Overexpression of factor inhibiting HIF-1 enhances vessel maturation and tumor growth via platelet-derived growth factor-C

Authors


Abstract

Recent studies have revealed that the maturation state of vessels in tumors, in addition to vascularity, is a critical determinant of tumor growth. The role of oxygen-dependent signaling pathways in hypoxia-stimulated angiogenesis is well established, however, little is known about their impact on vessel maturation in tumors. Here, we have studied the function of the cellular oxygen sensor, factor inhibiting HIF-1 (FIH), which controls the activity of hypoxia-inducible factor-1. FIH silencing in mouse LM8 osteosarcoma stimulated angiogenesis but did not influence tumor growth. In contrast, FIH overexpression led to increased pericyte coverage of the tumor vasculature, reduced vessel leakiness and enhanced tumor growth. Vessel maturation was paralleled by up-regulation of platelet-derived growth factor (PDGF)-C in tumors and expression of PDGF receptor-α on pericytes. Ablation of PDGF-C in FIH-overexpressing tumor cells reduced pericyte coverage and tumor growth. Our data suggest that FIH-mediated PDGF-C induction in LM8 osteosarcoma stimulates the recruitment of PDGFR-α positive pericytes to the tumor vasculature, leading to vessel maturation and enhanced tumor growth.

Solid tumors need an adequate supply with oxygen and nutrients to grow beyond a size of approximately 1 mm in diameter.1–3 Therefore, tumors can induce the formation of a functional vascular network for efficient blood delivery, through a process termed angiogenesis. However, the tumor vasculature displays structural and functional abnormalities, such as dilations, incomplete endothelial lining, arteriovenous shunts, blind ends and leakiness.1, 4, 5 These abnormalities lead to irregular blood flow resulting in low oxygen tension (hypoxia) in tumor tissue. Hypoxia consequently evokes changes in the expression of genes promoting cell survival, angiogenesis and anaerobic metabolism.6 Hypoxia-inducible factor (HIF)-1 is the key transcriptional regulator of the cellular response to hypoxia and has been implicated in tumor progression.7 Malignant lesions such as breast carcinoma, prostatic intraepithelial neoplasia and colonic adenoma were found to overexpress HIF-1; this emphasizes the importance of HIF in cancer growth and its impact on patient survival.8, 9 Consistently, HIF-1 ablation in different tumor models resulted in slower tumor growth.10, 11

HIF-1 and other members of the HIF family consist of an oxygen-regulated HIF-α subunit and a constitutive HIF-β subunit. The stability of the HIF-α subunit is regulated by enzymes known as prolyl hydroxylase domain proteins (PHDs). Under normoxic and mild hypoxic conditions, PHDs hydroxylate conserved prolyl residues of the HIF-α subunit, resulting in its rapid proteasomal degradation.12, 13 PHDs have recently been implicated in tumor progression. PHD1 acts as tumor suppressor in colon carcinoma, and PHD2 in human colon and pancreatic carcinoma.14, 15 However, we have recently shown that PHD2 silencing in murine LM8 osteosarcoma and Lewis lung carcinoma (LLC) can also inhibit tumor growth due to the anti-proliferative effect of TGF-β.16

The transcriptional activity of HIF represents another level of control exerted by the oxygen-dependent enzyme, factor inhibiting HIF (FIH). FIH hydroxylates an asparaginyl residue within the HIF-α subunit, leading to a steric clash that prevents the recruitment of the transcriptional co-activators p300 and CBP.13, 17 FIH remains active at even lower oxygen concentrations than PHDs and might therefore suppress the activity of HIF-α molecules that escape destruction in moderate hypoxia.18, 19 The biological role of FIH in mouse development and physiology has been described recently. Mice lacking a functional hif1an gene encoding FIH exhibit reduced body weight, elevated metabolic rate, as well as improved glucose and lipid homeostasis, which demonstrates its essential role in the regulation of metabolism.20 This study revealed also that FIH has no discernable role in altering the classical aspects of HIF function, in particular angiogenesis and erythropoiesis. However, little is known about the role of FIH in tumors. FIH expression has been demonstrated in human renal cell carcinoma, follicular lymphoma, non-small cell lung carcinomas and invasive breast cancer.21–23 The increase in FIH immunostaining observed in certain neoplasia compared to corresponding normal tissues shows that FIH is regulated not only at the level of activity but also at the level of protein expression.21 The currently available knowledge about FIH function in tumor cells is based on in vitro studies only, performed with several human cancer cell lines.18, 24 These analyses confirm that FIH regulates HIF-1α transcriptional activity, but not HIF-1α protein stability. Silencing of FIH by RNA interference induced HIF-1α-dependent target gene expression, whereas FIH overexpression had the opposite effect. FIH can interact also with proteins other than HIF, e.g., the von Hippel-Lindau protein, histone deacetylases, certain ankyrin-repeat domain proteins (e.g., IkBα, p105, Notch1, -2 and -3) and SOCS box protein 4.25–28 This indicates that FIH has also HIF-independent functions. However, the relevance of these interactions remains largely unclear.

