Background
Endothelial cells (ECs) form new blood vessels by migration of collective sprouts of cells that maintain cell-cell junctions [
1]. Vascular sprouts are guided by a “pathfinder” tip cell that responds to environment guidance cues, thereby determining vascular patterning [
2]. Single mature ECs are believed to migrate by mesenchymal type of motility [
3]. In 3D matrices, such motility implies an elongated spindle-like shape of the cell body whose translocation requires the formation of actin-rich lamellipodia and filopodia at the leading edge of the EC: this process is driven by the small GTPases of the Rho family, Rac for lamellipodia and CDC42 for filopodia [
4]. Both the leading and trailing edges of the EC establish adhesive interactions with the extracellular matrix (ECM), that serve as attachments for the actin stress fibers to generate forces required to translocate the trailing edge in the direction of the cell movement [
5]. Mesenchymal motility is characterized by the activity of membrane-associated proteases: integrins give rise to focal adhesions that recruit proteases thus opening a new path to invading tip cells [
3,
6].
The protease-independent amoeboid migration (named after the motility of the amoeba
Dictyostelium Discoideum) is characterized by fast cycles of contraction and expansion of the cell body, which is round or ellipsoid, obtained by contraction of the cortical actin and myosin filaments with the creation of cell “blebs” [
7]. This type of movement, observed also in hematopoietic stem cells and certain tumor cell [
1,
8], consists in a sort of “crawling” through less dense compartments of the ECM, driven by short-lived weak interaction of the amoeboid cell with the substrate. The enhanced contractility that enables cells that use the amoeboid strategy to squeeze into gaps of the ECM is promoted by the Rho/ROCK signaling pathway [
9].
In tumor cells the amoeboid and mesenchymal type of movement are interchangeable, thus defining the mesenchymal-amoeboid transition (MAT) and the amoeboid-mesenchymal transition (AMT) respectively, that represent a rapid response of cancer cells to microenvironment properties [
1,
10].
While mature ECs have never been shown to exhibit an amoeboid behavior, endothelial progenitor cells (EPCs) leave bone marrow as amoeboid cells characterized by a roundish shape and a cortical actin cytoskeleton lacking stress fibers [
11]. Of course, EPCs need to migrate through the blood vessel basement membrane and through ECM to home to sites where there is the need to form new vessels. At this step integrins provide a “grip” to EPC migration and proteases open the way across anatomical barriers to cells migrating toward the chemotactic source of pro-angiogenic factors [
12]. However, this observation is circumstantial and unable to explain plasticity of EPCs in terms of shifting from mesenchymal to amoeboid movement and viceversa.
Here we show that both mature ECs and endothelial colony forming cells (ECFCs), an EPC subpopulation with robust clonal proliferative potential and the ability to form de novo vessels in vivo, are able to migrate by mesenchymal and amoeboid style in vitro and in vivo. For this purpose we have used a mixture of physiologic inhibitors of serine-proteases, metallo-proteases and cysteine-proteases, thus mimicking a possible alternative physiological environment that ECs and ECFCs may encounter during their migration within the angiogenesis sites. As previously shown for malignant melanoma and prostate cancer cells [
13], the receptor of the urokinase-plasminogen-activator (uPAR, CD87), is indispensable for ECs and ECFCs to perform an efficient amoeboid angiogenesis, in terms of cell migration and capillary morphogenesis in vitro and in vessel formation within Matrigel plugs in mice.
Methods
Cell lines and culture conditions
Endothelial Colony Forming Cells (ECFCs) were isolated from > 50 ml human umbilical cord blood (UCB) of healthy newborns, as described previously [
14,
15], after maternal informed consent and in compliance with Italian legislation, and analyzed for the expression of surface antigens (CD45, CD34, CD31, CD105, ULEX, vWF, KDR, uPAR) by flow-cytometry [
14]. ECFCs were grown in EGM-2 culture medium (Lonza), supplemented with 10% FBS (Euroclone) onto gelatin coated dishes. Human microvascular endothelial cells (HMVEC) were purchased from Lonza and were grown in the same conditions of ECFCs.
