Organ-specific features of human kidney lymphatics are disrupted in chronic transplant rejection

Abstract

Lymphatic vessels maintain tissue fluid homeostasis and modulate inflammation, yet their spatial organization and molecular identity in the healthy human kidney, and how these change during chronic transplant rejection, remain poorly defined. Here, we show that lymphatic capillaries initiate adjacent to cortical kidney tubules and lack smooth muscle coverage. These vessels exhibit an organ-specific molecular signature, enriched for CCL14, DNASE1L3, and MDK, with limited expression of canonical immune-trafficking markers found in other organ lymphatics, such as LYVE1 and CXCL8. In allografts with chronic mixed rejection, lymphatics become disorganized and infiltrate the medulla, with their endothelial junctions remodeling from a button-like to a continuous, zipper-like, architecture. Lymphatics in rejecting kidneys localize around and interconnect tertiary lymphoid structures at different maturation stages, with altered intralymphatic and perilymphatic CD4+ T cell distribution. The infiltrating T cells express IFN-γ, which upregulates coinhibitory ligands in lymphatic endothelial cells, including PVR and LGALS9. Simultaneously, lymphatics acquire HLA class II expression and exhibit C4d deposition, consistent with alloantibody binding and complement activation. Together, these findings define the spatial and molecular features of human kidney lymphatics, revealing tolerogenic reprogramming accompanied by structural perturbations during chronic transplant rejection.

Keywords: Adaptive immunity; Immunology; Lymph; Nephrology; Organ transplantation; Vascular biology.

Figure 1. 3D imaging of lymphatics and their spatial relationships in the human kidney.

(A) Representative maximum intensity z-projection, from low-resolution confocal tile scans, of n = 3 human kidney tissues labeled for PDPN and UMOD, demonstrating PDPN⁺ lymphatics (arrowheads). Scale bar: 2,000 μm. (B) Segmented and rendered light-sheet imaging of lymphatics from the same kidney tissue in A, representative of n = 3 images. 3D color renderings represent vessel branch radii, with blue representing the smallest radius (<3.5 μm, asterisks) and red representing the largest radii (>50 μm, arrowheads). (C and D) Representative 3D reconstruction of cortical regions from n = 2 human kidney tissues labeled for PDPN and either PROX1 or LYVE1. The PROX1 signal and LYVE1 signal are masked to only include expression from within the vessel, demonstrating expression of PDPN⁺ cells. Sparse membrane localization of LYVE1 is demonstrated (arrowheads). Representative of 5 regions of interest imaged. Scale bars: 30 μm. (E–G) Regional localization of lymphatics (arrowheads) in the human kidney using LTL (cortex), UMOD (medulla), and UAE-I (with dotted lined delineating the capsule). Regional structures are indicated with asterisks, including proximal tubules in E, loops of Henle in F, and glomeruli in G. Scale bars: 70 μm (E), 150 μm (F), 100 μm (G). (H–K) Spatial relationships of lymphatics (arrowheads) relative to UAE-I⁺ renal arterioles (RA) and glomeruli (G) in H, LRP2⁺ proximal tubules (PT) in I, CALB1⁺ distal nephron tubules (DT) in J, and DBA⁺ collecting ducts (CD) in K. Scale bars: 50 μm (H), 80 μm (I and J), 300 μm (K). (L) Schematic depicting the spatial relationships of lymphatics (arrowheads) to nephron segments. All imaging from E–K representative of 5 regions of interest imaged across n = 2 kidneys.

Figure 2. Profiling kidney lymphatics and their molecular signature through scRNA-Seq of human kidney tissue.

