Endothelial cell HIF-1α and HIF-2α differentially regulate metastatic success

Abstract

The hypoxia inducible transcription factors (HIFs) control many mediators of vascular response, including both angiogenic factors and small molecules such as nitric oxide (NO). In studying how endothelial HIF response itself affects metastasis, we found that loss of HIF-1α in endothelial cells reduces NO synthesis, retards tumor cell migration through endothelial layers, and restricts tumor cell metastasis, and that loss of HIF-2α has in each case the opposite effect. This results from differential regulation of NO homeostasis that in turn regulates vascular endothelial growth factor expression in an NO-dependent feedback loop. These opposing roles for the two HIF factors indicate that both they and endothelial cells regulate metastasis as malignancy progresses.

Figure 1

HIF-1α Deletion in Endothelial Cells Leads to Reduction in Metastatic Incidence in Transgenic and Xenograft Models (A) Effect of Tie2-driven deletion of HIF-1α on primary tumor growth in the spontaneous breast tumor model PyMT (HIF-1α^df/Tie2Cre/PyMT) during different stages during tumor development, at 16 and 18 weeks, and at endpoint (when one tumor reaches diameter of 1 cm). (B) Effect of Tie2-driven deletion of HIF-1α on spontaneous breast tumor model PyMT (HIF-1α^df/Tie2Cre/PyMT) tumor metastases at 16 wks of age (left panel), and at endpoint (when one tumor reaches diameter of 1 cm, right panel). (C) LLC^GFP cells were implanted in cleared mammary pads of endothelial cell-specific deletion of HIF-1α (HIF-1α null) obtained by regenerating WT bone marrow into HIF-1α^df/Tie2cre+ mice. Effect of endothelial HIF-1α deletion on frequency of lung metastasis 3 weeks after tumor cell implantation (left) and on blood genomic GFP levels, a measure of relative abundance of circulating tumor cells (right); Data in (A–C) are average ± SEM. (D and E) Boyden assay for tumor cell migration during 9h periods of normoxia (21% O₂) or hypoxia (1% O₂) through endothelial monolayers of WT, HIF-1α null, and HIF-2α null cells (D) and VEGF null and iNOS null cells (E). Data in (D) and (E) are average ± SD of ratio of cells migrated through endothelium divided by the mean number of migrated cells through WT endothelium during normoxic conditions. An asterisk is placed above differences with a significance exceeding p < 0.05; ^∗∗p < 0.01; ^∗∗∗p < 0.001. NS indicates differences that do not reach a p < 0.05 level of significance.

Figure 2

Deletion and Specific Inhibition of iNOS Result in Decreased VEGF Expression and Downstream Activity (A–C) Quantification of relative abundance of iNOS mRNA in VEGF null endothelial cells (A) and VEGF mRNA abundance in iNOS KO cells (B) or WT cells treated with 15 μM of specific iNOS inhibitor 1400W (right panel) by quantitative RT-PCR (C). (D) VEGF mRNA levels in WT and iNOS null endothelial cells upon addition of 5 mM NO donor DETA NONOate (dNO). (E) Representative western blot of immunoprecipitated VEGFR2 (Flk1) probed for total Flk1 and P-tyrosine to detect activated form (left) and signal quantification of multiple blots (n = 3) by phosphor imaging (right). (F) Tube formation assay in collagen I matrix, representative pictures of WT, WT + 30 μM 1400W, and iNOS null are shown (left) (scale bar represents 100 μm) and quantification of length of detected networks (right) was performed using ImageJ. Data is average ± SEM of lengths obtained from at least three pictures of at least three wells. Experiment shown is representative and was repeated three times with cells no older than passage 2. In this figure, data shown is average ± SEM, An asterisk is placed above differences with a significance exceeding p < 0.05; ^∗∗p < 0.01; ^∗∗∗p < 0.001.