In vitro capillary morphogenesis
In vitro capillary morphogenesis was performed as described [
14,
15] in tissue culture wells coated with Matrigel (BD Biosciences). ECFCs or HMVECs were plated (18 × 10
3 /well) in EGM-2 medium, supplemented with 2% FCS and incubated at 37 °C-5% CO2. Results were quantified at 6 h by measuring the branching points of capillary projections. Six to nine photographic fields from three plates were scanned for each point. Results were expressed as % increase/decrease of branching points/field ± SD with respect to control fixed at 100%.
Speed of capillary structures formation was measured by time-lapse capillary morphogenesis assay (Additional file
1).
3D-invasion assay with Boyden chambers
Invasion was studied in Boyden chambers in which the upper and lower wells were separated by 8 μm–pore size polycarbonate filters coated with Matrigel (BD Biosciences), as previously described [
13]. For details, see (Additional file
1).
Induction of the amoeboid phenotype and cell viability assay
Protease-indipendent angiogenic properties and invasion were evaluated by in vitro capillary morphogenesis and 3D-Boyden chamber assays, as described above, with Matrigel coating in the presence of a physiological protease inhibitor cocktail, consisting of α2-antiplasmin (plasmin inhibitor; 5 μg/ml), Cystatin (cysteine protease inhibitor; 5 μM), PAI 1 (plasminogen activator inhibitor; 10 ng/ml), TIMP1, TIMP2 and TIMP3 (metallo-protease inhibitors; 0,5 μg/ml each one) purchased by Abcam. A completely artificial protease inhibitor cocktail (composed by Ilomastat, leupeptin, pepstatin A, E-64 and aprotinin; Sigma Aldrich) used in a previous study [
13] was also used. Concentrations used were selected according to the manufacturer’s instructions and literature [
16,
17]. Protease inhibitor cocktails were added to un-polymerized Matrigel solution on the upper surface of the porous filter. To induce the amoeboid phenotype, cells were treated overnight with the protease inhibitor cocktails at the same concentrations used in the invasion assay. Cell viability upon protease inhibitors treatment was evaluated by Trypan blue dye (Sigma) exclusion assay.
Collagen degradation assay
ECFC and HMVEC cell suspensions were co-polymerized with Matrigel containing 2% FITC-labeled collagen monomers (Molecular Probes). Digestion was allowed for 40 h at 37 °C and solid-phase Matrigel containing the cells was pelleted, whereas FITC released into the supernatant was analyzed by spectrofluorometry. One hundred percent values were obtained by complete collagenase digestion of cell-free Matrigel lattices. Background fluorescence was obtained by pelleting non-digested cell-free FITC-collagen-enriched Matrigel layers.
RhoA and Rac1 activity assay
Cells from different experimental conditions (control, physiological protease inhibitor cocktail) were lysed in radio-immunoprecipitation assay buffer, the lysates were clarified by centrifugation, and RhoA GTP or Rac1 GTP was quantified. Briefly, lysates were incubated with 10 μg rhotekin–glutathione S-transferase (GST) fusion protein (Millipore) or p21-activated kinase-GST fusion protein, both absorbed on glutathione–Sepharose beads for 1 h at 4 °C. Ratios between activated (GTP-bound) RhoA and Rac1 adsorbed to the beads were quantified by Western blot densitometry.
Western blotting
Specific electrophoretic conditions and the source of used antibodies are reported Additional file
1.
Semiquantitative reverse transcription–polymerase chain reaction (PCR) analysis
Total RNA preparation and reverse transcription were performed as previously reported. The levels of messenger RNA for the integrin chains were determined by an internal-based semiquantitative RT-PCR, using procedures and primers previously described [
18].
Treatment of cells with M25 peptide
Inhibition of uPAR-integrin interaction was obtained with the M25 peptide, previously identified in a phage display library [
19], able to uncouple uPAR interaction with integrin α-chain. The peptide was produced in collaboration with PRIMM srl, Milan, Italy. In the β-propeller model of α-chain folding, the sequence of this peptide (STYHHLSLGYMYTLN) spans an exposed loop on the ligand-binding surface of α-chain, thus impairing integrin α chain-uPAR interaction. In cell culture both M25 and scramble-M25 (sM25) were used at 50 μM for 15 h at 37 °C.
