Research ArticleFibrosis

Targeting the vascular and perivascular niches as a regenerative therapy for lung and liver fibrosis

See allHide authors and affiliations

Science Translational Medicine  30 Aug 2017:
Vol. 9, Issue 405, eaai8710
DOI: 10.1126/scitranslmed.aai8710

To fix fibrosis, nix the niche

Organs such as the liver can mount a regenerative response to acute injury, but during chronic injury, this innate repair capacity is overwhelmed, leading to scarring and dysfunction. Cao et al. investigated whether targeting the microenvironment could improve therapeutic cell engraftment to reinstate regeneration in fibrotic organs. Treating mice with a dual therapy—an inhibitor of an enzyme up-regulated in fibrotic fibroblasts and viral particles that promoted endothelial-specific secretion of a growth factor—improved hepatocyte engraftment in fibrotic livers and epithelial cell engraftment in fibrotic lungs. Targeting both the vascular and perivascular regions could be key for restoring regeneration to fibrotic organs.


The regenerative capacity of lung and liver is sometimes impaired by chronic or overwhelming injury. Orthotopic transplantation of parenchymal stem cells to damaged organs might reinstate their self-repair ability. However, parenchymal cell engraftment is frequently hampered by the microenvironment in diseased recipient organs. We show that targeting both the vascular niche and perivascular fibroblasts establishes “hospitable soil” to foster the incorporation of “seed,” in this case, the engraftment of parenchymal cells in injured organs. Specifically, ectopic induction of endothelial cell (EC)–expressed paracrine/angiocrine hepatocyte growth factor (HGF) and inhibition of perivascular NOX4 [NADPH (reduced form of nicotinamide adenine dinucleotide phosphate) oxidase 4] synergistically enabled reconstitution of mouse and human parenchymal cells in damaged organs. Reciprocally, genetic knockout of Hgf in mouse ECs (HgfiΔEC/iΔEC) aberrantly up-regulated perivascular NOX4 during liver and lung regeneration. Dysregulated HGF and NOX4 pathways subverted the function of vascular and perivascular cells from an epithelially inductive niche to a microenvironment that inhibited parenchymal reconstitution. Perivascular NOX4 induction in HgfiΔEC/iΔEC mice recapitulated the phenotype of human and mouse liver and lung fibrosis. Consequently, EC-directed HGF and NOX4 inhibitor GKT137831 stimulated regenerative integration of mouse and human parenchymal cells in chronically injured lung and liver. Our data suggest that targeting dysfunctional perivascular and vascular cells in diseased organs can bypass fibrosis and enable reparative cell engraftment to reinstate lung and liver regeneration.


The self-repair capacity of liver and lung tissue is sometimes prohibited by overwhelming or persistent injury (119). Transplantation of parenchymal stem cells might aid in reinstating the regenerative ability (2038), but efficient engraftment of parenchymal cells is frequently handicapped by the prohibitive fibrotic microenvironment in diseased organs. Therefore, designing effective cell therapy strategies requires understanding how microenvironmental cues regulate parenchymal regeneration and fibrosis in the damaged lung and liver (3947).

Surgical resection of liver or lung lobes by partial hepatectomy (PH) or pneumonectomy (PNX) triggers facultative regeneration without fibrosis (4852). Liver and lung can also resolve fibrosis after acute injury (5355). In contrast, chronic injury stimulates exuberant scar formation and fibrosis, leading to hepatic and pulmonary failure (5663). How fibrosis is resolved after PH, PNX, or acute injury but triggered by chronic or overwhelming injury remains incompletely known (6466). Because fibroblasts are the major cell type that produces matrix proteins during scar formation (59, 6769), we hypothesized that modulating the interplay between perivascular fibroblasts and endothelial cells (ECs) might obviate fibrosis in the liver (2, 48, 49, 52) and lung (47, 7073), forming an epithelially inductive niche for parenchymal cell reconstitution and regeneration (74, 75).

Here, we find that EC-expressed hepatocyte growth factor (HGF) prevents aberrant activation of NOX4 [NADPH (reduced form of nicotinamide adenine dinucleotide phosphate) oxidase 4] (56, 76, 77) in perivascular fibroblasts after lung and liver injury. Induction of HGF and inhibition of NOX4 in damaged organs promote the incorporation of regenerative parenchymal cells. We also devised a strategy to edit both vascular and perivascular cells by combining endothelial Hgf gene delivery with NOX4 inhibition. This dual niche-editing strategy enhanced functional reconstitution of mouse and human parenchymal cells, inducing fibrosis-free organ repair. Our data suggest that targeting vascular and perivascular cells in diseased organs might transform the prohibitive microenvironment to an epithelially inductive niche that bypasses fibrosis and facilitates engraftment of regenerative progenitor cells.


Repeated lung and liver injuries prohibit the incorporation of grafted parenchymal cells

We first tested the efficiency of parenchymal cell engraftment in both normal and injured mouse lung and liver. Noninjured and injured lungs were transplanted with type 2 alveolar epithelial cells (AEC2s), cells that contribute to lung epithelialization (Fig. 1, A and B, and fig. S1A) (14, 21, 24, 26), and livers were grafted with hepatocytes mediating hepatic reconstitution (Fig. 1, C and D, and fig. S1B) (27, 34, 78). Lung injury was induced by intratracheal injection of bleomycin (Bleo) or hydrochloric acid (Acid) (46), and liver repair was triggered by intraperitoneal injection of carbon tetrachloride (CCl4). To trace the in vivo incorporation of transplanted parenchymal cells, we bred AEC2-specific surfactant protein C–CreERT2 (Sftpc-CreERT2) mice (14) and hepatocyte-specific Albumin-Cre mice with TdTomato reporter mice. Isolated TdTomato+ AEC2s or hepatocytes were transplanted into mice via intratracheal or intrasplenic injection, respectively. We found that there was little parenchymal cell incorporation in the noninjured lung or liver (fig. S1, A and B). In contrast, AEC2s and hepatocytes integrated into the injured lung or liver after the third Bleo, Acid, or CCl4 injection (Fig. 1, B and D).

