Pro-fibrotic VEGFR3-dependent lymphatic vessel growth in autoimmune valvular carditis

Abstract

Background:

Rheumatic heart disease (RHD) is the major cause of valvular heart disease in developing nations. Endothelial cells (ECs) are considered crucial contributors to RHD, but greater insight into their roles in disease progression is needed.

Methods:

We used a Cdh5-driven EC lineage tracing approach to identify and track ECs in the K/B.g7 model of autoimmune valvular carditis. Single cell RNA sequencing (scSeq) was used to characterize the EC populations in control and inflamed mitral valves. Immunostaining and conventional histology were used to evaluate lineage tracing and validate scSeq findings. The effects of vascular endothelial growth factor receptor 3 (VEGFR-3) and VEGF-C inhibitors were tested in vivo. The functional impact of mitral valve disease in the K/B.g7 mouse was evaluated using echocardiography. Finally, to translate our findings, we analyzed valves from human patients with RHD undergoing mitral valve replacements.

Results:

Lineage tracing in K/B.g7 mice revealed new capillary lymphatic vessels arising from valve surface ECs during the progression of disease in K/B.g7 mice. Unsupervised clustering of mitral valve scSeq data revealed novel lymphatic valve ECs (Lymph-VECs) that express a transcriptional profile distinct from other VEC populations including recently identified Prospero homeobox protein 1 (PROX1)+ Lymph-VECs. During disease progression, these newly identified Lymph-VECs expand and upregulate a pro-fibrotic transcriptional profile. Inhibiting VEGFR3 through multiple approaches prevented expansion of this mitral valve lymphatic network. Echocardiography demonstrated that K/B.g7 mice have left ventricular dysfunction and mitral valve stenosis. Valve lymphatic density increased with age in K/B.g7 mice and correlated with worsened ventricular dysfunction. Importantly, human rheumatic valves contained similar lymphatics in greater numbers than non-rheumatic controls.

Conclusion:

These studies reveal a novel mode of inflammation-associated, VEGFR3-dependent postnatal lymphangiogenesis in murine autoimmune valvular carditis, with similarities to human RHD.

Graphical Abstract

INTRODUCTION

While the general link between autoimmune disease and cardiovascular disease is appreciated, the exact mechanisms by which autoantibody-associated rheumatic diseases lead to tissue damage in the heart remain poorly understood. Diseases such as rheumatoid arthritis, acute rheumatic fever (ARF), and systemic lupus erythematosus are each associated with increased risk of cardiovascular disease^1,2. In the acute phase of ARF, antibodies directed against Group A streptococcus cross-react with antigens in the patient, leading to arthritis, myocarditis, and valvular carditis³. Unrecognized or untreated ARF can progress to RHD predominantly affecting the mitral valves. In 2019, there were an estimated 40 million people living with RHD and 306,000 deaths attributed to RHD worldwide⁴. There are no options to medically manage severe RHD prior to surgical intervention, so improved understanding of the mechanisms driving valve inflammation hold the potential to improve therapy for patients with RHD and other rheumatic diseases.

To study how autoantibody-associated rheumatic diseases drive cardiovascular inflammation, we utilize the K/B.g7 mouse model of autoimmune valvular carditis. K/B.g7 mice spontaneously develop autoantibodies that recognize glucose-6-phosphate isomerase (GPI), a ubiquitously expressed glycolytic enzyme⁵. The mice develop arthritis and valvular carditis similar to the aforementioned human autoantibody-mediated diseases^5,6. Importantly, we recently showed that EC-specific tumor necrosis factor receptor 1 (TNFR1) expression is necessary for inducing valve disease in K/B.g7 mice⁷. We also reported that Lymphatic vessel endothelial hyaluronan receptor 1 (LYVE1)+VEGFR3+ lymphatic vessels are present in inflamed mitral valves⁷. This finding was particularly intriguing given recent studies demonstrating that cardiac valve-expressed PROX1, a key regulator of lymphatic cell identity, negatively regulates pathogenic extracellular matrix (ECM) remodeling to prevent myxomatous valve disease^8,9. In the present study, we sought to characterize valve ECs (VECs) more deeply, identify factors that regulate the growth of lymphatics in the inflamed mitral valve, and study their relevance to human RHD.

METHODS

Data Availability

The authors will make data, protocols, analytic methods, and study materials available to other researchers for purposes of reproducing the results or replicating the procedure upon request. The scRNA seq data reported in this study has been uploaded to the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) for public access under Accession code GSE221197.

Murine studies

Mouse lines

All animal experiments were conducted in accordance with the University of Minnesota Institutional Animal Care and Use Committee (IACUC) guidelines for the ethical care and treatment of laboratory animals (IACUC protocol # 2102-38871A). Standard housing in a specific pathogen free (SPF) facility was employed in all studies. All mouse lines, the number of mice, and sex of mice used in experiments are specified in the Major Resources Table in the Supplemental Materials. When conducting experiments, two different breeding schemes were utilized: 1) for scRNA seq and non-lineage-traced imaging studies, KRN mice crossed with NOD mice, designated “K/BxN”, and 2) for EC-lineage-tracing and VEGF-C-VEGFR3 blockade studies, KRN mice were crossed with C57BL/6J mice congenically expressing the NOD-derived H-2 locus (H-2^g7), resulting in mice designated “K/B.g7”. The autoimmune valvular carditis phenotype depends only on expression of KRN and the NOD-derived class II major histocompatibility molecule I-A^g7. We have observed no differences in arthritis or valvular carditis between K/B.g7 and K/BxN mice, so for simplicity refer to both here as “K/B.g7”.

