Oncogenic events regulate tissue factor expression in colorectal cancer cells: implications for tumor progression and angiogenesis

Cancer is believed to arise and progress toward increasing malignancy as a result of cumulative genetic “hits” sustained by the tumor cell genome. Paradigmatic in this regard is the development of colorectal carcinoma (CRC), where sequential transition through clinical stages of the disease is paralleled by a series of well-characterized alterations in proto-oncogenes and tumor suppressor genes.¹  In this tumor type, activation of mutant K-ras and subsequent inactivation/loss of p53 are key changes, which drive many interrelated aspects of the malignant phenotype including aberrant mitogenesis and survival.²  Moreover, both of these genetic alterations are thought to contribute to proangiogenic properties of affected cancer cells,^3,4  and thereby enable them to exploit the host vascular system to advance malignant growth and metastasize in vivo.⁵ 

The involvement of the vascular system in malignancy encompasses not only angiogenesis but also systemic hypercoagulability. Blood clotting abnormalities are detected in up to 90% of patients with metastatic disease, and thrombosis represents the second most frequent cause of cancer-related mortality.⁶  Cancer coagulopathy is often linked to up-regulation of tissue factor (TF), the primary cellular initiator of the blood coagulation cascade.^7,8  Interaction of coagulation factor VIIa with TF on the cell surface leads to activation of factor X and generation of thrombin, with subsequent involvement of platelets and formation of a fibrin clot.⁹  Remarkably, as a member of the class II cytokine receptor family, TF is also capable of transducing intracellular signals and regulating gene expression.^10,11  Interestingly, elements of the coagulation/fibrinolytic system in general,¹²  and TF in particular, have been implicated in regulation of angiogenesis,^13,14  as well as tumor growth¹⁵  and metastasis¹⁶  in various experimental settings. This is consistent with the observed up-regulation of TF in human malignancies and its elevation with advancing disease.^17,18  For instance, in human CRC, TF positivity correlates with clinical stage, histologic grade, poor prognosis, and vascularity.^19-21  Collectively, these observations suggest that TF is not only an important element of cancer-related coagulopathy but is also a correlate and indeed a likely determinant of malignant behavior of tumor cells.

In this context 2 main questions remain unanswered. First, what causes the up-regulation of TF in human cancer cells, with the subsequent increase in their malignancy? Second, how are the consequences of TF expression related to the phenotypic changes induced by underlying “cancer-causing” genetic alterations? Here we report that in CRC, TF expression by cancer cells is directly linked to their genetic status; for example, activation of the K-ras oncogene and loss of p53 are involved in TF regulation. These transforming alterations influence the level and activity of TF not only on the cell membrane but also on the surface of microvesicles shed by cancer cells into the circulation. We therefore suggest that both local and systemic hypercoagulability in cancer may have a hitherto unappreciated genetic cause or component. Finally, we demonstrate that TF is required for full expression of the K-ras-dependent tumorigenic and angiogenic phenotype of CRC cells in vivo, but not for cellular transformation in vitro.

The human CRC cell lines HCT116 and DLD-1, and their K-ras-deleted sublines (HKh-2, DKs-8, DKO-1, DKO-3) have been previously characterized.²²  The p53^-/- subline of HCT116 (379.2)²³  was kindly provided by Dr Bert Vogelstein (Johns Hopkins University, Baltimore, MD). The 528ras+mTF cell line was produced by transfecting transformed murine 528ras fibroblasts²⁴  with a full-length mouse TF expression vector to produce cells expressing high levels of mouse tissue factor (mTF) (J.L.Y. and J.W.R., unpublished data, 2004). The A549 human non-small cell lung carcinoma cell line was purchased from American Type Culture Collection (Manassas, VA). All cell lines were maintained in Dulbecco modified Eagle medium (DMEM; HyClone, Logan, Utah) with 10% fetal bovine serum (FBS; Gibco BRL/Invitrogen, Carlsbad, CA). For determination of cell growth in monolayer culture, cells were seeded into 24-well plates in complete medium. At each time point, cells (3 wells/cell line) were detached by trypsinization and cell number determined using a hemocytometer. The dominant-negative H-Ras-N17 mutant expression construct^25,26  was a generous gift from Dr Abhijit Guha (University of Toronto, Toronto, ON, Canada). PD98059, LY294002, and AG1296 were purchased from Biomol Research Laboratories (Plymouth Meeting, PA). CI-1033 was made available by Pfizer, courtesy of Dr Christian Marsolais. All inhibitor stock solutions were prepared in dimethyl sulfoxide (DMSO).

