Identification of Ephrin type-B receptor 4 as a critical mediator of tissue fibrosis

Pathological overactivation of tissue repair facilitates fibrosis, causing irreversible tissue damage that can lead to organ failure and patient mortality (1). Pulmonary fibrosis (PF) is characterized by excessive extracellular matrix (ECM) deposition that severely impairs oxygen exchange and deteriorates health. Idiopathic PF (IPF) and systemic sclerosis–related interstitial lung disease (SSc-ILD) are 2 diseases where PF is a primary pathology causing patient mortality (2). Advancements in PF research have led to the approval of growth factor receptor inhibitors nintedanib and pirfenidone for the treatment of IPF, along with the subsequent repurposing of nintedanib for the treatment of progressing fibrotic ILDs (3–5). Both nintedanib and pirfenidone inhibit multiple receptors in epidermal and connective tissue cells (fibroblasts), which has accelerated the investigation of epithelial and connective tissue in fibrogenesis (3–5). Despite significant therapeutic advancements, current interventions neither halt nor reverse PF. Deciphering disease mechanisms underlying fibrosis will help uncover novel therapeutic avenues to improve patient outcomes.

Erythropoietin-producing hepatocellular receptor interacting protein (Ephrin) and Eph receptors are unique protein families that undergo bidirectional signaling, where ligand-receptor interactions initiate downstream signaling from both the receptor (forward signaling) and ligand (reverse signaling) (6). The Ephrin-B2 ligand and EphB4 receptor signaling axis mediates angiogenesis and vasculogenesis in prenatal development by regulating cell migration via actin cytoskeletal remodeling pathways (6). Further evidence suggests that Ephrin-B2 may play a key role in pulmonary, cardiac, skin, and hepatic fibrosis (7–9). Release of the soluble Ephrin-B2 active ectodomain into lung tissue, modulated by a disintegrin and metalloproteinase domain-containing protein 10 (ADAM10), was shown to promote fibrosis (7). Furthermore, siRNA targeting of EphB4 abrogates soluble Ephrin-B2–mediated fibrotic responses in pulmonary fibroblasts in vitro (7). While EphB4 was proposed as a target of Ephrin-B2–mediated profibrotic responses in vitro, to the best of our knowledge, the exact role of EphB4 signaling in PF is yet to be investigated.

To study the role of EphB4 receptor in PF, we used a preclinical PF mouse model, total RNA-Seq and snRNA-Seq, NanoString, previously published single cell data, integrative computational biology analysis, and functional assays using primary mouse and human lung fibroblasts (Figure 1). We generated mice with conditional deletion of Ephb4 in Col1a2-expressing cells, including fibroblasts, to show that Ephb4 deletion reduced the severity of bleomycin-induced PF. This protective antifibrotic phenotype was replicated using the pharmacological EphB4 inhibitor NVP-BHG712. High-throughput RNA-Seq of lung fibroblasts derived from bleomycin-treated mice showed that Ephb4 deletion induced downregulation of key genes associated with ECM regulation, protein trafficking, and endoplasmic reticulum (ER) cargo concentration. NanoString analysis on mouse lung fibroblasts treated with EphB4 inhibitor also indicated a reduction in the expression of key ECM, protein trafficking, and ER cargo genes. snRNA-Seq identified 4 fibroblast subclusters in the lungs of bleomycin-treated mice. Elastin (ELN) expression was higher in identified fibroblast subclusters 1 and 2, and ELN expression was reduced with EphB4 inhibition under bleomycin condition. Total RNA-Seq identified ELN as the most differentially expressed ECM gene in primary human lung fibroblasts isolated from patients with IPF, which also exhibited significant downregulation in response to EphB4 inhibition. Integrated computational analysis of mouse and human fibroblast total RNA-Seq data identified ELN as a crucial EphB4 signaling target. TGF-β1–induced ELN expression was reduced by inhibition or silencing of EphB4 in human IPF fibroblasts. Similarly, ELN was reduced in bleomycin-treated mouse lungs following pharmacological inhibition or genetic Ephb4 deletion. This study suggests what we believe to be a novel mechanism by which EphB4 signaling contributes to PF.

Figure 1

Methodological approaches used to assess the role of EphB4 in pulmonary fibrosis. A schematic representation of the methods and experimental approach used in this study.

Genetic deletion of Ephb4 in Col1a2-expressing cells to generate Ephb4 conditional knockout mice. As part of the pathology of PF, fibroblasts are responsible for the deposition of excessive ECM components, like type I collagen. To examine whether Ephb4 expression in Col1a2-expressing cells (such as fibroblasts) contributes to PF development, floxed Ephb4 (Ephb4^tm1.1Jwu) mice were crossed with B6.Cg-Tg(Col1a2-Cre/ERT,-ALPP)7Cpd mice to generate Ephb4^f/f;Col1a2-Cre^ERT mice (detailed pedigree for the generation of the mouse line is included in Supplemental Figure 1; supplemental material available online with this article; https://doi.org/10.1172/jci.insight.189156DS1). Col1a2-driven expression of Cre in mice has been used in previous studies to ablate specific genes in fibroblasts (7, 10, 11). PCR genotyping was performed to confirm the presence of Col1a2-Cre^ERT in mice exhibiting homozygosity of the floxed Ephb4 sites (500 bp) compared with the WT Ephb4 gene (400 bp) (Figure 2A). At 4–5 weeks of age, Ephb4^f/f;Col1a2-Cre^ERT mice were treated with 4-hydroxytamoxifen (1 mg/day for 5 days) to induce Ephb4 gene deletion in Col1a2-expressing cells, including fibroblasts, or with corn oil (vehicle). Sustained knockdown of Ephb4 expression was confirmed at both mRNA and protein levels in primary lung fibroblasts isolated 7 days after the final 4-hydroxytamoxifen administration (Figure 2B). No significant reduction in the expression of EphB4 was observed in other cell types examined, such as macrophages and endothelial cells (Supplemental Figure 2, A and B). Additionally, cell culturing of primary lung fibroblasts to passage 1 (P1) yielded high purity of fibroblast populations (Supplemental Figure 2, C and D). Thus, Ephb4 conditional KO (CKO) mice permit 4-hydroxytamoxifen–inducible deletion of Ephb4 in Col1a2-expressing cells (herein referred to as Ephb4-CKO mice), including fibroblasts, compared with control mice treated with corn oil (Ephb4 control mice).

Figure 2

Ephb4 CKO mice exhibit a protective phenotype against the development of pulmonary fibrosis. Ephb4^f/f mice were crossed with Col1a2-Cre^ERT mice to generate a line with conditional Col1a2-driven deletion of Ephb4 (Ephb4 CKO). (A) Genotyping was performed to confirm the presence of homozygous floxed Ephb4 alleles and Cre. (B) Mice treated with 4-hydroxytamoxifen to induce the CKO were sacrificed and primary lung fibroblasts were cultured to first passage. RNA was collected and Ephb4 was quantified by qPCR (n = 6–7). Protein lysates were collected from cell cultures and EphB4 protein was quantified by Western blotting (n = 5). (C) A mouse model for inducing fibrosis in Ephb4-CKO mice was performed using intratracheal bleomycin challenge or PBS (control) following 4-hydroxytamoxifen-induced deletion. (D) Mouse lungs were collected at end point (9 weeks of age) and stained using Gomori’s One Step collagen stain (blue). Representative histology images from 2 separate animals/group are shown. Fibrosis was quantified either by modified Ashcroft scoring or with ImageJ total lung collagen content quantification (n = 8 mice/group). (E) αSMA immunostaining was performed in lung tissue sections to assess fibrotic activity in Ephb4-CKO and control mice with and without bleomycin challenge (n = 5). Semiquantitative scoring (scores 0–3) of the staining intensity was assigned to each mouse lung. Scale bar: 400 µM; original magnification, x 10. (F) Primary lung fibroblasts were isolated from Ephb4 Col1a2-Cre^ERT mice treated with 4-hydroxytamoxifen or corn oil and then bleomycin. RNA was isolated from fibroblasts, and qPCR was performed to detect Col1a1 and Acta2 (n = 4). For qPCR (B and F), 2^–ΔCT values were used to represent relative expression of the gene of interest normalized to Gapdh. For all graphs, data are expressed as mean ± SD where differences (*P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001) were identified using either 2-tailed Student’s t test (B) or 2-way ANOVA with Tukey’s post hoc tests (D–F).

