Predicting Prognosis in DLBCL: The Role of Immune-Related Gene Signatures

In an era of increasingly personalized oncology, a recent bioinformatics study published in BMC Immunology has spotlighted four immune-related genes—CXCL9, CCL18, C1QA, and CTSC—as potential prognostic biomarkers in diffuse large B-cell lymphoma (DLBCL).

Using an integrative machine learning pipeline, the authors have constructed a predictive model that not only identifies these key genes but also underscores their interplay with immune cell infiltration and survival outcomes.

From Expression Profiles to Prognostic Signatures
Drawing on three publicly available GEO datasets (GSE25638, GSE12195, and GSE12453), the research team implemented a combination of differential expression analysis and weighted gene co-expression network analysis (WGCNA), narrowing 5,000 highly variable genes down to 95 key genes enriched in inflammatory and chemokine signaling pathways.

Notably, KEGG and GO enrichment analyses revealed signatures associated with CXCR chemokine receptor binding, mineral absorption, and immune cell regulation, setting the stage for further refinement.

Random forest (RF) algorithms ranked feature importance across all datasets, leading to the convergence on CXCL9, CCL18, C1QA, and CTSC as top predictors of DLBCL risk. These genes were incorporated into a diagnostic nomogram—a scoring model that achieved an AUC of 1.000 in training cohorts and 0.839 in external validation (GSE83632), demonstrating robust predictive accuracy.

Survival Analysis and Immune Cell Correlation
Kaplan-Meier curves drew a sharp distinction in survival probability: high expression of the four genes was consistently associated with poorer overall survival. Moreover, the prognostic separation between germinal center B-cell-like (GCB) and non-GCB subtypes became more pronounced with the expression levels of these genes, with non-GCB patients faring worse, mirroring known clinical trends.

Immune deconvolution via single-sample gene set enrichment analysis (ssGSEA) painted a nuanced picture of the tumor microenvironment (TME). CXCL9 expression in particular correlated strongly with CD56^bright NK cells and activated CD4 T cells, while negatively associated with macrophages and immature B cells. The immune contexture further revealed elevated regulatory T cells (Tregs) and myeloid-derived suppressor cells (MDSCs), hinting at immune suppression layered atop immune activation.

CXCL9 Leads the Charge
Among the four hub genes, CXCL9 emerges as a frontrunner, validated by both qRT-PCR and immunohistochemistry (IHC). Elevated CXCL9 levels were not only statistically significant in patient samples, but they also aligned with previous reports on its role in immune recruitment and inflammation.

Interestingly, the study connects CXCL9’s expression to known pathways involved in CAR-T cell trafficking and response durability, suggesting possible intersections between this marker and immunotherapy responsiveness.

The other three genes contribute distinct immunological signatures. CCL18, often produced by tumor-associated macrophages, plays a dual role in modulating immune suppression and cancer progression. C1QA, a component of the classical complement pathway, has been linked to rituximab response and may originate from immune cells within the tumor microenvironment. And CTSC, a lysosomal protease, is involved in neutrophil extracellular trap (NET) formation, further tying it to pro-tumor immune mechanisms.

Clinical Implications and Limitations
While the study offers compelling biomarker candidates and an internally validated risk model, its reliance on retrospective GEO datasets and modest clinical validation sample size (13 DLBCL cases) introduces limitations. External validation in larger cohorts and mechanistic follow-up studies will be necessary to confirm clinical utility.

Still, the integration of immune profiling and machine learning offers a glimpse into the future of hematologic oncology—where diagnostic models are not only data-driven, but also biologically informed. As immune-based therapies expand, markers like CXCL9 and its co-players could serve as both indicators of prognosis and functional targets for intervention.

Reference
Wang S, Tao H, Zhao X, et al. Exploration of biomarkers for predicting the prognosis of patients with diffuse large B-cell lymphoma by machine-learning analysis. BMC Immunol. 2025;26(1):61. Published 2025 Aug 20. doi:10.1186/s12865-025-00738-z