Abstract
Integrin β3 is seen as a key anti-angiogenic target for cancer treatment due to its expression on neovasculature, but the role it plays in the process is complex; whether it is pro- or anti-angiogenic depends on the context in which it is expressed. To understand precisely β3’s role in regulating integrin adhesion complexes in endothelial cells, we characterised, by mass spectrometry, the β3- dependent adhesome. We show that depletion of β3-integrin in this cell type leads to changes in microtubule behaviour that control their migration. β3- integrin regulates microtubule stability in endothelial cells through Rcc2/Anxa2 driven control of Rac1 activity. Our findings reveal that angiogenic processes, both in vitro and in vivo, are more sensitive to microtubule targeting agents when β3-integrin levels are reduced.
INTRODUCTION
Angiogenesis, the formation of new blood vessels from those that already exist, plays an essential role in tumour growth (Hanahan and Weinberg, 2011). As such, targeting angiogenesis is seen as crucial in many anti-cancer strategies (Zhao and Adjei, 2015). Therapies directed against vascular endothelial growth factor (VEGF) and its major receptor, VEGF-receptor-2 (VEGFR2), whilst effective in a number of cancers, are not without side-effects due to the role this signaling pathway plays in vascular homeostasis (Chen and Hung, 2013). Fibronectin (FN)-binding endothelial integrins, especially αvβ3- and α5β1-integrins, have emerged as alternative anti-angiogenic targets because of their expression in neovasculature (Brooks et al., 1994; Kim et al., 2000). However, neither global nor conditional knockouts of these integrins block tumour angiogenesis long-term (Murphy et al., 2015; Reynolds et al., 2002; Steri et al., 2014), and clinical trials of blocking antibodies and peptides directed against these extracellular matrix (ECM) receptors have been disappointing (Schaffner et al., 2013; Stupp et al., 2014).
To gain novel insight into how αvβ3-integrin regulates outside-in signal transmission (Hynes, 2002), we have undertaken an unbiased analysis of the molecular composition of the mature endothelial adhesome (the network of structural and signaling proteins involved in regulating cell-matrix adhesions (Zaidel-Bar et al., 2007)), and profiled changes that occur when β3-integrin function or expression are manipulated. In so doing, we have uncovered β3-integrin dependent changes in microtubule behaviour that regulate cell migration.
RESULTS AND DISCUSSION
The isolation and analysis of integrin adhesion complexes (IACs) by mass-spectrometry (MS) is difficult because of the low affinity and transient nature of the molecular interactions occurring at these sites. However, using cell-permeant chemical crosslinkers improves recovery of IAC proteins bound to either FN-coated microbeads (Horton et al., 2015) or plastic dishes(Schiller et al., 2011). These advances have led to the characterisation of IACs from a number of cell types. Whilst a core consensus adhesome (the network of structural and signaling proteins involved in regulating cell-matrix adhesion (Zaidel-Bar et al., 2007)) can be defined (Horton et al., 2015), the composition and stoichiometry of the meta-adhesome depends on the cell-type being analysed, the integrin-receptor repertoire expressed by that cell type, and on any imposed experimental conditions.
To examine the composition of the endothelial adhesome we isolated lung microvascular endothelial cells (ECs) from C57BL6/129Sv mixed background mice and immortalised them with polyoma-middle-T-antigen by retroviral transduction (May et al., 2005). As our main interest was in establishing how β3-integrin influences the endothelial adhesome, we adhered cells to FN for 90 minutes, which allows β3-rich (mature) focal adhesions (FAs) to form (Schiller et al., 2013). To distinguish integrin-mediated recruitment of proteins from non-specific background, we also plated cells on poly-L-lysine (PLL) as a negative control (adhesion to PLL does not depend on integrins). Visualisation of neuropilin-1 staining in whole cells showed that this protein, which we previously demonstrated is present in the mature EC adhesome (Ellison et al., 2015), co-localises with talin-1 in FAs when cells are plated on FN, but not PLL (Fig. 1a). For all proteomics experiments, we crosslinked FAs using the cell permeant and reversible cross-linkers DPDPB and DSP (see materials and methods) for 5 minutes. Cells were lysed and subjected to a high sheer flow water wash to remove non-crosslinked material. Crosslinking was reversed, and samples were precipitated and eluted for analyses. Prior to MS, samples were quality controlled by SDS/PAGE and silver-staining to ensure efficient removal of non-crosslinked material had occurred (Fig. 1b).
