Increased Expression of α6 Integrin in Endothelial Cells Unveils a Proangiogenic Role for Basement Membrane Free

Angiogenesis, the development of new vessels from preexisting ones, plays a critical role in cancer progression. Tumor cells elicit angiogenesis through both enhanced production of proangiogenic factors such as vascular endothelial growth factor-A (VEGF-A) and fibroblast growth factor-2 (FGF-2), and decreased generation of angiogenesis inhibitors. In transgenic mouse models of multistage carcinogenesis—RIP-Tag, which develop spontaneous pancreatic islets (1, 2), enhanced angiogenesis in the premalignant stages, precedes tumor formation and the transition to an invasive carcinoma. Both genetic and pharmacologic inhibition of angiogenesis impaired disease progression (3–5). Thus, an angiogenic switch precedes and is potentially rate-limiting for tumor invasion and growth.

Although growth factors are required to elicit new blood vessel growth, adhesion to provisional extracellular matrix (ECM) proteins such as fibronectin and type I collagen is required for endothelial cell (EC) survival, proliferation, and motility during angiogenesis (6). Integrins mediate cell adhesion to the ECM, and multiple integrins are likely to contribute to angiogenesis (7).

VEGF-A and FGF-2 enhance the expression and activity of endothelial integrins, whereas negative regulators of angiogenesis, such as class 3 semaphorins, inhibit integrin function (8–10). Among many integrins expressed by EC, α5β1 and αvβ3 integrins have received special attention because of their involvement in the angiogenic process. Studies with adhesion-blocking reagents and knockout mice have been used to prove the involvement of α5β1 and αvβ3 integrins in angiogenesis (11–14). However, genetic studies in mice showed that integrin β4 signaling also plays a key role in tumor angiogenesis (15). Integrin β4 forms the laminin receptor α6β4 that is only expressed in a subset of ECs (15). In addition, integrin α6 could also form α6β1 heterodimers, which are expressed in the vascular endothelium. However, neither β4 nor α6 integrin knockout mice displayed vascular embryonic development defects but died immediately after birth, in part, due to severe skin blistering caused by passage through the birth canal (16, 17). To date, evidence for the role of α6 integrin in angiogenesis has been conflicting and requires further investigation (18, 19).

In this study, we show that α6 integrin is upregulated both in cultured EC stimulated by angiogenic growth factors and on the angiogenic vasculature of Rip-Tag2 tumors. Furthermore, here we show an important role of α6 integrin in regulating the onset of angiogenesis both in vitro and in animal models.

For in vitro morphogenesis on basement membrane extract (BME), Matrigel (BD Biosciences) or Cultrex (Trevigen) were added to each well at a concentration of 8 mg/mL and incubated at 37°C for 30 minutes to allow gel formation. EC (2 × 10⁴/well) were plated onto BME in the presence of VEGF-A (20 ng/mL; R&D Systems) and FGF-2 (10 ng/mL; R&D Systems). After 8 hours of incubation in 5% CO₂ humidified atmosphere at 37°C, cell organization was examined.

Collagen sprouting assay was performed as previously described (20). Briefly, EC spheroids were suspended in medium with or without 20 ng/mL of VEGF-A, and mixed with an equal volume of diluted collagen solution (0.6 mg/mL), collagen solution plus 50 μg/mL of laminin (Sigma) or BME gel (8 mg/mL). Antibodies were applied before gel solidification and administrated in medium with a final concentration of 20 mg/mL. Capillary-like sprouts were examined with inverted-phase contrast microscope (Leica Microsystems) and photographed. The lengths of the capillary-like structures were quantified with the imaging software winRHIZO Pro (Regent Instruments, Inc.).

The mouse aortic ring assay was performed as previously described (21) with modifications. Briefly, thoracic aortas were removed from 8- to 12-week-old wild-type C57/BL6 mice (Charles River) and fibroadipose tissue was dissected away. Aortas were sectioned in 1-mm-long aortic rings and incubated for 2 days in serum-free medium with antibiotics, polybrene, and lentiviral supernatant. Forty-eight–well culture dishes were coated with 100 μL of type I collagen (from rat tail; Roche) or BME and allowed to solidify. These were then sealed in place with an overlay of 70 μL of collagen or BME and covered with 300 μL of endothelial basal medium (Clonetics) 5% FCS or M199 10% FCS with VEGF-A (final concentration, 20 ng/mL; R&D Systems) and FGF-2 (final concentration, 10 ng/mL; R&D Systems). In the case of functional blocking antibody treatment, the final concentration of GOH3 or rat IgG was 10 μg/mL. Tubular structures were examined with inverted-phase contrast microscope (Leica Microsystems) and photographed. Lengths and projected areas of the capillary-like structures were quantified with the imaging software winRHIZO Pro (Regent Instruments).

