P53-induced microRNA-107 inhibits HIF-1 and tumor angiogenesis

Abstract

The pathway involving the tumor suppressor gene TP53 can regulate tumor angiogenesis by unclear mechanisms. Here we show that p53 regulates hypoxic signaling through the transcriptional regulation of microRNA-107 (miR-107). We found that miR-107 is a microRNA expressed by human colon cancer specimens and regulated by p53. miR-107 decreases hypoxia signaling by suppressing expression of hypoxia inducible factor-1β (HIF-1β). Knockdown of endogenous miR-107 enhances HIF-1β expression and hypoxic signaling in human colon cancer cells. Conversely, overexpression of miR-107 inhibits HIF-1β expression and hypoxic signaling. Furthermore, overexpression of miR-107 in tumor cells suppresses tumor angiogenesis, tumor growth, and tumor VEGF expression in mice. Finally, in human colon cancer specimens, expression of miR-107 is inversely associated with expression of HIF-1β. Taken together these data suggest that miR-107 can mediate p53 regulation of hypoxic signaling and tumor angiogenesis.

Many solid tumors require angiogenesis to grow beyond a certain size (1, 2). Tumor neovascularization is driven in part by hypoxia, which stimulates tumor cell production of angiogenic factors such as VEGF-A, basic FGF (bFGF), and placental growth factor (PlGF) (1–8). The hypoxia inducible factor-1 (HIF-1) heterodimer plays a critical role in hypoxic signaling in tumors (8–15). Mutations in genes regulating the HIF-1 pathway might alter tumor cell responses to hypoxia and tumor cell survival.

One of the genes most commonly mutated in human cancers is TP53 (16–21). When induced by cellular stress such as DNA damage or hypoxia, the tumor suppressor protein p53 normally inhibits cell growth and stimulates apoptosis. However, mutations in p53 or other disruptions in the p53 pathway are associated with tumor growth and angiogenesis (20, 21). Several studies suggest that p53 can regulate hypoxic signaling, but the signaling pathways are unknown (20, 22–24).

MicroRNA (miRNA) are critical components of the p53 signaling pathway. p53 activates transcription of a subset of miRNA that in turn suppresses transcription of genes that regulate apoptosis, DNA repair, and cell-cycle progression (25–28). Because p53 also can regulate hypoxic signaling by unknown pathways, we hypothesized that miRNA can mediate p53 regulation of tumor angiogenesis. Accordingly, we searched for miRNA that are expressed by human colon cancer specimens and are regulated by p53 and then screened this subset for miRNA that regulate hypoxia signaling.

Results

We characterized miRNA expression in colon cancers harvested from four patients at the Johns Hopkins Hospital, using microarray (Table S1). Of the 30 miRNA most highly expressed in human colon cancer specimens, we noted four miRNA that we and others had identified previously as being induced by p53 (26–30). To investigate the role of these miRNA in hypoxia signaling, we employed a bioinformatics approach. We found that miR-107, one of the members of this set of miRNA regulated by p53, is a potential regulator of HIF-1β. The HIF-1β and HIF-1α subunits form the heterodimer HIF-1, a transcription factor that mediates the transcriptional response to hypoxia (8–14, 31).

We first characterized p53 transcriptional regulation of miR-107. miR-107 is encoded within an intron of the gene for pantothenate kinase enzyme 1, PANK1. Computer analysis reveals a potential p53 binding site 1,811 bp upstream of the PANK1/miR-107 transcriptional start site (Fig. 1A). The DNA-damaging agent etoposide increased miR-107 levels in HeLa cells (Fig. 1B). Although etoposide also increased miR-107 in wild-type human colon cancer HCT116 cells [HCT116(WT)], etoposide failed to increase miR-107 in HCT116 cells lacking both p53 alleles [HCT116(p53 KO)] (Fig.1C). Expression of the parent gene PANK1 also is partly dependent upon p53 (Fig. 1D and Fig. S1).

