Single-Cell RNA Sequencing Identifies Response of Renal Lymphatic Endothelial Cells to Acute Kidney Injury

Visual Abstract

Abstract

Significance Statement

The renal lymphatic vasculature and the lymphatic endothelial cells that make up this network play important immunomodulatory roles during inflammation. How lymphatics respond to AKI may affect AKI outcomes. The authors used single-cell RNA sequencing to characterize mouse renal lymphatic endothelial cells in quiescent and cisplatin-injured kidneys. Lymphatic endothelial cell gene expression changes were confirmed in ischemia–reperfusion injury and in cultured lymphatic endothelial cells, validating renal lymphatic endothelial cells single-cell RNA sequencing data. This study is the first to describe renal lymphatic endothelial cell heterogeneity and uncovers molecular pathways demonstrating lymphatic endothelial cells regulate the local immune response to AKI. These findings provide insights into previously unidentified molecular pathways for lymphatic endothelial cells and roles that may serve as potential therapeutic targets in limiting the progression of AKI.

Background

The inflammatory response to AKI likely dictates future kidney health. Lymphatic vessels are responsible for maintaining tissue homeostasis through transport and immunomodulatory roles. Owing to the relative sparsity of lymphatic endothelial cells in the kidney, past sequencing efforts have not characterized these cells and their response to AKI.

Methods

Here, we characterized murine renal lymphatic endothelial cell subpopulations by single-cell RNA sequencing and investigated their changes in cisplatin AKI 72 hours postinjury. Data were processed using the Seurat package. We validated our findings by quantitative PCR in lymphatic endothelial cells isolated from both cisplatin-injured and ischemia–reperfusion injury, by immunofluorescence, and confirmation in in vitro human lymphatic endothelial cells.

Results

We have identified renal lymphatic endothelial cells and their lymphatic vascular roles that have yet to be characterized in previous studies. We report unique gene changes mapped across control and cisplatin-injured conditions. After AKI, renal lymphatic endothelial cells alter genes involved in endothelial cell apoptosis and vasculogenic processes as well as immunoregulatory signaling and metabolism. Differences between injury models were also identified with renal lymphatic endothelial cells further demonstrating changed gene expression between cisplatin and ischemia–reperfusion injury models, indicating the renal lymphatic endothelial cell response is both specific to where they lie in the lymphatic vasculature and the kidney injury type.

Conclusions

In this study, we uncover lymphatic vessel structural features of captured populations and injury-induced genetic changes. We further determine that lymphatic endothelial cell gene expression is altered between injury models. How lymphatic endothelial cells respond to AKI may therefore be key in regulating future kidney disease progression.

Introduction

AKI affects up to 50% of intensive care unit patients and is a significant contributor to the future development of CKD.^1–3 The initial injury extent, inflammatory resolution, and specific immune responses may play a role in the AKI-to-CKD transition.^4,5 Increasingly, studies have demonstrated that the renal lymphatic vasculature expands in response to the increased expression of lymphangiogenic ligands on injury and may regulate the AKI immune response through antigen, solute, fluid, and immune cell transport.^6–9 Lymphatic endothelial cells (LECs) have also demonstrated direct immunomodulatory roles during inflammation.^10,11 Owing to the relative sparsity of renal LECs compared with other cells, however, the heterogeneity and specific molecular roles of renal LECs in AKI have remained largely unexplored as the cells have been undercharacterized or excluded in past sequencing efforts.^12–14 Several studies have characterized the heterogeneity of LECs along the lymphatic network from the capillary tips to conducting vessels and highlight the fluid and solute transport and immunomodulatory roles of LECs in other tissues.^11,15

The general anatomy and physiology of the renal lymphatic vasculature have been elegantly reviewed and characterized previously.^16–18 The network develops from the renal hilus in a vasculogenic manner similar to the dermis.¹⁶ The adult renal lymphatic network in the mouse was recently elegantly visualized after tissue clearing and confirmed the appearance of valved collecting lymphatic vessels along the renal arteriole network and limited lymphatic capillaries throughout the kidney parenchyma.¹⁹ The molecular and genetic heterogeneity of renal LECs, however, is largely unknown.

After an AKI, lymphangiogenesis is initiated in response to inflammatory stimuli and the upregulation of lymphatic growth factors. Previous studies investigating whether renal lymphangiogenesis is beneficial to the recovery of kidney function and clearance of immune cells have remained inconclusive.²⁰ Some studies have indicated renal lymphangiogenesis as being a potential driver of kidney inflammation.²¹ However, other evidence suggests that lymphangiogenesis has a largely beneficial effect on the recovery of kidney function.^6,9,22 For example, our past work initiating renal lymphangiogenesis before injury altered the CD4:CD8 T-cell ratio and demonstrated improved long-term kidney functional and fibrotic outcomes.²² These findings indicate that renal lymphatics and LECs may respond dynamically to various injury stimuli and directly alter the postinjury immune response through lymphangiogenesis and changes in LEC biology.

In this study, we isolated murine renal LECs and used single-cell RNA (scRNA) sequencing to identify LEC vessel subpopulations in quiescence and determined distinct molecular responses for each subset in response to cisplatin-induced AKI. We analyzed global renal LEC-specific roles, evaluated their molecular identities and functions, and assessed how renal LEC functions alter gene and pathway expression after an AKI. Renal LEC composition was identified to have few lymphatic capillaries, and the network was strongly lymphangiogenic in response to AKI. The predominant immunomodulatory pathways in AKI were found to be in LEC T-cell interactions. These findings thus provide a novel characterization of the renal lymphatic vasculature and identify potentially how LECs affect the progression of AKI.