To investigate the role of FIH during tumor growth, we performed FIH loss-of-function and gain-of-function experiments. FIH inhibition in murine LM8 osteosarcoma cells did not influence tumor growth although it induced angiogenesis. In contrast, FIH overexpression accelerated tumor growth, an effect that was accompanied by enhanced maturation of tumor blood vessels, as indicated by increased pericyte coverage and reduced leakiness. This correlated with up-regulation of platelet-derived growth factor (PDGF)-C and its cognate receptor, PDGFR-α; and silencing of PDGF-C in FIH-overexpressing tumor cells reduced tumor growth and vessel maturation. Our results suggest that this signaling system induces vessel maturation and normalization in FIH-overexpressing tumors, resulting in increased tumor growth.

Abbreviations

FIH: factor inhibiting HIF-1; LLC: Lewis lung carcinoma; NEAA: non-essential amino acid; PDGF: platelet-derived growth factor; PECAM: platelet/endothelial cell adhesion molecule; PHD: prolyl hydroxylase domain protein; RT-PCR: reverse transcriptase-polymerase chain reaction; α-SMA: α-smooth muscle actin

Material and Methods

Cell culture

Murine LM8 osteosarcoma cells (kind donation of Dr. C. Beltinger, Ulm, Germany) were cultured in MEM-α medium (Gibco, Darmstadt, Germany) supplemented with 10% fetal bovine serum, 1% non-essential amino acids (NEAA) and 1% L-glutamine. Hypoxia experiments were performed as described.16

Gene silencing

FIH-silenced LM8 clones (LM8-shFIH) were generated as follows: oligonucleotides encoding two independent short hairpin RNA (shRNA) sequences specific for FIH (shFIH#1: 5′-TTATACCAGAAGTTCACAG-3′ and shFIH#2: 5′-GGGAGGAAATTAAATTTCA-3′) were ligated into the lentiviral vector pLVTHM. Packaging of the vector was achieved by cotransfection of 293T cells with pLVTHM, psPAX2 and pMD2.G (kind donation of Dr. D. Trono, Geneva, Switzerland). Subsequently, LM8 cells were transduced with recombinant lentiviral particles, and stable clones expressing the pLVTHM-encoded green fluorescent protein were selected. To generate control clones, the scramble sequence (shScr) 5′-AGTCGCTTAGAAACGAGAA-3′ was used which possesses no significant homology to murine sequences.

To generate FIH-overexpressing LM8 clones in which PDGF-C was silenced (LM8-pcFIH + shPDGF-C), clone LM8-cl.11 (pcFIH) (see below) was transduced with recombinant lentiviral particles. The oligonucleotide encoding shRNA sequence specific for PDGF-C was 5′-GACGATATATGCAAGTATG-3′. Control clones were generated as described above.

Generation of cell lines stably overexpressing FIH (LM8-pcFIH)

Full-length FIH cDNA was amplified from murine embryonic endothelial progenitor cells by reverse transcriptase-polymerase chain reaction (RT-PCR) using primers 5′-CGTCCCTAGAGTAGAGATG-3′ and 5′-CCGTTACAACTAACCTGCC-3′ and cloned into pcDNA3.1 vector (Invitrogen, Karlsruhe, Germany). LM8 cells were transfected with the FIH expression vector or empty control vector, and neomycin resistant clones were selected in medium supplemented with 0.5 mg/mL of G418 (Sigma, Munich, Germany).

In vitro proliferation assay

Fifty thousand cells were plated in triplicates in T25 flasks in complete culture medium. After 24–48 hr, cells were detached with trypsin and counted using a CASY cell counter (Innovatis, Bielefeld, Germany).

Mice and tumor models

2 × 106 LM8 tumor cells each were injected subcutaneously into both flanks of female C3H mice (University of Technology, Dresden, Germany). Every 2–3 days, the tumor volume was determined as (length × width2)/2. Tumors were collected before they reached the maximum ethically allowed volume and either frozen for RNA and protein preparation or embedded in OCT (Tissue-Tek, Sakura, Staufen, Germany) for histological examination. LLC sections were obtained as described.16

RNA isolation and reverse transcription (RT)-PCR

RNA isolation and RT-PCR analysis were performed as described previously.29 Sequences of PCR primers are shown in Supporting Information Table 1. The intensity of the bands was quantified using BioRad densitometer and Quantity One analysis software (Hercules, CA, USA).