Fig. 1.

p53 increases miR-107 expression. (A) Human PANK1 gene structure. miR-107 is encoded within an intron in the PANK1 gene. The gene encoding miR-107 is shown in black; the PANK1 gene also includes untranslated exons (white) and translated exons (gray). (B) Exposure to etoposide for 24 h increases miR-107 expression in HeLa cells by Northern blotting. (C) Exposure to etoposide for 24 h increases miR-107 only in HCT116 cells that express p53 by quantitative real-time PCR. (D) PANK1/pri-miR-107 but not PANK2/pri-miR-103–2 or PANK3/pri-miR-103–1 behaves like a p53 target gene. HCT116(WT) and HCT116(p53 KO) cells were exposed to Adriamycin for 24 h before transcript abundance was measured by quantitative PCR and normalized to U6. (E) Schematic diagram of the various PANK1 promoter reporter constructs containing a putative p53 binding element cloned into the pGL3 basic luciferase reporter plasmid. (F) Exposure to DNA-damaging agents for 24 h increases miR-107 promoter activity in HCT116 cells transfected with miR-107 reporter vector (n = 3 ± SD; *, P < 0.05). (G) Etoposide transactivates a miR-107 reporter vector containing a p53 binding site but fails to transactivate a shorter vector lacking the p53 binding site (n = 3 ± SD; *, P < 0.05 vs. control). (H) Etoposide fails to transactivate a miR-107 reporter vector with a specific mutation in the p53 binding site at −1,811 bp (n = 3 ± SD; *, P < 0.05 vs. control). (I) p53 interacts with miR-107 promoter in the region of the p53 binding site as demonstrated by ChIP. (J) p53 interacts with miR-107 promoter in the region of the p53 binding site as shown by quantitative PCR analysis of ChIP (n = 3 ± SD; *, P < 0.05)

To identify the regions upstream of PANK1/miR-107 that mediate p53 induction, we made reporter constructs consisting of 2,500 bp upstream of the PANK1/miR-107 gene driving expression of firefly luciferase (Fig. 1E). Genotoxic stress increases transactivation of the full-length reporter construct in HCT116(WT) cells (Fig. 1F). However, genotoxic stress fails to increase expression of the reporter construct in HCT116(p53 KO) cells (Fig. 1G). Mutation of the p53 binding element at −1,811 to −1,782 bp blunts the ability of etoposide to activate luciferase expression (Fig. 1H). ChIP confirms that p53 interacts with the 5′ UTR region from −1,922 to −1,696 bp containing the p53 binding site (Fig. 1 I and J). Taken together, these data suggest that p53 directly regulates miR-107 expression.

We next explored miR-107 regulation of HIF-1β. Overexpression of the precursor to miR-107 (pre-miR-107) decreases HIF-1β protein levels (Fig. 2 A–D). Also, miR-107 does not affect the partner of HIF-1β, namely HIF-1α (Fig. 2D). Conversely, knockdown of miR-107 increases HIF-1β expression in a variety of cell lines (Fig. 2E). Additionally, overexpression of other miRNA does not affect HIF-1β (Fig. 2F).

Fig. 2.