(A) Uniform manifold approximation and projection (UMAP) of an integrated atlas of 217,411 cells, including 151,058 control cells from live biopsies or nephrectomies, 46,540 cells from different etiologies of graft injury, and 19,813 from chronic kidney disease. TREM, triggering receptor expressed on myeloid cells 2. (B) Dot plot of top 20 markers of lymphatic profiles across all control cell types in the atlas. Groupings for each cell type are shown on the right. (C–F) Analysis of nonlymphatic expression of PROX1 and LYVE1 using 3D imaging. Arrowheads show the expression of each marker relative to CDH1⁺ medullary tubules (C), CD31⁺ vasa recta (D), or CD68⁺ macrophages (E) and peritubular capillaries (F). Scale bars: 50 μm (C, E, and F), 30 μm (D). (G and H) Examination of ACTA2 expression relative to PDPN⁺ lymphatics (arrowheads) in the renal hilum (G) and cortex (H). Scale bars: 50 μm (G), 100 μm (H). (I) Subclustering analysis of n = 452 lymphatic endothelial cells (LECs) derived from human control kidney datasets reveals 2 transcriptionally distinct clusters, which we term LEC₁ and LEC₂. (J) Feature plots demonstrating expression of markers of all LECs (PROX1, PDPN), lymphatic capillaries (CCL21, LYVE1), and lymphatic collecting vessels (GATA2, FOXC2). (K) Volcano plot showing differentially expressed genes (DEGs) between the 2 lymphatic subclusters, with each point representing a gene. The x axis represents average log-fold change (log₂FC), whereas the y axis represents –log₁₀ of the adjusted P value of the Wilcoxon rank-sum test for differential expression. Blue dots represent genes that meet significance. Selected marker genes for each cluster are shown in boxes.

Figure 3. A single-cell atlas of human organ lymphatics reveals organ-specific molecular heterogeneity of kidney lymphatic endothelial cells.

(A) Integrated UMAP featuring 13,454 cells from a total of 7 human organs incorporating kidney, skin, breast, heart, lung, small intestine, and large intestine. Unsupervised clustering resulting in 5 transcriptionally distinct clusters of lymphatic cells, which we designate LEC_1, LEC₂, LEC₃, and LEC₄, all of which have capillary identity, and a fifth cluster representing valve LECs. (B) UMAPs highlighting the cells corresponding to each organ and where they are represented within the dataset. Based on this analysis, LEC₁ and LEC₂ are dominated by cells from visceral organs, including kidney, heart, lung, and intestines. Conversely, LEC₃ and LEC₄ are dominated by cells from superficial organs, the skin and breast tissue. All organs show cells mapping to valve LECs. (C) Heatmap showing the top 35 differentially expressed genes (DEGs) enriched in kidney lymphatic cells versus top 35 genes that have low expression by kidney lymphatics compared with other organs. (D) Dot plot of differentially expressed chemokines, interleukins, and immune trafficking receptors across lymphatics of different organs. (E) Expression of DNASE1L3 and MDK (F) at the RNA level in the tubulointerstitium of patients within the publicly available NephroSeq database. Number of patients per condition are shown as follows for DNASE1L3: healthy (n = 8), diabetic kidney disease (DKD, n = 11), focal segmental glomerulosclerosis (FSGS, n = 22), lupus nephritis (n = 31, **P = 0.0013), minimal change disease (MCD, n = 9), and MDK: healthy (n = 14), DKD (n = 10, ****P < 0.0001), FSGS (n = 18, ****P < 0.0001), lupus nephritis (n = 31, ****P < 0.0001), MCD (n = 5). For both genes, significance values represent increase relative to healthy samples.

Figure 4. Structural remodeling of lymphatics in chronic transplant rejection.