For detection of surface TF antigen, cells were detached with 2 mM EDTA (ethylenediaminetetraacetic acid) to obtain a single-cell suspension. Cells (1.5 × 10⁶) were washed in phosphate-buffered saline (PBS) with 1% FBS and 0.1% sodium azide and stained for 30 minutes at 4°C with a monoclonal antibody against human TF (American Diagnostica, Greenwich, CT). After washing, samples were incubated with Alexa Fluor 488 goat anti-mouse secondary antibody (Molecular Probes, Eugene, OR) for 30 minutes at 4°C, washed, and fixed in 1% paraformaldehyde before being acquired on a FACScalibur (BD Biosciences, Mountain View, CA). For cell sorting experiments, 2 × 10⁷ HKh-2 cells were stained for TF as described, and a sterile sort performed on a FACStar Plus (BD Biosciences). Gates were set to collect the most TF⁺ and TF^- viable cells; 3.5 × 10⁵ cells from each side of the expression spectrum. The cells were collected and cultured in complete media containing penicillin-streptomycin (Gibco BRL/Invitrogen).

TriZol (Gibco BRL/Invitrogen) was used to isolate total RNA from cells or homogenized tumor tissue. Northern blotting was performed as described previously,²⁴  using vascular endothelial growth factor (VEGF), TF, thrombospondin 1 (TSP-1), or TSP-2 cDNA fragments as probes. For Western blotting, cells were lysed with nonidet-P40 (NP-40) lysis buffer (1% NP-40, 10% glycerol, 20 mM Tris (tris(hydroxymethyl)aminomethane)-HCl, pH 7.5, 137 mM NaCl, 100 mM NaF, 1 mM sodium vanadate, 1 mM phenylmethylsulfonyl fluoride [PMSF]), supplemented with complete protease inhibitor cocktail (Roche, Indianapolis, IN). Protein was quantified by Bradford assay (Bio-Rad, Hercules, CA), resolved by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), and transferred to Immobilon-P membrane (Millipore, Billerica, MA). Membranes were probed with rabbit anti-human TF IgG (American Diagnostica) followed by goat antirabbit secondary antibody (Jackson ImmunoResearch Labs, West Grove, PA). To confirm equal loading, membranes were probed with anti-ERK1/2 antibody (Upstate Biotechnology, Lake Placid, NY). The IMUBIND TF enzyme-linked immunosorbent assay (ELISA) kit (American Diagnostica) was used to quantify TF protein levels in plasma or conditioned medium.

TF activity was measured as described previously.²⁷  Standard curves were prepared using different dilutions of rabbit brain thromboplastin (Thromboplastin C Plus; Dade Behring, Marburg, Germany), and 1 U TF activity was defined as the activity of the 1:10⁴ dilution of thromboplastin standard. Cell number was used to normalize the TF activity. Detection of TF activity in conditioned media was performed similarly, by adding the Tris-buffered saline (TBS) reaction mixture to 100 μL media (or supernatant, or resuspended microvesicles after centrifugation).

All in vivo experiments were performed in 6- to 8-week-old severe combined immunodeficiency (SCID) mice (5-10/group; Charles River, Saint-Coustant, QC, Canada). Briefly, 1 to 5 × 10⁶ CRC cells were injected subcutaneously in 0.2 mL PBS. Blood was collected from mice by cardiac puncture, into one-tenth volume 3.8% sodium citrate. Platelet-free plasma was prepared by centrifugation at 2000g for 15 minutes, 2000g for 5 minutes, and 16 000g for 5 minutes and stored at -80°C until use. Tumor growth was monitored by measurements with Vernier calipers, and tumor volume (mm³) estimated using the standard formula (length × width²  × 0.52). Animal studies were approved by the Animal Research Ethics Board at McMaster University.

Cells were washed and fresh complete medium added for 24 hours. Conditioned medium was removed and centrifuged to eliminate debris (500g for 15 minutes, then 800g for 20 minutes), and then ultracentrifuged for 90 minutes at 4°C (100 000g) to pellet microvesicles, which were resuspended in PBS.