Fig. 1. EC-produced HGF promotes reconstitution of transplanted parenchymal cells in the injured lung and liver in mice.

(A) Schema illustrating the strategy to test the incorporation of transplanted alveolar epithelial progenitor in normal and injured lungs. TdTomato-expressing AEC2s (red) were instilled into the recipient lungs via the trachea. To induce lung repair, we subjected mice to multiple intratracheal injections of Acid or Bleo. (B) Immunostaining of SFTPC performed to visualize endogenous (TdTomatoSFTPC+; indicated by arrowheads in the inset) and grafted (TdTomato+SFTPC+; labeled with arrows in the inset) AEC2s in mice after three Bleo or Acid injections. Result of AEC2 transplantation in normal mouse lungs is shown in fig. S1A. DAPI, 4′,6-diamidino-2-phenylindole. (C) Approach to examine the incorporation of hepatocytes in normal and injured mouse livers. Hepatocytes were transplanted to recipient mice via intrasplenic injection of TdTomato+ hepatocytes, and sections were costained with the hepatocyte marker hepatic nuclear factor 4α (HNF4). (D) Immunostaining showing the incorporation of transplanted HNF4+TdTomato+ hepatocytes in the liver after three injections of CCl4. Incorporation of hepatocytes transplanted after the eighth CCl4 and data showing hepatocytes transplanted into normal mice are presented in fig. S1 (B and C). (E) Schema illustrating the approach to test organ regeneration, fibrosis, and the incorporation of parenchymal cells in mice with EC-specific deletion of Hgf (HgfiΔEC/iΔEC). HgfiΔEC/iΔEC and control HgfiΔEC/+ mice were subjected to lung and liver injury and transplanted with TdTomato+ AEC2s or hepatocytes. Regeneration, fibrosis, and the incorporation of injected parenchymal cells (engraftment) were compared between HgfiΔEC/iΔEC and control mice. (F) Survival rate of HgfiΔEC/iΔEC and HgfiΔEC/+ mice after Bleo injection every 12 days. (G) Sirius red staining of lung tissue from mice in (F). (H) Survival rate of HgfiΔEC/iΔEC and HgfiΔEC/+ mice after CCl4 injection every 3 days. (I) Sirius red staining of liver tissue from mice in (H). (J) Sirius red staining depicting collagen deposition and immunostaining illustrating transplanted AEC2s (red) in the injured mouse lungs. (K) Sirius red staining and immunostaining showing transplanted hepatocytes (red) in the injured mouse livers. Five mice per group were analyzed for qualitative staining experiments. Scale bars, 50 μm.

Injured lung and liver either undergo fibrosis-free repair or develop fibrosis. A single Bleo or Acid injection triggers re-epithelialization in the injured lung without triggering fibrosis, but fibrosis starts to occur after repetitive injury (15). Similarly, CCl4 injection stimulates hepatic regeneration after three injections, but the injured liver ceases to resolve fibrosis after multiple CCl4 injections (16). Parenchymal cells failed to incorporate into the liver and lung after more than six injections of Bleo, Acid, or CCl4 (Fig. 1, B and D, and fig. S1C), the stage at which the injured liver and lung developed fibrosis (fig. S1, D and E).

EC-produced HGF mitigates expression of profibrotic NOX4 protein in perivascular fibroblasts

We then defined the microenvironment-derived cues that foster the incorporation of transplanted hepatocytes and AEC2s. Vascular ECs lining liver sinusoids or pulmonary capillaries were shown to elicit parenchymal regeneration (2, 16, 52, 79, 80). EC-produced HGF stimulates proliferation of parenchymal cells for organ repair (16, 48, 8083). Thus, we tested whether endothelial HGF influences parenchymal cell engraftment. Mice expressing EC-specific vascular endothelial (VE)–cadherin–driven CreERT2 [Cdh5-(PAC)-CreERT2] (84) were bred with floxed Hgf mice (Fig. 1E). Mice were injected with tamoxifen to induce EC-specific ablation of Hgf (HgfiΔEC/iΔEC), and EC-specific Hgf heterozygous knockout (HgfiΔEC/+) mice were used as controls (fig. S2A).

Lung and liver repair were analyzed in HgfiΔEC/iΔEC and control mice after the third Bleo or CCl4 injection (Fig. 1E). Compared with control mice, there was increased lethality in HgfiΔEC/iΔEC mice after liver or lung injury, which was associated with elevated tissue destruction and increased fibrosis in HgfiΔEC/iΔEC liver and lung (Fig. 1, F to I, and fig. S2, B and C). Consequently, incorporation of transplanted parenchymal cells was suppressed in the injured HgfiΔEC/iΔEC lung or liver, as compared with that of controls (Fig. 1, J and K). Hence, EC-produced HGF promotes organ repair, resolves fibrosis, and facilitates the engraftment of parenchymal cells in the injured lung and liver.

HgfiΔEC/iΔEC mice were then characterized in another liver regeneration model induced by PH. Qualitative immunostaining analysis suggested that hepatocyte proliferation after PH was lower in HgfiΔEC/iΔEC mice than in controls (Fig. 2A), and collagen deposition and cell apoptosis were elevated in the liver of hepatectomized HgfiΔEC/iΔEC mice compared to controls (Fig. 2, B and C). Immunostaining and enzyme-linked immunosorbent assay (ELISA) analysis of malondialdehyde (MDA) revealed markedly higher peroxide formation in HgfiΔEC/iΔEC mouse liver after PH (Fig. 2, D and E). Hepatectomized HgfiΔEC/iΔEC mice also exhibited increased lethality and liver damage than controls, as evidenced by higher serum bilirubin concentration (fig. S2, D and E). Thus, HGF expressed by ECs stimulates parenchymal cell expansion and resolves fibrosis during PH-induced liver regeneration.

