Ductular reaction promotes intrahepatic angiogenesis through Slit2–Roundabout 1 signaling

Human biopsies and samples

Liver tissue samples were obtained from fragments of nontumoral cirrhotic liver tissue surrounding colon metastasis collected at the time of liver resection or from explants from liver transplantation due to ARLD, NASH, and HCV. The study was approved by the Ethics Committee of the Hospital Clinic of Barcelona, and all patients included in this study provided written informed consent.

Patient cohorts

Animals

Because complete or partial genetic deletion of Robo1 is embryonically lethal, we explored the role of the Slit2–Robo1 pathway in mice with partial genetic deletion of both Robo1 and Robo2 which are viable. ROBO1^−/+ROBO2^−/+ (also referred to as ROBO1/2^−/+) mice with a BALB/c background were kindly supplied by Dr. Jian-Guo Geng from the University of Michigan. Littermates of Robo1/2^−/+ mice were used in all studies. Details on chronic liver injury and regeneration mouse models are described in the Supporting Information. All animal experiments were approved by the Ethics Committee of Animal Experimentation of the University of Barcelona and conducted in accordance with the National Institutes of Health’s Guide for the Care and Use of Laboratory Animals.

Gene ontology analysis

A microarray analysis of yellow fluorescent protein–positive (YFP⁺) cells isolated from HNF1βCreER^YFP and wild-type (WT) mice receiving a 3,5-diethoxycarbonyl-1,4-dihydrocollidine (DDC) or a choline-deficient, ethionine-supplemented (CDE) diet for 3 weeks (n = 3 in all groups) was performed previously.^[23] In order to assess new biological processes related to DR, we performed a functional analysis using the transcriptomic data generated and deposited in the National Center for Biotechnology Information’s Gene Expression Omnibus (GSE51389). A description of the Gene Ontology (GO) analysis (Geneontology.org) performed is detailed in the Supporting Information.

Liver tissue and organoid analysis

Details on immunostaining of liver tissue and organoids as well as gene expression, western blot, and hydroxyproline analysis of liver tissue samples are included in the Supporting Information.

In vitro angiogenesis assay

Growth factor–reduced Matrigel (BD Biosciences, San Jose, CA) was completely thawed at 4℃ for 16–24 h before conducting the tube formation assay. Liquid Matrigel was added to a previously chilled 24-well plate (300 μl/well). The plate was then incubated for 1 h at 37℃ to allow Matrigel solidification. Human umbilical cord endothelial cells (HUVECs; 60,000 cells/well) were resuspended in basal medium (DMEM-F12, 1% Glutamax, 1% 4-(2-hydroxyethyl)-1-piperazine ethanesulfonic acid [Hepes], and 1% penicillin–streptomycin) or basal medium with recombinant SLIT2 protein (2 ng/ml). To further explore the role of organoid-derived SLIT2 in promoting angiogenesis, we preincubated HUVECs for 2 h with or without 10 mg/ml of sheep antihuman ROBO1 antibody (R&D Systems) in basal medium. Following this, HUVECs were stimulated with conditioned medium with and without αROBO1 antibody (10 mg/ml). In all conditions, HUVECs were incubated for 12 h at 37℃, 5% CO₂. Then, the medium was removed, and the endotubes were rinsed with PBS and fixed with 4% paraformaldehyde for 15 min. Tubules were visualized by phase-contrast microscopy, and six images per well at ×4 were taken. Tube formation was evaluated using the Angiogenesis Analyzer tool (ImageJ software).

