 and MG_2.

(A, B) Volcano plots showing differentially expressed genes (FDR corrected p-value <0.05, indicated by colors) in DCs (A) and MdMs (B) from tumor periphery versus center. (C) Volcano plot showing differentially expressed genes (FDR corrected p-value <0.05, indicated by colors) in DCs from tumor periphery versus PBMC. (D) Heatmap representation of GSEA between DCs from tumor periphery and PBMC using Hallmark collection of major biological categories. (E) Heatmap displaying the cluster-specific genes identifying MG_1 and MG_2 subclusters. Columns are ordered by site and cell type, and rows show centered and scaled normalized average expression values, hierarchically clustered. (F, G) Violin plots showing average expression levels (log2 counts) of selected markers among mononuclear phagocyte populations. (H) Frequencies of MG_1 and MG_2 subpopulations among total microglia between center and periphery, shown as boxplots. Symbols represent individual patients (n=5) and paired samples are indicated by connecting lines. Statistical significance was assessed by diffcyt-DA-voom method (Supplementary file 5), ***FDR corrected p value p≤0.001, ****FDR corrected p-value ≤0.0001.

Figure 4 with 1 supplement
The peripheral cytotoxic cell compartment exhibits an impaired activation signature.

(A, B) Volcano plots showing differentially expressed genes (FDR corrected p-value <0.05, indicated by blue and yellow) in CD8+ T cells (A) and NK cells (B) from tumor periphery versus tumor center. Colored rings mark genes belonging to selected GSEA Hallmark or Gene Ontology (GO) pathways as indicated. For a complete list of differentially expressed genes per cell cluster between tumor periphery and center, please refer to Supplementary file 4.

Figure 4—figure supplement 1
Differential expression analysis between tumor center and peripheral CD4+T cells.

(A) Volcano plot showing differentially expressed genes (FDR corrected p-value <0.05, indicated by blue and yellow) in CD4+T cells from tumor periphery versus tumor center.

Figure 5 with 1 supplement
CX3CR1 labels a specific CD8 +T cell population in the circulation of grade 4 glioma patients.

(A) CD8+ T cell cluster highlighted on tSNE map (left). CD8+ T cell cluster colored by site of origin (right). (B) Volcano plot showing differentially expressed genes (FDR corrected p-value <0.05, indicated by blue and green) in CD8+ T cells from tumor-periphery versus PBMC. For a complete list of differentially expressed genes per cell cluster between tumor periphery and PBMC, please refer to Supplementary file 6. (C) Frequency of CX3CR1+ CD8+ T cells among all CD8+ T cells in flow cytometry data (Figure 5—source data 1). (D) Unsupervised hierarchical sub-clustering of CD8+ T cells from PBMC and Periphery revealed two transcriptionally distinct subsets of PBMC CD8+ T cells, displayed on the tSNE map. (E) Expression of CX3CR1 overlaid on tSNE CD8+ T cell cluster. (F) Expression of genes associated with effector memory phenotype overlaid on tSNE CD8+ T cell cluster. Displayed genes are significantly, differentially expressed genes (DEGs) between tumor periphery and PBMC, as identified by differential gene expression analysis shown in panel (B). (G) Expression of selected genes associated with naive phenotype overlaid on tSNE CD8+ T cell cluster (H) Gating procedure applied to identify CD3+ CD8+ naive, T effector cells (Teff), effector memory (Tem), peripheral memory (Tpm) and central memory (Tcm), eluted from PBMCs. (I) Expression of CX3CR1 in PBMC CD8+ T cell subpopulations identified in (H) (Figure 5—source data 2). n=6 donors (C), n=11 donors (I). Statistics: Wilcoxon matched-pairs signed rank test (C); repeated measures one-way ANOVA with post-hoc Šidák’s correction for multiple comparisons (I). *p≤0.05, **p≤0.01, ***p≤0.001, ****p≤0.0001, no brackets indicate no significant difference.

Figure 5—source data 1

Related to Figure 5C.

CX3CR1+ CD8+ T cells between PBMC and tumor periphery.

https://cdn.elifesciences.org/articles/92678/elife-92678-fig5-data1-v1.xlsx
Figure 5—source data 2

Related to Figure 5I.

