Endothelial Reprogramming Stimulated by Oncostatin M Promotes Inflammation and Tumorigenesis in VHL-Deficient Kidney Tissue

A novel mechanism of cross-talk between ECs and VHL-deficient kidney tubules that stimulates inflammation and tumorigenesis is discovered, suggesting OSM could be a potential target for ccRCC intervention.

Abstract

Clear-cell renal cell carcinoma (ccRCC) is the most prevalent subtype of renal cell carcinoma (RCC), and its progression has been linked to chronic inflammation. About 70% of the ccRCC cases are associated with inactivation of the von Hippel–Lindau (VHL) tumor-suppressor gene. However, it is still not clear how mutations in VHL, encoding the substrate-recognition subunit of an E3 ubiquitin ligase that targets the alpha subunit of hypoxia-inducible factor-α (HIFα), can coordinate tissue inflammation and tumorigenesis. We previously generated mice with conditional Vhlh knockout in kidney tubules, which resulted in severe inflammation and fibrosis in addition to hyperplasia and the appearance of transformed clear cells. Interestingly, the endothelial cells (EC), although not subject to genetic manipulation, nonetheless showed profound changes in gene expression that suggest a role in promoting inflammation and tumorigenesis. Oncostatin M (OSM) mediated the interaction between VHL-deficient renal tubule cells and the ECs, and the activated ECs in turn induced macrophage recruitment and polarization. The OSM-dependent microenvironment also promoted metastasis of exogenous tumors. Thus, OSM signaling initiates reconstitution of an inflammatory and tumorigenic microenvironment by VHL-deficient renal tubule cells, which plays a critical role in ccRCC initiation and progression.

Significance:

A novel mechanism of cross-talk between ECs and VHL-deficient kidney tubules that stimulates inflammation and tumorigenesis is discovered, suggesting OSM could be a potential target for ccRCC intervention.

Graphical Abstract

Introduction

Many forms of cancer, including renal cell carcinoma (RCC), have been linked to chronic tissue inflammation (1,2,3). While immune cells and myofibroblasts have been recognized as major contributors to chronic inflammation, the endothelium has not been widely regarded as an active player in the pathogenic process. Recent findings, however, indicate that vascular endothelial cells (EC) in fact play a crucial role in the inflammatory response and in tumor progression, beyond the scope of the well-known hypoxia-induced neoangiogenesis for oxygen and nutrient supply (4).

Clear-cell RCC (ccRCC) is the most prevalent histologic subtype of RCC, and is resistant to most conventional chemo- and radiation therapies. Interestingly, about 70% of the ccRCC cases are associated with genetic or epigenetic defects in the von Hippel–Lindau (VHL) tumor-suppressor gene (5, 6). The important role of loss of VHL function in tumorigenesis is associated with upregulation of hypoxia-inducible factor-α (HIFα; refs. 5, 7). Upregulation of HIF transcription factor has been linked to neoangiogenesis, metabolic imbalance, and malignant progression. However, how loss of VHL function in parenchymal cells influences the surrounding microenvironment, and thus initiates the formation of ccRCC, still remains unclear.

We previously generated a conditional Vhlh (the mouse VHL allele) knockout (KO, exon 1 deleted) mouse model (Hoxb7-Cre-GFP; Vhlh^loxP/loxP) in a subpopulation of kidney tubule cells, including those in the collecting ducts, a subset of distal and proximal tubules, and ascending Henle's loops (8). The KO kidney tissue exhibits severe interstitial inflammation and fibrosis in addition to hyperplasia and appearance of transformed clear cells, but no full-blown ccRCC (8, 9). Furthermore, the inflammatory response of the VHL-deficient tubule cells is the result of endoplasmic reticulum (ER) stress caused by metabolic imbalance (9).

During the course of studying the etiology of tissue inflammation in the conditional Vhlh KO kidney, we noticed expression of proinflammatory markers such as phosphorylated c-Jun N-terminal kinase (p-Jnk) in the vascular EC, even though the ECs were not subject to genetic manipulations in the Vhlh KO mice. This prompted us to propose that in the Vhlh KO kidney, ECs are activated and/or inflamed, and that these activated ECs may play a critical role in the protumorigenic microenvironment. We report here that these ECs undergo profound changes in the gene expression program, showing characteristics of inflammatory response and mesenchymal transition, both in vitro and in vivo. Furthermore, these changes are induced by oncostatin M (OSM) emanating from the VHL/Vhlh-deficient cells. These dysfunctional, or activated, ECs reciprocally help establish a microenvironment that promotes tissue inflammation, tumor-cell proliferation, and metastasis.

Materials and Methods

A detailed description of cellular and biochemical experimental procedures is described in Supplementary Materials and Methods. The primers for real-time PCR are listed in Supplementary Table S1. Antibodies and staining conditions used in immunofluorescence and IHC are listed in Supplementary Table S2.

Cell culture

All cell lines used were purchased from cell-line repositories [ATCC, or Bioresource Collection and Research Center (BCRC)] that perform authentication. Standard practice in our laboratory consists of the following: Upon receiving the samples from the suppliers, the cells were grown to approximately 70% to 80% confluency, expanded, examined for the presence of mycoplasma using EZ-PCR Mycoplasma Detection Kit (from Biological Industries, SKU: 20–700–20), and collected for storage in liquid N2 using standard protocol. Depending on the proliferative characteristics of the cell lines, the expanded cells were rinsed and divided into 15 to 30 1 ml aliquots each containing an equivalent of approximately 25% to 40% confluent cell number in a 10-mm dish. For experimentation, each frozen aliquot was used and propagated for no more than 10 passages depending on the viability of the cell lines. A portion of the regrown cells was stained with DAPI to ensure the lack of Mycoplasma contamination. Other experimental details are described in Supplementary Materials and Methods.

Animals

The Hoxb7-Cre-GFP; Vhlh^loxP/loxP mouse strain has been described previously (8). OSM receptor KO mouse (Osmr^−/−) was purchased from RIKEN [B6.129S-Osmr<tm1Mtan> (RBRC02711)] and has been described (10). NOD/SCID mice (NOD.CB17-Prkd^cscid/NcrCrl) were purchased from BioLASCO. All experiments with mice were performed in accordance with the protocol approved by the Institutional Animal Care and Use Committee of National Central University.

