CXCL16/CXCR6/TGF-β Feedback Loop Between M-MDSCs and Treg Inhibits Anti-Bacterial Immunity During Biofilm Infection

Abstract

Staphylococcus aureus (S. aureus) is a leading cause of Periprosthetic joint infection (PJI), a severe complication after joint arthroplasty. Immunosuppression is a major factor contributing to the infection chronicity of S. aureus PJI, posing significant treatment challenges. This study investigates the relationship between the immunosuppressive biofilm milieu and S. aureus PJI outcomes in both discovery and validation cohorts. This scRNA-seq analysis of synovium from PJI patients reveals an expansion and heightened activity of monocyte-related myeloid-derived suppressor cells (M-MDSCs) and regulatory T cells (Treg). Importantly, CXCL16 is significantly upregulated in M-MDSCs, with its corresponding CXCR6 receptor also elevated on Treg. M-MDSCs recruit Treg and enhance its activity via CXCL16-CXCR6 interactions, while Treg secretes TGF-β, inducing M-MDSCs proliferation and immunosuppressive activity. Interfering with this cross-talk in vivo using Treg-specific CXCR6 knockout PJI mouse model reduces M-MDSCs/Treg-mediated immunosuppression and alleviates bacterial burden. Immunohistochemistry and recurrence analysis show that PJI patients with CXCR6^high synovium have poor prognosis. This findings highlight the critical role of CXCR6 in Treg in orchestrating an immunosuppressive microenvironment and biofilm persistence during PJI, offering potential targets for therapeutic intervention.

Keywords: Immunosuppression; M‐MDSCs; Periprosthetic joint infection; Staphylococcus aureus; Treg.

Figure 1

The single‐cell landscape of PJI, AF and OA knee synovial tissues. a) Schematic illustration of the experimental strategy. The human knee joint synovium scRNA‐seq data was utilized as a discovery cohort. b) Atlas of 151185 single cells collected from 6 PJI, 4 AF and 3 OA knee synovial tissues, displaying seven major cell types by Uniform Manifold Approximation and Projection (UMAP) plot. c,d) The UMAP plot showed commonalities of each sample and different groups after correcting for batch effects by Harmony. e,f) The violin plot and UMAP plot showed the expression level of biomarkers for major cell types. g, The bar chart showed the proportion of each cell type in PJI, AF and OA. The color panel indicated different cell types in the scRNA‐seq data. h,i) Milo analysis revealed differential abundance of cell neighborhoods in PJI versus AF and PJI versus OA in synovial cell subpopulations.

Figure 2

PJI samples exhibited increased infiltration of M‐MDSCs accompanied by enhanced immunosuppressive activity. a) The UMAP plot showed detailed annotation of the myeloid subpopulations after re‐clustering. Seven myeloid clusters were visualized. Macro: Macrophage, Mono: Monocyte, MSTR: Monocyte‐stress response, DCs: Dendritic cells, OCs: Osteoclasts. b) The UMAP plot showed the expression level of biomarkers for myeloid cell subpopulations. c) Changes in the composition of the myeloid compartment among PJI, AF and OA were visualized as cell density. d) The bar chart showed the proportion of each myeloid subpopulation in different samples (PJI, n = 6; AF, n = 4; OA, n = 3). e) The scatter plot compared the relative abundance of myeloid cell subpopulations in PJI versus AF and OA. The x and y axes represented the log2 fold change and p‐value, respectively, based on the Mann‐Whitney U test. Each dot represented a specific cell type. f) Violin plots showed the MDSCs scores in seven myeloid subpopulations, as well as in PJI, AF and OA. g,h) Milo analysis revealed differential abundance of cell neighborhoods in PJI versus AF and PJI versus OA in the myeloid subpopulations. i) Gene set variation analysis (GSVA) showed different enriched pathways of the seven myeloid subpopulations. j) The flow cytometry plots depicted the gating strategy employed to identify M‐MDSCs derived from PJI samples (left), AF samples (middle), and OA samples (right). Percentages were calculated over target populations. k,l) Quantitative analysis of M‐MDSCs and myeloid cells in PJI, AF, and OA (PJI, n = 5; AF, n = 5; OA, n = 5). Unpaired t test; ***p < 0.001, ****p < 0.0001.

