|Home | About | Journals | Submit | Contact Us | Français|
Recruiting a new blood supply is a rate-limiting step in tumor progression. In a 3D model of breast carcinogenesis, disorganized, proliferative transformed breast epithelial cells express significantly higher expression of angiogenic genes compared to their polarized, growth arrested non-malignant counterparts. Elevated VEGF secretion by malignant cells enhanced recruitment of endothelial cells (EC) in heterotypic co-cultures. Significantly, phenotypic reversion of malignant cells via re-expression of HoxD10, which is lost in malignant progression, significantly attenuated VEGF expression in a HIF1α-independent fashion and reduced EC migration. This was due primarily to restoring polarity: forced proliferation of polarized, non-malignant cells did not induce VEGF expression and EC recruitment, whereas disrupting the architecture of growth-arrested, reverted cells did. These data demonstrate that disrupting cytostructure activates the angiogenic switch, even in the absence of a proliferation and/or hypoxia and restoring organization of malignant clusters reduces VEGF expression and EC activation to levels found in quiescent non-malignant epithelium. These data confirm the importance of tissue architecture and polarity in malignant progression.
It is well established that solid tumors require angiogenesis to survive (1, 2) and changes in the breast tumor microenvironment promote formation of new blood vessels (3-5) with high angiogenic potential linked to poor prognosis (4, 6).
In breast tumors, a number of key angiogenic factors have been identified; the most prominent being vascular endothelial growth factor (VEGF), which acts on adjacent endothelial cells (ECs) through the VEGF receptor 2, initiating growth, migration and invasion into adjacent tumor stroma. Recent clinical studies have shown that function blocking antibodies against VEGF significantly impairs tumor progression (7).
In the earliest stages of malignant breast cancer (i.e. ductal carcinoma in situ, DCIS), VEGF may be induced by the increased metabolic demand and hypoxia (8) as low oxygen tension stabilizes HIF1α which binds to, and activates, transcription of the VEGF promoter (for review, see (9)). Yet, paradoxically, most cells suspend mRNA translation and protein synthesis when faced with either nutrient depletion or hypoxia (10, 11) and suggests additional changes in the breast tumor microenvironment may facilitate expression of angiogenic factors.
Sustained signaling through α6ß4 integrin and elevated PI3 Kinase (PI3K) levels also enhance translation of VEGF mRNA in carcinoma cells (12) and even in the absence of hypoxia, both PI3K and MAPK signaling are elevated in breast tumors, and can increase the transcription and secretion of VEGF, independent of HIF1α (13, 14). Expression of the transcription factor HoxB7 also increases VEGF expression in both breast and other epithelial tumor cells; similarly, ß-catenin activation can induce the expression of angiogenic factors, including VEGF (15, 16). Thus, a variety of changes in breast tumor cells conspire to activate expression of angiogenic factors.
Identifying features of the normal breast microenvironment, which collectively normalize MAPK, PI3K, sequester ß-catenin and/or attenuate α6ß4 expression may prove to be key in inhibiting the angiogenic switch. A number of studies by our group have shown that breast tissue architecture plays a fundamental role in mediating each of these pathways(17-20). Further, despite many genetic defects, malignant breast epithelial cells, which exhibit a proliferative, unpolarized morphology when grown in three-dimensional (3D) cultures (21), can be reverted to a polarized, acinar morphology and growth arrested by agents targeting the MAPK and PI3K pathways (17-19, 22). We have also shown that restoring expression of a key morphoregulatory gene, HoxD10, lost in tumorigenic breast epithelial cells reverts tumorigenic breast cells to a growth-arrested and organized phenotype (23). Whether the paralogous HoxA10 gene, which is also lacking in some breast tumors(24), also stabilizes breast tissue architecture is not known. Nonetheless, considering the remarkable dominance of the reverted tumor cell phenotype over the tumor cell genotype, we hypothesized that proper organization of breast epithelial cells may suppress expression of angiogenic factors, thus implicating loss of tissue organization as a key activator of the angiogenic switch in breast cancer.
Immortalized human dermal microvascular ECs HMEC-1 (a gift from T. Lawley (25) Emory University, Atlanta, GA) the human breast epithelial cell line MDA-MB-231 (American Type Culture Collection, Manassas, VA) and epithelial cell lines HMT-3522 T4-2, S-1 and S1 EGFR were grown and maintained in 2D cultures and 3D cultures as previously described (18, 19).