Echocardiography

Echocardiography was utilized to characterize in vivo hemodynamics of autoimmune carditis. Mice were anesthetized with 1-4% inhaled isoflurane and affixed to a heated imaging platform. An 18-38 MHz linear array transducer (MS400, Vevo 2100, VisualSonics, Toronto, CA) was used to capture echocardiograms. Left ventricular ejection fraction was measured using the LV Trace two-dimensional edge-tracking software package with cine from the parasternal long axis window. Left ventricular diastolic function was measured using pulsed wave Doppler of mitral valve inflow (E/A ratio) and tissue Doppler of the septal mitral valve annulus (e’). For assessment of mitral valve stenosis, the pulsed wave Doppler-quantified velocity of blood flow across the mitral valve, in the apical four chamber view, was referenced to the blood flow across the pulmonary valve (reference valve), in the parasternal short axis view, to calculate a dimensionless index¹⁰. Lower dimensionless index indicates more severe valvular stenosis.

Treatments

EC lineage-tracing was performed using B.g7 Cdh5-Cre/ER^T2 ROSA^mT/mG and K/B.g7 Cdh5-Cre/ER^T2 ROSA^mT/mG mice. Male and female mice of each genotype were used for all harvest time points and are reported in Table S1. At 3 weeks of age, mice were given tamoxifen (Sigma T5648) resuspended in corn oil intraperitoneally at a dose of 2 mg per mouse per day for 3 consecutive days. Tissues were then harvested from mice at 4-, 8-, 12-, and 20-weeks of age. VEGFR3-blocking studies were performed using the VEGFR3 small molecule inhibitor MAZ51 (MedChemExpress HY-116624) reconstituted in the vehicle dimethylsulfoxide (DMSO). K/B.g7 mice were treated intraperitoneally at a dose of 10 mg/kg per day^11-13. VEGF-C trap studies were performed using AAV8 packaged with VEGFR3[1-4]-Ig¹⁴. Control mice were administered AAV8-VEGFR3[4-7]-Ig containing transmembrane domains 4-7, which do not interact with VEGF-C. Both viruses were purchased from the Helsinki Virus Core (University of Helsinki). AAV8-recipient mice were given 5x10¹¹ viral particles diluted in PBS via a single intraperitoneal injection at 6 weeks of age. An equal number of male and female mice were used in each experimental group of each treatment regimen. Arthritis was measured prior to the start of treatment and weekly during each treatment. Serum anti-GPI IgG titers were quantified at the end of the study. Expanded methods for measuring arthritis and antibody titers are in the Supplemental Materials.

Human tissue collection and processing

De-identified human cardiac valve samples were acquired by CardioStart International during charitable missions in countries with high incidences of RHD. Additional de-identified control cardiac valve samples were acquired from the University of Minnesota Clinical & Translational Science Institute’s Specimen & Data Repository or were taken from valves explanted during surgical correction (all harvesting followed University of Minnesota’s patient consent procedures directed by UMN IRB protocols 1307M39481 and 0305M47681). Histological analysis of these samples was deemed exempt from University of Minnesota Institutional Review Board oversight due to the de-identified nature of the samples. Following surgical resection, samples were placed in 10% neutral buffered formalin for short- term storage and for transportation to the laboratory. Each sample was subsequently placed in 70% ethanol until paraffin embedding and sectioning. Formalin-fixed, paraffin-embedded (FFPE) samples were sectioned at 7 μm, deparaffinized in xylene, and rehydrated using a decreasing ethanol/water gradient.

scRNA sequencing

A complete description of the approaches used to isolate cells, sequence RNA, process data, and conduct downstream analysis can be found in the Supplemental Materials. Briefly, live cells were sorted and sequenced using the Single Cell 3’ Reagent Kit and NovaSeq 6000 Sequencing System. Data were processed using Cell Ranger and then analyzed with the Seurat package.

Tissue histology

Tissue staining and imaging

A complete description of protocols used for immunofluorescence, whole mount, picrosirius red, and pentachrome staining, imaging, and quantification of murine and human tissue samples is in the Supplemental Methods.

Statistical Analysis

Statistics were calculated using Prism 9 software (GraphPad Software, Inc). Because sample n < 15 for all murine experiments, normal distribution could not be determined. Male and female mice were pooled for all statistical analyses and each sex represented by different symbols in data graphs because it was determined through 2-way ANOVA analyses that sex did not contribute to differential phenotypes in any treatment cohorts. Mann-Whitney tests were used to compare 2 experimental groups and 2-way ANOVAs were used to compare 3 or more experimental groups with 2 independent variables. Following 2-way ANOVA, post hoc Sidak multiple comparisons test (with single pooled variance) was run to compare experimental groups. Normality and equal variance were not assessed as a precondition for this analysis. Data are expressed as mean ± SD unless otherwise specified. P values and the specific statistical analyses used are specified in the figures and figure legends.