Methods

Animal Injury Models

All sequencing experiments were performed in male mice on a C57BL6/J mice background, aged 10–12 weeks, and assignment to injury was not randomized. For cisplatin AKI, a single injection of cisplatin (Vet Tech, 68001-283-27) or saline control was administered to mice via a single intraperitoneal injection at a dose of 20 mg/kg (n=5 mice per group). Bilateral ischemia–reperfusion injury (IRI) procedures were conducted as previously described by Texas A&M University Comparative Medicine Program veterinary staff.²² In brief, mice were anesthetized by ketamine/xylazine intraperitoneal injection, and both renal pedicels were revealed, and blood flow halted by atraumatic vascular clamps (Fine Science Tools, 18055-05) for 20 minutes. Mice were maintained at 37°C throughout the procedure, and renal ischemia was confirmed visually through loss of kidney color. Restoration of kidney color confirmed renal reperfusion before suturing. All mice were monitored daily to assess body weight loss, behavior, or major signs of distress. Mice were euthanized 72 hours after AKI (n=5 mice per group). Mice were perfused via the left ventricle with 5 ml of ice-cold HBSS (14025092, Gibco), after which, perfused kidneys were collected and placed on ice in HBSS. To prevent batch effects, all kidneys were collected and processed on the same day. Mice for increased renal lymphatic density flow cytometry studies have been previously described.^8,23 In brief, the kidney-specific overexpression of vascular endothelial growth factor D (VEGFD) mouse (KSP-rtTA xTRE-VEGFD) (KidVD) were maintained on a C57/Bl6J background and with littermates lacking transgenes necessary for VEGFD overexpression serving as controls. Male mice approximately 8–12 weeks of age were administered doxycycline hyclate (0.2 mg/ml; Sigma) for 3 weeks before the administration of cisplatin to induce kidney injury. Mice were euthanized 72 hours postkidney injury, consistent with sequencing studies. Mice were provided ad libitum access to water and standard chow throughout the study. All animal protocols and housing for the study were approved by the Institutional Animal Care and Use Committee at Texas A&M University.

Kidney Whole-Mount and iDISCO Clearing

For whole-mount imaging, kidneys from C57BL6/J mice were prepared by cardiac perfusion with warm saline, perfusion fixation with 37°C 1% paraformaldehyde, followed by postfixation overnight at 4°C in 3% paraformaldehyde. For whole-mount immunohistochemistry, tissues were incubated for 20 minutes in increasing percentage methanol washes (20, 40, 60, and 100) before permeabilization with a 1:4 (volume/volume) DMSO/methanol solution (Dent's). Tissue bleaching was accomplished with 20-minute incubations of ice-cold H₂O₂ (0.1%, 0.3%, and 1% in Dent's solution). Tissue rehydration was performed by 20-minute incubation in decreasing percentage ethanol (100%, 75%, 50%, and 25%), before blocking and permeabilization, as described.^24,25 Tissues were incubated with rabbit anti-mouse LYVE1 (1:500, 103-PA50AG, Relia Tech) for 5 days and secondary donkey anti-rabbit-Cy5 (1:500, Jackson 711-605-152). Tissues were cleared according to iDISCO protocol. Tissues were imaged by ultramicroscope II (LaVision BioTec) using ImspectorPro software (LaVision BioTec). The light sheet was generated by a laser (wavelengths 640 nm, Coherent Sapphire Laser, LaVision BioTec) focused using two cylindrical lenses. A binocular stereomicroscope (MXV10, Olympus) with an x2 objective (MVPLAPO, Olympus) was used at ×0.8 magnification. A dipping cap protective lens, including correction optics for MVPLAPO ×2 objective, was applied for working distances inferior or equal to 5.7 mm. A Pioneering in Cameras and Optoelectronics Edge SCMOS CCD camera (2560×2160 pixel size, LaVision BioTec) was used to capture images. The step size between each image was fixed at 6 μm. Image processing used Imaris x64 software (version 8.0.1, Bitplane). Z stack light sheet images were first converted to imaris file (.ims) using ImarisFileConverter, and 3D reconstruction was performed using the “volume rendering” function. To facilitate image processing, images were converted to 8-bit format. Optical slices were obtained using the “orthoslicer” tool, 3D pictures.

Cell Isolation and Tissue Collection

For each condition, five mice were used in single-cell enrichment and sequencing experiments, as determined by previous isolation and flow cytometry experiments. Kidney capsules were retained, and kidneys were minced into small pieces and placed into 5 ml digestion buffer, 15 mg/ml collagenase D (Roche Applied Science, 11088858001), 10 mg/ml collagenase type 4 (Gibco, 17-105-019), 2 mg/ml dispase II (Roche Applied Biosciences, 4942078001), BSA Fraction V (Fisher Scientific, BP1605-100), and HBSS (Thermo Fisher Scientific, 14025092). The samples were then incubated in a 37°C water bath with manual agitation every 10 minutes. Digestion was halted by adding 10 ml of ice-cold HBSS. Digested samples were filtered through both 70 and 40 µm filters. Endothelial cells (ECs) were enriched magnetically using a mouse CD31 biotinylated antibody (BD Pharmingen, 553371) and a cleavable Biotin Binder Kit (Invitrogen, 11533D). After the linker was cleaved, the isolated cells were subsequently enriched using an anti-mouse podoplanin (Pdpn) biotinylated antibody (BAF3244, R&D Systems). CD31⁺Pdpn⁺ cells were placed into ice-cold RPMI 1640+10% FBS before flow cytometry confirmation of CD45⁻CD31⁺Pdpn⁺ cells. Cell fractions were resuspended in RPMI1640+10% FBS, counted, and diluted to a concentration of approximately 1000 viable cells/μl.

scRNA Sequencing Bioinformatics Pipeline and Analysis

Library Generation and Quality Control

CD31⁺Pdpn⁺ cell count and viability were confirmed on the Thermo Fisher Countess and the Moxi Go II (Orflo, MXG102). Samples were then loaded onto the Chromium X Controller using the Chromium Next GEM Single-Cell 3′ Reagent Kit V3.1. Sample cDNA quality was checked using the Agilent High-Sensitivity D5000 ScreenTape assay on the Agilent 4200 TapeStation System (Agilent Technologies, G2991BA). cDNA samples were then quantified and normalized before the library sequencing. Libraries were sequenced on a NovaSeq S4 Flow cell according to 10x Genomics recommendations. Single-cell RNA sequencing reads were processed using 10x Genomics cloud analysis. In brief, fastq files were uploaded to the 10x Genomics Analysis cloud and aligned to the reference mouse transcriptome (mm10) using Cell Ranger Count v6.0.1. A digital gene expression matrix, including raw unique molecular identifier counts, was assigned to each cell in the respective sample.