Immunoblot analysis

Cell or tissue lysates were separated on 10% NuPAGE® gels and transferred by semi-dry blotting onto nitrocellulose membrane (Whatman, Dassel, Germany). Membranes were incubated with primary antibodies for FIH (sc-26219, Santa Cruz, Heidelberg, Germany), HIF-1α (10006421, Cayman, Michigan, USA), HIF-2α (NB-100-122, Novus, Littleton, CO, USA), GLUT1 (NB300-666, Novus, Littleton, CO, USA), PHD2 (NB100-2219, Novus, Littleton, CO, USA), PDGFRα (3164, Cell Signalling, Frankfurt, Germany), PDGF-C (AF-1447, R&D systems, Wiesbaden, Germany), β-tubulin (RB-9249-P1, Neomarkers, Fremont, CA, USA) and β-actin (A5060, Sigma, Munich, Germany), followed by incubation with horseradish peroxidase-conjugated secondary antibodies. Proteins were visualized with enhanced chemiluminescence detection system (Thermo Scientific, Kalamazoo, MI, USA). Quantification was performed using Quantity One analysis software.

Immunohistochemistry and histology

Before tumor isolation, mice were injected intravenously with Hoechst33342 dye (Sigma, Munich, Germany) to determine vessel perfusion. Ten micrometer frozen sections were stained for platelet/endothelial cell adhesion molecule (PECAM) to visualize blood vessels.30 Perfused vessels were identified as double positive for PECAM and Hoechst33342. Vessel density was evaluated by counting PECAM-positive vessels per viable tumor area. Staining of pericytes was performed with antibodies for α-SMA (F3777, Sigma, Munich, Germany), desmin (ab15200, Abcam, Cambridge, MA, USA) and NG2 (AB5320, Millipore, Schwalbach, Germany). Vessel coverage was calculated as percentage of pericyte-covered vessels compared to the number of PECAM-positive vessels. Immunofluorescence staining was also performed to detect PDGF-C (AF-1447, R&D systems, Wiesbaden, Germany) and PDGFR-α (3164, Cell Signalling, Frankfurt, Germany; cross-reactivity with PDGFR-β according to the manufacturer). Proliferating tumor cells were visualized by Ki67 staining (DAKO, Hamburg, Germany). Light microscopy was performed using Axioplan-2 microscope (Carl Zeiss, München, Germany). Images were captured with Axiocam MRc5 camera and AxioVision AC Rel. 4.5 acquisition software (Carl Zeiss, München, Germany). Counting of vessels was performed using the NIH ImageJ software.

Detection of tumor hypoxia, necrosis and vessel leakiness

To detect hypoxic areas in the tumors, mice were injected intraperitoneally with 60 mg/kg pimonidazole (Chemicon, Hofheim, Germany) prior tumor collection. Pimonidazole adducts were immunostained on cryosections with Hypoxyprobe-1-Mab1 (Hypoxyprobe kit, Chemicon, Hofheim, Germany) according to the manufacturer's instructions. Tumor sections were stained with hematoxylin and eosin (H&E) to identify necrotic areas. H&E positive areas were considered as viable tumor area. Vessel leakiness was analyzed after intravenous injection of 0.25 mg Texas Red-conjugated Dextran 70 kD (Molecular Probes, Darmstadt, Germany).31 Ten minutes later, mice were perfused by intracardiac perfusion (20–25 mL in 5 min) of PBS and 2% paraformaldehyde. Tumors were harvested and frozen in OCT medium.

Statistics

Statistical analysis was performed using statistical software SigmaPlot 12 (Systat software). Statistical significance was evaluated by t-test, with p value < 0.05 considered as statistically significant.