Co-immunoprecipitation
For co-immunoprecipitation, ECFCs and HMVECs were plated at 500 × 10
3 cells/100 mm dish in complete medium. One of two dishes for each cell line was treated with M25 peptide at 50 μM for 15 h. After two washes in ice-cold PBS, cells were lysed on ice with Ripa buffer, centrifuged (15,000 rpm, 15 min) and the supernatant was used for co-immunoprecipitation. For details, see Additional file
1.
Immunofluorescence analysis
Immunofluorescence and confocal microscopy were performed as previously described [
18]. For details, see Additional file
1.
siRNA uPAR knock-down and quantitative real-time PCR analysis
See details under Additional file
1.
In vivo Matrigel plug assay
All procedures involving animals were performed in accordance with the ethical standards and according to the Declaration of Helsinki and to national guidelines approved by the ethical committee of Animal Welfare Office of Italian Health Ministry and conformed to the legal mandates and Italian guidelines for the care and maintenance of laboratory animals.
Two groups of 8 and 10 four-week-old male SCID beige mice (two for each experimental condition) were purchased from Charles River. The Matrigel plug assay was used to study a possible role of the amoeboid movement in vivo as previously described [
14]. VEGFA (10 ng/ml), was added to unpolymerized Matrigel at 4 °C at a final volume of 0.6 ml. Protease inhibitors were added to unpolymerized Matrigel at the same concentrations used for in vitro assays. Heparin (50 U/ml) was added to each solution. To study the role of mice vessels, the Matrigel suspension was carefully injected subcutaneously into both flanks of mice using a cold syringe. As the Matrigel warms to body temperature, it polymerizes to a solid gel, which then becomes vascularized within five days in response to the angiogenic substance. The extent of vascularization was quantified by measuring the hemoglobin content of the recovered plugs. Groups of four pellets were injected for each treatment. The eight animals were subdivided as follows: two controls (Matrigel alone); two animals injected with a Matrigel plug containing VEGF-A; two injected with Matrigel plus the physiologic protease-inhibitor cocktail; and two with Matrigel, the physiologic protease-inhibitor cocktail plus VEGFA. The reagents for all treatments were added to the Matrigel solution prior to injection. Five days after injection, the pellets were removed, minced and diluted in water to measure the hemoglobin content with a Drabkin reagent kit (Sigma). Vascularization was evaluated by sight taking a representative photograph of individual Matrigel plugs recovered at autopsy for the corresponding condition. Samples were also fixed in formalin, embedded in paraffin for histological analysis and stained with hematoxylin-eosin.
The second group of 10 mice was used to verify the role of uPAR in amoeboid angiogenesis in vivo. Before implantation we added to the plugs, composed of Matrigel plus physiologic protease-inhibitor cocktail, murine uPAR-aODN (ISIS Pharmaceuticals, Carlsbad Research Center, California) or M25 peptide used in vitro. At the third day, another dose of ODNs and M25 was injected subcutaneously into the plugs. Treatments consisted in plug administration of liposome-encapsulated vehicle alone (DOTAP; Roche, Germany), DOTAP + scramble ODN, DOTAP + uPAR-aODN, scramble M25 peptide and M25 peptide. Upon sacrifice (fifth day), isolated plugs were stained with hematoxylin-eosin.
Statistical analysis
Unless otherwise stated, all the experiments were performed five times in duplicate for a reliable application of statistics. Statistical analysis was performed with GraphPad Prism5 software. Results are expressed as means ± SD. Multiple comparisons were performed by Anova and paired Student T test. Statistical significances were accepted at p < 0.05. (*p < 0,005, **p < 0,001, ***p < 0,0001).
Discussion
We have shown that ECs and ECFCs enter into a so far undescribed program of vessel formation, that we call amoeboid angiogenesis, when protease inhibitors overwhelm their proteolytic potential.
