DLC1 negatively regulates angiogenesis in a paracrine fashion

Abstract

The Rho GTPase activating protein DLC1 is a tumor suppressor that is often deleted in liver cancer and downregulated in other cancers. DLC1 regulates the actin cytoskeleton, cell shape, adhesion, migration, and proliferation through its RhoGAP activity and focal adhesion localization. In this study we silenced DLC1 in non-malignant prostate epithelial cells to explore its tumor suppression functions. shRNA-mediated silencing of DLC1 was insufficient to promote more aggressive phenotypes associated with tumor cell growth. In contrast, DLC1 silencing promoted pro-angiogenic responses through VEGF upregulation, accompanied by the accumulation of hypoxiainducible factor HIF1α and its nuclear localization. Notably, modulation of VEGF expression by DLC1 was dependent on EGFR-MEK-HIF1 signaling but on RhoA pathways. Clinically, VEGF upregulation is a highly significant event in prostate cancers where DLC1 is downregulated. Thus, our results strongly suggest that loss of DLC1 may serve as a “second hit” in promoting angiogenesis in a paracrine fashion during tumorigenesis.

INTRODUCTION

Deleted in liver cancer 1 (DLC1) is a tumor suppressor that was originally identified in primary hepatocellular carcinoma(1). In addition to liver cancer, loss or reduced of DLC1 expression due to gene deletion or promoter methylation has been reported in lung, prostate, breast, kidney, colon, uterus, ovary, and stomach cancers(2-4). Mutations that altered the expression and function of DLC1 were detected in pancreas(5), colon, and prostate cancers(6). DLC1 is shown to regulate actin cytoskeleton and focal adhesion organizations, cell shape, adhesion, migration, proliferation, and apoptosis(2-4). These functions may directly contribute to DLC1's suppressive activities in tumorigenicity and metastasis(2-4). The RhoA pathway negatively regulated through DLC1's RhoGAP domain is believed to be critical for these functions, which were mainly analyzed by ectopical expression approaches in cancer cell lines.

Angiogenesis is the formation of new blood vessels from the existing vasculature and is essential for the growth of the primary cancer and for the formation of metastasis(7). Vascular endothelial growth factor (VEGF) plays a major role in tumor angiogenesis. It can promote the proliferation, survival, and migration of endothelial cells and is essential for blood vessel formation(8). Hypoxia is the strongest stimulus for triggering the VEGF expression in cancer cells. Nonetheless, many cancer cell lines express high levels of VEGF in normoxia(9).

Here, we have discovered a novel function of DLC1 in regulating angiogenesis. Silencing of DLC1 in non-malignant prostate epithelial cells leads to up-regulation of VEGF that promotes angiogenesis in vivo and in vitro. This up-regulation of VEGF is mediated through the EGFR-MEK-HIF1 pathway. Clinically, up-regulation of VEGF is highly associated with decreased DLC1 in prostate cancer.

MATERIALS AND METHODS

Cell culture and reagents

MLC-SV40 kindly provided by Dr. Johng Rhim (Center for Prostate Disease Research, Bethesda, Maryland)(10), and RWPE-1 cells purchased from the American Type Culture Collection (ATCC, CRL-11609) were cultured in keratinocyte serum-free medium (Invitrogen). Human vascular endothelium cells (HUVEC) from ATCC (CRL-1730) were cultured in Endothelial Cell Growth Medium (Genlantis). Cell lines were used within 3 months after receipt or resuscitation of frozen aliquots. The authentication of these cell lines was assured by the provider by cytogenetic analysis. No additional test was done specifically for this study. Lipofectamine-2000 (Invitrogen) was used for transfections. Stable shGFP or shDLC1 cells were generated by infection with shRNA lentiviruses against GFP or DLC1 (Sigma-Aldrich), followed by puromycin (2.5ug/ml) selection. ELISA kit (R&D) was used to determined VEGF levels in conditioned media. RhoA activity was measured with RhoA Activation Assay Kit (Cytoskeleton).

Xenograft assay

Growth factor-reduced Matrigel (BD Biosciences) containing 60 U/ml heparin (Sigma-Aldrich) was mixed with 2×10⁶ cells, and subcutaneously injected into nude mice. After 5 days, cell plugs were harvested and embedded in OCT for immunohistochemical staining using CD31 antibody and VEGF antibody.