RESULTS

Endothelial cell lineage-tracing identifies emergent vessels in inflamed mitral valves

To study how ECs behave during the progression of autoimmune valvular carditis, we treated 3-week-old K/B.g7 Cdh5-CreER^T2 ROSA^mTmG/+ mice with tamoxifen to label ECs prior to the development of valvular carditis (Figure S1A, Table S1). The ROSA^mTmG reporter drives expression of membrane Tomato in all cells; Cre-mediated recombination switches the reporter to membrane GFP¹⁵. Tamoxifen did not disrupt the development of anti-GPI autoantibodies, arthritis, or valvular carditis (Figure S1B-D). Furthermore, we verified that random Cre-mediated recombination did not occur in VECs in mice treated with vehicle (Figure S1E). We analyzed mGFP-expression at 4, 8, 12, and 20 weeks of age (Figure S1F). With this strategy, we efficiently labeled VECs on the surface of the mitral valves by 4 weeks of age, immediately prior to the development of valvular carditis⁶ (Figure S2). At later stages of disease progression, we identified collections of mGFP+ lineage-traced cells located deeper within the inflamed mitral valves (Figure 1A).

Figure 1. Identification of lymphatic vessels in inflamed mitral valves with in vivo lineage-tracing.

A. Fluorescence images of mTom and mGFP labeling in 12-week-old B.g7 and K/B.g7 lineage-traced mice. Boxes highlight identified mGFP+ vessels which are displayed at higher magnification in the right-most boxes. Scale bars are 50 μm. LA: left atrium, LV: left ventricle. B. Immunostaining of endothelial markers CD31 and ERG (first and second rows), lymphatic marker LYVE1 (middle row), hematopoietic marker CD45 (fourth row), and fibroblast marker Periostin (bottom row) in 20 week old K/B.g7 lineage-traced mitral valves. Arrows point to mGFP+ vessels. C. Quantification of mGFP+ vessel numbers in mitral valves of 4-, 8-, 12-, and 20-week old B.g7 and K/B.g7 lineage-traced mice. Each data point represents a different mouse. Statistics were calculated with Mann Whitney U tests. D. Quantification of mGFP+LYVE1+ vessel numbers in mitral valves of 4-, 8-, 12-, and 20-week old B.g7 and K/B.g7 lineage-traced mice (top graph) and proportion of mGFP+ vessels that were LYVE1+ in each genotype and age group (bottom graph). In C-D, circles indicate female mice and triangles indicate male mice. E. Paired bar graphs of the distance of mGFP+ vessels from the atrial surface and total valve thickness at the location of each vessel for 12- and 20-week old B.g7 and K/B.g7 lineage-traced mice. Each data point represents a different mouse and lines connect paired metrics per mouse. Schematic depicts how measurements in Figure 1F were taken: a line oriented perpendicularly to the valve length was drawn through each mGFP+ vessel and both the total length of the line (valve thickness) and the distance from the middle of the vessel to the atrial surface along that line (atrial distance) were quantified. F. Whole mounted image of a 17-week-old K/B.g7 lineage-traced mitral valve. Anatomical features of the valve are labeled with an area of lymphatic vessels enlarged in the righthand image. LA: left atrium, LV: left ventricle.

These groups of mGFP+ cells formed vessel-like structures based on their morphology and expression of CD31 and the transcription factor ETS-related gene (ERG) (Figure 1B). Supporting our prior observations⁷, these mGFP+ vessels also expressed LYVE1 (Figure 1B). Importantly, they did not express markers of hematopoietic cells (CD45) or fibroblasts (periostin) (Figure 1B), confirming their endothelial phenotype. Furthermore, these mGFP+ cells did not express the macrophage marker CD68; however, immunostaining revealed that many LYVE1+CD68+ macrophages co-localized with LYVE1+mGFP+ lymphatics located at the annulus of uninflamed and throughout inflamed mitral valves (Figure S3).

Vascularization increases during the progression of valvular carditis and remains localized near the atrial surface of mitral valves

Next, we quantified the number of lineage-traced vessels to determine how systemic inflammation and age affected vascularization. In control, non-inflamed mice (B.g7 Cdh5-CreER^T2 ROSA^mTmG/+), we identified a small number of lineage-traced vessels whose numbers did not change with age (Figure 1C). In contrast, in mice with valvular carditis (K/B.g7 Cdh5-CreER^T2 ROSA^mTmG/+), the number of lineage-traced vessels increased with age, with significantly more of them at 12 and 20 weeks of age compared to control mice (Figure 1C). Additionally, the number of mGFP+ vessels correlated with overall valve thickness (Figure S1F). We found that the majority of these mGFP+ vessels expressed LYVE1 (Figure 1D) indicating that lymphatic but not blood capillary growth occurs in inflamed mitral valves of K/B.g7 mice.

The mGFP+ vessels were localized close to the atrial surface of the valve, regardless of valve thickness (Figure 1E). Morphologically, these LYVE1+mGFP+ cells formed lymphatic vessel structures that spread across much of the leaflet faces (Figure 1F, S4). We did not observe this morphology in either control B.g7 mitral valves (Figure S4C) or K/B.g7 tricuspid valves (Figure S4D) which are not affected in this model. Additionally, these vessels lacked markers of lymphatic muscle cells, such as alpha-smooth muscle actin (αSMA), supporting a capillary identity (Figure S5). Taken together, these data reveal that mitral valve lymphatic vascularization increases during the progression of autoimmune valvular carditis, correlates with valve thickening, and remains localized near the atrial surface of the valve.