Fig. 2. HGF from ECs reduces NOX4 expression in perivascular fibroblasts to bypass fibrosis during liver regeneration.

(A to E) Liver sections from HgfiΔEC/+ (control) and HgfiΔEC/iΔEC mice after PH showing (A) Ki67 staining for proliferation; (B) Sirius red staining for collagen; (C) immunostaining for terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick end labeling (TUNEL), VE-cadherin, and HNF4; (D) MDA staining for peroxide formation; and (E) MDA quantity. n = 7 HgfiΔEC/iΔEC mice and 8 control mice. (F and G) mRNA expression of NOX family members in liver tissue from HgfiΔEC/iΔEC and control mice (F) 10 days after PH and (G) 12 days after CCl4 injections. n = 10 control and 11 HgfiΔEC/iΔEC mice in PH group and 12 mice in CCl4 group. (H) Western blot and quantification of NOX4 protein in liver tissue from HgfiΔEC/iΔEC and control mice after PH. n = 8 mice per group. (I) Immunostaining of the fibroblast marker desmin, VE-cadherin, and NOX4 in liver sections from mice 10 days after PH. Insets show colocalization of NOX4 with desmin+ fibroblasts adjacent to VE-cadherin+ liver ECs. (J and K) Western blot and quantification of NOX4 protein in liver tissue from HgfiΔEC/iΔEC and HgfiΔEC/+ (control) mice at day 14 after BDL. n = 8 mice per group. (L and M) Quantity (L) and immunostaining (M) of MDA in liver tissue from HgfiΔEC/iΔEC and controls. n = 10 HgfiΔEC/iΔEC mice and 12 HgfiΔEC/+ mice. (N and O) Sirius red (N) and TUNEL (O) staining of the liver sections from (L). Statistical difference between two experimental groups was determined by two-tailed t test. Scale bars, 50 μm.

Increased fibrosis and oxidative damage in HgfiΔEC/iΔEC mice led us to identify the influence of endothelial HGF on perivascular fibroblasts. HGF was shown to attenuate the expression of the NOX family of proteins that regulate redox balance (85), and elevated NOX activity after injury stimulates fibrosis (56, 76). After PH and three CCl4 injections, NOX4 expression was increased in the liver of HgfiΔEC/iΔEC mice compared with that of controls (Fig. 2, F to H). Furthermore, NOX4 protein was found to be preferentially distributed in perivascular fibroblasts of hepatectomized HgfiΔEC/iΔEC mice (Fig. 2I). Hence, development of fibrosis in HgfiΔEC/iΔEC mice in liver regeneration is associated with NOX4 induction in perivascular fibroblasts.

To establish the correlation between endothelial HGF and perivascular NOX4 in fibrogenesis, we adopted a liver-specific gene transduction system (86). Bolus injection of a large volume of gene material via tail vein caused gene transduction in the mouse liver (fig. S2F). This system allowed us to study the contribution of NOX4 to liver regeneration in HgfiΔEC/iΔEC mice. Silencing NOX4 expression in hepatectomized or CCl4-injured HgfiΔEC/iΔEC mice promoted liver regeneration and blocked fibrosis (fig. S2, G to J). Thus, endothelial HGF stimulates fibrosis-free liver repair at least partially by suppressing NOX4 up-regulation in perivascular fibroblasts.

We then sought to define the hepatic-protective effect of endothelial HGF in a liver cholestasis model, bile duct ligation (BDL). The common bile duct was ligated and resected to induce biliary epithelial injury. BDL caused perivascular NOX4 protein up-regulation and stimulated qualitatively higher degrees of cell apoptosis, peroxide formation, and collagen deposition in HgfiΔEC/iΔEC livers than controls (Fig. 2, J to O, and fig. S2K). Hence, loss of HGF expression in EC might stimulate perivascular fibroblast activation in cirrhotic liver.

The clinical relevance of the HGF-NOX4 axis was investigated in samples of liver from human cirrhotic patients. Immunostaining of NOX4 and desmin showed that the extent of fibrosis positively correlated with NOX4 protein expression in the cirrhotic livers (Fig. 3A). NOX4 was qualitatively up-regulated in perivascular desmin+ fibroblasts adjacent to VE-cadherin+ ECs (Fig. 3, B to D, and fig. S3, A to H). Thus, NOX4 up-regulation in perivascular fibroblasts of injured HgfiΔEC/iΔEC liver was reminiscent of aberrant expression of perivascular NOX4 in human cirrhotic livers.

Fig. 3. EC-expressed HGF abrogates profibrotic NOX4 induction in human perivascular fibroblasts.

(A) Correlation between expression of perivascular NOX4 and fibrosis score in individual human patients. A fibrosis score of 2, 4, or 10 was used, with a higher score denoting more severe fibrosis and larger fibrotic area in the observed slide. Each dot in the plot represents an individual patient. (B to D) Representative human liver section immunostaining images from (A). Insets show higher magnification. (E and F) Western blot and quantification of NOX4 protein in human LX-2 cells and mouse stellate cells treated with TGF-β (20 ng/ml) ± HGF (40 ng/ml). Representative blot image is shown in (E), and each lane represents one tested biological sample. n = 6 samples for each group. Statistical difference was determined by one-way analysis of variance (ANOVA) followed by Tukey’s post hoc test. (G and H) Representative immunofluorescence image of LX-2 cells cultured with human ECs on Matrigel. (I and J) Western blot and quantification of NOX4 protein in LX-2 cells incubated with human ECs. n = 6 samples per group. Statistical difference between experimental groups was calculated by two-tailed t test. Scale bars, 50 μm.