Liver organoids generated from cirrhotic liver tissues

Liver organoids were generated from cirrhotic liver tissue samples and cultured following the protocol previously described with minor modifications.^[8] Briefly, liver tissue was digested using Collagenase XI (Sigma-Aldrich) and Dispase II (ThermoFisher) for 30 min at 37℃. After washing and erythrocyte lysis steps, pelleted cells were embedded in basement matrix extract (BME; Ambio, Abingdon, UK) and seeded in 24-well plates. Following BME solidification, organoid expansion medium was added. Organoid expansion medium consisted of basal medium (Advanced DMEM/F12 [ThermoFisher], 1% Glutamax [ThermoFisher], 1% Hepes [Life Technologies, Carlsbad, CA], and 1% penicillin–streptomycin [Lonza, Basilea, Switzerland]) supplemented with % N2 and 1% B7 without vitamin A (both from Life Technologies); 1.25 mM N-acetylcysteine, 10 mM nicotinamide, and 10 nM gastrin (Sigma-Aldrich); 50 ng/ml EGF, 100 ng/ml FGF-10, 25 ng/ml HGF, 25 ng/ml Noggin, and 500 ng/ml Rspo1 (Peprotech, London, UK); and 5 μM A8301, 0.5 μM CHIR99021, 10 μM forskolin, and 10 μM of Y27632 (Axon Medchem, Groningen, the Netherlands). Y27632 was added only the first 3 days after isolation. The expansion medium was replaced every 2–3 days, and cells were split once a week. At a confluence of 70%–80%, organoid expansion medium was replaced by basal medium for 18 h. Organoid conditioned medium was collected and stored at −80℃ until the HUVECs tubulogenic assay and ELISA SLIT2 quantification were performed.

Serum SLIT2 detection

SLIT2 serum levels from healthy individuals and patients with AH as well as from supernatants of liver organoids generated from patients with cirrhosis were evaluated with a specific ELISA kit for SLIT2 (Cusabio Biotech Co., Wuhan, China) following the manufacturer's protocol.

Statistical analysis

Values are expressed as mean ± SEM. Statistical differences between groups were analyzed using the Student t test, two-way ANOVA with Bonferroni correction, or the Mann-Whitney U test when appropriate with GraphPad Prism 5.0 (San Diego, CA). p ≤ 0.05 was considered statistically significant. Correlations between variables were evaluated using Spearman's rho or Pearson's r, when appropriate.

Mouse DR cells exhibit enriched expression of genes related to angiogenesis

In order to identify biological functions associated with DR cells, we evaluated transcriptomic data obtained from biliary cells, isolated as YFP⁺ cells by fluorescence-activated cell sorting, from HNF1βCreER^YFP mice. DR cells were isolated from mice fed with a diet enriched in DDC, a CDE diet, or a chow diet for 3 weeks and from control mice. GO analysis revealed that the top five enriched biological processes associated with DR cells were inflammatory response (gene ratio 37/315, p = 1 × 10⁻⁹), followed by regulation of cell adhesion (gene ratio 31/315, p = 8 × 10⁻⁹), angiogenesis (gene ratio 25/315, p = 9 × 10⁻⁸), tube morphogenesis (gene ratio 33/315, p = 3.7 × 10⁻⁷), and blood vessel morphogenesis (gene ratio 26/315, p = 1.18 × 10⁻⁶) (Figure 1A). Because very little is known regarding the capacity of DR cells to participate in angiogenesis, we decided to focus on this key repair mechanism. In order to investigate the mechanism by which DR contributes to liver angiogenesis, we examined the molecules underlying the GO terms related to angiogenesis (Figure 1B). Slit2, a well-described proangiogenic factor, was the gene included in all the categories. Moreover, Slit2 was found to be significantly up-regulated in YFP⁺ cells isolated from the DDC and CDE models compared to those sorted from control mice (fold change [FC] = 2.2, p = 0.004 and FC = 7.48, p = 0.002, respectively) (Figure 1C).

