Single-cell analysis of the multiple myeloma microenvironment after γ-secretase inhibition and CAR T-cell therapy

Abstract

Chimeric antigen receptor (CAR) T cells and bispecific antibodies targeting B-cell maturation antigen (BCMA) have significantly advanced the treatment of relapsed and refractory multiple myeloma. Resistance to BCMA-targeting therapies, nonetheless, remains a significant challenge. BCMA shedding by γ-secretase is a known resistance mechanism, and preclinical studies suggest that inhibition may improve anti-BCMA therapy. Leveraging a phase 1 clinical trial of the γ-secretase inhibitor (GSI), crenigacestat, with anti-BCMA CAR T cells (FCARH143), we used single-nuclei RNA sequencing and assay for transposase-accessible chromatin sequencing to characterize the effects of GSI on the tumor microenvironment. The most significant impacts of GSI involved effects on monocytes, which are known to promote tumor growth. In addition to observing a reduction in the frequency of nonclassical monocytes, we also detected significant changes in gene expression, chromatin accessibility, and inferred cell-cell interactions after exposure to GSI. Although many genes with altered expression are associated with γ-secretase-dependent signaling, such as Notch, other pathways were affected, indicating GSI has far-reaching effects. Finally, we detected monoallelic deletion of the BCMA locus in some patients with prior exposure to anti-BCMA therapy, which significantly correlated with reduced progression-free survival (PFS; median PFS, 57 vs 861 days). GSIs are being explored in combination with the full spectrum of BCMA-targeting agents, and our results reveal widespread effects of GSI on both tumor and immune cell populations, providing insight into mechanisms for enhancing BCMA-directed therapies.

Graphical abstract

Figure 1.

Overview of study design. (A) GSI increase antigen density on MM cells by inhibiting γ-secretase–mediated cleavage of BCMA, thereby improving recognition of anti-BCMA CAR T-cell therapy for MM. (B) Timeline of a prior phase 1 clinical trial that assessed the safety of GSI for patients with MM receiving anti-BCMA CAR T-cell therapy, serving as the source for our study samples. (C) The design of this study was to interrogate the bone marrow microenvironment using single-cell multiome ATAC plus gene expression of 16 clinical trial patients before and after exposure to 3 doses of oral GSI (crenigacestat). QOD, every other day.

Figure 2.

Bone marrow nonclassical monocytes are significantly reduced after GSI exposure. (A) Comparison of snRNA-Seq and snATAC-seq data using uniform manifold approximation and projection (UMAP) illustrating the automated classification of 26 distinct cell types. (B) Plasma cells were defined by their gene expression profile and the presence of aneuploidy distinguished tumor cells from normal cells. (C) Stacked bar plot comparing the cell type frequency for all samples. The percentage of nontumor cells is shown in color bars, whereas the tumor cell fraction is represented by black bars. (D) Volcano plot illustrating the results of a differential abundance analysis comparing the frequency of cell types before and after exposure to GSI. Labeled points are cell types with a P value < .05. (E) After adjustment for multiple comparisons, only nonclassical monocytes remained statistically significant, with an FDR of 0.05. DC, dendritic cell; HSC/MPP, hematopoietic stem cells/multipotent progenitor cells; pDC, plasmacytoid dendritic cell; NK, natural killer; Tcm, central memory T cell; Temra, terminally differentiated effector memory T cell; Trm, tissue-resident memory T cell.

Figure 2.

Bone marrow nonclassical monocytes are significantly reduced after GSI exposure. (A) Comparison of snRNA-Seq and snATAC-seq data using uniform manifold approximation and projection (UMAP) illustrating the automated classification of 26 distinct cell types. (B) Plasma cells were defined by their gene expression profile and the presence of aneuploidy distinguished tumor cells from normal cells. (C) Stacked bar plot comparing the cell type frequency for all samples. The percentage of nontumor cells is shown in color bars, whereas the tumor cell fraction is represented by black bars. (D) Volcano plot illustrating the results of a differential abundance analysis comparing the frequency of cell types before and after exposure to GSI. Labeled points are cell types with a P value < .05. (E) After adjustment for multiple comparisons, only nonclassical monocytes remained statistically significant, with an FDR of 0.05. DC, dendritic cell; HSC/MPP, hematopoietic stem cells/multipotent progenitor cells; pDC, plasmacytoid dendritic cell; NK, natural killer; Tcm, central memory T cell; Temra, terminally differentiated effector memory T cell; Trm, tissue-resident memory T cell.

Figure 3.

Cell-type–specific effects of GSI on gene expression and chromatin accessibility. (A) Volcano plot depicting differentially expressed genes from bone marrow samples of patients before and after GSI treatment. Genes encoding proteins that are known substrates of γ-secretase are highlighted. (B) Dot plot of the log-fold change and FDR of genes encoding γ-secretase subunits after GSI exposure in each cell type. (C-D) Gene set enrichment analysis (GSEA) of Reactome pathways using log2 fold change of differentially expressed genes (C) and gene activity scores between pre- and post-GSI exposure samples (D). (E) IL-10 and TNF-α levels in MM cell lines (H292, U266, and MOLP8) cocultured with or without allogeneic classical or nonclassical monocytes isolated from healthy donor peripheral blood mononuclear cells (PBMCs). Error bars represent standard error from 3 independent experiments using 3 MM cell lines and 3 donor PBMCs. P values from Wilcoxon rank-sum tests indicate significant differences. (F) Significantly differentially detected circulating cytokines in clinical trial participants before and after GSI exposure (n = 10). P values from Wilcoxon signed-rank tests are shown. CCL, chemokine (C-C motif) ligand; CSF, colony stimulating factor; DC2, dendritic cell type 2; DVL, disheveled; GPCR, G protein-coupled receptor; GTPase, guanosine triphosphatase; LTA, lymphotoxin-α; NES, normalized enrichment score; NGF, nerve growth factor; NCSTN, nicastrin; PCP, planar cell polarity; PTEN, phosphatase and tensin homolog; PSEN, presenilin; RAF, rapidly accelerated fibrosarcoma; RHOF, Ras homolog family member F; TGF, transforming growth factor; WNT, wingless-INT.