Bioinformatics Data Processing and in Silico Selection

Gene expression matrices were generated using CellRanger software (10x Genomics), and raw data reads were processed in R (version 4.2.1). Cells then underwent quality control steps before downstream analyses. No differences in cell cycle scores were observed (Supplemental Figure 1C).²⁶ Cells with fewer than 200 detected genes or more than 4000 genes or more than 20% mitochondria reads were excluded (Supplemental Figure 1D).²⁶ Count data were normalized using the NormalizedData function in the Seurat package (version 4.0), and highly variable genes were identified with FindVariableFeatures function. Comparable populations were identified in scRNA sequencing and anchored in both control and injury conditions. The ScaleData function was used to scale expression of genes to provide equivalent weight to all genes. Data were then summarized using principal component analysis in the Seurat package followed by visualization with uniform manifold projection. Cells were clustered according to their gene expression profile using the FindClusters function in Seurat with a canonical correlation analysis (CCA) clustering resolution of 1. CCA resolution of 1 was set after evaluation of principal component analysis elbow plots, uniform manifold projection assessments of cell quality effects, and presence of distinct top gene markers present for each population. Clusters were identified by the expression of known vascular EC marker genes Emcn, Cd34, and Kdr; hybrid vascular markers Emcn, Flt4, and lack of Pdpn; lymphatic marker genes Pdpn, Prox1, Flt4, and Sox18; and immune cell gene Ptprc. After identification of cluster identities and initial global analysis, all nonlymphatic cell clusters were removed from further downstream analyses. Gene ontology (GO) and molecular complex detection analyses were processed and visualized using the Metascape tool.²⁷ Pathway enrichment analyses used the GO, WikiPathway, and Reactome databases.

Subcluster Analysis and Identity Assignment

Control and cisplatin-injured LEC clusters were subclustered using Seurat's subset function. A new Seurat object was generated and processed as described above in bioinformatics data processing. Four new clusters were identified using the FindClusters function with a CCA resolution of 0.25. New LEC subcluster differentially expressed genes were determined using the FindAllMarkers function with an average log-fold change of ≥0.2. New LEC subcluster identities were assigned using previously published lymphatic data set genes that correspond to vessel regional identity.^15,26,28,29 For injury comparisons, a LEC subset was generated as described above using the subset function. Control and cisplatin injury LEC subset genes were compared using the FindMarkers function, and top significantly upregulated and downregulated between conditions were identified. GO analyses were generated as described above. Violin plots of significantly altered immune genes were generated using the VlnPlot function in the Seurat package.

In Vitro Human Dermal Lymphatic EC Cisplatin Treatment

Human dermal lymphatic ECs (HDLECs) were purchased from PromoCell (C-122217) and maintained in complete endothelial basal media (C-22022, PromoCell) at 37°C, 5% CO₂ conditions, as previously described.³⁰ Confluent HDLECs were then treated with either saline control, 4 µg/ml, or 8 µg/ml cisplatin for 24 hours according to previously described studies.^31–33 A dose of 12 µg/ml cisplatin was lethal to HDLECs. Cells were then isolated at the end of the 24-hour treatment period, and total RNA was extracted using the PureLinkRNA Mini Kit (12183018A, Invitrogen).

In Vivo LEC RNA Isolation for Validation

LEC enrichment was performed as above, and the RNA from the LEC-enriched EC isolation was isolated from mouse kidneys of control, cisplatin AKI, and IRI conditions using the Direct‐zol RNA Miniprep Plus kit (R2072, Zymo) following the manufacturer's instruction. LEC enrichment was used to validate isolation of LECs via flow cytometry and with quantitative PCR (qPCR) of known lymphatic markers. Healthy human LEC and AKI are reported as log2FC comparing between conditions (Supplemental Table 7).

Human Kidney LEC Genes

Data for identifying a population of human renal LECs were previously published and used for comparison with renal LEC injury response genes for this study.³⁴ The data are publicly available through the Kidney Precision Medicine Project Kidney Tissue Atlas.

Quantitative Reverse Transcription PCR

cDNA was synthesized from 0.5 μg total RNA for all in vivo and in vitro experiments using the iScript cDNA kit (1708891, Bio-Rad). Five microliter qPCR reactions were performed using iTaq Universal SYBR Green Supermix (1725124, Bio-Rad). Reactions were run in duplicate on a QuantStudio6 Flex Real-Time PCR system (Applied Biosystems). Fold changes compared with control samples were calculated using the 2^ΔΔCT method with Ubc and ACTB as endogenous control for the in vivo and in vitro samples, respectively (Supplemental Table 1). All in vivo data from the CD31⁺Pdpn⁺ cells were normalized by a lymphatic EC-specific marker Flt4 (vascular endothelial growth factor receptor-3 [VEGFR-3]). 2^ΔΔCT values of all the genes were divided by the 2^ΔΔCT value of Flt4.

Immunofluorescence

For tissue immunofluorescence analyses, kidneys were formalin fixed for 24 hours and, following a sucrose gradient, embedded in optimal cutting temperature compound (T4583, Tissue-Tek) for sectioning. Ten micromole kidney sections were incubated overnight at 4°C with selected primary antibodies (Supplemental Table 2). Sections were then washed with PBS and then blocked in 10% donkey serum for 30 minutes. Sections were then incubated with secondary antibodies. Slides were mounted with Prolong Gold Antifade with 4',6-diamidino-2-phenylindole (OB010020, SouthernBiotech) and visualized with an Olympus BX51 fluorescent microscope with an Olympus Q5 camera. Representative images were captured at figure-indicated magnification.

Flow Cytometry

The whole left kidney from injured KidVD mice was digested using the Miltenyi Multi Tissue Dissociation Kit 2 on a gentleMacs Octa Dissociator (Miltenyi Biotec). Digested tissue was filtered through sterile 100 and 40 μm strainers to obtain a single-cell suspension. A 0.1% BSA with 2 mM EDTA solution was used for cell resuspension. Renal immune cells were isolated using a magnetic pull down of CD45⁺ cells according to Miltenyi Microbead kit instructions (130-052-301, Miltenyi Biotec). After CD45⁺ isolation, cells were washed with PBS and then incubated with live dead blue dye. Nonspecific Fc binding was blocked with an anti-mouse CD16/CD32 antibody (BD Pharmingen, San Jose, CA) on ice for 10 minutes, and cells were then incubated with fluorescent conjugated antibodies (Supplemental Table 2). Cells were stained for intracellular markers according to the manufacturer's instruction (00-5523-00, Invitrogen). Data were acquired on a Cytek Aurora 5-Laser Spectral Cytometer. Cells are reported as percent of CD45⁺ or percent of parent population.

