Immunotherapy response: is it all in the DLBCL-IQs? Free

In this issue of Blood, Tumuluru et al¹ provide new insights into 2 critical areas: the heterogeneity of the molecular and immunological landscape of diffuse large B-cell lymphoma (DLBCL) and the influence of these tumor environments on the efficacy of emerging immunotherapies, such as bispecific antibodies (BsAbs) and chimeric antigen receptor T-cell (CAR-T) therapies.

DLBCL is the most common B-cell lymphoma—aggressive in nature, yet biologically complex and heterogeneous. For over 2 decades, researchers have grappled with the best approach to refine the molecular taxonomy of the oversimplified histological umbrella—DLBCL—moving beyond the binary gene-expression–based cell-of-origin (COO) classification (germinal center B [GCB] vs activated B cell [ABC]) toward molecular subgroups defined by distinct genetic features.^2-5 More recently, classification systems based on immune microenvironmental features have emerged, often derived from bulk RNA sequencing or deconvoluted single-cell RNA-sequencing data.^6,7 

Despite these advancements, the field still seeks a unified classification system for DLBCL. From a translational perspective, a key question remains: Can these classification frameworks realistically serve as biomarkers to guide therapy selection? This issue is especially pressing as the therapeutic arsenal for DLBCL continues to expand.

In this context, Tumuluru et al approached this problem by performing bulk transcriptomic analysis on over 500 DLBCL patients, calculating per-sample enrichment scores for 19 gene sets comprising immune-related and COO-related signatures. This analysis enabled them to classify DLBCL into distinct immune environments, termed DLBCL-immune quadrants (DLBCL-IQs). The authors identified 4 IQs based on COO and immune gene expression signatures: ABC hot; ABC cold; GCB hot; and GCB cold. These 4 IQs were validated in an additional cohort of nearly 300 patients. They also noted specific features of the IQs. For instance, "hot" DLBCLs exhibited higher CD8⁺ T-cell infiltration and activity compared with "cold" DLBCLs. These findings were confirmed through multispectral immunofluorescence imaging and immune-cell deconvolution analysis using CIBERSORTx, validating the robustness of the DLBCL-IQ framework.

With numerous existing DLBCL classification systems at play, where does the IQ framework fit? The authors compared the DLBCL-IQs with 2 other published microenvironment-based classifiers, revealing partial overlap.^6,7 For example, ABC hot DLBCL-IQs significantly overlapped with the inflammatory microenvironment (LME-IN) group from the Kotlov et al study, whereas GCB hot DLBCL-IQs showed minimal similarity with any of the 4 Kotlov lymphoma microenvironments. Similarly, when compared with the 9 lymphoma ecotypes described by Steen et al, more than half of the ecotypes showed little resemblance to the DLBCL-IQs.

Notably, the gene sets defining the IQs primarily focused on T-cell–related signatures, omitting contributions from other microenvironmental players, such as natural killer cells and stromal components. Genetics-based subtypes, such as those defined by the LymphGen classifier,³ also showed only partial overlap with the IQs. These findings suggest that the various classification approaches—including the DLBCL-IQ framework—may be capturing different aspects of DLBCL biology. Consequently, the field remains some way from nearing a comprehensive, all-encompassing classification system for this heterogeneous disease.

Interestingly, the study identified some notable genetic associations with DLBCL-IQs. For example, SOCS1 loss-of-function mutations were strongly enriched in GCB hot DLBCLs, whereas both ABC and GCB cold DLBCLs demonstrated higher MYC activity compared with their “hot” counterparts. This underscores the potential for intrinsic tumor features to shape the surrounding T-cell immune microenvironment.

While the immune composition has been shown to predict outcome previously,⁸ the hot or cold IQ cohorts did not significantly influence overall patient outcomes; however, they did impact outcomes to immunotherapies increasingly utilized in patients with relapsed/refractory DLBCL. Several biomarkers have been reported as associated with efficacy of CAR-T therapy in DLBCL, ranging from the features of the CAR-T product to the nature of the baseline circulating immune states,^9,10 but there remains a huge knowledge gap. In their study, Tumuluru et al demonstrate that patients with GCB hot DLBCLs exhibited superior progression-free survival with BsAb therapy (mosunetuzumab) compared to those with GCB cold DLBCLs. However, unlike BsAbs, outcomes with CAR-T therapies were less influenced by the immune microenvironment, suggesting that CAR T cells may still function effectively even in unfavorable "cold" tumor environments. It is important to note the relatively small patient cohorts for these therapy-specific analyses, which limits the generalizability of the findings. Validation in larger cohorts and exploration in additional therapeutic contexts will be essential to establish clinical applicability.

Despite these limitations, Tumuluru and colleagues strengthen the case that pretreatment immune environments may influence the effectiveness of a range of immunotherapies. Although it remains unclear how the immune environment in DLBCL is dynamically reshaped over time in response to various therapies, studies like this serve as a starting point to bridge the complex biological heterogeneity in DLBCL into meaningful, clinically applicable biomarkers.

Conflict-of-interest disclosure: J.O. declares no competing financial interests.