Several types of collagen contain cryptic antiangiogenic noncollagenous domains that are released upon proteolysis of extracellular matrix (ECM). Among those is Arresten, a collagen-derived antiangiogenic factor (CDAF) that is processed from α1 collagen IV. However, the conditions under which Arresten is released from collagen IV in vivo or whether the protein functions in tumor suppressor pathways remain unknown. Here, we show that p53 induces the expression of α1 collagen IV and release of Arresten-containing fragments from the ECM. Comparison of the transcriptional activation of COL4A1 with other CDAF-containing genes revealed that COL4A1 is a major antiangiogenic gene induced by p53 in human adenocarinoma cells. p53 directly activated transcription of the COL4A1 gene by binding to an enhancer region 26 kbp downstream of its 3′ end. p53 also stabilized the expression of full-length α1 collagen IV by upregulation of α(II) prolyl-hydroxylase and increased the release of Arresten in the ECM through a matrix metalloproteinase (MMP)-dependent mechanism. The resulting upregulation of α1 collagen IV and production of Arresten by the tumor cells significantly inhibited angiogenesis and limited tumor growth in vivo. Furthermore, we show that immunostaining of Arresten correlated with p53 status in human prostate cancer specimens. Our findings, therefore, link the production of Arresten to the p53 tumor suppressor pathway and show a novel mechanism through which p53 can inhibit angiogenesis. Cancer Res; 72(5); 1270–9. ©2012 AACR.
The tumor suppressor functions of p53 are primarily derived from its ability to act as a sequence-specific transcription factor and regulate expression of a diverse array of genes. The most extensively studied p53 target genes are those that mediate cell autonomous effects such as cell-cycle arrest, senescence, and apoptosis (1). However, p53 can also upregulate genes that limit tumor growth by inhibiting effects such as metastasis and angiogenesis.
The ability of p53 to limit angiogenesis has been shown in several studies in which p53 mutation in prostate (2, 3), colon (3–5), head and neck (6), and breast cancers (7) was shown to correlate with increased microvessel density (MVD). This notion was further supported by animal models showing that reversal of the angiogenic switch required p53 (8) and that p53 could induce tumor dormancy by limiting angiogenesis (9, 10).
p53 negatively regulates angiogenesis by both inhibiting the production of proangiogenic factors such as VEGF and by increasing production of antiangiogenic ones such as thrombospondin (TSP-1) (11). p53 has also been shown to induce production of antiangiogenic factors derived from collagen such as endostatin and tumstatin (12). There are seven collagens identified to date that contain antiangiogenic activity in their noncollagenous 1 (NC1) domains (13–20). Experiments using transgenic mice have suggested that even modest increases in the levels of these factors can significantly slow tumor growth (21). Conversely, mice lacking expression of the precursor of tumstatin, α3 collagen IV, showed increased rates of tumor growth (22).
Several studies suggest that production of collagen-derived antiangiogenic factors (CDAF) may function in the p53 tumor suppressor pathway. A rate-limiting enzyme in collagen biosynthesis, α(II) prolyl-hydroxylase, was shown to be a transcriptional target of p53 (12). In addition, genome-wide screens for p53-binding sites have suggested that two collagen genes, COL18A1 (23) and COL4A1 (24), are targets of p53. These genes encode α1 collagen XVIII and α1 collagen IV that include the CDAFs endostatin and Arresten, respectively. In the current study, we show that p53 upregulates the expression of the COL4A1 gene leading to release of the C-terminal noncollagenous domain of α1 collagen IV, which contains Arresten. We propose these effects to be a p53-mediated response to limit tumor angiogenesis.
Materials and Methods
Additional details are provided in the Supplemental online Materials.
Cell lines and virus
H1299 and PC3 cells were purchased from American Type Culture Collection. HCT116 cells (p53 wild-type and null) were a gift from Bert Vogelstein (25). Cell lines were routinely authenticated by immunoblot analysis of p53. Human umbilical vein endothelial cells (HUVEC) were bought from Clonetics. Ad-p53 and Ad-lacZ adenovirus have been described previously (12).
Whole-cell extracts were obtained by harvesting cells and boiling in 1X Laemmli buffer. Endogenous proteins in conditioned media were precipitated with 15% TCA. V5-His–tagged collagen fragments in the conditioned media were precipitated with TALON Superflow Metal Affinity Resin (BD Biosciences).