Data and Code Availability

The single-cell RNA sequencing data generated in this study are deposited in Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/) with accession number (GSE236199). No unique code was generated from this study.

Statistical Analysis

Statistical analysis of qPCR data was performed using GraphPad Prism software (version 9.5.1). An unpaired t test with unequal variance (Welch) was used, and P value < 0.05 was considered significant.

Results

Renal Lymphatic Endothelial Cell Isolation and scRNA Sequencing Identification of Lymphatic Endothelial Cell Populations

Lymphatic ECs are sparse compared with other cells in the kidney and often missing in past scRNA sequencing studies.^13,14 Renal lymphatics follow larger blood vessels and have sparse capillary networks (Figure 1A). To identify the heterogeneity of LECs in the kidney, we collected both kidneys from five mice, enriched LECs, and performed scRNA sequencing (Figure 1B). Enrichment of renal LEC identity was confirmed before scRNA sequencing library preparation by flow cytometry and quantitative reverse transcription PCR of isolated cells (Supplemental Figure 1A). On the basis of the isolation protocol used, approximately 99% of cells were CD45⁻ via flow cytometry. Of these, approximately 60% highly expressed CD31 and Pdpn, with nearly all of these cells also demonstrating LYVE-1 on their surface (Figure 1C). Sequenced samples were approximately 60% LECs by classically described surface expression. Unsupervised clustering analysis of the total cell pool assigned seven unique cell clusters to the anchored data (Figure 1D). After quality control filtering and exclusion of contaminating cells, we obtained a total of 2323 cells for analysis. To discriminate LEC populations from other CD31⁺PDPN⁺ cells sequenced, populations were identified by cells expressing the following markers: vascular genes Kdr, Emcn, and Cd34 and LEC genes Prox1, Flt4, and Sox18 (Figure 1E and Supplemental Figure 1B). A population of immune cells were identified (Cd52, Ptprc) but excluded from further analyses (Supplemental Figure 1B). As expected, scRNA sequencing populations were a mix of both LECs and other vascular phenotype-like cells. It is noted that previous studies have indicated the presence of other Pdpn⁺ expressing cells in the medullary region, such as urothelial cells and ascending and descending loop epithelium, in the kidney.^17,35–37 Three LEC populations (Prox1, Flt4, and Sox18) were identified in addition to three vascular populations (Kdr, Emcn, and Cd34) and one immune population (Ptprc and Cd52). LEC1 and LEC2 genetic populations were similar, while LEC3 was genetically more like the “hybrid endothelium” attributed populations.

Figure 1.

Isolation and identification of renal LECs. (A) Whole-mount image of murine uninjured renal lymphatic vasculature labeled by LYVE1 (green). (B) Diagrammatic outline of LEC isolation method and follow on bioinformatics analysis. (C) Flow cytometric verification of renal LEC populations captured with magnetic enrichment as verified by gating of CD45⁻ live cells, PDPN⁺CD31⁺, and LYVE1⁺PDPN⁺. (D) Anchored clustering analysis of renal LEC single-cell datasets identifies seven cell populations through CCA unsupervised clustering. (E) Lymphatic identity of cell populations verified through the expression of lymphatic genes. Light sheet images collected at 0.8× magnification and 6 μm with Z stack. Scale bar=500 µM. CCA, canonical correlation analysis; LEC, lymphatic endothelial cell.

Renal Lymphatic Endothelial Cell Heterogeneity Is Similar to Lymphatic Structural Components from Other Tissues

To define each cluster globally, we assessed the top five differentially expressed genes that identified each unique cluster in the full cell population (Figure 2, A and B, and Supplemental Table 3). We identified that LEC1 and LEC2 populations were distinct in quiescent conditions and tested whether these populations could be assigned to established lymphatic vessel components on the basis of differentially expressed genes. LEC1 was characterized by increased expression of genes Fabp4, Plvap, Rsad2, and Kdr, while LEC2 was defined by Eln, Plxnb2, and Fbln5 (Figure 2A). Genetic expression of Fabp4,^26,38 Plvap,³⁸ and Kdr³⁹ has been well annotated in lymphatic vessels. LEC2 genes Flbn5 and Eln have been described in cases of lymphatic hypoxia,⁴⁰ and Plxnb2 is a receptor for SEMA4C on LECs.⁴¹ In differentiating these LEC populations, there was no discernible association to specific lymphatic vessel anatomical features (e.g., lymphatic valves) or developmental lineage (e.g., sprouting versus lymphangioblasts) from global analysis. LEC3 cells in quiescent conditions were sufficiently different and small in population number such that they were not analyzed further as LECs are compared in injury conditions (Supplemental Table 3).

Figure 2.

Genetic signatures of renal LECs. (A) In control conditions, LEC populations 1 and 2 comprised a bulk of the population and isolated as a subset for LEC-specific analyses. In control conditions, 458 total LECs were used in downstream analyses. Scaled expression level heatmap of top five genes from the identified cell clusters. (B) UMAP of control isolated cell populations and selected LEC subset (composed of LEC 1 and LEC 2) used for subsequent downstream analysis. (C) LEC subset UMAP cluster identifying four distinct LEC populations. (D) Expression level–scaled heatmap of top five marker genes in LEC subcluster phenotypes. (E) LEC subclusters differentiate into four distinct anatomical locations. Previously published gene markers used to identify distinct lymphatic vessel subpopulation LECs. (F) Diagrammatic depiction of the anatomical location of captured LEC populations. UMAP, uniform manifold projection.