Slit2 is expressed by DR cells and Robo1 by vessels in DDC mice

Slit2 acts as a regulator of new vessel formation through interactions with either Robo1 or Robo4 receptor in the context of cancer and ischemic diseases.^[16-19] In order to assess the role of Slit2 in intrahepatic angiogenesis, we first examined the levels of Slit2 expression and its receptors in liver tissue from DDC and CDE mouse models. Slit2 was found to be up-regulated in DDC and CDE compared to control livers (FC = 44.8 ± 4.6 and FC = 8.2 ± 3.3, respectively) at similar levels to the DR markers epithelial cell adhesion molecule (Epcam) and keratin 19 (Krt19). We also observed a significant up-regulation of Robo1 in both models (FC = 3.9 ± 0.15 and FC = 2.25 ± 0.3, respectively), whereas Robo4 was not found to be differentially expressed (Figure 2A). In order to confirm that the expression of Slit2 was restricted to DR cells, we analyzed the gene expression levels of Slit2 in both YFP⁺ isolated cells and a hepatocyte fraction from DDC mice using quantitative real-time PCR (qPCR). While Slit2 was found to be significantly enriched in DR cells compared to total liver tissue, no expression was found in the hepatocyte fraction (Figure 2B). As expected, we did not find Robo1 expression in either of the examined cell fractions. Next, by performing immunohistochemistry in DDC livers, we confirmed that SLIT2 was expressed by the ductular structures (KRT19-positive cells), whereas its receptor, ROBO1, was expressed by the endothelial cells comprising the new vessels located near the periportal areas (Figure 2C). It is important to note that Robo1 expression was not detected in endothelial cells of the portal vein or hepatic artery, indicating that its expression is specific to the neoangiogenic endothelial cells. These results indicate that, in a DDC mouse model, DR structures express the proangiogenic factor SLIT2, which might interact with ROBO1 receptors expressed by the neo-vessels.

ROBO1/2^−/+ mice display reduced hepatic angiogenesis compared to WT mice in response to DDC-induced injury

In order to determine the functional role of Slit2–Robo1 signaling in the pathogenesis of chronic liver injury, we used Robo1/2 partial knockout mice (ROBO1/2^−/+) and WT littermates with DDC and CDE diets, both involving DR expansion, for 3 and 4 weeks, respectively. Immunohistochemistry of DR key markers such as KRT19 and EpCAM revealed that ROBO1/2^−/+ mice fed the DDC or CDE diet did not display a reduction in DR structures compared to WT mice, indicating that Slit2–Robo1 signaling is not directly involved in ductular cell proliferation (Figure 3A,B). In line with this result, we did not find significant differences in the hepatic expression of genes related to DR such as Krt19, Epcam, or SRY (sex determining region Y)-box 9 (Sox9) (Figures S1A and S2A). In addition, we observed that fibrosis assessed by Sirius red staining; by gene expression of actin alpha 2 smooth muscle, collagen type I alpha 1 chain, matrix metalloproteinase 2, and tissue inhibitor of metalloproteinase 1; and by liver hydroxyproline level determination was not altered in ROBO1/2^−/+ mice treated with DDC or CDE when compared to WT mice (Figures 3A,B, S1B, S2B and S3). Indeed, although a profibrogenic role of Slit2–Robo1 signaling has been previously reported in a mouse model of advanced fibrosis based on chronic CCl₄ administration, it has been attributed to Slit2 expression by HSCs. Interestingly, quantification of immunohistochemistry staining for CD31 showed a marked reduction of vessels surrounding periportal areas in ROBO1/2^−/+ compared to WT mice in both injury models. However, we did not observe a major gene expression reduction of key markers related to angiogenesis (Figures 3, S1C and S2C). Moreover, we did not observe any differences in liver enzymes in serum of ROBO1/2^−/+ fed the DDC or CDE diet compared to WT (Figure S4A,B). These results demonstrate that Slit2–Robo1 signaling regulates hepatic angiogenesis in both models of chronic liver disease involving DR expansion.