Human ccRCC samples

Human paraffin-embedded tissue array of ccRCC was purchased from US Biomax Inc. (KD241). Other human ccRCC samples were collected from patients at Triservice General Hospital, Taiwan, Institutional Review Board no. 2–106–05–079.

Statistical analysis

Statistical analysis was performed using unpaired two-tailed Student t test (two-group comparison) or one-way ANOVA with Tukey post hoc test (multiple-group comparison). The difference between groups was considered statistically significant when the P value was < 0.05.

Data availability

mRNA transcriptome of ECs has been uploaded to the GEO (#GSE159754). All other data in this study are available from the corresponding author upon request.

Results

Conditional Vhlh KO mice exhibit inflammatory, clear-cell, and adenoma phenotypes

For brevity, the conditional mutant mouse (Hoxb7-Cre-GFP; Vhlh^loxP/loxP) is named Vhlh^KO and the wild-type mouse (Hoxb7-Cre-GFP or Vhlh^loxP/loxP) is named WT in this report. As shown in Fig. 1A, the KO kidney shows prominent phenotypes at 3 months of age, including formation of clear-cell clusters with irregular cell shape, infiltration of immune cells, increased vasculature, fibrosis, and appearance of inflammatory foci (clusters of immune cells and dilated tubules) with a penetrance of more than 90%, consistent with previous findings (8). These inflammatory foci are often associated with increased number of enlarged blood vessels (insets, Fig. 1A). Since Hoxb7-Cre-GFP is marked by the GFP reporter, it is possible to confirm that the tubule phenotypes are associated with Vhlh inactivation. As shown in Fig. 1B, wild-type kidney (Vhlh^loxP/loxP) without Hoxb7-Cre-GFP shows no GFP staining, while wild-type kidney carrying only Hoxb7-Cre-GFP shows GFP-expressing cells in both cortical and medullary tubule epithelia that are morphologically normal. In contrast, in Vhlh^KO kidney, the GFP-expressing tubules in both cortex and medulla exhibit multilayering with presence of clear cells (Fig. 1C). Accumulation of immune cells is also notable in the vicinity of abnormal tubules.

Figure 1.

Conditional Vhlh KO kidney shows inflammatory and tumorigenic phenotypes. Representative images of 4 to 5 animals are shown. A, Hematoxylin and eosin staining of kidney sections from 3-month-old mice, comparing wild-type (WT, Hoxb7-Cre-GFP) and mutant (Vhlh^KO, Hoxb7-Cre-GFP; Vhlh^loxP/loxP). White block arrows, clear cell clusters; black arrows, infiltrated immune cells; red arrows, blood vessels (containing red blood cells); sharp green arrows, fibrosis (pinkish fibrous staining); blue block arrows, inflammatory foci. Note that the inflammatory foci often contain an increased number of blood vessels. Bars, 50 μm. B, GFP staining of kidney sections from 3-month-old wild-type mice of the genotype Vhlh^loxP/loxP or Hoxb7-Cre-GFP; Vhlh^loxP/loxP shows no GFP and Hoxb7-Cre-GFP shows GFP staining in morphologically normal tubules in cortex and medulla. Bars, 20 μm. C, GFP staining of kidney sections from 3-month-old Vhlh^KO mice. GFP is expressed in abnormal tubules in cortex and medulla. These mutant cells show adenoma-like multilayering (yellow block arrows) and clear-cell (white block arrows) phenotypes. There is also an increase in immune-cell infiltrates near the mutant tubules (black arrows). Bars, 20 μm.

Our previous study has indicated that tissue inflammation in the Vhlh KO kidney is correlated with atopic activation (phosphorylation) of Jnk and NF-kB in the VHL-deficient cells resulting from ER stress (9). However, we have subsequently noted that p-Jnk expression is also enriched in vasculatures near the abnormal tubules in Vhlh^KO, but not in WT (Fig. 2A), even though these ECs do not harbor genetic manipulations. Similarly, in human ccRCC samples, p-JNK is not present in the normal adjacent tissue (NAT), while strong p-JNK expression is detected in the ECs (CD31⁺) surrounding clear-cell clusters in the ccRCC proper (Fig. 2B). It is also notable that these p-JNK/p-Jnk–expressing blood vessels are dilated and disorganized in Vhlh KO mouse kidney and in human ccRCC.

Figure 2.

ECs in Vhlh^KO mouse or human ccRCC kidney tissues exhibit increased p-JNK/p-Jnk expression, vasculature density, and vessel permeability. A, Representative images of mouse kidney sections double stained for CD34 and p-Jnk (n = 6). Because of the antibody host–source compatibility in double staining, CD34, instead of the more common CD31, was used for marking mouse ECs. p-Jnk is detected in abnormal tubules (white arrows) and in the adjacent ECs (CD34⁺ cells, yellow arrows) in Vhlh^KO compared with WT. Bars, 20 μm. B, Representative images of human ccRCC tissue sections double stained for CD31 and p-Jnk (n = 6). p-JNK was not detected in NAT. In ccRCC tissue, p-JNK is colocalized with CD31 in both grade 1 and grade 3 tumors. Bars, 20 μm. C, The number of CD31⁺ cells is increased nearly 2-fold in Vhlh^KO tissues compared with WT. Bars are 50 μm. D, Kidney sections were stained for FITC after tail-vein injection of FITC-dextran. Diffused FITC is detected surrounding blood vessels near abnormal tubules (red arrows), while in WT tissues, FITC is detected only within the glomeruli (G). Bars, 50 μm. Each data point in the graphs represents an individual animal or human tissue sample, and is the average number of four to five microscopic fields from each section. The data are presented as mean ± SD, and the P values are calculated by two-tailed unpaired Student t test. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Also, as shown in Fig. 2C, vascular density of Vhlh^KO tissue is increased by nearly 2-fold compared with WT. We then assayed in vivo vascular permeability by injecting FITC-conjugated dextran (500 kDa) via tail vein. As shown in Fig. 2D, IHC shows diffusion of FITC around the blood vessels in the peritubular space in Vhlh^KO mice, indicating that FITC-dextran has leaked out of the blood vessels and therefore cannot be cleared from circulation. Note that the microvasculature in glomeruli shows substantial permeability even in WT. This is consistent with the characteristics of these fenestrated vessels. These results suggest that alteration in vascular integrity already occurs before frank ccRCC is formed.

