|Home | About | Journals | Submit | Contact Us | Français|
We have recently reported that elastin microfibril interface located protein 2 (EMILIN2), an extracellular matrix (ECM) glycoprotein, triggers cell death through a direct binding to death receptors. EMILIN2 thus influences cell viability through a mechanism that is unique for an ECM molecule. In the present work, we report an additional function for this molecule. First, we identify the region responsible for the proapoptotic effects, a 90-amino acid residue-long coiled-coil fragment toward the N-terminus of the molecule. The fragment recapitulates EMILIN2 proapoptotic mechanisms. In addition, using either the full molecule or the active fragment, for the first time, we demonstrate a significant antitumoral effect in vivo, likely due to a decrease in tumor cell viability. Unexpectedly, tumors treated with EMILIN2 or the deletion mutant display a significant increase of tumor angiogenesis. In view of this novel finding, the cotreatment of the growing tumors with an antiangiogenic drug led, in most cases, to a complete regression of tumor growth. These results grant further support to recent findings that pinpoint the microenvironment as an important regulator of cell fate under both physiological and pathological conditions and disclose the possibility of using EMILIN2 fragments as potent antineoplastic tools for cancer treatment.
Tumor growth is profoundly influenced by the neighboring cells and microenvironment [1,2]. In this scenario, tumors represent functional tissues interconnected with a large variety of cell types, blood and lymphatic vessels, embedded in a complex network of extracellular matrix (ECM) molecules composing the microenvironment, which are thought to sustain tumor cell survival, growth, and dissemination to distant organs . For this reason, in the recent years, researchers have focused the attention on ECM and cell components of the microenvironment as potential targets or tools for anticancer therapy. Endothelial cells (ECs), for instance, are promising targets because, unlike tumor cells, they are genomically stable and thus less prone to drug resistance. Judah Folkman  first foresaw the possibility to target blood vessels instead of the tumor itself with the aim of halting the nutrients and oxygen supply to the aberrantly proliferating cells. These efforts have led to the discovery of many antiangiogenetic molecules such as bevacizumab, an anti-vascular endothelial growth factor (VEGF) antibody currently used in clinics in combination with conventional chemotherapy for different types of tumors [5–9]. In addition, tumor growth depends on the cell's capability to proliferate and survive despite apoptotic stimuli. Resistance to apoptosis is indeed one of the hallmarks of cancer and is achieved through the inactivation of the finely regulated mechanisms that control this process [10,11]. Apoptosis is achieved through the initiator caspases of the so-called intrinsic and extrinsic apoptotic pathways . The extrinsic pathway is triggered through the engagement of specific receptors such as DR4 (TRAIL-R1)  and DR5 (TRAILR2)  by their respective death ligands [12,15]. Among these ligands, the tumor necrosis factor-related apoptosis-inducing ligand (TRAIL) is a promising cancer therapeutic tool owing to its selective action on neoplastic cells [16–18]. TRAIL binds to its receptors as a homotrimer-inducing death receptor clustering in lipid rafts followed by death-inducing signaling complex (DISC) assembly and caspase-8  and caspase-10 activation .
Conversely, tumor cells receive prosurvival cues from different constituents of the ECM through integrin receptors through which they sense their three-dimensional location . In contrast, other ECM molecules impair cell viability. Among these proteins, CCN1 can either induce or suppress apoptosis through the integrin α6β1 and syndecan-4 [22,23]. Thrombospondin-1 stimulates apoptosis through the CD36/p59fyn pathway . Endostatin halts ECs' viability through the suppression of apoptosis inhibitors such as Bcl-2 or through autophagy [25,26]. The secreted protein acidic and rich in cysteine (SPARC) functions as a tumor suppressor in various cancers  and induces caspase-8 activation in a death receptor-independent manner , whereas decorin induces caspase-3 activation . Elastin microfibril interface located protein 2 (EMILIN2), a member of the EMILIN family, is characterized by the cysteine-rich EMI domain at the N-terminus, the hallmark of the family , α-helical domains with high probability for coiled-coil structure formation and, at the C-terminus, a proline-rich motif adjacent to a collagenous stalk preceding the globular gC1q domain . Recently, we have demonstrated that EMILIN2, unique among the other ECM molecules, binds to TRAIL receptors, leading to apoptotic cell death .