Figure 3

PJI samples exhibited increased infiltration of Treg and low‐activated phenotype of other lymphocytes. a) The UMAP plot showed detailed annotation of the lymphocytes subpopulations after re‐clustering. Eleven lymphocytes clusters were visualized. Tex: Exhaust T cell, Treg: Regulatory T cell. b) The UMAP plot showed the expression level of biomarkers for lymphocytes subpopulations. c) Changes in the composition of the lymphocytes compartment among PJI, AF and OA were visualized as cell density. d) The bar chart showed the proportion of each lymphocyte subpopulation in different samples (PJI, n = 6; AF, n = 4; OA, n = 3). e) The scatter plot compared the relative abundance of lymphocytes subpopulations in PJI versus AF and OA. The x and y axes represented the log2 fold change and p‐value, respectively, based on the Mann‐Whitney U test. Each dot represented a specific cell type. f) Violin plots showed the naive, exhausted and regulatory scores of eleven lymphocytes subpopulations. g,h) Milo analysis revealed differential abundance of cell neighborhoods in PJI versus AF and PJI versus OA in the lymphocytes subpopulations. i) GSVA showed immune‐related signaling pathways of the lymphocytes subpopulations PJI versus AF, and PJI versus OA. j) The flow cytometry plots depicted the gating strategy employed to identify Treg derived from paired PJI samples (left), AF samples (middle), and OA samples (right). Percentages were calculated over target populations. k) Quantitative analysis of Treg in PJI, AF, and OA (PJI, n = 5; AF, n = 5; OA, n = 5). Unpaired t test; **p < 0.01, ***p < 0.001.

Figure 4

The interaction of MDSCs and Treg by CXCL16‐CXCR6 signaling was obvious in PJI. a) The interaction plot showed the cell communications between myeloid cells and lymphocytes by ligand‐receptor pair analysis. The thicker the line represented, the stronger the interaction weights/strength, and the more the number of interactions among the cell types. b) The bar chart showed the interaction strength and the number of interactions among PJI, AF and OA. c) The bar chart showed the strength comparison of the interaction pathways among PJI, AF and OA. d) The heatmap showed relative strength of all enriched signals (outgoing and incoming) across myeloid cells and lymphocytes. e) The circle plot showed the inferred intercellular communication network for CXCL signaling pathway. f) The heatmap showed the CXCL signaling pathway among myeloid cells and lymphocytes. g) The bar chart showed the ligand‐receptor pair of CXCL signaling pathway among myeloid cells and lymphocytes. h) The bar chart showed the relative contribution of ligand‐receptor pair to the overall communication network of CXCL signaling pathway. i) The dot plot showed the incoming communication patterns of CXCL signaling pathway from M‐MDSCs among PJI, AF and OA. j) Multiplex immunofluorescence image showed the co‐location of CXCR6+ Treg and CD68+ myeloid cells with excretive CXCL16. Yellow arrow highlighted the CD68+ myeloid cells expressing CXCL16, and red arrow highlighted the CD4+FOXP3+ Treg expressing CXCR6. Scale bar, 20 mm (PJI, n = 3; AF, n = 3; OA, n = 3). Unpaired t test; **p < 0.01, ***p < 0.001.

Figure 5

PJI patients with high CXCR6 showed a higher recurrence rate. a) The Volcano plot showed differentially expressed genes in PJI versus AF patients. The x and y axes represented the log2 fold change and p‐value, respectively, based on the Mann‐Whitney U test. Red dots: upregulated genes, blue dots: downregulated genes. b) The heatmap showed the distribution of 28 types of immune cells in PJI and AF. c) The bar chart showed the difference in proportion of M‐MDSCs and Treg between PJI and AF (PJI, n = 18; AF, n = 18). Unpaired t test; ***p < 0.001. d) The heatmap showed the characteristic genes of chemokines, PJI markers, cell markers and immune checkpoint between PJI and AF. e–l) The violin plots showed part of the differential genes (CXCL16, CXCR6, S100A8, S100A9, FOXP3, CTLA4, TIGIT and LAG3) between PJI and AF as shown in (d). Unpaired t test; ^ns p > 0.05, ***p < 0.001, ****p < 0.0001. m,n) Quantitative analysis of immunohistochemical staining for CXCL16 and CXCR6 (PJI, n = 80; AF, n = 25; OA = 25). 40x magnification. Scale bar, 20 mm. Unpaired t test; **p < 0.01, ***p < 0.001. o) The Kaplan‐Meier curve showed the recurrence curves of PJI patients characterized by either low or high expression of CXCR6 (CXCR6‐high PJI, n = 30; CXCR6‐low PJI, n = 30). Significance was calculated using the log‐rank test. p,q) Quantitative analysis of Harris Hip score and HSS knee score in CXCR6‐high PJI and CXCR6‐low PJI group (CXCR6‐high PJI, n = 30; CXCR6‐low PJI, n = 30). Unpaired t test; *p < 0.05.