Because transforming growth factor–β (TGF-β) stimulates NOX4 expression in fibroblasts (56, 76), we investigated whether endothelial HGF influences NOX4 expression in fibroblasts in the presence of TGF-β. Human and mouse hepatic stellate cells were treated with TGF-β with or without HGF. HGF ameliorated NOX4 expression and activity in human and mouse stellate cells after TGF-β stimulation (Fig. 3, E and F, and fig. S3, I and J). An endothelial-stellate cell coculture system was used to explore the cross-talk between endothelial HGF and fibroblast NOX4 (Fig. 3G). NOX4 protein expression was lower in stellate cells cultured with ECs overexpressing HGF (ECHGF) than in those with control ECs with scrambled sequence (ECSrb) (Fig. 3, H to J).

Targeting vascular and perivascular cells improves the engraftment of mouse and human parenchymal cells in fibrotic livers

Building on the finding that endothelial HGF promotes liver regeneration, we postulated that ectopic expression of HGF in the liver ECs (8789) could enhance hepatocyte engraftment in the injured liver. Pseudotyped virus system was used to conjugate with immunoglobulin antibody at the viral surface (90). Coupling of pseudotyped virus with antibody recognizing EC antigen CD31 allows for gene transfer in ECs (15). Mouse CD31 antibody (Mec13.3) was conjugated with virus encoding green fluorescent protein (Gfp), scrambled sequence (Srb), or Hgf, resulting in Mec13-Gfp, Mec13-Srb, and Mec13-Hgf viruses, respectively. Intrasplenic injection was used to localize Mec13-virus in the hepatic vascular bed. Compared to control (Mec13-Srb), Mec13-Hgf attenuated perivascular NOX4 expression in BDL-injured liver (Fig. 4, A and B, and fig. S4A). Subsequently, we tested whether combining Mec13-Hgf with the NOX4 inhibitor GKT137831 (GKT) would more efficiently provoke regeneration in the injured liver. GKT was administered to injured mice together with Mec13-Hgf (Mec13-Hgf + GKT). Mec13-Hgf + GKT substantially lowered peroxide formation and hydroxyproline amounts after BDL, more than any other tested treatments (Fig. 4, C to E). Moreover, Mec13-Hgf + GKT efficiently promoted the incorporation of grafted mouse hepatocytes (Fig. 4F) and induced the most efficacious hepatic repair in all tested approaches (Fig. 4, G to I).

Fig. 4. Expression of HGF in liver ECs cooperates with NOX4 inhibition to enhance engraftment of regenerative hepatocytes.

(A and B) Immunostaining (A) and Western blot (B) of NOX4 protein in the injured liver after injection of pseudotyped virus. Antibody Mec13.3 recognizing CD31 was conjugated with virus expressing Hgf (Mec13-Hgf) or scrambled sequence (Mec13-Srb). Virus was injected into the portal circulation. n = 8 mice in Mec13-Hgf group and 10 mice in Mec13-Srb group. (C to E) Liver hydroxyproline and MDA amounts in injured mice treated with Mec13-Hgf, NOX4 inhibitor GKT, or coinjection of Mec13-Hgf and GKT (Mec13-Hgf + GKT). n = 8, 10, 10, and 9 animals in individual groups from left to right in (C) and (D). Statistical analysis was performed with one-way ANOVA followed by Tukey’s post hoc test. (F) Incorporation of transplanted TdTomato+ hepatocytes in the liver tissue of mice 21 days after BDL. (G to I) Serum concentrations of aspartate aminotransferase (AST) (G), alanine aminotransferase (ALT) (H), and bilirubin (I) in indicated mouse groups after BDL. n = 12, 16, 14, 16, 15, and 12 animals in individual groups from left to right in (G) to (I). Statistical difference was determined by one-way ANOVA followed by Tukey’s post hoc test. Scale bars, 50 μm.

We then tested how dual niche-editing influences reconstitution of human hepatocytes (Fig. 5A). Immunodeficient NOD (nonobese diabetic)–Prkdcem26Cd52Il2rgem26Cd22/Nju (NCG) mice treated with Mec13-Srb, Mec13-Hgf, GKT, or Mec13-Hgf + GKT after BDL. Human hepatocytes were transplanted into the injured mice via intrasplenic injection. Mec13-Hgf + GKT enhanced the incorporation of GFP-labeled human hepatocytes in the damaged liver, and these grafted hepatocytes maintained the expression of hepatocyte marker cholesterol 7α-hydroxylase (CYP7A1) (Fig. 5B). NCG mice treated with Mec13-Hgf + GKT + human hepatocytes showed reduced cell death in the liver, regenerated hepatic architecture and function, and higher serum human albumin concentration than all other test groups (Fig. 5, C to G, and fig. S4B). Therefore, dual editing of vascular and perivascular cells bypasses fibrosis to enable the engraftment of mouse and human parenchymal cells, promoting hepatic regeneration (Fig. 5H).

Fig. 5. Dual editing of endothelial and perivascular cells promotes the incorporation of regenerative human hepatocytes in the liver.

(A) Schema depicting the strategy to transplant human hepatocytes into NCG mice. (B) Distribution of grafted GFP+ human hepatocytes in the recipient mouse liver after indicated treatments. GFP was costained with cholesterol 7α-hydroxylase (CYP7A1) in the liver sections from the recipient injured mice. (C and D) Cell apoptosis in NCG mice with indicated treatments and transplanted with hepatocytes. TUNEL was costained with HNF4 and VE-cadherin. n = 5 mice per group. (E to G) Hepatic pathology and quantity of serum human albumin and bilirubin in recipient mice after BDL. n = 5 mice per group. (H) Dual editing of vascular and perivascular niches by endothelial Hgf gene delivery and NOX4 inhibition facilitates parenchymal cell engraftment and promotes liver repair. Statistical analysis was carried out with one-way ANOVA followed by Tukey’s post hoc test. Scale bars, 50 μm.