In this study, we discover an additional property of EMILIN2; first, we identify the region responsible for the proapoptotic effects and demonstrate that, as well as the whole molecule, it exerts a potent antineoplastic action in vivo. Second, we found out that EMILIN2 positively affects tumor angiogenesis by acting on EC's behavior and obtained a total regression of tumor growth in combination with the anti-VEGF antibody bevacizumab. Given that neither EMILIN2 nor bevacizumab is toxic for normal cells, these results open the possibility to develop novel anticancer tools with putatively limited adverse effects.
HT1080 and HeLa cell lines were from American Type Culture Collection (ATCC, Manassas, VA) and cultured in Dulbecco's modified Eagle medium containing 10% FBS (GIBCO BRL, Milan, Italy). NHDF-Ad-Adult human dermal fibroblasts and human umbilical vein endothelial cells (HUVECs) were from Lonza S.r.l. (Milan, Italy) and were cultured in fibroblast growth medium or endothelial growth medium, respectively; embryonic kidney cells (293-EBNA cells) were a gift from Rupert Timpl (Max Planck, Munich, Germany) and were cultured in Dulbecco's modified Eagle medium containing 10% FBS and 250 µg/ml of G418 and 0.5 µg/ml of puromycin after transfection. All cells were maintained at 37°C in a humidified 5% CO2 atmosphere.
The 828B3B3 anti-EMILIN2 was obtained as previously described . The anti-Δ4 polyclonal antibody was obtained after the injection of 150 µg of the recombinant fragment in rabbit (three total boosters). The anti-cleaved Poly(ADP-ribose) polymerase polyclonal antibody, anti-FADD antibody, FLICE/caspase 8 fluorimetric Protease Assay Kit, and ApopTag Apoptosis Detection Kit were from Chemicon International (Temecula, CA). Anti-caspase 3 and anti-type I collagen antibodies were from Santa Cruz Biotechnology, Inc., Santa Cruz, CA. The anti-DR4, anti-β-actin, anti-FLAG antibodies, the FLAG peptide, the anti-FLAG M2 affinity gel, and the human fibronectin were from Sigma-Aldrich S.r.l. (Milan, Italy) and Alexis (Plymouth Meeting, PA). The secondary HRP-conjugated antibodies were from Amersham (GE Healthcare S.r.l., Milan, Italy). The secondary antibody conjugated with Alexa Fluor 680 was from Invitrogen S.r.l. (Milan, Italy). The anti-histidine monoclonal antibody was from Abgent (Resnova S.r.l., Rome, Italy). The anti-glutathione S-transferase (GST) monoclonal antibody was from Abcam (Cambridge, MA). The Ni-NTA agarose was from QIAGEN (Milan, Italy). The Cell Death Detection ELISAPLUS Kit (used exclusively for the experiment in Figure 3F) and In Situ Cell Death Detection Fluorescein Kit were purchased from Roche Diagnostics S.p.A. (Milan, Italy). BD Adeno-X Rapid Titer Kit was from BD (Buccinasco, Milan, Italy). Caspase-Glo 8 and Caspase-Glo 3/7 assays were from Promega S.r.l. (Milan, Italy) and were used on tumor lysates according to the manufacturer's instructions.
