CXCR4 orchestrates the TOX-programmed exhausted phenotype of CD8+ T cells via JAK2/STAT3 pathway

Abstract

Evidence from clinical trials suggests that CXCR4 antagonists enhance immunotherapy effectiveness in several cancers. However, the specific mechanisms through which CXCR4 contributes to immune cell phenotypes are not fully understood. Here, we employed single-cell transcriptomic analysis and identified CXCR4 as a marker gene in T cells, with CD8⁺PD-1^high exhausted T (T_ex) cells exhibiting high CXCR4 expression. By blocking CXCR4, the T_ex phenotype was attenuated in vivo. Mechanistically, CXCR4-blocking T cells mitigated the T_ex phenotype by regulating the JAK2-STAT3 pathway. Single-cell RNA/TCR/ATAC-seq confirmed that Cxcr4-deficient CD8⁺ T cells epigenetically mitigated the transition from functional to exhausted phenotypes. Notably, clinical sample analysis revealed that CXCR4⁺CD8⁺ T cells showed higher expression in patients with a non-complete pathological response. Collectively, these findings demonstrate the mechanism by which CXCR4 orchestrates CD8⁺ T_ex cells and provide a rationale for combining CXCR4 antagonists with immunotherapy in clinical trials.

Keywords: CD8 T cells; CXCR4; JAK2-STAT3 pathway; PD-1; T cell exhaustion; immune checkpoint blockade; immunotherapy; single-cell ATAC-seq; single-cell RNA-seq; tumor microenvironment.

Graphical abstract

Figure 1

Blocking CXCR4 exerted anti-tumor effects in syngeneic immunocompetent mice by enhancing the functional T cell phenotype and attenuating T_ex phenotype (A–C) Schematic representation of experimental design and treatment timeline (A). The representative in vivo images (B) and statistical analysis (C) show the effect of plerixafor on xenograft tumors in syngeneic immunocompetent (C57BL/6 or BALB/c) and immunodeficient (NOD-SCID) mice. p values were calculated by the Mann-Whitney test in GraphPad Prism. (D) Survival analysis of plerixafor on xenograft tumors in syngeneic immunocompetent (C57BL/6 or BALB/c) and immunodeficient (NOD-SCID) mice. p values were calculated by the log-rank test in GraphPad Prism. (E) Flow cytometry analysis of ID8-derived xenograft ascites with different treatment groups. p values were calculated by unpaired, parametric t test in GraphPad Prism. (F) Flow cytometry analysis of U14-derived xenograft tumors with different treatment groups. p values were calculated by unpaired, parametric t test in GraphPad Prism. (G) Statistical analysis of flow cytometry analysis from Cxcr4-cKO transgenic mice and control mice with U14-xenograft. p values were calculated by unpaired, parametric t test in GraphPad Prism. Data are presented as mean ± SD. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001; ns, no significance.

Figure 2

CXCR4 inhibition promoted anti-PD-1 immunotherapy by reducing the TOX-associated exhausted phenotype in vivo (A) Tumor volume curve for mice bearing U14 tumors treated with a control (saline and immunoglobulin G), anti-PD-1 antibody, plerixafor, and combination treatment. p values were determined using two-way ANOVA using Geisser-Greenhouse correction. (B) Tumor anatomy picture and statistical analysis for mice bearing TC-1 tumors with different treatments. p values were determined using the unpaired, parametric t test. (C) Kaplan-Meier survival curve for mice bearing ID8 tumors with different treatments. p values were determined using the log-rank test. (D) Mean volumes of 4T1 tumors in primary recipients with different treatments. p values were determined using two-way ANOVA using Geisser-Greenhouse correction. (E) Anatomical images of the abdominal wall for mice bearing ID8 tumors treated with a blank control and anti-PD-1 antibody in Cxcr4-cKO and control mice. (F) Statistical analysis of flow cytometry results in U14-derived xenograft tumors with different treatments. p values were calculated by unpaired, parametric t test in GraphPad Prism. (G) Flow cytometry and statistical analyses of ID8-derived xenograft tumors in Cxcr4-cKO and control mice with different treatment groups. p values were calculated by unpaired, parametric t test in GraphPad Prism. (H) Venn plots of upregulated DEGs of different treatments and DEGs of Tox^−/− mice from GEO: GSE131871. (I) Gene set enrichment analysis and normalized enrichment scores of transcriptional signatures using the DEGs in plerixafor-treated versus control and combination-treated versus control. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001.

Figure 3

scRNA-seq analysis identified that CXCR4 inhibition enhanced anti-PD-1 immunotherapy by attenuating the T_ex phenotype (A) Schematic representation of experimental design and treatment. (B) UMAP plot displaying the 11 major cell types of U14-derived xenograft tumors based on scRNA-seq. (C) UMAP plot displaying the 11 major cell types of U14-derived xenograft tumors in different treatment groups. (D) The proportions and cell number of cell types in each sample. (E) UMAP plot displaying the CD8⁺ T cells of U14-derived xenograft tumors in different treatment groups. (F) Volcano plot showing the DEGs of CD8⁺ T cells between combination-treated and control groups. (G) UMAP plot and RNA velocity displaying the four major cell types of CD8⁺ T cells based on scRNA-seq data. (H) Pseudotime heatmap displaying the genes involved in the developmental process of CD8⁺ T cells.