Tumors were excised and fixed in 10% formalin. Deparaffinized 4-μm sections were stained with hematoxylin and eosin for analysis of gross morphology, or immunostained for TF after microwave antigen retrieval. Sheep anti-human TF IgG was used as primary antibody (Affinity Biologicals, Ancaster, ON, Canada), followed by Alexa Fluor 488 donkey anti-sheep IgG (Molecular Probes).

Activating mutations in codon 13 of the K-ras gene were detected by polymerase chain reaction-restriction fragment length polymorphism (PCR-RFLP) analysis.²⁸  DNA was extracted using a DNeasy tissue kit (Qiagen, Valencia, CA). PCR was performed in a reaction volume of 50 μL containing 250 ng DNA, 1 × reaction buffer, 0.2 mM dNTP, 1.5 mM MgCl₂, 0.2 μM of each primer, and 2.5 U Taq polymerase (Invitrogen, Carlsbad, CA). As described in Schimanski et al,²⁸  the primers were RAS A (sense) 5′-ACTGAATATAAACTTGTGGTCCATGGAGCT-3′ and RAS B (antisense) 5′-TTATCTGTATCAAAGAATGGTCCTGCACCA-3′. Amplification was performed in a thermocycler (Progene; Techne, Minneapolis, MN), and consisted of 30 cycles of 94°C for 1 minute, 55°C for 1 minute, and 72°C for 2 minutes. Two rounds of PCR were performed to obtain a very clean 166-base pair (bp) product. Then, 10 μL of the PCR reaction was digested with XcmI (10 U; New England Biolabs, Beverly, MA) for 20 hours at 37°C, in a total volume of 40 μL. When codon 13 is wild type, the PCR product contains a restriction site for XcmI, and digestion yields bands of 138 and 28 bp. If there is a mutation in either of the first 2 bases of codon 13, the mutant PCR fragment will not be cut by XcmI and will remain at its original size of 166 bp. Bands were visualized by electrophoresis on a 3.5% agarose gel. DNA extracted from the whole blood of a healthy donor was used as the wild-type K-ras codon 13 GGC-Gly control. DNAfrom HCT116 cells (heterozygously GAC mutated in codon 13)²⁹  was used as the control for known K-ras mutation. Amplified products were confirmed by sequencing.

Cells grown in 6-well plates were transiently transfected with the dominantnegative H-Ras-N17 plasmid (1.5 μg) and 0.5 μg pGL2-hTF (full length: -2102 to +121 TF promoter sequence cloned into the pGL2 luciferase reporter plasmid), using Lipofectamine 2000 (Invitrogen). Empty pcDNA3.1 vector was used as a control for H-Ras-N17 transfection. LacZ expression vector (pcDNA3.1-LacZ, 0.25 μg) was cotransfected to monitor transfection efficiency. Luciferase activity was determined 24 hours after transfection using the Luciferase Assay System (Promega, Madison, WI) and TD-20/20 luminometer (Turner Designs, Sunnyvale, CA), according to supplied protocols. Alternatively cells transfected with pGL2-hTF were treated with PD98059 (50 μM) or LY294002 (20 μM) for 2 to 48 hours prior to read out. Values were normalized to the β-galactosidase activity measured in the lysates.

The small interfering RNA (siRNA) expression plasmid (pSN-TF) was constructed by annealing and cloning the following siRNA oligonucleotides into the pSuppressorNeo (pSN) vector (Imgenex, San Diego, CA): 5′-TCGAGGCGCTTCAGGCACTACAAATTCAAGAGATTTGTAGTGCCTGAAGCGC TTTTT-3′ (forward) and 5′-CTAGAAAAAGCGCTTCAGGCACTACAAATCTCTTGAATTT GTAGTGCCTGAAGCGC-3′ (reverse). These oligonucleotides use a previously validated siRNA sequence,³⁰  and encode the sense siRNA sequence followed by a short spacer, the reverse complement of the sense strand, and a transcriptional termination signal. Thus, transcription of pSN-TF generates hairpin RNAs that are processed into siRNAs in the cell. As a control, oligonucleotides incorporating an siRNA sequence known to be inactive in TF silencing³⁰  were used to construct a control expression vector (pSN-77): 5′-TCGAGTGGAGACCCCTGCCTGG CCTTCAAGAGAGGCCAGGCAGGGGTCTCCATTTTT-3′ (forward) and 5′-CTAGAAAAATGGAGACCCCTGCCTGGCCTCTCTTGAAGGCCAGGCAGGGGTCTCCA-3′. HCT116 cells were transfected with pSN-TF and pSN-77 using Lipofectamine 2000 (Invitrogen), and clones screened for integration of the siRNA sequence by PCR (of gDNA) using the following primers: 5′-AATACGTGACGTAGAAAGTA-3′ (forward) and 5′-CACATGTGATA TCCGCAG-3′ (reverse). TF gene down-regulation in positive clones was then assessed by Northern blotting and TF activity assay.

