The Oncogene Phosphatidylinositol 3′-Kinase Catalytic Subunit α Promotes Angiogenesis via Vascular Endothelial Growth Factor in Ovarian Carcinoma¹

PI3k³ is a novel intracellular transducer with lipid substrate specificity, which is involved in many cancer-associated signaling pathways (1, 2, 3, 4). PI3k comprises a M_r 110,000 catalytic subunit (PIK3CA) and a regulatory subunit of either M_r 85,000, M_r 55,000, or M_r 50,000 (5). The gene encoding PIK3CA is located at 3q26, a region with reportedly increased copy numbers in >40% of ovarian and other cancer specimens by comparative genomic hybridization (6, 7). PIK3CA gene amplification has also been found in a variety of human solid tumors (8, 9, 10, 11, 12, 13, 14, 15). Recently, PIK3CA has been implicated as an oncogene in ovarian cancer (1). Increased DNA copy numbers have been frequently found in human ovarian cancer cell lines, which was associated with increased PIK3CA transcription and expression of the p110α protein (1). Importantly, PIK3CA gene amplification leads to increased p110α-p85 heterodimer formation and increased PI3k activity in vitro (1). However, current knowledge about the role of PIK3CA as an oncogene in ovarian cancer has been derived from in vitro experiments and very little information is available on the expression and role of PIK3CA in ovarian cancer in vivo.

A critical effect of oncogene activation is to promote aberrant cellular mitogenesis. However, accumulating evidence indicates that oncogenes such as ras, src, myc, HER-2/neu, as well as human papilloma virus E6 (16, 17, 18, 19, 20, 21, 22, 23) or tumor suppressor genes such as p53, pVHL, and PTEN (24, 25, 26, 27, 28, 29, 30, 31) are implicated in the control of tumor angiogenesis via HIF and the downstream VEGF (32, 33). PI3k pathway may be involved in the physiological regulation of VEGF and may serve as an angiogenesis regulatory gene (4, 34, 35, 36, 37, 38, 39, 40, 41). In endothelial cells, PI3k partakes in several growth factor-dependent pathways regulating VEGF expression and angiogenesis (4, 36) via HIF-1α. Direct in vitro or in vivo evidence for the involvement of PIK3CA in the regulation of VEGF in human tumors is lacking to date.

In this study, we show that PIK3CA overexpression is indeed highly prevalent in ovarian cancer and occurs early in the process of malignant transformation. Its expression positively correlates with the expression of VEGF, both at the mRNA and the protein level. Additionally, we demonstrate that PI3k is directly implicated in the control of VEGF expression in ovarian carcinoma, an effect that is mediated via HIF-1α. Finally, we show that PIK3CA overexpression correlates with increased proliferation and decreased apoptosis of tumor cells in ovarian cancer in vivo, an effect that may be mediated jointly through an increase in angiogenesis as well as a direct effect of PI3k. Collectively, these data strongly support a role of PIK3CA in ovarian cancer progression.

Two normal human ovaries, 4 benign cystadenomas, 4 LMPs, and 42 malignant tumors, including 9 stage I and 33 stage III or IV carcinomas, were randomly selected from the specimen bank existing in our lab (44, 45) and analyzed by quantitative real-time RT-PCR. Thirty-nine advanced stage tumors from the same specimen bank were analyzed by immunostaining.

A total of 17 human ovarian cancer cell lines (45) was cultured in 5% CO₂ atmosphere at 37°C and in DMEM supplemented with 10% fetal bovine serum and 100 units/ml penicillin. Primary human ovarian surface epithelium cultures were established as described previously (45). In some experiments, cells were treated with the PI3k inhibitor Ly294002 (43) (10 or 50 μm; Calbiochem, La Jolla, CA). Cells were seeded in 6-well plates and cultured in complete growth media. Upon reaching 90% confluence, cells were cultured in serum-free media overnight. After washing with PBS, cells were cultured in glucose- and serum-free DMEM in the presence or absence of Ly294002. Control cells were treated with DMSO. All experiments were repeated three times.

Total RNA was isolated from 1 × 10⁶ cultured cells or 100–500 mg of frozen tissue with Trizol reagent (Invitrogen). After treatment with RNase-free DNase (Invitrogen), RNA was further purified with RNeasy RNA isolation kit (Qiagen, Valencia, CA). Total RNA was reverse transcribed using Superscript First-Strand Synthesis Kit for RT-PCR (Invitrogen) under conditions described by the supplier.

cDNA was quantified by real-time PCR on the ABI Prism 7700 Sequence Detection System (Applied Biosystems, Foster City, CA) as described previously (43). Specific oligonucleotide primers and probes (Table 1) were designed based on published sequences.