Endothelial HGF suppresses profibrotic expression of NOX4 in perivascular lung fibroblasts

To explore the generality of the observed regenerative effect of endothelial HGF, we tested HgfiΔEC/iΔEC mice in a lung alveolar regeneration model triggered by PNX. PNX stimulates functional growth of the intact right lung lobe after surgical resection of the left lung lobe (Fig. 6A). Compared to the controls, HgfiΔEC/iΔEC mice had increased peroxide formation and NOX4 protein expression after PNX (Fig. 6, B and C), which was accompanied by inhibited restoration of lung mass and function, and qualitatively elevated cell apoptosis (fig. S5, A to D). Notably, NOX4 was qualitatively up-regulated in perivascular fibroblasts in pneumonectomized HgfiΔEC/iΔEC mice (Fig. 6D). Hence, endothelial HGF has a similarly important role in driving lung alveolar regeneration and obviating fibrosis after PNX.

Fig. 6. EC-expressed HGF stimulates lung alveolar regeneration and resolves fibrosis.

(A) Schema illustrating the strategy to test lung alveolar regrowth after PNX. (B to D) Lung MDA quantity (B), Western blot and quantification (C), and immunostaining (D) of NOX4 protein the lung of HgfiΔEC/iΔEC and control mice after PNX. n = 10 HgfiΔEC/iΔEC and 16 controls. Insets show the distribution of NOX4 protein in the lungs. Statistical difference between experimental groups was assessed by two-tailed t test. (E and F) Western blot (E) and quantification (F) of NOX4 protein in human and mouse lung fibroblasts treated with TGF-β (20 ng/ml) ± HGF (40 ng/ml). n = 6 samples per group. Statistical difference was determined by one-way ANOVA followed by Tukey’s post hoc test. (G) Western blot of NOX4 protein in human lung fibroblasts incubated with ECHGF or ECSrb. (H) Representative immunofluorescence image of human lung fibroblasts (green) cocultured with human ECHGF or ECSrb (red) and treated with TGF-β (10 ng/ml). (I) Correlation between perivascular NOX4 protein quantity and lung fibrosis score in human patients. Each dot in the plot represents data from an individual patient. (J to L) Representative immunostaining image of samples from (I). Insets show the higher magnification. Scale bars, 50 μm.

To investigate the relationship between endothelial HGF and NOX4 in lung regeneration, we silenced NOX4 in injured mouse lungs (56). NOX4 short hairpin RNA was administered intratracheally into the pneumonectomized HgfiΔEC/iΔEC mice (fig. S5E). This method blocked the up-regulation of NOX4 in perivascular fibroblasts, prevented fibrosis, and restored alveolar function in pneumonectomized HgfiΔEC/iΔEC lungs (fig. S5, F to I). Notably, HGF blocked the up-regulation of NOX4 protein in cultured human and mouse lung fibroblasts after TGF-β treatment (Fig. 6, E and F, and fig. S6, A and B), and incubating human lung fibroblasts with ECHGF abrogated NOX4 protein induction by TGF-β (Fig. 6, G and H, and fig. S6C). The degree of NOX4 up-regulation in perivascular fibroblasts correlated with fibrosis grade in human fibrotic lung tissue samples (Fig. 6I), and perivascular NOX4 induction in HgfiΔEC/iΔEC mouse lungs after PNX recapitulated NOX4 expression in perivascular fibroblasts of human fibrotic lungs (Fig. 6, J to L, and fig. S6, D to K). Therefore, endothelial HGF suppresses expression of NOX4 in perivascular lung fibroblasts, which may contribute to fibrogenesis in diseased human lungs.

Targeting dysfunctional vascular and perivascular cells enhances engraftment of AEC2s in fibrotic lungs

We investigated the therapeutic effect of Mec13-Hgf and GKT after repeated Bleo or Acid injections using mouse models. Jugular vein injection of Mec13-Hgf (15) in mice inhibited NOX4 up-regulation in injured lung fibroblasts (Fig. 7, A and B). Qualitative Sirius red staining suggested that Mec13-Hgf + GKT reduced fibrosis in the injured lungs (Fig. 7C). Moreover, Mec13-Hgf + GKT enhanced the incorporation of AEC2s in damaged lungs, which stimulated lung regeneration more efficiently than any other treatments (Fig. 7, D to F). These data suggest that editing dysfunctional vascular and perivascular cells can subvert an epithelially prohibitive microenvironment to an epithelially active niche, enabling regenerative therapy for fibrosis (fig. S7A).

Fig. 7. Dual editing of vascular and perivascular niches facilitates the incorporation of AEC2s, stimulating fibrosis-free lung repair.

(A and B) Western blot (A) and quantification (B) of NOX4 protein amount in mouse lungs after Bleo injection and Mec13-Hgf or Mec13-Srb treatment. n = 8 mice per group. Statistical difference between two experimental groups was determined by two-tailed t test. (C) Sirius red staining of lung sections from mice injected with Bleo and treated with Mec13-Srb, Mec13-Hgf, GKT, or Mec13-Hgf + GKT. (D and E) Blood oxygenation in Bleo- or Acid-injured mice after treatment of Mec13-Srb, Mec13-Hgf, GKT, or Mec13-Hgf + GKT. AEC2s were transplanted to depicted groups. n = 9, 7, 11, 8, 9, and 7 animals in individual groups from left to right in (D). n = 10, 8, 9, 7, 8, and 8 mice in individual groups from left to right shown in (E). Statistical analysis was carried out with one-way ANOVA followed by Tukey’s post hoc test. (F) Localization of transplanted TdTomato+ AEC2s (red) in lung tissues from indicated mouse groups. Five mice per group were assessed for qualitative staining analysis. Scale bars, 50 μm.

We also tested the efficacy of Mec13-Hgf + GKT in promoting reconstitution of human AEC2s. NCG mice were subjected to either Bleo or Acid injury and transplanted with GFP-labeled human AEC2s after Mec13-Srb, Mec13-Hgf, GKT, or Mec13-Hgf + GKT treatment (Fig. 8A). Mec13-Hgf + GKT efficiently promoted human AEC2 incorporation in the recipient mouse lungs (Fig. 8B). This engraftment was accompanied by blunted cell death, restored alveolar architecture, and recovered gas exchange function (Fig. 8, C to E, and fig. S7B).