EMILIN2 deletion constructs Δ1 and Δ2 were generated using the EcoNI and NarI restriction enzymes and the available pcDNA3.1-EMILIN2 construct . Δ3 and Δ4 complementary DNA were amplified with the following oligonucleotides: forward 5′-CTAGCTAGCAGACACCACTGTTAGTGGA-3′ and reverse 5′-CGGGATCCTTATTCTTCAGCGTTCTTCTC-3′; forward 5′-CTAGCTAGCAGGCCCGCAGCCCGGGTAT-3′ and reverse 5′-CGGGATCCTTAAGGCTTGGGGTGCTGGCT-3′. EMILIN2 and Δ4 were subcloned in pDNR-CMV plasmid by the SalI and BamHI restriction enzymes. The following oligonucleotides were used: forward 5′-CCCAAGCTTGTCGACCCACATGAGGGCCTGG ATCTTCTT-3′ and the above two reverse oligonucleotides. Wild-type DR4 (wtDR4), truncated DR4 mutant (tDR4), and TRAIL were reverse-transcribed from total RNA extracted from the HeLa cell line and cloned into pcDNA3.1/Myc-His. The following oligonucleotides were used: forward 5′-CCCAAGCTTATGGCGCCACCACCAGCT-3′ and reverse 5′-TGGATATCTCTCCAAGGACACGGCAGAGC-3′; the reverse oligonucleotide for the tDR4 was as 5′-GCTCTAGAGAGACCCAAGCGCCAGAA-3′. The forward oligonucleotide for TRAIL containing a FLAG tag was 5′-CTAGCTAGCCGACTACAAGGACGACGATGACAAGACCTCTGAGGAAACCATTTC-3′ and the reverse was 5′-CGTGAGCGGCCGCCAGGTCAGTTAGCCAACT-3′. The fragment was cloned into the pCEP-Pu vector and expressed in 293-EBNA cells. The extracellular regions of DR4 were cloned into the pGEX-KG vector with the following oligonucleotides: forward 5′-CGGGATCCGCGAGTGGGACAGAGGCA-3′ and reverse 5′-GGAATT CATTATGTCCATTGCCTGA-3′.
Expression of the recombinant molecules was carried out after transfection of 293-EBNA cells as previously described . Proteins were purified by means of the Ni-NTA or anti-FLAG beads. HT1080 cell lines were transfected using FuGene6 reagent (Roche Diagnostics S.p.A., Milan, Italy) according to the manufacturer's instructions.
The pDNR-CMV constructs were transferred into the Adeno-X-TRE acceptor vector through Cre-loxP recombination. The recombination products have been analyzed by polymerase chain reaction and enzyme restriction. Adenovirus was produced by transfecting HEK 293 cells, and the virus titration was determined using the Adeno-X Rapid Titer Kit (Clontech, EuroClone S.p.a., Milan, Italy). HT1080 cells were cotransduced with a regulatory virus (Adeno-X Tet-On; Clontech) and the response of EMILIN2- and Δ4-recombinant viruses using 2 µg/ml of doxycyline for induction.
Proteins were resolved in 4% to 20% Criterion Precast Gels (Bio-Rad Laboratories, Milan, Italy) and transferred onto Hybond-ECL nitrocellulose membranes (Amersham, GE Healthcare). Membranes were blocked with 5% dry milk in Tris-buffered saline Tween-20, probed with the appropriate antibodies, and the blots were developed using Enhanced chemiluminescence (Amersham Biosciences, Cologno Monzese, Italy). Alternatively, the Odyssey Infrared Imaging System was used (Li-COR Biosciences, Lincoln, NE).
Immunoprecipitation of the DISC complex was carried out as previously described ; an equimolar amount (37.5 nM) of EMILIN2, Δ4, TRAIL, or type I collagen was used, and the cells were incubated for 30 minutes at 37°C. To prevent interference with the antibodies used for immunoprecipitation, the Trueblot secondary antibody (eBiosciences, San Diego, CA) was used. The GST pull-down assay was carried out as previously described  using in vitro transcribed and translated EMILIN2, the Δ1 and Δ4 and TRAIL, and 10 µg of the recombinant DR4 extracellular region fused to the GST.
Cellswere plated in triplicates; after the different treatments, 0.3 mg/ml of 3-(4,5-dimethylthiazol-2-yl)-2,5 diphenyltetrazoliumbromide (MTT) was added, and the cells were incubated for three extra hours. The formazan crystals were solubilized with DMSO, and the absorbance was detected at 620 nm.
Apoptotic cells challenged with the various proteins were detected with the Cell Death Detection ELISAPLUS terminal deoxynucleotide transferase dUTP nick end labeling (TUNEL) assay. For the TUNEL test on tumor sections, the specimens were embedded in OCT Embedding Matrix (Kaltek S.r.l. Padua, Italy), and 7-µm-thick cryostat sections were obtained using a MICRON cryostat (Heidelberg, Germany); apoptosis was detected with the ApopTag Peroxidase In Situ Oligo Ligation Apoptosis Detection Kit (Chemicon International) according to the manufacturer's instructions. Blood vessels were detected with an antimouse multimerin 2 monoclonal antibody that was produced after the injection of the purified recombinant protein in BALB/C mice . Cell nuclei were stained with the TO-PRO-3 fluorescent dye (Molecular Probes, Invitrogen S.r.l.). Images were acquired with a confocal system (Leica Microsystems, Milan, Italy).