CX3CR1+ CD8+ T cells subpopulations in patient PBMC.

https://cdn.elifesciences.org/articles/92678/elife-92678-fig5-data2-v1.xlsx
Figure 5—figure supplement 1
Phenotypic characterization of PBMC CD8+ T cells.

(A) Gating strategy for paired tumor-periphery and PBMC cells; after debris, doublet and dead cell removal, CD8 +T cells were identified as CD3+CD8+events. (B) Single-cell expression of additional markers associated with naive/memory.

Figure 6 with 1 supplement
CD8+ T cells in the tumor periphery share features with tissue-resident memory T cells (Trm).

(A) Representative dot plot of tumor-periphery CD8+ T cells stained for CD45RA and CD45RO. (B) Quantification of tumor-periphery CD8+ T cells expressing CD45RA or CD45RO (Figure 6—source data 1). (C) Expression of genes associated with tissue-resident memory (Trm) phenotype overlaid on tSNE CD8+ T cell cluster. (D) Average expression levels of selected Trm markers between CD8+ T cells from PBMC versus tumor-periphery. Significance testing based on differential gene expression analysis shown in panel (Figure 5B) (E) Representative dot plots of CD69 and CD103 co-expression in CD8+ T cells from PBMC and tumor-periphery. (F) Quantification of CD69 and CD103 co-expression revealed CD69- CD103- in PBMC and CD69+ CD103- and CD69+ CD103+ in tumor-periphery as the dominant phenotypes (Figure 6—source data 2). (G) Expression of selected markers associated with T cell exhaustion/dysfunction, shown as boxplots between CD8+ T cell from PBMC and tumor-periphery and overlaid on tSNE CD8+ T cell cluster. Significance testing based on differential gene expression analysis shown in panel (Figure 5B). n=6 donors (B, F). Statistics: Wilcoxon matched-pairs signed rank test (B); repeated measures one-way ANOVA with post-hoc Šidák’s correction for multiple comparisons (F). *p≤0.05, **p≤0.01, ***p≤0.001, ****p≤0.0001, no brackets indicate no significant difference.

Figure 6—source data 1

Related to Figure 6B.

CD45RA or CD45RO expressing CD8+ T cells in tumor periphery.

https://cdn.elifesciences.org/articles/92678/elife-92678-fig6-data1-v1.xlsx
Figure 6—source data 2

Related to Figure 6F.

CD69 CD103 co-expression analysis on CD8+T cells from PBMC and tumor periphery.

https://cdn.elifesciences.org/articles/92678/elife-92678-fig6-data2-v1.xlsx
Figure 6—figure supplement 1
Phenotypic characterization of Periphery CD8+ T cells.

(A–C) Single-cell expression of additional markers associated with tissue-resident memory (A), T cell co-stimulation (B) and T cell exhaustion/dysfunction (C) overlaid on tSNE CD8+T cell cluster. (D) Volcano plot showing differentially expressed genes (FDR corrected p-value <0.05, indicated by blue and green) in CD4+ (left) and CD8+ T cells from tumor-periphery versus PBMC.

Figure 7 with 1 supplement
Cell-cell communication analysis using CellChat reveals critical role for SPP1-mediated crosstalk in tumor periphery.

(A) Chord diagram showing significant interactions from microglia to lymphocyte cell clusters. The inner bar colors represent the targets that receive signal from the corresponding outer bar. The inner bar size is proportional to the signal strength received by the targets. Chords indicate ligand-receptor pairs mediating interaction between two cell clusters, size of chords is proportional to signal strength of the given ligand-receptor pair. (B) Comparison of incoming and outgoing interaction strength allows identification of main senders and receivers. (C) Violin plots showing the expression distribution of signaling genes involved in the inferred SPP1 signaling network.

Figure 7—figure supplement 1
Cell-cell communication analysis using CellChat.