Results

FIH inhibition stimulates angiogenesis but does not influence LM8 tumor growth

To study the function of FIH during cancer progression, we suppressed FIH expression in murine LM8 osteosarcoma cells by lentiviral transduction of LM8 cells with two independent shRNA sequences (shFIH#1 and #2). shFIH#2 showed the most profound silencing effect [Fig. 1a; clone (cl.) 7 and cl.8]. To explore whether FIH silencing affects the growth of these cell clones in vitro, we performed cell proliferation assays. No significant difference in cell proliferation was observed between the two LM8-shFIH clones (cl.7 and cl.8) and control clones containing scramble shRNA (shScr cl.2 and cl.3), irrespective of whether they were cultured in normoxic (21% O2) or hypoxic (1% O2) atmosphere (Supporting Information Fig. 1a, and data not shown). To examine the growth of these cell clones in vivo, we injected them into immunocompetent C3H mice. Again, no significant difference in tumor growth between the two LM8-shFIH clones and control clones (LM8-shScr) was observed (Fig. 1b). Western blot analysis of tissue lysates prepared from tumors isolated 3 weeks after inoculation confirmed that FIH expression was still down-regulated (Supporting Information Fig. 1b).

Figure 1.

FIH inhibition does not influence LM8 tumor growth. (a) LM8 tumor cells transduced with lentivirus encoding shFIH RNA were tested for FIH expression by Western blot. β-actin was used as loading control. Arrows point to two efficiently silenced clones (cl.7 and cl.8). (b) LM8-shFIH cl.7 and cl.8 and control LM8-shScr cells were inoculated in C3H mice, and tumor growth was measured every 2–3 days. The mean tumor volume (mm3) ± SEM is presented. Ten tumors (n = 10, n-number of samples) per group were analyzed. (c) PECAM staining of vessels on sections of LM8-shFIH tumors (cl.7 and cl.8) and control LM8-shScr (cl.2) tumors. Only one LM8-shScr control is shown because vessel density in both clones was highly similar. Vessel density was determined by counting vessel number per tumor area. Error bars represent mean ± SEM (n = 5) (***p < 0.001). Scale bar represents 100 μm. (d) PECAM-staining of vessels (red) and Hoechst33342-staining showing tissue perfusion (blue). PECAM+/Hoechst+ staining indicates perfused vessels, PECAM+/Hoechst− staining non-perfused vessels (n = 5). Bar, 100 μm.

Although their growth was not altered, LM8-shFIH tumors displayed a significantly increased vessel density (>45%) compared to scramble control tumors (Fig. 1c). This observation contrasts with numerous reports describing a positive correlation between tumor angiogenesis and growth3 and suggests that increased angiogenesis in FIH-silenced tumors does not enhance the overall functionality of the tumor vasculature compared to control tumors. To address this issue, we injected Hoechst33342 dye before tumor isolation and evaluated the percentage of perfused vessels per area. We found no difference in perfusion between both groups, >95% of the vessels were perfused (Fig. 1d). In line with the unaltered tumor growth, staining of tumor sections for the proliferation marker Ki67 failed to reveal a significant difference in the proliferation rates of FIH-silenced and control tumors (Supporting Information Fig. 1c). In addition, LM8-shFIH and LM8-shScr tumors showed a similar extent of necrosis (Supporting Information Fig. 1d).

FIH overexpression accelerates tumor growth

Next, we asked whether overexpression of FIH influences tumor growth. Therefore, LM8 cells were stably transfected with pcDNA3.1 expression vector encoding mouse FIH (pcFIH). Two independent LM8-pcFIH clones (cl.11 and cl.13) showed abundant FIH overexpression (Fig. 2a). As expected, elevated FIH levels did not influence HIF-1α protein expression when cells were cultured at 21% O2, 5% O2 or 1% O2 (Fig. 2a). Consistent with inhibition of HIF-1 activity by FIH, expression of certain HIF target genes, such as carbonic anhydrase 9, was reduced in mild hypoxia (5% O2, data not shown). HIF-2α was not detected under these conditions (data not shown).

Figure 2.

FIH overexpression enhances LM8 tumor growth. (a) LM8 cells stably transfected with pcFIH expression vector (cl.11 and cl.13) and control cells (pcEmpty cl.4 and cl.6) were cultured at different O2 levels for 24 hr and tested for expression of FIH and HIF-1α protein. (b) LM8-pcFIH (cl.11 and cl.13) or LM8-pcEmpty cells were injected into C3H mice and tumor size was measured every 2–3 days (n = 10) (*p < 0.05 and **p < 0.01). (c) Ki67 staining of proliferating cells on sections from LM8-pcFIH and LM8-pcEmpty control tumors (n = 5) (***p < 0.001). Bar, 50 μm.

