The endogenous T cell landscape is reshaped by CAR-T cell therapy and predicts treatment response in multiple myeloma

Abstract

While most patients initially respond to CAR-T cell treatment, responses often are not durable and subsequent lines of immunotherapy show diminishing success. In this study, we investigated the co-evolutionary dynamics between CAR-T cells and the immune microenvironment in myeloma patients undergoing anti-BCMA CAR-T cell therapy at single-cell resolution. Our findings highlight the transformative impact of CAR-T cell treatment on the endogenous T cell landscape. We identify a novel transitional CD8 + T cell population that is predictive of poor treatment outcomes. The emergence of this population coincides with the depletion of the endogenous T cell repertoire and compositional evolution of functional T cell subsets. These changes in the endogenous T cell compartment induced by CAR-T cell therapy may contribute to inadequate immune capacity and tumor control. Our findings highlight the potential of targeting TIM3/GAL9 interactions to mitigate T cell exhaustion, apoptosis and lack of persistence, offering promising avenues for optimizing T cell-based cancer immunotherapies. We provide a framework for assessing and manipulating the 'mileage' of the immune system as predictive marker and therapeutic opportunity to prevent repeated immunotherapies from becoming increasingly less successful, even when targeting distinct antigens.

Fig. 1. CAR-T treatment changes the endogenous T cell landscape.

A Schematic overview of workflow. B UMAP of CD8 + T cells and CAR-T cells colored by clusters. C UMAP plots highlighting cells from each timepoint or CAR-T cells. D Expression of marker genes in different clusters. E Heatmap displaying expression of published gene signatures. F Comparison of proportion of cell states in endogenous T cells and CAR-T cells at M2D1 for patients. p = 0.11 for cluster 0, p = 0.36 for cluster 2, *p = 0.014 for cluster 6, p = 0.85 for cluster 1, p = 0.98 for cluster 3, p = 0.35 for cluster 4 and ***p = 2e-04 for cluster 5, respectively, by Wilcoxon test. G Volcano plot comparing endogenous CD8 + T cells at M2D1 vs. CAR-T cells. H Scoring of indicated signatures in endogenous CD8 + T cells at M2D1 vs. CAR-T cells. p < 2.2e − 16, p < 2.2e − 16, p = 4.7e − 06, p = 0.00031 by Wilcoxon test, respectively.

Fig. 2. Competing differentiation states in the CD8 + T cell compartment of CAR-T-treated patients.

A, B Proportion of clusters over time in endogenous CD8 + T cells. p = 0.15 and p = 0.29 for cluster 0, p = 0.0036 and p = 0.5 for cluster 2, p = 0.0065 and p = 0.63 for cluster 6, p = 0.069 and p = 0.15 for cluster 1, p = 0.00052 and p = 1 for cluster 3, p = 0.015 and p = 0.0015 for cluster 4 and p = 0.0025 and p = 0.033 for cluster 5 by Wilcoxon test, respectively. C Heatmap showing gene expression for representative T cell–related genes, categorized according to their function. D GSEA in endogenous CD8 + T cells at screening vs. M2D1. E GSEA in endogenous CD8 + T cells at screening vs. M7D1. F CD8 activation, exhaustion and naïve-like gene signature scores in transitional T cell states T1 (top) and T2 (bottom) compared to activated, exhausted and naïve CD8 + T cells. ****p = 6e − 14, ****p < 2.2e-16 and ****p < 2.2e-16 for T1 and p = 0.055, ****p = 7.4e-13 and ****p = 2.3e-13 for T2, respectively, by Wilcoxon test. G CD8 activation score. ****p = 1.2e − 08 by Wilcoxon test. H GSEA in transitional T cell state T1 vs. activated effector cells. For B,F,G box plots show the median and the interquartile range, and whiskers extend to 1.5x the interquartile range.

Fig. 3. Divergent transcriptional states in T cells determine response.