Label-free proteomic analyses of the FN + VEGF, FN, and PLL adhesomes (Fig. 1c; Supplementary Table SI) initially detected and quantified 1468 proteins. Stringent filtering, requiring proteins to be detected in all 3 repeats of at least one condition, left 1064 proteins - a high confidence dataset that was used to define the endothelial adhesome. Hierarchical clustering based on average Euclidian distance identified 12 clusters (A-L) which could be considered VEGF-enriched proteins (A-C), FN-enriched (D-F), and PLL-enriched (G-L). Fisher’s exact test enrichment analysis was carried out to identify which pathway, process, or component proteins within these clusters belong to using Gene Ontology annotations. Cell projection (GOCC, p=8.62 × 10−5) and microtubule (GOCC, p=1.6 × 10−4) categories were significantly enriched when cells were treated with growth factor, suggesting they are important in VEGF-mediated processes. Leukocyte trans-endothelial migration (KEGG, p=9.71 × 10−5) proteins were enriched in the FN adhesome, but not in the VEGF-stimulated adhesome, suggesting our cells represent quiescent vasculature without VEGF-stimulation. This same category contains many endothelial specific proteins (e.g. ve-cadherin/Cdh5), further confirming that the cells have an endothelial identity. Focal adhesion (KEGG, p=9.31x10−7) proteins were enriched in the FN adhesome but depleted in the PLL adhesome, confirming the success of the adhesome enrichment process, MS, and downstream analysis. Other adhesion/migration associated categories: focal adhesion (GOCC, p=5.99 × 10−5), cell projection (GOCC, p=3.03 × 10−5), cell adhesion (GOBP, p=1.61 × 10−6) and lamellipodium (GOCC, p=1.38 × 10−4) were depleted in the PLL adhesome.
FN-based matrices are essential for angiogenesis (Zhou et al., 2008) and studies in fibroblasts have demonstrated that av-integrins and α5β1-integrin cooperate to direct cell migration on FN-based matrices(Schiller et al., 2013). Therefore, we wanted to profile changes in the FN-endothelial adhesome in response to αvβ3-integrin blockade. Thus, we first determined the adhesome composition of ECs treated with EMD66203 (cyclo-Arg-Gly-Asp-DPhe-Val), a conformationally constrained cyclic pentapeptide that selectively targets αvβ3-integrin (Pfaff et al., 1994). EMD66203 inhibited EC adhesion to FN (the effects of EMD66203 on adhesion to vitronectin, an αvβ3 specific ligand, was tested to ensure activity of the compound - Fig. 2a) but had no dramatic effect on the endothelial adhesome (Fig. 2b; Supplementary Table S2), suggesting that: (1) even if αvβ3-integrin is unable to bind ECM, it can localise to FAs and participate in adhesome assembly in the presence of EMD66203; or (2) under the experimental conditions used, only a small percentage of αvβ3-integrin is conformationally available for EMD66203 binding (Demircioglu and Hodivala-Dilke, 2016).
To test the consequences of excluding β3-integrin more efficiently from the EC adhesome, we decided to profile changes in β3-heterozygous (β3HET) ECs, which carry one wild-type allele of β3-integrin, and one knockout allele. These cells express 50% wild-type levels of β3-integrin and we have shown they are a good model for studying the role of αvβ3-integrin in cell migration, whilst evading changes arising from the complete loss of the integrin on both alleles (e.g. altered VEGFR2-mediated responses) (Ellison et al., 2015). Both wild-type (β3WT) and β3HET ECs adhere equally to saturating concentrations (10 μg/ml) of FN (see Ellison et al., 2015 (Ellison et al., 2015)). To compare the size distribution of FAs between β3WT and β3HET ECs (which might affect the stoichiometry of components in the adhesome), we seeded cells for 90 minutes on FN, immunostained for paxillin, and measured FA area; we noted no differences in the percentage of FA size distributions between the two genotypes (Fig. 2c). Therefore, MS analyses comparing the adhesome between β3WT and β3HET ECs were performed (Fig. 2d; Supplementary Table S3). Enrichment analysis showed a depletion of cytoskeletal components (GOCC, p = 4.73 × 10−5) in the β3WT adhesome when compared with the β3HET adhesome, despite the enrichment of adhesion/migration associated categories previously noted in the FN adhesome of β3WT ECs (Fig. 1c). Whilst a majority of individual FA components in the mature adhesome do not change upon β3-integrin depletion, downstream connections to cytoskeletal components do. We took a particular interest in microtubules (MTs) because all detected tubulins were significantly upregulated in the β3HET adhesome. To confirm this finding by other means, we probed Western blots for α-tubulin and showed a significant increase in FA-enriched samples from β3HET cells compared with β3WT cells (Fig. 2e).
