Elucidating the mechanism of resistance to CAR-T in Large B-cell Lymphomas treated at first relapse/progression

Background 

CD19-directed chimeric antigen receptor (CAR) T cells induce durable remissions in relapsed/refractory large B cell lymphomas (R/R LBCL), but a sizeable fraction of patients still relapse. Identifying factors leading to treatment failure has been a major focus of recent research. We and others have shown that disease response is determined by complex interactions between host immune system, cell product, and lymphoma characteristics, all of which are indistinctly interrelated and equally instrumental for the success or failure of CAR T therapy. To date, no specific biomarker is able to predict the outcome at the single patient level. 

Additionally, data have been generated in the context of heavily pre-treated patients, whereas now CAR T are available for patients at first relapse/remission. 

Hypothesis 

The study of the peripheral blood immune contexture coupled with the analysis of the tumor genetic alterations in cell-free DNA prior to CAR T manufacturing, could lead to the identification of clinically relevant biomarkers of response. The latter could be used to improve patient selection when combined with clinical information known to impact CAR T outcomes, but such an analysis should rely on explainable artificial intelligence (XAI) tools used to define algorithms able to predict CAR T efficacy at the single patient level. 

Aims 

• Develop systematic biobanking of LBCL patients receiving CAR T cells at first relapse/progression with paired well-annotated clinical data 
• Identify pre-treatment tumor- and immune-related biomarkers of response 
• Define the impact of lymphoma genetic abnormalities on the phenotype and dynamics of immune cells 
• Analyze CAR T phenotypes, expansion and persistence 
• Develop machine learning tools to create an algorithm to improve patient selection for a better outcome. 

Experimental Design 

We have planned an observational study comprising a prospective collection of biological samples and clinical data, from 340 consecutive R/R LBCL receiving CAR T after failing one prior treatment line. Genotyping of circulating tumor DNA will be done to identify genetic lesions predictive of response or resistance. To characterize the immune subset composition in leukapheresis and peripheral blood pre-CAR T and longitudinally after infusion, we will apply transcriptomic and spectral flow cytometry. XAI-tools will be used to integrate biological and clinical data. 

Expected Results 

• Real life results on efficacy and toxicity of CAR T in second line  
• characterization of pre-treatment lymphoma genetic abnormalities relevant for the outcome 
• characterization of immune cell lineages in leukapheresis and in peripheral blood relevant for CAR T 
• expansion and disease response 
• definition of the interplay between lymphoma genetic alterations and the peripheral immune system 
• identification of AI-based signatures relevant for response and survival 
• definition of XAI-based prognostic model. 

Impact On Cancer 

We plan to define risk categories necessary to guide more effective CAR T administration in lymphoma patients with the final goal of increasing cure rate and maintaining future healthcare sustainability. Our effort would also allow the identification of characteristics that facilitate immune escape leading to unresponsiveness or relapse post-CAR T, thus enabling the planning of trials employing novel treatment strategies such as bispecifc antibodies or allogeneic CAR T.