Quantitative real-time PCR
RNA was isolated using TRIzol (Sigma). Reverse transcription was conducted with QuantiTect Reverse Transcription Kit (Qiagen), and PCR reactions were carried out by QuantiTect SYBR Green PCR Kit (Qiagen). The 18S ribosomal RNA was used as internal control.
Chromatin immunoprecipitation (ChIP) experiments were conducted as previously described (26).
Analysis of tumors in mice
Animal experiments were approved by the McGill Animal Ethics Committee. A total of 10 × 106 H1299-COL4A1-V5 or control [empty vector (EV)] cells were injected subcutaneously into the back of nu/nu mice (Charles River). Tumor growth in vivo was measured as previously described (12). MVD was measured as previously described (27).
Collagen deposition assay
Immunofluorescence was carried out using a rabbit polyclonal antibody against total collagen IV (Abcam ab6586) or normal rabbit γ-globulin (Innovative Research).
Endothelial tube formation assay
HUVECs were plated onto Matrigel (BD Bioscience) in the presence of endothelial growth media (EGM-2) (Lonza) conditioned on H1299-COL4A1-V5 or H1299-EV stable cell lines. Tube complexity was measured as the number of branches in each field.
Ethics approval for sample collection included in this study was obtained from the ethics board of the Research Center of Université de Montréal Hospital Center (Montréal, QC, Canada). Patients signed an informed consent form.
Two-tailed Student t test was carried out for most experiments. One-way ANOVA was used for analysis of the quantitative real-time PCR (qRT-PCR) data. Analyses of the chromosome conformation capture (3C) data were done using the Kolmogorov–Smirnov test. For tissue microarray (TMA) Mann–Whitney and Kruskal Wallis tests were used to show mean expression differences between groups. Correlations between the various variables were evaluated using the Spearman rho. For all statistical tests, differences were considered significant when P value was less than 0.05.
COL4A1 is a major transcriptional target of p53
Several types of collagen have been reported to contain cryptic antiangiogenic domains (13–20). We have previously shown that p53 can enhance the secretion of these factors; however, the relative importance of the various collagens in the p53 pathway has yet to be elucidated. In Figure 1A, qRT-PCR analysis was used to compare the ability of p53 to induce expression of collagens known to contain antiangiogenic activity as well as the well-characterized antiangiogenic p53 target, TSP-1. An adenoviral vector expressing p53 (Ad-p53) was used to reintroduce p53 into H1299 cells that lack p53 expression. Figure 1A and C show that COL4A1 expression measured by qRT-PCR was induced over 200-fold by p53 relative to controls, which was far greater than all other antiangiogenic collagens tested.
COL4A1 expression was also compared in HCT116 cells that express wild-type p53 relative to HCT116 cells in which p53 has been deleted (28). Addition of 5-fluorouracil (5-FU) to these cells results in the stabilization of p53 and induction of p53-dependent genes such as p21 (Fig. 1D). Figure 1B shows that COL4A1 mRNA was the most highly induced antiangiogenic collagen in p53 wild-type but not p53−/− cells following treatment with 5-FU, confirming that induction occurs under conditions using endogenous p53. Surprisingly, induction of COL4A1 far exceeded that of the well-known antiangiogenic p53 target TSP-1 in both H1299 and HCT116 cells.
The mechanism by which p53 enhances expression of COL4A1 seems to be unique from other p53 targets. The collagen type IV family consists of six different genes (α1-α6) arranged in pairs with a head-to-head orientation on three different chromosomes in the human genome. The COL4A1 and COL4A2 gene are structured in this manner and share a common promoter sequence (29). Sequence analysis of the promoter of COL4A1/A2 did not reveal any canonical p53-binding site. However, genome-wide analysis for p53-binding sites reported the presence of a perfect p53-binding site 26 kbp downstream from the 3′ end of the COL4A1 gene (ref. 24; Fig. 2B). To confirm the occupancy of p53 on the putative-binding site, ChIP was carried out. Binding of p53 to the putative-binding site was significantly enriched compared with a gene desert control region (Fig. 2A). Because the identified p53-binding site is 26 kbp away from the 3′ end of COL4A1, we then asked whether the distant p53 site interacts with the promoter through long range looping interactions as are often observed with transcriptional enhancers. Using 3C technique (30), a greater proximity between the promoter of COL4A1 and its 3′ p53-binding site was observed upon p53 expression (Fig. 2B). This result suggests that p53 induces COL4A1 expression from a 3′ end enhancer by a long-range mechanism.