Fertilized chick embryos were incubated for 3 days in 70% humidified atmosphere at 37°C. A small hole was made over the air sac at the end of the egg and a second hole was made directly over the embryonic blood vessels. After 10 days, cortisone acetate–treated filter discs (5 mm) saturated with 200 ng of FGF-2 or saline were placed on the chorioallantoic membrane (CAM) in an area with a minimum of small blood vessels. The day after, 10 μg of anti-α6 (GOH3) or rat IgG was applied on the filter discs, and eggs were incubated for 2 days. CAMs were analyzed as previously described (22), with the imaging software winRHIZO Pro (Regent Instruments).

The dosage regimen used was 0.125 mg of antibody per mouse through tail vein injection. Antibody treatment started when mice reached 9 weeks of age and continued for 15 days. Control animals were treated with purified rat IgG (rat IgGA2; R&D Systems) at a dose of 0.125 mg per mouse every 2 days for 2 weeks. Cohorts of six mice were treated per each arm of the trial study.

The details for the Materials and Methods, including reagents, cell culture, mice, quantitative reverse transcription-PCR (qRT-PCR), cytofluorimetric analysis, immunoprecipitation and immunoblotting analysis, immunofluorescence staining, short hairpin RNA (shRNA) lentiviral preparation, migration assay i.v. injection of antibody and detection, and analysis of tumor vasculature are available online as Supplementary Data.

Following microarray gene expression analysis, we found that mRNA expression of α6 integrin was significantly upregulated in human EC upon stimulation with VEGF-A and FGF-2 (data not shown). We confirmed these results by performing qRT-PCR experiments. EC stimulated for 24 hours with VEGF-A and FGF-2 increased the expression level of integrin α6 by 13- and 10-fold, respectively, compared with unstimulated cells (Fig. 1A). In the same experimental conditions, angiopoietin-1 induced a limited upregulation of α6 integrin mRNA (Fig. 1A).

We examined the surface expression of α6 integrin by flow cytometry analysis observing an increase of α6-positive cells after 24 hours of stimulation with VEGF-A and FGF-2 (Fig. 1B). Interestingly, in the absence of growth factors, ∼20% of EC expressed detectable levels of α6 integrin, whereas upon treatment with angiogenic growth factors for 48 hours, almost 80% of the cells became positive (Fig. 1B). Although α6 could form both α6β1 and α6β4 heterodimers, human cultured EC do not express detectable levels of integrin β4, therefore, only the α6β1 heterodimer is expressed (data not shown; ref. 15). Although immunoprecipitation experiments indicated that both subunits of α6β1 integrin were upregulated upon treatment with VEGF-A and FGF-2, the protein and mRNA level of total β1 integrin was not modulated by growth factor stimulation (Fig. 1A and C). Therefore, the amount of β1 associated with α6 was only dependent on α6 integrin expression, although some differences in the association ratio have been observed (Fig. 1C; Supplementary Fig. S1). These results suggest that α6 integrin was highly expressed on activated EC, but was nearly absent on quiescent EC.

To examine whether in vivo angiogenic vessels express α6, we analyzed its expression in the vasculature of Rip-Tag2 tumor mice (1). This genetic model of pancreatic insulinoma exhibits a well-characterized angiogenic switch that has been shown to be critical for the progression from hyperplastic to tumor stages (2, 5). At the angiogenic stage, new vessels formed into a tumor mass allowing the progression from hyperplasia to tumor. Integrin α6 was expressed both in vessels of the endocrine and exocrine sites of the pancreas, and it was detectable in all the stages of RIP-Tag2 tumor progression (Fig. 2A). In addition, it was expressed on membranes of exocrine epithelial cells, but not on islet epithelial cells (Fig. 2A). This specific tissue distribution allowed us to quantify, using qRT-PCR, the level of α6 mRNA on isolated islets. The level of α6 integrin was significantly high at the angiogenic stage and the expression pattern was comparable to those of β3 and α5 integrins, two well-characterized markers of angiogenic vessels (Fig. 2B). Moreover, to determine whether the higher number of vessels in the angiogenic stage, compared with hyperplastic and tumoral stages, could contribute to the enhanced α6 integrin expression level, we normalized its value to the expression level of VEGF receptor-2 (VEGFR2), and we still observed a significant increase of α6 at the angiogenic stage compared with normal and hyperplastic tissues (Fig. 2B). This increase is exclusively attributable to EC, as shown by colocalization experiments with VEGFR2 and VE-cadherin, two well-characterized markers of EC (Fig. 2C; Supplementary Fig. S2). Confocal microscopy analysis of RIP-Tag2 tissue confirmed the results obtained by qRT-PCR, showing a significant increase of α6 integrin expression in vessels during the angiogenic stage (Fig. 2D). In contrast, the level of VEGFR2 did not significantly change from the normal to the tumoral stages (Fig. 2D).

