The immune contexture of primary central nervous system diffuse large B cell lymphoma associates with patient survival and specific cell signaling

Abstract

Rationale: Primary central nervous system diffuse large B-cell lymphoma (PCNSL) is a rare and aggressive entity that resides in an immune-privileged site. The tumor microenvironment (TME) and the disruption of the immune surveillance influence lymphoma pathogenesis and immunotherapy resistance. Despite growing knowledge on heterogeneous therapeutic responses, no comprehensive description of the PCNSL TME is available. We hence investigated the immune subtypes of PCNSL and their association with molecular signaling and survival. Methods: Analysis of PCNSL transcriptomes (sequencing, n = 20; microarrays, n = 34). Integrated correlation analysis and signaling pathway topology enabled us to infer intercellular interactions. Immunohistopathology and digital imaging were used to validate bioinformatic results. Results: Transcriptomics revealed three immune subtypes: immune-rich, poor, and intermediate. The immune-rich subtype was associated to better survival and characterized by hyper-activation of STAT3 signaling and inflammatory signaling, e.g., IFNγ and TNF-α, resembling the hot subtype described in primary testicular lymphoma and solid cancer. WNT/β-catenin, HIPPO, and NOTCH signaling were hyper-activated in the immune-poor subtype. HLA down-modulation was clearly associated with a low or intermediate immune infiltration and the absence of T-cell activation. Moreover, HLA class I down-regulation was also correlated with worse survival with implications on immune-intermediate PCNSL that frequently feature reduced HLA expression. A ligand-receptor intercellular network revealed high expression of two immune checkpoints, i.e., CTLA-4/CD86 and TIM-3/LAGLS9. TIM-3 and galectin-9 proteins were clearly upregulated in PCNSL. Conclusion: Altogether, our study reveals that patient stratification according to immune subtypes, HLA status, and immune checkpoint molecule quantification should be considered prior to immune checkpoint inhibitor therapy. Moreover, TIM-3 protein should be considered an axis for future therapeutic development.

Keywords: PCNSL; bioinformatics; cellular interactions; immune contexture; immunotherapy.

Figure 1

PCNSL with an immune-rich TME defines a patient subgroup with a better outcome. A. Three PCNSL clusters exist based on immune gene expression, denoted high, intermediate, and low. B. TME cell gene signatures across PCSNL (n = 20) and DLBCL (n = 48) tumor sample transcriptomes reveal three groups of tumors: the immune-rich, immune-poor and immune-intermediate subtypes (clusters 2, 1, 3 respectively). PCNSL were classified in agreement with the gene clusters of panel A. C. Applying identical TME cell gene signatures to PCNSL microarray data (n = 34) reveals four clusters. Merging the two small rightmost clusters yielded three groups of tumors (denoted 1-3) with immunological subtypes comparable to those of panel B. D. Quantification of activated CD8+ T cells, regulatory T cells, macrophages, and CAFs in the immune-rich tumors versus the other subtype tumors (Wilcoxon two-sided tests, n = 20 = 8+12). E. The immune-rich subtype features a more favorable outcome (Kaplan-Meier curve, log-rank test, n = 20 = 8+12). Activ CD4: activated CD4+ T cells, Activ CD8: activated CD8+ T cells, Tregs: regulatory T cells, Tfh: T follicular helper cells, Th1: type 1 helper cells, Th2: type 2 helper cells, Th17: type 17 helper cells, Activ DC: activated dendritic cells, CD56 bright: CD56 bright NK cells, CD56 dim: CD56 dim NK cells, MDSC: Myeloid-derived dendritic cells, Mono: monocytes, TAM: tumor-associated macrophage, EC: endothelial cells, CAF: cancer-associated fibroblast.