IHC was performed using the Vectastain ABC kit as described by the manufacturer (Vector, Burlingame, CA). All primary antibodies, mouse antihuman p110α (1:50; PharMingen, San Diego, CA), VEGF (1:200; PharMingen), CD31 (1:30; Dako, Carpinteria, CA), rabbit antihuman Ki-67 (1:200; Dako), and cytokeratin (1:1000; Dako), were incubated on sample sections for 1 h. The prevalence of immunoreactive VEGF and p110α in tumors was scored semiquantitatively and blindly by two independent investigators and confirmed by a third investigator as following: 0 = all cells in tumor islets were negative; 1 = <20% of cells in tumor islets were positive; 2 = between 20 and 80% of the cells in tumor islets were positive; and 3 = >80% of cells in tumor islets were positive. Ki-67 staining was quantified by image analysis. Images were collected through Cool Snap Pro color digital camera (Media Cybernetics), and staining index was analyzed using Image-Pro Plus 4.1 software (Media Cybernetics). Immunofluorescent double staining was performed as described previously (47).

Cultured cells were lysed in 1 ml of lysis buffer containing 50 mm Tris-HCl (pH 7.4), 150 mm NaCl, and 1% Triton X-100. Protein was separated by 12% SDS-PAGE under denaturing conditions and transferred to nitrocellulose membrane. Membranes were incubated with an anti-HIF-1α monoclonal antibody (1:500; PharMingen), followed by incubation in rabbit antimouse secondary antibody conjugated with horseradish peroxidase (1:5000; Sigma, St. Louis, MO). Immunoreactive proteins were visualized using enhanced chemiluminescence detection system (Amersham Biosciences, Piscataway, NJ).

The ApopTag in Situ detection kit (Intergen, Purchase, NY) was used to visualize apoptotic cells as described previously (47).

Cells were seeded at a density of 1–5 × 10³/well in 96-well plates in complete culture medium. After 24 h of culture, serum-free medium was added with or without Ly294002 (10 μm). Cell proliferation was assessed 72 h later using CellTiter 96 (Promega, Madison, WI). Cell proliferation was determined as the ratio of the absorbance of treated cells to the density of control untreated cells.

Annexin-V staining was detected by flow cytometry using an apoptosis detection kit (R&D Systems, Minneapolis, MN) as described previously (47).

Statistical analysis was performed using the SPSS statistics software package (SPSS, Chicago, IL). All results were expressed as mean ± SE, and P < 0.05 was used for significance.

We used quantitative real-time RT-PCR to assess mRNA levels in ovarian neoplasms, including benign cystadenomas (n = 4), LMP tumors (n = 4), and invasive ovarian carcinomas at early (I, n = 9) or advanced stage (III or IV, n = 33). PIK3CA mRNA was detectable in 6 of 9 (66.6%) stage I and 31 of 33 (93.9%) advanced-stage ovarian cancer specimens. As a whole, malignant ovarian tumors significantly overexpressed PIK3CA mRNA compared with the group of benign and LMP neoplasms (P = 0.007). No significant difference was seen between early and advanced stage carcinomas (P = 0.812; Fig. 1,A). Interestingly, mRNA levels of PIK3CA were similar in whole normal postmenopausal ovaries (n = 2) and invasive carcinomas (data not shown). PIK3CA mRNA was also detected in all 17 human ovarian cancer cell lines analyzed (Fig. 2 A). The mean expression of PIK3CA in ovarian cancer cell lines was 2.7-fold higher compared with primary cultures of human ovarian surface epithelium.

To examine the expression and localization of p110α protein in ovarian carcinoma, we performed IHC. Immunoreactive p110α was detectable in 22 of 39 (56.4%) ovarian cancer specimens, with moderate to strong staining localized to tumor cells of islets (Figs. 1, B and C, and 2, C and F). Weak or no staining was found in tumor cells in the remaining 17 samples. Immunostaining was localized mainly to the cytoplasm of tumor cells but not to their nucleus (Figs. 1, B and C, and 2, C and F). Interestingly, p110α expression was also detected in endothelial cells of medium caliber vessels of tumor stroma in 28 of 37 evaluable (75.7%) samples (Fig. 1,D), whereas it was absent from vessels in the remaining 9 (24.3%) samples (Fig. 1,C). In addition, there was no relationship of the p110α expression between tumor cells and tumor-associated endothelial cells (Fig. 1,E). p110α expression was not detectable in other types of stroma supporting cells (Figs. 1,B–D and 2, C and F). Interestingly, in normal postmenopausal ovaries, p110α immunoreactivity was localized exclusively to the muscularis layer of blood vessels but not to their endothelium (data not shown).