To examine whether α6 integrin was directly involved in the angiogenic process, we inhibited integrin function by blocking antibody and gene silencing. We assessed whether α6 integrin is required for in vitro angiogenesis by using the endothelial tube formation assay in which EC, placed on BME gels in the presence of angiogenic factors, self-organize into structures morphologically similar to capillaries.

The addition of an α6 integrin-blocking antibody completely inhibited the in vitro tube formation (Fig. 3A). Moreover, integrin α6 was stably downregulated in EC by shRNA lentivirus transduction, selecting two shRNAs that were able to silence >50% of the membrane protein compared with a scrambled shRNA (shScrl; Supplementary Fig. S3A and B). To rescue integrin expression, silenced EC were transduced with murine α6 integrin cDNA, which led to an expression of α6 at levels similar to wild-type cells (Supplementary Fig. S3B). As shown in Fig. 3B, EC silenced with either shITGA6_4 or shITGA6_5 failed to form tubular structures whereas shITGA6_4 EC expressing murine ITGA6 displayed a network similar to that of shScrl EC.

These results were not confirmed in another in vitro angiogenesis model, the collagen sprouting assay. In this model, spheroids of EC, embedded in collagen type I gel, sprout when stimulated by VEGF-A or FGF-2, forming tubular structures. EC spheroids with reduced ITGA6 expression, or treated with the α6 blocking antibody, sprouted to the same extent as spheroids transduced with shScrl or in the presence of rat IgG (Fig. 3C; Supplementary Fig. S4).

Nevertheless, the addition of laminin, an α6 integrin ligand, to collagen gel increased sprout formation whereas the blockade of α6 integrin significantly reduced sprout lengths (Fig. 3C). Similar results have been obtained by embedding spheroids in BME gel in which EC formed mainly cord-like rather than tube-like structures (Supplementary Fig. S4). Therefore, α6 integrin regulates in vitro angiogenesis in the presence of ECM-containing laminin but seems irrelevant when unbound to its specific ligand.

Although it is conceivable that these effects were consequences of loss of adhesion strengthening, experiments on different ECM indicated that the adhesive ability of α6 integrin-silenced cells was unaffected on collagen and BME, whereas it was only partially reduced on laminin (Supplementary Fig. S5A). To examine instead whether α6 integrin was involved in EC migration, we performed a wound/scratch assay on BME and collagen matrix. EC plated on BME, scratched and treated with anti-α6 blocking antibody, displayed a significantly decreased motility compared with EC treated with control IgG (Fig. 4A). Similar results were obtained with silenced cells, which showed reduced migration ability that was rescued by the re-expression of murine α6 integrin cDNA (Fig. 4B). In contrast, the absence of integrin α6 did not interfere in EC motility on collagen type I (Supplementary Fig. S5B). As expected, most of the focal adhesions at the leading edge of cell migrating on BME contained α6 integrin, suggesting its active role during directional motility, although it was not recruited on collagen-induced focal adhesions (Fig. 4C and D).

To exclude that levels of α6 integrin solely affect the angiogenic response of cultured EC, we applied lentivirus-mediated gene silencing to mouse aortic ring assays. The lentivirus transduction efficiency on aortic rings was evaluated with green fluorescent protein expression by fluorescence microscopy and cytofluorimetry. Approximately 50% of aortic ring cells were green fluorescent protein–positive when infected with >1 × 10⁶ viral particles (Supplementary Fig. S3A), and most of the vascular outgrowths were green fluorescent protein–positive, showing that sprouting EC were efficiently infected with lentiviral vectors (Supplementary Fig. S6B). By applying this method, we were able to silence α6 integrin expression in aorta and angiogenic sprouts, which are both positive for α6 integrin immunostaining (Supplementary Fig. S7A and B). A reduction of mItga6 expression, evaluated by RT-PCR on cells extracted from the gel, was observed in aortic rings infected with two different shRNA, shITGA6_48 and shITGA6_50 (Supplementary Fig.S7C). Aortic rings infected with the scrambled shRNA, after embedding in BME matrix and VEGF-A/FGF-2 stimulation, sprouted tubular structures (Fig. 5A). In contrast, aortic rings silenced for Itga6 displayed a reduced ability to make capillary-like structures: both shITGA6_48 and shITGA6_50 significantly inhibited sprout formation at days 4 and 6 after BME embedding (Fig. 5A). Blockade of integrin by the addition of anti-α6 antibody to BME gel resulted in similar levels of sprouting inhibition (Supplementary Fig. S8A). In accordance with our own in vitro observations, α6 integrin downregulation did not inhibit or delay the formation of capillary-like structures from aortic rings embedded in type I collagen gel (Fig. 5B). These results confirmed that α6 integrin was functionally involved in the process of endothelial sprouting in a ligand-dependent manner.

