Tumor Angiogenesis in the Absence of Fibronectin or Its Cognate Integrin Receptors

Binding of α5β1 and αvβ3/β5 integrin receptors on the endothelium to their fibronectin substrate in the extracellular matrix has been targeted as a possible means of blocking tumor angiogenesis and tumor growth. However, clinical trials of blocking antibodies and peptides have been disappointing despite promising preclinical results, leading to questions about the mechanism of the inhibitors and the reasons for their failure. Here, using tissue-specific and inducible genetics to delete the α5 and αv receptors in the endothelium or their fibronectin substrate, either in the endothelium or globally, we show that both are dispensable for tumor growth, in transplanted tumors as well as spontaneous and angiogenesis-dependent RIP-Tag-driven pancreatic adenocarcinomas. In the nearly complete absence of fibronectin, no differences in vascular density or the deposition of basement membrane laminins, ColIV, Nid1, Nid2, or the TGFβ binding matrix proteins, fibrillin-1 and -2, could be observed. Our results reveal that fibronectin and the endothelial fibronectin receptor subunits, α5 and αv, are dispensable for tumor angiogenesis, suggesting that the inhibition of angiogenesis induced by antibodies or small molecules may occur through a dominant negative effect, rather than a simple functional block.


Introduction
Extracellular matrix proteins and their adhesion receptors are enticing targets for the regulation of tumor angiogenesis. The recruitment of new blood vessels by tumors is an important bottleneck in tumor development, without which tumors fail to grow. Thus, targeting tumor angiogenesis has been a therapeutic goal. Endothelial cell migration and survival is strongly regulated in vitro by adhesion to extracellular matrix, mediated by integrin receptors on the endothelium. Since the endothelium and its underlying matrix are readily targeted with small molecules and antibodies, disrupting matrix-integrin interactions would seem to be a useful method of inhibiting tumor angiogenesis.
Interactions between the extracellular matrix protein Fibronectin (FN) and its integrin receptors were some of the first such proposed targets, since FN and its receptors are strongly expressed around the tumor vasculature, and both are essential for developmental angiogenesis. Embryos and embryoid bodies deficient in FN fail to form vascular networks, despite proper Here, to determine the absolute requirement for the FN-integrin interactions in tumor angiogenesis, we used genetic tools to remove the FN-binding integrin subunits α5 and αv from the endothelium, prior to tumor growth, in two transplant models and the RIP-Tag model of angiogenesis-dependent pancreatic cancer. We further tested the requirement for FN, from any source, in RIP-Tag tumors in mice with global post-natal deletion of FN.

Materials and Methods Mice
Integrin Itga5 f/f ; Itgav f/f (also called α5 f/f ; αv f/f ) mice were generated from an intercross of our existing Itga5 f/f and Itgav f/f lines, as we previously reported [11,21]. FN f/f mice were obtained from R. Fassler [22]. To obtain endothelial deletion of integrins α5 and αv, or FN, the respective lines were crossed with Cdh5-CreER T2 (obtained from R. Adams [23]) or ROSA-CreER T2 (obtained from T. Jacks [24]) mice. Cre-mediated excision efficiency was monitored with the fluorescent mT/mG Cre-reporter mice [25], crossed with the FN f/f mice. The reporter switches from red to green when activated by Cre. RIP1-Tag2 mice were obtained from the National Cancer Institute [26], and intercrossed with the α5 f/f ; αv f/f , FN f/f , Cdh5-CreER T2 , ROSA-CreER T2 , and mT/mG mice.
In experiments with Cdh5-CreER T2 mice, excision was induced with 3 x 1mg Tamoxifen by intraperitoneal injection one week before tumor transplant, or at weeks 5-6 in RIP-Tag mice. Excision in ROSA-CreER T2 mice was induced with 5 x 1mg Tamoxifen by intraperitoneal injection at weeks 5-6 in RIP-Tag mice, and then either continued with 1 x 1mg weekly, or not.
All mice were housed and handled in accordance with approved Massachusetts Institute of Technology Division of Comparative Medicine protocols (IACUC approval 0412-033-15).
In brief, animals are monitored daily by animal care staff and laboratory personnel and appropriate measures are taken if infection, inflammation or distress occur. If body condition score is <2 or if aggregate tumor burden exceeds 1 cm in diameter, mice are euthanized by CO2 narcosis. We use a regulated flow valve (in addition to cylinder regulator) that restricts the CO2 flow rate to 20% of chamber volume per minute. These methods were chosen because they cause minimal stress to the animal. All methods of euthanasia are consistent with the recommendations of the Panel on Euthanasia of the American Veterinary Medical Association.