Activation of ECs in Vhlh^KO tissue

To understand the potential role of ECs in the Vhlh^KO phenotypes, we isolated ECs from WT (Hoxb7-Cre-GFP) and Vhlh^KO tissues and performed transcriptome analysis using RNA sequencing (RNA-seq). The RNA-seq result has been deposited in Gene Expression Omnibus (GEO; #GSE159754).

Anti-CD31 antibody-conjugated magnetic beads were used to isolate kidney ECs. These isolated ECs show high viability (>99%), and the purity of isolated CD31⁺ cells is approximately 95% (Supplementary Fig. S1A). Although CD31 is also expressed by lymphatic ECs, the expression level is much lower in the lymphatic vessels than in the vasculature (11). We therefore used stringent conditions for purifying ECs and expected that our EC preparation would contain few contaminating lymphatic ECs. Indeed, the expression levels of lymphatic EC gene Podoplanin in the RNA-seq–based transcriptome are low in either WT or Vhlh^KO samples (40–60-fold lower than those of CD31; Supplementary Table S3). The transcriptome shows that the ECs from Vhlh^KO (ECs-KO) kidneys exhibit profound changes in the gene expression program, compared with the counterparts isolated from WT (ECs-WT) animals (Supplementary Fig. S1B and S1C).

Overall, there are 450 genes upregulated and 483 genes downregulated more than 1.5-fold (Log₂) and with fragments per kilobase of exon per million reads (FPKM) over 5 in ECs-KO compared with ECs-WT (Supplementary Fig. S1C and Supplementary Table S3). The most upregulated gene is Arginase 1 (Arg1). Upregulation of Arg1 in EC has been linked to imbalance of nitric oxide (NO) synthesis, referred to as “eNOS uncoupling,” and suggested to play an important role in EC dysfunction and activation (12). For brevity, we will describe these ECs-KO as activated in this report.

Bioprocess Enrichment analysis shows that these differentially expressed genes are mainly involved in tissue inflammation. In addition, the enriched bioprocesses in cell migration and cell-shape regulation indicate mesenchymal transition of these ECs-KO, which has been suggested to play an important role in fibrosis (13,14,15). The Disease and Function analysis reveals that ECs-KO exhibit increased function of EC migration, growth of connective tissue, proliferation, tumor progression, and development of vasculature, as well as decreased function of glucose metabolism and acute inflammatory (i.e., antipathogen and antitumor) response (Supplementary Fig. S1C). These gene expression patterns are therefore consistent with the fibrotic and tumorigenic phenotypes observed in the Vhlh^KO kidney.

Among the upregulated genes, several are typically associated with chronic inflammation and fibrosis (Supplementary Fig. S2A). In the context of inflammatory response, ECs-KO show high expression of Hepatitis A virus cellular receptor 1 (Havcr1) and Lipocalin 2 (Lcn2). Havcr1 is a biomarker for kidney injury (16) and early detection of kidney cancer (17). Lcn2 has been recognized as a biomarker for acute kidney injury and chronic kidney disease (18). Also, IL6 is the most upregulated cytokine in ECs-KO. IL6 is produced at the site of inflammation and plays a crucial role in inflammatory response (19). IL6 from ECs has been shown to induce alternative macrophage activation (20) and promote tumor growth (21). There are also many upregulated genes related to immune-cell infiltration, including chemokines (Ccl2 and Ccl7) that recruit monocytes/macrophages, and adhesion molecules that capture circulating leukocytes, such as E-selectin (Sele), P-selectin (Selp), Intercellular adhesion molecule 1 (Icam1), and Vascular cell adhesion molecule 1 (Vcam1). Interestingly, markers associated with mesenchymal transition are enriched, including Snail1 (Snai1), Fibronectin 1 (Fn1), Collagen 1 (Col1a1), and Chitinase-like-3 (Chil3), consistent with the Bioprocess Enrichment analysis described above. Overexpression of these markers was verified by qRT-PCR (Supplementary Fig. S2B). Consistent with these gene-expression patterns, significant tissue fibrosis in the Vhlh^KO kidney has been reported previously (8, 9) and is confirmed here (Supplementary Fig. S2C). Also, ECs-KO indeed express high levels of Snai1 (Supplementary Fig. S2D). In addition, Icam1 is highly expressed in the vessels of Vhlh^KO kidney (Supplementary Fig. S2E), and ICAM1-positive vessels are associated with ccRCC (Supplementary Fig. S2F).

Osm is a major regulator of EC activation in conditional Vhlh KO mice

We suspect that the changes in EC-KO transcriptome are likely the result of paracrine action directly or indirectly emanating from the Vhlh-mutant tubule cells. Ingenuity Pathway Analysis (IPA) was therefore employed to identify the potential upstream regulators. Supplementary Table S4 shows top 15 potential upstream regulators with high increase values (based on z-score) that are statistically significant. Among these, OSM signaling shows the highest score. It is noteworthy that a few other candidates also belong to or are related to OSM signaling, including STAT3, IL6, and OSM receptor (OSMR).

OSM and IL6 belong to the same family of cytokines (19). They regulate overlapping but nonidentical sets of downstream genes, mainly via the JAK signal transducer and activator of transcription proteins (STAT)–signaling pathway. OSM is likely a more upstream regulator than IL6 for the following reasons: First, IL6 is known to be an OSM target gene in ECs (22, 23) and our EC transcriptome analysis also shows upregulation of IL6 in ECs-KO (Supplementary Table S3). Second, Osm is not a downstream target of IL6 yet the target genes of Osm, but not of IL6, such as Icam1 and Sele are overexpressed in ECs-KO. Recently, OSM has emerged as a major mediator in tissue fibrosis (24, 25) and may play an important role in the development of certain cancers (26, 27). Taken together, we consider Osm the most relevant signaling pathway in the activation of ECs in Vhlh^KO tissue.