Cell migration assays were performed using a modification of procedures previously described (23, 26). Briefly, 6.5 mm transwell chambers (8 μm pore) (Corning, Acton, MA) were coated with 10 μg/mL of type I Collagen (Cohesion Tech, Palo Alto, CA) and 5 × 104 serum-starved HMEC-1 plated in 300 μL of fibroblast basal medium (FBM; Lonza, Basel, Switzerland) containing 0.5% bovine serum albumin (BSA).
For co-culture experiments, T4-2, S1 or MDA-MB-231 cells were cultured using polymerized laminin-rich extracellular matrix (lrECM) (Matrigel, BD Biosciences, Bedford, MA.) for 72 hours and 16 hours prior to assays, the media changed to serum-free FBM and transwell inserts with EC added to the upper chamber. After 4 hours at 37°C, EC on the upper surface were removed and HMEC-1 migrated onto the bottom of the membrane were stained with Diff-Quick (VWR Scientific Products, West Chester, PA) and five fields in each well were counted by phase-contrast microscopy (magnification, ×20). When indicated, HMEC-1 were pre-incubated for 30 minutes with 0.5 μg/mL of control IgG or a monoclonal antibody against anti-human VEGF (R&D, Minneapolis, MN).
The human 1,100 bp HoxD10 cDNA (GenBank accession no. X59373) was cloned into the EcoRI site of the pBABE retroviral vector (Clontech, Palo Alto, CA). T4-2 cells and T4-2 Rac1L61 were transduced with control plasmid (pBABE) or pBD10 and selected in 0.5 μg/mL Puromycin (Sigma) as previously described (23).
Cells grown in 3D were released from lrECM using previously described procedures (18, 23) and cell pellets resuspended in RNA lysis buffer and extracted using the RNeasy Mini isolation kit (Qiagen, Valencia, CA). One microgram of total RNA was reverse transcribed using MMLV reverse transcriptase (Qiagen) and 1/25 of this reaction was linearly amplified for 30 cycles (VEGF, HoxD10) of denaturation (30 seconds at 95°C), annealing (30 seconds at 51°C for VEGF; 58°C for HoxD10), and extension (30 seconds at 72°C) in a thermal cycler (PTC-200 Peltier Thermal cycler, M.J. Research, Cambridge, MA). The following primers were used: HoxD10 forward 5' CTGTCATGCTCCAGCTCAACCC 3', reverse 5' CTAAGAAAACGTGAGGTTGGCGGTC 3'; VEGF forward 5' CGAAACCATGAACTTTCTGC 3', reverse 5' CCTCAGTGGGCACACACTCC 3'. Total RNA was normalized using 18S internal standards at a 1:3 ratio (Ambion, Austin, TX).
Quantitative real-time PCR was carried out in triplicate with a 10-20-fold dilution of first-strand cDNA using human Taqman probes and primers purchased as Assays On Demand (Applied Biosystems, Foster City, CA) for β-glucuronidase (Gus - reference gene control) and VEGF using an ABI Prism SDS 7000 (Applied Biosystems, Foster City, CA) according to the manufacturer's instructions. Data was analyzed with ABI Prism SDS 7000 companion software.
For 2D cultures, T4-2 and S1 cells were plated at a density of 500,000 cells/well and for 3D cultures, cells were plated at 100,000 cells/well on top of 200 μL of polymerized lrECM and overlaid with 10% lrECM. 48 hours later media was changed to FBM + 0.5% BSA and 24 hours later secreted VEGF was assayed in triplicate by ELISA (DVE00, R&D Systems, San Diego, CA), according to the manufacturer's instructions.
The relative expression of 84 angiogenesis-related genes were evaluated using the Human Angiogenesis RT2 Profiler PCR Array system (SuperArray Bioscience Corporation, Frederick, MD) according to the manufacturer's instructions. DNase-treated total RNA was purified from T4-2 cells treated with either the EGFR blocking antibody mAb225 or an IgG control. cDNA was generated by reverse transcription from 1 μg of total RNA from each sample using the RT2 First Strand Kit, then combined with the RT2 qPCR Master Mix and added to lyophilized primer pairs in the 96 well arrays. Thermal cycling was performed in a Biorad iCycler (Biorad, Hercules, CA). Relative gene expression levels were calculated using the ΔΔCt method (27) with normalization to the average expression level of five common genes (ACTB, B2M, GAPDH, HPRT and RPL13A).
