Clinical Validation of a Personalized NHL-TME Model System: One-on-One Comparison between Modeled Drug Responses and Patients' Clinical Responses

Background Clinical outcomes for patients with non-Hodgkin lymphoma (NHL) are highly variable depending on patient characteristics, pathologic features, and genomic abnormalities. Despite efforts to personalize treatment decisions using next-generation sequencing, success has been limited due to the substantial genomic heterogeneity observed even among tumors of the same histologic subtype. To address the unmet needs of personalized treatment of NHL, we established ex vivo tumor microenvironment (TME) models by co-culturing bone marrow fibroblasts (BMF) plus B-cell growth factors with tumor cells from individual CLL patients (Lu et al. Blood Cancer J 2021). Proliferation centers in the lymph nodes are the major histological sites for CLL pathogenesis, progression, and transformation (Soma et al. Hum Pathol, 2006). Our ex vivo TME models, with their ability to promote CLL proliferation, recapitulate this most salient cell behavior observed in vivo. In our initial validation with a cohort of 31 CLL patients, ex vivo drug responses (EVR) demonstrated strong correlation with individual patients' clinical responses to ibrutinib. Sensitive patients' cells exhibited sensitivity in the system, while resistant patients' cells behaved resistant. In our current study, we present additional validations of this model system, incorporating genomic and cytogenetic analyses, along with clinical correlations in patients with other NHL histologies, including Richter-transformed CLL and MCL. Methods & Results We found in our previous study (Lu et al) that the co-cultured CLL cells show characteristics typical of lymph node-resident CLL cells, including larger cell size and decreased cell surface CXCR4. Single cell RNA sequencing technology confirmed the downregulation of CXCR4 at the mRNA level. In addition, the co-cultured CLL cells exhibited increased MYC and E2F gene signatures, further supporting the fidelity of the ex vivo system in mimicking the lymph node TME. Additionally, our chromosomal microarray analysis revealed high levels of ex vivo tumor cell proliferation are associated with complex cytogenetics (>3 cytogenomic abnormalities), a poor prognostic indicator (Pearson's Chi-squared test, p = 0.03893). In this study, we included 23 pts (16 CLL/SLL, 4 MCL, and 3 Richter's transformation), of which 18 received BTKi and/or venetoclax (Ven). Five patients received rituximab or observation alone and were not included. EVR responses with a readout in 5-7 days were correlated with clinical treatment data collected from the medical records months later. Clinicians were blinded to EVR results when making treatment decisions and responses were assessed independently. Clinical BTKi response was “adequate” if duration of response (DOR) is >18 m and “inadequate” if DOR is < 18 m. Patients with disease progression on BTKi prior to EVR testing were also assessed. Ven response was assessed in patients receiving Ven-based treatment as the immediate next line after EVR testing. Patients with 1) complete response per imaging and absolute lymphocyte count, and 2) MRD < 10e-5 at both end of treatment and 18 m post treatment were defined as “adequate” responders. Those not meeting these criteria or with persistent symptoms beyond 1 m of Ven-based treatment were deemed “inadequate”. Sensitive EVR was defined as <10% residual viable cells in exposure to Ven and <30% residual cell proliferation in exposure to BTKi. One-on-one correlation was analyzed using the Chi-Square method. For 14 BTKi treated patients, correlation was identified in 13 patients, meaning EVR sensitive patients showed adequate clinical responses and EVR non-sensitive patients showed inadequate clinical responses or clinical resistance (p<0.0001). For Ven-treated individuals 7/8 evaluable patients showed EVR and clinical responses correlated well, but did not meet statistical significance likely due to the small number of patients in this cohort (p=0.0285). Pooled analysis of both BTKi and Ven treated patients also showed a highly significant correlation (p < 0.0001). Conclusions The ex vivo TME model has been validated from several perspectives including gene expression, cellular behaviors as well as correlations with cytogenetic prognosticators and retrospective clinical therapeutic response. If validated prospectively in a larger study, it could serve as a valuable tool for guiding treatment selection in patients with NHL.