OSM and other IL6-related family of cytokines share some but not all receptor components. In human, OSM signaling is initiated by binding of OSM to its type-I receptor complex (LIFRβ/gp130) or type-II receptor complex (OSMRβ/gp130) whereas mouse Osm shows high affinity only to type-II receptor (28). We will refer to OSMRβ as OSMR henceforth. ECs are the most probable target of OSM signaling because ECs express the highest level of OSMR in normal tissues (23, 29), and the levels of Osmr and gp130, but not Lifr, are further upregulated by Osm signaling and in Vhlh^KO (Fig. 3A; ref. 30). In order to establish the interactive network in the microenvironment of Vhlh^KO kidney tissue, we first sought to identify the source of Osm. As shown in Fig. 3B, WT tissue does not express appreciable level of Osm in any cells, while in Vhlh^KO tissue, high levels of OSM are detected in morphologically abnormal Vhlh-mutant tubule cells (Fig. 3B, lower panels). In addition, in human ccRCC samples, OSM and carbonic anhydrase IX (CAIX, a marker for VHL-mutant ccRCC cells) are undetectable in the NAT, while high levels of OSM are detected in CAIX-positive ccRCC cells (Fig. 3C).

Figure 3.

Cell-type–specific expression of OSM and OSMR. A, Overexpression of Osmr components in activated ECs from Vhlh^KO kidney. Heatmap of the transcriptome data shows Osmr subunits, Osmr and gp130, are overexpressed, while the alternative receptor, Lifr, is not (left). Overexpression of Osmr is validated by qRT-PCR analysis of RNA purified from the primary ECs of Vhlh^KO or WT kidneys (right). B, Mouse kidney sections from WT and Vhlh^KO were stained for Osm, which is overexpressed in the abnormal kidney tubules from Vhlh^KO (top right; n = 4) but not in WT (top left; n = 4). Double staining for GFP (Vhlh KO cell reporter) and Osm shows enriched expression of Osm in Vhlh KO cells (bottom; n = 4). Bars, 50 μm in top panels and 20 μm in bottom panels. C, Double staining of OSM and CAIX (a HIF target and marker of VHL-mutant ccRCC cells) in human ccRCC samples (n = 6). OSM and CAIX are undetectable in NAT. In contrast, OSM is strongly expressed in CAIX⁺ cells in cancer tissue. Bars, 20 μm. D, Human ccRCC samples double stained for CD31 and OSMR (n = 6). OSMR is not expressed in NAT but overexpressed in the CD31⁺ ECs around the clear-cell clusters in cancer tissue. Bars, 20 μm. Each data point in the graphs represents an individual animal or patient sample, and is the average number of four microscopic fields from each section. The data are presented as mean ± SD, and the P values are calculated by two-tailed unpaired Student t test. **, P < 0.01; ***, P < 0.001.

Currently there is no suitable anti-Osmr antibody for mouse-tissue IHC. However, an approximately 3-fold increase in Osmr mRNA level can be detected in primary ECs isolated from Vhlh^KO kidney compared with WT (Fig. 3A, right panel). In human ccRCC samples, high levels of OSMR protein is detected in cancer-associated vessels (CD31⁺), but not in normal adjacent tissue (NAT; Fig. 3D). We therefore conclude that OSM signaling mediates the interaction between Vhlh-mutant tubule cells and the ECs. Analysis of data from The Cancer Genome Atlas (TCGA) reveals that high expression levels of OSM (Supplementary Fig. S3A) and OSMR (Supplementary Fig. S3B) in ccRCC are associated with poor survival.

The characteristics of OSM-activated ECs

Two types of ECs, HUVECs or HMEC-1, were cocultured with human noncancerous kidney tubule cell line HK-2 with or without VHL knockdown (KD). Noncancerous kidney tubule cells were used because we were interested in the early events in the initial stage of tumorigenesis. Established malignant ccRCC cell lines accumulate many genomic alterations that are subsequent to tumor initiation and therefore could not serve this purpose. The efficiency of VHL KD was monitored by the reduction of VHL protein and increased expression of HIF1α protein (Fig. 4A; Supplementary Fig. S4A and S4B). Importantly, VHL KD results in increased expression of OSM in HK-2 cells (Fig. 4A; Supplementary Fig. S4C) and in conditioned media (Fig. 4B). In order to examine the cross-talk between kidney tubule cells and ECs, we employed a coculture system using the Boyden chamber (Fig. 4C). When cocultured with VHL KD HK-2, as compared with wild-type HK-2, HMEC-1 exhibits increased levels of OSMR, OSM downstream signal transducer p-STAT3 (activated STAT3), and target gene products ICAM1 and SNAIL1 (Fig. 4D, compare lane 1 with lanes 2 and 3; Supplementary Fig. S5A–S5D). The increase of OSM downstream target genes (p-STAT3, SNAIL1, and ICAM1) in ECs activated by VHL-deficient HK-2 was reverted when OSMR was knocked down (Fig. 4D, compare lane 4 with lanes 5 and 6, and lane 7 with lanes 8 and 9; Supplementary Fig. S5A, S5C, and S5D). When neutralizing OSM antibody was added to the conditioned media, the expression of these target genes in ECs was also reduced to the background level (Fig. 4D, lanes 10–12; Supplementary Fig. S5A, S5C, and S5D).

Figure 4.

OSM signaling mediates interaction between human kidney tubule cells and ECs. A, HK-2 cells were transfected with lentiviral vectors expressing control shRNA (shLuc976) or one of the two VHL-specific shRNAs (shVHL25 and shVHL61). Western blot analysis of protein extracts from the three cell lines shows decreased expression of pVHL and increased expression of OSM and HIF1α. Uncropped Western blots and quantification are shown in Supplementary Fig. S4. B, Conditioned media from the above three cell lines were measured by ELISA for the presence of OSM. VHL KD results in significant increase in the secretion of OSM. shRNA, short hairpin RNA. C, Schematics of the Boyden chamber coculture system. D, HK-2 cells with or without VHL KD were cocultured with HMEC-1 with or without OSMR KD, or in the presence or absence of neutralizing anti-OSM antibody in the coculture media. The expression levels of OSM downstream signaling markers were analyzed by Western blot. Uncropped Western blots and quantification are shown in Supplementary Fig. S5. E, Morphology and VE-cadherin relocalization of EC (HUVEC and HMEC-1) monolayer cocultured with wild-type HK-2 (shLuc976) or VHL KD HK-2. White arrows point to cytoplasmic relocalization of VE-cadherin in HUVEC, while in HMEC-1, the overall VE-cadherin expression is reduced. Representative images of a total of three independent experiments are shown. Bars, 20 μm. F, Permeability of EC monolayer cocultured with wild-type HK-2 or VHL KD HK-2 or in the presence of neutralizing OSM antibody (αOSM). The data are presented as mean ± SD of three independent experimental replicates, and the P values are calculated by two-tailed unpaired Student t test. ns, no significance; **, P < 0.01; ***, P < 0.001.

