The thrombin receptor PAR1 orchestrates changes in lymphatic endothelial cell junction morphology to augment lymphatic drainage during lung injury

Abstract

The lung lymphatic vasculature is capable of remarkable increases in lymphatic drainage in settings of inflammation and edema; however, the mechanisms driving this are not clear. Here we show that lung injury transforms the configuration of lung lymphatic endothelial cell junctions from a continuous 'zippered' configuration to a discontinuous and permeable 'button' configuration. Despite similarity to the junctional changes often seen in leaky and dysfunctional blood vessels, we find that the shift to button junctions in the lymphatic vasculature has an opposite effect, resulting in augmented lung lymphatic drainage. Mechanistically, we demonstrate that lung lymphatic button junction formation in models of lung injury is dependent on the thrombin receptor protease-activated receptor 1, a known mediator of blood vessel permeability. These results uncover a previously unknown role for the thrombin receptor protease-activated receptor 1 in the lymphatic vasculature that promotes a similar change in junction morphology as seen in blood vessels, but with a disparate effect on lymphatic function.

Fig. 1. Lung lymphatic collector junctions transform to a buttoned morphology with lung injury.

a, Graphical depiction of buttoned lymphatic capillaries draining lymphatic fluid and cells to zippered lymphatic collecting vessels with valves. b, Representative PCLS image of pulmonary lymphatic vessels relative to lung structures (Prox1-EGFP, green; VE-cadherin, blue; smooth muscle actin (SMA), red). The boxes indicate areas shown in higher magnification in c–f. c–f, Visualization of LEC junctions in lung lymphatic collectors (c and d) and initial lymphatic capillaries (e and f) (Prox1-EGFP, green; VE-cadherin, black). c, GFP labeling of a lung lymphatic collector vessel. d, VE-cadherin staining of the lung lymphatic shown in c demonstrating continuous zipper junction morphology. e, GFP labeling of a lung initial lymphatic. f, VE-cadherin staining of the lung lymphatic shown in e with discontinuous button morphology. g, Quantification of junction morphology among visualized lymphatics (n = 5). Each dot represents junction quantification for an individual mouse, with 8–12 images used for each mouse. A total of 507 unique junctions were annotated for quantification. A one-way analysis of variance (ANOVA) was used; ****P < 0.0001. h–k, Representative PCLS images of LEC junctions in lymphatic collecting vessels from mice exposed to hyperoxia-induced lung injury or control room air (Prox1-EGFP, green; VE-cadherin, black). h, GFP labeling of a lung lymphatic vessel from a room air-exposed mouse. i, VE-cadherin staining of the lung lymphatic vessel shown in h with zippered junction morphology. j, GFP labeling of a lung lymphatic vessel from a mouse exposed to hyperoxia. k, VE-cadherin staining of the lung lymphatic vessel shown in j with buttoned morphology. l, Quantification of junction types in room air mice (n = 3) compared to mice exposed to hyperoxia for 84 h (n = 2). m, Percentage of button junctions in the experiment presented in l. Each dot represents junction quantification for an individual mouse, with 8–12 images used for each mouse. A total of 1,038 unique junctions were annotated for quantification. n, Representative in vivo imaging system (IVIS) images of ICG fluorescence detected over the chest of mice initially (0 h) or 6 h after intratracheal ICG instillation. o, Decrease in ICG fluorescence intensity over time relative to peak ICG fluorescence in room air mice (n = 9, solid line) compared to mice exposed to hyperoxia for 72 h (n = 4, dashed line). p, Quantification of ICG signal at 6 h relative to peak ICG fluorescence. Each dot represents an individual mouse and reports fluorescence intensity at 6 h relative to peak ICG fluorescence intensity for that same mouse. An unpaired, two-tailed t-test was used; *P = 0.0184. Data are representative of at least two independent experiments. All mice used for this experiment were control PAR1^loxP/loxP mice in a mixed genetic background. The black arrows point to zipper junctions. The black arrowheads point to button junctions. All error bars represent the s.e.m. h–k, Scale bars, 20 μm.