RNA isolation and quantitation
RNA was extracted using the Qiagen RNAeasy mini-column kit, after combining the chloroform extract from Trizol 1:1 with 70% ethanol. A Promega kit was used to random prime a cDNA library from each tumor before expression analysis using the Bio-Rad iQ sybr mix (Table 1).

Staining and fluorescence quantification
For quantification of fluorescence, tumors were fixed in 4% PFA, sucrose infused and snap frozen in Optimal cutting temperature compound (OCT, Tissue Tek) on a metal block cooled by liquid nitrogen, and then sectioned at 10μm thickness. Sections blocked in 2% BSA, 0.1% Tri-tonX-100 in PBS were stained using the primary antibodies indicated ( Table 2). All samples were stained and imaged in parallel for experiments in which comparisons are made.
Quantitation of staining area was performed in ImageJ. An arbitrary threshold was set, and the percentage of area above this threshold was quantified.

Statistics
Student's t test was used to compare means of two independent groups to each other.

Tumor growth is not affected by the absence of endothelial integrin subunits Alpha5 and AlphaV
To determine the requirement for α5 integrins in tumor angiogenesis, Tie2-cre mice were crossed with α5 f/f mice [11]. We previously found that this results in efficient deletion of endothelial α5 by embryonic day 10.5 [11]. Despite the absence of endothelial α5, we observed no defects in the final tumor mass of either subcutaneous Lewis Lung tumors, or B16 melanoma tumors (Fig. 1A&1B).
Since developmental deletion of α5 was tolerated, but we had previously found that the combined Tie2-Cre mediated deletion of α5 and αv in the endothelium resulted in vascular remodeling defects in the embryo, we hypothesized that αv might similarly compensate for the absence of α5 in tumor angiogenesis [11]. To test this hypothesis, we generated α5 f/f ; αv f/f mice, and crossed them with Cdh5(PAC)-CreERT2 (hereafter referred to as Cdh5-CreER) mice allowing inducible deletion in the endothelium. In these mice, Cre was activated by tamoxifen treatment 1 week prior to tumor inoculation. We found no defect in the growth of the implanted Lewis Lung tumors (Fig. 1C). Transplant models are highly selected and aggressive, therefore we also examined the requirement for endothelial α5 and αv in the RIP1-Tag2 (hereafter referred to as RIP-Tag) model of pancreatic cancer [27]. Transformation of the beta cells of the pancreas in this model is induced by insulin-promoter-driven SV40 expression. Cancer growth follows a well- characterized progression from hyperplastic islets at 4-5 weeks of age, to angiogenic islets at about 7 weeks of age, and encapsulated adenomas by 10 weeks of age, and then finally to invasive adenomas by 14 weeks of age [28]. This is a selective process, and of the initial~400 islets, 50% become proliferative,~10% become angiogenic and only 1-2% go on to become tumors [29]. The model allows the genetic detection of defects in the angiogenic stage [30] as well as a later growth stage [31].
We treated mice with Tamoxifen at 5-6 weeks of age, and counted the number of angiogenic islets at 10 weeks of age. Here again, we saw no obvious reduction in the number of angiogenic islets, a measure of tumor angiogenesis (Fig. 1D).
We confirmed that Cdh5-CreER effectively excises endothelial α5 and αv by several methods. First, we crossed the mT/mG Cre reporter line with the Cdh5-CreER; α5 f/f ; αv f/f mice. The reporter expresses membrane RFP in all cells from the ROSA locus, and converts to membrane GFP with Cre activation. We induced Cre activity by tamoxifen treatment 1 week prior to Lewis Lung tumor cell inoculation and examined tumor sections for Cre reporter activity, with co-staining for CD31 to confirm the location of endothelial cells. We found that Cdh5-CreER induced strong deletion (green) within the endothelium of Lewis Lung tumors (Fig. 1E). Cre activation strongly overlapped with CD31-labeled endothelium, whether or not the floxed α5 and αv alleles were included in the cross (Fig. 1E), suggesting that excised endothelial cells are not selected against in tumor growth. Staining of sections of these tumors for α5 showed strong α5 staining in the endothelium of controls, but very little staining in the marked endothelial cells of Cdh5-CreER; mT/mG; α5 f/f ; αv f/f mice (Fig. 1F). Furthermore, endothelial expression of α5 and αv mRNA was effectively depleted, as measured in RNA collected from the aortic endothelium of similarly treated mice (Fig. 1G).
Thus, we conclude that α5 and αv are dispensable for tumor angiogenesis. Other integrins (α4β1, α8β1 and α9β1) are able to bind FN, and both α4β1 and α9β1 are expressed on the endothelium, albeit at lower levels. Thus, we cannot exclude the possibility that they are able to compensate for the loss of endothelial α5 and αv.

