Phenotypic Analysis of Baseline T cell Subsets and Third Generation CD20 Targeted CAR T cell Products for Treatment of Relapsed/Refractory Follicular Lymphoma

Background:

Chimeric antigen receptor (CAR) T cell therapy achieves high response rates in the treatment of follicular lymphoma (FL), with two FDA-approved therapies to date. However, some patients (pts) fail to achieve a complete response and others experience early relapse, and the reasons for this are unclear. One potential explanation for treatment failure is poor persistence or exhaustion of CAR T cells. Indeed, in other settings of CAR T cell (CAR-T) therapy, phenotypic differences in the CAR-T product and even pre-manufacturing T cells can predict duration of anti-tumor CAR-T function and antigen-responsive cytotoxicity.

Methods:

We previously reported promising efficacy and safety in a cohort of pts with relapsed/refractory FL treated with a fully-human 3 ^rd generation CD20-targeted CAR (Shadman et al., ICML 2023). In pt samples from this trial, we quantified CD4 ⁺ and CD8 ⁺ T cell subsets in screening blood samples at study entry (n=20), based on CD45RA, CD45RO, CCR7, CD62L, CD27, and/or CD95 expression (naïve [T _N], central memory [T _CM], effector memory, stem cell memory, effector, etc.) We also analyzed various combinations of markers from exhaustion (PD-1, Tim-3, Lag-3, CD39, and CD57) and memory (CD45RA, CD45RO, CCR7, CD62L, and CD127) panels in CAR-T products (measured 1 day prior to infusion) (n=19; 1 pt had insufficient cells in the infusion product [IP] to analyze). These were recorded as both percent expression (of total CD4 ⁺ or CD8 ⁺ cells) compared to isotype controls and calculated absolute cell counts per µl at apheresis or in IP. We performed a univariate analysis of hypothesis-driven biomarkers (7 at baseline, 10 in IP) (Table 1) capturing combinations of T _N / T _CM markers (screening and IP) and absence of exhaustion markers (IP only) with endpoints of complete remission (CR) at 6 months, progression-free survival (PFS), CAR-T expansion (area under the curve [AUC] and peak), and cytokine release syndrome (CRS) (no pts developed neurologic toxicity). Due to the small sample size, we performed multivariable analysis using a novel high dimensional inference (HDI) approach to rank potential biomarkers of interest with primary endpoint CR at 6 months and secondary endpoint CAR-T cell AUC. HDI provides “de-biased” estimates for coefficients, confidence intervals, and p-values without refitting, which is easier to interpret. Biomarker combinations comprising <1% cells were excluded.

Results:

We did not find any significant association between baseline or IP hypothesis-driven markers and CR at 6 months or CRS in univariate analysis. We noted a possible association with baseline absolute counts of CD45RO ⁺CD62L ⁺CD8 ⁺ cells for PFS; however, this was largely driven by an outlier pt with a short remission due to CD20-negative relapse. Baseline % of CD45RA ⁺CD62L ⁺ in CD8 ⁺ cells was associated with higher peak CAR-T expansion, and infused absolute counts of CD8 ⁺ CAR T cells with CD45RA ⁺CD62L ⁺, CD45RO ⁺CD62L ⁺, CD39 ^-PD-1 ^-, or Lag3 ^-PD-1 ^- phenotypes were found to be associated with both higher CAR-T AUC and peak CAR-T expansion. The multivariable HDI analysis revealed a rank list of markers of interest associated with CR at 6 months and CAR-T AUC, with the top 5 combinations for each association shown in Table 2.

Conclusion:

We found that higher infused numbers of CD8 ⁺ CAR T cells with T _N or T _CM phenotypes, or low expression of certain exhaustion phenotype marker combinations, were associated with improved CAR-T expansion. We did not find any notable phenotypic signatures at baseline or in the IP that were predictive of prolonged CR, PFS, or toxicity, though the analysis was limited by a relatively small sample size. In FL, where a response to CAR-T therapy is almost universal, pre-manufacture and IP T cell phenotypes appear to have a lower importance for achieving an initial response than what has been reported in other settings with lower response rates (chronic lymphocytic leukemia, large B cell lymphoma). Our lack of clear association between T cell phenotypes and prolonged CR or PFS suggests that later treatment failures in FL may be driven primarily by factors other than T cell fitness, such as tumor biology. However, our exploratory multivariable analysis did identify some potential phenotypic signatures that may prove to be predictive with further validation, and future studies incorporating duration of B cell aplasia may show that phenotypes are important for functional CAR T cell persistence.