To verify the direct effect of OSM on ECs, human recombinant OSM (rOSM) was added to the media of EC single culture. p-STAT3, ICAM1, and VCAM1 show dose-dependent increase in ECs (Supplementary Fig. S6A). In addition, qRT-PCR also confirmed the rOSM effect on the expression of some of the target genes identified in the EC transcriptome, including CCL2, CCL7, SELE, and IL6 (Supplementary Fig. S6B). The uncropped protein blot and quantification in Fig. 6A are shown in Supplementary Fig. S7A–S7C. Together, these results indicate that OSM secreted by VHL-deficient kidney tubule cells is the major activator of the inflammatory program in the ECs.

Figure 6.

Kidney vascular phenotypes and proliferation index in Vhlh^KO are rescued by Osmr inactivation. A, Kidney sections of the four indicated mouse strains were double stained for Vhlh-mutant cells (GFP) and CD31. The vasculature shows dramatic morphologic changes with disorganized and swollen vessels in Vhlh^KO tissue as compared with others. This phenotype is rescued in DBKO. Bars, 20 μm. B and C, The percentages of CD31⁺ cells in whole kidney suspensions from the four genotypes (B) or from WT and Vhlh^KO pretreated or untreated with tofacitinib (Tofa; C) were quantified by flow cytometry. Each data point represents one animal. D, The four strains of mice as indicated were injected with FITC-dextran via tail vein. Significantly elevated diffused FITC signal is shown in Vhlh^KO tissue, which is not observed in all other genotypes. Right panel shows the quantification by colorimetric measurement using the ImageJ software. Bars, 20 μm. Each data point represents average of four 10× fields of view in one section from one animal. E, Kidney sections from the four mouse strains were double stained for Vhlh-mutant cells (GFP) and Ki-67. Red arrows point to Ki-67⁺ cells that are Vhlh mutant. White arrows point to Ki-67⁺ cells that are not. Right panel shows the quantification by counting total Ki-67⁺ cells and Ki-67⁺GFP⁺ cells per 10× field of view. Bars, 20 μm. Each data point is the average of four 10× fields of view in one section from one animal. The data are presented as mean ± SD, and the P values are calculated by one-way ANOVA (Tukey post hoc test). ns, no significance; *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Endothelial-to-mesenchymal transition signature of OSM-activated ECs

Endothelial-to-mesenchymal transition (EndoMT) has been demonstrated in various fibrotic conditions (13,14,15) and cancer progression (31). However, we have not been able to detect cells with dual endothelial–mesenchymal markers outside of the endothelium. While this may be because ECs outside of the endothelium rapidly lose the EC-specific markers such as CD31 and VE-cadherin, in our view, expression of EndoMT markers in ECs does not necessarily indicate transformation of the ECs into myofibroblast-like mesenchymal cells, but can result in reduced integrity of endothelial structure. Such activated state may be needed for vessel remodeling, allowing for immune-cell recruitment, extravasation, and wound healing or fibrosis (4).

To test this notion, we examined EC characteristics in the coculture system. When cocultured with wild-type HK-2, ECs remain a typical cobblestone-like monolayer morphology. In contrast, when cocultured with VHL KD HK-2, ECs become spindle-shaped (Fig. 4E), with accompanying overexpression of SNAIL1 (Fig. 4D). We also found loss of cell-junction integrity revealed by diffused VE-cadherin localization (Fig. 4E). FITC-conjugated dextran was added over the EC monolayer cocultured with wildtype HK-2 or VHL KD HK-2 cells (Fig. 4F). Coculturing with VHL KD HK-2 results in approximately 2-fold increase in FITC-dextran leakage to the lower chamber (Fig. 4F). The leakiness of EC monolayer was rescued by neutralizing OSM in the coculture system (Fig. 4F). When the endothelium was directly treated with increasing dosages of rOSM, the ECs expressed increasing amounts of mesenchymal markers SNAIL1 and SLUG proteins (Supplementary Fig. S8A). These gene-expression patterns are consistent with the change of EC monolayer morphology (Supplementary Fig. S8B) and VE-cadherin localization (Supplementary Fig. S8C). rOSM treatment also resulted in an approximately 2-fold increase in leakiness of the EC monolayer (Supplementary Fig. S8D). Finally, rOSM treatment resulted in an approximately 4-fold increase in the number of ccRCC cells 786-O invading through the endothelium (Supplementary Fig. S8E). The uncropped protein blot and quantification in Supplementary Fig. S8A are shown in Supplementary Fig. S9A–S9C.

Therefore, the OSM-activated ECs display cellular and molecular changes that are likely important for the protumorigenic role of these ECs in the microenvironment.

OSM-activated ECs induce macrophage polarization toward the M2 phenotype

The conditioned media from ECs with or without rOSM pretreatment were used to stimulate macrophages derived from THP-1 (partially differentiated to CD68⁺ M0 macrophages; Supplementary Fig. S10A). We also verified that OSM was not present in the conditioned media of activated ECs; therefore OSM had no direct activating functions on the monocytes/macrophages (Supplementary Fig. S10B and S10C). The populations of all macrophages, and classically activated (M1) and alternatively activated (M2) macrophages were detected by flow cytometry using CD68, CD86, and CD163 markers, respectively. Interestingly, conditioned media from activated ECs have no effect on the total number of CD68⁺ macrophages but significantly induce the population of CD163⁺ M2 (∼20–40-fold) compared with control (Supplementary Fig. S10D and S10E). The effect is specifically on M2 polarization because the M1 (CD86⁺) population is not increased.

