HIF induces human embryonic stem cell markers in cancer cells

Abstract

Low oxygen levels have been shown to promote self-renewal in many stem cells. In tumors, hypoxia is associated with aggressive disease course and poor clinical outcomes. Furthermore, many aggressive tumors have been shown to display gene expression signatures characteristic of human embryonic stem cells (hESC). We now tested whether hypoxia might be responsible for the hESC signature observed in aggressive tumors. We show that hypoxia, through hypoxia-inducible factor (HIF), can induce an hESC-like transcriptional program, including the induced pluripotent stem cell (iPSC) inducers, OCT4, NANOG, SOX2, KLF4, cMYC, and microRNA-302 in 11 cancer cell lines (from prostate, brain, kidney, cervix, lung, colon, liver, and breast tumors). Furthermore, nondegradable forms of HIFα, combined with the traditional iPSC inducers, are highly efficient in generating A549 iPSC-like colonies that have high tumorigenic capacity. To test potential correlation between iPSC inducers and HIF expression in primary tumors, we analyzed primary prostate tumors and found a significant correlation between NANOG-, OCT4-, and HIF1α-positive regions. Furthermore, NANOG and OCT4 expressions positively correlated with increased prostate tumor Gleason score. In primary glioma-derived CD133 negative cells, hypoxia was able to induce neurospheres and hESC markers. Together, these findings suggest that HIF targets may act as key inducers of a dynamic state of stemness in pathologic conditions.

Figure 1. Hypoxia induces hESC markers among cancer cells

(A) mRNA profiling of hESC and hypoxic cancer cell lines (2% O₂ 24h) shown in a heatmap representation of gene expression changes. The color bar represents log₁₀ expression ratio −0.7 (teal) to +0.7 (magenta) (P<0.01 in at least 9 experiments). Outlined in dotted red are the overlap between these two profiling experiments. Cluster #5 (common up-regulated genes) contains OCT4 and other stem cell markers. (B) iPSC inducers are upregulated in cancer cells in response to hypoxia. The color bar represents log₁₀ expression ratio (hypoxia relative to normoxia) −0.6 (teal) to +0.6 (magenta). (C) qPCR validation of OCT4 expression in various cancer cell lines after 24h in hypoxia (2% O₂). (D) Stem cell markers are up-regulated by hypoxia over time in ME180 cells (qPCR analysis). (E) OCT4 protein is up-regulated by hypoxia in ME180 cells (flow cytometry analysis). Incubation with the secondary antibody (2^nd AB) alone was used as a negative control. (F) hESC enriched miRNAs are up-regulated by hypoxia treatment (24h in 2% O2) in multiple tumor cell lines (qPCR analysis). p values were calculated using paired t-test to measure the difference between the microRNA level under hypoxia and normoxia conditions in all cell lines. (G) The miR-302 promoter is responsive to over-expression of non-degradable forms of HIFα in HeLa cells.

Figure 2. Dependence of HIF for hypoxia-induced expression of stem cell markers

(A) Overexpression of non-degradable (nd) forms of HIFα under normoxia was assessed by Western blot analysis in ME180 cells 3 days after transfection with either an empty vector control (EV) or ndHIF1α or ndHIF2α. (B) The HIF-sensor containing 6 repeats of the HIF binding site (HBR) upstream of a minimal promoter (mP) and the yellow fluorescence protein (YFP) coding gene (upper panel) is upregulated by overexpression of ndHIF1α or ndHIF2α. (C) OCT4 promoter-GFP reporter is activated in ME180 Oct4-GFP and U251 Oct4-GFP cells two days after transfection with ndHIF1α and ndHIF2α but not with EV. Percentage of GFP-positive cells are indicated. (D) Expression of stem cell markers in renal carcinoma cells expressing either no functional VHL (qPCR; 786-0-pBABE = 786) or a functional VHL (786-0-pBABE-VHL = 786+VHL). (E) Oct4-GFP is induced in 786 but not in 786+VHL cells after transfection with ndHIF1α or ndHIF2α. Percentages of GFP-positive cells are indicated. (F–G) Hypoxia-induced expression of OCT4, NANOG and SOX2 is dependent on both HIF1α and HIF2α in ME180 (F) and HCT116 cells (G). The cells were transfected with siRNA directed against HIF1α, HIF2α or HIFβ. 24h later, the cells were cultured in normoxia or hypoxia (2% O₂) for 24h before qPCR (F) or microarray (G) analysis. The color bar represents log₁₀ expression ratio (hypoxia vs. normoxia) −0.3 (teal) to +0.3 (magenta) (G).