Figure 3

Endothelial Cells Generate NO through iNOS Activation during Hypoxia (A) Relative abundance of iNOS and ARG1 transcripts in WT (left), HIF-1α null (center), and HIF-2α null (right) endothelial cells exposed to 21% O₂ (normoxia) or 1% O₂ (hypoxia) for 9 hr. Data is average ± SEM of each sample; mRNA levels are compared to those observed in WT normoxia (that was defined as 1.0-fold). (B) NO levels in endothelial cell conditioned medium from cells treated for 48 hr at 21% or 0.5% O₂; data presented are average ± SEM. An asterisk is placed above differences with a significance exceeding p < 0.05; ^∗∗p < 0.01; ^∗∗∗p < 0.001.

Figure 4

Inhibition of VEGF Expression upon iNOS Inhibition Is A Phenomenon Specific to Endothelial Cells (A) Quantitative PCR of wild-type mouse endothelial cells (mEC), human umbilical vascular endothelial cells (hUVEC), mouse embryonic fibroblasts (MEF), and Lewis lung cancer cells (LLC) exposed to normoxia or hypoxia (21% or 1% O₂, respectively) in the presence or absence of 15 μM 1400W, to assess mRNA levels of VEGF and PGK (hypoxia and HIF-1α control). Data is average ± SEM of ratio of mRNA levels in sample compared to mRNA levels in WT untreated during normoxia. (B) Quantitative PCR of RNA extracted from subcutaneous LLC^GFP tumors in WT mice. Mice were treated with two intraperitoneal injections: the first, a 15 mg/kg 1400W (or respective saline control of equal volume) at 8 hr before sacrifice, and the second, a 10 mg/kg injection at 4 hr prior to sacrifice. An asterisk is placed above differences with a significance exceeding p < 0.05; ^∗∗p < 0.01; ^∗∗∗p < 0.001.

Figure 5

Hypoxia-Induced Tumor Cell Migration Is Inhibited in the Absence of Active iNOS (A) Endothelial monolayers of different genotypes were incubated with 15 μM of 1400W for 3 hr, and then washed to remove inhibitor. They were then incubated with LLC cells (in the absence of inhibitor) for 9 hr at either 21% or 1% O₂, to assay for effects of inhibition of iNOS in endothelial cells on tumor cell migration. (B) Migration assay using WT and iNOS null EC monolayers. (C) Metastases in WT mice lungs 14 days posttail vein injection of 1 × 10⁶ LLC^GFP cells. Experimental mice received a total of two intraperitoneal injections of either control saline or 1400W, one at 8 hr, and one at 4 hr (the first of 15 mg/kg followed by 10 mg/kg) immediately prior to tumor cell injections. In this figure, data presented is average ± SEM. An asterisk is placed above differences with a significance exceeding p < 0.05; ^∗∗p < 0.01; ^∗∗∗p < 0.001.

Figure 6

Endothelium-Specific Deletions of HIF-1α and HIF-2α Have Opposing Effects on Tumor Metastases LLC^GFP cells (1 × 10⁶) were injected via tail vein into HIF-1α^df/Tie2Cre or HIF-2α^df/Tie2Cre mice. All Cre-expressing and WT mice were lethally irradiated, and then received WT bone marrow, in order to ensure endothelial-exclusive deletion of gene of interest. Lungs were collected 14 days postinjection, fixed, and sectioned for hematoxylin and eosin staining. (A) Representative images of mouse lungs 14 days post-LLC^GFP tail vein injection. Scale bars represent 5 mm. (B) Number of metastases foci found in lungs of HIF-1α null, HIF-2α null, and their respective WT (i.e., df = double floxed/cre negative) controls. (C) Average of relative tumor area in lung sections obtained by quantifying tumor area versus whole lung area in every third section of whole lungs. In this figure, data presented is average ± SEM. An asterisk is placed above differences with a significance exceeding p < 0.05; ^∗∗p < 0.01; ^∗∗∗p < 0.001.

Figure 7

Differential Functions of HIF-1 and HIF-2 during Metastasis Depend on Differential NO Synthesis As shown, HIF-1 and HIF-2 response in endothelial cells gives rise to differing levels of NO production following hypoxic induction of the factors: in this case, through hypoxic stress caused by tumor cell-induced blockage of a capillary. This, in turn, results in differing levels of metastatic success when one or the other HIF is genetically deleted from the endothelial cell. A remaining and important question is how hypoxic stress can cause differing levels of induction of the two HIF factors.