Single cell sequencing reveals novel VEGFR3+ Lymph-VECs distinct from previously identified cardiac Lymph-VECs

Hulin et al previously identified Prox1+ lymphatic VECs (lymph-VECs) in uninflamed mitral valves of young mice. These Lymph-VECs localize to the sites of disturbed flow: the ventricular side of the mitral valve¹⁶. To complement these findings, we performed scRNA seq on all cells isolated from mitral valves of non-lineage-tracing mice with autoimmune valvular carditis (K/B.g7) and control mice (B.g7). The data were reduced by principal component analysis (PCA), followed by uniform manifold approximation projection (UMAP) dimensionality reduction (Figure S6A). This produced 24 unique clusters, some of which were manually merged based on marker gene expression, resulting in 14 cell populations in the mitral valve (Figure S6B, Table S3). Based on gene expression, we identified three Cdh5+ VEC clusters enriched for Vwf, Hapln1, or Prox1 expression, respectively. Each of these clusters aligned with VEC populations previously described by Hulin and colleagues¹⁶ – vascular-VECs (Vwf), coapt-VECs (Hapln1), and lymph-VECs (Prox1) (Figure S6C, Table S3).

In addition to VECs, we identified smooth muscle cells (Acta2), two populations of fibroblasts (Tcf21+ and Wif1+), myeloid cells (Adgre1, Lyz2), T cells (Cd3d, Cd28), mesothelial cells (Msln), two populations of melanocyte-like cells (Kcna1+ and Pmel+), red blood cells (Hba-a1), B cells (Cd19, Ms4a1), and proliferating cells (Mki67) (Figure S6D-E, Table S3).

To characterize the VEC populations further, we applied PCA and UMAP dimensionality reduction to only the three VEC clusters. This resulted in seven clusters that we classified as: endocardial (VEC Cluster 0), lymphatic (VEC Clusters 1, 2, 3), coapt (VEC Cluster 4), arterial/capillary-like (VEC Cluster 5), and proliferating (VEC Cluster 6) (Figure S7A, 2A-B, Table S4). The endocardial population (VEC Cluster 0) expressed high levels of a Npr3, which is generally associated with this population^17,18, and Vwf, which was previously defined as vascular VECs (Figure 2Ci)¹⁶. We found that these cells localized to the surface of the valves and were the most abundant VEC population (Figure S7B). The presence of an endocardial signature within VECs is not surprising given the embryonic origin of the cardiac valves from the endocardium^19,20. Cluster 4 expressed genes including Prnp, Hapln1, Wnt9b, and Serpinb1a, which are enriched in coapt-VECs¹⁶ (Figure 2Cii, S7C). Interestingly, Cdh11, which has been reported to be expressed in endocardial cells^17,21,22, was also expressed in coapt-VECs in our dataset (Figure S7B-C). Cluster 5 expressed many genes characteristic of arterial ECs including Sema3g, Jag2, Sox17^18,23 (Figure S7D). Cluster 6 expressed genes related to cell proliferation such as Mki67, Top2a, and Cenpf²³ (Table S4).

Figure 2. Single-cell RNA sequencing reveals a heterogeneous cell population including Lymph-VEC subsets differentially localized in the MV.

A. UMAP plot of all VEC clusters in 3-, 8-, and 25-week old B.g7 and K/B.g7 mice (n = 2 per age and genotype, one male and one female in each experimental group) overlaid with Clusters 0-6. B. Table reporting the top 10 differentially expressed genes in each VEC cluster. C. Genes enriched in Hulin et al’s VEC, coapt-VEC, and Lymph-VEC populations overlaid on the UMAP plot. D. Prox1 and Lyve1 transcript expression levels overlaid on UMAP plots with respective immunostaining of PROX1 and LYVE1 in serial sections of a 12-week-old K/B.g7 MV immediately below. Arrowheads delineate PROX1+LYVE1^neg cells. Scale bars are 100 μm. A: atrium, V: ventricle. E. Ccl21a and Flt4 transcript expression levels overlaid on UMAP plots with respective immunostaining of CCL21 and VEGFR3 in serial sections of a 12-week-old K/B.g7 EC-lineage-traced MV immediately below. Scale bars are 50 μm.

