Chemotaxis overrides the killing response in alloreactive CTLs, providing vascular immune privilege during cellular rejection

Abstract

Graft endothelial cells (ECs) express donor alloantigens and encounter cytotoxic T lymphocytes (CTLs) but are generally spared during T cell-mediated rejection (TCMR), which predominantly affects epithelial structures. The mechanisms underlying this vascular immune privilege are unclear. Transcriptomics analyses and endothelial-mesenchymal transition assessments confirmed that the graft endothelium was preserved during TCMR. Coculture experiments revealed that endothelial and epithelial cells were equally susceptible to CTL-mediated lysis, ruling out cell-intrinsic protection. Intravital microscopy of murine kidney grafts and single-cell RNA-Seq of human renal allografts demonstrated that CTL interactions with ECs were transient compared with epithelial cells. This disparity was mediated by a chemotactic gradient produced by graft stromal cells, guiding CTLs away from ECs toward epithelial targets. In vitro, chemotaxis overrode T cell receptor-induced cytotoxicity, preventing endothelial damage. Finally, analysis of TCMR biopsies revealed that disruption of the chemotactic gradient correlated with endothelialitis lesions, linking its loss to vascular damage. These findings challenge the traditional view of cell-intrinsic immune privilege, proposing a cell-extrinsic mechanism, in which chemotaxis preserves graft vasculature during TCMR. This mechanism may have implications beyond transplantation, highlighting its role in maintaining vascular integrity across pathological conditions.

Keywords: Cellular immune response; Chemokines; Endothelial cells; Immunology; Nephrology; Transplantation.

Figure 1. Opposite profiles of damage compartmentalization in AMR and TCMR.

(A) Unsupervised clustering analysis of healthy donor kidney scRNA-Seq data was performed to determine endothelial (red) and tubular epithelial (blue) lineage-specific gene lists. (B–D) Comparison of transcriptomics profiles of AMR and TCMR samples. Microarray analyses of 105 AMR, 67 TCMR, and 501 NR graft biopsies were used to establish the transcriptomics profiles of AMR and TCMR using NR samples as a reference. DEGs in AMR and TCMR were filtered out using the endothelial and tubular epithelial lineage gene lists established in A. (B) Volcano plots showing the change in expression of endothelial and epithelial lineage genes in AMR and TCMR samples. (C) Comparison of the proportions of epithelial and endothelial genes in DELGs. **P < 0.01, by χ² test. (D) GSEA of endothelial (red) and tubular epithelial (blue tones) lineage gene sets in AMR and TCMR samples using NR as a reference. (E and F) Expression of EndMT markers (fascin, hsp47, vimentin) in biopsies of transplanted kidneys with no rejection features (NR, negative control, n = 9), TCMR (n = 22), or AMR (n = 12). (E) Representative microscopy features (scale bars: 40 μm). (F) Each EndMT marker and the EndMT score were assessed by a trained nephropathologist blinded to the clinical data using a semiquantitative scale in the PTC ECs identified using a CD34-based mask. Data represent the mean ± SEM. **P < 0.01, by unpaired, 2-sample Wilcoxon test.

Figure 2. Alloreactive CTLs interact with allogeneic ECs in vivo and mediate their destruction in vitro.

(A and B) Confocal microscopy analyses of renal allograft biopsies with TCMR. (A) The donor origin of graft cells was confirmed by the expression of donor-specific mismatched HLA A24 molecules (green) on CD31⁺ (red) ECs (glomerular and peritubular capillaries) as well as on E-cadherin⁺ (blue) tubular epithelial cells (scale bars: 20 μm). (B) Percentage of CD31⁺ peritubular capillary and glomerular ECs (upper row, red) and tubular epithelial cells (lower row, blue) expressing donor-specific HLA A24 molecules. Results are from the analysis of distinct fields of 4 independent biopsies (mean ± SEM). (C) IHC analyses of renal allograft biopsies with TCMR. CD8⁺ CTLs (brown) interacted with graft tubular epithelial cells (upper row) as well as the CD34⁺ ECs (red; lower row) of glomerular (left panels) and peritubular (right panels) capillaries (scale bars: 20 μm). (D) Electron micrographs of the interactions of CTLs with the various compartments of a renal allograft (middle panel). Magnification of the interactions of CTLs with, respectively, the EC (endoth cell) of a peritubular capillary (left panels) and a tubular epithelial cell (right panels) is shown (scale bars: 5 μm). (E–H) Quantification of allospecific, CTL-mediated killing of glomerular ECs (ciGENC, red) and PT epithelial cells (HK-2, blue). (E) Target cell destruction by nonspecific or allospecific CTLs was assessed using time-lapse microscopy. The decrease in live cell area (calcein, green) and acquisition of the apoptosis marker (annexin V, red) were quantified. Representative images from the end of cocultures are shown (scale bar: 50 μm). (F) Quantification of cell destruction from 2 independent time-lapse experiments (median ± IQR ). Significance was determined by unpaired, 2-sample Wilcoxon test. (G) Kinetics monitoring of ciGENC (red) and HK-2 (blue) target cell destruction by allospecific CTLs based on impedance reduction in culture wells. Impedance values were normalized to control conditions with nonspecific CTLs. (H) AUC of the impedance values for ciGENC and HK-2 cocultures at varying effector/target ratios. Results from 2 independent experiments are shown. Significance was determined by 1-way ANOVA.

Figure 3. Distinct contact durations of alloreactive CTLs with endothelial and tubular epithelial cell targets.