Soft agar colony assays were carried out using 0.5% low-melting agarose as previously described . HT1080 cells transduced with the Δ4 recombinant adenovirus were used in the presence or absence of 2 µg/ml doxycycline. Alternatively, equimolar amounts of recombinant fibronectin or of Δ4 (37.5 nM) were added. Pictures were obtained after 13 days of incubation, and the clones that formed by more that 10 cells were counted. The assay was performed in triplicate. For the three-dimensional Matrigel growth test, 5 x 105 HT1080 cells transduced either with the EMILIN2 or with Δ4 were embedded in 500 µl of Matrigel (BD Biosciences Euroclone S.p.a., Milan, Italy). Complete medium was added after solidification of the Matrigel drops containing the cells. After 10 days, pictures were obtained.
Two hundred HT1080 cells per well were plated on 12-well plates. Conditioned media from mock, EMILIN2, or Δ4 were used in the presence of 5% FBS. After 13 days, the cells were washed with PBS and stained with crystal violet; pictures were obtained and the clones were counted. Alternatively, HT1080 cells were transduced with the Δ4 adenoviral construct and were plated in the presence or absence of doxycycline.
Transwells with 8-µm pore membranes (BD Biosciences, Buccinasco, Italy) were coated with 10 µg/ml of denatured bovine serum albumin, type I collagen, or EMILIN2, and 150 x 103 HUVECs were used for each point. Fifty nanograms per milliliter of VEGF (BD Biosystems) was used as a chemoattractant. After 24 hours, cells were removed from the top of the membranes, and the migrated cells on the bottom were fixed, stained with crystal violet, and counted. Each experiment was performed in triplicate and repeated three times.
HUVECs cells were plated on gelatin-coated 24-well plates and, on reaching confluence, were starved for 3 hours in serum-free medium and scratched with a tip. Cells were then incubated with 10 µg/ml of EMILIN2 or type I collagen in complete diluted 1:1 with serum-free medium. Pictures were obtained over time with the Leica AF6000 Imaging System (Leica Microsystems). The percentage of cell migration was evaluated measuring the area of the scratch covered by the migrated cells by means of the Image Tool Software (http://ddsdx.uthscsa.edu/dig/itdesc.html; San Antonio, TX). Each experiment was performed in triplicate and repeated three times.
A total of 25 x 103 HUVECs per well were seeded on a 96-well plate on top of presolidified Matrigel (BD Biosciences Euroclone). Cells were treated with 2.5 µg/ml of purified recombinant EMILIN2 or Δ4; an equal volume of PBS was used as a control. Pictures were obtained over time with the Leica AF6000 Imaging System (Leica Microsystems). The experiment was repeated three times.
Five BALB/c nude mice were subcutaneously injected with 1 x 106 HT1080 cells stably transfected with pcDNA3.1 vector carrying the luciferase or EGFP coding sequence. The animals were treated every 3 days with 10 µg of purified EMILIN2 or Δ3 or with an equal volume of PBS. Tumor growth was evaluated over time either using the in vivo Imaging System (Caliper Life Sciences S.A., France) or measuring the growing tumors with a caliper; in this case, tumor volumes were calculated using the following formula: πLW2/6, where L indicates length and W indicates width.
Alternatively, two groups of four animals were subcutaneously injected with 1 x 106 EMILIN2- or EMILIN2-Δ4-transduced HT1080 cells in the right flank and with Tet-On-transduced cells, used as a control, in the left flank. The cell number was carefully evaluated by vital dye exclusion (Trypan blue; Invitrogen S.r.l.); thus, a comparable number of vital cells were injected in the different groups of animals. Drinkable water was added with 1 mg/ml doxycycline and 2.5% sucrose, given to the animals 2 days before tumor injection and changed every 2 days. In other experiments, 500 µg of bevacizumab per mouse (Roche Pharma AG, Germany) was intraperitoneally injected 2 days after cell injection.
Five other BALB/c nude mice were subcutaneously injected with 1 x 106 HT1080 cells and treated with 20 µg of purified EMILIN2 or Δ4 mutant or PBS at days 2, 5, and 9 after the implant.