(A) 3D renderings of segmented lymphatic networks from donor kidneys and rejecting kidney allografts using LSFM; n = 3 samples per group. Vessel branch radii are color-coded: blue is smallest radius (<3.5 μm; asterisks) and red the largest (>18 μm; arrowheads). (B) Quantitative analysis of lymphatic branching architecture. Vessel metrics are shown per kidney (scatterplot, n = 3 per group) and pooled across vessels (violin plots, n = 75,036 vessels). Vessel density was significantly increased in rejection (95.12 ± 49.21 vs. 690.3 ± 121.6 vessels/mm³, **P = 0.0014, unpaired t test). Vessel length, radius, and branching angle distributions were significantly shifted in rejection (****P < 0.0001 for each; Kolmogorov–Smirnov tests). (C and D) Confocal imaging of PDPN⁺ lymphatic vessels (arrowheads) in cortex adjacent to DBA⁺ tubules (C) and medulla adjacent to UMOD⁺ tubules (D), showing lymphatic expansion in cortex and infiltration into medulla. Representative of 6 regions across n = 3 kidneys/group. Scale bars: 200 μm (C), 100 μm (D). (E and F) 3D reconstruction of CDH5⁺ lymphatic endothelial junctions in control (E) and rejecting (F) kidneys (n = 2 kidneys/group). Junctions identified within PDPN⁺ lymphatics using surface rendering in Imaris. Scale bars: 30 μm. Below: surface-rendered high-magnification views of lymphatic vessel blind ends from E and F, showing discontinuous CDH5⁺ “button-like” junctions in controls and continuous “zipper-like” junctions in rejection. Scale bars: 4 μm (control), 10 μm (rejection). (G and H) Quantification of total PDPN⁺ lymphatic vessel volume per field (G) and density of discontinuous CDH5⁺ junctions per mm³ of vessel volume (H). Each point represents a single image; circles, Repeat 1 and squares, Repeat 2. Rejecting kidneys showed increased lymphatic volume (mean difference = 2.01 × 10^–4 ± 6.83 × 10^–5 mm³) and reduced density of discontinuous junctions (mean difference = 265,674 ± 73,557 discontinuous junctions per mm³).

Figure 5. Spatial association between lymphatics and maturation of tertiary lymphoid structure.

(A) Representative segmented confocal images of PDPN⁺ lymphatics (white arrowhead), CD20⁺ B cells, and CD4⁺ T cells in regions with evidence of ectopic lymphoid aggregation. A tertiary lymphoid structure (TLS) is shown (white asterisk). Representative image of 4 T cell– and B cell–rich TLSs taken from n = 2 rejecting allografts. Scale bar: 40 μm. (B) 3D rendering of TLS interconnected by lymphatics. Such interconnections (white arrowhead) were observed between TLSs in all (n = 3) rejecting allografts imaged. (C) Representative segmented confocal images of TLS, containing PDPN⁺ lymphatics (white arrow), CD21⁺ follicular DCs (FDCs) and peripheral lymph node addressin (PNAd⁺) high endothelial venules (HEVs). Nine TLSs were imaged across n = 3 rejecting allografts. Each image represents TLSs at different stages, with either HEVs absent (early stage; top image), scant (mid-stage; middle image), or present (late-stage, bottom image). Scale bar: 50 μm. (D) Comparison of distance between the CD21⁺ FDC core and lymphatic vessel (green) or HEVs (orange), with each data point representing an individual TLS imaged. Circles represent Repeat 1, squares Repeat 2, and triangles Repeat 3. Lymphatic vessels were significantly closer to CD21⁺ FDCs than HEVs (mean difference = 60.04, 95% CI = 24.33–95.76, **P = 0.0047, unpaired t test).

Figure 6. Molecular and spatial analyses indicate impaired T cell trafficking by kidney lymphatics in alloimmunity.

(A and B) Segmented (A) and rendered (B) confocal images of PDPN⁺ lymphatics (white arrow), CD20⁺ B cells (yellow asterisk), and CD4⁺ T cells (white asterisk). In B, the transparency of rendered lymphatics is increased to visualize intraluminal B cells and T cells. Scale bars: 30 μm. (C and D) Number of intraluminal CD20⁺ B cells (C) or CD4⁺ T cells (D), normalized by volume, was quantified and compared with that of the tissue parenchyma. Each point represents 1 volume of interest imaged, with circles representing Repeat 1 and squares representing Repeat 2. Luminal CD20⁺ B cell density was higher than that of the tissue parenchyma in both control kidneys and rejecting allografts. A similar trend was observed for intraluminal CD4⁺ T cells, with a greater magnitude in increase in density within rejection. (E and F) Spatial point-pattern of perilymphatic CD20⁺ cell (E) or CD4⁺ cell (F) density, where lymphatic branch points represent gray dots and CD20⁺ cells are color-coded according to their density around the lymphatic network. (G and H) Histograms of CD20⁺ cell (G) or CD4⁺ T cell (H) frequency as a function of distance from the nearest lymphatic vessel. P values demonstrate whether lymphocytes are clustered around lymphatics greater than would be expected under complete spatial randomness. The only significant association observed was between CD4⁺ T cells and lymphatics in donor kidneys (*P = 0.029). All imaging data are representative of n = 5 imaging volumes, each acquired from n = 2 allografts with chronic mixed rejection and n = 2 donor controls.