Our findings intimated that αvβ3-integrin drives MT localisation away from FAs. To increase the power of our studies, we included β3-integrin knockout (β3NULL) ECs in subsequent analyses. We examined MT organisation in β3WT, β3HET, and β3NULL ECs by immunolabeling for α-tubulin in whole cells (Fig. 3a). Although no gross changes in cell microtubule arrays were observed, the increased bundling of microtubules in the β3HET and β3NULL ECs suggests they could be crosslinked by microtubule stabilising proteins. Microtubule bundles are known to form in ECs to assist with directional migration by counteracting actomyosin generated contractile forces; these bundles often target adhesion sites in EC protrusions (Lyle et al., 2012). Furthermore, total cellular levels of α-tubulin were similar across the three genotypes (Fig. 3b). However, colocalisation of MTs with talin-1 at peripheral FAs was greater in β3HET and β3NULL ECs as was extension into lamellipodea (Fig. 3c). Overall, the findings suggest that β3-integrin limits the targeting of MTs to FAs.
Given that MTs can drive FA turnover, and thus cell migration(Kaverina and Straube, 2011), we next tested whether EC migration is differentially sensitive to microtubule targeting agents (MTAs) in β3HET and β3NULL ECs. For each MTA examined, we first determined the dose of the compound that allowed 90 percent survival of β3WT ECs (not shown), and then tested the effects of this dose on random migration in β3WT, β3HET, and β3NULL cells (Fig. 3d). Random migration was affected by MT stabilisers (Paclitaxel, Epothilone B) in cells of all three genotypes, although β3WT cells were generally less sensitive than β3HET and β3NULL cells. β3WT ECs were insensitive to MT destabilisers (Colchicine, Mebendazole, Fosbretabulin) and the mechanistically unique MTA Eribulin (which functions through an end poisoning mechanism (Jordan et al., 2005)), whilst β3HET and β3NULL ECs showed a greater sensitivity to these compounds; in general, β3NULL cells were more sensitive than β3HET cells. We extended these types of analyses in vivo to examine the effects of Eribulin on tumour growth and angiogenesis. We chose Eribulin because the agent is well tolerated in mice (Su et al., 2016). We settled on a suboptimal dose (0.15 mg/kg) that would allow us to observe potential synergy with endothelial depletion of β3- integrin. β3-integrin-floxed/floxed mice (Morgan et al., 2010) were bred with Tie1Cre mice (Gustafsson et al., 2001) to generate β3-integrin-floxed/floxed Cre-positive animals (Cre-negative littermates were used as controls). CMT19T lung carcinoma cells were injected subcutaneously and allowed to establish for 7 days, at which point Eribulin was administered i.v. One week later, a second dose of Eribulin was given, and then tumours were harvested one week later (see Fig. 3e for dosing regime). Eribulin had no effect on tumour growth in Cre-negative animals compared with vehicle treated animals, but tumour growth was significantly reduced in Eribulin-treated Cre-positive animals (Fig. 3e). Analysis of tumours by immunostaining for blood vessels showed a reduction in intratumoral microvascular density only in sections from Eribulin-treated Cre-positive animals (Fig. 3f). There have been reports that targeting β3-integrin increases the efficiency of drug delivery to tumours (Wong et al., 2015). To test whether this might be driving increased responses in Cre-positive animals, we conducted a similar experiment with doxorubicin, a DNA damaging agent. In these studies, we observed no difference in tumour growth or vessel density when comparing Cre-positive to Cre-negative animals (Fig. 3f). On the whole, the findings suggest β3- integrin depletion in ECs renders them more sensitive to MTAs both in vitro and in vivo.