Ephb4-CKO mice are protected against the development of PF. Ephb4-CKO and Ephb4 control mice were subjected to intratracheal instillation of bleomycin to induce PF, with PBS used as a corresponding bleomycin control (Figure 2C). Mouse lungs were collected 14 days later to assess collagen content using Gomori’s One Step Trichrome blue collagen stain. As shown in Figure 2D, collagen content was comparable in both Ephb4-CKO and Ephb4 control lungs treated with PBS. After bleomycin challenge, Ephb4-expressing control mice showed a striking increase in collagen staining, indicating development of PF. Interestingly, this effect was not readily observed in Ephb4-CKO mice treated with bleomycin. Instead, deletion of Ephb4 resulted in a marked decrease in PF severity and reduced the number of lobes exhibiting fibrosis, as determined by Ashcroft scoring and collagen staining quantification (Figure 2D and Supplemental Table 1) (12). Next, IHC staining of αSMA (gene name Acta2), a marker of activated fibroblasts associated with fibrosis (13), was examined. As expected, PBS control treatment yielded minimal αSMA staining in mouse lungs of Ephb4-CKO and control mice. Bleomycin treatment induced strong αSMA staining in Ephb4 control mice with substantially weaker staining in Ephb4-CKO mice (Figure 2E). Furthermore, periostin staining was also reduced in the lungs of Ephb4-CKO compared with control mice, indicating a protective phenotype in mouse lungs with Ephb4 ablation (Supplemental Figure 3A). Intriguingly, the antifibrotic phenomenon of Ephb4 deletion in lung fibroblasts after bleomycin challenge was maintained in vitro. Primary lung fibroblasts derived from bleomycin-treated Ephb4-CKO mice also exhibited reduced αSMA Acta2 and type I collagen (Col1a1) gene expression compared with bleomycin-treated EphB4 control mice (Figure 2F). Taken together, it appears that Ephb4 in lung fibroblasts plays an important role in PF development.

We next wanted to determine if Ephb4 CKO had a similarly altered response of fibroblasts in another form of fibrosis. To this aim, we examined a skin fibrosis model. We first confirmed knockdown of Ephb4 expression at both mRNA and protein levels in skin fibroblasts from Ephb4-CKO and Ephb4 control mice (Supplemental Figure 4A). After Ephb4 knockdown, mice received daily s.c. bleomycin injections (0.5 U/mL) for 21 days (Supplemental Figure 4B). Skin samples were then collected 22 days after Ephb4 knockdown and subjected to histological Gomori’s trichrome staining. No differences were observed in fibrosis severity and dermal thickening (Supplemental Figure 4C). Thus, the antifibrotic effect of Ephb4 CKO observed in PF was not observed in the bleomycin-induced model of skin fibrosis, suggesting that the effects of Ephb4 ablation in fibroblasts may depend on the type of tissue/organ and the tissue microenvironment.

Total RNA-Seq identifies downregulation of ECM, ER cargo concentration, and protein trafficking–related transcriptome in lung fibroblasts from Ephb4-CKO mice. To elucidate the effects of Ephb4 ablation on the lung fibroblast transcriptome in the context of PF, total RNA-Seq was performed on lung fibroblasts isolated from Ephb4 CKO (or Ephb4 control mice) 2 weeks after bleomycin (or PBS) challenge (Figure 3A). First, the effects of bleomycin on lung fibroblast transcriptome were examined by comparing Ephb4 control mice (bleomycin versus PBS) 2 weeks after bleomycin treatment. A total of 257 differentially expressed genes (DEGs; false discovery rate [FDR] q < 0.05) were identified (155 upregulated and 102 downregulated) in bleomycin-treated Ephb4 control mice compared with PBS-treated Ephb4 control mice. No DEGs were identified in lung fibroblasts from Ephb4 control and Ephb4-CKO mice after PBS treatment. We next examined the effect of Ephb4 ablation on bleomycin-induced gene expression in lung fibroblasts. When Ephb4-CKO and Ephb4 control mice challenged with bleomycin were compared, 317 DEGs were identified (103 upregulated and 214 downregulated) in lung fibroblasts from bleomycin-challenged Ephb4-CKO mice (Figure 3, B and C), suggesting that the effects of Ephb4 ablation in lung fibroblasts are unique to fibrotic conditions. The list of top DEGs identified can be found in Supplemental Table 2 and the full sequencing dataset can be accessed online on the GEO data repository (GSE205612).

Figure 3

Ephb4-CKO fibrotic mouse lung fibroblasts exhibit unique transcriptomic profiles. (A) Ephb4^f/f;Col1a2-Cre^ERT mice were treated with 4-hydroxytamoxifen (1 mg/day for 5 days) or corn oil control at 4 weeks of age and were then intratracheally instilled with PBS or bleomycin (1.2 U/kg) at 6 weeks of age. Mice were sacrificed 14 days after bleomycin or PBS challenge, and lung fibroblasts were isolated and cultured (n = 4 mice per group). Fibroblasts were enriched with a single passage followed by RNA extraction for total RNA-Seq. (B) Pairwise comparisons were made between bleomycin and PBS-treated Ephb4 control mice, or between Ephb4-CKO and control mice under bleomycin condition (n = 4 mice/group). Differentially expressed genes (DEGs) were determined using an FDR cutoff of q < 0.05 (red). In the bleomycin versus PBS comparison, 257 DEGs were identified in fibroblasts, including 155 upregulated and 102 downregulated DEGs. In the Ephb4-CKO versus control comparison, 317 DEGs were identified in fibroblasts, including 103 upregulated and 214 downregulated DEGs. (C) DEGs from both comparisons were visualized using heatmaps. (D) Lists of both upregulated and downregulated DEGs in the RNA-Seq dataset of Ephb4 CKO (n = 4) versus Ephb4 control mice (n = 4) treated with bleomycin were analyzed separately using pathDIP to identify subsets of genes enriched for unique pathways in the Reactome database. Gene-pathway associations were visualized in NAViGaTOR v.3.0.16. (E) Unique gene-pathway enrichments from the downregulated gene list visualized by dot plot, with significance determined using the Benjamini and Hochberg method (with cutoff of q < 0.01). (F) The expression of downregulated DEGs associated with ECM organization, protein transport, and ER cargo concentration visualized by Wideplot.

Next, we used pathDIP (14) to identify pathways enriched in the up- and downregulated DEG lists of Ephb4 CKO lung fibroblasts relative to their Ephb4 control counterparts under bleomycin challenge (Supplemental Table 3). A subset of 31 upregulated DEGs were uniquely enriched for 8 pathways, and a subset of 57 downregulated DEGs were uniquely enriched for 11 pathways. Upregulated DEGs were uniquely enriched for pathways associated with phospholipid and lipid metabolism, cytokine signaling, and IL signaling. Among the downregulated genes, 11 were associated with ECM organization, 7 with protein trafficking, 4 with ER cargo concentration, and 12 with class I–mediated antigen processing and presentation. Gene set enrichment analysis (GSEA) using both Reactome and Gene Ontology Biological Process databases showed that downregulated genes were also mainly enriched for ECM and collagen organization pathways, while upregulated genes showed enrichment for innate immune response and cytokine signaling (Supplemental Table 4). The unique enriched gene-pathway associations identified using pathDIP (14) were visualized in an interaction map using NAViGaTOR (15) (Figure 3D).