Figure 7. Interrogating kidney lymphatic–T cell crosstalk reveals a type 2 IFN-driven immunoinhibitory molecular landscape in alloimmunity.

(A) Violin plots showing upregulation of IFN-inducible genes IFITM2 and IFITM3 in LECs from rejecting allografts. (B) UMAP of the scRNA-Seq data showing enrichment of IFN-γ within the T/NK cell cluster. (C) UMAP showing enrichment of an IFN-γ signature, including IFNGR1, IFNGR2, IFITM2, and IFITM3. (D) CellPhoneDB interaction map depicting predicted lymphatic-CD4⁺ T cell crosstalk in rejection. Inhibitory interactions (blue) include PVR and LGALS9; stimulatory interactions (red) are also shown. Node size reflects expression frequency; line intensity indicates interaction strength. Ligands of interest, PVR and LGALS9, are highlighted. (E) Heatmap of immune checkpoint interactions between LECs and effector CD4⁺ T cells across disease states. Color indicates normalized CellPhoneDB interaction score. All scores were normalized for each ligand-receptor pair. (F) Immunofluorescence validation of PVR expression on PDPN⁺ lymphatics (arrowhead) in rejecting kidneys (n = 2); CD4⁺ T cell shown in contact (asterisk). Scale bar: 30 μm. (G) IFN-γ stimulation of cultured human LECs increases LGALS9 levels at 24 and 48 hours (qPCR; ***P = 0.0002, **P = 0.0093, respectively) relative to HPRT. (H) LGALS9 protein secretion increased at 48 and 72 hours (ELISA; ***P = 0.0002, ****P < 0.0001, respectively) after IFN-γ stimulation of cultured human LECs. qPCR and ELISA experiments were repeated 3 times, and all assays were performed in duplicate, with each dot on the graph representing the mean data obtained for each repeat.

Figure 8. Donor lymphatics upregulate MHC class II molecules and represent a target for the alloimmune response.

(A) Dot plot of the expression of transcripts encoding MHC class II molecules within lymphatics in the dataset. (B) Single nucleotide variant–based analysis of the origin of lymphatics in allograft tissues from the scRNA-Seq atlas. Cells are grouped by control, chronic rejection, or alternative causes of graft injury. (C) Representative optical z-sections from control and chronically rejecting renal tissue stained for HLA-DR. Isolated, discrete HLA-DR⁺ cells are shown with asterisks in both conditions, whereas in rejection there is also vascular staining (white arrowheads). Representative of 3 nonoverlapping fields of view per kidney, imaged across n = 2 kidneys per group. (D–F) 3D confocal images of HLA-DR expression (arrowheads) in CD31⁺ endothelia (D), CD68⁺ macrophages (E), and PDPN⁺ lymphatics (F). Images are representative of 5 regions imaged across n = 2 kidneys with chronic transplant rejection. All scale bars: 30 μm. (G) Representative 3D reconstructions of n = 2 transplant donor kidney tissues and n = 2 allograft tissues with chronic rejection stained using D2-40 and HLA-DR antibody. The HLA-DR signal is masked by D2-40 expression, such that only the signal inside lymphatics is visible. HLA-DR expression is observed in rejection (see white arrowheads). Three nonoverlapping fields of view per kidney were imaged. Scale bar: 50 μm. (H) 3D confocal images of C4d deposition, representative of 5 regions imaged across n = 2 kidneys with chronic transplant rejection. C4d deposition is observed in PDPN⁺ lymphatics (arrowheads) and presumptive blood capillaries (asterisks). Scale bar: 30 μm.