Slit2–Robo1 signaling does not participate in modulating physiological angiogenesis in response to partial hepatectomy

After partial hepatectomy, angiogenesis and angiocrine growth factors play a key role in the correct regeneration of the liver.^[25] Because Slit2–Robo1 has a role in regulating hepatic angiogenesis in chronic liver damage, we assessed whether this signaling pathway participates in the regeneration mechanisms of a healthy liver. With this objective, we first performed two-thirds partial hepatectomy in C57BL/6J mice. We assessed the hepatic gene expression of Slit2 and Robo1 together with DR markers (Epcam, Krt19, Sox9) at different stages of liver regeneration after partial hepatectomy, considering early (0, 24, 48, and 24 h) and late (Days 7 and 28) time points after surgery. Interestingly, we found increased Slit2 expression at 48 h, 72 h, and Day 7 after partial hepatectomy compared to intact liver at 0 h. Besides, liver expression of Robo1 was significantly up-regulated at 72 h after the surgery (Figure 4A). Next, to assess the potential role of Slit2–Robo1 signaling in physiologic angiogenesis, we performed partial hepatectomy in ROBO1/2^−/+ mice and WT littermates. We evaluated liver regeneration at 48 h and Day 7 after partial hepatectomy. We did not find significant differences in the number of proliferating cells within the parenchyma, assessed by immunohistochemistry of KI67 between ROBO1/2^−/+ and WT mice at early (48 h) and late (7 days) time points after surgery (Figures 4B and S6B). Likewise, liver regeneration index, calculated as liver weight/body weight ratio, did not change either (Figures S5A and S6A). Regarding liver progenitor cell expansion, we found a down-regulation of Epcam and Krt19 gene expression in ROBO1/2^−/+ mice at early, but not at late, time points after partial hepatectomy when compared to WT mice. However, we did not see any differences in DR expansion as assessed by immunohistochemistry of KRT19 at 48 and 7 days after surgery (Figures 4B,C, S5B and S6C), suggesting that liver regeneration is not affected by Robo1 deficiency in partial hepatectomy. To specifically investigate whether angiogenesis was impaired in ROBO1/2^−/+ mice, we quantified CD31-stained vessels within the liver tissue. We did not observe differences in the formation of new vessels at early or late time points after partial hepatectomy, indicating that Slit2–Robo1 signaling does not significantly contribute to physiological angiogenesis (Figure 4B,C). These results suggest that, although there is an activation of the Slit2–Robo1 pathway after partial hepatectomy, other mechanisms, possibly mediated by VEGF, are more relevant in driving angiogenesis during hepatic regeneration of healthy liver.

DR and the Slit2–Robo1 pathway are associated with ARLD progression

DR is a histological key feature of AH and increases along ARLD progression. Given the proangiogenic role of SLIT2 in mouse models involving DR expansion, we hypothesized that DR contributes to tissue repair along ARLD progression by promoting angiogenesis through Slit2–Robo1 signaling. In order to assess this hypothesis, we first showed that neoangiogenesis was taking place near the DR structures in liver tissues from patients with cirrhosis and AH by performing a double immunohistochemistry of von Willebrand factor (VWF) and KRT7 (Figure 5A). Next, we examined the levels of SLIT2 and ROBO1 expression as well as the gene expression profile of gene sets related to angiogenesis and DR in RNA sequencing data from total liver tissue from a cohort of patients encompassing the whole spectrum of ARLD.^[22] The patients included in this cohort were grouped according to ARLD stage: severe AH, patients with AH and MELD ≥ 21 (n = 11); nonsevere AH, patients with AH and MELD < 21 (n = 18); patients with compensated cirrhosis (n = 9); and healthy individuals (n = 10). Interestingly, the expression profile of the angiogenesis gene set markedly increased with ARLD progression, suggesting that intrahepatic angiogenesis is a relevant regeneration mechanism underlying ARLD. As expected, the DR gene signature increased in ARLD progression, and we found that it followed a similar gene expression pattern as angiogenesis. In addition, SLIT2 and ROBO1 expression displayed a similar trend to DR and angiogenesis sets, showing significant up-regulation in severe compared to nonsevere AH (Figure 5B). Next, we sought to determine if DR and angiogenesis were linked biological processes in patients with chronic liver diseases of different etiologies. For this analysis we included a group of patients with nonadvanced chronic liver disease with the following etiologies: ARLD with histologic criteria of steatohepatitis (n = 12), NAFLD (n = 10), and HCV-infected patients without cirrhosis (n = 10). The hepatic expression of DR markers (KRT7 and KRT19) strongly correlated with the expression of angiogenesis markers such as platelet/endothelial cell adhesion molecule and VWF (Figure 5C) as well as with SLIT2 and its receptor ROBO1 (Figure 5D). This result suggests that DR expansion and angiogenesis are activated not only along ARLD progression but also in chronic liver diseases of other etiologies. Moreover, our findings indicate that the Slit2–Robo1 pathway is associated with fundamental repair processes in chronic liver disease such as DR and angiogenesis. In addition, the expression of the angiogenic signaling factor VEGFA and the kinase insert domain receptor and Fms-related receptor tyrosine kinase 1 did not correlate with the hepatic expression of DR markers (Figure S7), suggesting that in chronic liver disease DR-mediated angiogenesis may not be related to VEGFA signaling.