The lymphatic VEC (Lymph-VEC) cells separated into three clusters. Lymph-VEC 1 expressed many genes found in the Lymph-VEC population reported by Hulin et al.¹⁶ (Figure 2Ciii) including Prox1 (Figure 2D). Interestingly, Foxc2, a key regulator of lymphatic valve formation, was significantly upregulated in Lymph-VEC 1 cells (Figure S7E), a point that we address in the Discussion. Importantly, the Lymph-VEC 1 population lacked expression of Lyve1, which was enriched in Lymph-VEC 2 and −3 (Figure 2D). Immunostaining demonstrated two separate populations of PROX1+ cells: one that did not express LYVE1 and localized to the ventricular side of the mitral valve similar to the previously-reported Lymph-VECs (here termed Lymph-VEC 1)¹⁶, and another PROX1+ population that did co-express LYVE1 and localized within the lymphatic vessels of K/B.g7 mitral valves (Figure 2D). Lymph-VEC 2 and −3 expressed additional lymphatic markers Ccl21a/CCL21 and Flt4/VEGFR3 that colocalized specifically with the mGFP+ vessels; these markers were absent in Lymph-VEC 1 (Figure 2E). This was further supported by pathway enrichment analysis of differentially expressed genes between Lymph-VEC 1 and Lymph-VEC 2/3, which revealed enrichment of pathways related to lymphatic EC differentiation and receptor tyrosine kinase signaling in Lymph-VEC 2/3 cells (Figure S7F, Table S5). We thus conclude that the two additional Lymph-VEC populations −2 and −3 identified in our scSeq data set correspond to the mitral valve lymphatic vessels that emerge during autoantibody-induced inflammation in K/B.g7 mice.

Lymph-VEC 2 cells upregulate a pro-fibrotic program with age during inflammation

Together, the number of Lymph-VEC 2 and −3 cells increased more substantially with age than Lymph-VEC 1 cells in the inflamed mitral valves (Figure S7G). More specifically, the fraction of Lymph-VEC 2 but not −1 or −3 increased proportionally with inflammation (Figure 3A). To further investigate the changes occurring in VEC populations during autoantibody-induced inflammation, we conducted pseudotime trajectory analyses of the total B.g7 and K/B.g7 VEC populations (Figure 3B). Outcomes demonstrated that Lymph-VECs 2 and 3 were transcriptionally distinct from the remaining VEC populations including Lymph-VEC 1 (delineated by an “*1” in Figure 3B). Furthermore, a subset of 25-week old Lymph-VEC 2 cells emerged from the Lymph VEC 2/3 cluster only in the K/B.g7 trajectory (circled in Figure 3B).

Figure 3. A distinct population of Lymph-VEC 2 cells adopt fibrotic characteristics during chronic inflammation.

A. Quantification of Lymph-VEC 1-3 frequencies amongst total VECs (top) and total Lymph-VECs (bottom) in 3-, 8-, and 25-week old K/B.g7 valves from the scRNA seq dataset. B. Individual pseudotime trajectories of B.g7 and K/B.g7 VECs superimposed with clusters (top) or mouse age (bottom). Trajectories were created using slingshot. “*1” indicate the location of Lymph-VEC 1 cells in each plot. Dotted circles highlight the subset of Lymph-VEC 2 cells that emerge in aged K/B.g7 valves. C. Pseudotime trajectories of Lymph-VECs from B.g7 or K/B.g7 valves superimposed with clusters (right) or mouse age (left). D. Bar plot of the negative log10-transformed p-values of the 10 most significant enriched pathways that are up-regulated (green) or down-regulated (red) in K/B.g7 Lymph-VECs found at the end of the trajectory shown in C. Tradeseq was used to identify genes that were differentially expressed at the end of the K/B.g7 Lymph-VEC trajectory compared to the start of the trajectory using a start-vs-end test. Significantly up-regulated (160) or down-regulated (158) genes were determined as those with an adjusted p-value < 0.05. Represented pathways are from Gene Ontology’s Biological Processes.

The clear separation of Lymph-VEC 2/3 cells from the remaining VECs (Figure 3B) suggests that these cells originate from their own cellular source. Lymph-VEC 2/3 cells are present in uninflamed control valves of B.g7 mice (Figure S7G). Immunostaining revealed isolated LYVE1+mGFP+ cells on the atrial surface of both B.g7 and K/B.g7 valves (Figure S8A). Budding of LYVE1+ lymphatic vessels from the atrial side of the valve was detected as early as 4 and 8 weeks of age in inflamed K/B.g7 valves (Figure S8B-C). In summary, we identified isolated LYVE1+mGFP+ cells in both uninflamed and inflamed valves. In the setting of chronic inflammation, these cells appear to give rise to valve lymphatic vessels consistent with our scSeq findings.

To better understand the changes occurring in the Lymph VEC 2/3 cells, we conducted pseudotime trajectories in just these two clusters (Figure 3C). There was a distinct difference in the trajectory pattern between B.g7 and K/B.g7 cells where K/B.g7 cells clearly separated based on Cluster (Lymph-VEC 2 vs. −3) as well as along an age-related trajectory, moving over time from Lymph-VEC 3 to 2 and from 3 weeks to 25 weeks. In contrast, Lymph-VEC 2 and −3 cells of B.g7 mice still separated by cluster as expected, but no correlation with age was appreciated (Figure 3C). Pathway analysis of differentially expressed transcripts along the Lymph-VEC K/B.g7 trajectory revealed progressive up-regulation of ECM-related pathways and down-regulation of pathways related to cell proliferation and motility (Figure 3D, Table S6). Assessment of individual differentially expressed genes revealed up-regulation of many ECM-associated genes (Col1a1, Vim, Loxl2, Timp2, Mfap4, Lyve1), as well as fibroblast-associated genes (Dcn, Gsn, Wif1, Fbln2), and chemokines (Ccl21a and Cxcl14). Significantly down-regulated genes included those related to endothelial-specific function and identity (Pecam1, Cdh5, Emcn, Tie1, Eng, Nos3, Tbx1), vascular growth (Pdgfb, Stab1), and lymphatic differentiation and function (Prox1, Gata2, Nfat5) (Figure 3D). Thus, this emergent population of aged Lymph-VEC 2 cells appears to lose traditional lymphatic and endothelial characteristics and upregulate transcripts related to ECM regulation, reflecting the fibrotic nature of aged K/B.g7 valves²⁴. Immunostaining of collagen type I in EC lineage-traced valves of aged K/B.g7 mice supported these findings: we identified many lymphatics co-localized near collagen (Figure S9). Together, scSeq analysis demonstrates that Lymph-VEC 2/3 cells retain a distinct identity compared to other VECs; this identity changes over time in the setting of inflammation to become more pro-fibrotic.