To determine whether the LECs were associated with a lymphatic vessel structural feature and assign their identities as such, we performed subset analysis on LEC1 and LEC2 (Figure 2B). LEC subset analysis revealed four subclusters and new defining differentially expressed genes (Figure 2C and Supplemental Table 4). We used previously published lymphatic markers to identify the following features: collecting LECs (cLECs) Igfbp5,⁴² Fabp4²⁶, and Plvap³⁸; valve LECs (vLECs) Calm1,⁴² Pdgfb²⁶, and Neat1⁴²; and precollecting LECs (pLECs) Cldn5²⁶, Gja4²⁶, and Bmx.^26,43 cLECs expressed genes as defined above, and vLECs split into two distinct populations. vLEC1 was defined by listed vLEC markers, while vLEC2 shared the expression of vLEC markers in addition to the expression of pLEC and capillary LEC markers Stmn2, Tppp3, and low expression of Ccl21a^26,44 (Figure 2, D–F). Interestingly, we did not isolate or identify a distinct lymphatic capillary population; however, past studies examining murine lymphatic anatomy, and our own imaging (Figure 1A), have found sparse lymphatic capillary structures within the kidney.^16,19,45,46

Renal Lymphatic Endothelial Cell Gene Expression Changes in Response to Cisplatin AKI

In AKI, whether the lymphatic system and LECs hold a beneficial or detrimental role in resolution or progression after injury is not yet clear.²⁰ To characterize renal LEC transcriptional adaptation to AKI, mice were injected with a single high dose of cisplatin and renal LECs were isolated 72 hours postinjury. After injury, there was an increase of total number of Pdpn⁺ expressing cells in the kidney. This was not due to lymphatic expansion (appreciable lymphangiogenesis is only significant around 7 days after cisplatin injury⁹). The populations of LEC1 and LEC2 were reduced in the total captured cells compared with control conditions (62%–7%) (Figure 3A). The number of “hybrid” cells increased relative to the total number of captured cells (35%–56%). De novo Pdpn expression was found in the inner medulla (Supplemental Figure 2A). Pdpn expression on injury has been previously reported in medullary collecting duct cells: Pdpn expression in the inner medulla was negative for potential vasa recta or endothelial markers PLVAP and SLC14a1 by immunofluorescence confirming their lack of “LEC-ness” (Supplemental Figure 2, B and C). An increase in isolated immune cells (7%–37%) was also noted in AKI (Figure 3A).

Figure 3.

LEC injury response demonstrates altered molecular roles. (A) UMAP split of control and cisplatin injury conditions demonstrated a shift toward increased podoplanin-expressing vessels and a distinct shift in LEC 1 and LEC 2 populations. (B) Heatmap of 283 most altered LEC subset genes between control and injury conditions. Fold change indicates increased (red) or reduced (blue) expression for indicated gene. Adjusted P < 0.05. (C) Pie chart of top 30 upregulated injury genes when compared with control conditions indicates that top changed genes are associated with biologic processes, such as increased angiogenic signaling, lymphatic vessel development, and increased inflammation-related processes. Graphic color corresponds to P value (indicated in legend), and size of graphic sections corresponds to the number of genes.

To understand how LECs respond to AKI, we compared LEC subset gene expression differences between injured and quiescent conditions (Supplemental Table 5). Changes in LEC 1 and LEC 2 log₂FC were examined briefly before LEC subset comparison between quiescent and injury conditions (Supplemental Figure 3, A–C). GO analyses identified the top changed injury genes when compared with quiescent conditions as Fabp4 (log₂FC 4.89), Kdr (log₂FC 3.24), and Plpp3 (log₂FC 3.19) (Figure 3B). These genes are highly involved in lymphatic vessel development, vascular endothelial growth factor (VEGF) receptor signaling, and immunoregulatory interactions, roles identified in the injury GO analyses (Supplemental Figure 4, A and C). Downregulated processes included oxidative stress and redox pathways and dicarboxylic acid transport, key pathways in metabolism that are altered with cisplatin injury^47,48 (Supplemental Figure 4C). Additionally the genes Cd300lg (log₂FC 2.37) and Tspo (log₂FC 2.11) were identified and have the respective previously demonstrated roles: lymphocyte homing and adhesion to ECs⁴⁹ and immune cell processes.⁵⁰ In a previous study, Tspo had an identified role in apoptosis prevention in renal IRI models (Figure 3B and Supplemental Table 5).⁵¹

To understand how highly changed genes corresponded to specific biologic processes, the top 30 upregulated genes in injury conditions, when compared with quiescence, were grouped and characterized according to GO biologic process. Most genes were associated with VEGF receptor signaling pathways, VEGF stimulus, and lymphatic vessel development: all priming LECs undergoing lymphangiogenesis. However, roughly a quarter of the genes were associated with either T-cell differentiation (Ctla2a, Id1, Cd38, Rsad2) or in the regulation of the inflammatory response (Plpp3, Ifitm3, Bst2) (Figure 3C). LECs have previously demonstrated T-cell–related immunoregulatory roles in other disease models, such as cancer, arthritis, and cardiac injury.^25,52–54 These data lend additional support to the finding that renal LECs are undergoing proliferation, upregulating their interaction with immune cells, and alter their molecular repertoire on kidney injury.

Renal Lymphatic Endothelial Cell Adaption to Injury Stimulus and Anatomical Composition

To assess LEC involvement with the immunologic transition between innate and adaptive immune processes, genes related to innate immune response processes were selected and plotted to compare expression between control and injury conditions. Genes related to integrin signaling and the innate immune response, such as Ltbp4 and Kitl, were decreased in LECs after injury along with Plscr2, a mast cell activating gene, and Foxp1, a negative regulator of T follicular helper cell differentiation. However, genes Spp1 and Igfbp7 were present almost exclusively in injury. The transcript of Spp1, osteopontin, acts as a key renal cytokine, while the transcript of Igfbp7, insulin-like growth factor-binding protein 7, has been demonstrated to have roles in stimulating cell adhesion, blocking angiogenic signaling, and changing cell sensitivity to chemotherapeutic drugs^55,56 (Figure 4A).

Figure 4.