The Slit2–Robo1 pathway is enhanced in patients with AH

To confirm the association of DR with the Slit2–Robo1 pathway, we evaluated the hepatic expression of SLIT2 and its receptor ROBO1 in a cohort of patients with AH (n = 29) and healthy individuals (n = 5). The characteristics of the patients are described in Table S1. SLIT2 and ROBO1 expression was increased in patients with AH compared to controls (FC = 3.4 ± 0.4, and FC = 2.35 ± 019, respectively) (Figure 6). As expected, the hepatic expression of SLIT2 significantly correlated with the expression of its receptor, ROBO1. Interestingly, in the same cohort of patients with AH, SLIT2 expression strongly correlated with the hepatic expression of KRT7, suggesting that DR cells are a cell source of SLIT2 in patients with AH (Figure 6A). Next, we investigated the serum levels of SLIT2 in a subset of patients with AH (n = 16) and healthy individuals (n = 6). We observed that circulating SLIT2 levels were significantly increased in patients with AH compared to healthy individuals (9.8 ± 0.9 versus 3.4 ± 1.1 ng/ml) (Figure 6B). Additionally, circulating levels of SLIT2 positively correlated with serum levels of aspartate aminotransferase and alkaline phosphatase, suggesting that SLIT2 increases together with liver injury in AH.

DR cells from cirrhotic livers secrete SLIT2 and induce tubulogenesis

We recently reported that liver organoids derived from cirrhotic liver tissue mimic DR cells and represent a relevant in vitro model to study DR.^[8] In order to confirm that human DR cells secrete SLIT2, we generated liver biliary organoids from patients with cirrhosis of different etiologies (NASH, n = 1; ARLD, n = 2; and HCV, n = 2). The cirrhotic tissues used to generate organoids expressed higher levels of SLIT2 and ROBO1 proteins compared to control livers (Figure 6C). We also confirmed the presence of DR in all the cirrhotic tissue samples examined by KRT7 immunohistochemistry (Figure S8A). As expected, we detected the SLIT2 protein in all the supernatants of the organoids generated (17 ± 2.8 ng/ml) (Table S2 and Figure S8B), and we observed SLIT2 expression by immunofluorescence in the liver organoid (Figure S8C). Next, with the purpose of demonstrating the proangiogenic potential of DR cells, we performed a tubulogenic assay by exposing HUVECs to conditioned medium derived from cirrhotic liver organoids (n = 3). As a positive control, we incubated HUVECs with recombinant SLIT2 protein (2 ng/ml). Compared to the basal medium, liver organoid conditioned medium significantly increased angiogenesis, as assessed by the quantification of the number of junctions formed by HUVECs. To further demonstrate that Slit2 secreted by DR cells mediates angiogenesis, we treated HUVECs with conditioned medium containing αROBO1 blocking antibody. Interestingly, we observed that the proangiogenic ability of DR cells was significantly attenuated after ROBO1 blockade (Figures 6D and S8D). Altogether, these results suggest that DR cells release the proangiogenic factor SLIT2, thus contributing to angiogenesis.