(A, B) Chord diagram showing significant interactions from microglia to all cell clusters in center (A) and periphery (B). The inner bar colors represent the targets that receive signal from the corresponding outer bar. The inner bar size is proportional to the signal strength received by the targets. Chords indicate ligand-receptor pairs mediating interaction between two cell clusters, size of chords is proportional to signal strength of the given ligand-receptor pair. (C, D) Violin plots showing the expression distribution of signaling genes (in log2 counts) involved in the inferred SPP1 signaling network in center (C) and periphery (D).

Tables

Key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
Biological sample (Human adult GBM tissue samples)Tumor Center; Tumor PeripheryNeurosurgical Clinic of the University Hospital of Basel, Switzerland
Biological sample (Human Peripheral Blood Buffy Coat)PBMCNeurosurgical Clinic of the University Hospital of Basel, Switzerland
AntibodyRat monoclonal anti-human/mouse CD11b (clone M1/70), FITCBioLegendCat# 101206FACS: 1:50
AntibodyMouse monoclonal anti-human CD45 (clone 2D1), FITCBioLegendCat# 368508FACS: 1:50
AntibodyFc-Block anti-human CD16/CD32, TruStain FcXBioLegendCat# 422302FACS: 1:50
AntibodyMouse monoclonal anti-human CD45RO (clone UCHL1), APCBioLegendCat# 304210FACS: 1:25
AntibodyMouse monoclonal anti-human CD45RA (clone HI100), PEBioLegendCat# 304108FACS: 1:25
AntibodyMouse monoclonal anti-human CD3e (clone UCHT1), BV650BioLegendCat# 300468FACS: 1:25
AntibodyMouse monoclonal anti-human CD8a (clone RPA-T8), BV421BioLegendCat# 301036FACS: 1:25
AntibodyMouse monoclonal anti-human CCR7 (clone G043H7), FITCBioLegendCat# 353216FACS: 1:25
AntibodyMouse monoclonal anti-human CD62L (clone DREG-56), AF700BioLegendCat# 304820FACS: 1:50
AntibodyMouse monoclonal anti-human CD69 (clone FN50), APC/Cy7BioLegendCat# 310914FACS: 1:25
AntibodyMouse monoclonal anti-human CD103 (clone Ber-ACT8), PerCP/Cy5.5BioLegendCat# 350226FACS: 1:25
AntibodyRat monoclonal anti-human CX3CR1 (clone 2A9-1), PE-Cy7BioLegendCat# 341612FACS: 1:25
Commercial assay or kitChromium Single Cell 3’ Reagent Kits v310 x GenomicsCat# CG000183
Commercial assay or kitBioAnalyzer High Sensitivity DNA Analysis kitAgilentCat# 5067–4626
Commercial assay or kitQubit dsDNA High Sensitiivity assay kitThermoFisherCat# Q33230
Chemical compound, drugCollagenase-4Worthington Biochemical CooperationCat# LS004188
Chemical compound, drugDNAse1RocheCat# 10104159001
Chemical compound, drugACK lysis bufferGibcoCat# A1049201
Chemical compound, drugBambankerNippon GeneticsCat# BB01
Chemical compound, drugSucorseSigma AldrichCat# 84100
Chemical compound, drugFicoll-Paque PLUSCytivaCat# 17144002
Chemical compound, drugLive/Dead Fixable Near IR Dead Stain Kit, APC-Cy7ThermoFisherCat# L34976
Chemical compound, drugZombie Aqua Fixable Viability KitBioLegendCat# 423102
Software, algorithmR environment, version 4.1R Core Teamhttps://www.r-project.org/
Software, algorithmGraphPad Prism 9GraphPad Software Inc.N/A
Software, algorithmFlow Jo, version 10.8.1Tree StarN/A

Additional files

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Philip Schmassmann
  2. Julien Roux
  3. Steffen Dettling
  4. Sabrina Hogan
  5. Tala Shekarian
  6. Tomás A Martins
  7. Marie-Françoise Ritz
  8. Sylvia Herter
  9. Marina Bacac
  10. Gregor Hutter
(2023)
Single-cell characterization of human GBM reveals regional differences in tumor-infiltrating leukocyte activation
eLife 12:RP92678.
https://doi.org/10.7554/eLife.92678.2