Renal LEC adaptation to injury stimulus and anatomical composition. (A) Violin plots of selected immune process–related genes altered between control and injury conditions. Genes selected represent changed genes between conditions which are additional related to innate and adaptive signaling processes. (B) UMAP of injured LEC subset demonstrating similar anatomical populations to control conditions. (C) Expression level–scaled heatmap of top five marker genes in LEC subcluster phenotypes; LECs differentiate into four distinct anatomical locations. (D) In injury conditions, lymphatic vessel anatomical identity GO analyses identify changes in defined genetic roles. GO, gene ontology.

Among the anatomical LEC subpopulations, cLEC and vLEC1 subpopulations do not have a response to injury that is specifically unique to their population (Figure 4, B and C). Notably, pLEC retains some defining genes from the control conditions (Mgp, Clu), unlike the cLEC, vLEC1, and vLEC2 populations. GO analysis was performed on LEC subpopulations to determine how injury alters the molecular functions of renal LEC anatomical locations. By gene pathway analysis, cLECs are involved in cell migration and transport processes and vLEC populations are involved in nuclear organization, apoptotic processes, and angiogenic signaling. In injury, pLECs are more highly involved with developmental processes and metabolism (Figure 4D). Overall, these data indicate that LECs maintain similar anatomical composition to quiescent LECs but alter their vasculogenic signaling and metabolism.

Validation of Renal Lymphatic Endothelial Cell Gene Response to Kidney Injury

To validate LEC-specific gene changes in injury, we used immunofluorescent labeling and qPCR on renal LECs. We identified Spp1, one of the highly variable genes in injury conditions and associated with the LEC injury response to colocalize with Pdpn⁺ lymphatic vessels exclusively in injury conditions (Figure 5A). We did not detect immunofluorescent colocalization of Spp1 in uninjured kidneys. We further confirmed colocalization of a second gene target in injured kidneys, Fabp4, which was one of the most changed in the injury LEC subset (Figure 5B). We did not note colocalization of either Spp1 or Fabp4 with Pdpn⁺ glomeruli, indicating that their Pdpn⁺ colocalized upregulation was lymphatic specific. As immunofluorescence options were limited, fresh LEC-enriched cell isolations were tested by qPCR in cisplatin and IRI-injured LECs. Tested genes were selected based on both significance in injury and the largest log₂FC between conditions. Gene expression between conditions was compared in a heatmap to determine the most altered genes between injury models (Figure 5C). Genes Ptprb, Mgp, Gng11, and Vim demonstrated a significant increase in relative expression validating scRNA sequencing findings (Figure 5D). In addition, while not significant, Kdr demonstrated a trend toward increased gene expression in injury by qPCR in the LEC-enriched cells. In IRI conditions, Ptprb was also significantly increased, but Vim and Fabp4 were significantly decreased, indicating an injury-specific effect (Figure 5E). These data indicate that the scRNA sequencing findings are reproducible and detectible in LECs.

Figure 5.

Validation of lymphatic genetic response to kidney injury. (A) Immunofluorescent labeling of Spp1 (red) and Pdpn (green) demonstrating increased Spp1 on cisplatin injury and colocalization with Pdpn⁺ lymphatic vessel (arrow). (B) Immunofluorescent staining of Fabp4 (red) and Ppdn (green) with colocalization in injury (arrow). Images collected at 40× on 10 µM sections. Scale bar=20 µM. (C) Heatmap of log2FC values of isolated renal LECs for highly changed genes as determined by scRNA sequencing data in cisplatin and IRI. Genes are normalized to respective condition controls. (D) qPCR relative expression values in cisplatin (Cis) isolated renal LECs confirm significant gene expression changes with injury compared with controls (Con) for Ptprb, Mgp, Gng11, and Vim as identified in scRNA sequencing results. Kdr (P = 0.0575) and both Slc34a1 and Fabp4 were not detected as significantly different. (E) qPCR relative expression values in renal LECs isolated from IRI confirm similar significant gene expression changes with injury for Ptprb and similar trends to cisplatin LECs for Slc34a1 and Kdr. In IRI, Fabp4 had significantly decreased relative expression when compared with control conditions. *P < 0.05, **P < 0.005, ****P < 0.0001. IRI, ischemia–reperfusion injury; qPCR, quantitative PCR; scRNA, single-cell RNA.

To determine whether detected gene expression genes were due to a direct cisplatin interaction with LECs, HDLECs were challenged with cisplatin in vitro. A heatmap of all tested genes normalized to saline controls indicated that genes (Ptprb, Vim, Kdr, and Fabp4) were significantly increased with 8 µg/ml cisplatin (Figure 6, A and B, Supplemental Table 6). However, Slc34a1 and Spp1 were significantly decreased with cisplatin treatment. These data indicate that cisplatin does not directly drive the in vivo–identified LEC genetic changes.

Figure 6.

Gene expression of cisplatin-treated HDLECs in vitro and human AKI. (A) Heatmap of log2FC values of HDLECs highly changed genes as determined by scRNA sequencing data. HDLECs genes compared in cultures treated for 24 hours with 4 µg/ml or 8 µg/ml cisplatin (Cis). Genes are normalized to saline-treated controls. (B) Relative expression values of PTPRB, VIM, KDR, FABP4, and RSAD2 in HDLECs confirming scRNA sequencing gene expression. (C) Heatmap of log2FC values of LEC genes in healthy human kidney and AKI single-cell dataset. *P < 0.05, **P < 0.005, ***P < 0.0005. HDLEC, human dermal lymphatic endothelial cell.

To assess the human AKI clinical relevance of our dataset, we used a recently published dataset that identified a subset of human renal LECs to examine gene expression.³⁴ Although the dataset had sparse LEC populations (healthy=12 cells and AKI=45), many of the genes trended in a similar manner to the dataset reported here. We found that FABP4 had the highest log2FC expression in LECs compared against all other clusters in both healthy and AKI human samples. Other LEC genes (PTPRB, CD9, KDR, and BMX) trended in similar directions to sequencing results (Figure 6C). These data indicate that genetic findings in murine LECs are potentially translatable to human clinical samples.