Endothelial deletion of Fibronectin does not inhibit tumor growth
Given the practical limitations of targeting all of the FN-binding integrins, we turned our attention to FN itself. To determine the requirement for endothelial FN in tumor growth, we used Cdh5-CreER to excise FN in the endothelium one-week prior to transplant of Lewis Lung tumor cells. We have previously found this results in >90% deletion of endothelial FN [32]. The absence of endothelial FN did not significantly affect the final Lewis Lung tumor mass ( Fig. 2A).
Similarly, when we treated RIP-Tag mice with Tamoxifen at 6 weeks of age, we observed no reduction in the number of angiogenic, red islets in Cdh5-CreER; FN f/f mice at 10 weeks of age (Fig. 2B).
Therefore, we conclude that endothelial FN is also not required for tumor angiogenesis. This, of course, does not exclude contributions of FN from other sources, since FN is produced by almost all of the cells in the tumor, and is also abundant in plasma.

Global deletion of Fibronectin delays the formation of angiogenic islets but does not affect final tumor mass
Total FN deletion is embryonic lethal, therefore we used Rosa-CreER to bypass embryonic lethality and excise FN globally in the post-natal mouse. We examined the requirement for FN in the RIP-Tag model system.
We deleted FN from the RIP-Tag tumors at the angiogenic stage (5-6 weeks) by tamoxifen treatment. To determine the effect of FN deletion at the angiogenic stage, we counted the number of angiogenic tumors at 10-11 weeks of age. We found that deletion of FN reduced (~30%) the total number of angiogenic islets (Fig. 3A-3C, P<0.004). To determine the effect on final tumor mass, we weighed all of the tumors from the pancreas of each individual mouse at 12-13 weeks. Surprisingly, we saw no significant effect on final tumor mass (Fig. 3D).
To confirm the efficiency of gene deletion, we examined the mT/mG Cre reporter in Rosa-CreER; mT/mG; FN f/f ; RIP-Tag mice. We found that excision in the islets and adjacent pancreas was robust 1 week after the first tamoxifen treatment (Fig. 3E). Initially, we had performed analysis with continuous weekly tamoxifen treatment, to prevent outgrowth of cells with intact FN floxed sites. However, analysis of the largest tumors from mice with either a single (5x 1mg, over one week) or continuous (5x 1mg + 1mg weekly thereafter) tamoxifen treatments showed no significant difference in final tumor mass (Fig. 3D), or mT/mG Cre-reporter activation at endpoint (Fig. 3E), suggesting that outgrowth of cells with intact FN floxed sites does not explain the maintenance of tumor growth.
To confirm the ablation of tumor FN, we examined sections of tumors from Rosa-CreER; mT/mG; FN f/f ; RIP-Tag mice and their Cre-negative or tamoxifen-negative controls at 12-13 weeks. We found the characteristic vascular pattern of FN staining was absent in the Rosa-CreER; mT/mG; FN f/f ; RIP-Tag mice with either single or continuous Tam treatment (Fig. 3E). Quantification of the stained areas revealed a nearly complete ablation of FN in the Rosa-CreER; mT/mG; FN f/f ; RIP-Tag tumors (Fig. 3F).
We asked whether there was any difference in the depletion of DOC-soluble FN versus the DOC-insoluble, or "fibrillar" FN. We found that, consistent with the histological immunofluorescence staining, both total and soluble FN were depleted by 12 days after the first tamoxifen   (Fig. 3G). FN was not entirely removed from the insoluble pool. It is not clear whether the remaining insoluble FN is derived from existing matrix, prior to FN excision, or FN deposited subsequent to FN excision and tumor growth.
We next asked whether FN expression by the tumor cells themselves was also reduced. First, we examined the efficiency in the deletion of FN in the tumors by quantitative PCR. We found that Cre activity resulted in a reduction of intact FN alleles by 40-75% with a single period of Tamoxifen treatment and 50-90% with continuous Tamoxifen treatments (Fig. 3H). RIP-Tag tumors were targeted less efficiently than the liver, which yielded a consistent >90% reduction in intact FN alleles with a single Tamoxifen treatment period (Fig. 3H, "liver"). There was no obvious increase in excision frequency with continuous Tamoxifen treatments.
We examined the levels of FN RNA, relative to 18s RNA in the largest tumors of individual mice. We found that FN expression was variable, but that several of the largest tumors isolated from Rosa-CreER; FN f/f mice had >150-fold depletion of FN transcript (Fig. 3I).
Thus, we conclude that the genetic deletion of almost all FN in RIP-Tag tumors slightly reduces the number of initial angiogenic islets, but does not significantly suppress final tumor mass.

Angiogenesis and deposition of vascular basement membrane is not suppressed by Fibronectin deletion
In vitro studies [9] and in vivo developmental studies [1,2] had suggested FN was essential for angiogenesis, yet apparently normal tumor growth occurred in the near absence of FN. Therefore, we examined tumor angiogenesis closely.
To determine whether the deletion of FN suppressed angiogenesis in the tumors of Rosa-CreER; FN ff ; RIP-TAg mice, we examined CD31 staining by immunofluorescence staining of histological sections of 12-13 week tumors from Rosa-CreER; FN f/f mice and FN f/f controls. We found no significant reduction in CD31-stained tumor area in the Rosa-CreER; FN f/f ; RIP--Tag mice (Fig. 4A-4C).
FN is typically found adjacent to CD31-stained blood vessels in RIP-Tag tumors, but perivascular FN staining is absent in Rosa-CreER; FN f/f ; RIP-Tag mice (Fig. 4D). Since FN assembly has been shown to be important in basement membrane assembly in vitro [10], we tested whether the absence of FN impaired the assembly of the basement membrane in tumors in vivo. However, despite the nearly complete absence of FN in the tumor, basement membrane components ColIV, laminin, Nid-1 and Nid-2 appeared to be properly assembled (Fig. 4E).
Thus, we conclude that nearly complete absence of FN in Rosa-CreER; FN ff ; RIP-Tag tumors does not reduce vessel density or appreciably alter the composition of basement membrane components.