Figure 2

HLA expression is related to an immune-rich TME in PCNSL. A. HLA class I and II gene expression in PCNSL. High/low HLA gene expression defined the HLA status: HLA normal, HLA I, II, and I & II down (see heatmap legend). Expression was assessed independently in each cohort; cohort 1 is mRNA-sequencing and cohort 2 is microarray data. B. Samples from the distinct immune subtypes were counted in each HLA gene expression cluster of panel A. C. Differentially expressed genes according to distinct HLA status (HLA normal in pink versus HLA I, II, and I & II down in green): 103 significantly deregulated genes were selected that perfectly segregated the samples according to their HLA status (FDR < 0.01, log₂-FC > 4 (absolute value), average read counts > 20). D. Main Gene Ontology Biological Process (GOBP) terms found significantly enriched in genes in panel C (hypergeometric test, FDR < 0.05, at least 3 deregulated genes in each GO term). E. HLA-down status (I, II, I & II) associates with earlier relapse in PCNSL (Kaplan-Meier curves, log-rank test, n = 14, normal = above median, down = below median). F. Differentially expressed genes between PCNSL with normal HLA class I gene expression (orange) and HLA I down (dark blue): 62 significantly deregulated genes were selected which perfectly segregated the two groups (FDR < 0.01, log₂-FC > 4 (absolute value), average read counts > 20). G. Main GOBP terms found significantly enriched in genes in (f) (hypergeometric test, FDR < 0.05, at least 3 deregulated genes in each GO term). H. HLA I-down status associates with earlier relapse in PCNSL (Kaplan-Meier curves, log-rank test, n = 17, normal = above median, down = below median).

Figure 3

The immune landscapes of PCNSL unravel distinct oncogenic signaling. A. Signature scores (average z-scores of signature genes) of different oncogenic signaling pathways according to the immune subtypes defined in Figure 1bc. B. Gene Set Enrichment Analysis performed between the immune-rich and other subtypes (intermediate & poor). We report the normalized enrichment score (NES) and indicate statistical significance (FDR < 0.05) by a pink dot (otherwise blue). C. Quantification of NOTCH, WNT, and HIPPO signature scores in the different immune subtypes (Wilcoxon one-sided tests, p-value < 0.05, n = 4+8+8 = 20, one immune-poor subtype outlier was removed (significant according to Grubbs and Dixon tests). D. Correlation between the main cell-type scores (EC, CAF, Act CD8, Tregs, Act DC, TAM, MDSC) and oncogenic signaling z-scores. Highly significant correlations are indicated only (spearman rank correlation coefficient, r > 0.6, p-value < 0.01). E. Correlation between STAT3 and IFNγ signature scores and CD274 gene expression in the PCNSL cohort (spearman rank correlation coefficient, r > 0.6, p-value < 0.0001). Note that correlation is observed with both mRNA-sequencing (RNA-seq) and microarray (MA) data. F. High INFγ and STAT3 signaling z-scores were associated with lower relapse-free survival (RFS) in PCNSL (Kaplan-Meier curves, log-rank test, n = 20, high = above median, low = below median).

Figure 4

Ligand-receptor interactions within the PCNSL microenvironment. A. Ligand-receptor (L-R) pair selection strategy: our algorithm selected 673 and 393 confident L-R pairs from the mRNA-sequencing (n = 20) and microarray (n = 34) datasets respectively. A total of 128 confident and PCNSL-specific L-R pairs were selected from both datasets. Out of these 128 L-R pairs, survival analysis found 26 L-R pairs significantly associated to RFS and 8 to overall survival (OS). B. Functional categories associated to the L-R pairs selected. C. Immune infiltrate-associated L-R pairs in our PCNSL cohort (n = 20). D. LGALS9/HAVCR2 L-R scores are higher in the immune-rich subtype of PCNSL (Wilcoxon one-sided tests, p-value < 0.01, n = 20 = 8+12 and n = 2+22 for mRNA-sequencing and microarrays data respectively). E. LGALS9/HAVCR2 above median L-R scores are associated with better RFS (Kaplan-Meier curve, log-rank test, p-value < 0.05).

Figure 5

Galectin-9/TIM-3 crosstalk is up-regulated and linked to immune activities in PCNSL. A. TIM-3 protein expression is up-regulated in PCNSL compared to normal brain tissue. B. Galectin-9 (Gal-9) protein expression is up-regulated in PCNSL tissue compared to normal brain tissue. C. TIM-3 and galectin-9 expression are up-regulated in PCNSL tumors in comparison to normal brain tissue (student test). D. TIM-3 and galectin-9 expression are correlated in our PCNSL cohort (spearman rank correlation, n = 20). E. Correlation between either HAVCR2 (TIM-3) or LGALS9 (galectin-9) gene expression and other immune checkpoints (spearman rank correlation, adjusted p-value < 0.01). F. Correlation between either HAVCR2 (TIM-3) or LGALS9 (galectin-9) expression and the cell-type z-scores (spearman rank correlation, adjusted p-value < 0.05). G. Correlation between either HAVCR2 (TIM-3) or LGALS9 (galectin-9) gene expression and PCNSL related signaling pathway z-scores (spearman rank correlation, adjusted p-value < 0.05).