The graphs and statistical analysis were generated using Microsoft Office Excel (Microsoft Italia, Segrate, Italy) and SigmaPlot software (Systat Software Inc., San Jose, CA). Densitometric analysis was performed using the Image Tool Software. Cell count and analysis of viability were performed with Countess Automated Cell Counter and the relative software (Invitrogen S.r.l.). In vivo imaging acquisitions were performed using the IVIS Imaging System 100 Series (Caliper Life Sciences). Evaluation of the light units for the luminescent substrates was carried out with a luminometer (PerkinElmer, Inc., Waltham, MA). Data analysis was performed using Living Image 3.0 software (Caliper Life Sciences). All images have been assembled using Adobe Illustrator CS and Adobe Photoshop CS (Adobe Systems Inc., San Jose, CA).
To investigate the proapoptotic effects of EMILIN2 in detail, we have generated an inducible adenoviral construct. This system allowed an efficient expression of the molecule, which can be turned on by the addition of doxycycline and is particularly suitable for both in vitro and in vivo studies (Figure 1A). As expected from our previous observations , this approach overexpressing EMILIN2 also led to a significant decrease in tumor cell viability, which correlated with a dramatic increase in the cell death rate. This effect is distinctive of EMILIN2 because treatment with other members of the family such as EMILIN1 or Multimerin2 did not affect tumor cell viability (Figure 1B). Moreover, only cancer cells were susceptible to the EMILIN2-driven proapoptotic activity (Figure 1C). Transduction of normal dermal fibroblasts, among other normal cells (data not shown), with the EMILIN2 adenoviral vector did not affect their viability as detected by the MTT assay, confirming the results obtained with another cell model  (Figure 1D). To determine which region of the molecule was responsible for these effects, we have generated a series of deletion mutants (Figure 2, A and B). In this case, the purified recombinant molecules were used to challenge HT1080 cells. As shown in Figure 2C, the Δ1 and the Δ2 deletion mutants did not affect cell viability, whereas the Δ3 and the Δ4 were as effective as the whole molecule. Because the Δ4 fragment was the smallest deletion retaining the proapoptotic activity, we have chosen this portion for most of the subsequent analysis. This recombinant fragment is soluble and likely preserves a proper three-dimensional structure because it self-assembled in vitro to form homotrimers, as demonstrated by native PAGE analysis (Figure W1). The treatment of HT1080 cells with Δ4 led to a dramatic increase in their apoptotic rate, comparable to that induced by the whole molecule or by TRAIL (Figure 2D). This effect was also evident after transduction of HT1080 cells with the Δ4 adenoviral construct (Figure 3, A and B). The mechanism of action activated by the fragment recapitulates that of the whole molecule described in our previous study . This fragment in fact binds to death receptor DR4 (Figure 3, C and D), leading to receptor activation, as demonstrated by the induction of DISC assembly followed by caspase-8 cleavage (Figure 3, D and E). The proapoptotic effect induced by Δ4 was abrogated by overexpressing in HT1080 cells a dominant-negative DR4 mutant lacking the cytoplasmic domain; thus, this function requires an intact receptor (Figure 3F).
To determine whether the EMILIN2 and Δ4 proapoptotic effects could lead to a decreased tumor growth in vitro, we first set up a clonogenic assay. As shown in Figure 4A, HT1080 cells challenged with EMILIN2 or Δ4 formed 50% less colonies in comparison with the control. Similar results were obtained with the use of the Δ4 adenoviral vector (Figure 4A, right panel). The Δ4 fragment also halted the number of colonies developed in soft agar either using the recombinant molecule or after transduction with the recombinant adenovirus (Figure 4B). Moreover, Δ4 impaired the growth of tumor cells in a three-dimensional context, despite the prosurvival stimuli provided by the growth factors and ECM Matrigel components (Figure 4C), as previously demonstrated for the entire molecule . Next, to verify whether EMILIN2 could also decrease tumor growth in vivo, EGFP-expressing HT1080 cells were injected in nude mice, and the tumor masses were then challenged with purified recombinant EMILIN2. The treatment led to a significant decrease of tumor growth (Figure 5A). Analogous results were obtained, challenging established tumors with recombinant EMILIN2 (Figure W2). Accordingly, overexpression of EMILIN2 after transduction with the adenoviral vector led to a remarkable inhibition of tumor growth (Figure 5B). Similar results were obtained treating the tumors with recombinant Δ3 (Figure 5C) or transducing the cells with the Δ4 adenoviral vector (Figure 5D).
