A phase 2 trial of CXCR4 antagonism and PD1 inhibition in metastatic pancreatic adenocarcinoma reveals recruitment of T cells but also immunosuppressive macrophages

Abstract

Pancreatic ductal adenocarcinoma (PDAC) is characterized by dense stroma and myeloid-rich microenvironment that confer resistance to immunotherapies. Previous studies demonstrated that disrupting the immune-stroma CXCR4-CXCL12 axis facilitates T cell recruitment and mobility to collaborate with anti-PD1/PD-L1 therapy. We sought to test the clinical viability of this immunotherapeutic strategy. 21 patients with metastatic PDAC were enrolled and treated in a phase 2 trial evaluating the effects of the plerixafor/AMD3100 and cemiplimab. Primary endpoint was objective response rate. Blood and tissue biospecimens were collected for correlative analyses. Parallel mouse studies were used to determine potential mechanisms of resistance. Treatments were well-tolerated, but only two patients demonstrated a best response of stable disease. Correlative analyses confirmed significant mobilization of immune cells into circulation as well as increased immune infiltration into the tumor. High-parameter imaging revealed higher levels of CD8⁺ T cells but also granulocytes and macrophages upon treatment. Spatial analysis showed that treatment resulted in closer proximity between macrophages and T cells but not between granulocytes and T cells. Mouse studies revealed that whereas total granulocyte depletion had no effect on immunotherapeutic efficacy, macrophage-targeting yielded significant benefit. Tumor growth measurements and immune profiling of immunotherapeutic combinations incorporating macrophage-targeting showed that despite the increased T cell infiltration, CXCR4 antagonism was in fact associated with enrichment of CD206^hiIA/IE^lo macrophage subtypes and modestly dampened efficacy. Our findings validate the utility of CXCR4 antagonism as an effective immune-recruiting platform but also underscores the need for strategies that better leverage its effects.

Keywords: CXCL12; CXCR4; Cancer immunology; checkpoint immunotherapy; mass cytometry; pancreatic cancer; tumor microenvironment.

Figure 1.

Clinical trial design and results. (A) AMD3100/plerixafor (CXCR4 antagonist) is infused intravenously on days 1–7 every cycle at 80mcg/kg/hr, and cemiplimab (anti-PD1) is administered on day 1 of every cycle at 350 mg IV. Each cycle is 21 days. Spider plots of percent change from baseline in (B) sum of longest diameters (SLD) by RECIST 1.1 and (C) serum measurement of carbohydrate antigen 19–9 (CA19–9) in individual patients over time. Each line represents a unique patient.

Figure 2.

Peripheral blood (PB) correlatives. (A) clinical blood count (K/uL) line plots of neutrophils, lymphocytes, and monocytes pre- and post-AMD3100 treatment show significant p-values (****p ≤ 0.0001, paired t-test). Each line represents a unique patient. (B) Representative flow plots of PB mononuclear cells (PBMC) immune-profiled with cytometry by time-of-flight (CyTOF) show CD34+ cell proportions. These plots compare timepoint-matched samples pre- and post-AMD3100 treatment. (C) Proportion line plot of CD34⁺ cells from CD45⁺ cell population pre- and post-AMD3100 treatment. Density line plot of CD34⁺ of K cells per μL of blood volume pre- and post-AMD3100 treatment with significant p-values (***p < 0.005, paired t-test). (D) UMAP visualization of major cell types identified by CyTOF profiling annotated from FlowSOM clustering. (E) CCR2, CCR5, CXCR2, and CXCR4 expression superimposed on UMAP. Abbreviations: B, B cell; Baso, basophil; DC, dendritic cell; NK, natural killer cell; pDC, plasmacytoid dendritic cell; Tc, cytotoxic T cell; Th, helper T cell; C1D1, cycle 1 day 1; C1D8, cycle 1 day 8; pre-tx, pre-treatment.

Figure 3.

Tumor tissue correlatives. (A) Representative H&E timepoint-matched images of a metastatic liver lesion PreTx and OnTx. All tumor biopsies received blinded inflammation scores by H&E review. Line plot of H&E inflammation scores of liver biopsies at PreTx and OnTx (*p ≤ 0.05, Wilcoxon matched-paired test). Each line represents a unique liver metastasis patient. (B) Paired metastatic liver biopsies were stained with an immune-focused 38-marker imaging mass cytometry (IMC) panel. Representative images of tumor immune microenvironments show αSMA, pan-keratin, CD8, CD15, and CD163 stain PreTx and OnTx. Scale bar: 100 μm. (C) Images of IMC-stained tumor biopsy samples were processed, segmented, and clustered with FlowSOM. A total of 14 annotated clusters (excluding unassigned, UA) were identified. Heatmap shows expression levels of canonical markers across the clusters. (D) Annotated cells are shown in density (cell count/image area) line plots PreTx vs. OnTx for Tc, Treg, and Mac (*p ≤ 0.05, paired t-test). Each line represents a unique liver metastasis patient. Based on X and Y coordinates of each annotated cell, the overall distribution of distances from (E) Th and (F) Tc to other annotated cell types is visualized as violin plots, comparing PreTx and OnTx (**p ≤ 0.01, ***p ≤ 0.001, unpaired t-test). (G) Spatial network visualization maps were generated by visualizing the average shortest distance relationships from each cell type to all other cell types at PreTx and OnTx timepoints. (H) Heatmap shows expression levels of key markers for macrophage phenotypes. Abbreviations: DC, dendritic cell; gran, granulocyte; Mac, macrophage; B, B cell; Tc, cytotoxic T cell; Th, helper T cell; Treg, regulatory T cell; NK, natural killer cells; PreTx, pre-treatment; OnTx, on-treatment.