HCT116, SI-2, and SI-3 cells were rendered mitotically incompetent by treatment for 2 hours with 10 μg/mL mitomycin C (Sigma, St Louis, MO), washed, and resuspended in Matrigel (BD Biosciences) at a concentration of 4 × 10⁶ cells/mL. A total of 0.2 mL of the mixture was injected subcutaneously in the flanks of SCID mice. As a negative control, cell-free Matrigel alone was injected. Plugs were excised after 11 days, placed in Carnoy fixative (60% methanol, 30% chloroform, 10% glacial acetic acid) for 4 hours, then transferred to ethanol. Paraffin-embedded specimens were cut and stained with hematoxylin and eosin or immunostained for von Willebrand Factor (VWF). The primary antibody was rabbit anti-human VWF (1:300; Dako, Carpinteria, CA), followed by goat antirabbit secondary antibody (1:1000, Zymed Laboratories, San Francisco, CA), and color was developed with 3,3′-diaminobenzidine tetrahydrochloride dihydrate (DAB; Dako). Angiogenic response was quantified by measurement of the area occupied by VWF⁺ endothelial networks per × 20 field. A total of 5 fields were analyzed from 5 sections of 5 different Matrigel plugs per group. Image capture and analysis was performed using Northern Eclipse software (Eurpix Imaging, Mississauga, ON, Canada).

Data are presented as means of several independent measurements ± SD. Statistical analyses (t test) were performed with GraphPad InStat software (GraphPad Software, San Diego, CA). Linear regression analysis was performed with Origin 6.1 (OriginLab, Northampton, MA).

Clinical progression of CRC is paralleled by sequential alterations in K-ras and p53 genes, as well as by increasing TF immunoreactivity.^1,21  Two independent human CRC cell lines, DLD-1 and HCT116, both harbor 1 mutant (and 1 wild-type) K-ras allele and express elevated levels of TF.²²  To determine whether mutation of the K-ras oncogene is directly linked to TF expression in CRC, we examined the consequences of selective disruption of the mutant K-ras allele by homologous recombination in previously characterized cellular variants.²²  This procedure produced the DKs-8 and DKO-3 sublines (from DLD-1), as well as the HKh-2 subline (from HCT116), all of which consequently express only the wild-type K-ras allele and a diminished level of malignancy.²²  DKO-1 cells were derived from DLD-1 after accidental disruption of the wild-type rather than the mutant K-ras allele (Figure 1).

Cell surface TF immunoreactivity was greatly diminished in cells without mutant K-ras (DKs-8, DKO-3, and HKh-2), relative to their respective isogenic but mutant K-ras-expressing DLD-1 and HCT116 counterparts (Figure 2A-B). TF mRNA, total TF protein, and cell surface TF activity were all down-regulated on mutant K-ras disruption (Figure 2C-E). These results suggest that TF up-regulation in CRC cells is directly linked to expression of the activated mutant K-ras oncogene.

During CRC progression, activation of K-ras is frequently followed by loss-of-function mutations in the p53 tumor suppressor gene. Therefore, we next investigated whether p53 status has an impact on TF expression in cells representative of more advanced CRC. HCT116 cells harbor 2 wild-type p53 alleles, both of which have been disrupted in a previously generated variant (designated 379.2).²³  Thus, 379.2 cells recapitulate natural loss-of-function alterations (loss of heterogeneity [LOH]) affecting this tumor suppressor in CRC. As with K-ras mutation, loss of p53 in 379.2 cells was associated with a marked further increase in cell surface TF activity, TF protein, and mRNA relative to their parental p53^+/+ HCT116 counterparts (Figure 2B-E). It is noteworthy that the TF profile of p53^-/- HCT116 (379.2) cells (which harbor mutant K-ras) reflects the cumulative impact of both transforming alterations (K-ras and p53); that is, TF expression is progressively up-regulated with these 2 sequential oncogenic events (HKh-2 < HCT116 < 379.2).