Fig. 8. Reconstitution of regenerative human AEC2s in the injured lungs after treatment with Mec13-Hgf and GKT.

(A) Schema describing the approach to graft human AEC2s into NCG mice. Bleo- or Acid-injured NCG mice were treated with Mec13-Srb, Mec13-Hgf, GKT, or Mec13-Hgf + GKT and transplanted with human AEC2s. (B) Localization of transplanted GFP+ human AEC2s in the injured lungs from indicated mouse groups. GFP was costained with SFTPC. Four mice per group were analyzed in qualitative staining analysis. (C and D) Histological and cell apoptosis analysis of lung sections of mice with the indicated injury and treatments. (E and F) Quantification of lung cell apoptosis and blood oxygenation in the injured mice with described treatments. Sr, Mec13-Srb; Hg, Mec13-Hgf. n = 5 mice in all groups except Mec13-Srb group transplanted with AEC2s (Sr group; n = 4). Statistical analysis was performed with one-way ANOVA followed by Tukey’s post hoc test. Scale bars, 50 μm.


The microenvironment of the recipient organ, the “soil,” can regulate the efficiency of cell engraftment. Establishing an epithelially inductive microenvironment in diseased organs might facilitate the engraftment of regenerative parenchymal cells. Here, we find that vascular and perivascular niche cells jointly regulate the incorporation of transplanted alveolar progenitor cells or hepatocytes in damaged organs. Parenchymal cell incorporation in injured lungs and livers requires expression of HGF in the endothelial niche and suppression of perivascular NOX4 activity. Moreover, incorporated parenchymal cells synergize with targeted niches to stimulate the most efficacious organ repair. Thus, reinstating the self-repair capacity of diseased lung and liver might require the cooperation between the proregenerative microenvironment and grafted parenchymal stem/progenitor cells.

Fibrosis in injured organs might hamper engraftment of parenchymal cells. Therefore, understanding how fibrosis is resolved during organ regeneration is helpful for devising cellular therapy for various diseases (1, 5). HGF has an important role in promoting parenchymal cell expansion (83, 9194). Here, we found that EC-expressed HGF contributes to resolving lung and liver fibrosis, which is at least partially mediated by suppressing NOX4 expression in perivascular fibroblasts. In PH and PNX models that induce parenchymal regeneration without fibrosis, deletion of Hgf in mouse EC blocked regeneration and caused fibrosis. Reciprocally, in mouse liver and lung fibrosis models, ectopic expression of Hgf in EC prevented activation of perivascular fibroblast and suppressed fibrosis. These genetic model–based in vivo and in vitro cultivation data implicate that in addition to stimulating parenchymal reconstitution, endothelial HGF also serves as a key molecule to prevent fibrosis.

Therapeutically, NOX4 inhibition synergized with endothelial HGF induction to stimulate efficacious parenchymal reconstitution and regeneration. This synergistic effect may be due to several layers of mechanisms. First, reduced NOX4 in perivascular cells by endothelial HGF might require less NOX4 inhibitor to reach sufficient inhibition. Second, reducing oxidative stress in the damaged organs by the NOX4 inhibitor may promote production or activation of HGF from the vascular niche. The additive effects by endothelial HGF and NOX4 inhibition might also rely on operative mechanisms independent of the HGF-NOX4 axis. Endothelial HGF can exert various regenerative functions (9194), and this dual-editing approach might use both NOX4-dependent and NOX4-independnet mechanisms. NOX4 inhibition prevents parenchymal cell death after injury (95), offering hepatogenic or alveologenic effects that extend beyond HGF signaling. Hence, our dual-editing approach might exploit various functions of HGF and NOX4 that are independent of the HGF-NOX4 axis.

HGF expression triggered in the vascular niche after parenchymal injury may act as a protective response. The vast surface area of hepatic or pulmonary microvasculature is ideal to deploy enormous amounts of paracrine growth factors within a short time period (2, 75, 9698). Several endothelial receptors were demonstrated to mediate HGF up-regulation in ECs, including vascular endothelial growth factor receptor 1 (VEGFR1) (48), VEGFR2 (80), and CXCR7 (16). Conceivably, these endothelial receptors sense tissue mass loss or chemical injury to trigger HGF expression. Loss of these pathways might not only impair regeneration but also cause fibrosis. It was shown before that activation of Smad2/3 blocks HGF transcription in keratinocytes (99). Whether the TGF/Smad pathway mediates the subversion of epithelially inductive vascular niche function during parenchymal repair can be explored in the future.

The translational value of the dual niche-editing system could be improved through alternate gene transfer methods or coupling to different inhibitors or therapeutics. Pseudotyped lentivirus accomplishes EC-selective gene transfer at the expense of low efficiency; a therapeutically applicable EC gene-editing approach might require a clinically tested gene transfer system with higher efficiency. Engineered adeno-associated virus (AAV) can be chemically conjugated with an endothelial targeting moiety via processes such as biotin-avidin interaction (87, 89). Administration of EC-targeted AAV might offer more efficacious and safer vascular gene editing in fibrotic organs. Other therapeutics can be exploited to target the dysfunctional perivascular niche in conjugation with a NOX4 inhibitor.

Together, we show that vascular and perivascular cells jointly influence the engraftment of parenchymal progenitor cells in the injured lungs and livers. Implementing cell transplantation with niche editing efficiently induces organ regeneration. Our proof-of-principle evidence may help develop cell therapy approaches to enable fibrosis-free repair in various organs.