To evaluate the effect of FIH overexpression on tumor growth, LM8-pcFIH clones cl.11 and cl.13 were inoculated into C3H mice. Both FIH-overexpressing clones grew significantly faster than the LM8-pcEmpty control clones (transfected with empty pcDNA vector) (Fig. 2b). Consistently, Ki67 staining of tumor sections revealed a clear increase in the number of proliferating cells in FIH-overexpressing tumors (Fig. 2c). Immunoblot analysis of representative tumor lysates isolated 16 days after inoculation (post injection, p.i.) of tumor cells, confirmed elevated FIH levels in these tumors (Fig. 3b). Unexpectedly, HIF-1α protein levels were elevated at this time point in tumors overexpressing FIH. In line with this observation, the expression of the HIF-regulated PHD2 and glucose transporter 1 (GLUT1) proteins, and of mRNA for CA9, ADM and Hk-2 was increased (Fig. 3b and Supporting Information Fig. 2a).

Figure 3.

HIF-1α induction in LM8-pcFIH tumors at Day 16 p.i. is paralleled by the occurrence of tumor hypoxia adjacent to necrotic areas. (a) Western blot detection of FIH, HIF-1α, GLUT1 and PHD2 in lysates of LM8-pcFIH (cl.11 and cl.13) and control LM8-pcEmpty tumors isolated 8 days p.i. (n = 3). LM8 cells cultured in 1% O2 were used as positive control for HIF-1α detection [indicated by (+)]. (b) Western blot analysis for FIH, HIF-1α, GLUT1 and PHD2 in lysates (Day 16 p.i.) from LM8-pcFIH or LM8-pcEmpty clones (n = 4) (*p < 0.05 and ***p < 0.001). β-actin levels were used as loading control. (c) H&E staining of sections from LM8-pcFIH and LM8-pcEmpty tumors (Day 16 p.i.) (n = 5). Necrosis was evaluated using ImageJ software (**p < 0.01). (d) Sections of LM8-pcFIH (cl.11 and cl.13) and LM8-pcEmpty (cl.4 and cl.6) tumors stained for PECAM (vessels, green), Hoechst33342 (perfused vessels, blue) and hypoxyprobe (red) (n = 5). Bar, 100 μm.

Next, we analyzed whether HIF-1α protein level and activity are elevated already in early stage tumors, when changes leading to altered tumor growth should be recognizable. At Day 8 p.i., we could not detect HIF-1α protein (Fig. 3a), and the expression of GLUT1 and PHD2 proteins was not altered (Fig. 3a). This might suggest that HIF-1 does not cause the initial growth advantage of FIH-overexpressing tumors, in line with our observation that HIF-1 acts as tumor suppressor, rather than as tumor promoter in this model.16 At Day 8 p.i., FIH-overexpressing and control tumors displayed a similar extent of necrosis (Supporting Information Fig. 2b). However, the necrotic areas were small, in line with our inability to detect HIF-1α protein in tumor lysates (Fig. 3a). The observation that HIF-1α levels were up-regulated in FIH-overexpressing tumors at later stages led us to analyze the extent of necrosis and hypoxia in 16-day p.i. tumor tissue. LM8-pcFIH tumors displayed large and significantly more necrotic areas compared to control tumors (Fig. 3c); the necrotic areas were surrounded by hypoxic regions, as shown by staining with hypoxyprobe (Fig. 3d). This suggests that elevated HIF-1α levels are a consequence of increased tissue hypoxia in 16-day p.i. tumors.

FIH overexpression in tumors enhances vessel maturation and normalization

In vitro proliferation assays showed that FIH overexpression did not accelerate the growth rate of LM8 cells (Supporting Information Fig. 3a). It is therefore likely that the increased growth of LM8-pcFIH tumors is the result of a cross-talk between tumor cells and host tissue. Therefore, we investigated the density and functionality of the tumor vasculature. PECAM staining of endothelial cells revealed no difference in vessel morphology and density between FIH-overexpressing and control pcEmpty tumors both at Day 8 p.i. (data not shown) and Day 16 p.i. (Supporting Information Fig. 3b).

Other vascular parameters known to influence tumor growth are vessel maturation and functionality.31–33 Therefore, we stained tumor sections for the pericyte markers α-smooth muscle actin (α-SMA), desmin and NG2. Vessels in LM8-pcFIH tumors were covered by pericytes which were positive for α-SMA, but not for desmin or NG2 (Fig. 4a and Supporting Information Fig. 4). This expression pattern is indicative of a less mature pericyte phenotype.34–37 Pericyte coverage changed during the progression of FIH-overexpressing tumors (Fig. 4a). At Day 8 p.i., LM8-pcFIH tumors showed only a slight increase in the percentage of α-SMA-positive vessels compared to the controls. However, at Days 11 p.i. and 16 p.i., the proportion of pericyte-covered vessels was significantly increased in FIH-overexpressing tumors. Thus, LM8-pcFIH tumors seem to harbor more mature, normalized vessels with improved functional capacity. To test this hypothesis, we analyzed vessel leakiness. After injection of tumor-bearing mice, Texas Red-conjugated dextran never drained off from α-SMA-positive vessels (Supporting Information Fig. 5), whereas a proportion of α-SMA-negative vessels in both groups showed dextran extravasation (Fig. 4b). However, the percentage of dextran-positive vessels was significantly lower in FIH-overexpressing tumors (23 ± 2%) compared to control tumors (37 ± 3%) (Fig. 4c). Therefore, while vessel density was not altered in LM8-pcFIH tumors, the vasculature was more mature, as indicated by increased pericyte coverage and reduced vessel leakiness.