An interesting paper by Stratman et al. [
22] reported that in 3D matrices ECs exploit MT1-MMP proteolytic activity to excavate vascular guidance tunnels during tube morphogenesis. Afterwards, ECs migrate within such tunnels even upon inhibition of MMP activity. However, a possible amoeboid movement of ECFCs has been suggested only on the basis of cell morphology [
1,
11,
29,
30]. A recent paper [
31] has described an amoeboid phase of endothelial cells engaged in the endothelial to mesenchymal transition in the frame of Systemic Sclerosis-associated fibrosis.
In a previous study we have shown that, in malignant melanoma and prostate cancer, the receptor of the urokinase-plasminogen-activator (uPAR, CD87) is indispensable for supporting a shift from a mesenchymal to an amoeboid movement [
13]: uPAR regulates cortical actin contraction/relaxation cycles by interaction with beta1 and beta3 integrins, that in turn connect uPAR to the actin in a RGD-independent fashion. This happens even in the presence of a full-range of synthetic proteases-inhibitor cocktail.
Here we show that both mature ECs and ECFCs can migrate by mesenchymal and amoeboid style. For this purpose, we used a mixture of physiologic inhibitors of serine-proteases, metallo-proteases and cysteine-proteases, thus mimicking a possible physiological environment that ECs and ECFCs may encounter during their migration within tissues. This study shows that, under the effect of the physiological inhibitor-MIX, both ECFCs and HMVECs performed a greatly enhanced Matrigel invasion, producing tube-like structure as in mesenchymal conditions. The protease inhibitor-MIX used in this study is composed by molecules that individually affect a broad range of substrates, including molecules other than proteases [
12,
32‐
35]. Here we demonstrated that the biological activity of the single inhibitors produced no or scarce decrease of cell invasion as compared to the intense invasion-promoting activity of the full-range cocktail. This is possibly related to the described cross-talks among the main families of proteases, whereby both serine-proteases and cathepsins may activate pro-uPA to uPA [
36], and plasmin is involved in proteolytic activation of pro-MMP1 [
37], pro-MMP3 [
38], pro-MMP9 [
39]. Therefore, we additionally demonstrated that the effect of the mix was due to the synergistic effect of all inhibitors mixed together and not to any biological activity of a single one. Moreover, a large body of literature indicates that each inhibitor exhibited protease inhibition-independent activities that often resulted into angiogenesis inhibition, and never in its stimulation. These observations indicate that the endothelial cell amoeboid invasion and angiogenesis are an environment-dependent escape mechanism.
We also had in vivo evidence of amoeboid angiogenesis in the Matrigel plug Assay confirming that the inhibitor-MIX stimulated the in vivo angiogenesis showing an increased number of vascular structures with lumens and red blood cells. We have been unable to avoid a hemorrhagic effect under amoeboid conditions in the presence of VEGF. We speculated this is related to the VEGF effect as vascular permeability factor acting on the tumultuous and fast angiogenesis dictated by amoeboid conditions.
These findings indicate the existence of a new type of neovascularization: the “amoeboid angiogenesis”. uPAR silencing with aODN and uPAR-integrin uncoupling with the M25 peptide abolished both mesenchymal and amoeboid invasion and angiogenesis of ECs and ECFCs in vitro and amoeboid murine angiogenesis in vivo, indicating a role of the uPAR-integrin-actin axis in the regulation of amoeboid angiogenesis.
However, since uPAR silencing either in melanoma and prostate cancer cells [
13] or in ECs and ECFCs is unable to exhaustively inhibit amoeboid movement, we speculate on a possible role of other protease receptors (in particular MMPs receptors), in the light of the recent observation indicating that also MMP9 regulates amoeboid migration of melanoma cells in a catalytic independent manner through regulation of actomyosin contractility via its CD44 receptor [
26].
Confocal immuno-fluorescence analysis of integrin αvβ3 and uPAR showed that uPAR-integrins interactions persist under both mesenchymal and amoeboid conditions. Treatment of cells with 50 μM peptide M25 uncoupled uPAR from integrins, in the absence and in the presence of the inhibitor cocktail demonstrating that in ECFCs and ECs, uPAR signaling is mediated by αvβ3 integrins also in amoeboid conditions.