To explore whether up-regulation of TF in tumor cells translates into systemic release of this procoagulant, TF antigen concentration in the plasma of tumor-bearing mice was measured using a human-specific TF ELISA. Indeed, human (tumor-derived) TF was detected in the plasma of mice with HCT116 tumors, at a concentration corresponding with increasing tumor volume (Figure 3A). The plasma TF level in mice with extremely large tumors was somewhat reduced, possibly reflecting tumor necrosis and inefficient tumor perfusion. More important, the level of circulating human TF was significantly higher in mice bearing highly TF⁺ 379.2 (p53^-/-) tumors, relative to mice with similarly sized HCT116 (p53^+/+) tumors; poorly tumorigenic HKh-2 cells were not included in this analysis (Figure 3B). Thus, the p53-dependent increase in TF expression in 379.2 cells translated into a higher level of TF in the circulation. Plasma of mice harboring large murine fibrosarcomas engineered to express high levels of mouse TF (528ras+mTF) did not contain any human TF immunoreactivity detectable by this ELISA and neither did plasma of tumor-free animals (Figure 3B). Hence, essentially all circulating TF antigen detected in mice harboring HCT116 and 379.2 tumors is attributable to tumor cell-derived TF.

Although this particular assay does not offer insight into TF procoagulant activity related to either tumor or host cells, such activity was detected in conditioned medium from HCT116 cells, as well as from its mutant K-ras and p53-disrupted sublines (Figure 3C). As with cell-bound TF, the secretion of cell-free TF activity corresponded with the genetic status of the respective CRC cells (ie, K-ras^wt/p53^wt < K-ras^mut/p53^wt < K-ras^mut/p53^mut). Surprisingly, all CRC cells expressed only minimal amounts of the recently described, soluble splice variant of TF (asHTF; data not shown),³¹  and instead most of the released cell-free TF activity could be isolated (pelleted) from the conditioned medium by ultracentrifugation (Figure 3C). Because viable cells are known to actively shed plasma membrane-derived vesicles (also called microvesicles or microparticles) that may contain procoagulant activity,^32-35  this result is consistent with the ability of CRC cells to shed TF-containing microvesicles both into the media and the circulation. Indeed, such structures were observed under electron microscope in supernatants from CRC cell lines (data not shown). Again, the amount of cell-free vesicle-associated TF activity released by each cell line was proportional to their total TF production, which was dictated by their oncogenic status (Figure 3C). These data suggest that K-ras and p53 influence not only cell surface TF expression, but also the global amount of TF activity released from cancer cells into their surroundings and the circulation.

Low TF-expressing, mutant K-ras^- HKh-2 cells are poorly tumorigenic in vivo (Figure 4A). Interestingly, HKh-2 tumors that eventually arise and assume more aggressive properties were found to display increased TF mRNA expression (50 days after cell injection), relative to their counterparts maintained in culture (Figure 4B top). Because analysis of early stage (day 7) HKh-2 tumors by TF immunostaining revealed the presence of extremely rare, but highly TF⁺ cells among the TF^- majority (Figure 4C-E), we examined whether this TF positivity might be a signature of a cell subset with an overt growth advantage, which would subsequently contribute to the gradual increase in overall tumor aggressiveness in vivo. First, HKh-2 cells were stained with a TF antibody, and cell sorting was performed to collect the fractions with the 5% highest and lowest TF expression; these were designated TF-positive (TF-POS) and TF-negative (TF-NEG), respectively. Both cell subsets were expanded in vitro and analyzed for expression of TF mRNA and tumor-forming ability in SCID mice. In this setting, the TF-NEG cell subset was poorly tumorigenic in SCID mice, whereas TF-POS cells grew more aggressively (Figure 4F), a finding that suggests close cosegregation between TF expression and malignancy within a single tumor cell line.