VEGFC-VEGFR3 signaling regulates lymphatic outgrowth in the valves

Due to high expression of Flt4/VEGFR3 in the emergent mitral valve lymphatics (Figure 2E), we hypothesized that VEGF-C/VEGFR3 signaling was important for lymphatic outgrowth in this disease setting. Indeed, the transcript encoding VEGFR3, Flt4, is expressed almost exclusively in the Lymph-VECs 2/3 in our dataset (Figure S10A). Secondarily, we sought to determine if the emergent valve lymphatic vessels contributed to progression of autoimmune valvular carditis. To explore these possibilities, we treated K/B.g7 mice with a VEGFR3 small molecular inhibitor MAZ51^12,13,25,26 or adeno-associated vector encoding the soluble VEGF-C-trap (AAV8-VEGFR3[1-4]-Ig)¹⁴ from 6- to 10-weeks of age (Figure 4A,F). Importantly, neither treatment approach interfered with anti-GPI autoantibody titers, body mass, or arthritis development (Figure S11).

Figure 4. Blockade of VEGFR3 signaling inhibits mitral valve lymphatic growth.

A. MAZ51 treatment scheme. B. Representative immunostaining of VEGFR3, LYVE1, and CD31 in mitral valves of MAZ51- (n = 9) or control- (n = 9) treated mice. A: atria, V: ventricle. Scale bars in left, lower magnification panels are 100 μm and in right, higher magnification panels are 25 μm. C. Bar graph of the number of VEGFR3+LYVE1+ lymphatic vessels in mitral valves of K/B.g7 mice treated with MAZ51 or vehicle control. D. Bar graph of the average mitral valve (MV) thickness of each mouse treated with control or MAZ51. E. Bar graph of the largest MV thickness measurement from each mouse treated with control or MAZ51. F. VEGF-C Trap treatment scheme. G. Representative immunostaining of VEGFR3, LYVE1, and CD31 in mitral valves of VEGF-C Trap- (n = 7) or control- (n = 8) treated mice. A: atria, V: ventricle. Scale bars in left, lower magnification panels are 100 μm and in right, higher magnification panels are 25 μm. H. Bar graph of the number of VEGFR3+LYVE1+ lymphatic vessels in mitral valves of K/B.g7 mice treated with VEGF-C Trap or control. I. Bar graph of the average mitral valve (MV) thickness of each mouse treated with control or VEGF-C Trap. J. Bar graph of the largest MV thickness measurement from each mouse treated with control or VEGF-C Trap. Mann Whitney U tests were used and circles indicate female mice and triangles indicate male mice in C-E and H-J.

Both MAZ51 and VEGF-C-trap treatments effectively blocked lymphatic growth in the mitral valves, as the number of VEGFR3+LYVE1+ valve lymphatic vessels was significantly reduced in MAZ51- and VEGF-C-trap-treated mice compared to controls (Figure 4B-C,G-H). We found that neither MAZ51 nor VEGF-C-trap treatment affected the number of existing VEGFR3+ lymphatic numbers in the uninflamed left atrium or ventricles (Figure S10C). Interestingly, treatment with MAZ51 significantly, albeit subtly, reduced both average and maximal mitral valve thickness (Figure 4D-E), while VEGF-C-trap treatment did not (Figure 4I-J). We observed increased collagen hybridizing peptide (CHP) binding in MAZ51-treated valves (Figures S12) reflecting an increase in collagen degradation²⁷. This finding supports the observed decreased valve thickness in MAZ51-treated valves as it is known that valve thickening is partially due to increased collagen production^24,28. Indeed, we found that the area of positive collagen staining decreased only in MAZ51-treated valves (Figure S12D) and collagen area significantly correlated with valve thickness (Figure S12E). These findings demonstrate that VEGFC-VEGFR3 signaling is important for the growth and/or maintenance of lymphatic vessels in the inflamed valve during peak inflammation, but loss of these lymphatics during that time frame does not directly affect valve disease progression.

Based on these findings, we investigated which cells express VEGFR3 ligand VEGF-C within the mitral valve. ScRNA seq data demonstrated that macrophages, ECs, and fibroblasts all expressed the Vegfc transcript (Figure S13A). Validation with flow cytometry confirmed that macrophages and CD45- stromal and vascular cells expressed VEGF-C (Figure S13B). The number of VEGF-C-expressing macrophages and stromal cells and the mean fluorescence intensity (MFI) of VEGF-C increased with inflammation (Figure S13C). We found that the folate receptor-β- and CD206-expressing resident macrophages expressed higher levels of VEGF-C than the remaining macrophage subsets (Figure 13D).