No significant defects in the recruitment of a FN-linked regulator of TGFbeta signaling
In vitro studies have shown that FN plays a critical role in the deposition of other ECM proteins, including collagens, fibrillins, fibulins, latent TGFβ-binding proteins (ltbps), and Tenascin-C [8]. However, in vivo studies are not as clear. In mice with Mx-Cre-mediated excision of floxed FN in the liver, injury still induces a robust increase in ColI, ColIII, Ltbp-1, Ltbp-3, and Ltbp-4 incorporation into the extracellular matrix [33,34].
We examined the expression of Fibrillin-1, Fibrillin-2, Ltbp-1, Ltbp-3 and Ltbp-4 by immunofluorescence in FN-deleted RIP-Tag tumors. We found abundant Fibrillin-1 expression, and no reduction in the percentage of tumor area with Fibrillin1 staining, relative to control tumors (Fig. 5A&5B). Only minimal levels of Fibrillin-2, Ltbp-1, Ltbp-3 and Ltbp-4 could be detected in tumors of either type, and were not obviously different (Fig. 5A and data not shown). Staining patterns for Fibrillin-1 were similar when detergent was excluded from the staining buffer, suggesting that the fibrillar staining is extracellular matrix and not intracellular deposits (S1 Fig.).
Thus, we conclude that nearly complete deletion of FN in Rosa-CreER; FN f/f ; RIP-Tag tumors does not appreciably affect Fibrillin-1 incorporation into the extracellular matrix.