The list of gene-pathway associations was used to identify cellular processes that Ephb4 might regulate under fibrotic conditions. Unique DEGs downregulated in lung fibroblasts from Ephb4-CKO mice challenged with bleomycin were associated with ECM-relevant processes, including collagen formation and crosslinking, elastic fiber formation, fibronectin matrix formation, or ECM organization as well as genes related to protein trafficking and ER cargo concentration–related processes (Figure 3E). Interestingly, of the 11 downregulated DEGs associated with ECM organization identified, many have been documented as prominent profibrotic mediators, including: prolyl 4-hydroxylase beta polypeptide (P4bh), integrin b1 (Itgb1), integrin a5 (Itga5), peroxidasin (Pxdn), lysl oxidase like 2 (Loxl2), lysl oxidase (Lox), type XI collagen (Col11a1), tolloid like 1 (Tll1), and Eln (16–21). Among the ECM, ER cargo and protein trafficking genes, 20 of 22 exhibited a > 1.5-fold change in gene expression in bleomycin-challenged Ephb4 CKO animals compared with bleomycin-challenged Ephb4 control mice, with Eln having the greatest mean fold change (Figure 3F).

Overall, total RNA-Seq showed that, under fibrotic stimulation (bleomycin treatment), fibroblasts from Ephb4-CKO mice compared with Ephb4 control mice exhibit reduced expression of genes related to fibrosis, ECM organization, ECM remodeling, ER cargo concentration, and protein trafficking.

Pharmacological EphB4 inhibition reduces mouse lung fibroblast expression of genes related to ECM organization, ER cargo concentration, and protein trafficking. We next examined whether pharmacological inhibition of EphB4 elicited comparable changes in gene expression (as observed in total RNA-Seq) in primary mouse lung fibroblasts. WT primary lung fibroblasts were isolated from C57BL/6J mice and stimulated with TGF-β1 (or PBS as control) in combination with the EphB4 small molecule inhibitor NVP-BHG712 (or vehicle) (Figure 4A). NVP-BHG712 has previously been shown to decrease EphB4 activity and phosphorylation in vitro and in vivo, including in mouse lung tissue (22). We first confirmed that NVP-BHG712 effectively inhibits phosphorylation of EphB4 in human lung fibroblasts at 250 nM (Supplemental Figure 5) and used this concentration for subsequent experiments. A NanoString CodeSet of 47 gene targets was designed using our previously identified downregulated DEGs involved in ECM organization, protein trafficking, ER cargo concentration, MHC Class I presentation, and posttranslational protein modification (remaining genes were excluded, including those related to translation). TGF-β1 significantly increased the expression of 25 of the 47 selected genes, with EphB4 inhibition significantly reducing the expression of 19 of 25 of these genes (Supplemental Table 5). Notably, EphB4 inhibition downregulated many genes associated with cellular processes involved in ECM organization (9 of 11), ER cargo concentration (3 of 4), and protein trafficking (6 of 7) (Figure 4B and Supplemental Figure 6). Thus, pharmacological inhibition of EphB4 in mouse lung fibroblasts recapitulated reduced expression of most genes identified in our in vivo Ephb4-CKO mouse lung fibrosis model.

Figure 4

NVP-BHG712 downregulates TGF-β1–induced expression of genes related to ECM organization, protein trafficking, and ER cargo concentration regulation. (A) Lungs were isolated from WT mice, and fibroblasts were cultured. Fibroblasts were then simultaneously treated with the profibrotic cytokine TGF-β (20 ng/mL; or PBS as control) and the EphB4 inhibitor NVP-BHG712 (250 nM; or DMSO as inhibitor control) for 24 hours (n = 9). RNA was isolated and NanoString was performed on a 47 gene CodeSet to examine the expression of downregulated DEGs identified by RNA-Seq in bleomycin-challenged Ephb4-CKO lung fibroblasts. (B) Transcript levels of a subset of the CodeSet of genes that were significantly upregulated with TGF-β1 treatment and downregulated with NVP-BHG712 treatment and were associated with ECM organization, protein trafficking, and ER cargo concentration regulation (remaining genes can be found in Supplemental Figure 6). Gene expression was normalized to housekeeping genes Rplp0 and Acad9 and displayed as log₂ transformed gene expression (mean ± SD). Wilcoxon signed rank test with FDR q ≤ 0.05 was used to identify significant changes in gene expression. Legend colors indicate the gene-associated pathDIP-identified pathways. *q < 0.05.

NVP-BHG712 attenuates bleomycin-induced PF in mice. Since NVP-BHG712 attenuated the expression of key genes involved in ECM organization, ER cargo concentration, and protein trafficking in cultured mouse lung fibroblasts, we next tested its antifibrotic potential in vivo using the bleomycin lung fibrosis mouse model. Seven-week-old WT C57BL/6J mice were intratracheally challenged with bleomycin or PBS (control) and, 5 days later, received NVP-BHG712 (10 μg/kg daily, p.o.) or vehicle (control; N-methyl-2-pyrrolidone plus polyethylene glycol 300 at ratio of 1:9) for 5 days. Lungs were harvested for Gomori’s blue collagen stain histological assessment 14 days after the start of the bleomycin challenge (Figure 5A). As observed in Ephb4-CKO mice, WT mice treated with NVP-BHG712 also exhibited reduced severity in the degree of bleomycin-induced lung fibrosis (Figure 5B). Ashcroft lung fibrosis score of affected areas showed reduced fibrosis severity in NVP-BHG712–treated mice, with a mean ± SD disease severity score of 2.48 ± 0.55 compared with 4.05 ± 0.61 in bleomycin challenged, vehicle-treated lungs (Figure 5C). Histomorphometric analysis of lung section images indicated that lung collagen content in NVP-BHG712 treated lungs decreased from 49.7% ± 8.8% to 37.81% ± 6.8% (mean ± SD) per NVP-BHG712–treated animal (Figure 5C). Notably, there was no significant change in Ashcroft scoring or lung collagen quantification in NVP-BHG712–treated mice compared with vehicle-treated mice in the PBS-challenged groups. αSMA and periostin staining was also markedly reduced in NVP-BHG712–treated lungs compared with vehicle-treated controls after bleomycin challenge, with no difference found between PBS-treated controls (Figure 5D and Supplemental Figure 3B). Collectively, these results show that EphB4 pharmacological inhibition in mice confers protection in bleomycin-induced lung fibrosis, similar to the genetic deletion in Ephb4-CKO mice.

Figure 5

Pharmacological inhibition of EphB4 protects against development of PF in mice. (A) Bleomycin sulphate (1.2 U/kg) or PBS (control) was intratracheally administered to WT C57BL/6J mice. Five days later, mice were given either NVP-BHG712 (10 μg/kg; p.o.) or vehicle (control) daily for 5 days (n = 6/group). Lungs were harvested 14 days after bleomycin challenge, and histology was performed. (B) To assess fibrosis, collagen fibers in mouse lungs were stained via Gomori’s One Step Trichrome (blue). Representative histology images from 2 separate animals/group are shown. (C) Images of mouse lung sections (n = 5 animals/group; 5 lobes/animal) were graded for severity of fibrosis using the Ashcroft scoring system (top graph). Images of lung sections underwent color deconvolution in Fiji V. 2.9.0 to isolate the blue channel as a representative of total lung collagen. Total lung collagen was normalized to total lung tissue and expressed as percent total lung collagen. Graphical representations indicate averages of 5 lobes per animal (bottom graph). (D) αSMA immunohistochemistry performed on mouse lung sections (n = 5 animals/group). Semiquantitative scores were assigned scores from 0–3 based on degree of staining. Original magnification, x 10). For all graphs, data are expressed as mean ± SD and analyzed by 2-way ANOVA followed by Tukey’s post hoc tests for multiple comparisons. *P < 0.05; **P < 0.01; ****P < 0.0001.