Fig. 2. Loss of PAR1 in lymphatics prevents button formation during lung injury.

a, Top: immunoblot for phospho-PKC (pPKC), pan-PKC and actin in lysates of lung-derived human microvascular endothelial cells (HMVEC-Ls) grown in culture and treated with 1 U ml⁻¹ thrombin for 10 min with or without pretreatment with 50 nM of the pan-PAR1 inhibitor vorapaxar for 1 h. Bottom: quantification performed using relative densitometric intensity of pPKC relative to pan-PKC. b, Schematic depiction of the hyperoxia experiments. Mice for the IVIS experiments underwent 72 h of hyperoxia to maximize survival during IVIS imaging. c, Top: immunoblot of whole-lung tissue from hyperoxia-injured mice (PAR1^loxP/loxP, n = 6) and room air mice (PAR1^loxP/loxP, n = 3) to detect the 55-kDa fibrin fragment. Bottom: quantification using the relative densitometric intensity of fibrin relative to actin. An unpaired two-tailed t-test was used; **P = 0.0045. d–k, Representative PCLS images of lung lymphatic junctions (Prox1-EGFP, green; VE-cadherin, black) from PAR1^loxP/loxP and PAR1^iLEC knockout (KO) mice exposed to room air (d–g) and PAR1^loxP/loxP and PAR1^iLEC KO mice exposed to hyperoxia (h–k). d, GFP labeling of a lung lymphatic from a PAR1^loxP/loxP mouse exposed to room air. e, VE-cadherin staining of the lung lymphatic shown in d with zippered junctions. f, GFP labeling of a lung lymphatic from a PAR1^iLEC KO mouse exposed to room air. g, VE-cadherin staining of the lung lymphatic shown in f with zippered junctions. h, GFP labeling of a lung lymphatic from a PAR1^loxP/loxP mouse exposed to hyperoxia. i, VE-cadherin staining of the lung lymphatic shown in h with buttoned morphology. j, GFP labeling of a lung lymphatic from a PAR1^iLEC KO mouse exposed to hyperoxia. k, VE-cadherin staining of the lung lymphatic shown in j with zippered morphology. The black arrowheads indicate button junctions. The black arrows indicate zipper junctions. l, Quantification of junction morphology in PAR1^loxP/loxP room air mice (n = 5), PAR1^iLEC KO room air mice (n = 3), PAR1^loxP/loxP hyperoxia mice (n = 3) and PAR1^iLEC KO hyperoxia mice (n = 4). m, Proportion of button junctions; 2,084 unique junctions were annotated for quantification in l and m. A one-way ANOVA was used; ****P = 0.0001, ****P < 0.0001. n, Representative IVIS images of ICG fluorescence detected after intratracheal ICG instillation. o, Decrease in ICG fluorescence intensity over time relative to peak ICG fluorescence in PAR1^loxP/loxP room air mice (n = 5, black solid line), PAR1^iLEC KO room air mice (n = 5, gray solid line), PAR1^loxP/loxP hyperoxia mice (n = 6, black dashed line) and PAR1^iLEC KO hyperoxia mice (n = 9, gray dashed line). p, Individual quantification of ICG retention at 6 h after peak ICG fluorescence. *P = 0.097. Data are representative of at least two independent experiments. All error bars represent the s.e.m. d–k, Scale bars, 20 μm. Source data

Fig. 3. PAR1 mediates zipper-to-button junction transformation in lung lymphatic collecting vessels after LPS-induced lung injury.