TUNEL assay analysis performed on the tumor sections revealed that the tumors treated with EMILIN2 or Δ4 displayed a significant increase in tumor cell apoptosis compared with the controls (Figure 6A). The augmented intratumoral apoptosis was prompted by the activation of extrinsic pathway initiator caspase-8, which in turn activates the effector caspase-3/7 likely accounting for the decreased tumor growth (Figure 6B). A more accurate analysis of the tumor sections highlighted the presence of an increased tumor vessels density in the tumors treated with EMILIN2 or Δ4 (Figure 6C). This finding prompted us to investigate the efficacy of EMILIN2 and Δ4 in halting tumor growth in combination with the antiangiogenetic antibody bevacizumab. The combination of Δ4 with bevacizumab was strikingly more effective than either treatment alone (Figure 6D). In most cases, a total regression of the tumors was observed, and the tumors failed to resume growth even after the withdrawal of bevacizumab treatment and of doxycycline. Only two of the treated tumors formed small masses of approximately 2 mm3 thus allowing the excision for further analysis. As shown in Figure 6D, treatment with bevacizumab was highly effective in reducing the Δ4-dependent vessel density. The depletion of the blood vessels might likely account for the remarkable results obtained with the combined treatment.
The finding that tumors treated with EMILIN2 and Δ4 displayed increased vascularization prompted us to investigate whether EMILIN2 could affect EC function. Whereas treatment with EMILIN2 or Δ4 failed to alter tubules formation in Matrigel (Figure 7A), they significantly increased their proliferation rate (Figure 7B). In addition, EMILIN2 was found to positively affect EC migration both in tests performed in transwells (Figure 7B) and in scratch tests (Figure 7, C and D). Thus, the increased EC proliferation and migration induced by EMILIN2 may likely account for the proangiogenic effects observed in vivo (Figure 6C). Taken together, these results unveil the possibility to develop new promising tools for cancer treatment.
In this study, we demonstrate a dual role for EMILIN2. First, we identify that the proapoptotic property resides in a region toward the N-terminus of the molecule: the Δ4 deletion mutant encompasses the features of the entire molecule. It binds to the DR4 receptor, inducing DISC assembly and consequent activation of the extrinsic apoptotic pathway, and it displays a significant antitumorigenic activity in vitro. It is interesting to note that in analogy with TRAIL, this fragment is homotrimeric. It is thus conceivable that it might engage death receptors mimicking TRAIL three-dimensional homotrimeric structure rather than the amino acid composition because the two molecules have no sequence homology. In addition, in this study, for the first time, we have analyzed the activity of EMILIN2 and Δ4 in vivo. The adenoviral system resulted particularly useful for these studies, especially those involving combinatorial treatments because it prevented the requirement of multiple injections. Our results indicate that tumor growth was significantly impaired by EMILIN2 or Δ4 either overexpressing the molecules or after treatment with the purified proteins. It is interesting to remark that, as shown in Figure 5B, the growth of Ad E2 cells in the first 6 days after the injection was comparable to that of the control cells. This indicates that the exogenous EMILIN2, which is removed after the trypsinization and washing procedures, must be resynthesized in vivo before exerting its proapoptotic effects. The growth impairment of the tumors challenged with Δ4 correlates with a significant increased activation of the extrinsic apoptotic pathway because activated caspase-8 and caspase-3/7 were enhanced in these samples. Taken together, these observations indicate that the EMILIN2 antitumoral properties can be primarily ascribed to its proapoptotic effects. It is interesting to point out that the proapoptotic activity exerted by this molecule was specific for tumor cells while leaving normal cells unharmed. In fact, the ectopic expression of EMILIN2 or Δ4 in normal skin fibroblasts did not lead to any change in their viability. This aspect is particularly advantageous when dealing with a putative antineoplastic molecule such as the case of TRAIL, which, for its specificity in killing tumor cells, has been proposed as a promising therapeutical tool for the cure of cancer. In fact, several preclinical studies pinpoint the successes encountered with the use of this molecule [18,34–36]. In this context, the EMILIN2 Δ4 region may also represent a promising approach for the development of novel nontoxic therapeutic tools. EMILIN2 is expressed by a variety of tissues during development, and its deposition decreases in adulthood . It was demonstrated to be expressed in Ewing sarcomas and to be induced by epidermal growth factor and insulin-like growth factor 1 . However, the low or negative expression of EMILIN2 in some tumors should not preclude its usefulness as a potential therapeutic agent. Whether the higher expression of EMILIN2 in some tumor types directly relates to their growth rate is a matter of further investigation provided that appropriate antibody reagents become available.