Transcriptomic differences in fibroblast subclusters using snRNA-Seq of mouse lungs treated with NVP-BHG712 in the bleomycin model. Since NVP-BHG712 treatment resulted in reduced PF severity in vivo and reduced expression of ECM genes in mouse lung fibroblasts in vitro, we next conducted snRNA-Seq and bioinformatic analyses on bleomycin or PBS instilled mouse lungs treated with NVP-BHG712 or vehicle to identify lung fibroblast subsets and interrogate the differences in the ECM transcriptome of these fibroblast subsets in response to EphB4 inhibition. A total of 12 mouse lungs (n = 3 PBS + vehicle, n = 3 PBS + NVP-BHG712, n = 3 bleomycin + vehicle and n = 3 bleomycin + NVP-BHG712) were subjected to snRNA-Seq. In total, 49,163 nuclei were sequenced and, after filtering, 39,611 nuclei were analyzed. Multiple cell types, including fibroblasts, macrophages, endothelial cells, type I alveolar cells, and type II alveolar cells, were identified based on canonical cell markers (Figure 6, A and B). We next subclustered fibroblasts across all 12 samples, identifying 4 fibroblast subclusters (Figure 6C), which were present in all treatment conditions (Supplemental Figure 7A). A list of genes (q < 0.05 and log₂FC > 0.5) associated with each fibroblast subcluster can be found in Supplemental Table 6. Interestingly, the cellular proportion of fibroblast subcluster 2 increased with the treatment of bleomycin, whereas the proportion of subcluster 0 fibroblasts was decreased with bleomycin treatment. These changes in fibroblast subcluster proportions were ameliorated with NVP-BHG712 treatment. To better understand the cellular annotations of the 4 identified fibroblast subclusters in our dataset, we mapped our data to the markers identified in the previously published single cell studies (23, 24). Using the markers identified in these studies, we mapped the key transcriptomic profiles to our 4 fibroblast subclusters. Our analysis suggests that fibroblast subcluster 0 overlaps with the transcriptomic profiles associated with alveolar fibroblasts, subcluster 1 with adventitial fibroblasts, subcluster 2 with fibrotic fibroblasts, and subcluster 3 with peribronchial fibroblasts (Figure 6D). Furthermore, several genes previously reported to be associated with activated fibroblasts (25) mapped to fibroblast subcluster 2 in our dataset (Supplemental Figure 7B).

Figure 6

snRNA-Seq profiling of mouse lungs. (A) UMAP projection of distinct cell populations identified in mouse lung tissue. (B) Dot plot showing expression of canonical markers used to annotate the identified cell clusters. The identity of each cell cluster was assigned by matching the cluster expression profile with established cell-specific marker genes. Dot color represents average expression, and dot size represents the fraction of expressing cells. (C) UMAP projection of fibroblast subclusters from all treatment groups and the proportion of cells that contributed to each cluster by treatment group (veh, vehicle; Inh, NVP-BHG712). (D) Dot plot showing the expression of marker genes used to annotate the fibroblast subclusters (23–25). (E) Pathway analysis of upregulated and downregulated genes in bleomycin mice treated with NVP-BHG712 compared with vehicle. (F) Bar plots show total number of interactions and strength of interactions in fibroblasts in collagen signaling pathways in bleomycin mice treated with NVP-BHG712 (Inh) compared with vehicle; left panel). Circle plots showing cell-cell communication between fibroblast subclusters associated with collagen signaling in bleomycin mice treated with vehicle or NVP-BHG712. The plot shows interaction between ligands from sender and receptors from receiver. Each plot has fibroblast subcluster 0, 1, 2, and 3 as receiver and ligands from all other subclusters as senders. (G) Differential expression of 11 ECM genes in bleomycin-treated groups in all 4 fibroblast subclusters in NVP-BHG712 compared with vehicle (left panel). Violin plots showing the expression of Eln in fibroblast subclusters 1 and 2 (right panel).

Next, we performed pathway enrichment analysis using pathDIP (14) on up-and downregulated genes in response to EphB4 inhibition under bleomycin condition. Results show that both up- and downregulated genes were enriched with multiple pathways (Figure 6E). Notably, downregulated genes were associated with ECM-associated pathways, consistent to our previous findings in this study.

CellChat (26) was next used to investigate the signaling strength and interactions between fibroblast subclusters in NVP-BHG712 compared with vehicle-treated groups under bleomycin condition. CellChat analysis identified multiple ligand-receptor interactions across distinct fibroblast subclusters (Supplemental Figure 8A). Interestingly, collagen-based interactions exhibited one of the highest interaction strengths (Supplemental Figure 8, A and B). While investigating collagen-based interactions specifically in NVP-BHG712 versus vehicle-treated groups under bleomycin condition, the total strength and number of interactions were reduced in response to EphB4 inhibition (Figure 6F)

We then evaluated the expression of all identified downregulated ECM-related genes in each fibroblast subcluster in response to NVP-BHG712 treatment under bleomycin condition. Interestingly, ELN expression was identified in all fibroblast subclusters, with subcluster 1 and 2 showing the highest expression. In response to NVP-BHG712 treatment, the expression of ELN was reduced (Figure 6G).

Overall, our data have identified ECM transcriptomic changes associated with distinct fibroblast subsets in response to NVP-BHG712 treatment during fibrosis.

Identifying ELN as an EphB4 signaling partner using tanscriptomic differences in control donor versus IPF human lung fibroblasts treated with NVP-BHG712. To investigate the relevance of EphB4 inhibition in human fibroblasts, we treated primary human lung fibroblasts from control donors (HLF) and patients with IPF (Supplemental Table 7) with NVP-BHG712 or vehicle for 24 hours and subjected each sample to total RNA-Seq (total 16 samples; n = 4/group; Figure 7A). When IPF and HLF fibroblasts were compared under vehicle conditions, 3,108 DEGs (1,569 downregulated and 1,539 upregulated) were identified in the IPF group (Figure 7B). It is notable that the expression of COL1A1 was significantly increased, and ELN was determined as the most upregulated ECM gene.

Figure 7

ELN is a key signaling partner of EphB4 in human IPF. (A) Schematic detailing total RNA-Seq pipeline of human lung fibroblasts (HLF) from control donors or patients with IPF subjected to NVP-BHG712 (250 nM) or vehicle (n = 4/group). (B) Differentially expressed genes (DEGs) comparing IPF to HLF fibroblasts using an FDR q < 0.05 (red). Heatmap depicting 3,108 DEGs (1,569 downregulated and 1,539 upregulated) identified in IPF versus HLF fibroblasts. Dot plot showing expression of top ECM genes in IPF versus HLF fibroblasts per donor. (C) DEGs comparing IPF fibroblasts treated with NVP-BHG712 or vehicle (red). Heatmap depicting 85 DEGs (61 downregulated and 24 upregulated) identified in response to EphB4 inhibition of human IPF fibroblasts. Dot plot showing expression of ELN comparing NVP-BHG712 versus vehicle treatment in IPF fibroblasts per donor. (D) EphB4 was queried with ELN, as identified across human and mouse RNA-Seq, using the Integrated Interaction Database to find direct human PPIs and visualized using NAViGaTOR.

Next, the transcriptomic differences elicited by EphB4 inhibition were determined by comparing IPF fibroblasts treated with NVP-BHG712 compared with corresponding vehicle controls. In total, 85 DEGs (24 upregulated and 61 downregulated) were identified (Figure 7C; a list of the top 50 DEGs is provided in Supplemental Table 8). Notably, expression of ELN was significantly downregulated by NVP-BHG712 inhibition in IPF fibroblasts.