HIF-1β is a target of miR-107. (A) Manipulation of miRNA-107 in HCT116. Transfection of HCT116 cells with anti-sense miR-107 decreases endogenous miR-107 levels (± SD, *P < 0.05) (Left). Transfection of HCT116 cells with pre-miR-107 increases miR-107 levels (Right). (B) Transfection of HCT116 cells with pre-miR-107 suppresses HIF-1β protein levels as shown by immunoblotting. (C) Quantification of dose-dependent suppression of HIF-1β protein levels by miR-107 in HCT116 cells by densitometry (n = 3 ± SD; *, P < 0.05). (D) HCT116 cells were transfected with pre-miR-control or pre-miR-107 and exposed to normoxia or hypoxia, and the levels of HIF-1β and HIF-1alpha were probed by immunoblotting. miR-107 affects only HIF-1β expression, not HIF-1α. (E) Transfection of human colon cancer cell lines HCT116 and SW480 or the human epithelial cancer HeLa cell line with antisense (AS) to miR-107 slightly decreases HIF-1β protein levels. (F) Pre-miR-107 decreases HIF-1β expression in HCT116 cells, but other miRNA do not. (G) The HIF-1β 3′ UTR contains a binding site for miR-107. (Top) The HIF-1β 3′ UTR contains a potential miR-107 binding site with an exact 8-nt match. (Middle) Mutation of the miR-107 binding site in the HIF-1β 3′ UTR. (Bottom) Schematic of reporter vector containing luciferase followed by the HIF-1β 3′UTR. (H) miR-107 decreases transactivation of HIF-1β 3′ UTR. HCT cells were transfected with a reporter vector containing luciferase followed by 1,000 bp of the 3′ UTR of HIF-1β, containing either a WT UTR or a mutant UTR lacking the miR-107 binding site. Cells were transfected with control siRNA or pre-miR-107, and luciferase activity was measured (n = 3 ± SD; *, P < 0.05 vs. mutant binding site).

The 3′ UTR of HIF-1β contains a potential binding element for miR-107 with an 8-nt match to the miR-107 seed region (Fig. 2G). This potential binding site in the 3′ UTR of HIF-1β is broadly conserved among vertebrates (32). To test whether the HIF-1β 3′ UTR is a target of miR-107, we constructed a reporter vector with a constitutive promoter driving expression of a coding sequence containing the luciferase cDNA fused to the HIF-1β 3′ UTR (Fig. 2G). Overexpression of miR-107 decreases transactivation of the luciferase reporter (Fig. 2H). However, miR-107 does not affect expression of the luciferase reporter with a mutated miR-107 binding site (Fig. 2H).

Because miR-107 regulates HIF-1β, we proceeded to test the effect of miR-107 upon hypoxic signaling. As expected, hypoxia or the hypoxia mimetic desferrioxamine (DFX) increases HCT116 tumor cell levels of VEGF protein and mRNA (Fig. 3A and Fig. S2). Overexpression of miR-107 decreased DFX-induced VEGF expression (Fig. 3B). Conversely, knockdown of endogenous miR-107 increased DFX-induced VEGF levels (Fig. 3C). To confirm that miR-107 affects the ability of cells to regulate VEGF release, we treated HCT116 cells with pre-miR-107 or a scrambled oligonucleotide, exposed cells to hypoxia, and tested the conditioned media for the ability to stimulate endothelial proliferation, using a 3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay. Pre-miR-107 decreases the ability of HCT116 cells to stimulate endothelial proliferation (Fig. 3D).

Fig. 3.

miR-107 regulates hypoxia signaling. (A) Hypoxia (1% O₂) for 24 h or the hypoxia mimetic DFX for 24 h increases secretion of VEGF protein from HCT116 cells as measured by ELISA (n = 3 ± SD; *, P < 0.05 vs. control). (B) Exogenous miR-107 decreases secretion of VEGF from HCT116 cells transfected with siRNA control oligo or pre-miR-107 and then treated with DFX for 24 h (n = 3 ± SD; *, P < 0.05). (C) Endogenous miR-107 decreases secretion of VEGF from HCT116 cells transfected with siRNA control oligo or anti-sense (AS)-miR-107 and then treated with DFX for 24 h (n = 3 ± SD; *, P < 0.05). (D) miR-107 affects cellular stimulation of endothelial proliferation. HCT116 cells were transfected with control or pre-miR-107 and exposed to hypoxia for 16 h. Then the conditioned media were transferred to endothelial cells. Human umbilical vein endothelial cell proliferation was measured after 48 h by the MTT assay (n = 3 ± SEM; *, P < 0.05). (E) Manipulation of HIF-1β in HCT116 cells. HIF-1β siRNA transfection decreases HIF-1β protein expression in HCT116 cells (Upper rows). HIF-1β expression vector (pCMV-HIF-1β) transfection increased HIF-1β protein expression in HCT116 cells (Lower rows). (F) HIF-1β rescues cells from miR-107 inhibition of hypoxia signaling. Overexpression of HIF-1β in HCT116 cells restores DFX stimulation of VEGF protein secretion measured by ELISA (n = 3 ± SD; *, P < 0.05 for HIF-1β vs. control vector). (There is no significant difference between pre-miR-scr + DFX and pre-miR-107 + DFX in cells transfected with HIF-1β vector. The cDNA encoding HIF-1β lacks an miR-107 response element, so pre-miR-107 does not affect plasmid expression of HIF-1β.)