Discussion
FN and the FN integrin receptors α5β1 and αvβ3/β5 have been thought to be critical for new angiogenesis. Although antibody and small molecule targeting of the adhesion of these receptors to their FN ligand was a promising target for anti-angiogenic therapies, early optimism has been tempered by a series of disappointing clinical trials. However, since lower doses of FN-integrin blocking peptides have been shown to promote, rather than suppress angiogenesis [16], it has not been clear whether the initial pre-clinical success and subsequent clinical failures reflect targeting difficulties or true biological redundancy. Here, using temporally regulated and tissue-specific genetic deletion, we demonstrate that endothelial expression of α5 and αv is dispensable for tumor growth in two transplant models as well as a "spontaneous," RIP-Tag-driven model. Furthermore, we demonstrate that endothelial FN is not required for tumor growth. In fact, nearly complete ablation of FN by ROSA-CreER-mediated global deletion does Fibronectin and Its Receptors are Dispensable in Tumor Angiogenesis not significantly impede tumor angiogenesis, deposition of basement membrane proteins, or recruitment of matrix-linked proteins Fibrillin-1 and -2 (Figs. 4 & 5). Together, our results demonstrate that the growth of these tumor models is not significantly dependent on FN or its α5β1 and αvβ3/β5 integrin receptors, and suggest that other mechanisms may explain the effects of high-dose RGD inhibitors and integrin-blocking antibodies on angiogenesis observed in pre-clinical studies.
Our data suggest that the antibodies targeting the interaction between FN and its receptors may be doing more than simply blocking this interaction. Blockage of α5β1 binding to FN is a common feature of blocking antibodies-but is not necessarily their only function. Early preclinical work showed that antibodies blocking α5β1 binding to FN interfered with adhesion and migration of HUVECs in vitro, and potently inhibited FGF-induced angiogenesis in a chick CAM assay [35]. Volociximab, a humanized antibody that interferes with the binding of α5β1 to FN, blocks endothelial cell binding to FN in vitro and inhibits angiogenesis in animal models [36,37]. A similar antibody, designed to bind to murine α5β1 and block its adhesion to FN, blocked angiogenesis in xenografted human tumors, suggesting that effects were mediated by interfering with murine host α5β1, rather than the human α5β1, which was not recognized by the antibody [38]. Yet, our results show that genetic deletion of either endothelial α5 or total FN has little effect on tumor angiogenesis. Differences in models may explain the differential effects. However, a more interesting possibility is that antibodies to α5β1 antagonize tumor angiogenesis not only by a simple block of the interaction between Fn and the α5β1 receptor, but rather by inducing some anti-angiogenic function in the integrin. The genetic deletion of β3 integrin (Itgb3) results in increased angiogenesis and tumor growth [39], while antibody targeting of αvβ3/β5 integrins or mutations of a tyrosine phosphorylation site that leave the β3 receptor intact but disrupt downstream signaling suppresses angiogenesis and tumor growth, suggesting a similar antagonistic effect [40]. While the findings in the RIP-Tag and transplant tumors we describe may not necessarily reflect the requirement for α5β1 and αvβ3/β5 integrins in human tumors, our results suggest that a better understanding of the effects of the blocking antibodies to α5β1 and αvβ3/β5 integrin receptors might be a fruitful path forward towards inhibiting tumor angiogenesis [41].