Lymphatic density correlates with mitral valve stenosis and left ventricular dysfunction in aged mice

To date, the impact of mitral valve inflammation and thickening on cardiac function has not been investigated in the K/B.g7 model. We performed echocardiography on K/B.g7 mice aged 6 to 9 months as well as uninflamed control mice matched by age and sex. K/B.g7 mice exhibited no change in left ventricular ejection fraction (Figure 5A) or E/A ratio (Figure 5B); however, the E/e’ ratio increased (Figure 5C) and pulmonary valve to mitral valve dimensionless index decreased in K/B.g7 mice (Figure 5D). The dimensionless index has been demonstrated as a reliable metric for identifying hemodynamically significant mitral valve stenosis²⁹. Importantly, both valve thickness (Figure 5E) and lymphatic density (Figure 5F) were greater in these aged K/B.g7 mice than in controls. Together, these results suggest that inflammation and resultant fibrosis that occurs in the mitral valves of K/B.g7 mice induces measurable left ventricular diastolic dysfunction and mitral valve stenosis, and these metrics correlate with increased valve thickness and lymphatic density.

Figure 5. Valve thickness and lymphatic density correlate with worsened indices of cardiac function in aged K/B.g7 mice.

A-D. Echocardiograph was used to quantify the left ventricular ejection fraction (A), E/A ratio (B), E/e’ ratio (C), and PA/MV dimensionless index (D) in K/B.g7 and uninflamed controls aged 6-9 months. Using histological approaches, valve thickness (E) and lymphatic density (F) were quantified in each mouse. Circles indicate female mice and triangles indicate male mice. Mann Whitney U tests were used to calculate statistics.

Lymphatic structures are present in human rheumatic mitral valves

To validate our findings in human disease, we investigated whether lymphatic vessels could be identified in mitral valves from patients with RHD (Table S7, Figure S14A). Pentachrome staining of valves revealed that RHD valves appeared more vascularized than control valves (Figure 6A). Further immunostaining demonstrated the presence of lymphatic vessels expressing VEGFR3 and LYVE1 as well as VWF in rheumatic valves (Figure 6B, Figure S14B). We found that LYVE1− and VEGFR3-positive vessels also expressed VWF, but there were also occasional vessels that expressed only VWF (Figure S14C). As observed in the mice, human valves also contained LYVE1-expressing macrophages, which localized perivascularly in some instances (Figure S14D). We quantified the frequency at which valve samples had lymphatic vessels present and found that rheumatic valves contained LYVE1+ and VEGFR3+ vessels much more frequently than did non-rheumatic controls (Figure 6C). When interpreting these results, it should be noted that the area of the valve obtained per subject could not be standardized. These data demonstrate that human RHD mitral valves contain lymphatic vessels that express LYVE1 and VEGFR3, paralleling our observations in the K/B.g7 mouse model.

Figure 6. Lymphatic vessels identified more frequently in human rheumatic valve samples.

A. Images of pentachrome staining on rheumatic and non-rheumatic human valve sections. Scale bars of lower magnification images are 500 μm and higher magnification are 100 μm. B. Immunostaining of case/rheumatic and control valves for LYVE1, VEGFR3, and VWF using serial sections from the same region of each indicated valve. Rightmost images are enlarged from the area delineated with a white box in each merged image. Scale bars are 100 μm. C. Immunostained rheumatic and non-rheumatic valves were reviewed for presence of VEGFR3+ or LYVE1+ lymphatic vessels and percentage of valves with lymphatics present were calculated.

DISCUSSION

Here we used EC lineage-tracing and scRNA Seq to study the localization and characteristics of VECs, and more specifically lymphatic VECs, and how they contribute to autoimmune valvular carditis. Although many studies have reported the importance of lymphatics in regulating cardiac development³⁰, responding to ischemia during myocardial infarction^31-34, and regulating atherosclerosis development^35-37; our study is the first to investigate the role of lymphatics in the mitral valve.

We identified and characterized cardiac valve lymphatic vessels that 1) derived from Cdh5-expressing VEC, 2) were rare in healthy murine mitral valves and increased in number in a VEGFR3-dependent manner, and 3) correlated with worsened cardiac function. Importantly, assessment of human mitral valves indicated an increased presence of lymphatics in rheumatic valves compared to non-rheumatic controls. This work substantiates previous studies reporting individual instances of lymphatic vessels in canine, pig, and human mitral valves in a variety of disease states^38-44.