p53 expression stimulates the production of Arresten
We then examined the effect of p53 expression on the production of the full-length α1 collagen IV. Figure 3A shows that expression of p53 in H1299 cells results in no significant change in full-length α1 collagen IV protein levels in whole-cell extracts. However, because previous studies have shown that soluble antiangiogenic fragments of α1 collagen IV can be released from the extracellular matrix (ECM; ref. 15), we analyzed the conditioned media by Western blotting. Interestingly, the conditioned media from p53-expressing cells showed the appearance of NC1 containing proteins of approximately 80 and 30 kDa in size (Fig. 3A, right). To confirm that the bands in the conditioned media are derived from α1 collagen IV, a stable cell line was derived expressing full-length COL4A1 containing the V5-His epitope tag fused to the C-terminus. Using this cell line (H1299/COL4A1-V5), NC1-containing peptides were purified from conditioned media using metal affinity resin and detected with anti-V5 antibody. Figure 3B shows that COL4A1-expressing cells produce an 80 kDa product similar to that observed with the endogenous protein. Expression of p53 in the H1299/COL4A1-V5 cells results in enhanced production of Arresten (Fig. 3B), suggesting that p53 can also stimulate the production of the protein post-transcriptionally as the transgene is under the control of an exogenous promoter. These results show that p53 expression can enhance the expression and processing of Arresten.
To confirm that p53-dependent induction of Arresten could be observed without the need for overexpression of p53 or COL4A1, we used the p53 wild-type and p53−/− HCT116 cell lines described earlier. Figure 3C shows that when wild-type HCT116 cells are treated with 5-FU, an 80 kDa Arresten band is observed in the conditioned media, which is barely detectable in conditioned media from p53−/− cells. The increase in α1 collagen IV protein is specific to the soluble C-terminal fragment as no significant increase in the full-length protein is observed in whole-cell extracts (Fig. 3C, top).
p53-induced Arresten production is independent of caspase activation and requires matrix metalloproteinase activity
Expression of p53 can result in the induction of apoptosis, and therefore, the activation of caspases. To determine whether the processing of α1 collagen IV into Arresten was dependent upon caspase activity, H1299 cells were treated with the caspase inhibitor zVAD-fmk. Supplementary Figure S1 shows that zVAD-fmk prevented the activation of caspase-3 but had no effect on the appearance of Arresten in the conditioned media in response to p53.
Release of the antiangiogenic peptides from parent collagen requires proteolysis. Previous studies have implicated matrix metalloproteinase (MMP) in the production of tumstatin, which is derived from α3 collagen IV (22). To determine whether MMP activity is required for Arresten production, cells were treated with the MMP inhibitor, GM6001. Figure 3D shows that addition of GM6001 reduced levels of Arresten in the conditioned media of cells expressing p53 but did not affect full-length collagen in whole-cell extract. These data indicate that production of Arresten is not a consequence of apoptosis induced by p53 but rather is part of an antiangiogenic program whereby α1 collagen IV expression is increased and processed into Arresten by MMP activity.
Prolyl-hydroxylase activity stimulates the production of α1 collagen IV and Arresten
We have previously reported that a rate-limiting enzyme in collagen production, α(II) collagen prolyl-hydroxylase [α(II)PH], is a target gene of p53 (12). We, therefore, asked whether α(II)PH could potentiate the production of Arresten. Figure 4A shows that cell lines overexpressing α(II)PH were capable of producing more full-length α1 collagen IV than control cells when transfected with a plasmid encoding the COL4A1 gene. Moreover, the conditioned media of α(II)PH-expressing cells contained more of the fully processed Arresten peptide (Fig. 4A, right).
To confirm that prolyl-hydroxylase activity is essential for Arresten production, we examined the effect of an inhibitor of α(II)PH, ethyl-3,4-dihydroxy benzoate (EDHB), on α1 collagen IV and Arresten production. Figure 4B shows that EDHB reduced levels of Arresten in conditioned media after p53 expression. Taken together, these results suggest that p53 increases production of Arresten in a process that is dependent upon the α(II)PH enzyme.
α1 collagen IV expression is antiangiogenic and inhibits tumor formation
The H1299/COL4A1-V5 cells release an NC1-containing peptide of approximately 80 kDa in size into the conditioned media (Fig. 3B). We, therefore, determined whether these cells secrete antiangiogenic activity. Figure 5A and B show that conditioned media from H1299/COL4A1-V5 cells significantly inhibited tube formation of HUVECs and increased HUVEC cell death in vitro. It should be noted that even though this is a stable population of COL4A1-expressing cells, the level of Arresten released by these cells is less than that observed when endogenous Arresten is induced by p53 (Supplementary Fig. S2). The antiangiogenic effect of COL4A1 expression in this system is, therefore, unlikely to be an artifact of overexpression.
