Molecular and Clinical Determinants of CAR-T Therapy Response in DLBCL

Introduction:

Diffuse large B-cell lymphoma (DLBCL) is one of the most common forms of blood cancer worldwide. Outcomes for patients with relapsed or refractory (r/r) DLBCL have remained dismal. Chimeric antigen receptor (CAR) T-cell therapy targeting the B-cell surface marker CD19 has recently emerged as a novel, effective approach capable of producing durable responses in r/r DLBCL patients. Indeed, nearly all r/r DLBCL patients in the United States are considered candidates for FDA-approved CAR-T therapy. However, CAR-T therapy has a number of major shortcomings. First, it has limited efficacy. The median progression-free survival (PFS) of r/r DLBCL patients treated with CAR-T therapy is roughly six months. Second, it has major toxicities. CAR-T therapy is associated with risk of life-threatening side effects such as cytokine release syndrome (CRS) and CAR-T cell-related encephalopathy syndrome (CRES). Third, it is highly expensive. Strategies for identifying, a priori, those patients likely to respond to this therapy could provide an important tool for clinical decision-making and improving outcomes in patients with r/r DLBCL.

Methods & Results:

We enrolled a real-world cohort of 161 r/r DLBCL patients treated with CD19-directed CAR-T therapy and applied whole exome and transcriptome sequencing on their pre-treatment biopsies to identify predictors of outcome.

A number of clinical variables were associated with worse PFS or overall survival (OS) including a higher International Prognostic Index (IPI) score at diagnosis, involvement of multiple extranodal sites, poor performance status at the time of apheresis, high-grade CRES, and the need for ICU admission. Conversely, lower IPI score, activated B cell type (ABC) subtype of DLBCL, need for tocilizumab administration, and achievement of complete remission trended with more favorable survival outcomes.

As expected, the most frequently mutated genes were involved in histone modification (e.g. KMT2D, CREBBP, EP300), B-cell signaling (e.g. MYD88, PIM1, CD79B) and DNA damage-response/repair (e.g. TP53, DDX3X, BCL6, ATM). Gene mutations associated with the least favorable prognosis were SOCS1, BCL7A, and MYC, while HIST1H1E, CD79B, and MPEG1 mutations were associated with a favorable prognosis.

We further developed a supervised approach for analysis and validation of gene expression profiles and their association with outcome. We first divided the patients into independent training and test sets. In the training set, we applied the Cox proportional hazards model to identify individual genes that were associated with PFS in CAR-T therapy treated patients. We identified two gene expression signatures that were associated with superior outcomes and poor outcomes, respectively. These two signatures were combined to generate a single prognostic score that comprised the genomic model that was validated in the independent test set. The first “low-risk” signature appeared to be derived predominantly from tumor cells with a number of B-cell related genes including CD19, PAX5, CD79A and CXCR5 as well as genes associated with the NF-kB pathway. The second signature included macrophage-related genes such as CD163, FCGR1A (CD64) and MSR1, as well as genes associated with MYC targets, ferroptosis and iron transport.

We found that the model distinguished patients with significantly different outcomes. Patients in the lowest quartile of prognostic scores had a median PFS of 2.8 months, while those in the highest quartile of scores had a significantly longer median PFS of 18 months (Wilcoxon rank-sum test p< 0.001). 73% of patients identified as having favorable risk achieved complete remission.

Conclusion:

The management of CAR-T treated patients remains challenging owing to limitations in efficacy, cost and side effects. Our study indicates that genomic features of the tumor prior to treatment are significantly associated with response to CAR-T therapy in DLBCL. We anticipate that the translation of these findings into clinical practice will enable the development of precision medicine approaches using CAR-T therapy in in this challenging disease.