T Cell Determinants of Response and Resistance to PD-1 Blockade in Richter's Transformation

Despite advances in the treatment of chronic lymphocytic leukemia (CLL), the transformation of CLL to an aggressive lymphoma, or Richter's transformation (RT), remains a clinical challenge, as it responds poorly to standard therapies and shortens survival. Recent studies demonstrate that RT, but not underlying CLL, responds to PD-1 checkpoint blockade (CPB) with an overall response rate of 43-65%. Given the central role of T cells in anti-tumor immunity, we hypothesized that differences in T cell populations underlie response and resistance to CPB in RT.

We focused on a discovery cohort of 6 patients with RT (4 responders, 2 non-responders) and 2 patients with relapsed/refractory CLL enrolled on a study in which patients were initiated with anti-PD1 therapy (nivolumab 3 mg/kg every 2 weeks), with subsequent concurrent ibrutinib (420 mg daily)(NCT 02420912). We examined a total of 15 serial study marrow specimens collected at treatment initiation and 3 month response evaluation, as well as 2 healthy marrow donors.

To systematically discover the T cell populations and states associated with CPB response in RT, we performed single-cell RNA-sequencing (scRNA-seq, 10x Genomics) of non-lymphoma (CD5-CD19-) cells isolated by flow cytometry from marrow samples. A total of 60,727 T and NK cells were captured with average detection of 1001 genes/cell. Using the novel joint clustering approach Conos, 11 transcriptionally distinct clusters of lymphocytes were identified. We first contrasted baseline RT/CLL with normal marrow and observed differences across T cell populations, which we confirmed through the examination of publicly available marrow scRNA-seq data from 28 healthy donors. Compared to normal marrow, RT/CLL marrow was enriched for cytotoxic populations, including both CD8 effector/effector memory (E/EM) (p=0.001, t-test) and cytotoxic CD4 (p=0.001) T cells as well as for cells expressing multiple exhaustion markers, including PDCD1, LAG3 and TIGIT (p=0.001). In contrast, normal marrow contained increased T cells with a naïve-like phenotype (p=0.06).

When we focused on the pre-treatment samples from RT patients, RT responders had a larger CD8 E/EM population (p=0.04) and fewer T regulatory cells (p=0.006, t-test) than RT non-responders. Using DESeq2 to compare clusters from all samples, we evaluated if there were differences in gene expression between RT responders and non-responders. CD8 E/EM T cells of RT non-responders showed increased expression of TOX, a recently uncovered master regulator of cell exhaustion (p_adj =0.00016), while this cell subtype in RT responders upregulated a contrasting program of activating transcription factors as well as the co-stimulatory gene CD226 (p_adj =0.04). As for CD4 T cells, RT responders revealed an enriched cytotoxic gene program compared to RT non-responders (p_adjPRF1 5.9 x 10^-10, GZMH 6.0 x 10^-6, NKG7 6.4 x 10^-19).

To investigate whether response to CPB therapy for RT was associated with changes in the T cell receptor (TCR) repertoire, and to obtain protein-level validation of transcriptional signatures, we performed single-cell TCR sequencing with paired gene and protein expression (10x Genomics) on pre- and post-therapy samples from a RT responder and a non-responder. Indeed, we confirmed our gene expression findings, including validation of cytotoxic CD4 T cells and the enrichment of CD226 protein in E/EM CD8 T cells in the RT responder. TCR clonal expansion was observed in the RT responder at baseline with persistence of enriched clonotypes following CPB, suggesting the presence of tumor-reactive T cell clones. In contrast, the RT non-responder displayed higher TCR diversity with enriched clonotypes showing increased exhaustion post-CPB (p<0.001, Poisson rate test). Ongoing validation studies include characterizing the phenotypes of corresponding peripheral blood T cells and confirming our findings in an independent, larger cohort of RT patients using multiplex immunofluorescence and flow cytometry.

In conclusion, we identified marrow T cell populations enriched in RT patients and described distinct T cell transcriptional programs that delineate RT responders from non-responders. We have thus discovered candidate gene biomarkers that may identify patients likely to respond to CPB therapies and uncovered a CD8 E/EM T cell population that is likely to underlie response to PD-1 CPB.