Figure 4

CXCR4 inhibition promoted CD8⁺ T cell activation by mediating abundant CD8⁺ TCR clonotypes and increasing MHC-I molecule expression (A) UMAP plot displaying the distribution of TCR types in cell types based on scTCR-seq. (B) The number of CDR3 (AA), TCRα, and TCRβ in different treatment groups. (C) The number of clonotypes shows differences among different treatment groups. (D) UMAP plot displaying the distribution of TCR types in CD8⁺ T cells based on scTCR-seq. (E) Column chart showing the number of clone types in different treatment groups. (F) Network plot showing the TCR specificity groups with at least one cell from the different treatments. (G) Violin plots of MHC-I expression levels in dendritic cells, monocytes/macrophages, and B cells within different treatment groups. p values were calculated by the Mann-Whitney test in GraphPad Prism. Each group was compared with the control group, and a significant p value is indicated by ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001.

Figure 5

CXCR4 inhibition mitigated the TOX-mediated T_ex phenotype by JAK2-STAT3 pathway regulation to reduce p-STAT3 and TOX interactions (A) Western blots showing the results of the JAK-STAT pathway and the exhausted markers after plerixafor or NUCC-390 treatment in Jurkat T cells. The JAK-STAT pathway and the exhausted markers were then detected by western blot after use of the CXCR4 antagonist (plerixafor) and JAK2-STAT3 inhibitor (SD-1029) in Jurkat T cells. (B) Western blots showing the results of the T_ex phenotype in the JAK2-STAT3 pathway treated by motixafortide in Jurkat and mouse T cells. (C) Co-immunoprecipitation assay showing the interaction between p-STAT3 and TOX in human and mouse T cells. (D) Co-immunoprecipitation assay and western blot results of the spleens in Cxcr4^flox/floxLck^Cre and Cxcr4^flox/flox mice show the interactions between p-STAT3 and TOX, and the changes of T_ex phenotype and the phosphorylation of the JAK2-STAT3 pathway. (E) Mean enrichment profiles along gene bodies (±3 kb) of peaks and TSS profile plots for p-STAT3 and TOX ChIP-seq treated by plerixafor in Jurkat T cells. TSS, transcription start site; TES, transcription end site. (F) Mean enrichment profiles along gene bodies and TSS profile plots of peaks for p-STAT3 and TOX ChIP-seq treated by plerixafor in mouse spleen. (G) TSS profile plots of ICGs for p-STAT3 and TOX ChIP-seq treated by plerixafor in Jurkat T cells. (H) Schematic representation of the experimental design. (I) Pictures of the anatomy of intestinal tissues with different treatments. (J) Kaplan-Meier analysis of mice bearing tumors transplanted with CD8⁺ T cells pretreated with vehicle control, SD-1029, or STAT3-IN-9. p values were determined using a log-rank test in GraphPad Prism.

Figure 6

Single-cell chromatin accessibility profiles revealed that CXCR4 deficiency epigenetically orchestrated the exhausted and functional phenotype of CD8⁺ T cells (A) Schematic representation of the experimental design for scATAC-seq and scRNA-seq. (B) UMAP plot displaying the 14 major cell types of spleen based on scATAC-seq and scRNA-seq. (C) Violin plots showing the expression levels of Cxcr4 in the CD8⁺ T cell cluster. (D) mIHC staining of spleen tissues validated the expression levels of CXCR4 and exhausted markers in CD8⁺ T cells. (E) The UMAP plot displays eight subtypes of CD8⁺ T cells based on scATAC-seq and scRNA-seq. The violin plot shows the marker genes associated with eight subtypes of CD8⁺ T cells. (F) The proportions of eight subtypes of CD8⁺ T cells in each sample. (G) Peaks plots in exhausted markers show the peaks in CD8⁺ T cells and other cell clusters excluded T cells in each sample. (H) Peaks plots in functional markers show the peaks in CD8⁺ T cells and other cell clusters excluded T cells in each sample. (I) Assessment of chromatin accessibility levels in the CD8⁺ T cell cluster via Cicero. (J) Statistical analysis of IHC staining with different treatments. p values were determined using an unpaired, parametric t test in GraphPad Prism. ∗∗∗∗p < 0.0001.

Figure 7

Clinical applications demonstrated that CXCR4 expression was associated with ICB response status in cancers (A–D) mIHC staining in surgical samples from cervical cancer undergoing NACT with cisplatin plus paclitaxel and combined with anti-PD-1 treatment (A). Statistical analysis of the expression of p63 (B), percentage of CXCR4⁺CD8⁺ and GZMB⁺CD8⁺ T cells (C), and number of CD8⁺ T cells (D) in samples. p values were determined using the Mann-Whitney test in GraphPad Prism. (E) The expression levels of CXCR4 in responder (R) and non-responder (NR) groups from ICBatlas. Patients who achieved complete response or partial response were categorized as R, while those experiencing stable disease or progressive disease were classified as NR. (F) CXCR4 expression in melanoma before (pre-) and after (post-) the anti-CTLA4 therapy (ipilimumab). The DEGs in RNA-seq were calculated by DESeq2 with parameters (|log₂ fold change| > 1 and p < 0.05). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001.