Gene expression analysis was performed on samples of RNA purified from S1 and T4-2 cells grown in 3D lrECM cultures in the presence of various signaling inhibitors or vehicle controls. The Affymetrix High Throughput Array (HTA) GeneChip system, with HG-U133A chips mounted on pegs in a 96 well format was used for the analysis, as described (28). Data were imported into the Partek Genomics Suite (Partek Inc., St. Louis, MO) and normalized using RMA (29).
Cells were cultured in 3D lrECM for 72 hours, released and lysed in 10 mM Tris-HCl (pH 7.4), 1 M Sodium Chloride, 1% Triton-X, 50 mM sodium fluoride, 1 mM sodium orthovanadate, and protease inhibitor cocktail. Total protein was determined using the BCA assay (Pierce, Rockford, IL) and 40 micrograms electrophoresed on SDS-PAGE, transferred to PVDF membranes and blocked with 5% milk. The following antibodies were used: HoxD10 (E-20) polyclonal antibody (sc-33005, Santa Cruz Biotechnology, Inc., Santa Cruz, CA), a polyclonal phospho-AKT antibody ser473 (193H12, Cell Signaling, Danvers, MA), a mouse monoclonal HIF1α antibody (NB 100-105, Novus Biologicals, Littleton, CO) and detected with enhanced chemiluminescence system (Amersham BioSciences, Piscataway, NJ). Relative protein loading was assessed by β-actin (Ab8227, Abcam, Cambridge, MA). Nuclear extracts were isolated from T4-2 and HoxD10 reverted T4-2 cells and electrophoretic mobility shift assays (EMSA) were performed using procedures as described (30).
After release from 3D lrECM, cells were smeared onto slides, fixed in cold 1:1 methanol-acetone as previously described (23). After blocking with 10% goat serum, cells were incubated overnight with a 1:100 dilution of antibodies against β4 integrin (mAb1964, Chemicon, Temecula, CA), washed with immunofluorescence buffer, followed by a 1:400 dilution of goat anti-mouse Alexa Fluor 546 IgG (H+L) (Invitrogen) and nuclei counterstained with 1:1000 dilution of 4'6-diamidino-2-phenylindole (DAPI, Sigma). Slides were mounted in Fluoromount G (Southern Biotechnology Associates, Inc., Birmingham, AL) and images were collected with a Nikon Eclipse TE300 fluorescence microscope.
Proliferation was assessed by KI-67 immunostaining using a modification of the previous method (23) with a 1:500 dilution of KI-67 Ab (VP-K451, Vector Laboratories, Burlingame, CA) overnight at 4°C, followed by a 1:400 dilution of goat anti-rabbit Alexa Fluor 546 IgG (H+L) (Invitrogen), counterstained with DAPI and mounted in Fluoromount G. Proliferation was determined by visually counting at least 300 DAPI-labeled nuclei and thereafter scoring KI-67 positive cells as a percentage of a total cell number.
Malignant breast epithelial cells (T4-2) cultured in 3D lrECM exhibit a disorganized morphology compared to their growth arrested, polarized non-malignant counterparts (S1). Disrupting either β1 integrin, EGFR, MAPK, or PI3K-mediated signaling restores basolateral polarity, growth arrest and reverts malignant cells to a phenotype similar to non-malignant cells and fail to form tumors in vivo (18, 19, 31, 32). We utilized microarray analysis to determine whether malignant T4-2 cells display enhanced expression of angiogenic factors and whether phenotypic reversion reduces expression of angiogenic factors. We analyzed RNA samples from S1, T4-2, and T4-2 cells reverted with the EGFR inhibitor, AG1478, (in duplicate), and T4-2 cells reverted with an EGFR blocking antibody mAb225, a β1-integrin blocking antibody AIIB2, or inhibitors of MEK (PD98059) and TACE (TAPI-2). Unsupervised hierarchical clustering of all samples (Fig. 1A) revealed significantly different transcriptional profiles of S1, T4-2 and reverted T4-2 cells. To identify gene expression changes associated with polarity, we selected genes differentially expressed between disorganized T4-2 colonies, organized non-malignant S1 colonies and reverted T4-2 cells (Fig. 1B, t-test, P < 0.001). Differentially expressed genes that were significantly higher in T4-2 compared to non-malignant or reverted T4-2 included several of the Gene Ontology angiogenesis class (GO:0001525)(Fig. 1C).