Study design

We combined vascular Hgf gene expression with NOX4 inhibition via GKT to edit the vascular and perivascular niches in fibrotic lung and liver. The contribution of vascular and perivascular niches was demonstrated by suppressed parenchymal cell engraftment in host mice lacking endothelial HGF and overexpressing perivascular NOX4. The role of endothelial HGF in promoting parenchymal regeneration and bypassing fibrosis in liver and lung was established using (i) EC-specific knockout (HgfiΔEC/iΔEC) mice and (ii) complementary lung and liver regeneration and repair models. Both AEC2s and hepatocytes were grafted into the recipient mice after treatment with EC-targeted Hgf, GKT, or combination therapy. We analyzed the extent of the incorporation of transplanted mouse and human parenchymal cells in the recipient mice and tested the efficacy of functional organ repair in liver and lung injury models. Investigators who performed mouse experiments and analyzed the pattern and extent of cell distribution and tissue pathology were randomly assigned with animal or human samples. Investigators were blinded to the phenotypes of samples from different groups. For quantitative experiments consisting of n < 20 animals per group, individual subject-level data are shown in table S1.

Genetic mouse models

The AEC2-specific Sftpc-CreERT2 mouse line was provided by B. Hogan (Duke University), and hepatocyte-specific Albumin-Cre mouse and ROSA26-TdTomato reporter mouse expressing TdTomato after floxed stop codon were purchased from The Jackson Laboratory. Sftpc-CreERT2 and Albumin-Cre mice were bred with ROSA26-TdTomato mice, and Sftpc-CreERT2 expression in the offspring mouse was induced by six consecutive intraperitoneal injections of tamoxifen (100 mg/kg) (Sigma-Aldrich). Expression of Cre induced excision of the stop codon preceding TdTomato and triggered expression of red fluorescent protein in AEC2s and hepatocytes. Floxed Hgf mice were obtained from mutant mouse regional resource centers (no. 000423). Mice expressing EC-specific Cdh5-(PAC)-CreERT2 (84) were provided by R. Adams. This mouse line was crossed with floxed Hgf mice to generate HgfiΔEC/iΔEC mice and control HgfiΔEC/+ mice after treatment of tamoxifen at a dose of 150 mg/kg for 6 days and interrupted for 3 days after the third dose. Deletion of target genes in ECs was corroborated by quantitative polymerase chain reaction. Six- to 10-week-old sex- and weight-matched HgfiΔEC/iΔEC and HgfiΔEC/+ mice and male wild-type mice were used for mouse PH, PNX, and liver and lung injury models. Wild-type mice were purchased from the Jackson Laboratory and Chengdu Dossy Biological Technology Company. Six- to 10-week-old immunodeficient NCG mice from Nanjing Biomedical Research Institute of Nanjing University were used for transplantation of human hepatocytes and AEC2s. All animal experiments were carried out by protocols approved by the Institutional Animal Care and Use Committee at Sichuan University and Weill Cornell Medicine.

Liver and lung injury models

Injections of CCl4 were used to induce liver injury as previously described (16). CCl4 was diluted in olive oil (Sigma-Aldrich) to yield a concentration of 40% (0.64 g/ml) and intraperitoneally injected to mice at a dose of 1.6 g/kg. Mice were subjected to BDL to induce cirrhotic liver injury (16). To perform BDL, we anesthetized mice with a cocktail of ketamine (100 mg/kg) and xylazine (10 mg/kg). Ketamine and xylazine were provided by an approved veterinary service at Weill Cornell Medicine and Sichuan University (China). Repetitive intratracheal Bleo or Acid injection models were used to induce lung injury (100). At the described time points, oxygen tension in arterial blood of treated mice was measured using i-Stat (Abbott Laboratories).

Transplantation of mouse and human parenchymal cells

AEC2s and hepatocytes expressing TdTomato were isolated as described (21, 51, 80). To isolate AEC2s, we removed the mouse lungs and digested them in a cocktail solution containing collagenase A (2 mg/ml) and dispase (1 mg/ml) (Roche Life Science) in Hanks’ balanced salt solution. One milliliter of the digestion solution was directly instilled via the trachea and used to perfuse via pulmonary arteries to accelerate the digestion process. Perfused mouse lungs were removed from the chest cavity, transferred to the EC growth medium (Sigma-Aldrich), minced, and disrupted. Lungs were then suspended in the digestion cocktail (2.5 ml/100 mg) at 37°C for 15 min. Digested lungs were filtered through a 40-μm nylon mesh (cell strainer) and centrifuged. Cell suspension was centrifuged via Percoll gradient. The AEC2 band was harvested, and TdTomato+ AEC2s were sorted by flow cytometry. Human AEC2s were purchased from iCell Bioscience Inc.

For mouse AEC2 transplantation, 3 million isolated mouse AEC2s were infused into mouse lungs via the trachea 3 days after the third or sixth Bleo or Acid treatment. Mice were also treated with Mec13-Srb, Mec13-Hgf, GKT, or Mec13-Hgf + GKT after the third Bleo or Acid injection, and 3 million mouse AEC2s were transplanted after the fifth Bleo or sixth Acid injection. For human AEC2 transplantation, 10 million human AEC2s were labeled with GFP and transplanted into immunodeficient NCG mice after the indicated treatments. Recipient mice were sacrificed 10 days after transplantation to determine the incorporation of transplanted cells in the injured lungs. Blood oxygen tension was analyzed 20 days after the last Bleo or Acid injury.

Hepatocytes were purified as described via a two-step perfusion and digestion procedure (51, 80). Human hepatocytes were obtained from BioreclamationIVT. For mouse hepatocyte transplantation, 2.5 million isolated mouse hepatocytes were transplanted to the injured mouse liver via intrasplenic injection (80) 3 days after the third or eight injection of CCl4. Intrasplenic injection was also performed as described (80). Ten million human hepatocytes were transplanted into NCG mice at day 10 after BDL injury. Recipient mice were sacrificed at day 21 after BDL to assess the extent of hepatocyte incorporation in the injured liver.