Blockade of lymphatic growth by impeding VEGFC-VEGFR3 binding did not alter valve thickening in the K/B.g7 model. This finding may indicate that growth of lymphatics into inflamed mitral valves is simply an epiphenomenon that occurs during autoimmune valvular carditis. However, it is important to address the factors of time and function in our interpretation of these results. First, the 6–10-week treatment time point we assessed would not address the possibility that very early or late, chronic stage signaling within lymphatic VECs promotes disease. Lymphatic-expression of the chemokine CCL21 is one candidate that could influence disease progression by recruiting or retaining proinflammatory immune cells within the valves. This may be important later in disease if these valve lymphatics do not function properly as is observed in other chronic inflammatory diseases^45-50. Specifically, chemokine signaling could attract immune cells near the lymphatics and, without proper trans-endothelial migration and/or draining functions, compound inflammation^46,51. Alternatively, valve Lymph-VECs could be activating immune cells to promote an inflammatory environment early in disease. Indeed, we observed close interactions between lymphatics and macrophages in uninflamed mitral valves, and previous work demonstrates that lymphatic ECs can regulate the behavior of macrophages⁵². In turn, these macrophages have the capacity to promote lymphatic growth and fibrosis via mechanisms including VEGF-C, TNF, and TGFβ^53-56. We validated that macrophages express VEGF-C in the mitral valve, and previously demonstrated they produce TNF which then acts on ECs to promote disease progression²⁴. Whether lymphatic specific TNFR1 signaling is necessary for valve disease remains unclear. We have shown that TGFβ signaling via the TGFβR2 on ECs does not affect valve disease progression, but it remains possible that TGFβ action on fibroblasts regulates disease progression⁷. Future lymphatic inhibitor and genetic studies (including macrophage and fibroblast-specific knockouts) with discrete timing considerations will be needed to test these hypotheses and to understand more fully the functional roles of lymphatics, fibroblasts, and macrophages in the inflamed mitral valve.

Our Cdh5-dependent lineage-tracing approach showed that these new lymphatic vessels arose from VECs present prior to the development of autoimmune valvular carditis. Most postnatal lymphangiogenesis occurs via proliferation of lymphatic ECs with subsequent migration away from pre-existing vessels, a process termed sprouting^57-59. In our study, however, we did not identify such pre-existing vessels in normal cardiac valves, although we did identify individual VECs with a lymphatic signature. This raises the possibility that certain isolated VECs have the potential to give rise to new lymphatic capillaries. Our trajectory analysis shows that Lymph-VECs maintained an identity distinct from the other VEC populations, consistent with this notion that specialized, non-vessel-constituent Lymph-VECs can give rise to valve lymphatic capillaries. Future studies tracking clonality within the VEC and Lymph-VEC populations are needed to directly test our hypothesis.

In addition to the lymphatic capillaries (Lymph-VEC clusters 2 and −3), we also identified a distinct population of VECs that express PROX1 but not LYVE1 (Lymph-VEC 1). These PROX1+LYVE1^neg VECs appear similar in gene expression and localization to those initially described by Hulin and colleagues¹⁶ and subsequently demonstrated to regulate pathogenic ECM remodeling in a PROX1-dependent manner in aortic valves^8,9. Although these cells express PROX1, they do not express other typical lymphatic markers, so their exact identity remains unclear. PROX1 is also expressed in venous ECs and plays an important role in the development of lymphovenous valves⁶⁰, providing precedence for the existence and importance of non-lymphatic PROX1+ ECs. Interestingly, in our dataset, these PROX1+LYVE1^neg VECs express Foxc2, a transcription factor crucial for the development and maintenance of the lymphatic collecting system^60-63. Foxc2 is upregulated in response to oscillatory flow^60,61,63, which is experienced by ECs on the ventricular side of the mitral valve. Additionally, PROX1 promotes Foxc2 expression through interaction with beta-catenin during lymphatic differentiation⁶⁴ and FOXC2 cooperates with PROX1 to promote lymphatic valve formation⁶². We speculate that these PROX1+LYVE1^neg VECs on the ventricular surface of the mitral valve represent mechanosensing cells that respond to oscillatory shear stress, a hypothesis we are keenly interested in exploring.

In summary, we have identified and studied the role of cardiac VECs in a mouse model of systemic autoimmune valvular carditis using in vivo lineage-tracing and single-cell RNA sequencing. We showed that Cdh5-expressing ECs give rise to a network of capillary lymphatic vessels during the progression of autoimmune valvular carditis through a VEGFR3-dependent mechanism. Finally, we demonstrated increased frequency of VEGFR3+ lymphatic vessels in mitral valves of human RHD patients, suggesting that targeting of maladaptive lymphangiogenesis may be therapeutically beneficial or serve as a novel biomarker for severity of RHD. Overall, this study provides novel single-cell characterization of healthy and inflamed adult mitral valves and broadens our understanding of the tissue-specific roles that lymphatics play in disease.

HIGHLIGHTS.

• We identified two additional populations of lymphatic valve endothelial cells (Lymph-VECs), distinct from the recently identified PROX1-expressing Lymph-VECs that control pathogenic thickening and extracellular matrix remodeling in myxomatous disease.

• In the setting of inflammation, these novel Lymph-VECs form lymphatic vessels in a VEGFR3-dependent manner on the atrial side of the mitral valve and correlate with worsened indices of cardiac function.

• The number of lymphatic vessels in mitral valves increases in the setting of human rheumatic heart disease compared to non-rheumatic controls.

NON-STANDARD ABBREVIATIONS AND ACRONYMS

ARF

Acute rheumatic fever

ECM

Extracellular matrix

Lymph-VECs

Lymphatic valve endothelial cells

LYVE1

Lymphatic vessel endothelial hyaluronan receptor 1

PROX1

Prospero homeobox protein 1

RHD

Rheumatic heart disease

VECs

Valve endothelial cells

VEGFR3

Vascular endothelial growth factor receptor 3