a, Schematic depiction of intratracheal LPS-induced lung injury. b, Left: immunoblot of whole-lung tissue from LPS-injured mice (PAR1^loxP/loxP mice, n = 4) and PBS controls (PAR1^loxP/loxP mice, n = 3) to detect a 55-kDa fibrin fragment. Right: quantification performed using relative densitometric intensity of fibrin to actin. An unpaired, two-tailed t-test was used; ***P = 0.0005. c–j, Representative PCLS images of lung lymphatic junctions (Prox1-EGFP, green; VE-cadherin, black) in PAR1^loxP/loxP and PAR1^iLEC KO mice given intratracheal LPS or PBS. c, GFP labeling of a lung lymphatic from a PAR1^loxP/loxP mouse treated with PBS. d, VE-cadherin staining of the lung lymphatic shown in c with zippered junctions. e, GFP labeling of a lung lymphatic vessel from a PAR1^iLEC KO mouse treated with PBS. f, VE-cadherin staining of the lung lymphatic vessel shown in e with zippered junctions. g, GFP labeling of a lung lymphatic vessel from a PAR1^loxP/loxP mouse treated with LPS. h, VE-cadherin staining of the lung lymphatic vessel shown in g with buttoned junctions. i, GFP labeling of the a lung lymphatic from a PAR1^iLEC KO mouse treated with LPS. j, VE-cadherin staining of the lung lymphatic vessel shown in i with zippered junctions. The black arrows indicate zipper junctions. The black arrowheads indicate button junctions. k, Quantification of junction types in PAR1^loxP/loxP (n = 3) and PAR1^iLEC KO (n = 5) mice treated with PBS, compared to PAR1^loxP/loxP (n = 8) and PAR1^iLEC KO (n = 4) mice after LPS-induced lung injury. l, Proportion of the button junctions presented in k. Each dot represents junction quantification of an individual mouse, with 8–12 images used for each mouse. A total of 2,723 unique junctions were annotated for k and l. A one-way ANOVA was used; ****P < 0.0001. Data are representative of at least two independent experiments. All error bars represent the s.e.m. c–j, Scale bars, 20 µm. Source data

Fig. 4. Blockade of PAR1/G_q signaling inhibits lung lymphatic button formation.

a, Schematic depiction of LPS-induced lung injury with PM2, a selective allosteric inhibitor of PAR1/G_q, in Prox1-EGFP lymphatic reporter mice. Mice for the IVIS experiments were shaved and allowed to recover an additional 24 h after LPS instillation to maximize mouse survival during IVIS imaging. b–g, Representative PCLS images of lung lymphatic junctions (Prox1-EGFP, green; VE-cadherin, black) mice treated with PBS, intratracheal LPS, or intratracheal LPS with preinjection of PM2. b, GFP labeling of a lung lymphatic vessel of a mouse treated with PBS. c, VE-cadherin staining of the lung lymphatic shown in b with zippered junctions. d, GFP labeling of a lung lymphatic from a mouse treated with LPS. e, VE-cadherin staining of the lung lymphatic shown in d with buttoned junctions. f, GFP labeling of a lung lymphatic from a mouse treated with LPS with preinjection of PM2. g, VE-cadherin staining of the lung lymphatic shown in f with zippered junctions. The black arrows indicate zipper junctions. The black arrowheads indicate button junctions. h, Quantification of junction types in PBS control mice (Prox1-EGFP mice, n = 4) compared to LPS-treated mice (Prox1-EGFP mice, n = 5) and LPS-treated mice preinjected with PM2 (Prox1-EGFP mice, n = 6). i, Proportion of the button junctions presented in h. A total of 2,252 unique junctions were annotated for h and i. A one-way ANOVA was used; ****P < 0.0001. j, Representative IVIS images of ICG fluorescence detected over the chest of mice after intratracheal ICG instillation, demonstrating the retention of ICG in PM2-treated mice. k, Decrease in ICG fluorescence intensity over time relative to peak ICG fluorescence in PBS control mice (Prox1-EGFP mice, n = 5, solid line), LPS-treated mice (Prox1-EGFP mice, n = 6, black dashed line) and LPS-treated mice preinjected with PM2 (Prox1-EGFP mice, n = 6, gray dashed line). l, Quantification of ICG retention 6 h after peak ICG fluorescence in k. A one-way ANOVA was used; from left to right, *P = 0.0386 and P = 0.0324. m, Left: proposed model of PAR1-dependent plasticity of pulmonary lymphatic collecting vessel junctions after lung injury. Right: lymphatic-endothelial-cell-specific loss of PAR1 (PAR1^iLEC KO) and inhibition of PAR1/G_q signaling with PM2 prevented zipper-to-button junction transformation in lymphatic collecting vessels and blunt lymphatic drainage. Data are representative of at least two independent experiments. All error bars represent the s.e.m. b–g, Scale bars, 20 μm.