In vitro matrigel angiogenesis assay

Growth factor–reduced Matrigel was used to coat 96 wells plate (50ul/well) and HUVECs (20,000 cells/well) were seeded with conditioned medium (200ul). After 4hrs of incubation, capillary-like structures were scored by measuring lengths of tubules per field in each well at 100X magnification with Image J software (NIH).

Aortic ring assay

Thoracic aortas from C57BL/6 mice were dissected and transferred to ice-cold PBS. The fat tissue was removed and 1mm long aortic rings were sectioned and embedded in growth factor-reduced Matrigel. Rings were co-cultured with 500 ul conditioned medium with or without 1 ug anti-VEGF blocking antibody (R&D, clone26503) for 8 days, and the outgrowth of endothelial tubes was counted.

Migration assay

HUVECs (80000 cells) were added to the upper chamber in each transwell. Conditioned media (400μl) were added to the lower chamber. Cells were fixed and stained 5hrs later. Cells migrated to lower surface were visualized microscopically and photographed. For VEGF blockade, conditioned medium was incubated with 1 ug anit-VEGF blocking antibody (R&D) at room temperature for 1hr prior to experiments.

Adenoviruses

Human DLC1 cDNA was subcloned into pENTR1A vector and the DLC1/pENTR clone was used in a site-directed recombination reaction to place DLC1 cDNA into the pAD/CMV/V5-DEST vector (Invitrogen). The adenoviral expression clone was transfected into 293A cells. After 10-12 days, the crude viral lysate was harvested and used for infection.

Immunohistochemical Staining, Scoring and MVD counting

Prostate normal/cancer tissue arrays (Imgenex) were dewaxed and rehydrated. After antigen retrieval, endogenous peroxidase activity was blocked with 3% hydrogen peroxide in methanol for 20min followed by normal serum blocking. Slides were incubated at 4 °C overnight with anti-DLC1 (1:50, BD Biosciences, Clone3), anti-VEGF (1:50, Santa Cruz biotechnology, SC152), or anti-CD31 (1:200, Pharmingen, 01951D). Signal was detected with Vectastain ABC Elite Kit (Vector Laboratories) and diaminobenzidine substrate. Slides were counterstained with hematoxylin. Images were observed by Zeiss Axioplan2 microscope. For MVD counting, the vessels were detected by CD31 immunostaining and mean values of the vessel count were calculated using the average of three most intense vascularization areas at 200X magnification. Tissue array immunoreactivities were scored by the intensity of the staining (0, no staining; 1, weak; 2, moderate; 3, strong) and the percentage of stained cells (0, no staining; 1, 1-10%; 2, 10-50%; 3, 50-80%; 4, >80%). By multiplication of both values a final score ranging between 0 and 12 was obtained.

RESULTS

Silencing of DLC1 in non-malignant prostate epithelial cells does not promote the tumorigenicity but does enhance angiogenesis

To investigate the loss of function of DLC1 in normal prostate epithelial cells, we have generated DLC1 knockdown (shDLC1) by shRNA in the non-malignant prostate epithelial cell line, MLC-SV40. DLC1 protein level was significantly reduced in shDLC1 than control shGFP cells and as expected the RhoA activity was enhanced due to the reduction of the negative regulator DLC1 (fig. 1A). To examine whether the loss of DLC1 enhanced their tumorigenicity, we injected shDLC1 or shGFP cells into nude mice. None of them developed tumors during three months observation (not shown). Interestingly, we detected increased small blood vessels (CD31 positive) around injected shDLC1 cells at 5 days post-injection and this was confirmed by significantly higher microvascular density (MVD) in shDLC1 (fig. 1B&C). We have generated additional pair of shDLC1 and shGFP in another non-malignant prostate cell line RWPE-1. Similar results were observed using this pair of RWPE-1 cells throughout this project (fig. 1A) and only data from MLC-SV40 cells are shown. These results suggest that silencing of DLC1 did not promote the proliferation of prostate epithelial cells but somehow it enhanced angiogenic responses of endothelial cells.

Figure 1. Silencing of DLC1 in prostate epithelial cells promotes angiogenesis in xenograft assays.