For independent confirmation, we performed an RT2 Profiler PCR array for 84 different angiogenic mediators with RNA isolated from control (IgG) T4-2 and T4-2 reverted with a function blocking antibody against EGFR (Mab225) (Fig. 1D). Expression of VEGF, and other angiogenic factors including FGF-2, MMP-9, placental growth factor (PGF) and VEGF-C (33) were significantly reduced in reverted cells. Thus expression of angiogenic genes may be restrained when cells adopt a polarized tissue morphology.
To assess a functional role for angiogenic factors linked to tissue polarity, we focused on VEGF. As most soluble inhibitors that revert tumor cells also inhibit signaling in adjacent ECs, we transfected T4-2 cells with HoxD10, which we previously showed attenuates growth and restores polarity in metastatic breast tumor MDA-MB-231 cells(23). Although T4-2 cells express low levels of HoxD10 mRNA and protein (Fig. 2A), restoring expression of HoxD10 did not induce any observable differences in morphology when grown on conventional polystyrene tissue culture plates (Fig. 2B upper panel). However, when cultured within a 3D lrECM, HoxD10-expressing T4-2 cells formed organized, growth-arrested structures resembling their non-malignant counterparts, as shown by phase microscopy and immunofluorescent imaging of ß4 integrin in which total levels were reduced (not shown) and the remaining re-distributed to the basal surface (Fig. 2B lower panels). In 3D lrECM, both non-malignant S1 and HoxD10-reverted T4-2 cells produced significantly less VEGF compared to control T4-2 cells (Fig. 2C). In contrast, when cultured on polystyrene, VEGF mRNA and protein were similar in S1, T4-2, and HoxD10-expressing T4-2 cells (data not shown).
We developed a modified Boyden invasion assay (Fig. 2D, inset) to quantify EC migration in response to malignant or non-malignant cells cultured in lrECM and observed that ECs migrated significantly more in response to co-culture with T4-2 cells than S1 cells (Fig. 2D). Pre-treatment of ECs with VEGF-blocking antibodies reduced EC migration to the basal levels seen in co-cultures with S1 cells (Fig. 2D), implicating VEGF as the primary mediator of EC migration in EC/T4-2 co-cultures. Notably, migration of ECs co-cultured with reverted HoxD10-expressing T4-2 cells was significantly reduced to levels observed by blocking VEGF or co-culturing with non-malignant cells (Fig. 2D). No differences in EC migration were observed when co-cultured with S1, T4-2 or HoxD10-expressing T4-2 cells grown in tissue culture plastic (data not shown).
Together these data suggest that differential expression of VEGF by non-malignant and tumorigenic epithelial cells is influenced by their respective tissue architecture.
We also compared expression of VEGF in MDA-MB-231 cells, HoxD10-expressing MDA-MB-231 cells and MDA-MB-231 expressing the paralogous HoxA10 (Fig. 3A, upper). In contrast to HoxD10, restoring HoxA10 did not induce organized structures in MDA-MB-231 cells in3D lrECM, as evidenced by diffuse IF staining of ß4 integrin (Fig. 3A, lower panel) and growth was not significantly reduced as determined by Ki-67 labeling (Fig. 3A). Importantly, levels of VEGF mRNA remained high in non-polarized control or HoxA10-expressing cells compared to polarized HoxD10-expressing cells (Fig. 3B). EC migration during co-culture with MDA-MB-231 could be blocked by antibodies against VEGF or by re-expression of HoxD10, but not by HoxA10 expression (Fig. 3C). Thus, only HoxD10, which reverts MDA-MB-231 to a growth-arrested and basally polarized phenotype, attenuates VEGF expression and EC migration.