Figure 3.

Cell-type–specific effects of GSI on gene expression and chromatin accessibility. (A) Volcano plot depicting differentially expressed genes from bone marrow samples of patients before and after GSI treatment. Genes encoding proteins that are known substrates of γ-secretase are highlighted. (B) Dot plot of the log-fold change and FDR of genes encoding γ-secretase subunits after GSI exposure in each cell type. (C-D) Gene set enrichment analysis (GSEA) of Reactome pathways using log2 fold change of differentially expressed genes (C) and gene activity scores between pre- and post-GSI exposure samples (D). (E) IL-10 and TNF-α levels in MM cell lines (H292, U266, and MOLP8) cocultured with or without allogeneic classical or nonclassical monocytes isolated from healthy donor peripheral blood mononuclear cells (PBMCs). Error bars represent standard error from 3 independent experiments using 3 MM cell lines and 3 donor PBMCs. P values from Wilcoxon rank-sum tests indicate significant differences. (F) Significantly differentially detected circulating cytokines in clinical trial participants before and after GSI exposure (n = 10). P values from Wilcoxon signed-rank tests are shown. CCL, chemokine (C-C motif) ligand; CSF, colony stimulating factor; DC2, dendritic cell type 2; DVL, disheveled; GPCR, G protein-coupled receptor; GTPase, guanosine triphosphatase; LTA, lymphotoxin-α; NES, normalized enrichment score; NGF, nerve growth factor; NCSTN, nicastrin; PCP, planar cell polarity; PTEN, phosphatase and tensin homolog; PSEN, presenilin; RAF, rapidly accelerated fibrosarcoma; RHOF, Ras homolog family member F; TGF, transforming growth factor; WNT, wingless-INT.

Figure 4.

Perturbations in tumor cell transcriptional regulatory network after the inhibition of γ-secretase. (A) Network diagram illustrating tumor cell transcription factor binding motifs (green) overrepresented in differentially accessible promotor regions after γ-secretase inhibition. Target genes are represented by ovals, in which the red fill color corresponds to increased expression after GSI and blue indicates reduced expression. Edges represent the expected transcriptional regulation curated from a database of human transcriptional regulatory networks (TRRUST v2). (B) Pseudobulk normalized frequency of Tn5 insertion events (a measure of chromatin accessibility) for ATAC-seq peaks encompassing the NFKB1 transcription factor binding motif (MA0105.4). Only peaks within 3 kilobases of a known target gene’s promoter that are statistically correlated with its expression are shown. The violin plots to the right of the tracks illustrate the normalized mRNA expression. Complete results of statistical testing are provided in supplemental Table 6. Chr, chromosome.

Figure 5.

GSI alters predicted cell-cell interactions within the tumor microenvironment. (A-B) Bar plots showing the total number of significant interactions between pre- and post-GSI treatment samples. Only cell-cell interactions that significantly increased (Fisher exact test P < .05) (A) or decreased (B) after GSI treatment are shown. (C-D) Dot plot illustrating the average expression and interaction specificity of ligand-receptor pairs that significantly increased (C) or decreased (D) in frequency after GSI. Only interactions in which the ligand or receptor are recognized substrates of γ-secretase are shown (bold text).

Figure 6.

GSI reduced BCMA shedding from plasma cells in patients without effect on TNFRSF17 mRNA expression. (A-B) Box plots comparing the percentage of tumor and normal plasma cells expressing TNFRSF17 (A) and the normalized TNFRSF17 expression (B). Paired Wilcoxon rank-sum P values are shown for each comparison between time points. (C) Box plots comparing the percentage of plasma cells expressing BCMA using flow cytometry. (D) Box plots comparing the percentage of tumor and normal plasma cells with accessible chromatin in the TNFRSF17 gene body or promoter regions before and after exposure to GSI. (E) Normalized TNFRSF17 gene activity (counts within gene body or promoter regions) before and after exposure to GSI. (F) Box plots comparing BCMA MFI before and after exposure to GSI. (G) Density plots of normalized TNFRSF17 expression for each tumor cell subclone inferred by clustering of copy number variants. An asterisk to the left of the density plot indicates that its TNFRSF17 expression significantly differed from normal plasma cells (eg, plasma cells without aneuploidy) using the MAST statistical framework. (H) Density plots of normalized TNFRSF17 expression according to the number of copies of TNFRSF17. Amp, amplification; Del, deletion; Neu, neutral.

Figure 7.

TNFRSF17 copy number and BCMA protein expression correlate with PFS. (A) Tumor cell TNFRSF17 mRNA expression, gene activity score (normalized counts within chromatin accessible gene promoters), copy number, BCMA MFI, and sBCMA in relation to PFS in patients receiving anti-BCMA CAR T-cell therapy (n = 15). (B) Kaplan-Meier survival curve by TNFRSF17 copy number status detected in plasma cells before GSI exposure. (C) Percent change in sBMCA levels measured in the serum after GSI treatment compared with pretreatment baseline levels. All comparisons to baseline except the 1-hour measurement were statistically significant (Wilcoxon rank-sum test, P < .05).