The increased sensitivity to destabilising MTAs suggested to us that there is an increased population of stable MTs in β3HET and β3NULL ECs compared with their wild-type counterparts. We explored this premise by exposing ECs to cold temperatures (which destabilises MTs), washing out tubulin monomers (Ochoa et al., 2011), followed by immunolabeling for α-tubulin. We noted elevated stable MTs in both β3HET and β3NULL cells (Fig. 4a). We also measured this biochemically by separately extracting both cold-sensitive (Fig. 4b) and cold-stable (Fig. 4c) MTs from cold-treated cells and Western blotting for α-tubulin. β3HET and β3NULL ECs showed decreased cold-sensitive and increased cold-stable MTs compared with β3WT ECs.
To gain mechanistic insight into how β3-integrin at FAs might be regulating MT function, we delved deeper into our β3-dependent adhesome data. We noted that Rcc2 clusters with β3-integrin in the β3WT adhesome, but is significantly decreased in that of β3HET ECs. Rcc2 (also known as telophase disk protein of 60 kDa, TD-60) has previously been shown to associate with integrin complexes(Humphries et al., 2009) and to regulate MTs (Mollinari et al., 2003). We therefore examined whether Rcc2 was regulating MT stability in ECs. Knocking down Rcc2 by siRNA in β3WT ECs elicited a significant increase in cold-stable MTs (Fig. 5a). This finding suggested to us that Rcc2 plays a β3-dependent role in regulating MTs in ECs, but does not do so in isolation. We therefore cross-referenced our adhesome data with an Rcc2 pull-down assay performed from HEK-293T cells (Supplementary Table S4) (Williamson et al., 2014). Some obvious potential candidates (e.g. Coronin-1C) were present in both the β3WT and β3HET adhesomes, but at the same level, so were ruled out from further analysis. However, annexin-a2 (Anxa2) co-precipitates with Rcc2 in HEK-293T cells and, like Rcc2, was reduced in the β3HET adhesome. Therefore, we examined whether Anxa2 was also regulating MT stability in ECs via siRNA-mediated knockdown. Like Rcc2 knockdown, even a relatively small (~30%) Anxa2 knockdown in β3WT ECs elicited a significant increase in cold-stable MTs (Fig. 5b).
Both Rcc2 (Humphries et al., 2009; Williamson et al., 2014) and Anxa2 (Hansen et al., 2002) have been identified as regulators of Rac1, and work by a number of groups has demonstrated that cortical Rac1 activity promotes MT stability (Banerjee et al., 2002; Daub et al., 2001; Wittmann et al., 2004). Because total Rac1 stoichiometry was unchanged when comparing β3WT and β3HET EC adhesomes, we hypothesised that Rcc2/Anxa2-dependent alterations in Rac1 activity were responsible for altered MT stability in β3HET and β3NULL ECs. First, we tested the premise that Rac1 plays a differential role in regulating MT stability in β3WT and β3-depleted ECs by testing the effects of the Rac1 inhibitor NSC23766. NSC23766 had no effect on MT stability in β3WT cells, but cold stable MTs in both β3HET and β3NULL ECs were significantly reduced in the presence of the inhibitor (Fig. 5c). We next demonstrated that the increases observed in MT stability upon Rcc2 or Anxa2 knockdown were abrogated in the presence of NSC23766 (Fig. 5d), suggesting that both proteins regulate MT stability in ECs via Rac1.
Rcc2 has been reported to bind directly to the nucleotide-free form of Rac1 and it has been suggested that it functions as a Rac1 guanine nucleotide exchange factor (GEF)(Humphries et al., 2009). Indeed, Rcc2 can guide mesenchymal cell migration by trafficking Rac1 and controlling its exposure to GEFs (Williamson et al., 2014). We therefore tested whether there were differences in Rcc2/Anxa2/active-Rac1 associations between β3WT and β3-depleted ECs. PAK-PBD pull-downs of GTP-bound Rac1 showed co-association of all three proteins in β3WT, β3HET and β3NULL ECs (Fig. 5e), so we concluded that changes in Rac1 activity alone were not responsible for alterations in MT stability in β3-depleted cells. Humphries et al. showed that Rcc2 is recruited to α5β1-FN complexes but not α4β1-Vcam1 (vascular cell adhesion molecule-1) complexes in cells expressing both α4- and α5-integrins (Humphries et al., 2009). Thus, we also tested associations between Rcc2, Anxa2 and α5-integrin in β3WT and β3-depleted ECs by PAK-PBD pull-downs. Rcc2, Anxa2, β3- integrin and α5-integrin were pulled down with Rac1-GTP in β3WT ECs. Rcc2, Anxa2 were also pulled down with Rac1-GTP in β3HET and β3NULL ECS, whilst β3-integrin-Rac1-GTP associations were lost and α5-integrin-Rac1-GTP associations were increased (Fig. 5e). Given the stoichiometry of α5-integrin in the β3-depleted adhesome is unchanged compared to the β3WT adhesome (Fig. 2) whilst Rcc2 and Anxa2 levels are decreased, we speculate that a substantial proportion of the observed increase in Rcc2/Anxa2/active-Rac1/Itga5 associations in β3-depleted cells occurs away from FAs, perhaps in recycling endosomes.