Among the upregulated genes in the ECs-KO, IL6 is the most abundant cytokine (Supplementary Table S3), which is known to promote alternative activation of macrophages (32). Indeed, when anti-IL6 neutralizing antibody was added to the conditioned media of rOSM-activated ECs, the proportion of CD163⁺ macrophages was reduced to the control level (Supplementary Fig. S10D and S10E).

To directly compare the impact of IL6 and OSM on THP-1–derived macrophages, we treated these cells with recombinant IL6 (rIL6) or rOSM. As shown in Supplementary Fig. S11A, rOSM does not induce polarization of THP-1–derived macrophages. In contrast, rIL6 promotes THP-1–derived macrophages polarization toward the M2 type (CD163⁺) in a dosage-dependent manner (Supplementary Fig. S11B). Taken together, these results support the notion that OSM secreted by the epithelial cells does not activate the monocytic population directly but instead signals indirectly through activation of ECs.

The above analyses reveal a network of interacting cells in the tumorigenic microenvironment containing VHL-deficient tubule cells—i.e., VHL-deficient tubule cells activate ECs via OSM signaling, and ECs in turn produce at least IL6 to activate immune cells such as macrophages.

Osm signaling is important for the Vhlh-mutant phenotypes in vivo

To test the function of OSM signaling in vivo, we generated Vhlh-Osmr double mutant. A systemic KO strain of Osmr was used for the following reason: first, tissue-specific double KO requires a combination of at least 4 transgenic markers, which are cumbersome to maintain and the mice carrying double Cre transgenes are often unhealthy; second, Osmr homozygous mutant is viable and fertile with no overt developmental and health problems (10); and third, our results indicate that Osmr is expressed at a negligible level in the WT kidney but at a high level preferentially in the EC compartment in Vhlh^KO kidney.

The efficacy of Osmr KO was monitored by examining the Osmr protein levels in the kidney extracts from mice of the four different genotypes: Hoxb7-Cre-GFP (WT), Osmr^−/− (Osmr^−/–), Hoxb7-Cre-GFP; Vhlh^loxP/loxP (Vhlh^KO), and Hoxb7-Cre-GFP; Vhlh^loxP/loxP; Osmr^−/− [double KO (DBKO)]. The Osmr protein expression is eliminated in Osmr KO and in DBKO (Fig. 5A, top panel). There are no significant changes in the body weight and mobility of all 4 genotypes at 3 months of age (Fig. 5A, bottom panel), indicating that loss of Osmr function does not affect the general health of the animals. Also, Vhlh KO dramatically increased the levels of Osmr, GP130, and IL6rα as compared with WT (Fig. 5A; Supplementary Fig. S12A–S12C). In DBKO tissue, the increased expression of GP130 and IL6rα as a result of Vhlh KO is reduced to the WT level, while Lifr remains unchanged (Fig. 5A; Supplementary Fig. S12A–S12D). These results suggest the involvement of Osmr type II but not type I receptor in Vhlh^KO tissue.

Figure 5.

Osmr inactivation rescues the Vhlh KO phenotypes in vivo. A, Four mouse strains of WT, Osmr^−/–, Vhlh^KO, and Vhlh^KO-Osmr^−/– DBKO were compared for their body weights at 3 months of age (bottom). The expression of OSMR components (Osmrβ, GP130, IL6rα, and Lifr) in whole kidney extracts was monitored for validating the Osmr KO and potential feedback regulation (top). Uncropped blots are shown in Supplementary Fig. S12. B, The expression of p-Jnk in ECs of the four mouse strains shows an increase in Vhlh^KO and a rescue effect in DBKO. C, The sera of the four mouse strains show the amount of sSele is increased in Vhlh^KO compared with WT, Osmr^−/–, and DBKO. D, The numbers of macrophages (F4/80⁺) in the kidneys of the four mouse strains measured using flow cytometry (Supplementary Fig. S14) show increased infiltration in Vhlh^KO compared with WT, Osmr^−/–, and DBKO. E, Kidney sections of the four strains of mice were stained with the trichrome reagents. The amount of collagen fiber deposit (purple stain) was quantified by color metrics measurement using the ImageJ software (right panel). Each data point represents the average number per 10× view of one section from one animal. Bars, 50 μm. F, The concentration of Osm in sera of the four strains of mice. G-I, Quantification of the concentration of sSele in sera (G), the numbers of macrophages (H), and the amount of collagen fiber deposit (I) under pharmacologic treatment (tofacitinib, Tofa). Bars, 50 μm. The data are presented as mean ± SD, and the P values are calculated by one-way ANOVA (Tukey post hoc test). ns, no significance; *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Importantly, EC activation markers in the Vhlh^KO animals, including EC expression of p-Jnk (see Fig. 2A) and serum level of secreted selectin E (sSele), are ameliorated when Osmr is inactivated in DBKO (Fig. 5B and C). In addition, the expression of p-STAT3 (Y705), which is downstream of activated JAKs, is strongly expressed in ECs in Vhlh^KO but not in WT or Osmr^−/- tissue (Supplementary Fig. S13A). Similarly, in human ccRCC tissue strong expression of p-STAT3 is found in ECs of ccRCC tissue (both grade 1 and grade 3) but not in NAT (Supplementary Fig. S13B). Interestingly, expression of p-STAT3 is also seen in some tubule cells (white arrows, Supplementary Fig. S13A) and in ECs (yellow arrows, Supplementary Fig. S13A) in Vhlh^KO tissue; but in DBKO, p-STAT3 is lost in ECs but remains in some tubule cells. This is consistent with the notion that activation of ECs in Vhlh^KO is largely mediated via Osm signaling, while in the mutant epithelial cells, the inflammatory response is autonomously regulated via the intrinsic ER stress–induced unfolded protein response (9).

The extent of macrophage infiltration was monitored by counting the number of F4/80⁺ cells using flow cytometry, which is reduced in DBKO compared with that in Vhlh single KO (Fig. 5D; Supplementary Fig. S14). The fibrotic phenotype associated with Vhlh^KO is also rescued in DBKO (Fig. 5E).

The level of Osm in sera is below the detection limits of the ELISA assay in WT and Osmr^−/– mice, while it is significantly higher in Vhlh^KO and DBKO mice (Fig. 5F). This indicates that the serum level of Osm may serve as a diagnostic marker for ccRCC development.