To identify common signaling partners involved in EphB4 signaling in mouse and human fibroblasts during lung fibrosis, the DEG list from human IPF RNA-Seq analysis was compared with the DEG list from Ephb4-CKO mouse RNA-Seq. Four common gene targets were identified that included 2 upregulated genes (low density lipoprotein receptor [LDLR] and transmembrane protein 38b [TMEM38B]), and 2 downregulated genes (ELN and glycyl-TRNA synthetase 1 [GARS1]) (Supplemental Table 9). In our analysis using pathDIP database, LDLR and TMEM38B did not associate with any pathways, ELN associated with ECM organization and elastin synthesis pathways, and GARS1 associated with protein translation. Among these 4 genes, ELN stood out as a prime target for investigating EphB4 signaling since ELN was consistently downregulated with Ephb4 CKO (total RNA-Seq) and pharmacological EphB4 targeting in both mouse (snRNA-Seq and NanoString) and human IPF fibroblasts (total RNA-Seq).

To further elucidate a potential relationship between EphB4 signaling and ELN, we queried physical protein-protein interactions (PPIs) between the 2 entities in the Integrated Interaction Database (Figure 7D and Supplemental Table 10) (27). We identified 4 proteins that can physically interact with EphB4 and ELN, including amyloid precursor protein (APP), fibronectin (FN1), perlecan (HSPG2), and TGF-β1. It is notable that a key profibrotic cytokine, namely TGF-β1, linked EphB4 to ELN (13). Indeed, GO analysis also indicated that ELN, among 16 other DEGs from the human IPF total RNA-Seq data, was enriched for TGF-β1 regulation of ECM (Supplemental Figure 9), and the downregulated genes in our sequencing analysis were enriched for ECM-related pathways (Supplemental Table 11). Overall, PPI analysis coupled with our sequencing data suggests that ELN may be a crucial signaling partner involved in EphB4 signaling during lung fibrosis.

Inhibition of EphB4 mitigates fibrotic upregulation of ELN in lung fibroblasts and fibrotic lung tissue. To further investigate the regulatory relationship between EphB4 and ELN, and their potential link to TGF-β1 signaling, lung fibroblasts from patients with IPF were subjected to TGF-β1 stimulation followed by EphB4 inhibition using NVP-BHG712 or EPHB4 siRNA treatment. As shown in Figure 8, A and B, TGF-β1 stimulation markedly increased the expression of ELN in lung fibroblasts, while subsequent treatment with NVP-BHG712 or EPHB4 siRNA moderately or significantly attenuated-induced ELN expression, respectively. Stress fiber formation induced by TGF-β1, as indicated by rhodamine-phalloidin staining, and αSMA staining was also reduced by NVP-BHG712 (Supplemental Figure 10).

Figure 8

EphB4 signaling in PF. (A and B) Human IPF lung fibroblasts were treated with TGF-β1 or PBS for 24 hours. Cells (total 16 samples; n = 4 patients/group) were then treated with either NVP-BHG712 (250 nM) (A) or Ephb4 siRNA (B), or their respective controls, for an additional 24 hours. Protein lysates were collected, and Western blotting was performed to detect ELN protein levels. ImageJ densitometry was performed and normalized to GAPDH (mean ± SD). *P < 0.05; **P < 0.01, as determined by paired 2-way ANOVA followed by Tukey’s post hoc test adjusted for multiple comparisons. (C) Western blotting of siRNA treated IPF cells to detect P-SMAD 2/3 and total SMAD 2/3 (n = 4/group). (D and E) Verhoeff Van Gieson staining was performed on lung sections from mice with Ephb4 conditional knockout or respective control mice to visualize ELN (n = 5/group) following bleomycin or PBS instillation (D), or mice with bleomycin or PBS instillation treated with EphB4 inhibitor NVP-BHG712 (10 mg/kg) or vehicle (n = 5/group) (E). ELN staining was semiquantitatively scored (scores 0–3). ****P < 0.0001 as determined by 2-way ANOVA followed by Tukey’s post hoc tests adjusted for multiple comparisons. (F) Schematic depicting the role of EphB4 signaling in PF.

As a key regulatory cytokine for fibrogenesis, TGF-β1 typically relies on canonical signaling associated with phosphorylation of the transcription factors SMAD2/3. To assess whether SMAD2/3 activity affects the regulatory effect of EphB4 on ELN expression, human lung fibroblasts were treated with TGF-β1 or control for 24 hours, before being treated with EPHB4 siRNA or control for an additional 24 hours (Figure 8C). As predicted, TGF-β1 treatment increased phosphorylation of SMAD2/3 in human lung fibroblasts; however, silencing of EPHB4 did not decrease phosphorylation levels of SMAD2/3, suggesting that EphB4 likely does not modify fibrotic pathways through canonical TGF-β1 signaling.

To assess the link between EphB4 and ELN in vivo, ELN staining, assessed using Verhoeff Van Gieson ELN stain, was examined in the lungs of bleomycin-challenged Ephb4-CKO and Ephb4 control mice (Figure 8D) and NVP-BHG712–treated mice (Figure 8E). Both Ephb4-CKO and NVP-BHG712 mouse lung tissue showed qualitative decreases in ELN protein after bleomycin challenge, with no changes observed in PBS controls. Thus, our data suggest that EphB4 expression is crucial for ELN expression in the context of fibrosis in both mouse tissue and human lung fibroblasts. Overall, our data support a mechanism involving EphB4-ELN signaling in lung fibrosis.

Ephrins and Ephs are key regulators of embryogenesis, prenatal development, and wound healing (28). While Ephrin-B2 is known to participate in the dysregulation of tissue repair during fibrosis (7, 9, 29), the role of EphB4 signaling in fibrosis remains unclear. To comprehensively explore the role of EphB4 receptor in PF development, we employed a multifaceted approach that included the generation of Ephb4-CKO mice in conjunction with a preclinical animal model of lung fibrosis, total RNA-Seq and snRNA-Seq, NanoString, previously published single cell data, advanced computational biology analysis, and functional assays using mouse and human (HLF and IPF) lung fibroblasts.

To our knowledge, we are the first to demonstrate that pharmacologic inhibition or genetic deletion of EphB4 reduces the expression of fibrotic ECM-related transcriptome and attenuates the development of PF. We showed that Col1a2-driven deletion of Ephb4-CKO or EphB4 pharmacological inhibition using NVP-BHG712 attenuated bleomycin-induced PF in vivo. Our results indicate that fibrotic injury caused transcriptomic changes in lung fibroblasts, along with the transcriptomes of specific fibroblast subpopulations. Consistent with the reduced pathological histomorphometry, we also found that genetic or pharmacological targeting of EphB4 ameliorated fibrotic injury–induced changes through reduction of select ECM genes in fibroblasts. A subset of genes identified in our mouse studies were also elevated in human lung fibroblasts from patients with IPF when compared with HLFs, with ELN identified as a prominent interaction partner of EphB4 signaling in mouse and human lung fibroblasts, and across genetic and pharmacological EphB4 targeting.

To date, the role of EphB4 in injury and fibrosis remain elusive, particularly in lung tissue. EphB4 mutations causing increased expression in lung tumors has been linked to critical cell processes including tumor cell survival and proliferation (30, 31). Interestingly, we found that reduction of EphB4 in lung tissue markedly protects against development of PF and that this antifibrotic phenotype garners profound changes in the transcriptome (317 DEGs) of mouse pulmonary fibroblasts as determined though total RNA-Seq. A subset of the 214 downregulated DEGs was linked to ECM organization, protein transport, or ER cargo regulation in fibroblasts. While the exact mechanisms of EphB4 signaling remains to be fully elucidated, downstream targets of EphB4 mediate cytoskeletal rearrangements via actin GTPase activity, which is typically associated with cell motility (29, 32). EphB4 deficiency has also been linked to accumulation of type IV collagen in endothelial cell ER through impaired protein trafficking, illuminating the potential of EphB4 as a regulator of protein transport and secretion (33). However, our in vivo data also suggest that the role of EphB4 may be context specific, as we did not observe protection from bleomycin-induced skin fibrosis in Ephb4-CKO mice. Overall, the cytoskeletal role of EphB4 signaling may intersect with ECM secretion and protein trafficking genes observed in our study; however, further investigation is required to parse the biological and pathological signaling mechanisms for EphB4 in various tissues.