(A) DLC1 and RhoA expression levels in indicated cell lines were analyzed by immunoblotting. The amounts of active GTP-bound Rho GTPases were determined by GST-RBD pull-down assay followed by anti-RhoA immunoblotting. (B) Cells were injected subcutaneously into nude mice. Cell plugs were removed 5 days post-injection and processed for IHC staining using anti-CD31 and anti-VEGF antibodies. MVD and VEGF levels were scored and representative IHC images were shown (C). Arrow and arrowhead show CD31 and VEGF positive cells, respectively. Bar=50μm.

DLC1 negatively regulates VEGF expression

Since VEGF is the major pro-angiogenic factor, we examined VEGF levels in xenografts by immunohistochemical (IHC) staining and found significantly elevated VEGF staining in injected shDLC1 cells, which were accompanied with CD31 positive endothelial cells (fig. 1B&C). In cultured cells, VEGF mRNA and protein levels were up-regulated in shDLC1 cells and the conditioned medium from shDLC1 also contained higher levels of VEGF (fig. 2A), confirming that lack of DLC1 promotes VEGF expression and secretion. Furthermore, shDLC1 conditioned medium enhanced angiogenic related responses of endothelial cells by forming more tube-like networks, faster cell migration, and more sprouting of capillary (fig. 2B). These effects were significantly reduced when anti-VEGF blocking antibody was applied (fig. 2B). However, by adding a similar amount of recombinant VEGF (500 pg/ml) to fresh medium did not have same effects as detected with shDLC1 conditioned medium, suggesting that either recombinant VEGF was not as potent as endogenous VEGF or there might be additional factor(s) in the conditioned medium that also contributes to the observed effects. Re-expression of shRNA-resistant DLC1 in shDLC1 markedly suppressed VEGF expression (fig. 2C). A similar effect was detected when re-expression of DLC1 in prostate cancer cell lines, including LnCap, DU145, and even VEGF low-expressing PC3 cells (fig. 2C), confirming that DLC1 negatively regulates VEGF expression.

Figure 2. DLC1 regulates the expression of functional VEGF.

(A) VEGF mRNA, total protein, and secreted protein levels in MLC-SV40 system were determined by RTPCR, immunoblot, and ELISA assays. (B) Representative images and results of HUVEC tube formation (upper), aortic ring (middle), and migration (lower) assays using conditioned media from shGFP, shDLC1, or shDLC1 incubated with anti-VEGF blocking antibody. (C) VEGF levels of shDLC1, PC3, LnCap, and Du145 cells infected with adenovirus expressing LacZ (AdLacZ as control) or DLC1 (AdDLC1) were measured by ELISA. All data are mean SD from triplicate experiments. * p<0.05.

DLC1-regulated VEGF expression is mediated through the EGFR-MEK-HIF1 but not RhoA pathways

Since the RhoGAP activity is critical for a variety of DLC1's regulatory function, including cell shape, migration, and tumorigenicity, we test whether VEGF up-regulation in shDLC1 cells is regulated through the RhoA pathway. By overexpression of RhoA wild-type, constitutive active and dominant negative mutants in MLC-SV40 cells, or silencing of RhoA expression in shDLC1 cells, VEGF levels were not affected (fig. S1), suggesting that DLC1-mediated VEGF expression is not regulated by the RhoA pathway.

To investigate the potential pathway(s) involved, several pharmacological inhibitors were used. While Jak (AG490) inhibitor had no effect on VEGF expression, inhibitors to EGFR (AG1478, AG1517) and MEK (PD98059, U0126) markedly reduced VEGF expression in shDLC1 cells in a dose-dependent manner (fig. 3A). These inhibitor dosages have no effect on cell proliferation (fig. S2). Since hypoxia-inducible factor 1 (HIF1) is a major transcription factor that mediates VEGF expression, we measured the protein and subcellular localization of HIF1α. Indeed, HIF1α protein level was increased and accumulated in the nuclei of shDLC1 cells (fig. 3B&C). Concomitantly, increased EGFR(Tyr1068), MEK(Ser217/221), and ERK(Thr202/Tyr204) phosphorylation levels were detected (fig. 3B). In addition, HIF1α protein levels and nuclear localization were markedly reduced by EGFR or MEK inhibitors (fig 3B-D). Furthermore, silencing of EGFR significantly decreased MEK and ERK phosphorylation, and VEGF expression in shDLC1 (fig. 3D). These data suggest that DLC1-mediated VEGF expression is modulated through the EGFR-MEK-HIF1 pathway.