In another course of the experiment, we used tofacitinib to block JAK-STAT signaling in Vhlh^KO tissue to test the possibility of pharmacologic treatment. As expected, tofacitinib can markedly alleviate observed phenotypes in Vhlh^KO tissue by reducing concentration of sSele in serum (Fig. 5G), the number of F4/80-positive cells (Fig. 5H), and fibrosis (Fig. 5I).

The vessels in Vhlh^KO kidney show dramatic swollen and disorganized morphology in areas both juxtaposing and away from the mutant tubules (Fig. 6A), consistent with the activated EC phenotypes in inflammatory or tumor tissues. This phenotype is rescued in the DBKO (Fig. 6A).

The levels of neoangiogenesis and neovascularization were examined by measuring the total number of CD31⁺ cells in the kidney suspension (Fig. 6B; Supplementary Fig. S15). The result shows that Vhlh^KO can increase the number of CD31⁺ ECs and introduction of Osmr KO ameliorates this phenotype. We also found that the angiogenic phenotypic of EC in Vhlh^KO tissue could be significantly reduced by tofacitinib (Fig. 6C).

The vascular permeability was then examined by measuring the amount of FITC-dextran in the kidney tissue after tail-vein injection. In Vhlh^KO kidney, there is a significant level of FITC signals leaking into the kidney tissue (Fig. 6D). The extent of FITC-dextran leakage is reduced to the normal level in DBKO (Fig. 6D). In addition, the number of Ki67-positive cells in both the Vhlh-mutant cell population (GFP⁺) and in the rest of the kidney is increased in Vhlh^KO, consistent with a hyperplastic phenotype described previously (8). This hyperplastic phenotype is completely rescued by Osmr KO (Fig. 6E). Note that the Ki-67–positive cells outside of the Vhlh-mutant tubules include both interstitial cells (probably proliferating immune cells and ECs) and those in normal tubules. This indicates that hyperplasia may be induced in genetically normal tubule cells by paracrine mitogenic factors secreted by other cell types in the microenvironment downstream of Osm signaling.

Microenvironment reconstituted by the Vhlh-mutant tubule cells promotes metastasis

Finally we tested whether the changes induced by OSM had any pathologic significance related to cancer progression. We used xenograft of an exogenous but syngeneic mouse melanoma cancer cell line B16 for this purpose (Fig. 7A). Human ccRCC cells cannot be used in this immune competent mouse model, and the existing mouse RCC cell line such as Renca is of a different genetic background (BALB/c) and does not grow in our Vhlh KO mice of the C57BL/6 background. In addition, the use of nonrenal cancer cells allowed us to test whether the OSM-activated ECs could have a general effect on tumor progression, not limited to autochthonous kidney cancer. The tumor cells were injected under the renal capsule using a Hamilton syringe as described in Supplementary Material and Methods, and we ensured that the single-cell suspension (compare Supplementary Fig. S16A with S16B) and viability (Supplementary Fig. S16C) of the injected cells were not affected. Subsequent nephrectomy examination also did not indicate leakage of the injected cells locally into normal tissues (see below). The tumor was harvested at day 14, and lung tissue was fixed and stained for GFP (marker gene transfected into B16) to quantify metastatic foci and area. Tumor growth was monitored by IVIS at day 7 and day 14 after implantation, and the actual tumor size was quantified by weight at day 14. As shown in Fig. 7B, B16 cells grow readily in the kidney of our mouse models and there is no difference in the growth rate of primary tumor, which may be because the high proliferative rate of these malignant cells cannot be further increased by the microenvironment. However, metastasis to lung is greatly enhanced in Vhlh^KO animals, and is reduced to the WT level in DBKO (Fig. 7C and D). Local invasiveness of the injected B16 tumor cells is greatly increased in Vhlh^−/− background, to the extent that no clear boundary between tumor and normal tissue can be drawn. The invasiveness is reverted to normal in DBKO (Fig. 7E). We therefore conclude that the microenvironment reconstituted by VhlhI-mutant cells can not only facilitate tumor initiation, but also promote metastasis of malignant cells, and that this property is dependent on OSM signaling. The metastasis-promoting activity of the Vhlh^−/− microenvironment correlates with the intratumoral vascular density (Fig. 7E).

Figure 7.

Vhlh^KO-reconstituted kidney microenvironment promotes metastasis. A, Mouse B16 melanoma cells (marked with GFP and luciferase) were injected under the renal capsule of the four strains of mice. Bar, 100 μm. B, Luciferase-carrying tumors were imaged by IVIS at 7 days after implantation. Left, representative images. Right panels show no difference in the luciferase activities (7 days) and in weights of the resected primary tumors (14 days) from the four strains of mice. Each data point represents one animal. C, Images of lungs from the tumor-carrying mice. Metastasized B16 foci (black spots) can be clearly seen. The number of foci on the surface is significantly higher in Vhlh^KO mice and can be brought down to the WT level by Osmr inactivation. D, Lung sections from the four mouse strains stained for GFP (marking the metastatic B16 cells). Bars, 100 μm. The average area of metastasized cancer per section is significantly higher in Vhlh^KO than in other strains. Each data point represents one section from one animal. E, Local invasiveness and intratumoral vascular density of the injected B16 tumor cells in different host genetic backgrounds. White dash lines mark the boundaries of tumors (T) and normal host tissues (N). Right, quantification of intratumoral ECs (CD31⁺). Each data point represents one section from one animal. The data are presented as mean ± SD, and the P values are calculated by one-way ANOVA (Tukey post hoc test). ns, no significance; **, P < 0.01; ***, P < 0.001.

The role of OSM signaling in promoting metastasis was further tested using a xenograft model of human ccRCC cell 786-O. Implantation of 786-O into kidney tissue of immune-compromised NOD/SCID mice (NOD.CB17-Prkdc^scid/NcrCrl) was conducted (Fig. 8A). In this malignant tumor xenograft model, anti-OSM antibody treatment (Fig. 8A) does not affect primary tumor growth (Fig. 8B–D) but can inhibit metastasis (Fig. 8C and E), thus in agreement with the result with the B16 model described above.

Figure 8.