Expanded Renal Lymphatic Density before Injury Alters T Regulatory Cell Populations

To assess whether genetically induced renal lymphatic networks captured lymphatic capillary networks, we used KidVD⁺ mice and performed image and single-cell analyses. Immunofluorescent labeling of LYVE-1 in a whole-mount cleared KidVD⁺ kidney demonstrates increased lymphatic networks primarily in the kidney cortex (Figure 7A) compared with control kidneys (Figure 1A). KidVD⁺ isolated LECs were integrated and anchored to control and cisplatin-injured datasets and demonstrated an increased percentage of cLECs (58% compared to control 22% and injury 20%) and pLECs (4% compared with control 3% and injury 2%) (Figure 7B). Capillary LEC genes (Stmn2, Tppp3, Ccl21a, Reln, Ptx)^26,44 were detected in vLEC2 and pLEC but did not resolve into distinct capillary population (Figure 7C).

Figure 7.

Increased renal lymphatic density increases intrarenal T regulatory cells in cisplatin injury. (A) Whole-mount kidney image labeled with LYVE1 (green) of KidVD⁺ mouse demonstrating expanded renal lymphatic network. (B) LEC subset clustering from isolated KidVD⁺ mice demonstrating four distinct populations. (C) Dot plot of previously published LEC genes confirming cluster population assigned lymphatic anatomical locations. (D) Example gating CD45⁺ cells from single live cells. (E) Total CD45⁺ cells in control and KidVD⁺ kidneys. (F) Percent of CD8⁺ T cells and (G) CD19⁺ B cells. (H) Example flow cytometry gating of CD4⁺ population. (I) Percent of CD4⁺ T cells, (J) percent of CD4⁺CD25⁺ T cells, and (K) percent of T regulatory cells (Tregs, CD4⁺CD25⁺FoxP3⁺). Arrows indicate lymphatic network expansion in the cortex. Light sheet images collected at 0.8× magnification and 6 μm with Z stack. Scale bar=300 µM. *P < 0.05; **P < 0.005.

To determine whether expanded renal lymphatic density before injury onset altered immune populations, KidVD mice were injured with cisplatin and assessed via flow cytometry. There was no significant difference between KidVD⁺ and control mice for total CD45⁺ cells, percent CD8⁺, or percent CD19⁺ immune cells (Figure 7, D–G). KidVD⁺ mice did demonstrate increased intrarenal CD4⁺ T cells with no difference in CD4⁺CD25⁺ subsets and a significantly increased population of T regulatory (Tregs, CD4⁺CD25⁺FoxP3⁺) cells (Figure 7, H–K). Interestingly, these findings support the GO analyses of injured LECs indicating LECs regulate regulatory T-cell differentiation (Figure 3C).

Discussion

Several physiologic and pathophysiologic roles have been described for lymphatic vessels in the kidney, but renal LECs and their response to injury remain largely unstudied, in part due to their relative scarcity compared with other renal cells. Our study applied scRNA sequencing technology to characterize the renal LECs in quiescence and injury and has identified distinct LEC populations, potential molecular roles, and changes in genetic expression with kidney injury, uncovering previously undefined roles for lymphatic vessels in the kidney injury response. We have also confirmed, as indicated in previous studies, that kidney injury tubules increase Pdpn expression.⁵⁷ Our data further confirm that the LEC genetic response is likely injury-context dependent, supported by our genetic confirmation in cisplatin and IRI injury models.

In most tissues, lymphatic vessels begin as blind-ended capillaries that coalesce into precollector vessels, possess unidirectional valves, and evolve further into larger conducting vessels. Several studies have characterized the heterogeneity of LECs along this network from the capillary tips to conducting vessels and highlight the transport and immunomodulatory roles of LECs.^15,26 Past studies examining the anatomy of lymphatics in the kidney have identified a network of lymphatics along the interlobular and corticomedullary vasculature with some vessels noted in the murine cortex and subcapsular space; recent light sheet microscopy of adult mouse kidneys beautifully demonstrated renal lymphatic anatomy.^17–19 Using gene expression from past LEC sequencing efforts, we were able to confirm populations of collecting vessel LECs, pLECs, and LECs from lymphatic valves. Interestingly, indications of capillary LECs were few but conform to the past anatomical descriptions of the network. These findings corresponded with our images demonstrating few lymphatic capillary networks but extensive cortical and hilar lymphatic vessels.

The growth of new lymphatic vessels, lymphangiogenesis, occurs postinjury as an often-necessary step to resolve inflammation and restore tissue homeostasis.^58–60 Several recent elegant studies have characterized lymphatics in the developing kidney.^61,62 The importance of renal lymphatic vessels in pathologic progression in the adult kidney, however, and how manipulation of lymphangiogenic signaling through VEGFR-3 affects functional outcomes is less clear. What is clear is that the lymphatic vasculature and VEGFR-3 signaling is involved in transplant rejection, renal fibrotic remodeling, immune accumulation, and microvascular rarefaction.^21,63–65 After AKI insult, the expression of the lymphatic growth factors VEGF-C and VEGF-D is increased within the kidney, and most models (and clinical samples) demonstrate increased lymphangiogenesis at some point postinjury; however, the timing and extent of injury may likely regulate the importance of endogenous postinjury LEC responses and lymphatic impact.^6,9,66,67 Here, we also identify increased Ptprb, implicated in vasculogenesis, and increased Fabp4, fatty acid oxidation, may also aid to drive lymphangiogenesis and is required for differentiation of LECs from precursor cell subset.⁶⁸ It is thus not surprising that a series of vasculogenic pathways were some of the most upregulated in LECs 3 days after injury in our data.