Recent studies have identified distinct fibroblast cell subsets and transcriptomic profiles that may play a key role in PF development (23–25, 34–36). Using snRNA-Seq, we identified 4 fibroblast subclusters within mouse lung tissues. Mapping of transcriptomic profiles associated with peribronchiolar, fibrotic, alveolar, and adventitial fibroblasts, identified by previous studies (23, 24), showed enrichment of these transcriptomic profiles in distinct fibroblast subsets. Fibroblast subcluster 2 from our dataset was enriched with genes associated with the fibrotic fibroblasts. It is interesting to note that subcluster 2 showed increased cellular proportionality among all fibroblasts with bleomycin treatment and was attenuated with EphB4 inhibition. Furthermore, ELN was identified in all fibroblast subsets with higher expression in subcluster 2 and 1, and its expression was reduced in response to NVP-BHG712 treatment under bleomycin condition.

Among the ECM genes identified in mouse lung tissue, a subset was also elevated in lung fibroblasts from patients with IPF compared with donor controls; however, when IPF fibroblasts were treated with EphB4 inhibitor, ELN was the only ECM gene that was significantly downregulated. ELN also emerged as a gene that was consistently responsive to EphB4 deletion and inhibition across total RNA-Seq of IPF and control donor fibroblasts, total RNA-Seq and snRNA-Seq of mouse lung fibroblasts, and NanoString in mouse lung fibroblasts. Our PPI results highlight ELN as a key signaling interaction partner of EphB4 during PF. PF pathological impairment of tissue function is largely driven by excessive ECM accumulation, to which elastin reorganization and protein deposition are known indicators of disease in patients with PF (19). PPI and GO analyses suggested that EphB4 and elastin signaling both interact with TGF-β1. TGF-β1 is well established to increase elastin expression in lung fibroblasts (37); however, we showed that this induction is ameliorated by EPHB4 silencing. Furthermore, both genetic and pharmacological blockade of EphB4 was sufficient to reduce elastin in mouse lung tissue following bleomycin-induced PF, further demonstrating a crucial link between EphB4 and elastin signaling during PF

In this study, we used the bleomycin-induced model of lung fibrosis to assess the role of EphB4 in the development of PF. However, the results from this study do not indicate that EphB4 is targetable to reverse or ameliorate existing fibrosis as a therapeutic, which would require future investigation. Furthermore, while the bleomycin preclinical model of lung fibrosis is important in understanding mechanisms of fibrosis, it does not fully capture all clinical manifestations of human IPF. For example, strong acute inflammatory phase observed in the bleomycin model is typically not present in patients with IPF (1, 2); thus, future studies should be directed toward understanding the role of Ephb4 signaling in other relevant animal models to further understand its role in PF.

Together, our Ephb4 CKO, small molecule EphB4 inhibitor NVP-BHG712, and EPHB4 siRNA data coupled with transcriptomic data demonstrate an important role of EphB4 signaling in PF (Figure 8F). Targeting EphB4 signaling requires extensive additional investigation due to the vital functions of this molecule in developmental and physiological processes, including angiogenesis (38–40), and blood pressure and heart rate regulation (41). Additional studies are also warranted to elucidate the fibrotic signaling pathways from EphB4 to the expression of its effector molecules (e.g., elastin). Signaling molecules in these pathways may provide alternative therapeutic targets for the treatment of PF and other pathological fibrotic conditions in which EphB4 signaling is implicated.

Sex as a biological variable. For our animal studies, we used male and female animals for the CKO mice subjected to the bleomycin model of PF. For our human cell studies involving RNA-Seq and functional assays, we obtained lung fibroblasts from both male and female subjects.

Generation of Ephb4-CKO mouse. Ephb4^tm1.1Jwu mice, donated by Jianping Wu (University of Montreal), were crossed with B6.Cg-Tg(Col1a2-cre/ERT,-ALPP)7Cpd mice (The Jackson Laboratory; strain #029567) to generate Ephb4^loxP/loxP;Col1a2-Cre^ERT mice on a C57BL/6J background. Mouse lines were backcrossed for 5 generations prior to the initiation of mouse experiments. To induce conditional Ephb4 knockdown, 4-hydroxytamoxifen was administered by i.p. injection (1 mg in 100 μL 20% ethanol/80% corn oil vehicle) for 5 consecutive days. Control mice were administered vehicle comprising 20% ethanol and 80% corn oil. Lung fibroblasts were isolated and cultured from Ephb4-CKO mice and Ephb4 control mice, and qPCR and Western blotting was performed to assess successful deletion of EphB4.

Mouse genotyping. Ear punches were collected, and alkaline lysis was performed at 90°C for 1 hour to extract DNA. DNA was subjected to PCR against the Ephb4 and Cre genes (100 V for 35 minutes), and gels were imaged using Sybr Safe (Thermofisher) in a BIO-RAD Chemidoc apparatus. Primers used can be found in Supplemental Table 12.

Lung fibrosis model. To induce PF, bleomycin sulphate (1.2 U/kg in 50 μL PBS; Bioshop) was intratracheally instilled in 6- to 7-week-old mice that were anesthetized with isoflurane (1.5%) and xylazine/ketamine cocktail (10 mg/kg xylazine:100 mg/kg ketamine). Bleomycin control mice were administered PBS only. Instillation was initiated 7 minutes following xylazine/ketamine administration. Mice were sacrificed by anesthetization and CO₂ asphyxiation 14 days following bleomycin challenge. For experiments assessing the effects of NVP-BHG712 on the development of bleomycin-induced PF, mice were intratracheally administered bleomycin sulphate (1.2 U/kg) at 7 weeks of age. At 5 days after bleomycin challenge, the treatment with NVP-BHG712 (Sigma Aldritch; SML0333; 10 mg/kg in 150 μL N-methyl-2-pyrrolidone and polyethylene glycol 300 at ratio of 1:9) or vehicle control via oral gavage was initiated. Mice were given NVP-BHG712 once daily for a total of 5 days. Mice were sacrificed and lungs collected 2 weeks after bleomycin challenge.

Skin fibrosis model. To induce skin fibrosis, mice were treated with bleomycin sulphate by s.c. injection. Briefly, bleomycin sulphate was reconstituted to 1 U/mL in PBS, aliquoted, and stored at –80°C. Ephb4-CKO and Ephb4 control mice at 8 weeks of age were anesthetized with xylazine and ketamine (10 mg/kg xylazine:100 mg/kg ketamine), and back hair was removed using Nair. Bleomycin stock solution was diluted to a working concentration (0.1 U/mL in 100 μL volume), and mice were s.c. treated with bleomycin daily for 21 days. Mice were sacrificed 1 day following treatment end point and 4 skin sections were collected for each animal.

Lung histology. After mice were sacrificed, the heart was carefully removed and the lungs were inflated with 1 mL 10% formalin via the trachea, which was then tied off. The entire lung was excised, rinsed in PBS under gentle shaking for 1 minute, and fixed in 10% formalin at room temperature with shaking for 48 hours. After rinsing the lungs in PBS for 30 minutes, lobes were individually excised, dehydrated in reagent alcohol for 1 minute with vigorous agitation, and incubated for 24 hours in a minimum of 50 mL JFC solution (Milestone) per mouse lung for completion of tissue dehydration and tissue clearing. Processed lobes were paraffin embedded with the ventral side facing the surface of the paraffin block. For staining, 4 μm lung sections were collected after removal of 80–150 μm of tissue. Gomori’s One Step Trichrome stain (Thermo) was performed according to manufacturer’s instructions to assess lung collagen content. To assess the degree of PF in affected lung tissue, Ashcroft scoring was performed across total lung sections with a minimum of 50 nonoverlapping fields of view (12). For quantification of lung collagen staining across total lung tissue, lung sections were imaged at 4X magnification and stitched together using Autostitch (42) to form full lobe images. Lobes were color corrected in GIMP image editing software V.2.10.34. Quantification of lung collagen staining was performed in Fiji V. 2.9.0 (43). Color deconvolution was performed selecting the preset Mason’s trichrome filter to separate blue, green, and red color channels (43). The blue color channel was selected and thresholding was performed using the filter “mean.” Normalization to total lung tissue was performed by converting total lobe images to RGB stacks and quantifying the total pixel count using the threshold filter “default.” Lung collagen was expressed as percent lung collagen (%) by normalizing the blue color channel to the total lobe. Lung sections from C57BL/6J mice treated with bleomycin, then NVP-BHG712 or vehicle were normalized to total tissue and collagen content was expressed as a percentage. Verhoeff Van Gieson stain (Abcam) was performed according to manufacturer’s instructions to assess elastin in lung tissue. Lung sections were assigned elastin expression scores on a scale of 0–3 (absent, mild, moderate, or severe staining) based on the level of staining of elastin fibers against pink and yellow counterstain.