Extended Data Fig. 1. Thrombin activates RhoA and PKC signaling in LECs in a PAR1-dependent manner.

(a) Immunofluorescence staining for RhoA-GTP (green) in cell cultures of Human Lung Microvascular Lymphatic Endothelial Cells (HMVEC-LLy) treated with 1 unit/ml thrombin for 10 min +/- pretreatment with 50 nM of the pan-PAR1 inhibitor vorapaxar for 1 hour. Quantification by mean corrected total cell fluorescence. DAPI staining shown in blue. Unpaired 2-tailed t-test; **, p = 0.0028. (b) Immunofluorescence staining for Rac1-GTP (red) in cell cultures of Human Lung Microvascular Lymphatic Endothelial Cells (HMVEC-LLy) treated with 1 unit/ml thrombin for 10 min +/- pretreatment with 20 nM APC for 1 hour. Quantification by mean corrected total cell fluorescence. DAPI staining shown in blue. (c) Left: Western blot for phospho-PKC (P-PKC) and actin in lysates of Human Lung Microvascular Lymphatic Endothelial Cells (HMVEC-LLy) grown in culture and treated with 1 unit/ml thrombin for 10 min +/- pretreatment with 10 μM parmodulin2 (PM2) for 1 hour. Right: Quantification by relative densitometric intensity of phospho-PKC relative to actin. Each point represents an individual well. Scale bars = 100 μm. All error bars represent SEM. Source data

Extended Data Fig. 2. Generation of Par1^flox mice.

(a) Wild type PAR1 allele (top), floxed PAR1+Neo^R allele (middle), and floxed PAR1 allele (bottom) are shown with location of probes and restriction sites relevant to the Southern blot. The location and direction of loxP sites and primers A, B, C, and D are shown with expected band sizes (dashed lines). The 5’ and 3’ homology regions and the floxed Neo^R and exon2 cassettes were from the targeting vector. The floxed PAR1+ Neo^R mouse line was therefore obtained using the targeting vector, whereas the floxed PAR1 mouse line was obtained by crossing the floxed PAR1+ Neo^R mouse line with MeuCre40 mice to perform in vivo excision of the Neo^R cassette. The resulting floxed PAR1 mouse line was subsequently backcrossed with C57BL/6 mice for 7 generations. (b) Southern blot performed from EcoRI-digested genomic DNA isolated from the ES cell clone that was microinjected into blastocysts to obtain the floxed PAR1+NeoR mice strain. The left panel shows Southern blot using the 5’ probe, which should produce bands of 26 and 20 kb. The right panel shows Southern blot using the 3’ probe, which should yield bands of 26 and 6.9 kB. (c) The gel image (black/white inverted) shows examples of genotyping of mice homozygous (hom) and heterozygous (het) for the floxed PAR1+NeoR allele as well as wild type mice (wt) using a multiplex PCR protocol with primers A, B, C, and D. The presence of the Neo^R cassette generates a band of 2.3 kb, and the presence of the floxed exon 2 generate bands of 1223 bp and 269 bp (small band absent from gel image) following digestion with MluI. The wild type PAR1 allele does not contain a MluI site and generates a single band of 1490 bp. Digested PCR products were run on 0.8% agarose gels containing ethidium bromide and photographed. M indicates the 1 kb DNA ladder (New England Biolabs). (d) The gel image shows examples of PCR genotyping of mice homozygous (hom) and heterozygous (het) for the floxed PAR1 allele as well as wild type mice (wt) using the same protocol as in (c).