Figure 4.

FIH overexpression leads to increased tumor vessel maturation. (a) Immunofluorescence co-staining for PECAM (red, endothelial cells) and α-SMA (green, pericytes). The percentage of α-SMA positive vessels is indicated (n = 5) (***p < 0.001). Bar, 100 μm. (b) PECAM-staining of blood vessels (green). On injection of Texas Red-conjugated dextran (red), the number of leaky vessels was determined in both FIH-overexpressing (cl.11 and cl.13) and control tumors (pcEmpty) (n = 5). Orange arrows on merged images point to leaky vessels. Bar, 40 μm. (c) The percentage of dextran-positive vessels was determined by comparing the number of PECAM+/dextran+ vessels and PECAM+/dextran− vessels. Bars represent mean ± SEM (n = 5) (***p < 0.001).

Vessel maturation in FIH-overexpressing tumors is accompanied by PDGF-C induction

The Ang/Tie2 and PDGF/PDGF-receptor pathways are known to influence tumor vessel maturation.32, 33, 37–42 Therefore, we examined the expression of the respective receptors and ligands. By RT-PCR, Ang-1 mRNA expression was not detected in LM8-pcFIH tumors or control tumors (data not shown). Transcript levels of Ang-2, which is known to destabilize blood vessels, were similar at 8-day p.i. in LM8-pcFIH and control tumors, and at Day 16 p.i. its expression was reduced in tumors of only one of the two LM8-pcFIH clones analyzed (cl.13), compared to control tumors (Supporting Information Fig. 6a). It is therefore unlikely that Ang-2 has a major impact on vessel maturation in LM8-pcFIH tumors. Of the PDGF family members, only PDGF-C mRNA was increased in FIH-overexpressing tumors (Supporting Information Fig. 6b). Moreover, the protein levels of PDGF-C and its cognate receptor, PDGFR-α, were significantly increased in LM8-pcFIH tumors both at Day 8 p.i. and 16 p.i. (Figs. 5a–5c). Up-regulation of PDGF-C and PDGFR-α was also detected in LM8-pcFIH tumor cells in vitro (Supporting Information Fig. 6c). Our observation that tumor cell proliferation in vitro was not increased (Supporting Information Fig. 3a) suggests that this signalling system stimulates LM8-pcFIH tumor growth not via an autocrine mechanism, but rather indirectly via a paracrine regulation mechanism.

Figure 5.

PDGF-C induction parallels vessel maturation. (a) Immunoblot for PDGF-C and PDGFR-α (Day 8 and 16 p.i.) on tumor lysates from both FIH-overexpressing (cl.11 and cl.13) and control (pcEmpty) tumors (n = 3). (b) PDGF-C protein band intensity from tumor lysates from FIH-overexpressing and control tumors isolated at 8 days p.i. and 16 days p.i. Bars represent mean ± SEM (n = 3) (***p < 0.001). (c) PDGFR-α protein band intensity from FIH-overexpressing and control tumors isolated at Day 8 and 16 p.i. (n = 3) (***p < 0.001). (d) Immunofluorescence double staining for PDGF-C (red) and PECAM (green) (n = 5). Orange arrows on merged images indicate sites of co-localization. Bar, 40 μm. (e) Immunofluorescence co-staining for PDGFR-α (red) and α-SMA (green) to detect mature vessels (n = 5). Orange arrows point to sites of co-localization. Bar, 20 μm.

In line with this hypothesis, several studies point out that overexpression of PDGF-C in tumors results in vessel maturation.35, 36 By immunostaining of tumor sections, we found that PDGF-C is localized to tumor and endothelial cells in FIH-overexpressing and control tumors (Fig. 5d). In both tumor groups, expression of PDGFR-α and PDGFR-β was stronger in α-SMA-positive pericytes than in tumor cells (Fig. 5e). Taken together, our results and those of others suggest that endothelial-pericyte interaction is mediated through PDGF-C and its cognate receptor, PDGFR-α, in tumors.