The second aspect of our findings was that the treatment with EMILIN2 and Δ4, despite the robust reduction of tumor growth, led to a significant increase in the blood vessel density, as observed after a deeper analysis of the cancer lesions. This discovery was supported by in vitro experiments where EMILIN2 and Δ4 significantly increased EC proliferation and motility, which likely led to the increased tumor vascularization. Interestingly, the increase in motility is specifically induced on EC only because preliminary data demonstrate that EMILIN2 impinges cell migration of different tumor cell lines (data not shown). The molecular mechanisms, based on the increased EC proliferation and motility induced by EMILIN2, were not investigated here. The α4β1 integrin is a receptor for ECM molecules such as fibronectin and EMILIN1 [38–40]. One possibility would be that it may also function as a receptor for EMILIN2: this integrin is also expressed by ECs , and similar to EMILIN1, EMILIN2 bares an inserted sequence in the C1q domain that contains amino acid residues compatible with an α4β1 binding site . Thus, the proangiogenic activity of EMILIN2 might be related to integrin ligation. However, this possibility seems unlikely because the Δ4 mutant is as effective as the whole EMILIN2 molecule in inducing angiogenesis. Alternatively, EMILIN2 and Δ4 might engage a still unknown cell surface receptor. Finally, EMILIN2 and Δ4 may function as a reservoir for cytokines that activate specific receptor tyrosine kinases at the EC surface driving in turn the angiogenic response. The combined treatment with EMILIN2 or Δ4 and bevacizumab [43,44] successfully improved the cure rate and, in most cases, led to the complete regression of the tumors. These results indicate that the residual tumor growth observed after the treatment with EMILIN2 or Δ4 was sustained by increased tumor perfusion. It is known that VEGF, which is recognized to display antiapoptotic effects , as well as its receptor, is expressed by tumor cells thus establishing a positive loop of stimulation. By removing a key prosurvival stimulus, the blockage of VEGF with bevacizumab may shift the balance toward the proapoptotic action of EMILIN2 and Δ4. In addition, increased blood vessel perfusion prompted by EMILIN2/Δ4 may favor drug delivery, leading to a paradoxical higher efficiency in blocking tumor vessel formation by antiangiogenic or conventional cytotoxic drugs. Conversely, normalization of tumor vasculature induced by antiangiogenic therapy  may enhance Δ4 delivery within the tumor thus allowing a more effective proapoptotic therapy.
In conclusion, there are at least three aspects that render the cotreatment with EMILIN2 and bevacizumab particularly promising. First, both molecules do not harm normal cells. Second, bevacizumab, despite the fact that it enhances tumor hypoxia, abolishes the undesired increase in blood vessel formation induced by EMILIN2, simultaneously sensitizing the cells to the EMILIN2 proapoptotic action, thus improving the efficacy of the treatments . Third, the initial increased tumor perfusion induced by EMILIN2, as well as the vasculature normalization induced by bevacizumab, may facilitate drug delivery. Taken together, these results highlight a potent antitumoral effect in vivo of a 90-residue long EMILIN2 region. Owing to the lack of cytotoxicity for normal cells, it is conceivable that this molecule or specific peptide encompassing the proapoptotic deletion mutant might be applied in the future as an antitumoral agent especially in combination with antiangiogenic drugs such as bevacizumab.
1The authors thank the ISS-ACC Program 2, the Associazione Italiana per la Ricerca sul Cancro and MIUR Fondi di Investimento per laRicerca di Base (grant no. RBRN07BMCT) for supporting this work. No potential conflicts of interest are disclosed.