In conclusion, by mining the FN-3-integrin EC adhesome, not only have we generated a valuable tool for the integrin and angiogenesis communities, we have also utilised the data to uncover a novel role for β3-integrin in regulating MT function/stability during EC migration. We previously showed that endothelial Rac1 is only required for tumour growth and angiogenesis when β3-integrin is absent (D’Amico et al., 2010), but the underlying mechanism for this observation has remained unclear. Our working hypothesis is that engagement of αvβ3-integrin with FN at mature FAs localises an Rcc2/Anxa2/Rac1 containing complex to these sites, either preventing GTP-Rac1 from participating in MT stability, or actively destabilising MTs (our experiments do not allow us to distinguish between these two possibilities), perhaps by controlling its exposure to GEFs. When αvβ3 is not present, this complex associates with α5β1-integrin instead, where it now has the opposite effect on MTs (Fig. 5f). This re-positioning of Rac1 activity means that it plays a role in MT-linked EC migration only when αvβ3 is not present in mature FAs. There is certainly precedence for β3-integrin regulating spatial distribution of signaling pathway components in cells. For example, we previously showed that β3-integrin plays a role in locally suppressing β1-integin in fibroblasts to promote persistent cell protrusion and migration by regulating interactions between vasodilator-stimulated phosphoprotein (Vasp) and Rap1-GTP-interacting adaptor molecule (Apbb1ip/RIAM) (Worth et al., 2010). Moreover, MTs have recently been shown to target active β1-integrins (Byron et al., 2015). Thus, it will be particularly pertinent to next determine the full composition of the Rcc2/Anxa2/Rac1-GTP complex as many of the proteins that might be suspected to play a role in MT capture (e.g. Clip170 and Clasps) do not appear to be present in the EC adhesome (Fukata et al., 2002); to gain a full picture of how MT stability/FA targeting are regulated in ECs it will be essential to establish how this complex behaves in α5β1-deficient ECs. Finally, whilst our findings reinforce the concept that integrin inhibitors for use as anti-angiogenic agents need rethinking, they nevertheless suggest that once effective αvβ3-integrin antagonists are available (e.g. ProAgio (Turaga et al., 2016)), they may be particularly useful as anti-angiogenic agents when used in combination with already approved MTAs, such as Eribulin.
Competing Interests
The authors declare no competing financial interests.
Author Contributions
SJA designed and performed experiments, analyzed data, and helped write and edit the manuscript. AMG, TSE, and RTJ designed and performed experiments, analyzed data, and helped edit the manuscript. BMK, AA, WJF and BCS performed experiments, analyzed data, and helped edit the manuscript. JGS, KW and MDB provided essential data and helped edit the manuscript. MMM and DRE analyzed data and helped edit the manuscript. SDR designed experiments, performed experiments, analyzed data, and wrote the manuscript.
Funding
This work was part funded by BBSRC DTP PhD studentships to SJA, RTJ, BMK, and WJF. The work was also part funded by BigC PhD studentships to TSE and AMG and by British Heart Foundation and Breast Cancer Now project grants to SDR and DRE, respectively. The work was also supported by charitable donations from Norfolk Fundraisers, Mrs Margaret Doggett, and The Colin Wright Trust.
Acknowledgements
We thank Maddy Parsons (Kings College London, London, UK) for her gift of the paxillin-GFP construct. A special thanks to both Dr Sophie Akbareian and Peng Liu for their undying enthusiastic and critical support of this project.