In ovarian cancer cell lines, PIK3CA amplification leads to increased PI3k activity (1). Because PI3k has been shown to affect the expression of HIF-1α, an upstream regulator of VEGF (35), we investigated whether the observed overexpression of PIK3CA correlated with overexpression of VEGF in ovarian cancer. We analyzed the mRNA levels of PIK3CA and VEGF in 33 advanced stage ovarian cancer specimens, as well as 17 established human ovarian cancer cell lines using quantitative real-time RT-PCR. A significantly positive correlation was found between VEGF and PIK3CA mRNA levels in both tissue specimens (P = 0.008), as well as established cell lines (P = 0.026, Fig. 2 A).

To confirm the correlation between PIK3CA and VEGF expression, we examined the expression of p110α and VEGF protein by IHC in serial sections of ovarian cancer specimens. VEGF immunolocalized predominantly to cells located within tumor islets but not in tumor stroma (Fig. 2, D and G). Select tumors exhibited strong immunostaining for VEGF. A significant correlation (P < 0.01) was found between the semiquantitative score of VEGF and that of p110α protein expression (Fig. 2,B), with two representative cases shown in Fig. 2,C–G. Furthermore, microvascular development, as assessed by CD31 staining, was markedly more pronounced in the tumors with high p110α expression as compared with tumors with low p110α expression. Two representative cases are shown in Fig. 2, E and H. These data collectively suggest a possible functional relationship between PIK3CA kinase activity and VEGF expression in ovarian cancer.

To test whether PI3k is involved in the regulation of VEGF expression in ovarian cancer, we examined the ability of the PI3k inhibitor Ly294002 (43) to block VEGF or HIF in established cell lines. Our previous work demonstrated that glucose starvation can dramatically up-regulate VEGF expression in ovarian cancer cell lines (46). We therefore subjected cells to glucose starvation in the presence or absence of Ly294002 (10 μm) for 24 h. VEGF was analyzed by quantitative real-time RT-PCR and IHC. In all of the five cell lines tested, VEGF mRNA was up-regulated by glucose starvation; this effect was largely negated by the presence of Ly294002 (Fig. 3,A). These results were confirmed at the protein level by IHC (Fig. 3,B). To test whether PI3k regulates VEGF via HIF, cells were treated at resting conditions with 10 μm Ly294002 for 36 h. Incubation with the inhibitor significantly suppressed the mRNA levels of HIF-1α as assessed by quantitative real-time RT-PCR in all of the five ovarian cancer cell lines (Fig. 3,C). In addition, Ly294002 suppressed the induction of HIF-1α by glucose starvation (data not shown). These results were additionally confirmed by Western blot analysis. HIF-1α protein levels markedly decreased after exposure of A2008 or A2780 cells to Ly294002 (Fig. 3 D). Collectively, these data indicate that PIK3CA might play an important role in the regulation of the constitutive as well as induce expression of HIF-1α and VEGF in ovarian cancer.

Because PIK3CA promotes angiogenesis and because it is a putative oncogene in ovarian cancer (1), we hypothesized that PIK3CA overexpression is associated with enhanced tumor cell survival and proliferation in vivo. Using the median of PIK3CA mRNA level (21.8 relative expression units) as the cutoff point, all advanced stage tumors were separated into two groups: one with relatively high and one with relatively low expression of PIK3CA mRNA. Cell proliferation was detected by Ki-67 staining. Ki-67-positive cells were found predominantly in tumor islets and, occasionally, in tumor stroma. Double immunofluorescent staining of Ki-67 and cytokeratin revealed that 99% of Ki-67-positive cells were also cytokeratin positive and were therefore tumor cells (Fig. 4,A), whereas <1% of cells within tumor islets were Ki-67 positive and cytokeratine negative. The ratio of Ki-67-positive cells over total cells was used to compute the tumor proliferation index. Three random fields of tumor islets/tumor were analyzed by computer-assisted imaging analysis (examples of tumor with low or high PIK3CA expression are provided in Fig. 4, C and D, respectively). The proliferation index was significantly higher in tumors with high PIK3CA expression compared with those with low PIK3CA expression (P < 0.05, Fig. 4,B). We additionally assessed the viability of tumor cells by examining cell proliferation and apoptosis in the same specimens by combined fluorescent immunostaining and ApopTag in Situ TUNEL assay. Consistent with the result from IHC, higher prevalence of proliferating cells and lower prevalence of apoptotic cells were detected in tumors with high PIK3CA expression compared with those with low PIK3CA expression (Fig. 4, E and F).