We next performed a rescue experiment to confirm that HIF-1β mediates the effects of miR-107. First we confirmed that HIF-1β mediates the transcriptional response to hypoxia mimetics: DFX increases VEGF expression, but silencing HIF-1β blocks the effect of DFX upon VEGF (Fig. S3). Next we constructed a vector that expresses HIF-1β (Fig. 3E, Lower panels). Our HIF-1β expression vector contains the cDNA for HIF-1β but does not contain the 3′ UTR region responsive to miR-107. Therefore miR-107 will inhibit endogenous HIF-1β but will not affect expression of HIF-1β from the expression vector. DFX increases VEGF production, and miR-107 inhibits VEGF production (Fig. 3F, Left). However, overexpression of HIF-1β blocks the effects of miR-107 (Fig. 3F, Right). miR-107 also has no effect on HIF-1α protein levels (Fig. S4). These data suggest that miR-107 blocks hypoxic signaling by suppressing HIF-1β expression.

We constructed viral vectors to explore the role of miR-107 in tumor angiogenesis (33). First we made a control lentiviral vector expressing enhanced GFP (eGFP) (LV-GFP) and an experimental lentiviral vector expressing both eGFP and pri-miR-107 (LV-GFP-miR-107) (Fig. 4A). More than 90% of HCT116 cells transfected with LV-GFP are positive for GFP (Fig. 4B). The levels of miR-107 are 15-fold increased in HCT116 cells transduced with LV-GFP-miR-107 as compared with cells transduced with LV-GFP (Fig. 4C).

Fig. 4.

miR-107 decreases tumor volume and vessels. (A) Schematic of LV-GFP-miR-107. (B) GFP fluorescence of HCT116 cells transduced with control media (Left) or with lentivirus expressing GFP (Right) analyzed by FACS. (C) LV-GFP-miR-107-transduced cells (solid bars) contain 10-fold more miR-107 than LV-GFP–infected cells (white bars) by quantitative real-time PCR (n = 3–10 ± SD; *, P < 0.05). (D) miR-107 decreases tumor volume. Nude mice were injected with 2 million HCT116 cells after transduction with lentiviral vectors. Four different HCT116 cell types were injected: HCT116(WT) and LV-GFP; HCT116(WT) and LV-miR-107; HCT116(p53 KO) and LV-GFP; and HCT116(p53 KO) and LV-miR-107. Tumor dimensions were measured for 31 d. Expression of miR-107 decreases the size of HCT116(WT) and HCT116(P53 KO) tumors (n = 10; *, P < 0.05 vs. LV-miR-107). (E) miR-107 decreases the density of blood vessels in tumors, as assessed by immunohistochemistry for VWF. (F) miR-107 decreases the number of tumor blood vessels, as measured by counting the number of vessels per high-power field for 12–15 fields for three tumors in each of the four groups (n = 3 ± SD; *, P < 0.001). (G) miR-107 decreases expression of the hypoxia-regulated gene VEGF. Quantitative real-time PCR was used to analyze expression of VEGF in tumors from nude mice implanted with HCT116 cells transduced with lentiviral vectors (n = 3–10 ± SD; *, P < 0.05). (H) miR-107 is inversely related to VEGF expression in human colon cancer. We measured miR-107 expression and VEGF expression by quantitative real-time PCR in 32 specimens of colon cancer harvested from patients at the Johns Hopkins University School of Medicine. We grouped the 32 human colon cancer specimens into the 16 specimens with the lowest expression of miR-107 and the 16 specimens with the highest expression of miR-107 (n = 16 ± SD; *, P < 0.0001). VEGF expression is higher in the low miR-107 group than in the high miR-107 group (n = 16 ± SD; *, P < 0.03). (I) Schematic for miR-107 regulation of hypoxia signaling.