One obvious way by which genetically unstable HKh-2 cells (K-ras^del/+) could reacquire increased aggressiveness is by re-expression of a dominant genetic lesion, for example, mutation in the remaining wild-type K-ras allele.³⁶  Indeed, using a PCR-RFLP assay to detect mutations in K-ras, we found that HKh-2 cells expressed mostly wild-type K-ras allele, as expected. In contrast, HKh-2 cells isolated from tumors that eventually emerged in SCID mice contained a new K-ras mutation (Figure 4B bottom). Moreover, analysis of TF-POS and TF-NEG cells collected from 2 independent sorting experiments (including the one depicted in Figure 4F) revealed that even in culture, the more aggressive TF-POS subpopulation of HKh-2 cells (expressing high TF levels) is enriched in “revertants” expressing mutant K-ras (Figure 4G). In contrast, low TF-expressing and poorly tumorigenic TF-NEG cells continue to express the wild-type K-ras allele. These results suggest that either TF is a consistent molecular “marker” of greater malignancy (eg, K-ras mutation), or a more direct effector of the malignant phenotype, at least in the context of mutant K-ras.

To reinforce the role of mutant ras in up-regulation of TF expression, we examined the effect of expressing dominant inhibitory Ras-N17 mutant^25,26  on TF promoter activity in Ras-driven cells. Even transient transfection of the Ras-N17 construct into HCT116 and DLD-1 CRC cells and A549 lung carcinoma cells diminished TF promoter activity relative to cells transfected with a control plasmid (Figure 4H). Expression of the dominant-negative Ras-N17 mutant also decreased TF promoter activity in HKh-TUM2 cells (Figure 4H), which were isolated from HKh-2 tumors (50 days after cell injection) and found to be revertants harboring de-novo K-ras mutation (Figure 4B). Moreover, pharmacologic inhibition of the MEK1/mitogen-activated protein kinase (MAPK) pathway (by PD98059) downstream of Ras resulted in similar inhibition of TF promoter activity in both HCT116 and 379.2 cells. Collectively, these results substantiate a causal role of the activated Ras pathway in up-regulation of TF in CRC cells (both with and without defects in p53) and point to at least partial contribution of a transcriptional mechanism in this process (Figure 4H).

To further investigate some of the potential downstream events involved in TF up-regulation by mutant K-ras, we treated HCT116 cells with a panel of pharmacologic inhibitors. Again, selective inhibition of MEK1 (and MAPK pathway) with PD 98059 resulted in decreased TF protein levels (Figure 5A). Interestingly, blockade of phosphatidylinositol 3′-kinase (PI3K) activity also led to greatly diminished TF expression, to levels comparable to those of wild type K-ras-expressing Hkh-2 cells (compare Figures 1 and 5). In contrast, treatment of HCT116 cells with a pan-ErbB inhibitor (CI-1033) only slightly decreased TF expression, whereas a platelet-derived growth factor receptor (PDGFR) signaling inhibitor (AG1296) or vehicle (DMSO) had no effect. Similar results were observed following treatment of 379.2 cells with the same inhibitors (Figure 5B). These data suggest that the PI3K and MAPK pathways both contribute to up-regulation of TF in CRC cells expressing mutant K-ras even when this effect is amplified by the simultaneous loss of p53.

To determine the extent (if any) to which TF expression may be required for K-ras-driven tumorigenicity, we used an RNA interference (RNAi) approach to knock down TF expression in HCT116 cells. Previously validated small interfering RNA (siRNA) sequences³⁰  were used to construct a TF siRNA expression plasmid suitable for generation of permanent cell lines, to achieve the long-term effects necessary for in vivo studies. Several TF siRNA-transfected HCT116 clones integrated the siRNA sequence (Figure 6A inset) and exhibited diminished TF activity and expression (Figure 6A and data not shown). Three of these clones (SI-2, SI-3, and SI-9) were selected for in vivo studies on the basis of their stable TF-suppressed phenotype.