Because the effect of α6 integrin inhibition in EC was dependent on ECM context, we sought to investigate its role in animal models. The well-establish CAM angiogenesis model aimed to test the function-blocking antibody anti-α6 integrin during angiogenic growth factor–stimulated vessel formation. First, we looked at the expression of α6 integrin in CAM, observing elevated levels of α6 both in EC lining the internal face of blood vessels and in other sparse cells, probably epithelial cells forming the CAM sheets (Supplementary Fig. S8B). Filter discs saturated with FGF-2 were applied to CAMs and treated 24 hours later with function-blocking anti-α6β1 antibody. FGF-2 promoted vessel-branching and remodeling from pre-existing blood vessels, whereas saline treatment did not induce angiogenesis (Fig. 5C). Anti-α6β1 partially, but significantly, abrogated FGF-2–induced angiogenesis whereas immunoglobulin treatment was ineffective (Fig. 5C).

Nevertheless, the observation that α6-null embryos do not display any macroscopic defects during vasculogenesis and developmental angiogenesis argues against a role for α6 integrin in mammalian angiogenesis (16). To examine the potential role of α6 in adult neovascularization, we studied the effect of anti-α6 blocking antibody in the Rip-Tag2 mouse model. We first evaluated the accessibility of the blocking antibody to pancreas islet vasculature (23). Mice injected with anti-α6 antibody showed, 10 minutes after treatment, a clear localization of this antibody in the islets vessels (Fig. 6A). The blocking antibody also reacted with the vessels of others tissues, although the detected level of anti-α6 was lower than in pancreatic islets vasculature (Fig. 6A).

To evaluate the effects of anti-α6 blocking antibody on the angiogenic switch, we initiated the treatment of animals at 8 weeks of age (the beginning of the angiogenic stage) and continued treatment for 2 weeks. We treated the animals with different concentrations of antibody and we found that the dose with the lowest toxicity and maximum effect was 0.125 mg/mouse. Although in mock-treated animals, the angiogenic switch resulted in a higher number of hypervascularized islets, α6 integrin blockade produced a significant reduction (∼40%) in the number of angiogenic islets (Fig. 6B). Histologic analysis of tumors showed that integrin α6-blocking treatment significantly reduced vessel density, as shown by anti-VEGFR2 immunostaining (Fig. 6C). Interestingly, the vessel diameter of treated mice was reduced compared with control mice, suggesting that α6 integrin could be involved in the vascular maturation process (Fig. 6C). To address one possible explanation for this observation, we double-stained the pancreatic islets with antibodies to ECs and pericytes showing that endothelium coverage by pericytes was dramatically increased by anti-α6 integrin treatment (Fig. 6D; Supplementary Fig. S9). Taken together, these observations suggest that α6 integrin promotes both physiologic and pathologic angiogenesis.

Tumor cells, macrophages, and fibroblasts within tumors could secrete factors such as VEGF-A and FGF-2, which induces blood vessel growth in tumors (24, 25). These growth factors activate or upregulate the expression of vascular integrins such as α1β1, α2β1, α4β1, α5β1, and αvβ3 (18), which in turn promote EC migration and survival during invasion of tumor tissue. Although integrins binding to provisional ECM have been extensively studied, the role of integrins that bind components of vascular BM such as laminins is less clear. In fact, laminin-binding integrins such as α6β1 and α3β1 are considered important for the process of endothelial tube stabilization, and their role in regulating sprouting angiogenesis is uncertain (6). Here, we show that α6 integrin is upregulated by angiogenic factors in EC and is strongly expressed in the angiogenic vasculature. Although it is known that α6 integrin is expressed in human and mice endothelium, there has been no evidence thus far of its upregulation during angiogenesis (15, 26). It is likely that this depends on the broad expression of α6 integrin in different tissues, including most epithelial and carcinoma cells. We overcame this limitation using the Rip-Tag2 tumor model, which is negative for α6 integrin expression in tumor cells. This transgenic mouse develops pancreatic islet tumors through specific stages of neoplastic progression and is characterized by an angiogenic switch activated in premalignant stages and concomitant with tumor growth. mRNA expression analysis of isolated islets shows that α6 is strongly induced during the angiogenic stage and its modulation is comparable to that of α5 and β3 integrins, which were previously reported to be upregulated on angiogenic endothelium in both this model and in human tumors (23).