IHC. IHC methodology can be found in the Supplemental Methods.

Skin histology. Mouse epidermis and dermis sections were collected using surgical scissors and forceps. Mouse skin was fixed in 10% formalin at room temperature for 48 hours, before being incubated for 24 hours in a minimum of 15 mL JFC solution per 4 mouse skin sections. Skin sections were embedded in paraffin, and 4 μm sections were collected. Mouse skin sections were stained with Gomori’s Onse Step Trichrome. Skin fibrosis scoring and dermal thickness analyses were performed as previously described (7).

Lung fibroblast isolation. Mouse lungs were perfused with 10 mL PBS at 37°C through the right ventricle of the heart and out the abdominal aorta until completely blanched. Lungs were excised and rinsed in PBS. After blotting away excess PBS, lung tissue was minced by cutting with scissors 50 times. In total, 1 mL collagenase (1 mg/mL in DMEM; Sigma) was then added, and tissue pieces were cut an additional 50 times until they were no larger than 1 mm in diameter. The lung tissue plus collagenase slurry was collected and volume adjusted with collagenase to 5 mL. Tissue pieces were digested by incubating at room temperature for 15 minutes. Tissue digest was deactivated with 1 mL FBS, and the cells were immediately passed through a 70 μm cell filter and rinsed with 10 mL DMEM. Cells were centrifuged at 300g for 8 minutes and resuspended in 1 mL complete DMEM (10% FBS, 1% penicillin/streptomycin). Cells were seeded in a 6 cm plate in 4 mL complete DMEM for 24 hours, at which point 50% of the media was changed. Cells were lifted with trypsin (0.05%) and passaged when they reached 80% confluence (3–7 days). For the preparation of RNA for NanoString gene analysis, mouse lung fibroblasts from C57BL/6J mice were treated with TGF-β1 (20 ng/mL) or PBS (control), and NVP-BHG712 (250 nM) or DMSO (vehicle control), for 24 hours.

Skin fibroblast isolation. Ephb4 control (n = 6) and Ephb4 CKO (n = 6) mice were used to isolate dermal fibroblast cells. After sacrifice, mouse skin was sterilized with 70% ethanol and fur was removed with Nair. The skin was excised, minced, and digested in collagenase for 2 hours at 37°C. Cell suspension was collected and centrifuged at 300g for 8 minutes. Cell pellets were resuspended in DMEM (10% FBS, 1% penicillin/streptomycin) and cultured in 60 mm culture dishes for 24 hours. Mouse fibroblasts were passaged to P1 and cell supernatants were collected for RNA and protein collection.

Primary human lung fibroblast culture and treatment. Primary human lung fibroblasts from either patients with IPF (n = 4; Lonza, CC-7231) or donors with no history of lung fibrosis (n = 5; Lonza, CC-2512) were purchased at P2 (Supplemental Table 7). Fibroblasts were seeded, expanded, and passaged to P4. The cells were starved overnight and subsequently treated with TGF-β1 or PBS (control) for 48 hours. After 24 hours of TGF-β1 treatment, NVP-BHG712 (250 nM) or DMSO vehicle was spiked into the fibroblast cultures for an additional 24 hours, and RNA was isolated for qPCR. For RNA-Seq, P4 primary IPF lung fibroblasts were treated with NVP-BHG712 or DMSO vehicle for 24 hours. IPF lung fibroblasts were alternatively treated with silencer select siRNA (Thermo) targeted to EphB4 using RNAiMax lipofectamine according to manufacturer’s instruction. siRNA treatment was performed for 24 hours at the recommended concentration of 25 pmol per well in a 6-well plate.

Western blot. Western blot methodology can be found in the Supplemental Methods.

Immunofluorescence. Immunofluorescence methodology can be found in the Supplemental Methods.

Flow cytometry. Flow cytometry methodology can be found in the Supplemental Methods.

RNA isolation and qPCR. RNA isolation and qPCR methodology can be found in the Supplemental Methods. Primer sequences can be found in Supplemental Table 13.

NanoString. Mouse lung fibroblasts treated with TGF-β1 and NVP-BHG712, or human lung fibroblasts from IPF donors treated with NVP-BHG712, were lysed in Trizol reagent and frozen at –80°C. Total RNA was isolated using RNeasy MinElute columns (74204; Qiagen), per the manufacturer’s instruction. RNA was quantified by Nanodrop and 100 ng of total RNA was hybridized overnight at 65°C on the nCounter instrument (NanoString Technologies), using custom designed panels for mouse or human transcripts of interest. Primers used for the NanoString gene panels can be found in Supplemental Table 14. Results were assessed using the nCounter analysis system on log-transformed expression counts, with normalization performed using the standard housekeeping controls Acad9 and Rplp0. Paired 2-tailed t test and generalized least squared regression were performed for all analyses. The resulting P values were further adjusted for multiple comparisons according to the Benjamini-Hochberg method with q < 0.05 being determined as significant.

Total RNA-Seq and analyses. Quality of total RNA extracted from either cultured lung fibroblasts of Ephb4-CKO mice treated with bleomycin (n = 4) or PBS (n = 5), Ephb4 control mice treated with bleomycin (n = 4) or PBS (n = 4), or human IPF lung fibroblasts treated with NVP-BHG712 or vehicle (n = 4 per treatment group), were assessed using an RNA Nano chip on an Agilent Bioanalyzer (Agilent). RNA integrity numbers for all samples were > 9.5. Samples were fluorometrically quantified using the Qubit RNA BR assay (ThermoFisher) on the Denovix DS-11 spectrophotometer (Denovix). For each sample, 200 ng RNA was used to prepare sequencing libraries using the TruSeq Stranded Total RNA with RiboZero following the low sample protocol as per manufacturer’s recommendations (Illumina), as previously described (44). Library quality was assessed on a high sensitivity DNA Bioanalyzer chip (Agilent). In total, 17 mouse libraries or 8 human libraries were volumetrically pooled and sequenced on an Illumina NextSeq 550 sequencer for 75 paired end read-cycles at the Centre for Arthritis Diagnostic and Therapeutic Innovation (CADTI, Schroeder Arthritis Institute, Krembil). Sequencing data are deposited in the gene expression omnibus (GEO) data repository under the accession codes GSE205612 and GSE205613 for the mouse and human lung fibroblasts, respectively.

Demultiplexing of samples was performed using bcl2fastq conversion tool (v2.19.1.403). Quality assessment of each sample revealed high quality of reads. To maintain minimum read length of 25 bp after trimming of adapters, Cutadapt (v2.5) software was executed (very few reads being contaminated ~3%) along with trimming of Ns. Splice-aware alignment of reads using a Hierarchical Graph FM index (HGFM) method was performed using HISAT2 software (v2.1.0 with parameters –rna-strandness RF –dta) against mouse reference genome (vGRCm38) and human reference genome (vGRCh38) for mouse and human samples respectively. To populate the abundance of transcripts based on the reference genome and transcriptome, StringTie (v2.0.3 with parameters -e -B -G referencetranscripts.gtf) was run to generate table format output files. These files were processed using the script prepDE.py from StringTie software to collect the raw gene expression levels for downstream processing.