We used these lentiviral vectors to examine the role of miR-107 in tumor angiogenesis in mice. We injected nude mice (n = 10 mice per group) with 2 million HCT116(WT) cells or HCT116(p53 KO) cells that had been transduced with LV-GFP or with LV-GFP-miR-107. We measured tumor size over 31 days. Transduction of HCT116(WT) cells with LV-miR-107 decreased tumor size (Fig. 4C, red symbols). Overexpression of miR-107 also decreased the size of tumors from HCT116(p53 KO) cells (Fig. 4C, green symbols). These data show that miR-107 decreases tumor size in the analyzed model.

We next measured the effect of miR-107 on the number of vessels inside tumors. Tumors were sectioned and stained with antibodies to platelet endothelial cell adhesion molecule (PECAM) and von Willebrand factor (VWF) (Fig. 4D). We then counted the number of vessels in 10–15 high-power fields per section (n = 3 per group). Overexpression of miR-107 decreased the number of vessels in HCT116(p53 KO) (3.9 ± 2.4 vessels for LV-miR-107 vs. 6.8 ± 3.1 vessels for LV-GFP; P = 0.00002) (Fig. 4 E and F). Overexpression of miR-107 decreased vessel number in HCT116(WT) by a nonsignificant amount. Tumors from HCT116(WT) cells had fewer vessels than tumors from HCT116(p53 KO) cells (P < 0.001).

We also measured tumor expression of the hypoxia-regulated gene VEGF. Transduction of tumors with LV-miR-107 decreased VEGF expression in tumors in mice (Fig. 4G).

Finally we examined human colon cancer specimens to explore the association between miR-107 and VEGF. We measured miR-107 and VEGF expression by quantitative real-time PCR in 32 human colon cancer specimens from patients at the Johns Hopkins Hospital. We compared VEGF expression in the 16 human colon cancer specimens with the highest levels of miR-107 and in the 16 specimens with the lowest miR-107 levels. Expression of VEGF is higher in the low miR-107 group and is lower in the high miR-107 group (Fig. 4H).

We previously had shown that the p53-regulated miR-34a promotes apoptosis. We therefore explored the effect of miR-107 upon proliferation and the cell cycle. We transfected HCT116 cells with a retroviral vector expressing miR-107 or with empty vector and measured proliferation and the cell cycle by flow cytometric measurement of DNA content. Overexpression of miR-107 decreases proliferation of both HCT116(WT) and of HCT116(P53 KO) cells (Fig. S5A). HCT116(WT) cells transfected with the control vector exhibited low levels of apoptosis (Fig. S5B, Upper Row). In contrast, overexpression of miR-107 in HCT116(WT) cells induced apoptosis. In cells synchronized by nocodazole, overexpression of miR-107 also increased the percentage of cells in G1 (Fig. S5B, Lower Row). These data suggest that miR-107 in part mediates p53 regulation of the cell cycle. Taken together, these data also suggest that miR-107 can regulate tumor progression by two distinct mechanisms, regulation of the cell cycle and control of hypoxic signaling.

Discussion

The major finding of our study is that miR-107 regulates hypoxic signaling (Fig. 4I). Several aspects of our study are intriguing. First, although it is widely assumed that HIF-1 signaling is regulated primarily through the HIF-1α subunit, our data suggest that regulation of HIF-1β also may modulate hypoxic responses. In addition, our studies suggest that specific miRNA can modulate the cellular response to hypoxia. Our data reinforce the work of others who have shown that hypoxia can alter the expression of miRNA and that miRNA can alter HIF-1α expression in vitro (34, 35). Others have shown that, compared with WT tumors, tumors that lack HIF-1α are poorly vascularized but are faster growing, perhaps because of a loss of dependency upon neovascularization (36, 37).