In contrast to previously reported observations in brain microvascular EC which showed that α6 integrin was solely induced by VEGF-A (27), we show that other angiogenic factors stimulate α6 expression, suggesting that high levels of α6 integrin on tumor vessels might represent one of the early changes that occur on the vascular cell surface during angiogenesis. In fact, in EC, in which α6 integrin is downregulated, defects in the ability to migrate and form tubular structures on laminin-containing matrix are prominently displayed. However, when similar experiments were performed on ECM ligands other than laminin, the α6 integrin downregulation did not modify the EC response.

It is known that the effects of ECM on vascular cell walls differs greatly, depending on the state of the vessel and, to a lesser extent, on the vessel type. For these reasons, the role of α6 integrin has been studied in the aortic ring model. By setting up the knockdown of α6 integrin in aortic rings using a lentiviral vector strategy, we further showed that endothelial α6 was required for sprouting from the aortic ring toward BME, but not type I collagen.

It is evident that switching from quiescent endothelium to an angiogenic endothelium implies marked changes in ECM interactions in which EC are involved. Interestingly, the ECM features change along the angiogenic process from an established laminin and collagen IV–enriched to a provisional matrix, mainly constituted by collagen I, vitronectin, and fibronectin (28). It is likely that modulation of α6 integrin levels could play a role in this switch. The high levels of α6 integrin could allow EC to invade BM, probably counteracting the effect of “stabilizing” integrins, whereas invading EC, which interact with collagen, are not dependent on this integrin. Therefore, we propose that α6 integrin is required only in the early phases of EC sprouting from a mature vessel.

However, the phenotype of α6 integrin knockout mice does not support the proangiogenic role of this integrin. Actually, mice deficient in α6 show no obvious defects in developmental angiogenesis, but they have not been extensively tested for other forms of angiogenesis. In contrast, our results provide evidence that α6 integrin plays an unexpected role in adult pathologic angiogenesis, showing that blockade of α6 integrin inhibits vessel formation both in CAM and in mammalian tumor transgenic models. These conflicting results could be explained by the observation that α6 levels are low in the embryo vasculature and increase during vessel maturation (26).

However, treatment with blocking antibody does not completely inhibit the FGF-2 and tumor-induced formation of blood vessels but seems likely to “normalize” the angiogenic vasculature. The effects on the cancer vasculature, which we noticed after blockade of α6 integrin, recall the notion of vascular normalization proposed by Jain (29), whereby antiangiogenic drugs prune, remodel, and increase pericyte coverage of otherwise abnormal tumor vessels, which in turn, become more efficient in blood flow and consequently in delivering cytotoxic drugs and oxygen for radiotherapy. The reduced tumor vessel diameter, together with the relative abundance of pericytes in mice treated with anti-α6 integrin antibody, strongly suggests that the increase of α6 on EC contributes to the morphologic and functional defects described in tumoral vessels. Although it is likely that α6 is involved in the process of EC migration and invasion, as shown by in vitro experiments, its role in the mechanism of vessel stabilization by pericytes is less clear.

Notably, we showed that α6 integrin on angiogenic vessels is easily accessible, similar to α5 integrins (23), and a very low dosage regimen is effective in normalizing the pathologic vasculature.

In conclusion, our results show that α6 integrin is expressed on angiogenic ECs in both culture and mouse tumors, and it plays an important role in vascular sprouting and tumoral angiogenesis.

No potential conflicts of interest were disclosed.

We thank Drs. Guido Serini and Daniela Taverna for critically reading the manuscript, and Dr. Arthur M. Mercurio (University of Massachusetts Medical School, Worcester, MA) for providing reagents.

Grant Support: Italian Association for Cancer Research (F. Bussolino, E. Giraudo, and L. Primo), Regione Piemonte (Finalized Health Research 2006, 2008, 2008bis, and 2009; Industrial Research and Precompetitive Development 2006: grants PRESTO and SPLASERBA; Technological Platforms for Biotechnology: grant DRUIDI; Converging Technologies: grant PHOENICS; Industrial Research 2009: grant BANP; F. Bussolino and L. Primo). CRT Foundation (F. Bussolino) and Italian Ministry of Health (Oncological Research Program 2006; Finalized Research 2006; F. Bussolino and E. Giraudo); L. di Blasio was supported by a Italian Foundation for Cancer Research fellowship.

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.