That robustly vascularized tumors grow in the near absence of FN was surprising, since embryos in which FN has been completely deleted have severe defects in developmental angiogenesis and die by embryonic day 8.5 to 9.5 [1,2]. Subsequent work showed a requirement for FN in vascular network formation in embryoid bodies [3], and a requirement for the assembly of soluble FN in angiogenesis [9]. Why are the pancreatic tumors, but not the embryo, able to establish vascular networks in the absence of FN? One potential explanation is that FN is not entirely deleted from the tumors, and that the remaining FN is sufficient. Analysis of genomic DNA revealed that deletion ranged from 40-90% in isolated tumors, and reduced DOCinsoluble FN could be detected in Western blot analyses. However, there is effectively no detectable extravascular FN in the tumors, suggesting that if FN is playing a role, it does so as an organizer, acting at low concentrations, rather than simply a "building-block" of the fibrils. Alternatively, the longer time-course (several weeks for tumor angiogenesis, versus several days for embryo angiogenesis) and growing genetic heterogeneity may allow tumors time to compensate for the loss of FN in ways that the developing embryo cannot. This hypothesis is supported by an initial delay (at 10 weeks) in tumor growth in mice deleted for FN that is eventually overcome (Fig. 3). Potential compensating RGD-containing extracellular matrix proteins include the collagens, fibrillins and nidogens, all expressed around RIP-Tag tumor vessels, as we show here. Which, if any, are important in compensating for the loss of FN remains unclear.
In vitro studies have suggested a critical function for FN as an organizer of the extracellular matrix, necessary for the integration of other matrix proteins, including ColIV, Laminin, Fibrillin-1 and -2 [8,10], yet we observed no significant reduction or alteration in the vascular deposition pattern in any of these matrix components in RIP-Tag tumors with nearly complete FN deletion. It is possible that, as with the requirement for FN in angiogenesis, the FN requirement in fibril organization can be achieved at a low concentration, or that its absence only slows, but does not halt matrix assembly. Subtle differences in assembly speed would be more clearly appreciated in vitro than in the weeks-long development of RIP-Tag tumors. Another possible explanation is that other matrix proteins could compensate for the function of FN in the deposition of other matrix proteins, as suggested by the initial delay in tumor growth that is overcome at later stages. The ECM produced by growing tumors is diverse, with dynamic changes in the ECM repertoire according to tumor phenotype [42][43][44][45][46]. In liver fibrosis, Col5 has been suggested to compensate for the deletion of liver FN [33]. Whether Col5, or other matrix proteins are able to compensate for the absence of FN in RIP-Tag tumor growth is unclear. Alternatively, compensation may be through changes in ECM-linked signaling pathways, such as Hippo [47]. Further analysis of compensatory measures in this model may reveal new modifiers of vascular matrix assembly.

Conclusions
Taken together, our results demonstrate that FN and the FN receptors are dispensable for tumor angiogenesis, raising new questions about the mechanisms underlying the anti-angiogenic activities of the blocking antibodies to the FN receptors and underscoring the importance of in vivo genetic studies to test the complex regulation of endothelial basement membrane assembly.