For mouse lung fibroblast RNA-Seq analyses, raw gene expression levels comprised 55,421 genes measured across fibroblasts obtained from 17 mice. Lowly expressed genes were filtered out based on genes with less than 2 samples with a count per million less or equal to 10. This filter kept 10,481 genes for further analysis. For human IPF fibroblast RNA-Seq analyses, 60,675 genes were measured across 8 samples in the original matrix of counts. Low-expressed genes were filtered out based on genes with fewer than 2 samples with a count per million less or equal to 10. This resulted in retaining 9,619 genes for further analysis.

Changes in gene expression were evaluated using the method proposed by Anders and Huber (45), which assumes a negative binomial distribution. Function DESeq in R package DESeq2 was used with default parameters to perform differential expression analyses (46). Pairwise comparisons among groups of interest were calculated and tested for statistical differences in expression levels. All analyses include batch number as a categorical covariate in the model to control for potential batch effects. P values were adjusted for multiple comparisons via FDR correction proposed by Benjamini and Hochberg (47). An FDR of 5% (q < 0.05) was statistically significant. All analyses were performed in R version 3.5.0 (48).

Computational analyses. Identified DEGs were analyzed using pathDIP (14) (https://ophid.utoronto.ca/pathDIP) v.5 selecting mouse as the organism, Reactome as a source, and Ortholog Pathways members as annotation, retaining only pathways with q < 0.01(48). Up- and downregulated genes were subjected to pathway analysis, and unique pathways were further considered. GO pathway enrichment analysis was performed using Enrichr web portal (June 1, 2022) and BioPlanet 2019 pathway source (49). For network annotation, each gene was annotated with the pathway with the lowest q value it belonged to, and a network with the gene-pathway interactions was built using NAViGaTOR v.3.0.16 (15). After validation through qPCR of genes involved in pathways linked to ECM, PPIs were retrieved using mouse Integrated Interaction Database (IID; https://ophid.utoronto.ca/iid) v.2021-05 (27), and shortest paths between Ephb4 and the confirmed genes were calculated using igraph_1.2.6 in R 4.0.3 (48). Node color depicts Gene Ontology Biological Process. Network image was exported in SVG format and finalized in Adobe Illustrator version 26.3. Gene Ontology and Reactome GSEA were performed using clusterProfiler_4.6.2 and ReactomePA_1.42.0 in R 4.2.3. GSEA was performed using all deregulated genes combined as well as upregulated and downregulated genes, individually.

EPHB4 and ELN were used to query IID v.2021-05 to retrieve their direct, human, physical protein interactions (27). We used NAViGaTOR v.3.0.16 to visualize the resulting network and highlight connecting proteins and those confirmed in RNA-Seq data (red label) (15). Node color depicts Gene Ontology Cellular Component. Network image was exported in SVG format and finalized in Adobe Illustrator version 26.3.

snRNA-Seq and bioinformatics analysis. C57BL/6J mice were subjected to intratracheal instillation of bleomycin or PBS control, then treated with NPV-BHG712 or vehicle control as described above. Mouse lungs were perfused with PBS, isolated, snap frozen, and stored at –80°C. Nuclei were isolated from frozen mouse lungs (n = 12; n = 3/group) by physical dissociation using a glass dounce homogenizer and were sorted using fluorescence-activated nuclei sorting (FANS) based on DAPI⁺ fluorescence to obtain a single nuclei population. Nuclei concentration was determined using DAPI staining and hemocytometer, then adjusted to 1,000 nuclei/μL. SnRNA-Seq libraries were generated using the 10X Genomics Chromium Next GEM Single Cell 3′ Reagent Kits (v3.1 Dual Index). Sequencing was performed on the Illumina NextSeq 550 system (2x150 bp High Output Kit) at the CADTI. For data analysis of the single nuclei sequencing experiments, sequencing data were processed using Cell Ranger pipeline (v 7.2.0) by 10x Genomics (https://www.10xgenomics.com). Reads were aligned to the mouse transcriptome (mm10), followed by filtering and correction of cell barcodes and Unique Molecular Identifiers (UMIs). Reads associated with retained barcodes were quantified and used to build expression matrices. The standard procedures of filtering, normalization, dimensionality reduction, and clustering were performed using R package Seurat (v5.0.3) (50). Low-quality cells expressing fewer than 200 genes and low-expressed genes present in fewer than 3 cells were removed. DecontX from Celda suite (celda_1.22.1) was used to remove ambient RNA contamination in individual cells. Cells with high mitochondrial DNA, high number of unique feature counts (potential doublets or multiplets) were excluded. We used DoubletFinder for the detection of doublets (51). Samples were merged for subsequent clustering and visualization. The batch correction between samples was done with the Harmony R package (v1.0) (52).

Data were log normalized and highly variable genes were detected using variance-stabilizing transformation (vst). The normalized data were centered and scaled. Dimensionality reduction was performed using principal component analysis (PCA) and the most significant PCs were chosen for subsequent clustering. Graph-based clustering was implemented by calculating k-nearest neighbors, followed by modularity optimization to clusters cells. Nonlinear dimensionality reduction and visualization was performed using Uniform Manifold Approximation and Projection (UMAP). Clusters were annotated on the basis of canonical markers and differential gene expression testing was used to determine a gene set signature for each cluster using the Wilcoxon Rank Sum test. Pathway enrichment analysis was performed on upregulated and downregulated genes using the Reactome database through pathDIP, and the enriched pathways were subsequently visualized using dot plots.

To explore potential cell-cell interactions across fibroblast subclusters in bleomycin samples, we employed CellChat v2.0 R package where each fibroblast subcluster was chosen as “receiver” cell type and the other subclusters as “sender” cell type (26). The “netVisual_chord_gene” function in CellChat was used to generate a chord diagram (circle plots) to visualize the complex ligand-receptor interactions between source and target cell types. The total number of interactions is defined as the sum of ligand-receptor pair interactions across all cell groups for each condition within the inferred pathway. The interaction strength for a given pathway is defined as the sum of communication probabilities across all cell-group pairs in the inferred network.

Statistics. qPCR and Western blot data are presented as interleaved scatter plots with horizontal lines representing the mean. The upper and lower bounds in the scatter plots correspond to the SD of the data set. Statistical analysis identifying the effects of the following utilized 2-way ANOVA followed by Tukey’s honest significant difference post hoc test: determining the severity of fibrosis using either the Ashcroft scoring system or through ImageJ (NIH) computational analysis, quantification of IHC scoring for αSMA in lung tissue, evaluation of the mRNA expression of Col1a1 and Acta2 after treatment with PBS versus bleomycin and Ephb4 control versus Ephb4 CKO in mouse lung fibroblasts via qPCR, evaluation of the protein expression of ELN via Western blot after PBS versus TGF-β1 and Vehicle versus NVP-BHG712/EPHB4 siRNA, and quantification of ELN stain scoring for the Van Gieson stain in lung tissue. Statistical analysis identifying the effects of the following utilized 2-tailed Student’s t test: determining EphB4 expression or phosphorylation of EphB4 either through qPCR or Western blot comparing Ephb4 control to Ephb4 CKO. In all comparison tests, a value of P < 0.05 was considered statistically significant. Graphs and analysis were generated in GraphPad Prism version 10.

Study approval. Animal experiments were approved by the Animal Resources Centre under the animal use protocol number 3730, University Health Network.

Data availability. All datasets for RNA-Seq can be accessed online as part of the GEO super series GSE273468. Datasets for total RNA-Seq of mouse lung fibroblasts can be accessed online under accession no. GSE205612. Datasets for snRNA-Seq of mouse lung fibroblasts can be accessed online under accession no. GSE273467. Datasets for RNA-Seq of human lung fibroblasts can be accessed online under accession number GSE205613. Full access to all data points, means, P-values, and sample size numbers can be found in the Supporting Data Values file.