The p53-dependent induction of COL4A1 expression and production of Arresten may represent a mechanism to limit tumor angiogenesis and growth. To test this possibility, H1299/COL4A1-V5 or vector control cells were transplanted on the backs of nude mice. Figure 5D shows that H1299/COL4A1-V5 tumors displayed significantly slower rates of tumor growth relative to vector alone (EV). This effect was not due to differences in cell growth or apoptosis as no major changes in these parameters were observed (Fig. 5C). Histologic examination of tumors revealed that MVD within the α1 collagen IV–expressing tumors was significantly lower suggesting that the slower tumor growth was due to an antiangiogenic effect (Fig. 5E).
Expression of p53 remodels collagen IV matrix and inhibits tube formation
Type IV collagen serves as a major reservoir for antiangiogenic factors that can be released by proteolysis, but conversely, stable collagen IV matrix is an essential component of vascular basement membranes and is, therefore, essential for angiogenesis (31, 32). Because p53 expression in tumor cells increases the production of peptides derived from α1 collagen IV, this suggests that p53 may promote remodeling of collagen IV matrices to prevent angiogenesis. To form stable protomers and assemble into a matrix, both α1 and α2 collagen IV must be present. These proteins associate in a heterotrimeric triple-helix and subsequently assemble into a large-scale matrix through interactions with the globular NC1 domains (32). We, therefore, examined the effect of p53 on collagen IV deposition of tumor cells. H1299 cells do not form any appreciable collagen IV matrix due to lack of baseline expression of both α1 and α2 collagen IV (see Fig. 3 and Supplementary Fig. S3). The prostate cancer cell line PC3, however, expresses both α1 and α2 collagen IV and forms a robust collagen IV matrix (Fig. 6B). Importantly, PC3 cells also completely lack expression of p53, which allows the effects of p53 reexpression to be addressed.
Figure 6A shows that both α1 and α2 collagen IV are present in PC3 cells and when p53 is reintroduced, NC1 fragments are released in the conditioned media. The extent of collagen deposition on the surface of these cells was examined by immunofluorescence. Figure 6B shows that p53 expression following matrix deposition completely abolished the collagen IV matrix surrounding these cells. In this experimental system, p53 expression did not result in any appreciable induction of apoptosis or changes in cell morphology (Supplementary Fig. S4).
Remodeling of the collagen IV matrix by p53 could represent a potential antiangiogenic mechanism destabilizing the basement membrane essential for angiogenesis. We, therefore, investigated the effect of p53-conditioned media on tube formation of endothelial cells growing on reconstituted basement membranes (Matrigel). Matrigel monolayers were incubated with conditioned media from p53-expressing or control cells. After 72 hours, the conditioned media was removed and HUVECs were added to the treated Matrigel and assessed for tube formation. Figure 6C shows that Matrigel treated with p53-conditioned media was unable to support tube formation. Intriguingly, addition of the MMP inhibitor GM6001 to the p53-conditioned media during Matrigel treatment blocked this effect, suggesting that the inhibition of tube formation was dependent upon MMP activity.
A previous study has shown that p53 can transcriptionally activate expression of the MMP2 gene, which encodes a collagenase well known to degrade type IV collagen (33). We, therefore, determined whether p53 is able to induce MMP2 expression in PC3 cells. Supplementary Figure S5A shows that infection of PC3 cells with Ad-p53 resulted in a 4-fold induction of the MMP2 gene. To determine whether MMP2 activity is able to release the Arresten NC1 domain from PC3 cells, a hemagglutinin (HA)-tagged cDNA expressing MMP2 was transfected into PC3 cells. Supplementary Figure S5B shows that expression of MMP2 induced the appearance of a similar 80 kDa Arresten band observed upon p53 expression.