PDGF-C knockdown reduces vessel maturation and growth of FIH-overexpressing tumors

To test the hypothesis that PDGF-C stimulates the recruitment of pericytes to the vasculature of FIH-overexpressing tumors and enhances tumor growth, we silenced PDGF-C in an FIH-overexpressing clone. The resulting clone (LM8-cl.11 (pcFIH) + shPDGF-C); Fig. 6a) was injected into mice. It showed reduced growth compared to the control clone (LM8-cl.11 (pcFIH) + shScr) (Fig. 6b). Moreover, the percentage of mature vessels was significantly lower in LM8-cl.11 (pcFIH) + shPDGF-C) tumors (3.4 ± 0.6%) than in control tumors (14.6 ± 0.5%) (Fig. 6c). These results demonstrate that PDGF-C enhances vessel maturation in FIH-overexpressing tumors, which consequently leads to improved tumor growth.

Figure 6.

PDGF-C knockdown inhibits vessel maturation and growth of FIH-overexpressing tumors. (a) LM8-pcFIH cl.11 tumor cells transduced with lentivirus encoding shPDGF-C RNA [LM8-cl.11 (pcFIH) + shPDGF-C] were tested for PDGF-C and FIH expression by Western blot. As controls, we used LM8-pcFIH cl.11 transduced with lentivirus encoding shScramble RNA [LM8-cl.11 (pcFIH) + shScr] and untransfected LM8-pcFIH cl.11 cells. β-actin levels served as loading control. Arrow points to the PDGF-C silenced clone. (b) LM8-cl.11 (pcFIH) + shPDGF-C and LM8-cl.11 (pcFIH) + shScr cells were inoculated in C3H mice, and tumor growth was measured every 2–3 days (n = 10) (*p < 0.05 and **p < 0.01). (c) After immunofluorescence staining for α-SMA to visualize pericytes, the percentage of α-SMA positive vessels was determined (n = 5) (***p < 0.001).

FIH-silenced LM8 tumors showed no alteration in PDGF-C protein expression compared to control (LM8-shScr) tumors (Supporting Information Fig. 7a) and no difference in tumor-associated vessel maturation (Supporting Information Fig. 7b). This observation might indicate that basal PDGF-C expression in LM8 cells is not regulated by FIH.

Discussion

Hypoxia is a common feature of many solid tumors leading to HIF-1α stabilization, which in turn stimulates the growth of new blood vessels, facilitates the adaptation of tumor cells to changes in the microenvironment and promotes their invasion and survival.10, 43–46 Overall, hypoxia can be considered as one of the most important environmental factors contributing to tumor progression. Accordingly, many experimental studies support the hypothesis that HIF-1 acts as tumor promotor, although in certain models, the opposite effect was observed.47, 48 The recent characterization of HIF hydroxylases that regulate the stability or activity of the HIF-α subunit and act as cellular oxygen sensors has raised the question regarding their function in tumors. Ablation of PHD2, which is considered as the main prolyl hydroxylase regulating HIF stability, has recently been shown to promote tumor angiogenesis, although the overall effects on tumor growth can vary, depending on the tumor model used.15, 16 Full silencing of HIF-1α in normoxic and mildly hypoxic cells requires the cooperation of PHD2 with FIH, whose activity prevents the interaction of HIF-1α subunits that have escaped destruction with transcriptional coactivators, such as p300. FIH is expressed in malignancies like human renal cell carcinoma, follicular lymphoma, non-small cell lung carcinomas and invasive breast cancer.21–23 However, the role of FIH in tumor progression has not been addressed experimentally.

In our article, we demonstrate that ablation of FIH in the murine osteosarcama cell line LM8 does not affect tumor growth. Even though FIH-deficient tumors displayed increased angiogenesis, vessel perfusion, as measured by penetration of Hoechst33342 dye into the tumor tissue was comparable to the control tumors. Thus, angiogenesis and growth which are considered to be closely linked10, 43, 44 are apparently uncoupled in this tumor model. A similar result was obtained previously when we studied the effect of PHD2 inhibition in LM8 and LLC tumors; this treatment led to increased vessel density but inhibited tumor growth as a consequence of the anti-proliferative activity of TGF-β.16 Together, these observations suggest that, within certain limits, vessel density in LM8 tumors is not the primary determinant of tumor growth. Uncoupling of tumor growth and angiogenesis has been first described by Noguera-Trois et al. who showed that increased angiogenesis caused by blockage of Delta-like ligand 4 inhibits tumor growth due to reduced perfusion of the blood vessels.49