Lymphatic vessels are increasingly recognized as active regulators of the immune response.^69–71 For example, in cancer models, LECs can upregulate PD-L1, present antigens through major histocompatibility complex expression, and affect tissue cytotoxic CD8⁺ T-cell numbers homeostasis.^70,72 Studies in other disease models have demonstrated that LECs have a molecular repertoire that specifically alters immune cell activation, differentiation, trafficking, chemokine receptor expression, and antigen presentation.^10,70,73 Past work from our laboratory and others has demonstrated that manipulating renal lymphatic density alters immune cell populations postinjury.^8,9,22,65 We therefore hypothesized that renal LECs alter their molecular roles in response to injury. Here, we have confirmed renal LECs transcriptionally adapt to cisplatin and IRI, uncovering an underappreciated role of renal lymphatics as an active regulator of the response to injury stimuli. We find that renal LECs on injury stimuli have increased major histocompatibility complex gene expression (B2m, H2-D1) involved in biologic processes related to T-cell differentiation (Ctla2a, Id1, Cd38, Rsad2) and antigen presentation/cytokine signaling (Lyc61, Foxp1, Plscr2). We demonstrate that renal LEC genetic changes in AKI are translatable to human clinical samples and an increased murine renal lymphatic network directly altered T regulatory cell presence. These findings suggest that lymphatic vessels actively prime the immune environment and that increases in renal lymphatic density may support injury resolution by educating the adaptive immune response. Whether renal LECs maintain the same molecular functions and immunoregulatory roles determined in this study across other kidney injury models remains an open question, but at least some of the repertoire was present in LEC-enriched cells from IRI kidneys. We propose that renal LECs are potential regulators of inflammatory resolution postkidney injury, and importantly, studies targeting LEC genes associated with immune cell interactions should be conducted to determine postinjury outcomes.

There are several potential limitations for the study. First, we used only male mice for renal LEC enrichment studies. It is likely gender plays a role in LEC-specific responses as previous studies have identified female mice of C57Bl6/J background to have a degree of protection from AKI, likely an estrogen-mediated effect.^74,75 Second, owing to the lack of renal LEC databases, we identified and characterized clusters with the understanding that they are CD31⁺Pdpn⁺ and used Sox18, Prox1, and Flt4 as additional identification markers. To our knowledge, there is currently no database or resource to specifically catalog genes associated with renal LECs, although some studies have identified small LEC populations.^12,13 Our study uses the LEC genetic signatures currently available; however, with the increasing capabilities of next-generation sequencing, new lymphatic populations may be identified. Specifically, once de novo lymphatic vessels are formed postinjury (after day 7 according to other cisplatin studies), they would likely be of the capillary type, which has previously demonstrated a strong immunoregulatory pathway expression.¹⁵ Finally, the cisplatin injury model introduces systemic and local effects that may contribute to the overall finding of the scRNA sequencing dataset. We mitigate this shortcoming by assessing gene expression of isolated renal LECs in an IRI model and confirming cisplatin-specific effects with HDLEC studies. Future studies would focus on applying scRNA sequencing in various injury models to determine whether LECs alter response to injury dependent on the injury stimulus.

Lymphatic vessels and the LECs that make them up provide for an intriguing target to manipulate tissue pathologic responses. The immunologic mechanisms by which AKI progresses to CKD may be tied to LEC-immune interactions. The renal network of lymphatic vessels and LECs may provide a future target to change the inflammatory progression of AKI.

Supplementary Material

jasn-35-549-s001.pdf ^{(1.5MB, pdf)}

jasn-35-549-s002.xlsx ^{(148.1KB, xlsx)}

Disclosures

E. Brakenhielm reports advisory or leadership role (only academic commission work) for German research council (DGF), Germany, Inserm CSS3 commission, France, and Vetenskapsradet (VR), Sweden, and other interests or relationships with French Alliance du Coeur, Healthy Heart Fund, and KU Leuven, Belgium. B.M. Mitchell reports employment with Texas A&M University, research funding from NIH, and patents or royalties from Texas A&M University. B.L. Tate reports employment with ProScribe. All remaining authors have nothing to disclose.

Funding

J.M. Rutkowski: National Institute of Diabetes and Digestive and Kidney Diseases (R01 DK119497). B.M. Mitchell: National Institute of Diabetes and Digestive and Kidney Diseases (R01 DK120493). H.A. Creed: National Institute of Diabetes and Digestive and Kidney Diseases (F31 DK132838) and American Heart Association (AHA 916334). S. Chakraborty: Cancer Prevention and Research Institute of Texas (RP210213).

Author Contributions

Conceptualization: Priyanka Banerjee, Sanjukta Chakraborty, Heidi A. Creed, Saranya Kannan, Brett M. Mitchell, Joseph M. Rutkowski.

Data curation: Heidi A. Creed.

Formal analysis: Heidi A. Creed, Joseph M. Rutkowski.

Funding acquisition: Sanjukta Chakraborty, Heidi A. Creed, Brett M. Mitchell, Joseph M. Rutkowski.

Investigation: Priyanka Banerjee, Sanjukta Chakraborty, Heidi A. Creed, Saranya Kannan, Brittany L. Tate.

Methodology: Priyanka Banerjee, Heidi A. Creed, David Godefroy, Saranya Kannan, Brittany L. Tate.

Project administration: Joseph M. Rutkowski.

Resources: Sanjukta Chakraborty, Joseph M. Rutkowski.

Supervision: Ebba Brakenhielm, Joseph M. Rutkowski.

Visualization: Ebba Brakenhielm, Heidi A. Creed, David Godefroy.

Writing – original draft: Heidi A. Creed, Saranya Kannan, Joseph M Rutkowski.

Writing – review & editing: Heidi A. Creed, Joseph M. Rutkowski.

Data Sharing Statement

(1) scRNA sequencing data for all samples utilized in the study were deposited in the Gene Expression Omnibus (GSE236199). https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE236199, (2) sfcnuqaihngffud. (1) https://doi.org/10.1038/s41586-023-05769-3, (2) Data from published study available at KPMP. https://atlas.kpmp.org/

Supplemental Material

This article contains the following supplemental material online at http://links.lww.com/JSN/E585 and http://links.lww.com/JSN/E586.

Supplemental Table 1. qPCR Primers.

Supplemental Table 2. Antibodies used.

Supplemental Table 3. Control conditions global genes.

Supplemental Table 4. Control LEC subcluster differentially expressed genes.

Supplemental Table 5. Control versus injury differentially expressed genes.

Supplemental Table 6. qPCR data values with averages.

Supplemental Table 7. Human healthy and AKI gene mean expression and log2FC.

Supplemental Figure 1. QC and identification markers for scRNA sequencing clusters.

Supplemental Figure 2. Podoplanin labeling of other renal cells.

Supplemental Figure 3. LEC population global genetic response in quiescent and injury conditions.

Supplemental Figure 4. Renal LEC GO biologic pathways in injury.