These data suggest that p53 affects collagen IV by two mechanisms that can inhibit angiogenesis. First, p53 enhances production of the antiangiogenic factor Arresten. Second, p53-dependent induction of MMP2 or other ECM proteases remodel existing collagen IV networks into a destabilized state that cannot support attachment and migration of endothelial cells.
p53 mutation correlates with Arresten expression in human prostate cancer
Because p53 increases production of Arresten in vitro, we then asked whether p53 status in primary human cancers correlate with expression of Arresten. A TMA containing 99 cases of prostate cancer was used for these studies. A summary of the clinicopathological parameters of the patient cohort present on this TMA is shown in Supplementary Table S1. Even though p53 mutations are not as common in prostate cancer as in other carcinomas, when present they correlate with poor patient survival (34). Prostate cancers carrying p53 mutations are believed to be a small but highly aggressive subgroup that has a high risk of progression even after radical prostatectomy. To identify tumors that contain p53 mutations, immunohistochemistry was carried out to visualize nuclear accumulation of p53. Nuclear accumulation of p53 correlates with the presence of missense mutations in the TP53 gene (35–37). A duplicate set of TMAs was also stained with the same monoclonal antibody against Arresten used in the earlier experiments.
In agreement with previous studies, we find that the p53 mutation correlates with several negative diagnostic outcomes including bone metastasis (correlation coefficient = 0.345, P < 0.001) and overall mortality (correlation coefficient = 0.348, P < 0.001). Mutant p53 was detected in 17% of the cases examined, which fits well with previously reported p53 mutation frequencies in prostate cancer (36). Interestingly, presence of mutant p53 significantly correlated with lack of Arresten staining within the tumor (correlation coefficient = −0.211, P = 0.036). In normal prostate tissue, very little staining was observed for either p53 or Arresten (Supplementary Fig. S6A). Representative tumors are shown in Fig. 7. These data provide clinical relevance to the in vitro findings we have described thus far and show that loss of p53 activity in human tumors results is reduced levels of Arresten.
Several types of collagen have been shown to have potent antiangiogenic domains that can be released by proteolysis of ECM (31). Our previous work has suggested that these factors can be mobilized downstream of the p53 pathway (12). The relative importance of each of the various types of antiangiogenic collagen in the p53 pathway remains unclear. In the current study, we show that the collagen gene COL4A1, which contains the antiangiogenic peptide Arresten, is the most highly induced CDAF containing collagen in response to p53. We show that p53 is able to increase Arresten production, through three distinct mechanisms. First, p53 is able to directly increase transcription of the COL4A1 gene that contains Arresten. Second, α(II)PH, which we have previously characterized as a p53 target gene also potentiates the production of full-length α1 collagen IV and Arresten (12). Finally, p53 promotes the MMP-dependent remodeling of the collagen IV matrix in the ECM, which further enhances the processing of α1 collagen IV. Because a stable collagen IV matrix in vascular basement membranes is required to form vessels (31, 32), remodeling of collagen IV in the ECM would prevent the association of endothelial cells with the destabilized matrix and be potentially antiangiogenic.
We observe a predominant 80 kDa C-terminal fragment of α1 collagen IV released upon p53 activation in H1299, HCT116, and PC3 cells. This peptide is considerably larger than the 26-kDa NC1 domain that has been previously shown to have antiangiogenic activity (15). The actual size of Arresten present in vivo has never been reported. Although the 80-kDa species is the predominant form of Arresten released from cells in culture, we cannot rule out that additional processing steps could take place in vivo.
Our data support an antiangiogenic function for p53 in which collagen IV is shed from cells into the tumor microenvironment. This would result in Arresten and possibly other collagen IV–derived antiangiogenic factors inhibiting endothelial cell growth and tube formation. Of the six type IV collagens, five of these have been reported to possess domains that are antiangiogenic (18). These collagens represent an enormous quantity of antiangiogenic potential that can be released by proteolysis of basement membrane with MMPs or other ECM-degrading enzymes. Measurement of physiologic serum concentrations of these factors in humans has never been reported and it would be interesting to determine whether serum levels of Arresten or other factors correlate with clinical parameters such as overall survival in patients with cancer.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
This research was supported by grants from the Canadian Institute of Health Research (CIHR) to J.G. Teodoro (MOP-86752 and MOP-115195) and J. Dostie (MOP-86716). J.G. Teodoro and J. Dostie are both CIHR New Investigators and FRSQ Research Scholars. S. Assadian. was supported by a Canderel/FRSQ studentship and CIHR doctoral award. W. El-Assaad was supported by a postdoctoral fellowship from the CIHR cancer training program. X.Q.D. Wang is supported by a scholarship from the Cole Foundation.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The authors thank Bert Vogelstein for cell lines and Isabelle Gamache for invaluable technical assistance.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
- Received July 15, 2011.
- Revision received November 28, 2011.
- Accepted December 12, 2011.
- ©2012 American Association for Cancer Research.