Figure 3. Overexpression of OCT4, SOX2, NANOG, LIN28 and HIFs in lung cancer cells

(A) Experimental flowchart depicting the generation of iPS-like cells from A549 lung adenocarcinoma cells. (B–C) Bright field images of A549 cells infected with a pBABE empty vector control (EV, B) and of a colony obtained 12 days after infection of A549 with Oct4/Sox2, Lin28/Nanog (OSLN), ndHIF1α and ndHIF2α (C). (D–E) Bright field images of picked colonies amplified on feeders obtained from A549 infected with OSLN (D) and with OSLN+ndHIFs (E). (F) HIF1 overepression enhances the induction of iPSC-like colonies. Colonies were counted on day 20 after lentiviral transduction. (G) Alkaline phosphatase staining of A549(OSLN+HIFs) colonies grown on feeders. (H) RT-qPCR analysis of endogenous NANOG and OCT4 mRNA expression in A549(OSLN) and A549(OSLN+HIFs) clones compared to H1 cells. (I) Methylation analysis of OCT4 promoter region. White circles represent unmethylated; black circles methylated cytosine-phosphate-guanisine (CpG). Numbers indicate the percentage of unmethylated CpG. (J–K) A549(OSLN+HIFs) tumors present malignant characteristics. Representative H&E-stained section of the xenograft A549(OSLN+HIFs) showing high mitotic index (J) and local invasion (K). (L) Tumors generated by injection of A549(OSLN+HIFs) cells present regions of differentiation. Paraffin embedded sections were subjected to H&E (upper panels) or Alcian blue-PAS (lower panels) staining. Bar right panels equals 100 μm.

Figure 4. Overlap between HIF, NANOG and OCT4 staining in prostate tumors

(A–C) Venn diagram representations of the staining analysis of specimen 09066C (A), 07-020C (B) and 07-091LA (C). (D–F) Quantification of HIF and NANOG, and HIF and OCT4 colocalization by correlation analysis in specimen 09-066C (D), 07-020C (E) and 07-091LA (F). Each dot represents 200 to 500 cells. The y-axis shows the ratio of HIF positive staining, and the x-axis the ratio of NANOG or OCT4 positive staining. (G–N) Representative images of HIF1α (G–J) and NANOG (K–N) immunohistochemistry staining on serial sections of specimens 09-066C, 07-020C and 07-091LA. Antibody used for NANOG staining is indicated. High magnification images are presented in the lower right corner. (O–P) CD107b (a marker for a tumor region; O) and CD104 (a marker for normal region; P) staining on serial section of the specimen 09-066C. (Q) Representative images of OCT4 immunohistochemistry staining on serial sections of specimen 07-020C. (R) NANOG and OCT4 expressions are enriched in Gleason 5 only glands, compared to the levels seen in overall glands.

Figure 5. Hypoxia induces stem cell markers and enhances neurosphere formation in gliomas

(A) Adjacent serial sections from a GBM patient tissue sample demonstrates nuclear staining of NANOG and OCT4 in a tumoral region with HIF1α immunoreactivity. Magnifications are 40X. Lower magnifications of the same region are presented in Suppl. Fig.11. (B–C) Hypoxia enhances neurosphere formation. Following collection from T4121 patient specimen, glioma stem and non-stem cells were plated in a 24 well plate at a density of 10 cells/well and cultured in either 2% O₂ or 20% O₂. After 12 days of treatment, representative phase contrast images were taken at 20X magnification (B). The number of neurospheres per well is presented (C). **, P < 0.01 with ANOVA.. (D) Hypoxia up-regulates hESC enriched microRNA in neurospheres generated from the non-stem population of primary gliomas. Following 12 days of exposure to normoxia or hypoxia (2% O₂), the 24 wells of neurospheres derived from non-stem primary glioma cells were pooled for each condition, and the expression of hESC-enriched miRNAs was quantified by RT-qPCR.