Extended Data Fig. 3. Validation of PAR1^{iLEC KO} mice.

(a-d) Tracing assay for VEGFR3Cre^ERT2 using Lox-stop-lox;TdTomato;VEGFR3Cre^ERT2 mice. (a) Merged image of Lox-stop-lox;TdTomato;VEGFR3Cre^ERT2 mice after treatment with tamoxifen stained for anti-RFP (red, b) and LEC marker anti-CCL21 (green, c) demonstrating colocalization of TdTomato (red) and LECs (green) without significant TdTomato (Cre reporter) expression outside of LECs. (d) Merged image of Lox-stop-lox;TdTomato;VEGFR3Cre^ERT2 mice after treatment with tamoxifen stained for anti-RFP (red) and anti-VEGFR3 (LEC marker) also demonstrating colocalization of TdTomato (red) and LECs (green) without significant TdTomato expression outside of LECs. (e) Lung lymphatic endothelial cells (LECs) were isolated by flow cytometry for qPCR to assess knock-down efficiency of F2r (PAR1) in PAR1^iLECKO mice (n = 6) compared to PAR1^fl/fl mice (n = 3) with Gapd (GAPDH) as the reference housekeeping gene. Unpaired 2-tailed t-test; **, p = 0.0093. (f) BAL cell counts in room air PAR1^fl/fl (n = 4) and PAR1^{iLEC KO} (n = 4) mice compared to PAR1^fl/fl mice (n = 8) and PAR1^{iLEC KO} mice (n = 18) after 84 hours of hyperoxia exposure. One-way ANOVA; *, p = 0.0124; **, p = 0.0022; * p = 0.0332 from left to right. (g, h) Representative PCLS images demonstrating accumulation of CD45R⁺ leukocytes (blue) in the peri-interstitial alveolar spaces adjacent to lymphatic collectors (Prox1-EGFP, green) in PAR1^{iLEC KO} mice after hyperoxia-induced lung injury. BAL count and imaging data representative of at least 2 independent experiments. All error bars represent SEM. All scale bars = 200 µm. Source data

Extended Data Fig. 4. Thrombin inhibition with dabigatran partially inhibits zipper-to-button junction transformation in lymphatic collecting vessels after LPS-induced lung injury.

(a) Elevated BAL protein concentration in mice after hyperoxia-induced lung injury (PAR1^fl/fl mice, n = 14) compared to room air mice (PAR1^fl/fl mice, n = 6). Unpaired 2-tailed t-test; ****, p < 0.0001. (b) Absence of BAL protein concentration elevation in mice after LPS-induced lung injury (PAR1^fl/fl mice, n = 7) compared to control mice after PBS instillation (PAR1^fl/fl mice, n = 3). Unpaired 2-tailed t-test; p = 0.4488. (c-h) Representative PCLS images demonstrating LEC junctions in lymphatic collecting vessels in Prox1-EGFP reporter mice (Prox1-EGFP in green, VE-Cadherin in black) in control mice (c, d), intratracheal LPS-treated mice (e, f), and mice treated with LPS and dabigatran (g, h). Black arrows point to zipper junctions. Black arrowheads point to button junctions. (i) Quantification of junction types in PBS control mice (Prox1-EGFP mice, n = 3) compared to LPS-treated mice (Prox1-EGFP mice, n = 6) and LPS-treated mice given dabigatran (Prox1-EGFP mice, n = 6). (j) Graph of proportion of button junctions presented in i. Each dot represents junction quantification of an individual mouse, with 8–12 images used for each mouse. 3151 unique junctions were annotated for figures i and j. One-way ANOVA; **, p = 0.0022; ***, p = 0.003; ****, p < 0.0001. Scale bars = 20 μm. All error bars represent SEM. Source data