To additionally confirm the effects of PIK3CA-signaling pathway on ovarian cancer cells, we treated human ovarian cancer cell lines A2008, A1847, A2780, and its platinum-resistant clone A2780/CP70, with Ly294002. It was found that Ly294002 significantly decreased cell proliferation in vitro at a low dose (10 μm; Fig. 4,G) while potently induced cell apoptosis at a high dose (50 μm; Fig. 4 F). Collectively, these data are consistent with the reports from other groups (1), supporting the function of PIK3CA as an important oncogene in ovarian carcinoma.

Although in vitro evidence strongly suggests that PIK3CA functions as an important oncogene in ovarian cancer (1), the prevalence of PIK3CA overexpression in vivo, especially in early stage ovarian cancer, remains to date unknown. Furthermore, the effects of PIK3CA overexpression on the tumor microenvironment have not been elucidated. In this study, we report that PIK3CA overexpression is highly prevalent in ovarian carcinoma. Our results are in agreement with the previously reported increased copy number of PIK3CA gene and increased PIK3CA transcription in ovarian cancer cell lines (1). Our findings also suggest that PIK3CA overexpression may occur early during malignant transformation in the ovary because mRNA levels were similar in early- and late-stage tumors. These findings are in agreement with findings that gain in 3q26 region, where PIK3CA is located, are encountered frequently in early-stage ovarian carcinoma (48, 49).

In this study, we provide direct evidence that PIK3CA overexpression induces increased angiogenesis in ovarian carcinoma because of up-regulation of VEGF via HIF. These data provide an explanation for the reported complete abrogation of ascites during in vivo treatment with Ly294002 in a xenograft i.p. model of ovarian carcinoma (50) because VEGF is directly implicated in ascites formation in ovarian carcinoma. These data are also in agreement with evidence that PIK3CA^del/del embryos demonstrated areas of defective angiogenesis (51). PIK3CA is therefore added to the list of other oncogenes overexpressed in ovarian carcinoma, which may initiate angiogenesis through HIF activation.

Interestingly, immunoreactive p110α was strongly expressed in tumor stroma endothelial cells in 75.7% of tumors examined. Tumor vasculature claims a morphologically and functionally unique phenotype, greatly distinct in growth properties, functions, and responses to angiogenic factors (52). The PI3k-signaling pathway plays an important role in endothelial cell migration and proliferation during angiogenesis (4, 36, 39, 40, 41). Expression of p110α at levels detectable through IHC may therefore reflect activation of tumor endothelial cells and enhanced angiogenesis. In fact, vascular endothelium from postmenopausal ovaries was negative for p110α expression. Conversely, strong expression of p110α was detected in the muscularis layer of vessels from postmenopausal ovaries, possibly reflecting a physiological role of PIK3CA in menopause-induced vascular smooth muscle changes (53, 54).

The PI3k-signaling pathway is involved in multiple aspects of malignant transformation (1, 2, 3, 55, 56, 57, 58). We found that PIK3CA overexpression was associated with increased cell proliferation and decreased apoptosis, suggesting an important association with overall enhanced tumor growth. This effect may be partly attributed to the observed effect of PIK3CA on angiogenesis because increased angiogenesis and microvascular density closely correlates with tumor cell survival and proliferation and tumor growth (59). In agreement with a previous article (1), we found that PIK3CA promotes also cell survival and proliferation directly in ovarian cancer in vitro.

In summary, this study provides strong evidence that PIK3CA overexpression plays an important role in ovarian cancer. Collectively, our data indicate that PIK3CA promotes tumor growth through at least two distinct pathways: it enhances cell survival and proliferation through a direct mechanism and it up-regulates VEGF via HIF to promote angiogenesis. PI3k and its associated pathway have been considered as a potential antitumor therapeutic target (4, 50, 60, 61, 62, 63), and PI3k blockade has shown efficacy in a xenograft model of ovarian carcinoma (50, 62). Our data support the notion that therapy strategies targeting PI3k or directly PIK3CA (e.g., antisense or interfering RNA therapy) should contribute not only to the inhibition of oncogenic signaling pathways but also to the inhibition of tumor angiogenesis.

We thank Drs. Kang-Sheng Yao and Steven Johnson (Department of Pharmacology, University of Pennsylvania) for donating the human ovarian cancer cells. We also thank Ann O’Brien-Jenkins and Alisha Mohamed-Hadley for their excellent technical support.