Folkman (1) and others (38–40) proposed that some tumors can undergo an angiogenic switch, during which avascular tumors recruit new blood vessels, increasing the supply of oxygen and nutrients and permitting further tumor growth. Tumor angiogenesis may be driven not only by the tumor microenvironment but also by tumor and stromal mutations. One of the genes most commonly mutated in human cancers is TP53 (16–18) (19), and mutations in TP53 are associated with an increase in tumor angiogenesis (20, 21). Some studies suggest that p53 regulates hypoxic signaling and that tumor cells lacking TP53 are resistant to hypoxia (20, 22–24); conventional explanations for this phenomenon include decreased apoptosis and senescence, and our data suggest an additional mechanism, namely increased levels of HIF-1 signaling. Our data provide one pathway through which p53 can regulate hypoxic signaling and tumor angiogenesis. Cancers with mutations in TP53 and increased hypoxia signaling may be more susceptible to antiangiogenic therapy than tumors with an intact p53 pathway (20, 41).

Materials and Methods

Cell Culture.

HCT116 (p53 WT and p53 KO) cells were a gift from Bert Vogelstein, Baltimore, MD.

Transfection.

HCT116 cells were transfected using siPORT NeoFX (Applied Biosystems) or Lipofectamine 2000 (Invitrogen) with precursor miRNA or with antisense miRNA (0–20 nM) and were harvested 48 or 72 h later.

Quantitative Real-Time PCR.

To analyze miRNA expression, TaqMan MicroRNA assays (Applied Biosystems) were used to quantify levels of mature miRNAs following the manufacturer's instructions. The primers for quantitative real-time PCR and the PCR mix were purchased from Applied Biosystems.

Plasmid Construction.

The 3′ UTR of HIF-1β (960 bp) containing the HIF-1β-miR-107 response element was cloned into pMIR-REPORT Luciferase vector (Applied Biosystems). For reporter assays, the 5′ UTR region of Pank1/miR-107 extending upstream from the transcriptional start site (0 bp) to −2,500 bp was amplified from genomic human DNA and cloned into the pGL3 promoter plasmid (Promega).

Cell Proliferation Assay.

Cell proliferation was measured by MTT assay. Briefly, we plated each group of cells (1 × 104) on 96-well plates, added MTT, and incubated the cells for 3 h. The cells were dissolved by adding HCl/SDS, and absorbance was measured at 570 nm.

Apoptosis Analysis.

HCT116 cells were transfected with precursor miRNA or scrambled oligonucleotides as a control. After a further incubation of 72 h, the cells were harvested, stained with propidium iodide and anti-annexin-V antibody, and analyzed by FACS.

Lentivirus Construction and Transduction.

Third-generation lentiviral vectors were a gift of I. M. Verma (La Jolla, CA) (33). The lentiviral vectors were constructed and prepared according to previously published protocols (33). HCT116 cells were transduced with lentiviral vectors at an MOI of 10, and expression of GFP was confirmed by FACS.

Murine Tumor Model.

Tumors were implanted into mice as described previously (42). Briefly, nude mice were anesthetized and injected in the flank with 2 million HCT116 cells. Mice were divided into four groups of 10 mice: group 1 received HCT116(WT) transduced with LV-GFP; group 2 received HCT116(WT) transduced with LV-miR-107; group 3 received HCT(p53 KO) transduced with LV-GFP; and group 4 received HCT116(p53 KO) transduced with LV-miR-107. Tumor diameters were measured with an electronic caliper over 31 d. Mice were killed at 31–38 d, tumors were harvested, and sections of tumors were stained for PECAM and VWF. The number of vessels per high powered field was counted in 12–15 fields for three tumors in each of the four groups.