Identification of CDCP1 as a hypoxia-inducible factor 2α (HIF-2α) target gene that is associated with survival in clear cell renal cell carcinoma patients

Abstract

CUB domain-containing protein 1 (CDCP1) is a transmembrane protein that is highly expressed in stem cells and frequently overexpressed and tyrosine-phosphorylated in cancer. CDCP1 promotes cancer cell metastasis. However, the mechanisms that regulate CDCP1 are not well-defined. Here we show that hypoxia induces CDCP1 expression and tyrosine phosphorylation in hypoxia-inducible factor (HIF)-2α-, but not HIF-1α-, dependent fashion. shRNA knockdown of CDCP1 impairs cancer cell migration under hypoxic conditions, whereas overexpression of HIF-2α promotes the growth of tumor xenografts in association with enhanced CDCP1 expression and tyrosine phosphorylation. Immunohistochemistry analysis of tissue microarray samples from tumors of patients with clear cell renal cell carcinoma shows that increased CDCP1 expression correlates with decreased overall survival. Together, these data support a critical role for CDCP1 as a unique HIF-2α target gene involved in the regulation of cancer metastasis, and suggest that CDCP1 is a biomarker and potential therapeutic target for metastatic cancers.

Fig. 1.

Hypoxia activates the CDCP1-Src pathway. (A) HCT116 cells were exposed to 21% O₂ (normoxia; N) or 1% O₂ (hypoxia; H) for 12, 18, 24, 36, or 64 h. The anti-CDCP1 antibody was used for immunoprecipitation (IP) and immunoblotting. An anti-phosphotyrosine antibody was used for immunoblotting. (B) HCT116 cells were exposed to N or H for 48 h. Lysates were immunoblotted for the phosphorylation of CDCP1 and Src-family kinases using a specific CDCP1 P-734 antibody that recognizes CDCP1 when phosphorylated on Tyr734 and a P-416 SFK antibody that recognizes the activation loop of Src-family kinases when phosphorylated at Tyr416. An anti-SFK antibody was used as loading control. (C) DLD1 and HCT116 cells cultured in 0.1% or 10% FBS were treated ±100 μM DFO and/or 50 nM dasatinib for 24 h. The anti-CDCP1 antibody was used for immunoprecipitation and immunoblotting. An anti-phosphotyrosine antibody was used for immunoblotting. (D) Two retroviral shRNAs targeted against CDCP1 were used to stably knock down CDCP1 in MCF10A cells. An anti-tubulin antibody was used as loading control. Stable cell lines were exposed to N or H for 48 h. Lysates were immunoblotted for the phosphorylation of CDCP1 and SFKs as in B using the specific CDCP1 P-734 antibody and the anti-SFK antibody for loading control. (E) Stable CDCP1 knockdown MCF10A cell lines were serum-starved and exposed to 1% O₂ for 24 h; subsequently, the cells were seeded in transwells and returned to 1% O₂ for 24 h. Cells were fixed and stained with crystal violet (0.1%; Lower). The number of cells that migrated to the bottom of the filter was counted and the data are reported as fold induction over the pRenilla control cells. The data presented are the result of triplicate analyses and the error bars indicate SEM. *P = 0.088, ***P = 0.0007.

Fig. 2.

Hypoxic activation of CDCP1 is HIF-2α–dependent. (A) Stable knockdown of HIF-1α, HIF-2α, and ARNT in MCF10A cells. (B) Stable cell lines exposed to 21% O₂ (N) or 1% O₂ (H) for 48 h. Lysates were immunoblotted for CDCP1, P-734 CDCP1, P-416 SFK, SFK, and α-tubulin for loading control. (C) Quantitative real-time PCR for Cdcp1 on mRNA isolated from MCF10A stable cell lines cultured in N or H for 24 h. Data are represented as the means ± SEM (n = 3). (D) MCF10A cells were exposed to N or H for 24 h and subsequently ChIP was carried out on DNA–protein complexes with anti–HIF-1α, anti–HIF-2α, anti-ARNT, or control IgG antibodies followed by qRT-PCR. Antibodies used are indicated on the x axis. Data are represented as the means ± SEM (n = 3). ***P < 0.0001, two-tailed Student’s t test. (E) Alignment of Cdcp1 (chromosome 3) with the predicted HRE/ARNT binding site using MAPPER2 (http://genome.ufl.edu/mapper). Red indicates the predicted HRE/ARNT binding site. (F) HT1080 cells were transfected with the R01-scrambled control vector or with the Cdcp1-promoter vector. At 24 h after transfection, cells were exposed to 21%O₂ (N) ±100 μM DFO or 1 mM DMOG. RenSP luciferase was measured and the data are reported as fold induction over untreated normoxic R01 control vector cells. The data presented are the result of triplicate analyses and the error bars represent SEM. *P = 0.02, ***P = 0.0002. (G) Stable MCF10A cell lines were serum-starved and exposed to 1% O₂ for 24 h; subsequently, the cells were seeded in transwells and returned to 1% O₂ for 24 h. CUB1 mAb was added at the time of seeding to the top and bottom chambers. Cells were fixed and stained with crystal violet (0.1%; Lower). The number of cells that migrated to the bottom of the filter was counted and the data are reported as fold induction over the pLK0.1 control cells. The data presented are the result of triplicate analyses and the error bars indicate SEM. **P < 0.001, two-tailed Student’s t test.

Fig. 3.

HIF-2α is sufficient to activate CDCP1 and promote tumor growth in xenografts. (A) Tumor formation over time in nude mice injected with the A375 cancer cell line expressing GFP or HIF-2αDPA. Doxycycline chow was used to induce expression of GFP or HIF-2αDPA. Error bars are SEM. (n = 6). (B) Representative images of tumors; GFP control cells (left flank) or HIF-2αDPA (right flank) after mice were euthanized. (C) Protein was isolated from tumors and immunoblotted for CDCP1, P-734 CDCP1, HA, and α-tubulin for loading control. Shown are two representative tumors from the mice. (D) Scatterplot shows that overexpression of HIF-2α significantly enhanced lung metastases in NOD/SCID mice. The number of mice with surface lung metastases was counted 9 wk after tail-vein injections (n = 4). (E) Heat map of CDCP1 and HIF-2α expression based on the microarray data of 732 unique cancer cell lines from the SCLP. (F) Scatterplots comparing the expression levels of CDCP1 with HIF-2α (Left), EGFR (Center), and MET (Right), all showing a strong positive correlation in their expression across the 732 cancer cell lines (P = 1 × 10⁻²⁰, Pearson’s correlation coefficient analysis). (G) Heat map of CDCP1, HIF-2α, EGFR, and MET expression across the SCLP (n = 732).

Fig. 4.

High CDCP1 expression in ccRCC is associated with poor survival. Tissue microarray provided by the Kidney Cancer Tissue Acquisition, Pathology and Clinical Data Core Facility at The Dana-Farber Cancer Institute. The anti-CDCP1 antibody (Cell Signaling Technology) was used 1:500 with TSA. Pictures were taken under 20× magnification on a Leica microscope. (A) Representative pictures of total CDCP1 levels in RCC tumors (low and high expression of CDCP1, left to right). (B) Histogram of IHC expression of CDCP1. Numbers on top of the columns indicate the number of patients. (C) Kaplan–Meier survival analysis for patients with low vs. high CDCP1 expression. DOD, date of diagnosis. (D) Lysates from RCC cell lines (RCC4, A498, and 786-0) were immunoblotted for the phosphorylation (P-734) of CDCP1, total CDCP1, HIF-2α, and α-tubulin for loading control under normoxia (21% O₂).