To evaluate whether growth arrest or re-establishment of basolateral polarity was suppressing VEGF, we compared non-malignant S1 cells with S1 cells constitutively expressing the EGF receptor (S1-EGFR) that we previously showed maintained basal polarity in lrECM despite a ten-fold increase in proliferation (19). VEGF expression by proliferating, polarized S1-EGFR cells was not significantly different from growth-arrested S1 cells (Figs. 4A). Moreover, there were no significant differences in EC migration in response to S1 or S1-EGFR cells (Fig. 4B) indicating that growth arrest was not sufficient to suppress VEGF expression or EC activation.
VEGF can be induced by HIF1α, α transcription factor stabilized by hypoxic tumor microenvironments (34, 35). We investigated whether reduced VEGF expression in reverted tumor cells was accompanied by reduced HIF1α. Western blot analysis shows that relative levels of HIF1α were unchanged in 3D cultures of control T4-2 or reverted T4-2 cells (Fig. 4C). We also assessed HIF1α binding to consensus sites in target DNA using EMSA and observed similar levels of bound HIF1α in both control and reverted T4-2 cells (Fig. 4C). Thus, reduced VEGF in reverted, polarized tumor cells are not linked to changes in HIF1α expression or activity
Given that neither changes in HIF1α or continued growth in polarized cells induced VEGF expression, we exploited our previous findings that proliferation and polarity are mediated by distinct signaling pathways (22) to establish whether polarity alone regulated VEGF. A dominant-active Rac mutant (Rac1L61) was introduced into HoxD10-expressing T4-2 cells (Fig. 5A), which disrupted acinar polarity, as indicated by diffuse ß4 integrin staining, without significantly affecting proliferation (Fig. 5B). The loss of polarity was accompanied by increased production of VEGF to levels found in non-polarized control T4-2 cells (Fig. 5C) and significantly increased EC migration compared to polarized T4-2 cells expressing only HoxD10-expressing (Fig. 5D).
Together, these data emphasize that maintaining tissue structure is critical for suppressing VEGF expression and subsequent activation of adjacent EC by malignant cells in the breast tumor microenvironment (Fig. 6).
In the present study we show that production of angiogenic factors and EC activation is primarily controlled by breast epithelial cell polarity and tissue architecture. In conventional (2D) tissue culture, non-malignant and malignant breast epithelial cells do not differentially secrete VEGF or recruit ECs. However, once provided with 3D lrECM, non-malignant cells organize into quiescent, polarized acini, produce low levels of VEGF and limit migration of co-cultured ECs. Disorganized, malignant epithelial cells, on the other hand, produce more VEGF and significantly increase recruitment of ECs, mimicking the stromal angiogenic response of malignant breast tumors in vivo. Significantly, phenotypically reverting malignant epithelial cells to polarized acini via reintroduction of the HoxD10 tumor suppressor or by treatment with various signaling inhibitors(20) reduces VEGF and reduces migration of co-cultured ECs to levels observed with non-malignant cells. Hence, differential expression of VEGF by non-malignant and malignant breast epithelial cells is evident only when cells undergo characteristic morphological changes in 3D lrECM. Morever, reduced VEGF transcription in tumor cells re-polarized by HoxD10 is unlikely a direct effect as HoxD10 did not influence VEGF production when cells were unable to re-organize into polarized structures in 2D cultures, although it remains possible that accompanying morphological changes in 3D unmask binding sites within the VEGF promoter. Still, these findings indicate that activation of the angiogenic switch is not simply due to genetic changes within the tumor cells, but rather linked to how cells sense their architecture and interact with their microenvironment.
Importantly, we also show that in tumor cells, which characteristically exhibit increased growth and metabolic demand compared to non-malignant cells, restoring normal tissue architecture is essential for attenuating their angiogenic potential. Indeed, a ten-fold increase in proliferation of non-malignant cells was not sufficient to explain resistance to chemotherapeutic agents(17) and is also not sufficient to activate VEGF expression when growing cells maintain a polarized architecture.
Many studies have established that increased proliferation, metabolic demand and hypoxia in tumor cells stabilizes HIF1α and drives VEGF expression (3, 6, 34-36). Blocking HIF1α expression/activity reduces VEGF expression, and HIf1α null mice display defective angiogenesis and fail to support tumor growth (8),(35). However, our results show that neither differences in HIF1α expression or activity underly reduced VEGF in reverted, polarized tumor cells and implicate other pathways and/or transcriptional mediators in regulating the angiogenic switch in this system.