Whereas the in vitro growth properties of HCT116 cells were unaffected by down-regulation of TF even in hypoxic or spheroid cultures (Figure 6B and data not shown), the growth of tumors established from TF-suppressed clones was markedly retarded relative to parental HCT116 cells (Figure 6C). This effect was highly specific because growth of HCT116 cells transfected with an expression plasmid encoding an siRNA sequence known to be ineffective in TF gene silencing³⁰  were indistinguishable from parental HCT116 (data not shown). Because TF status only affected the behavior of HCT116 cells in vivo, that is, in a manner that was likely host-dependent, this could be indicative of the proangiogenic role of TF in the context of mutant K-ras-driven tumorigenicity of these CRC cells. Hence, we assessed the global capacity of a fixed number of HCT116 cells or their TF-deficient variants to recruit host endothelial cells into Matrigel plugs in vivo. Cells were treated briefly with mitomycin C to prevent changes in cell number, resuspended in Matrigel, and injected subcutaneously into mice, and plugs were recovered 11 days later. Quantification of VWF⁺ vascular structures by morphometry indicated a consistent, and statistically significant, 3-fold difference in the proangiogenic potential of HCT116 cells relative to their TF down-regulated SI-2 and SI-3 counterparts (Figure 6D-F). In SI-2 and SI-3 plugs, VWF⁺ vasculature was found merely at the outer margins and rarely contained red blood cells (Figure 6E). Plugs consisting of Matrigel alone showed only minimal background invasion (Figure 6F). Because HCT116 cells ultimately depend on mutant K-ras for their tumorigenic and angiogenic properties,²²  these results raise the possibility that TF is required for the expression of the angiogenic phenotype driven by this oncogene.

In keeping with this possibility, transcripts encoding 2 angiogenesis inhibitors (and oncogene targets), TSP-1 and TSP-2, were expressed at much higher levels in TF down-regulated cell lines than in their parental HCT116 counterparts (Figure 6G-H). Interestingly, this difference was only observed in vivo (and not between the respective cell lines in culture). This raises a possibility (but does not prove) that tumor-derived TF may interact with host-dependent ligands (eg, factor VIIa) to execute the “proangiogenic switch” in tumors harboring mutant K-ras. Expression of VEGF in HCT116 cells was minimally affected by their TF status (Figure 6I). Collectively, these results suggest that at least a portion of K-ras-dependent angiogenic activity expressed by CRC cells in vivo may be mediated through TF up-regulation and its multiple downstream angiogenic targets exemplified by TSP-1 and TSP-2 (Figure 6J).

The intriguing properties of TF, which acts as a procoagulant and proangiogenic cellular receptor in various pathologic contexts, have recently been a subject of intense interest.¹³  Indeed, TF is consistently up-regulated in human cancer cells,¹³  but the causes and mechanisms of this change have remained unknown. In this regard our study led to several new findings.

First, our results suggest that TF is a target of at least 2 of the most common genetic alterations in human malignancy: inactivation of p53 and mutation of K-ras. Moreover, action of these respective transforming events appears to be cumulative in nature. Thus, in a unique and well-characterized series of isogenic human CRC cell lines engineered genetically to recapitulate 2 distinct steps in genetic tumor progression,^1,22,23  activation of mutant K-ras and subsequent inactivation of p53 up-regulated TF expression and activity in a cooperative manner. There is a remarkable similarity between this stepwise increase in TF levels and the reported pattern of gradual increase in TF immunoreactivity in clinical specimens of human CRC,^19-21  as well as the increasing frequency of systemic coagulopathy associated with advanced malignancy.^6,37  Both K-ras and p53 likely act at the level of TF gene expression (rather than de-encryption) because TF mRNA, total protein, immunoreactivity, and procoagulant activity (both cell-associated and released in membrane vesicles) were affected in a parallel fashion by the genetic status of CRC cells. Taken together, our observations suggest the possibility that TF up-regulation and hypercoagulability in patients with cancer may have a hitherto unappreciated genetic cause.

The second major point presented in our study is that TF up-regulation is not merely a correlate (“marker”) or consequence of oncogenic alterations, but also an important effector of K-ras-dependent tumorigenesis and angiogenesis, at least in human CRC. This finding causally links 2 hitherto separately studied solitudes, namely, the coagulation system and genetic tumor progression. In particular, tumor growth, angiogenesis, and metastasis have long been attributed, at least in part, to both canonical and noncanonical TF activities,^15,38,39  including generation of thrombin, deposition of fibrin, and activation of proangiogenic platelets.¹²  In some of these instances, elevated TF expression in cancer cells elicited such cellular responses as constitutive up-regulation of VEGF, down-regulation of TSP-2,¹⁵  and increased metastatic³⁸  and angiogenic capacity.¹⁵ 