Figure 6

CXCR6 is a key gene for Treg recruitment in the synovial microenvironment of PJI. a) The UMAP plot showed detailed annotation of the lymphocytes subpopulations in synovial cells of mice knee joint from NC (n = 6), PJI (n = 6) and CXCR6‐KO PJI (n = 6) group. Twelve myeloid clusters were visualized. Treg Regulatory T cell, Th17 T helper cell 17, Pro_T Proliferative T cell. b) The UMAP plot showed the expression level of biomarkers for lymphocytes subpopulations. c) Changes in the composition of the lymphocytes subpopulations compartment among NC, PJI and CXCR6‐KO PJI group were visualized as cell density. d) The heatmap showed the average expression of inhibitory factors (Foxp3, Ctla4, Il10, Tgfb1, Ebi3, Tigit and Il12a) and cytokines (Ifng, Il2 and Il17a) in NC, PJI and CXCR6‐KO PJI group. e–g) Expression of iNOS, IL10 and Arg‐1 protein in M‐MDSCs between CXCR6‐KO PJI and PJI group (CXCR6‐KO PJI [flox/flox], n = 5; PJI [+/+], n = 5). Unpaired t test; *p < 0.05, **p < 0.01. h) T cell suppression assay to evaluate the capacity of M‐MDSCs to suppress T cells (n = 3). Unpaired t test; ***p < 0.001, ****p < 0.0001. i) The Venn plot illustrated the number of differentially expressed genes of Treg in the PJI group compared with the NC and CXCR6‐KO PJI groups. j) GSVA showed the upregulated pathways of Treg in the PJI group compared with the NC and CXCR6‐KO PJI groups. k) The number of synovial M‐MDSCs with or without the Treg cell‐related cytokines TGF‐β (n = 3). Unpaired t test; ***p < 0.001. l–n) IL‐10, iNOS, and Arg‐1 levels in supernatant of M‐MDSCs in different experimental groups. Anti‐TGF‐β antibody (10 mg/mL) was used to neutralize TGF‐β. Data were representative of three independent experiments (n = 3). Unpaired t test; ^ns p > 0.05, *p < 0.05, **p < 0.01.

Figure 7

Target CXCR6 inhibit PJI progression in Treg CXCR6 specific KO PJI mouse model. a) Representative sagittal X‐ray images, 3D reconstruction images of horizontal views and sagittal views demonstrating knee synovial tissues and bones changes from NC, PJI, and CXCR6‐KO group at different time points (7day, 35day) (n = 6, per group). b,c) Quantitative micro‐CT analysis of bone volume fraction (BV/TV) and Trabecular Number (TB.N). Unpaired t test; ^ns p > 0.05, *p < 0.05, **p < 0.01. d) The representative images depicted colony forming units (CFUs) of S. aureus on the surface of the prosthesis, which were diluted 10‐fold (10X) and 100‐fold (100X) in the original solution (1X). (n = 6, per group). e) The numbers of bacteria isolated from the surface of the prosthesis. Unpaired t test; ^ns p > 0.05, *p < 0.05, **p < 0.01. f) The formation of S. aureus biofilm on the surface of prosthesis under SEM. The red star represented Ti particles; The yellow arrow represented S. aureus biofilm. g) Representative H&E‐stained photomicrographs of knee joint sagittal sections (n = 6, per group). h–k) Quantitative analysis of histopathological scoring including necrosis, inflammation, fibrovascular tissue formation and implant osseointegration score. Unpaired t test; ^ns p > 0.05, *p < 0.05, **p < 0.01. l) Weight‐bearing was assessed to demonstrate the motor function of mice (n = 6, per group). Forelimbs (dark blue ink); hindlimbs (red ink). m) The grade of weight‐bearing activity among different groups. Unpaired t test; ^ns p > 0.05, *p < 0.05, **p < 0.01.