In contrast to FIH inhibition, its overexpression led to increased tumor growth. This result supports the hypothesis that enhanced expression of FIH in tumor cells can stimulate tumor progression.21–23 Unexpectedly, HIF-1α protein levels were elevated in FIH-overexpressing tumors at Day 16 p.i., consistent with the presence of extended hypoxic areas around necroses. HIF-1 up-regulation did not occur in early-stage tumors or in cultured LM8-pcFIH cells and can therefore be considered as a secondary effect of increased tumor cell proliferation, resulting in inadequate perfusion of tumor tissue. It is unlikely that HIF-1α up-regulation is responsible for the increased tumor growth in the LM8-pcFIH tumors because HIF-1α silencing accelerates, rather than inhibits LM8 tumor growth.16

Since FIH overexpression did not alter the proliferation rate of LM8 osteosarcoma cells in vitro, we sought to identify influences from the host that promote tumor cell growth in vivo. Maturation of tumor vessels is an important parameter known to influence tumor progression.31–33 In addition, it has been described that α-SMA-positive pericyte coverage facilitates perfusion of tumor vessels.31 In our study, while we did not observe differences in blood vessel density, FIH-overexpressing tumors contained a higher proportion of mature vessels than control tumors. These vessels were typically covered with α-SMA-positive pericytes. The fact that we did not detect NG2 and desmin expression suggests that the pericytes in FIH-overexpressing tumors have a somewhat immature phenotype because capillary pericytes of mature vessels generally express NG2 and desmin and down-regulate α-SMA.34–37 It therefore appears that the signaling pathways that are activated in FIH-overexpressing LM8 cells are sufficient to stimulate the recruitment of immature pericytes that are associated with enhanced tumor growth,37 but fail to induce a more mature pericyte phenotype. Nevertheless, α-SMA positive mural cells were able to reduce vessel leakiness: dextran did not extravasate from α-SMA-positive vessels. Moreover, FIH-overexpression reduced the overall percentage of leaky vessels in the tumors. Therefore, we conclude that enhanced vessel maturation in the LM8-pcFIH tumors leads to a reduction in overall vessel leakiness, consequently facilitating delivery of oxygen and nutrients to the tumor tissue and finally resulting in increased tumor growth.

When searching for the mechanisms that promote vessel maturation in LM8-pcFIH tumors, we investigated the potential role of angiopoietins and PDGF family members37–42 which are known to regulate vessel maturation in tumors. Reduction of Ang-2 expression is expected to result in vessel stabilization; however, lower Ang-2 mRNA levels were observed only in late stage LM8-pcFIH tumors, suggesting that this change is not the main trigger of vessel maturation. In contrast, PDGF-C was consistently up-regulated in LM8-pcFIH tumors at all stages examined. Overexpression of PDGF-C has been demonstrated to stimulate tumor-associated vessel maturation, induction of α-SMA-positive mature vessels and accelerated tumor growth.40, 41 Our study shows that loss of PDGF-C in LM8-pcFIH tumors leads to reduction of vessel maturation and tumor growth, supporting the hypothesis that this factor is the major player facilitating vessel maturation and growth of FIH-overexpressing tumors. Moreover, PDGFR-α was co-localized with α-SMA on pericytes, whereas the cognate ligand, PDGF-C, was distributed throughout the whole tumor tissue. The antibody used can also cross-react with PDGFR-β, yet the clear effect of the PDGF-C knock-down indicates that the pericytes express functional PDGFR-α. Together, these findings support the idea that elevated PDGF-C expression stimulates the recruitment of pericytes to the vasculature in FIH-overexpressing tumors, leading to enhanced vessel maturation and tumor growth.

In conclusion, we show for the first time that FIH overexpression in tumor cells accelerates tumor growth. Our results demonstrate that FIH is a crucial factor controlling the maturation of tumor vessels. The mature phenotype correlated with the recruitment of pericytes and, as has also been observed in other tumors, the upregulation of PDGF-C and PDGFRα. Future experiments are necessary to investigate the pathways that lead to the induction of this signaling system. Altogether, our findings might have implications for the development of new therapies.

Acknowledgements

The authors thank Dr. Christian Beltinger (Ulm, Germany) for providing the LM8 cells, Dr. Didier Trono (Lausanne, Switzerland) for providing lentiviral vectors and Dr. Anne Klotzsche-von Ameln (Dresden, Germany) for providing murine LLC tumor sections.

Ancillary