Such TF-induced changes resemble those described previously in the context of mutant oncogenes (eg, K-ras⁴ ). Our present results indicate that this is not merely a coincidence because we find TF expression to be required for full manifestation of the K-ras-dependent angiogenic and aggressive tumor cell phenotype in vivo (but not in vitro). Thus, enforced TF down-regulation in K-ras-expressing CRC cells by RNAi profoundly reduced global angiogenic capacity, led to increased levels of angiogenesis inhibitors (TSP-1 and TSP-2), and diminished growth of tumors in vivo without affecting the transformed phenotype and mitogenic properties of the cancer cells themselves (in vitro). Although TF does not seem to possess any overt cell autonomous transforming or growth regulatory properties in this setting, it clearly possesses the ability to modify tumor cell behavior in vivo, likely through interaction with host-derived entities. The nature of the latter remains unknown at present but the likely candidates include factor VIIa, factor Xa, and other putative TF ligands (eg, plasminogen, TF pathway inhibitor [TFPI]).^40,41  Regardless of the molecular mechanism, the tumor-inhibitory effects of TF down-regulation highlight the possible validity of targeting TF in the context of K-ras-driven human tumors, such as pancreatic, colorectal, and lung cancer.²  In this context, TF-dependent regulation of the angiogenic phenotype in cancer cells must be distinguished conceptually and mechanistically from the role TF appears to play in the context of angiogenic host endothelial cells. The latter was recently linked with signaling properties of the TF cytoplasmic domain.¹⁴ 

At variance with earlier reports,¹⁵  we have not observed any significant link between the status of TF in HCT116 cells and expression of VEGF in vivo or in vitro. It is possible that VEGF in the context of HCT116 cells may be under TF-independent regulatory control (eg, by mutant K-ras itself⁴ ), a notion consistent with the pleiotropic and tumor/oncogene-specific nature of proangiogenic gene expression profiles characterized to date.²⁴  This pleiotropism appears to extend to angiogenic mediators downstream of TF because our findings suggest that TF is linked to the regulation of at least 2 different endogenous angiogenesis inhibitors (TSP-1 and TSP-2). A more complete survey of TF-dependent and TF-independent effectors of oncogene-driven tumor angiogenesis is warranted, using approaches of functional genomics²⁴  and proteomics.

Although many influences (including inflammation, hypoxia, and other) could contribute to TF up-regulation by cancer cells, our findings suggest that oncogenic events play an important and hitherto unappreciated role in this regard. It follows that oncogene-targeted therapies might be useful in reversing this increase. Unfortunately, no potent, specific, and proven pharmacologic modulators of K-ras and p53 status in the context of CRC are currently available. In contrast, recent clinical successes have been achieved with inhibitors of ErbB receptor kinases.⁴²  We have observed that in A431 epithelial carcinoma cells, the malignant properties of which are mainly driven by overexpression/activation of EGFR, both epidermal growth factor receptor (EGFR)/ErbB1 and pan-ErbB inhibitors markedly diminished TF expression and reduced the levels of cancer cell-derived (human) TF in the circulation of tumor-bearing mice (J.L.Y. and J.W.R., unpublished data, 2004). It is worth considering whether TF levels in biopsy material or plasma of patients receiving certain oncogene-directed agents could provide a measure of the agent's biologic, and possibly also therapeutic, efficacy.

Because of technical considerations (xenograft model), our experiments did not provide direct evidence of the contribution of tumor cell-associated TF, and its oncogenic up-regulation, to the overall hypercoagulability associated with cancer (ie, with inclusion of host-dependent and inflammatory events). However, an observation that all-trans-retinoic acid (ATRA) attenuates coagulopathy in patients with acute promyelocytic leukemia (APL) serves as a thought-provoking precedent, as ATRA acts essentially by blocking the oncogenic action of the promyelocytic/retinoic acid receptor α (PML/RARα) gene product on the malignant phenotype of APL cells.⁴³ 

Collectively, TF expressed by cancer cells appears to act as both a regulatory target and an important mediator of oncogenedriven tumor growth and neovascularization. Future studies will determine the extent to which targeting and monitoring TF expression may be useful, from a diagnostic, prognostic, and therapeutic standpoint. A better understanding of the oncogenic defects driving malignant progression will likely provide important clues in this regard.

We thank Sarah Gilpin for technical assistance and our colleagues Dr Sarka Lhotak, Alan Stafford, and Lindsay Watson for technical suggestions.