Figure 4.

Establishing a murine model of PDAC for recapitulating clinical and immunological responses seen in patients (A) Cell lines KPCY 2838c3 (c3) and KPCY 6419c5 (c5) were assayed for CXCL12 gene expression using nanostring, yielding a significant difference across both cell lines (****p ≤ 0.0001, unpaired t-test). (B) Experimental schema for KPCY c5 tumors in C57BL/6J treated with vehicle (VEH) or AMD3100 (day 3 through 17 using 14-day osmotic pumps) and isotype (ISO) or αPD1 started on day 4 given biweekly. N = 5–10 mice per group. (C) Tumor volumes were tracked and measured with a caliper (**p ≤ 0.01, non-linear regression). (D) Subcutaneous tumors were harvested and dissociated into single cells to be analyzed by CyTOF, yielding 12 annotated key clusters (excluding unassigned, UA). Heatmap shows the expression profiles of the annotated clusters. Inset heatmap shows expression profiles of macrophage clusters based on macrophage classification markers. (E) Tc and Th abundance boxplots are shown (*p ≤ 0.05, **p ≤ 0.01, unpaired t-test, N = 5–8 tumors per group). Heatmap shows fold-changes in the abundance of monocyte and macrophage clusters normalized by the VEH+ISO group, compared together across the treatment groups (Friedman test). Pairwise analysis with Wilcoxon test (F) Experimental schema for KPCY c5 tumors in C57BL/6J treated with VEH/BL8040 daily and ISO/αPD1 biweekly starting on day 5. N = 5 mice per group. (G) Tumor volumes were tracked and measured with a caliper (**p ≤ 0.01, non-linear regression). (H) CyTOF analysis of tumors showing fold-changes in the abundance of macrophage and granulocyte clusters normalized by the VEH+ISO group and compared across the groups (Friedman test). Abbreviations: Tc, cytotoxic T cell; Th, helper T cell; DC, dendritic cell; gran, granulocyte (N, neutrophil); mono, monocyte; Mac, tumor associated macrophage; Endoth, endothelial cell; Non_Imm, nonimmune cell; UA, unassigned.

Figure 5.

Determining the role of granulocytes and macrophage in the treatment responses. (A) experimental schema for KPCY c5 tumors in C57BL/6J treated with VEH/BL8040 daily from day 5, ISO/αPD1+αCTLA4 biweekly from day 4, αLy6G on days − 2, 0, and three times weekly thereafter, αCSF1R on days 7 and 14, and αCD40 once on day 7. N = 5 mice per group. (B) On day 21, cheek bleeds were performed to obtain peripheral blood for flow cytometry to verify granulocyte depletion. Representative peripheral blood flow plots show gating of granulocytes (CD11b⁺Ly6G⁺) for four treatment groups. Bar plot shows granulocyte abundance as a percentage of CD45⁺ cells in the blood (***p ≤ 0.001, ****p ≤ 0.0001, unpaired t-test). (C) Tumor volumes were tracked and measured with a caliper (****p ≤ 0.0001, non-linear regression). (D) Tumors were profiled with CyTOF. Abundance boxplots show changes in macrophages and granulocytes within the tumors (*p ≤ 0.05, ***p ≤ 0.001, ****p ≤ 0.0001, unpaired t-test). (E-G) Violin plots show the distribution of per-cell expression of functional markers (CD40, CD74, PD1, ICOS, PDL1, IA/IE, CD69) within key immune clusters compared among the treatment groups (*p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, ****p ≤ 0.0001, unpaired t-test). Group color legends are as shown for panel D. (H) Heatmap shows expression levels of key markers for myeloid phenotypes. Abbreviations: DC, dendritic cell; gran, granulocyte; Mac, macrophage; mono, monocyte; MonoMac, shares markers for both monocytes and macrophages; TcN, naïve cytotoxic T cell; ThN, naïve helper T cell.

Figure 6.

Determining the differential effects of macrophage-targeting agents (A) experimental schema for KPCY c5 tumors in C57BL/6J treated with VEH/BL8040 daily from day 5, ISO/αPD1+αCTLA4 biweekly from day 4, αCSF1R on days 7 and 14, and αCD40 once on day 7. N = 5 mice per group. (B) Tumor volumes were tracked and measured with a caliper (*p ≤ 0.05, **p ≤ 0.01, ****p ≤ 0.0001, non-linear regression). (C-D) Tumors were profiled using CyTOF to identify subtypes of macrophages, dendritic cells, and T cells. Abundance boxplots of all macrophage phenotypes, CD40hi_DC, TcM, and ThM are shown for each treatment group (*p ≤ 0.05, **p ≤ 0.01, unpaired t-test). Group color legends are as shown for panel B. for annotating comparisons, magenta line compares groups with ISO vs. αCSF1R; red line compares groups with ISO vs. αCD40; green line compares groups with αCSF1R vs. αCD40+αCSF1R, and blue line compares groups with VEH vs BL8040. Abbreviations: DC, dendritic cell; Mac, macrophage; TcM, memory cytotoxic T cell; ThM, memory helper T cell.