Figure 3. VEGF up-regulation in shDLC1 cells is mediated through EGFR-MEK-HIF1 pathway.

(A) VEGF levels in conditioned media from shDLC1 cells treated with indicated inhibitors at various concentrations for 24 hrs were measured by ELISA. (B) Cell lysates were analyzed by immunoblotting with indicated antibodies. (C) Cells grown on coverslips treated with indicated reagents for 24 hrs were labeled for HIF1α and nuclei (Propidium Iodide) and visualized with Zeiss LSM 510 confocal microscope. Bar=10 μm. (D) VEGF levels in conditioned media from cells treated with control or EGFR siRNAs for 24 hrs were determined. The levels of EGFR, phosphor-ERK, phosphor-MEK in these cells were detected by immunoblotting. * p<0.05.

Down-regulation of DLC1 is associated with up-regulation of VEGF in prostate cancer

To examine DLC1 and VEGF expression patterns in clinical samples, we have stained DLC1 and VEGF in consecutive prostate tissue sections. By IHC analysis, DLC1 is highly expressed in normal prostate epithelial cells, whereas VEGF expression is very weak if any (fig. S3). In prostate cancer specimens, DLC1 expression was reduced in 86.3% (107/124), whereas VEGF protein level was increased in 75.8% (94/124) of tumor samples (table 1). When comparing the expression of VEGF in DLC1 down-regulated versus no change prostate cancer samples, increased VEGF expression is a statistically significant event in DLC1 down-regulated prostate cancer.

Table 1. DLC1 and VEGF expressions in prostate cancers.

IHC analysis comparing prostate cancer to normal tissues indicated that increased VEGF expression in DLC1 down-regulated prostate cancer was statistically significant. Fisher's exact test p=0.046.

Total 124 case	VEGF up (94)	VEGF down or no change (30)
DLC1 down (107)	78/124 (62.9%)	29/124 (23.4%)
DLC1 up or no change (17)	16/124 (12.9%)	1/124 (0.8%)

DISCUSSION

Our current studies demonstrate that loss of DLC1 tumor suppressor alone is not sufficient for tumorigenesis. This is in agreement with the finding that DLC1 knockdown had little impact on the colony formation of liver progenitor cells in vitro but had efficiently promoted tumor development in Myc expressing and lacking p53 liver progenitor cells engrafted into the mouse livers(11). Although lack of DLC1 does not increase cell tumorigenicity, it enhances the angiogenesis, which plays an essential role in cancer progression. Tumor growth and metastasis are dependent on angiogenesis, and preventions of tumor angiogenesis and/or metastasis are perceived as attractive approaches in regulation tumor progression. Since DLC1 also prevents cancer cell migration and metastasis(12-13), understanding DLC1-mediated various pathways might offer potential targets for both angiogenesis and metastasis by enhancing DLC1's function or suppressing its negatively regulated downstream molecules. Interestingly, our recent finding(14) showed that DLC2 knockout mice display enhanced angiogenic responses induced by matrigel or tumor cells. These data suggest that the DLC family may play various roles in angiogenesis.

Other focal adhesion molecules may regulate VEGF expression. For example, overexpression of ILK (integrin-linked kinase) stimulates VEGF expression and silencing of ILK reduces VEGF production in prostate cancer cells(15). TM4SF5, a transmembrane protein that interacts with integrins, regulates VEGF expression through an integrin α5-Src-STAT3 pathway(16). Both ILK and TM4SF5 positively regulate VEGF expression, while DLC1 negatively controls VEGF production. This negative regulation of DLC1 appears to be mediated thought the EGFR signaling, which is known to regulate the synthesis and secretion of VEGF in cancer cells. Maity et al (17) showed that EGFR activation regulates VEGF in glioblastoma cells through the PI3K pathway but not the MAPK pathway. In head and neck squamous carcinoma, it is dependent on MEK/MAPK but not PI3K pathways(18). These findings suggest that different mechanisms are involved in the ability of EGFR signaling in modulating VEGF expression in various types of cancer cells. Our data indicate that lack of DLC1 up-regulates VEGF in an EGFR-MEK-HIF1 dependent fashion. Currently, we are investigating how and whether DLC1 directly or indirectly regulates the activity of EGFR.