Metastasis of human ccRCC cells is inhibited by blocking Osm in vivo. A, Schematic illustrating the experimental design. B, Percentage growth of primary tumor in animal injected (αOSM) or uninjected (IgG) with antihuman OSM antibody over time. Day 0 is set at 100%. There is no difference in primary tumor growth with or without blocking OSM. C, Tumor weights at day 28 comparing animals treated or untreated with αOSM. D, IVIS imaging at day 28 in prone position (left) and supine position (right) of treated and untreated with αOSM. E, The invasiveness of tumor cells and intratumor vascular density in primary tumor of treated and untreated with αOSM. Tumor cells (786-O^LUC) were detected with human mitochondria (red), and ECs (CD31⁺) are in green. White dash lines mark the boundaries of tumor (T) and normal host tissues (N). Right, quantification of intratumoral vascular density (CD31⁺). Bars, 100 μm. F, The metastasis of 786-O^LUC treated or untreated with αOSM in lung tissue similarly stained as in E. Bars, 50 μm. Intratumoral vascular density, number of metastasis foci, and average metastasis foci area were quantified by ImageJ. Each data point represents one animal. The data are presented as mean ± SD, and the P values are calculated by two-tailed unpaired Student t test. *, P < 0.05 **, P < 0.01.

Discussion

The influence of the inflammatory components on progression of ccRCC has been noted (1, 2). Our mouse model of Vhlh conditional KO in a subset of kidney tubule cells also exhibits highly penetrant inflammatory phenotypes, adenoma, and hyperplasia (Fig. 1). During analysis of the inflammatory phenotypes in the Vhlh^KO mice, we noted expression of inflammatory marker p-Jnk in the endothelial compartment of the KO kidney tissue, although these ECs were not genetically manipulated, indicating that these ECs are activated by secreted factors emanating from the Vhlh-mutant cells. In this report, we analyzed the transcriptome of primary ECs from the Vhlh^KO mouse kidney. Our results show that the gene expression program in ECs, a quiescent component in normal tissues, is profoundly altered in kidney tissue containing Vhlh deletion in the tubule cells (Supplementary Fig. S1 and S2). The gene expression program in ECs-KO is consistent with the observed tissue fibrosis and vascular abnormalities in Vhlh^KO kidney (Figs. 1 and 2).

Activated ECs in Vhlh^KO mice overexpress Sele, Selp, Vcam1, and Icam1. Sele and Selp are critical molecules for capturing leukocytes, and Vcam1 and Icam1 are important for arresting and promoting crawling of captured leukocytes before extravasation into target tissue. Also, there is increased expression of the chemoattractants Ccl2 and Ccl7 in ECs-KO, indicating that Vhlh-mutant epithelial cells induce the activation in ECs and activated ECs, in turn, recruit circulating immune cells. This finding also explains the increased infiltration of immune cells surrounding the Vhlh-mutant tubules (Fig. 1).

The transcriptome results show that activation of ECs in kidney tissues is mediated mainly through the OSM pathway. OSM is overexpressed and secreted by the VHL-deficient kidney tubule cells and the receptor OSMR is expressed and induced by OSM in the ECs thus establishing an intercellular feed-forward signaling mechanism (Figs. 3, 4D, and 5A). Significantly, deletion of Osmr in Vhlh^KO could rescue all the inflammatory and hyperplastic phenotypes, including immune-cell infiltration, tissue fibrosis, neoangiogenesis, vascular permeability, and proliferation index (Figs. 5 and 6). It is interesting to note that Osmr KO in Vhlh^KO mice could normalize the vascular density of Vhlh^KO kidney to the basal level (Fig. 6). This is somewhat unexpected since the neoangiogenic phenotype in VHL/Vhlh-mutant tissues is normally attributed to the VHL-HIF-VEGF signaling axis. It is possible that the VEGF action may require priming by OSM signaling; e.g., in activating migratory and/or EndoMT capacity. Whether OSM plays a dominant role in neoangiogenesis in ccRCC, at least at the very early stage, requires further studies.

On the other hand, OSM has been shown to induce EndoMT and contribute to fibrosis (33). Our data demonstrate that OSM induces expression of mesenchymal transition markers such as SNAIL1 and SLUG in ECs and alters the membrane localization of VE-cadherin (Fig. 4; Supplementary Fig. S8). Attenuating the OSM pathway diminishes fibrosis in the Vhlh^KO tissue (Fig. 5). It is therefore likely that in our system, the expression of EndoMT markers promotes loosening of the barrier function of blood vessels while inducing overproduction of matrix proteins such as collagen 1 and fibronectin in situ (Fig. 1; Supplementary Fig. S2A and S2B).

Furthermore, recruitment and alternative activation of macrophage promoted by OSM signaling is in large part mediated via IL6 produced by activated ECs (Supplementary Fig. S10D and S10E). As such, our findings describe a network of reciprocal, cross-talking constituent cells in the tumor microenvironment that is important for ccRCC formation.

Most importantly, the microenvironment generated by the Vhlh-mutant cells can promote metastasis of an exogenous cancer cell line in Vhlh^KO host, and this property is dependent on the function of OSM signaling (Fig. 7). The Osm-dependent metastasis potential of a human ccRCC line is also demonstrated in immune-deficient host (Fig. 8). There have so far been only a few reports describing the cancer-promoting function of OSM signaling (27, 34), and those few reports mostly describe the autocrine action of OSM secreted by the cancer cells in vitro. Our report here is a rare example that takes into account the tissue context of OSM and OSMR expression in vivo, and describes the important mechanism of tumor cell–EC cross-talk that is relevant to tumor formation. Since OSM-signaling functions in only a limited number of tissues (25) and mice with systemic KO of Osmr are viable and fertile (10), inhibition of OSM signaling may represent an amenable therapeutic option for ccRCC and other cancers.

The above observations are summarized in the Graphic Abstract. Our discovery demonstrates that microenvironment reconstituted by the original VHL genetic defect in the epithelial cells is a critical factor in ccRCC progression, and provides a useful framework for further elucidation of the molecular and cellular mechanisms involved in this process.

Authors' Disclosures

No disclosures were reported.

Authors' Contributions

H.-H. Nguyen-Tran: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. T.-N. Nguyen: Investigation, methodology. C.-Y. Chen: Investigation, methodology. T. Hsu: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, project administration, writing–review and editing.