In a mouse model of pancreatic cancer, hypoxia-independent activation of the angiogenic switch occurs via increased MMP-9 activity, which liberates bioactive VEGF trapped within the tumor matrix (37). Despite the fact that MMP-9 is highly expressed in disorganized T4-2 cells and reduced in reverted T4-2 (A. Beliveau, and MJ Bissell submitted), elevated secretion of VEGF is reflected by increased VEGF mRNA. Whether reduced VEGF transcription by re-polarized breast tumor cells increases inhibitory factors (38) and/or attenuates other positive inducers of VEGF requires further investigation.
Previous studies also demonstrated that PI3K signaling induces VEGF (13, 39), and our current study suggests that the Rac1 branch of this pathway directly drives VEGF expression. We reported PI3K inhibition phenotypically reverts tumorigenic breast epithelial cells and is accompanied by down regulation of both the Akt and Rac1 effector pathways (22). However, while attenuation of Akt reduces proliferation, it did not restore a polarized phenotype. Instead, suppression of Rac1 activity was necessary for re-establishing an organized polarized phenotype, and selective re-activation of Rac1 disrupts polarity without re-initiating growth when Akt remained blocked (22). Moreover, whereas activation of Rac1 in non-malignant breast epithelial cells normally induces apoptosis, in breast tumor epithelial cells lacking scribble, Rac activation promotes tumorigenesis and loss of cell polarity (40). In the present study, we demonstrate that re-activation of Rac1 and disruption of polarity restores VEGF expression and induces EC migration despite the fact that growth remains suppressed via low pAkt levels (latter data not shown).
Active Rac1 directly binds and phoshphorylates STAT3 (41), which in turn forms Sp1/STAT3 complexes that bind the VEGF promoter (42) and the SP1 site within the -86- -66 region of the VEGF promoter is linked to HIF1α-independent transcription of VEGF (10, 39, 43). Phosphorylated STAT3 can also bind to HIF1α and prevent its degradation (44). Although we observed an increase in phospho-STAT3 in Rac1-overexpressing cells, reversion of malignant cells by HoxD10 did not reduce phospho-STAT3 (data not shown) and HIF1α protein levels were similar in both tumorigenic and HoxD10-reverted tumor cells. Thus binding of STAT3 to either SP1 or HIF and VEGF expression may be context dependent as previously suggested by investigations of pathways impinging upon VEGF expression (39), and our current study also emphasizes the critical role tissue polarity in mediating VEGF expression in breast tumor cells.
It is noteworthy that inhibiting EGF receptors represses VEGF in both a HIF-dependent and HIF-independent manner, with the latter attributed to reduced phosphorylation and binding of Sp1 to the VEGF promoter (43). A number of EGFR inhibitors, including those used in this study, not only reduce VEGF expression and inhibit breast tumor angiogenesis in vivo, but also phenotypically revert tumor cells to a polarized morphology in culture (19, 43, 45, 46). Finally, mice lacking LKB1, a tumor suppressor which directs cell polarity and when mutated, produces epithelial cancers, show markedly increased VEGF (47, 48).
The present study addresses breast tumor epithelial organization and VEGF production, but it is important to note that other components of the tumor microenvironment, including fibroblasts and macrophages, are also rich sources of angiogenic factors. In DCIS where newly formed capillaries appear immediately adjacent to the basal surface of the breast epithelium (5), the BM is largely intact but a majority of epithelial cells have lost their characteristic polarized morphology and display attenuated expression of the HoxD10 tumor suppressor (3, 23). As macrophage recruitment typically occurs in later stages of malignant progression accompanied by BM degradation (49), it is likely that the loss of epithelial cell polarity marks an early and critical step in activation of the angiogenic switch.
Supported by CBCRP Award 101B-0157 (NJB). MJB is supported by U.S. Department of Energy, OBER Office of Biological and Environmental Research (DE-AC02-05CH1123), a Distinguished Fellow Award and Low Dose Radiation Program and the Office of Health and Environmental Research, Health Effects Division, (03-76SF00098); National Cancer Institute awards R01CA064786, R01CA057621, U54CA126552 and U54CA112970; U.S. Department of Defense (W81XWH0810736) and (W81XWH0510338).