A, B Clinical response. Patient cohort was divided by median progression-free survival (PFS) into long-term responders (LTR) and short-term responders (STR) (20.14 months, red line). C Comparison of proportion of cells per state in normal donor CD8 + T cells, T cells from patients at screening timepoint and CAR-T cells from LTR and STR. D Volcano plot showing differential gene expression in CAR-T cells from LTR vs. STR. Selected canonical exhaustion markers are highlighted. E Pearson correlation between activation and exhaustion signatures in CAR-T cells. F CAR-T cell persistence in LTR vs. STR at M2D1, determined by the percentage of CAR-T cells of total CD3 + T cells in peripheral blood. Box plot shows the median and the interquartile range, and whiskers extend to 1.5x the interquartile range. p = 0.035 by Wilcoxon test. G Correlation of CAR-T persistence, as shown in F with percentage of exhausted CAR-T cells. H Heatmap showing gene expression in endogenous CD8 + T cells at screening, M2D1 and M7D1 in long-term responders (LTR) and short-term responders (STR) for representative T cell-related genes, categorized according to their function. I Gene expression scores comparing LTR and STR in endogenous CD8 + T cells at M2D1 for cytolytic markers, exhaustion markers and memory-associated markers. p = 0.019, p = 0.00064 and p = 0.00071 by Wilcoxon test, respectively. J Terminal differentiation in M7D1 endogenous CD8 + T cells. p = 0.0067 by Wilcoxon test. K Percentage of CD8 + T cells in transitional T cell T1 cluster (cluster 3) per timepoint by response. p = 0.89, p = 0.11 and p = 0.024 by Wilcoxon test for comparisons between LTR and STR, respectively. L Transitional T1 signature scored in endogenous CD8 + T cells in LTR and STR across timepoints. p = 0.0036, p = 0.073 and p = 9.1e − 09 by Wilcoxon test, respectively. M Correlation plot showing correlation of transitional T1 signature with apoptosis signature in endogenous CD8 + T cells.

Fig. 4. State-specific clonal expansion in the endogenous CD8 + T cell compartment of CAR-T-treated patients.

A Schematic overview of the droplet-based single-cell RNA sequencing workflow for bone marrow aspirates. B Comparison of the proportion of expanded clonotypes - CDR3s (top) and cells (bottom) - at screening and M7D1 in CD8 + T cells. C Proportion of cell states in CD8 + T cells with singletons and expanded clonotypes. D Comparison of TCR clone size of clonotypes shared between screening and M7D1. ***p = 2e-05 by paired Wilcoxon test. E Proportions of cell states in pre-existing clonotypes, defined as clonotypes detected in screening and M7D1. F Proportion of cells per cell state in the individual TCR clonotype 1. G Stacked bar plots showing disappearing and emergent expanded clonotypes. H RNA velocity projected on UMAP. Cells of clonotype 1 are highlighted in color of their cell state. I Model of CD8 + T cell differentiation.

Fig. 5. Reciprocal cell-cell interactions promote immunosuppressive microenvironment.

A Exhaustion score. n = 3491 cells. ****p < 2.2e-16 by Wilcoxon test for all comparisons. B HAVCR2 expression. ****p < 2.2e-16 by Wilcoxon test for all comparisons. C Correlation between percentage of exhausted CAR-T cells and exhausted endogenous CD8 + T cells at M2D1. D Correlation between expression of an apoptosis gene signature in CAR-T cells and endogenous CD8 + T cells at M2D1. E Interaction count between CAR-T cells and endogenous T cells at M2D1. F Heatmap showing significant inferred cell-cell interactions between CAR-T cells and endogenous CD8 + T cells at M2D1. Only significant interactions (-log(pvalue) >3) are displayed. G TIM3 surface expression compared between CD8 + T cells from different timepoints and CAR-T cells generated by CITE-seq staining from the droplet-based scRNA-Seq dataset of bone marrow aspirates of 4 patients. n = 5158 cells. ****p = 5.9e-06 for M2D1, ****p = 3.5e-05 for M7D1 and *p = 0.026 for CAR-T cells by Wilcoxon test. H Boxplots showing persister scores depending on Galectin-9 plasma levels measured by ELISA. *p = 0.017 by Wilcoxon test. I Proportion of live T cells following pre-incubation with Gal9 in untransduced CD8 + T cells and CAR-T cells. p < 0.0001 by unpaired t-test. J Model illustrating the effect of GAL9 binding to TIM3 on CAR-T cell persistence and exhaustion. For A,G,H box plots show the median and the interquartile range, and whiskers extend to 1.5x the interquartile range.