The targeting of vascular endothelial growth factor A (VEGFA), a crucial regulator of both normal and pathological angiogenesis, resulted in innovative therapeutic approaches in oncology and ophthalmology [
40,
41] and, in combination with chemotherapy and radiation, is able to correct leaking vessels, to decrease tumor interstitial pressure, and to inhibit vessels development. Although VEGF-targeted therapies currently are standard of therapy for multiple tumor types, many patients develop resistance and progress toward metastasis. In this context, targeting both VEGF in parallel with other pathways implicated in angiogenesis should result into more effective tumor growth inhibition. Here we tested VEGF stimulation under amoeboid conditions and observed an “indifference” of ECs and ECFCs to VEGF treatment. Maybe it could be ascribed to the endothelial cells capacity to migrate and differentiate into tubular structures at the maximum levels under amoeboid conditions thus justifying the limited efficacy of VEGF-targeted therapies as detailed above. Furthermore, surprisingly, our results reveal that VEGF stimulation induce an amoeboid-like signaling pattern in both mesenchymal and amoeboid conditions.
Synthetic metalloproteinase inhibitors (MPIs) were developed and utilized in human clinical trials but the results were disappointing [
42]. MPIs are now viewed only as potential antiangiogenic agents for primary tumors [
43] and as a therapy able to maintain small clusters of metastatic cells in a dormant state. In light of what we have shown up in this study, we hypothesized that the failure of treatment after the initial stages of tumor development could be ascribed also to the onset of the angiogenic transition, during which the tumor microenvironment is able to skip the attack of the MPI therapy by allowing blood vessel formation using the “amoeboid” strategy.
So far, there are no additional explanations about the mechanisms that determine the transition from one migration mode to another, excluding the presence or absence of protease inhibitors. It is known that the so-called cell “protease web” (both membrane-associated and released into the ECM), reaches a high level of complexity, involving the whole range of proteases and about 140 substrates, many of which deal with angiogenesis [
44]. In particular, MMPs may release from ECM many biologically active protein fragments and the main growth factors involved in angiogenesis, VEGF, FGF2, TGF-β, that pave the way to angiogenic endothelial cells, also providing a chemotactic gradient for the growing vessel. All such factors induce actin stress fibers organization and the release of pro-angiogenic proteases, features that define a “mesenchymal” mode of endothelial cell invasion [
45‐
47]. The presence of a full-range protease inhibitor cocktail blocks availability of ECM-trapped angiogenic molecules, an event that triggers the amoeboid migration, an ancestral escape mechanism of movement ranging from amoebae to vertebrates [
48].
Supporting our thesis, recent studies demonstrate that TIMP family members (TIMP1, TIMP2, TIMP3), as well as other physiological protease inhibitors (PAI-1, alpha2-antiplasmin, cystatin), are regarded as negative prognostic factors in patients and in experimental animals showing increased plasma and intra-tumor concentrations [
49‐
55]. Even the assumption that the natural protease inhibitors could exhibit an anti-metastatic effect had been challenged in the past few years. For example, despite its function as an inhibitor of urokinase and tissue-type plasminogen activator (PA), PA inhibitor-1 (PAI-1) has a paradoxical pro-tumorigenic role in cancer, promoting angiogenesis and tumor cell survival and, as a biomarker in breast cancer, is validated for prognostic use in level-of-evidence-1 studies [
35]. Our results, along with the results obtained in other studies [
56,
57], show that uPAR is a molecular mediator of plasticity in angiogenesis, in concert with integrins [
58]. Our results, along with the results obtained in other studies [
56,
57], show that uPAR is an indispensable molecular mediator of plasticity in angiogenesis, in concert with integrins [
58]. Besides, Rao and coworkers [
59] demonstrated that a cooperation between uPA/uPAR and MMP-9 is required for breaching of the vascular wall, a rate-limiting step for intravasation, and consequently for tumor progression and metastasis. Therefore, uPAR silencing or the block of its interaction with integrins, together with standard treatment against VEGF, could be a possible therapeutic strategy impairing vascular growth and cancer cell invasion at the same time, overcoming resistance to anti-VEGF and anti-protease therapy.