Mouse liver and lung injury models

A mouse PH model was used to induce liver regeneration (80). Left lung PNX model and measurement of alveolar regeneration (79, 98) and lung injury (15) were adapted. Animal experiment procedures are described in Supplementary Materials and Methods. Liver and lung fibrotic responses were determined at the indicated times after injection of Bleo, Acid, or CCl4, or after BDL. Collagen distribution was assessed by Sirius red staining (15). Hydroxyproline amount was quantified in the liver and lung to determine the extent of fibrosis (15, 55). To measure peroxide formation in the tissue, we performed immunostaining and ELISA of MDA using antibody against MDA (Abcam) and a lipid peroxidation assay kit (Abcam). Generation of H2O2 in cultured stellate cells and lung fibroblasts was performed as described (76).

Culture of hepatic stellate cells and lung fibroblasts

For coculture experiments with human ECs, 200 μl of Matrigel (BD Biosciences) was incubated in a 24-well plate, and the fluorescently labeled human stellate cell line LX-2 or lung fibroblasts and mCherry-labeled ECs were seeded at 1:1 ratio on Matrigel (BD Biosciences). Formation of three-dimensional vascular tubes was imaged by fluorescence microscopy. The human stellate cell line LX-2 was provided by S. Friedman (Mount Sinai Hospital), and the mouse hepatic stellate cell line was obtained from ScienCell Research Laboratories. Human and mouse lung fibroblasts were purchased from Science Cell Inc. Human ECs were obtained from Angiocrine Bioscience. Cultured stellate and fibroblast cells were treated with recombinant TGF-β (20 ng/ml) and HGF (40 ng/ml) (PeproTech Inc.) and retrieved for NOX4 protein analysis by Western blot.

NOX4 inhibitor GKT and EC-specific delivery of Hgf

Pseudotyped viral particles containing Hgf were conjugated with Mec13.3 antibody to induce HGF expression in EC (Mec13-Hgf). Virus containing Srb construct was similarly processed with Mec13.3 as a control group (Mec13-Srb). After the third Bleo or Acid injection or BDL, 6- to 8-week-old mice were subjected to Mec13-Hgf or Mec13-Srb every 3 days at a dose of 75 μg of p24 capsid protein. The effect of GKT was tested in lung and liver fibrosis models. GKT was dissolved in an aqueous solution (0.5% carboxymethylcellulose and 0.25% Tween 20). GKT (10 mg/kg) was given to mice twice weekly by oral gavage (started together with Mec13-Hgf) (56, 76). The effect of combination treatment was compared with vehicle, Mec13-Hgf, or GKT alone.

Immunostaining and histological analysis of mouse and human cryosections

Mouse liver and lung tissues were harvested for histological analysis (16, 80). Human sample sources are described in the Supplementary Materials. For immunofluorescence microscopy, the liver sections (10 μm) were blocked (5% donkey serum/0.3% Triton X-100) and incubated in primary antibodies: anti–VE-cadherin polyclonal antibody (pAb) (2 μg/ml; R&D Systems), anti-NOX4 (pAb, 5 μg/ml; Abcam), anti-MDA antibody (5 μg/ml; Ab6463, Abcam), and anti-desmin (pAb, 2 μg/ml; Abcam). After incubation in fluorophore-conjugated secondary antibodies (2.5 μg/ml; Jackson ImmunoResearch), sections were counterstained with DAPI (Invitrogen). Five sections were analyzed for each animal.

Image acquisition and analysis

Histology analysis and Sirius red staining of liver or lung sections were captured with an Olympus BX51 microscope (Olympus America). Densitometry analysis of the Western blot image was performed with ImageJ software using calibrated standard curve of optical density, and fluorescence images were recorded on an AxioVert LSM710 confocal microscope (Zeiss).

Statistical analysis

All data were presented as means ± SEM. For statistical analysis of experiments where there are more than two treated groups, ANOVA was performed, followed by Tukey’s post hoc test. Comparison of statistical difference between two experimental groups was determined by two-tailed t test.


Materials and Methods

Fig. S1. Characterization of parenchymal cell incorporation in the mouse lungs and livers.

Fig. S2. EC-expressed HGF is required for bypassing fibrosis and stimulating regeneration after liver and lung injury.

Fig. S3. Expression of NOX4 in liver fibroblasts.

Fig. S4. Dual editing of vascular and perivascular niches promotes liver repair.

Fig. S5. EC-expressed HGF suppresses perivascular NOX4 expression to stimulate lung alveolar regeneration and attenuate fibrosis.

Fig. S6. Expression of NOX4 in lung fibroblasts.

Fig. S7. Dual editing of vascular and perivascular cells blocks fibrosis and facilitates lung repair.

Table S1. Individual subject-level data.


  1. Acknowledgments: We are grateful to S. L. Friedman (Mount Sinai Hospital, New York, NY) for providing the human hepatic stellate cell line LX-2. We thank I. S. Y. Chen (University of California, Los Angeles, CA) for the pseudotype virus vector. We are indebted to R, H. Adams (Max Planck Institute, Germany) and B. L. M. Hogan (Duke University, Durham, NC) for EC-specific Cdh5-(PAC)-CreERT2 mice and AEC2-specific Sftpc-CreERT2 mouse line. Funding: This work was supported by the National Scientist Development Grant from the American Heart Association (12SDG1213004), National Heart, Lung, and Blood Institute (R01HL097797, R01HL119872, and R01HL130826), National Natural Science Foundation of China (91639117), and National Key Research and Development Program focused on Stem Cell and Translational Research (2016YFA0101600). S.R. is supported by the Empire State Stem Cell Board and New York State Department of Health grants (C024180, C026438, C026878, and C028117). Author contributions: Z.C. conceived the project, performed the experiments, analyzed the results, and wrote the paper. T.Y., Y.S., and G.J. designed and carried out the experiments, analyzed the data, and edited the manuscript. K.S., Y.C., L.L., F.N., X.L., and Z.H. performed the experiments. J.L.K., V.M., L.Q., and C.C. analyzed the data. F.J.M. commented on the study and edited the manuscript. S.R. commented on the study. B.-S.D. designed and carried out the experiments, interpreted the results, and wrote the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All materials are contained within the article and the Supplementary Materials.
View Abstract

Navigate This Article