(A) Schematic representation of the murine model of TCMR of renal allografts. (B and C) Representative findings of intravital microscopy analysis of OVA-specific OT-I (yellow) or control P14 (cyan) CTLs trafficking within B6-OVA renal allografts. The vascular compartment is identified by fluorescent dextran (red). (B) Global view. The tracks of the cells were color-coded according to their position: intravascular (red) or extravascular (blue). Scale bars: 50 μm. (C) Representative behavior of OT-I (left columns) and P14 (right columns) cells in the intravascular (upper rows) and extravascular (lower rows) compartments (scale bars: 10 μm; time stamps format = mm:ss). (D–F) Comparison of the trafficking behavior of OT-I and P14 cells in intravascular (red) and extravascular (blue) compartments of the graft. Results in D and E represent independent experiments. (D) OVA-specific OT-I and control P14 CTLs were cotransferred into a mouse that had received a B6-OVA renal allograft (n = 1). Each symbol corresponds to a tracked cell. (E) OVA-specific OT-I CTLs were transferred alone into mice that had received a B6-OVA renal allograft (n = 2 experiments involving 2 animals each; each shape represents an individual animal). Individual cells are represented by a small symbol, and the larger symbol is the mean for the animal. Data represent the mean ± SEM. **P < 0.01 and ***P < 0.001, by 2-sided Student’s t test. (F) Overlay of individual OT-I T cell tracks plotted after aligning their starting positions. Cells were tracked over a 30-minute period in the intravascular (red) or extra vascular (blue) compartments. (G and H) Quantification of infiltrating CTLs in vascular and tubular epithelial compartments of 5 renal allograft biopsies with TCMR. (G) Computer-assisted quantification of CTLs (CD8⁺, brown) in the tubular epithelial (COL-IV, red, left column) and vascular (CD34⁺, red, right column) compartments of rejected renal allografts. CTLs were automatically counted within (green circles) and outside (black circles) each compartment (scale bars: 50 μm). (H) The density of CTLs in intravascular (red) and tubular epithelial (extravascular, blue) compartments was compared. E/T, effector/target ratio. ***P < 0.001, by Wilcoxon signed-rank test.

Figure 4. Chemotaxis protects graft endothelium against alloreactive T cell–mediated cytotoxicity in vitro.

(A) UMAP plot of 4 cell clusters (ECs, epithelial cells, stromal cells, and infiltrating T cells) identified by scRNA-Seq analysis of a rejected kidney allograft. (B) Bubble plots comparing the expression of 18 cytokines genes by ECs (end), epithelial cells (tub), and stromal (str) cells of the rejected graft. Bubble size is proportional to the percentage of cells in a cluster expressing the gene, and color intensity is proportional to the average scaled gene expression level (avg. expr). (C) Relative expression of CXCL12 (left panel) and CXCR4 (right panel) overlaid on UMAP plot. Color intensity is proportional to average scaled gene expression. (D–H) In vitro modeling of dynamic and static CTL–target cell interactions. (D) Schematic representation of the dynamic and static assays. chemok., chemokine. (E) ciGENC target cells were cultured with 80 ng/mL CXCL12 in static (sta) or dynamic (dyn) assay, with nonspecific (left column) or allospecific (right column) CTLs (CTV stained, blue). Target cell destruction was monitored by the decrease in live cell area (calcein, green) and acquisition of the apoptosis marker (ethidium bromide, red) measured by fluorescence microscopy. Representative images at the end of cocultures are shown (scale bars: 50 μm). (F) Quantification of ciGENC target cell destruction. Results (mean ± SEM) of 2 independent experiments are shown. ****P < 0.0001, by unpaired, 2-sample Wilcoxon test. (G) The same experiments were conducted with increasing concentrations (0–320 ng/mL) of CXCL12 and other chemokines (CXCL9 and CCL2). Where indicated, the chemokine was placed in the upper (top, 320 ng/mL) chamber of the assay. Each symbol shape represents an individual experiment (CXCL12, n = 4; CXCL9, n = 3; CCL2, n = 3). Small symbols are replicates, and larger symbols indicate the mean. (H) The same experiments (n = 2) were performed using the tubular epithelial cell line HK2 as target cells. ****P < 0.0001, by two-sided Student’s t test. Data are presented as mean ± SEM (G and H).

Figure 5. Disruption of the chemokine gradient associates with endothelial damage in clinical TCMR.

(A–C) ciGENC cells were cultured to confluence under flow conditions, and their morphology was monitored in real time by video microscopy. (A) Representative bright-field images. The first column (No effect) shows ciGENC morphology in the absence of CTLs, while the other columns show their response to the addition of either nonspecific CTLs or allospecific CTLs. Where indicated, CXCL12 was incorporated into the coating beneath the ciGENC. Interactions between CTLs and ciGENC are marked with orange arrowheads. (B) Magnified time-lapse sequence illustrating the morphological changes of a ciGENC engaged by an allospecific CTL (red) in the absence of CXCL12. (C) Quantification of ciGENC morphological changes over time across the different experimental conditions. (D–F) Comparison of the CXCL12 gradient in 12 TCMR kidney graft biopsies with (v >0, purple, n = 6) or without (v = 0, orange, n = 6) endothelialitis (D) Banff i (faded color) and t (dark color) scores for each biopsy are plotted. (E) Representative images of CXCL12 staining (left column) (scale bar: 75 μm) and corresponding computer-assisted quantification of the chemotactic gradient (right column). Percentages of biopsy surface area (F) and staining intensity (G) for CXCL12 were compared between TCMR kidney graft biopsies with (purple, End⁺) and without (orange, End^–) endothelialitis. *P < 0.05, by nonparametric Mann-Whitney U test. Data are presented as median ± IQR (F and G).